2010-04-06 22:14:15 +00:00
|
|
|
#include <linux/ceph/ceph_debug.h>
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
#include <linux/fs.h>
|
|
|
|
#include <linux/kernel.h>
|
2017-02-02 18:15:33 +00:00
|
|
|
#include <linux/sched/signal.h>
|
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 08:04:11 +00:00
|
|
|
#include <linux/slab.h>
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
#include <linux/vmalloc.h>
|
|
|
|
#include <linux/wait.h>
|
2010-01-18 00:53:08 +00:00
|
|
|
#include <linux/writeback.h>
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
#include "super.h"
|
2010-04-06 22:14:15 +00:00
|
|
|
#include "mds_client.h"
|
2013-08-21 21:29:54 +00:00
|
|
|
#include "cache.h"
|
2010-04-06 22:14:15 +00:00
|
|
|
#include <linux/ceph/decode.h>
|
|
|
|
#include <linux/ceph/messenger.h>
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Capability management
|
|
|
|
*
|
|
|
|
* The Ceph metadata servers control client access to inode metadata
|
|
|
|
* and file data by issuing capabilities, granting clients permission
|
|
|
|
* to read and/or write both inode field and file data to OSDs
|
|
|
|
* (storage nodes). Each capability consists of a set of bits
|
|
|
|
* indicating which operations are allowed.
|
|
|
|
*
|
|
|
|
* If the client holds a *_SHARED cap, the client has a coherent value
|
|
|
|
* that can be safely read from the cached inode.
|
|
|
|
*
|
|
|
|
* In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the
|
|
|
|
* client is allowed to change inode attributes (e.g., file size,
|
|
|
|
* mtime), note its dirty state in the ceph_cap, and asynchronously
|
|
|
|
* flush that metadata change to the MDS.
|
|
|
|
*
|
|
|
|
* In the event of a conflicting operation (perhaps by another
|
|
|
|
* client), the MDS will revoke the conflicting client capabilities.
|
|
|
|
*
|
|
|
|
* In order for a client to cache an inode, it must hold a capability
|
|
|
|
* with at least one MDS server. When inodes are released, release
|
|
|
|
* notifications are batched and periodically sent en masse to the MDS
|
|
|
|
* cluster to release server state.
|
|
|
|
*/
|
|
|
|
|
2016-07-04 10:06:41 +00:00
|
|
|
static u64 __get_oldest_flush_tid(struct ceph_mds_client *mdsc);
|
2016-07-07 10:34:45 +00:00
|
|
|
static void __kick_flushing_caps(struct ceph_mds_client *mdsc,
|
|
|
|
struct ceph_mds_session *session,
|
|
|
|
struct ceph_inode_info *ci,
|
|
|
|
u64 oldest_flush_tid);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Generate readable cap strings for debugging output.
|
|
|
|
*/
|
|
|
|
#define MAX_CAP_STR 20
|
|
|
|
static char cap_str[MAX_CAP_STR][40];
|
|
|
|
static DEFINE_SPINLOCK(cap_str_lock);
|
|
|
|
static int last_cap_str;
|
|
|
|
|
|
|
|
static char *gcap_string(char *s, int c)
|
|
|
|
{
|
|
|
|
if (c & CEPH_CAP_GSHARED)
|
|
|
|
*s++ = 's';
|
|
|
|
if (c & CEPH_CAP_GEXCL)
|
|
|
|
*s++ = 'x';
|
|
|
|
if (c & CEPH_CAP_GCACHE)
|
|
|
|
*s++ = 'c';
|
|
|
|
if (c & CEPH_CAP_GRD)
|
|
|
|
*s++ = 'r';
|
|
|
|
if (c & CEPH_CAP_GWR)
|
|
|
|
*s++ = 'w';
|
|
|
|
if (c & CEPH_CAP_GBUFFER)
|
|
|
|
*s++ = 'b';
|
|
|
|
if (c & CEPH_CAP_GLAZYIO)
|
|
|
|
*s++ = 'l';
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
|
|
|
const char *ceph_cap_string(int caps)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
char *s;
|
|
|
|
int c;
|
|
|
|
|
|
|
|
spin_lock(&cap_str_lock);
|
|
|
|
i = last_cap_str++;
|
|
|
|
if (last_cap_str == MAX_CAP_STR)
|
|
|
|
last_cap_str = 0;
|
|
|
|
spin_unlock(&cap_str_lock);
|
|
|
|
|
|
|
|
s = cap_str[i];
|
|
|
|
|
|
|
|
if (caps & CEPH_CAP_PIN)
|
|
|
|
*s++ = 'p';
|
|
|
|
|
|
|
|
c = (caps >> CEPH_CAP_SAUTH) & 3;
|
|
|
|
if (c) {
|
|
|
|
*s++ = 'A';
|
|
|
|
s = gcap_string(s, c);
|
|
|
|
}
|
|
|
|
|
|
|
|
c = (caps >> CEPH_CAP_SLINK) & 3;
|
|
|
|
if (c) {
|
|
|
|
*s++ = 'L';
|
|
|
|
s = gcap_string(s, c);
|
|
|
|
}
|
|
|
|
|
|
|
|
c = (caps >> CEPH_CAP_SXATTR) & 3;
|
|
|
|
if (c) {
|
|
|
|
*s++ = 'X';
|
|
|
|
s = gcap_string(s, c);
|
|
|
|
}
|
|
|
|
|
|
|
|
c = caps >> CEPH_CAP_SFILE;
|
|
|
|
if (c) {
|
|
|
|
*s++ = 'F';
|
|
|
|
s = gcap_string(s, c);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (s == cap_str[i])
|
|
|
|
*s++ = '-';
|
|
|
|
*s = 0;
|
|
|
|
return cap_str[i];
|
|
|
|
}
|
|
|
|
|
2010-06-17 23:16:12 +00:00
|
|
|
void ceph_caps_init(struct ceph_mds_client *mdsc)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
2010-06-17 23:16:12 +00:00
|
|
|
INIT_LIST_HEAD(&mdsc->caps_list);
|
|
|
|
spin_lock_init(&mdsc->caps_list_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
2010-06-17 23:16:12 +00:00
|
|
|
void ceph_caps_finalize(struct ceph_mds_client *mdsc)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
|
|
|
struct ceph_cap *cap;
|
|
|
|
|
2010-06-17 23:16:12 +00:00
|
|
|
spin_lock(&mdsc->caps_list_lock);
|
|
|
|
while (!list_empty(&mdsc->caps_list)) {
|
|
|
|
cap = list_first_entry(&mdsc->caps_list,
|
|
|
|
struct ceph_cap, caps_item);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
list_del(&cap->caps_item);
|
|
|
|
kmem_cache_free(ceph_cap_cachep, cap);
|
|
|
|
}
|
2010-06-17 23:16:12 +00:00
|
|
|
mdsc->caps_total_count = 0;
|
|
|
|
mdsc->caps_avail_count = 0;
|
|
|
|
mdsc->caps_use_count = 0;
|
|
|
|
mdsc->caps_reserve_count = 0;
|
|
|
|
mdsc->caps_min_count = 0;
|
|
|
|
spin_unlock(&mdsc->caps_list_lock);
|
2010-02-17 18:02:43 +00:00
|
|
|
}
|
|
|
|
|
2010-06-17 23:16:12 +00:00
|
|
|
void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta)
|
2010-02-17 18:02:43 +00:00
|
|
|
{
|
2010-06-17 23:16:12 +00:00
|
|
|
spin_lock(&mdsc->caps_list_lock);
|
|
|
|
mdsc->caps_min_count += delta;
|
|
|
|
BUG_ON(mdsc->caps_min_count < 0);
|
|
|
|
spin_unlock(&mdsc->caps_list_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
2013-06-26 03:15:27 +00:00
|
|
|
void ceph_reserve_caps(struct ceph_mds_client *mdsc,
|
2010-06-17 23:16:12 +00:00
|
|
|
struct ceph_cap_reservation *ctx, int need)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
|
|
|
int i;
|
|
|
|
struct ceph_cap *cap;
|
|
|
|
int have;
|
|
|
|
int alloc = 0;
|
|
|
|
LIST_HEAD(newcaps);
|
|
|
|
|
|
|
|
dout("reserve caps ctx=%p need=%d\n", ctx, need);
|
|
|
|
|
|
|
|
/* first reserve any caps that are already allocated */
|
2010-06-17 23:16:12 +00:00
|
|
|
spin_lock(&mdsc->caps_list_lock);
|
|
|
|
if (mdsc->caps_avail_count >= need)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
have = need;
|
|
|
|
else
|
2010-06-17 23:16:12 +00:00
|
|
|
have = mdsc->caps_avail_count;
|
|
|
|
mdsc->caps_avail_count -= have;
|
|
|
|
mdsc->caps_reserve_count += have;
|
|
|
|
BUG_ON(mdsc->caps_total_count != mdsc->caps_use_count +
|
|
|
|
mdsc->caps_reserve_count +
|
|
|
|
mdsc->caps_avail_count);
|
|
|
|
spin_unlock(&mdsc->caps_list_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
for (i = have; i < need; i++) {
|
|
|
|
cap = kmem_cache_alloc(ceph_cap_cachep, GFP_NOFS);
|
2013-06-26 03:15:27 +00:00
|
|
|
if (!cap)
|
|
|
|
break;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
list_add(&cap->caps_item, &newcaps);
|
|
|
|
alloc++;
|
|
|
|
}
|
2013-06-26 03:15:27 +00:00
|
|
|
/* we didn't manage to reserve as much as we needed */
|
|
|
|
if (have + alloc != need)
|
|
|
|
pr_warn("reserve caps ctx=%p ENOMEM need=%d got=%d\n",
|
|
|
|
ctx, need, have + alloc);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2010-06-17 23:16:12 +00:00
|
|
|
spin_lock(&mdsc->caps_list_lock);
|
|
|
|
mdsc->caps_total_count += alloc;
|
|
|
|
mdsc->caps_reserve_count += alloc;
|
|
|
|
list_splice(&newcaps, &mdsc->caps_list);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2010-06-17 23:16:12 +00:00
|
|
|
BUG_ON(mdsc->caps_total_count != mdsc->caps_use_count +
|
|
|
|
mdsc->caps_reserve_count +
|
|
|
|
mdsc->caps_avail_count);
|
|
|
|
spin_unlock(&mdsc->caps_list_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
ctx->count = need;
|
|
|
|
dout("reserve caps ctx=%p %d = %d used + %d resv + %d avail\n",
|
2010-06-17 23:16:12 +00:00
|
|
|
ctx, mdsc->caps_total_count, mdsc->caps_use_count,
|
|
|
|
mdsc->caps_reserve_count, mdsc->caps_avail_count);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
2010-06-17 23:16:12 +00:00
|
|
|
int ceph_unreserve_caps(struct ceph_mds_client *mdsc,
|
|
|
|
struct ceph_cap_reservation *ctx)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
|
|
|
dout("unreserve caps ctx=%p count=%d\n", ctx, ctx->count);
|
|
|
|
if (ctx->count) {
|
2010-06-17 23:16:12 +00:00
|
|
|
spin_lock(&mdsc->caps_list_lock);
|
|
|
|
BUG_ON(mdsc->caps_reserve_count < ctx->count);
|
|
|
|
mdsc->caps_reserve_count -= ctx->count;
|
|
|
|
mdsc->caps_avail_count += ctx->count;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
ctx->count = 0;
|
|
|
|
dout("unreserve caps %d = %d used + %d resv + %d avail\n",
|
2010-06-17 23:16:12 +00:00
|
|
|
mdsc->caps_total_count, mdsc->caps_use_count,
|
|
|
|
mdsc->caps_reserve_count, mdsc->caps_avail_count);
|
|
|
|
BUG_ON(mdsc->caps_total_count != mdsc->caps_use_count +
|
|
|
|
mdsc->caps_reserve_count +
|
|
|
|
mdsc->caps_avail_count);
|
|
|
|
spin_unlock(&mdsc->caps_list_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2014-04-18 01:57:11 +00:00
|
|
|
struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
|
|
|
|
struct ceph_cap_reservation *ctx)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
|
|
|
struct ceph_cap *cap = NULL;
|
|
|
|
|
|
|
|
/* temporary, until we do something about cap import/export */
|
2010-06-29 16:28:39 +00:00
|
|
|
if (!ctx) {
|
|
|
|
cap = kmem_cache_alloc(ceph_cap_cachep, GFP_NOFS);
|
|
|
|
if (cap) {
|
2012-11-03 02:32:37 +00:00
|
|
|
spin_lock(&mdsc->caps_list_lock);
|
2010-06-17 23:16:12 +00:00
|
|
|
mdsc->caps_use_count++;
|
|
|
|
mdsc->caps_total_count++;
|
2012-11-03 02:32:37 +00:00
|
|
|
spin_unlock(&mdsc->caps_list_lock);
|
2010-06-29 16:28:39 +00:00
|
|
|
}
|
|
|
|
return cap;
|
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2010-06-17 23:16:12 +00:00
|
|
|
spin_lock(&mdsc->caps_list_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
dout("get_cap ctx=%p (%d) %d = %d used + %d resv + %d avail\n",
|
2010-06-17 23:16:12 +00:00
|
|
|
ctx, ctx->count, mdsc->caps_total_count, mdsc->caps_use_count,
|
|
|
|
mdsc->caps_reserve_count, mdsc->caps_avail_count);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
BUG_ON(!ctx->count);
|
2010-06-17 23:16:12 +00:00
|
|
|
BUG_ON(ctx->count > mdsc->caps_reserve_count);
|
|
|
|
BUG_ON(list_empty(&mdsc->caps_list));
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
ctx->count--;
|
2010-06-17 23:16:12 +00:00
|
|
|
mdsc->caps_reserve_count--;
|
|
|
|
mdsc->caps_use_count++;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2010-06-17 23:16:12 +00:00
|
|
|
cap = list_first_entry(&mdsc->caps_list, struct ceph_cap, caps_item);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
list_del(&cap->caps_item);
|
|
|
|
|
2010-06-17 23:16:12 +00:00
|
|
|
BUG_ON(mdsc->caps_total_count != mdsc->caps_use_count +
|
|
|
|
mdsc->caps_reserve_count + mdsc->caps_avail_count);
|
|
|
|
spin_unlock(&mdsc->caps_list_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
return cap;
|
|
|
|
}
|
|
|
|
|
2010-06-17 23:16:12 +00:00
|
|
|
void ceph_put_cap(struct ceph_mds_client *mdsc, struct ceph_cap *cap)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
2010-06-17 23:16:12 +00:00
|
|
|
spin_lock(&mdsc->caps_list_lock);
|
ceph: fix iterate_caps removal race
We need to be able to iterate over all caps on a session with a
possibly slow callback on each cap. To allow this, we used to
prevent cap reordering while we were iterating. However, we were
not safe from races with removal: removing the 'next' cap would
make the next pointer from list_for_each_entry_safe be invalid,
and cause a lock up or similar badness.
Instead, we keep an iterator pointer in the session pointing to
the current cap. As before, we avoid reordering. For removal,
if the cap isn't the current cap we are iterating over, we are
fine. If it is, we clear cap->ci (to mark the cap as pending
removal) but leave it in the session list. In iterate_caps, we
can safely finish removal and get the next cap pointer.
While we're at it, clean up put_cap to not take a cap reservation
context, as it was never used.
Signed-off-by: Sage Weil <sage@newdream.net>
2010-02-16 19:39:45 +00:00
|
|
|
dout("put_cap %p %d = %d used + %d resv + %d avail\n",
|
2010-06-17 23:16:12 +00:00
|
|
|
cap, mdsc->caps_total_count, mdsc->caps_use_count,
|
|
|
|
mdsc->caps_reserve_count, mdsc->caps_avail_count);
|
|
|
|
mdsc->caps_use_count--;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
/*
|
2010-02-17 18:02:43 +00:00
|
|
|
* Keep some preallocated caps around (ceph_min_count), to
|
|
|
|
* avoid lots of free/alloc churn.
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
*/
|
2010-06-17 23:16:12 +00:00
|
|
|
if (mdsc->caps_avail_count >= mdsc->caps_reserve_count +
|
|
|
|
mdsc->caps_min_count) {
|
|
|
|
mdsc->caps_total_count--;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
kmem_cache_free(ceph_cap_cachep, cap);
|
|
|
|
} else {
|
2010-06-17 23:16:12 +00:00
|
|
|
mdsc->caps_avail_count++;
|
|
|
|
list_add(&cap->caps_item, &mdsc->caps_list);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
2010-06-17 23:16:12 +00:00
|
|
|
BUG_ON(mdsc->caps_total_count != mdsc->caps_use_count +
|
|
|
|
mdsc->caps_reserve_count + mdsc->caps_avail_count);
|
|
|
|
spin_unlock(&mdsc->caps_list_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
2010-04-06 22:14:15 +00:00
|
|
|
void ceph_reservation_status(struct ceph_fs_client *fsc,
|
2010-02-17 18:02:43 +00:00
|
|
|
int *total, int *avail, int *used, int *reserved,
|
|
|
|
int *min)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
2010-04-06 22:14:15 +00:00
|
|
|
struct ceph_mds_client *mdsc = fsc->mdsc;
|
2010-06-17 23:16:12 +00:00
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
if (total)
|
2010-06-17 23:16:12 +00:00
|
|
|
*total = mdsc->caps_total_count;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
if (avail)
|
2010-06-17 23:16:12 +00:00
|
|
|
*avail = mdsc->caps_avail_count;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
if (used)
|
2010-06-17 23:16:12 +00:00
|
|
|
*used = mdsc->caps_use_count;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
if (reserved)
|
2010-06-17 23:16:12 +00:00
|
|
|
*reserved = mdsc->caps_reserve_count;
|
2010-02-17 18:02:43 +00:00
|
|
|
if (min)
|
2010-06-17 23:16:12 +00:00
|
|
|
*min = mdsc->caps_min_count;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Find ceph_cap for given mds, if any.
|
|
|
|
*
|
2011-11-30 17:47:09 +00:00
|
|
|
* Called with i_ceph_lock held.
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
*/
|
|
|
|
static struct ceph_cap *__get_cap_for_mds(struct ceph_inode_info *ci, int mds)
|
|
|
|
{
|
|
|
|
struct ceph_cap *cap;
|
|
|
|
struct rb_node *n = ci->i_caps.rb_node;
|
|
|
|
|
|
|
|
while (n) {
|
|
|
|
cap = rb_entry(n, struct ceph_cap, ci_node);
|
|
|
|
if (mds < cap->mds)
|
|
|
|
n = n->rb_left;
|
|
|
|
else if (mds > cap->mds)
|
|
|
|
n = n->rb_right;
|
|
|
|
else
|
|
|
|
return cap;
|
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2010-06-30 19:44:34 +00:00
|
|
|
struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci, int mds)
|
|
|
|
{
|
|
|
|
struct ceph_cap *cap;
|
|
|
|
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_lock(&ci->i_ceph_lock);
|
2010-06-30 19:44:34 +00:00
|
|
|
cap = __get_cap_for_mds(ci, mds);
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
2010-06-30 19:44:34 +00:00
|
|
|
return cap;
|
|
|
|
}
|
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
/*
|
2010-05-26 21:31:27 +00:00
|
|
|
* Return id of any MDS with a cap, preferably FILE_WR|BUFFER|EXCL, else -1.
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
*/
|
2010-06-10 21:21:36 +00:00
|
|
|
static int __ceph_get_cap_mds(struct ceph_inode_info *ci)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
|
|
|
struct ceph_cap *cap;
|
|
|
|
int mds = -1;
|
|
|
|
struct rb_node *p;
|
|
|
|
|
2010-05-26 21:31:27 +00:00
|
|
|
/* prefer mds with WR|BUFFER|EXCL caps */
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) {
|
|
|
|
cap = rb_entry(p, struct ceph_cap, ci_node);
|
|
|
|
mds = cap->mds;
|
|
|
|
if (cap->issued & (CEPH_CAP_FILE_WR |
|
|
|
|
CEPH_CAP_FILE_BUFFER |
|
|
|
|
CEPH_CAP_FILE_EXCL))
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
return mds;
|
|
|
|
}
|
|
|
|
|
|
|
|
int ceph_get_cap_mds(struct inode *inode)
|
|
|
|
{
|
2011-11-30 17:47:09 +00:00
|
|
|
struct ceph_inode_info *ci = ceph_inode(inode);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
int mds;
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_lock(&ci->i_ceph_lock);
|
2010-06-10 21:21:36 +00:00
|
|
|
mds = __ceph_get_cap_mds(ceph_inode(inode));
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
return mds;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2011-11-30 17:47:09 +00:00
|
|
|
* Called under i_ceph_lock.
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
*/
|
|
|
|
static void __insert_cap_node(struct ceph_inode_info *ci,
|
|
|
|
struct ceph_cap *new)
|
|
|
|
{
|
|
|
|
struct rb_node **p = &ci->i_caps.rb_node;
|
|
|
|
struct rb_node *parent = NULL;
|
|
|
|
struct ceph_cap *cap = NULL;
|
|
|
|
|
|
|
|
while (*p) {
|
|
|
|
parent = *p;
|
|
|
|
cap = rb_entry(parent, struct ceph_cap, ci_node);
|
|
|
|
if (new->mds < cap->mds)
|
|
|
|
p = &(*p)->rb_left;
|
|
|
|
else if (new->mds > cap->mds)
|
|
|
|
p = &(*p)->rb_right;
|
|
|
|
else
|
|
|
|
BUG();
|
|
|
|
}
|
|
|
|
|
|
|
|
rb_link_node(&new->ci_node, parent, p);
|
|
|
|
rb_insert_color(&new->ci_node, &ci->i_caps);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* (re)set cap hold timeouts, which control the delayed release
|
|
|
|
* of unused caps back to the MDS. Should be called on cap use.
|
|
|
|
*/
|
|
|
|
static void __cap_set_timeouts(struct ceph_mds_client *mdsc,
|
|
|
|
struct ceph_inode_info *ci)
|
|
|
|
{
|
2010-04-06 22:14:15 +00:00
|
|
|
struct ceph_mount_options *ma = mdsc->fsc->mount_options;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
ci->i_hold_caps_min = round_jiffies(jiffies +
|
|
|
|
ma->caps_wanted_delay_min * HZ);
|
|
|
|
ci->i_hold_caps_max = round_jiffies(jiffies +
|
|
|
|
ma->caps_wanted_delay_max * HZ);
|
|
|
|
dout("__cap_set_timeouts %p min %lu max %lu\n", &ci->vfs_inode,
|
|
|
|
ci->i_hold_caps_min - jiffies, ci->i_hold_caps_max - jiffies);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* (Re)queue cap at the end of the delayed cap release list.
|
|
|
|
*
|
|
|
|
* If I_FLUSH is set, leave the inode at the front of the list.
|
|
|
|
*
|
2011-11-30 17:47:09 +00:00
|
|
|
* Caller holds i_ceph_lock
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
* -> we take mdsc->cap_delay_lock
|
|
|
|
*/
|
|
|
|
static void __cap_delay_requeue(struct ceph_mds_client *mdsc,
|
|
|
|
struct ceph_inode_info *ci)
|
|
|
|
{
|
|
|
|
__cap_set_timeouts(mdsc, ci);
|
|
|
|
dout("__cap_delay_requeue %p flags %d at %lu\n", &ci->vfs_inode,
|
|
|
|
ci->i_ceph_flags, ci->i_hold_caps_max);
|
|
|
|
if (!mdsc->stopping) {
|
|
|
|
spin_lock(&mdsc->cap_delay_lock);
|
|
|
|
if (!list_empty(&ci->i_cap_delay_list)) {
|
|
|
|
if (ci->i_ceph_flags & CEPH_I_FLUSH)
|
|
|
|
goto no_change;
|
|
|
|
list_del_init(&ci->i_cap_delay_list);
|
|
|
|
}
|
|
|
|
list_add_tail(&ci->i_cap_delay_list, &mdsc->cap_delay_list);
|
|
|
|
no_change:
|
|
|
|
spin_unlock(&mdsc->cap_delay_lock);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Queue an inode for immediate writeback. Mark inode with I_FLUSH,
|
|
|
|
* indicating we should send a cap message to flush dirty metadata
|
|
|
|
* asap, and move to the front of the delayed cap list.
|
|
|
|
*/
|
|
|
|
static void __cap_delay_requeue_front(struct ceph_mds_client *mdsc,
|
|
|
|
struct ceph_inode_info *ci)
|
|
|
|
{
|
|
|
|
dout("__cap_delay_requeue_front %p\n", &ci->vfs_inode);
|
|
|
|
spin_lock(&mdsc->cap_delay_lock);
|
|
|
|
ci->i_ceph_flags |= CEPH_I_FLUSH;
|
|
|
|
if (!list_empty(&ci->i_cap_delay_list))
|
|
|
|
list_del_init(&ci->i_cap_delay_list);
|
|
|
|
list_add(&ci->i_cap_delay_list, &mdsc->cap_delay_list);
|
|
|
|
spin_unlock(&mdsc->cap_delay_lock);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Cancel delayed work on cap.
|
|
|
|
*
|
2011-11-30 17:47:09 +00:00
|
|
|
* Caller must hold i_ceph_lock.
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
*/
|
|
|
|
static void __cap_delay_cancel(struct ceph_mds_client *mdsc,
|
|
|
|
struct ceph_inode_info *ci)
|
|
|
|
{
|
|
|
|
dout("__cap_delay_cancel %p\n", &ci->vfs_inode);
|
|
|
|
if (list_empty(&ci->i_cap_delay_list))
|
|
|
|
return;
|
|
|
|
spin_lock(&mdsc->cap_delay_lock);
|
|
|
|
list_del_init(&ci->i_cap_delay_list);
|
|
|
|
spin_unlock(&mdsc->cap_delay_lock);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Common issue checks for add_cap, handle_cap_grant.
|
|
|
|
*/
|
|
|
|
static void __check_cap_issue(struct ceph_inode_info *ci, struct ceph_cap *cap,
|
|
|
|
unsigned issued)
|
|
|
|
{
|
|
|
|
unsigned had = __ceph_caps_issued(ci, NULL);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Each time we receive FILE_CACHE anew, we increment
|
|
|
|
* i_rdcache_gen.
|
|
|
|
*/
|
2010-05-27 17:40:43 +00:00
|
|
|
if ((issued & (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO)) &&
|
2013-08-21 21:29:54 +00:00
|
|
|
(had & (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO)) == 0) {
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
ci->i_rdcache_gen++;
|
2013-08-21 21:29:54 +00:00
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
/*
|
2013-03-13 11:44:32 +00:00
|
|
|
* if we are newly issued FILE_SHARED, mark dir not complete; we
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
* don't know what happened to this directory while we didn't
|
|
|
|
* have the cap.
|
|
|
|
*/
|
|
|
|
if ((issued & CEPH_CAP_FILE_SHARED) &&
|
|
|
|
(had & CEPH_CAP_FILE_SHARED) == 0) {
|
|
|
|
ci->i_shared_gen++;
|
2013-02-18 08:38:14 +00:00
|
|
|
if (S_ISDIR(ci->vfs_inode.i_mode)) {
|
|
|
|
dout(" marking %p NOT complete\n", &ci->vfs_inode);
|
2013-03-13 11:44:32 +00:00
|
|
|
__ceph_dir_clear_complete(ci);
|
2013-02-18 08:38:14 +00:00
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Add a capability under the given MDS session.
|
|
|
|
*
|
|
|
|
* Caller should hold session snap_rwsem (read) and s_mutex.
|
|
|
|
*
|
|
|
|
* @fmode is the open file mode, if we are opening a file, otherwise
|
|
|
|
* it is < 0. (This is so we can atomically add the cap and add an
|
|
|
|
* open file reference to it.)
|
|
|
|
*/
|
2014-04-18 01:57:11 +00:00
|
|
|
void ceph_add_cap(struct inode *inode,
|
|
|
|
struct ceph_mds_session *session, u64 cap_id,
|
|
|
|
int fmode, unsigned issued, unsigned wanted,
|
|
|
|
unsigned seq, unsigned mseq, u64 realmino, int flags,
|
|
|
|
struct ceph_cap **new_cap)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
2010-04-06 22:14:15 +00:00
|
|
|
struct ceph_mds_client *mdsc = ceph_inode_to_client(inode)->mdsc;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
struct ceph_inode_info *ci = ceph_inode(inode);
|
|
|
|
struct ceph_cap *cap;
|
|
|
|
int mds = session->s_mds;
|
|
|
|
int actual_wanted;
|
|
|
|
|
|
|
|
dout("add_cap %p mds%d cap %llx %s seq %d\n", inode,
|
|
|
|
session->s_mds, cap_id, ceph_cap_string(issued), seq);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If we are opening the file, include file mode wanted bits
|
|
|
|
* in wanted.
|
|
|
|
*/
|
|
|
|
if (fmode >= 0)
|
|
|
|
wanted |= ceph_caps_for_mode(fmode);
|
|
|
|
|
|
|
|
cap = __get_cap_for_mds(ci, mds);
|
|
|
|
if (!cap) {
|
2014-04-18 01:57:11 +00:00
|
|
|
cap = *new_cap;
|
|
|
|
*new_cap = NULL;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
cap->issued = 0;
|
|
|
|
cap->implemented = 0;
|
|
|
|
cap->mds = mds;
|
|
|
|
cap->mds_wanted = 0;
|
2013-02-27 01:26:09 +00:00
|
|
|
cap->mseq = 0;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
cap->ci = ci;
|
|
|
|
__insert_cap_node(ci, cap);
|
|
|
|
|
|
|
|
/* add to session cap list */
|
|
|
|
cap->session = session;
|
|
|
|
spin_lock(&session->s_cap_lock);
|
|
|
|
list_add_tail(&cap->session_caps, &session->s_caps);
|
|
|
|
session->s_nr_caps++;
|
|
|
|
spin_unlock(&session->s_cap_lock);
|
2013-11-24 06:44:38 +00:00
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* auth mds of the inode changed. we received the cap export
|
|
|
|
* message, but still haven't received the cap import message.
|
|
|
|
* handle_cap_export() updated the new auth MDS' cap.
|
|
|
|
*
|
|
|
|
* "ceph_seq_cmp(seq, cap->seq) <= 0" means we are processing
|
|
|
|
* a message that was send before the cap import message. So
|
|
|
|
* don't remove caps.
|
|
|
|
*/
|
|
|
|
if (ceph_seq_cmp(seq, cap->seq) <= 0) {
|
|
|
|
WARN_ON(cap != ci->i_auth_cap);
|
|
|
|
WARN_ON(cap->cap_id != cap_id);
|
|
|
|
seq = cap->seq;
|
|
|
|
mseq = cap->mseq;
|
|
|
|
issued |= cap->issued;
|
|
|
|
flags |= CEPH_CAP_FLAG_AUTH;
|
|
|
|
}
|
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
if (!ci->i_snap_realm) {
|
|
|
|
/*
|
|
|
|
* add this inode to the appropriate snap realm
|
|
|
|
*/
|
|
|
|
struct ceph_snap_realm *realm = ceph_lookup_snap_realm(mdsc,
|
|
|
|
realmino);
|
|
|
|
if (realm) {
|
|
|
|
spin_lock(&realm->inodes_with_caps_lock);
|
|
|
|
ci->i_snap_realm = realm;
|
|
|
|
list_add(&ci->i_snap_realm_item,
|
|
|
|
&realm->inodes_with_caps);
|
|
|
|
spin_unlock(&realm->inodes_with_caps_lock);
|
|
|
|
} else {
|
|
|
|
pr_err("ceph_add_cap: couldn't find snap realm %llx\n",
|
|
|
|
realmino);
|
2010-07-16 19:00:02 +00:00
|
|
|
WARN_ON(!realm);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
__check_cap_issue(ci, cap, issued);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If we are issued caps we don't want, or the mds' wanted
|
|
|
|
* value appears to be off, queue a check so we'll release
|
|
|
|
* later and/or update the mds wanted value.
|
|
|
|
*/
|
|
|
|
actual_wanted = __ceph_caps_wanted(ci);
|
|
|
|
if ((wanted & ~actual_wanted) ||
|
|
|
|
(issued & ~actual_wanted & CEPH_CAP_ANY_WR)) {
|
|
|
|
dout(" issued %s, mds wanted %s, actual %s, queueing\n",
|
|
|
|
ceph_cap_string(issued), ceph_cap_string(wanted),
|
|
|
|
ceph_cap_string(actual_wanted));
|
|
|
|
__cap_delay_requeue(mdsc, ci);
|
|
|
|
}
|
|
|
|
|
2013-05-31 08:37:11 +00:00
|
|
|
if (flags & CEPH_CAP_FLAG_AUTH) {
|
|
|
|
if (ci->i_auth_cap == NULL ||
|
2014-03-18 02:15:29 +00:00
|
|
|
ceph_seq_cmp(ci->i_auth_cap->mseq, mseq) < 0) {
|
2013-05-31 08:37:11 +00:00
|
|
|
ci->i_auth_cap = cap;
|
2014-03-18 02:15:29 +00:00
|
|
|
cap->mds_wanted = wanted;
|
|
|
|
}
|
2013-11-24 06:44:38 +00:00
|
|
|
} else {
|
|
|
|
WARN_ON(ci->i_auth_cap == cap);
|
2013-01-04 06:28:07 +00:00
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
dout("add_cap inode %p (%llx.%llx) cap %p %s now %s seq %d mds%d\n",
|
|
|
|
inode, ceph_vinop(inode), cap, ceph_cap_string(issued),
|
|
|
|
ceph_cap_string(issued|cap->issued), seq, mds);
|
|
|
|
cap->cap_id = cap_id;
|
|
|
|
cap->issued = issued;
|
|
|
|
cap->implemented |= issued;
|
2013-11-13 06:47:19 +00:00
|
|
|
if (ceph_seq_cmp(mseq, cap->mseq) > 0)
|
2013-02-27 01:26:09 +00:00
|
|
|
cap->mds_wanted = wanted;
|
|
|
|
else
|
|
|
|
cap->mds_wanted |= wanted;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
cap->seq = seq;
|
|
|
|
cap->issue_seq = seq;
|
|
|
|
cap->mseq = mseq;
|
2009-11-09 20:05:48 +00:00
|
|
|
cap->cap_gen = session->s_cap_gen;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
if (fmode >= 0)
|
|
|
|
__ceph_get_fmode(ci, fmode);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Return true if cap has not timed out and belongs to the current
|
|
|
|
* generation of the MDS session (i.e. has not gone 'stale' due to
|
|
|
|
* us losing touch with the mds).
|
|
|
|
*/
|
|
|
|
static int __cap_is_valid(struct ceph_cap *cap)
|
|
|
|
{
|
|
|
|
unsigned long ttl;
|
2009-11-11 00:02:23 +00:00
|
|
|
u32 gen;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2012-01-13 01:48:10 +00:00
|
|
|
spin_lock(&cap->session->s_gen_ttl_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
gen = cap->session->s_cap_gen;
|
|
|
|
ttl = cap->session->s_cap_ttl;
|
2012-01-13 01:48:10 +00:00
|
|
|
spin_unlock(&cap->session->s_gen_ttl_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2009-11-09 20:05:48 +00:00
|
|
|
if (cap->cap_gen < gen || time_after_eq(jiffies, ttl)) {
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
dout("__cap_is_valid %p cap %p issued %s "
|
|
|
|
"but STALE (gen %u vs %u)\n", &cap->ci->vfs_inode,
|
2009-11-09 20:05:48 +00:00
|
|
|
cap, ceph_cap_string(cap->issued), cap->cap_gen, gen);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Return set of valid cap bits issued to us. Note that caps time
|
|
|
|
* out, and may be invalidated in bulk if the client session times out
|
|
|
|
* and session->s_cap_gen is bumped.
|
|
|
|
*/
|
|
|
|
int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented)
|
|
|
|
{
|
2014-04-18 01:57:11 +00:00
|
|
|
int have = ci->i_snap_caps;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
struct ceph_cap *cap;
|
|
|
|
struct rb_node *p;
|
|
|
|
|
|
|
|
if (implemented)
|
|
|
|
*implemented = 0;
|
|
|
|
for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) {
|
|
|
|
cap = rb_entry(p, struct ceph_cap, ci_node);
|
|
|
|
if (!__cap_is_valid(cap))
|
|
|
|
continue;
|
|
|
|
dout("__ceph_caps_issued %p cap %p issued %s\n",
|
|
|
|
&ci->vfs_inode, cap, ceph_cap_string(cap->issued));
|
|
|
|
have |= cap->issued;
|
|
|
|
if (implemented)
|
|
|
|
*implemented |= cap->implemented;
|
|
|
|
}
|
2013-07-02 04:40:20 +00:00
|
|
|
/*
|
|
|
|
* exclude caps issued by non-auth MDS, but are been revoking
|
|
|
|
* by the auth MDS. The non-auth MDS should be revoking/exporting
|
|
|
|
* these caps, but the message is delayed.
|
|
|
|
*/
|
|
|
|
if (ci->i_auth_cap) {
|
|
|
|
cap = ci->i_auth_cap;
|
|
|
|
have &= ~cap->implemented | cap->issued;
|
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
return have;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Get cap bits issued by caps other than @ocap
|
|
|
|
*/
|
|
|
|
int __ceph_caps_issued_other(struct ceph_inode_info *ci, struct ceph_cap *ocap)
|
|
|
|
{
|
|
|
|
int have = ci->i_snap_caps;
|
|
|
|
struct ceph_cap *cap;
|
|
|
|
struct rb_node *p;
|
|
|
|
|
|
|
|
for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) {
|
|
|
|
cap = rb_entry(p, struct ceph_cap, ci_node);
|
|
|
|
if (cap == ocap)
|
|
|
|
continue;
|
|
|
|
if (!__cap_is_valid(cap))
|
|
|
|
continue;
|
|
|
|
have |= cap->issued;
|
|
|
|
}
|
|
|
|
return have;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Move a cap to the end of the LRU (oldest caps at list head, newest
|
|
|
|
* at list tail).
|
|
|
|
*/
|
|
|
|
static void __touch_cap(struct ceph_cap *cap)
|
|
|
|
{
|
|
|
|
struct ceph_mds_session *s = cap->session;
|
|
|
|
|
|
|
|
spin_lock(&s->s_cap_lock);
|
ceph: fix iterate_caps removal race
We need to be able to iterate over all caps on a session with a
possibly slow callback on each cap. To allow this, we used to
prevent cap reordering while we were iterating. However, we were
not safe from races with removal: removing the 'next' cap would
make the next pointer from list_for_each_entry_safe be invalid,
and cause a lock up or similar badness.
Instead, we keep an iterator pointer in the session pointing to
the current cap. As before, we avoid reordering. For removal,
if the cap isn't the current cap we are iterating over, we are
fine. If it is, we clear cap->ci (to mark the cap as pending
removal) but leave it in the session list. In iterate_caps, we
can safely finish removal and get the next cap pointer.
While we're at it, clean up put_cap to not take a cap reservation
context, as it was never used.
Signed-off-by: Sage Weil <sage@newdream.net>
2010-02-16 19:39:45 +00:00
|
|
|
if (s->s_cap_iterator == NULL) {
|
2009-12-22 04:40:34 +00:00
|
|
|
dout("__touch_cap %p cap %p mds%d\n", &cap->ci->vfs_inode, cap,
|
|
|
|
s->s_mds);
|
|
|
|
list_move_tail(&cap->session_caps, &s->s_caps);
|
|
|
|
} else {
|
|
|
|
dout("__touch_cap %p cap %p mds%d NOP, iterating over caps\n",
|
|
|
|
&cap->ci->vfs_inode, cap, s->s_mds);
|
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
spin_unlock(&s->s_cap_lock);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check if we hold the given mask. If so, move the cap(s) to the
|
|
|
|
* front of their respective LRUs. (This is the preferred way for
|
|
|
|
* callers to check for caps they want.)
|
|
|
|
*/
|
|
|
|
int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int touch)
|
|
|
|
{
|
|
|
|
struct ceph_cap *cap;
|
|
|
|
struct rb_node *p;
|
|
|
|
int have = ci->i_snap_caps;
|
|
|
|
|
|
|
|
if ((have & mask) == mask) {
|
|
|
|
dout("__ceph_caps_issued_mask %p snap issued %s"
|
|
|
|
" (mask %s)\n", &ci->vfs_inode,
|
|
|
|
ceph_cap_string(have),
|
|
|
|
ceph_cap_string(mask));
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) {
|
|
|
|
cap = rb_entry(p, struct ceph_cap, ci_node);
|
|
|
|
if (!__cap_is_valid(cap))
|
|
|
|
continue;
|
|
|
|
if ((cap->issued & mask) == mask) {
|
|
|
|
dout("__ceph_caps_issued_mask %p cap %p issued %s"
|
|
|
|
" (mask %s)\n", &ci->vfs_inode, cap,
|
|
|
|
ceph_cap_string(cap->issued),
|
|
|
|
ceph_cap_string(mask));
|
|
|
|
if (touch)
|
|
|
|
__touch_cap(cap);
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* does a combination of caps satisfy mask? */
|
|
|
|
have |= cap->issued;
|
|
|
|
if ((have & mask) == mask) {
|
|
|
|
dout("__ceph_caps_issued_mask %p combo issued %s"
|
|
|
|
" (mask %s)\n", &ci->vfs_inode,
|
|
|
|
ceph_cap_string(cap->issued),
|
|
|
|
ceph_cap_string(mask));
|
|
|
|
if (touch) {
|
|
|
|
struct rb_node *q;
|
|
|
|
|
2011-03-31 01:57:33 +00:00
|
|
|
/* touch this + preceding caps */
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
__touch_cap(cap);
|
|
|
|
for (q = rb_first(&ci->i_caps); q != p;
|
|
|
|
q = rb_next(q)) {
|
|
|
|
cap = rb_entry(q, struct ceph_cap,
|
|
|
|
ci_node);
|
|
|
|
if (!__cap_is_valid(cap))
|
|
|
|
continue;
|
|
|
|
__touch_cap(cap);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Return true if mask caps are currently being revoked by an MDS.
|
|
|
|
*/
|
2013-07-02 04:40:21 +00:00
|
|
|
int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
|
|
|
|
struct ceph_cap *ocap, int mask)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
|
|
|
struct ceph_cap *cap;
|
|
|
|
struct rb_node *p;
|
|
|
|
|
|
|
|
for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) {
|
|
|
|
cap = rb_entry(p, struct ceph_cap, ci_node);
|
2013-11-22 05:50:45 +00:00
|
|
|
if (cap != ocap &&
|
2013-07-02 04:40:21 +00:00
|
|
|
(cap->implemented & ~cap->issued & mask))
|
|
|
|
return 1;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
2013-07-02 04:40:21 +00:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
int ceph_caps_revoking(struct ceph_inode_info *ci, int mask)
|
|
|
|
{
|
|
|
|
struct inode *inode = &ci->vfs_inode;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
spin_lock(&ci->i_ceph_lock);
|
|
|
|
ret = __ceph_caps_revoking_other(ci, NULL, mask);
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
dout("ceph_caps_revoking %p %s = %d\n", inode,
|
|
|
|
ceph_cap_string(mask), ret);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
int __ceph_caps_used(struct ceph_inode_info *ci)
|
|
|
|
{
|
|
|
|
int used = 0;
|
|
|
|
if (ci->i_pin_ref)
|
|
|
|
used |= CEPH_CAP_PIN;
|
|
|
|
if (ci->i_rd_ref)
|
|
|
|
used |= CEPH_CAP_FILE_RD;
|
2015-06-16 12:48:56 +00:00
|
|
|
if (ci->i_rdcache_ref ||
|
|
|
|
(!S_ISDIR(ci->vfs_inode.i_mode) && /* ignore readdir cache */
|
|
|
|
ci->vfs_inode.i_data.nrpages))
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
used |= CEPH_CAP_FILE_CACHE;
|
|
|
|
if (ci->i_wr_ref)
|
|
|
|
used |= CEPH_CAP_FILE_WR;
|
2011-05-11 10:29:54 +00:00
|
|
|
if (ci->i_wb_ref || ci->i_wrbuffer_ref)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
used |= CEPH_CAP_FILE_BUFFER;
|
|
|
|
return used;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* wanted, by virtue of open file modes
|
|
|
|
*/
|
|
|
|
int __ceph_caps_file_wanted(struct ceph_inode_info *ci)
|
|
|
|
{
|
2016-06-06 08:01:39 +00:00
|
|
|
int i, bits = 0;
|
|
|
|
for (i = 0; i < CEPH_FILE_MODE_BITS; i++) {
|
|
|
|
if (ci->i_nr_by_mode[i])
|
|
|
|
bits |= 1 << i;
|
|
|
|
}
|
|
|
|
if (bits == 0)
|
|
|
|
return 0;
|
|
|
|
return ceph_caps_for_mode(bits >> 1);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Return caps we have registered with the MDS(s) as 'wanted'.
|
|
|
|
*/
|
2017-01-29 14:15:47 +00:00
|
|
|
int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
|
|
|
struct ceph_cap *cap;
|
|
|
|
struct rb_node *p;
|
|
|
|
int mds_wanted = 0;
|
|
|
|
|
|
|
|
for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) {
|
|
|
|
cap = rb_entry(p, struct ceph_cap, ci_node);
|
2017-01-29 14:15:47 +00:00
|
|
|
if (check && !__cap_is_valid(cap))
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
continue;
|
2014-03-08 01:51:45 +00:00
|
|
|
if (cap == ci->i_auth_cap)
|
|
|
|
mds_wanted |= cap->mds_wanted;
|
|
|
|
else
|
|
|
|
mds_wanted |= (cap->mds_wanted & ~CEPH_CAP_ANY_FILE_WR);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
return mds_wanted;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2011-11-30 17:47:09 +00:00
|
|
|
* called under i_ceph_lock
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
*/
|
|
|
|
static int __ceph_is_any_caps(struct ceph_inode_info *ci)
|
|
|
|
{
|
2014-04-18 01:57:11 +00:00
|
|
|
return !RB_EMPTY_ROOT(&ci->i_caps);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
2013-11-30 04:47:41 +00:00
|
|
|
int ceph_is_any_caps(struct inode *inode)
|
|
|
|
{
|
|
|
|
struct ceph_inode_info *ci = ceph_inode(inode);
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
spin_lock(&ci->i_ceph_lock);
|
|
|
|
ret = __ceph_is_any_caps(ci);
|
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2015-03-23 12:12:20 +00:00
|
|
|
static void drop_inode_snap_realm(struct ceph_inode_info *ci)
|
|
|
|
{
|
|
|
|
struct ceph_snap_realm *realm = ci->i_snap_realm;
|
|
|
|
spin_lock(&realm->inodes_with_caps_lock);
|
|
|
|
list_del_init(&ci->i_snap_realm_item);
|
|
|
|
ci->i_snap_realm_counter++;
|
|
|
|
ci->i_snap_realm = NULL;
|
|
|
|
spin_unlock(&realm->inodes_with_caps_lock);
|
|
|
|
ceph_put_snap_realm(ceph_sb_to_client(ci->vfs_inode.i_sb)->mdsc,
|
|
|
|
realm);
|
|
|
|
}
|
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
/*
|
2010-05-12 03:56:31 +00:00
|
|
|
* Remove a cap. Take steps to deal with a racing iterate_session_caps.
|
|
|
|
*
|
2011-11-30 17:47:09 +00:00
|
|
|
* caller should hold i_ceph_lock.
|
2010-02-22 21:59:00 +00:00
|
|
|
* caller will not hold session s_mutex if called from destroy_inode.
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
*/
|
2013-09-22 02:15:58 +00:00
|
|
|
void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
|
|
|
struct ceph_mds_session *session = cap->session;
|
|
|
|
struct ceph_inode_info *ci = cap->ci;
|
2010-03-26 09:40:33 +00:00
|
|
|
struct ceph_mds_client *mdsc =
|
2010-04-06 22:14:15 +00:00
|
|
|
ceph_sb_to_client(ci->vfs_inode.i_sb)->mdsc;
|
2010-05-12 03:56:31 +00:00
|
|
|
int removed = 0;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
dout("__ceph_remove_cap %p from %p\n", cap, &ci->vfs_inode);
|
|
|
|
|
ceph: fix iterate_caps removal race
We need to be able to iterate over all caps on a session with a
possibly slow callback on each cap. To allow this, we used to
prevent cap reordering while we were iterating. However, we were
not safe from races with removal: removing the 'next' cap would
make the next pointer from list_for_each_entry_safe be invalid,
and cause a lock up or similar badness.
Instead, we keep an iterator pointer in the session pointing to
the current cap. As before, we avoid reordering. For removal,
if the cap isn't the current cap we are iterating over, we are
fine. If it is, we clear cap->ci (to mark the cap as pending
removal) but leave it in the session list. In iterate_caps, we
can safely finish removal and get the next cap pointer.
While we're at it, clean up put_cap to not take a cap reservation
context, as it was never used.
Signed-off-by: Sage Weil <sage@newdream.net>
2010-02-16 19:39:45 +00:00
|
|
|
/* remove from session list */
|
|
|
|
spin_lock(&session->s_cap_lock);
|
|
|
|
if (session->s_cap_iterator == cap) {
|
|
|
|
/* not yet, we are iterating over this very cap */
|
|
|
|
dout("__ceph_remove_cap delaying %p removal from session %p\n",
|
|
|
|
cap, cap->session);
|
|
|
|
} else {
|
|
|
|
list_del_init(&cap->session_caps);
|
|
|
|
session->s_nr_caps--;
|
|
|
|
cap->session = NULL;
|
2010-05-12 03:56:31 +00:00
|
|
|
removed = 1;
|
ceph: fix iterate_caps removal race
We need to be able to iterate over all caps on a session with a
possibly slow callback on each cap. To allow this, we used to
prevent cap reordering while we were iterating. However, we were
not safe from races with removal: removing the 'next' cap would
make the next pointer from list_for_each_entry_safe be invalid,
and cause a lock up or similar badness.
Instead, we keep an iterator pointer in the session pointing to
the current cap. As before, we avoid reordering. For removal,
if the cap isn't the current cap we are iterating over, we are
fine. If it is, we clear cap->ci (to mark the cap as pending
removal) but leave it in the session list. In iterate_caps, we
can safely finish removal and get the next cap pointer.
While we're at it, clean up put_cap to not take a cap reservation
context, as it was never used.
Signed-off-by: Sage Weil <sage@newdream.net>
2010-02-16 19:39:45 +00:00
|
|
|
}
|
2010-05-12 03:56:31 +00:00
|
|
|
/* protect backpointer with s_cap_lock: see iterate_session_caps */
|
|
|
|
cap->ci = NULL;
|
2015-05-14 09:22:42 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* s_cap_reconnect is protected by s_cap_lock. no one changes
|
|
|
|
* s_cap_gen while session is in the reconnect state.
|
|
|
|
*/
|
|
|
|
if (queue_release &&
|
|
|
|
(!session->s_cap_reconnect || cap->cap_gen == session->s_cap_gen)) {
|
|
|
|
cap->queue_release = 1;
|
|
|
|
if (removed) {
|
|
|
|
list_add_tail(&cap->session_caps,
|
|
|
|
&session->s_cap_releases);
|
|
|
|
session->s_num_cap_releases++;
|
|
|
|
removed = 0;
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
cap->queue_release = 0;
|
|
|
|
}
|
|
|
|
cap->cap_ino = ci->i_vino.ino;
|
|
|
|
|
ceph: fix iterate_caps removal race
We need to be able to iterate over all caps on a session with a
possibly slow callback on each cap. To allow this, we used to
prevent cap reordering while we were iterating. However, we were
not safe from races with removal: removing the 'next' cap would
make the next pointer from list_for_each_entry_safe be invalid,
and cause a lock up or similar badness.
Instead, we keep an iterator pointer in the session pointing to
the current cap. As before, we avoid reordering. For removal,
if the cap isn't the current cap we are iterating over, we are
fine. If it is, we clear cap->ci (to mark the cap as pending
removal) but leave it in the session list. In iterate_caps, we
can safely finish removal and get the next cap pointer.
While we're at it, clean up put_cap to not take a cap reservation
context, as it was never used.
Signed-off-by: Sage Weil <sage@newdream.net>
2010-02-16 19:39:45 +00:00
|
|
|
spin_unlock(&session->s_cap_lock);
|
|
|
|
|
2010-05-12 03:56:31 +00:00
|
|
|
/* remove from inode list */
|
|
|
|
rb_erase(&cap->ci_node, &ci->i_caps);
|
|
|
|
if (ci->i_auth_cap == cap)
|
|
|
|
ci->i_auth_cap = NULL;
|
|
|
|
|
|
|
|
if (removed)
|
2010-06-17 23:16:12 +00:00
|
|
|
ceph_put_cap(mdsc, cap);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2015-03-23 12:12:20 +00:00
|
|
|
/* when reconnect denied, we remove session caps forcibly,
|
|
|
|
* i_wr_ref can be non-zero. If there are ongoing write,
|
|
|
|
* keep i_snap_realm.
|
|
|
|
*/
|
|
|
|
if (!__ceph_is_any_caps(ci) && ci->i_wr_ref == 0 && ci->i_snap_realm)
|
|
|
|
drop_inode_snap_realm(ci);
|
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
if (!__ceph_is_any_real_caps(ci))
|
|
|
|
__cap_delay_cancel(mdsc, ci);
|
|
|
|
}
|
|
|
|
|
2016-11-10 12:42:03 +00:00
|
|
|
struct cap_msg_args {
|
|
|
|
struct ceph_mds_session *session;
|
|
|
|
u64 ino, cid, follows;
|
|
|
|
u64 flush_tid, oldest_flush_tid, size, max_size;
|
|
|
|
u64 xattr_version;
|
|
|
|
struct ceph_buffer *xattr_buf;
|
|
|
|
struct timespec atime, mtime, ctime;
|
|
|
|
int op, caps, wanted, dirty;
|
|
|
|
u32 seq, issue_seq, mseq, time_warp_seq;
|
2016-11-10 12:42:06 +00:00
|
|
|
u32 flags;
|
2016-11-10 12:42:03 +00:00
|
|
|
kuid_t uid;
|
|
|
|
kgid_t gid;
|
|
|
|
umode_t mode;
|
|
|
|
bool inline_data;
|
|
|
|
};
|
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
/*
|
|
|
|
* Build and send a cap message to the given MDS.
|
|
|
|
*
|
|
|
|
* Caller should be holding s_mutex.
|
|
|
|
*/
|
2016-11-10 12:42:03 +00:00
|
|
|
static int send_cap_msg(struct cap_msg_args *arg)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
|
|
|
struct ceph_mds_caps *fc;
|
|
|
|
struct ceph_msg *msg;
|
2014-11-14 14:39:13 +00:00
|
|
|
void *p;
|
|
|
|
size_t extra_len;
|
2016-11-10 12:42:05 +00:00
|
|
|
struct timespec zerotime = {0};
|
2017-04-13 15:07:04 +00:00
|
|
|
struct ceph_osd_client *osdc = &arg->session->s_mdsc->fsc->client->osdc;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
dout("send_cap_msg %s %llx %llx caps %s wanted %s dirty %s"
|
2015-06-10 03:09:32 +00:00
|
|
|
" seq %u/%u tid %llu/%llu mseq %u follows %lld size %llu/%llu"
|
2016-11-10 12:42:03 +00:00
|
|
|
" xattr_ver %llu xattr_len %d\n", ceph_cap_op_name(arg->op),
|
|
|
|
arg->cid, arg->ino, ceph_cap_string(arg->caps),
|
|
|
|
ceph_cap_string(arg->wanted), ceph_cap_string(arg->dirty),
|
|
|
|
arg->seq, arg->issue_seq, arg->flush_tid, arg->oldest_flush_tid,
|
|
|
|
arg->mseq, arg->follows, arg->size, arg->max_size,
|
|
|
|
arg->xattr_version,
|
|
|
|
arg->xattr_buf ? (int)arg->xattr_buf->vec.iov_len : 0);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2015-06-10 03:09:32 +00:00
|
|
|
/* flock buffer size + inline version + inline data size +
|
|
|
|
* osd_epoch_barrier + oldest_flush_tid */
|
2016-11-10 12:42:05 +00:00
|
|
|
extra_len = 4 + 8 + 4 + 4 + 8 + 4 + 4 + 4 + 8 + 8 + 4;
|
2014-11-14 14:39:13 +00:00
|
|
|
msg = ceph_msg_new(CEPH_MSG_CLIENT_CAPS, sizeof(*fc) + extra_len,
|
|
|
|
GFP_NOFS, false);
|
2010-04-01 23:06:19 +00:00
|
|
|
if (!msg)
|
|
|
|
return -ENOMEM;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2016-11-10 12:42:05 +00:00
|
|
|
msg->hdr.version = cpu_to_le16(10);
|
2016-11-10 12:42:03 +00:00
|
|
|
msg->hdr.tid = cpu_to_le64(arg->flush_tid);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2009-12-22 19:24:33 +00:00
|
|
|
fc = msg->front.iov_base;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
memset(fc, 0, sizeof(*fc));
|
|
|
|
|
2016-11-10 12:42:03 +00:00
|
|
|
fc->cap_id = cpu_to_le64(arg->cid);
|
|
|
|
fc->op = cpu_to_le32(arg->op);
|
|
|
|
fc->seq = cpu_to_le32(arg->seq);
|
|
|
|
fc->issue_seq = cpu_to_le32(arg->issue_seq);
|
|
|
|
fc->migrate_seq = cpu_to_le32(arg->mseq);
|
|
|
|
fc->caps = cpu_to_le32(arg->caps);
|
|
|
|
fc->wanted = cpu_to_le32(arg->wanted);
|
|
|
|
fc->dirty = cpu_to_le32(arg->dirty);
|
|
|
|
fc->ino = cpu_to_le64(arg->ino);
|
|
|
|
fc->snap_follows = cpu_to_le64(arg->follows);
|
|
|
|
|
|
|
|
fc->size = cpu_to_le64(arg->size);
|
|
|
|
fc->max_size = cpu_to_le64(arg->max_size);
|
|
|
|
ceph_encode_timespec(&fc->mtime, &arg->mtime);
|
|
|
|
ceph_encode_timespec(&fc->atime, &arg->atime);
|
|
|
|
ceph_encode_timespec(&fc->ctime, &arg->ctime);
|
|
|
|
fc->time_warp_seq = cpu_to_le32(arg->time_warp_seq);
|
|
|
|
|
|
|
|
fc->uid = cpu_to_le32(from_kuid(&init_user_ns, arg->uid));
|
|
|
|
fc->gid = cpu_to_le32(from_kgid(&init_user_ns, arg->gid));
|
|
|
|
fc->mode = cpu_to_le32(arg->mode);
|
|
|
|
|
|
|
|
fc->xattr_version = cpu_to_le64(arg->xattr_version);
|
|
|
|
if (arg->xattr_buf) {
|
|
|
|
msg->middle = ceph_buffer_get(arg->xattr_buf);
|
|
|
|
fc->xattr_len = cpu_to_le32(arg->xattr_buf->vec.iov_len);
|
|
|
|
msg->hdr.middle_len = cpu_to_le32(arg->xattr_buf->vec.iov_len);
|
2016-11-10 12:42:02 +00:00
|
|
|
}
|
|
|
|
|
2014-11-14 14:39:13 +00:00
|
|
|
p = fc + 1;
|
2016-11-10 12:42:05 +00:00
|
|
|
/* flock buffer size (version 2) */
|
2014-11-14 14:39:13 +00:00
|
|
|
ceph_encode_32(&p, 0);
|
2016-11-10 12:42:05 +00:00
|
|
|
/* inline version (version 4) */
|
2016-11-10 12:42:03 +00:00
|
|
|
ceph_encode_64(&p, arg->inline_data ? 0 : CEPH_INLINE_NONE);
|
2014-11-14 14:39:13 +00:00
|
|
|
/* inline data size */
|
|
|
|
ceph_encode_32(&p, 0);
|
2017-04-13 15:07:04 +00:00
|
|
|
/*
|
|
|
|
* osd_epoch_barrier (version 5)
|
|
|
|
* The epoch_barrier is protected osdc->lock, so READ_ONCE here in
|
|
|
|
* case it was recently changed
|
|
|
|
*/
|
|
|
|
ceph_encode_32(&p, READ_ONCE(osdc->epoch_barrier));
|
2016-11-10 12:42:05 +00:00
|
|
|
/* oldest_flush_tid (version 6) */
|
2016-11-10 12:42:03 +00:00
|
|
|
ceph_encode_64(&p, arg->oldest_flush_tid);
|
2014-11-14 14:39:13 +00:00
|
|
|
|
2016-11-10 12:42:05 +00:00
|
|
|
/*
|
|
|
|
* caller_uid/caller_gid (version 7)
|
|
|
|
*
|
|
|
|
* Currently, we don't properly track which caller dirtied the caps
|
|
|
|
* last, and force a flush of them when there is a conflict. For now,
|
|
|
|
* just set this to 0:0, to emulate how the MDS has worked up to now.
|
|
|
|
*/
|
|
|
|
ceph_encode_32(&p, 0);
|
|
|
|
ceph_encode_32(&p, 0);
|
|
|
|
|
|
|
|
/* pool namespace (version 8) (mds always ignores this) */
|
|
|
|
ceph_encode_32(&p, 0);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* btime and change_attr (version 9)
|
|
|
|
*
|
|
|
|
* We just zero these out for now, as the MDS ignores them unless
|
|
|
|
* the requisite feature flags are set (which we don't do yet).
|
|
|
|
*/
|
|
|
|
ceph_encode_timespec(p, &zerotime);
|
|
|
|
p += sizeof(struct ceph_timespec);
|
|
|
|
ceph_encode_64(&p, 0);
|
|
|
|
|
|
|
|
/* Advisory flags (version 10) */
|
2016-11-10 12:42:06 +00:00
|
|
|
ceph_encode_32(&p, arg->flags);
|
2016-11-10 12:42:05 +00:00
|
|
|
|
2016-11-10 12:42:03 +00:00
|
|
|
ceph_con_send(&arg->session->s_con, msg);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2010-02-22 21:59:00 +00:00
|
|
|
* Queue cap releases when an inode is dropped from our cache. Since
|
2011-11-30 17:47:09 +00:00
|
|
|
* inode is about to be destroyed, there is no need for i_ceph_lock.
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
*/
|
|
|
|
void ceph_queue_caps_release(struct inode *inode)
|
|
|
|
{
|
|
|
|
struct ceph_inode_info *ci = ceph_inode(inode);
|
|
|
|
struct rb_node *p;
|
|
|
|
|
|
|
|
p = rb_first(&ci->i_caps);
|
|
|
|
while (p) {
|
|
|
|
struct ceph_cap *cap = rb_entry(p, struct ceph_cap, ci_node);
|
|
|
|
p = rb_next(p);
|
2013-09-22 02:15:58 +00:00
|
|
|
__ceph_remove_cap(cap, true);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Send a cap msg on the given inode. Update our caps state, then
|
2011-11-30 17:47:09 +00:00
|
|
|
* drop i_ceph_lock and send the message.
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
*
|
|
|
|
* Make note of max_size reported/requested from mds, revoked caps
|
|
|
|
* that have now been implemented.
|
|
|
|
*
|
|
|
|
* Make half-hearted attempt ot to invalidate page cache if we are
|
|
|
|
* dropping RDCACHE. Note that this will leave behind locked pages
|
|
|
|
* that we'll then need to deal with elsewhere.
|
|
|
|
*
|
|
|
|
* Return non-zero if delayed release, or we experienced an error
|
|
|
|
* such that the caller should requeue + retry later.
|
|
|
|
*
|
2011-11-30 17:47:09 +00:00
|
|
|
* called with i_ceph_lock, then drops it.
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
* caller should hold snap_rwsem (read), s_mutex.
|
|
|
|
*/
|
|
|
|
static int __send_cap(struct ceph_mds_client *mdsc, struct ceph_cap *cap,
|
2016-11-10 12:42:06 +00:00
|
|
|
int op, bool sync, int used, int want, int retain,
|
|
|
|
int flushing, u64 flush_tid, u64 oldest_flush_tid)
|
2011-11-30 17:47:09 +00:00
|
|
|
__releases(cap->ci->i_ceph_lock)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
|
|
|
struct ceph_inode_info *ci = cap->ci;
|
|
|
|
struct inode *inode = &ci->vfs_inode;
|
2016-11-10 12:42:03 +00:00
|
|
|
struct cap_msg_args arg;
|
|
|
|
int held, revoking, dropping;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
int wake = 0;
|
|
|
|
int delayed = 0;
|
|
|
|
int ret;
|
|
|
|
|
2010-02-09 21:41:47 +00:00
|
|
|
held = cap->issued | cap->implemented;
|
|
|
|
revoking = cap->implemented & ~cap->issued;
|
|
|
|
retain &= ~revoking;
|
|
|
|
dropping = cap->issued & ~retain;
|
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
dout("__send_cap %p cap %p session %p %s -> %s (revoking %s)\n",
|
|
|
|
inode, cap, cap->session,
|
|
|
|
ceph_cap_string(held), ceph_cap_string(held & retain),
|
|
|
|
ceph_cap_string(revoking));
|
|
|
|
BUG_ON((retain & CEPH_CAP_PIN) == 0);
|
|
|
|
|
2016-11-10 12:42:03 +00:00
|
|
|
arg.session = cap->session;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
/* don't release wanted unless we've waited a bit. */
|
|
|
|
if ((ci->i_ceph_flags & CEPH_I_NODELAY) == 0 &&
|
|
|
|
time_before(jiffies, ci->i_hold_caps_min)) {
|
|
|
|
dout(" delaying issued %s -> %s, wanted %s -> %s on send\n",
|
|
|
|
ceph_cap_string(cap->issued),
|
|
|
|
ceph_cap_string(cap->issued & retain),
|
|
|
|
ceph_cap_string(cap->mds_wanted),
|
|
|
|
ceph_cap_string(want));
|
|
|
|
want |= cap->mds_wanted;
|
|
|
|
retain |= cap->issued;
|
|
|
|
delayed = 1;
|
|
|
|
}
|
|
|
|
ci->i_ceph_flags &= ~(CEPH_I_NODELAY | CEPH_I_FLUSH);
|
2017-01-12 09:18:00 +00:00
|
|
|
if (want & ~cap->mds_wanted) {
|
|
|
|
/* user space may open/close single file frequently.
|
|
|
|
* This avoids droping mds_wanted immediately after
|
|
|
|
* requesting new mds_wanted.
|
|
|
|
*/
|
|
|
|
__cap_set_timeouts(mdsc, ci);
|
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
cap->issued &= retain; /* drop bits we don't want */
|
|
|
|
if (cap->implemented & ~cap->issued) {
|
|
|
|
/*
|
|
|
|
* Wake up any waiters on wanted -> needed transition.
|
|
|
|
* This is due to the weird transition from buffered
|
|
|
|
* to sync IO... we need to flush dirty pages _before_
|
|
|
|
* allowing sync writes to avoid reordering.
|
|
|
|
*/
|
|
|
|
wake = 1;
|
|
|
|
}
|
|
|
|
cap->implemented &= cap->issued | used;
|
|
|
|
cap->mds_wanted = want;
|
|
|
|
|
2016-11-10 12:42:03 +00:00
|
|
|
arg.ino = ceph_vino(inode).ino;
|
|
|
|
arg.cid = cap->cap_id;
|
|
|
|
arg.follows = flushing ? ci->i_head_snapc->seq : 0;
|
|
|
|
arg.flush_tid = flush_tid;
|
|
|
|
arg.oldest_flush_tid = oldest_flush_tid;
|
|
|
|
|
|
|
|
arg.size = inode->i_size;
|
|
|
|
ci->i_reported_size = arg.size;
|
|
|
|
arg.max_size = ci->i_wanted_max_size;
|
|
|
|
ci->i_requested_max_size = arg.max_size;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2010-08-22 22:16:41 +00:00
|
|
|
if (flushing & CEPH_CAP_XATTR_EXCL) {
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
__ceph_build_xattrs_blob(ci);
|
2016-11-10 12:42:03 +00:00
|
|
|
arg.xattr_version = ci->i_xattrs.version;
|
|
|
|
arg.xattr_buf = ci->i_xattrs.blob;
|
|
|
|
} else {
|
|
|
|
arg.xattr_buf = NULL;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
2016-11-10 12:42:03 +00:00
|
|
|
arg.mtime = inode->i_mtime;
|
|
|
|
arg.atime = inode->i_atime;
|
|
|
|
arg.ctime = inode->i_ctime;
|
|
|
|
|
|
|
|
arg.op = op;
|
|
|
|
arg.caps = cap->implemented;
|
|
|
|
arg.wanted = want;
|
|
|
|
arg.dirty = flushing;
|
|
|
|
|
|
|
|
arg.seq = cap->seq;
|
|
|
|
arg.issue_seq = cap->issue_seq;
|
|
|
|
arg.mseq = cap->mseq;
|
|
|
|
arg.time_warp_seq = ci->i_time_warp_seq;
|
|
|
|
|
|
|
|
arg.uid = inode->i_uid;
|
|
|
|
arg.gid = inode->i_gid;
|
|
|
|
arg.mode = inode->i_mode;
|
|
|
|
|
|
|
|
arg.inline_data = ci->i_inline_version != CEPH_INLINE_NONE;
|
2016-11-10 12:42:06 +00:00
|
|
|
arg.flags = 0;
|
|
|
|
if (sync)
|
|
|
|
arg.flags |= CEPH_CLIENT_CAPS_SYNC;
|
2014-11-14 14:39:13 +00:00
|
|
|
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2016-11-10 12:42:03 +00:00
|
|
|
ret = send_cap_msg(&arg);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
if (ret < 0) {
|
|
|
|
dout("error sending cap msg, must requeue %p\n", inode);
|
|
|
|
delayed = 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (wake)
|
2010-07-27 20:11:08 +00:00
|
|
|
wake_up_all(&ci->i_cap_wq);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
return delayed;
|
|
|
|
}
|
|
|
|
|
2016-07-04 10:06:41 +00:00
|
|
|
static inline int __send_flush_snap(struct inode *inode,
|
|
|
|
struct ceph_mds_session *session,
|
|
|
|
struct ceph_cap_snap *capsnap,
|
|
|
|
u32 mseq, u64 oldest_flush_tid)
|
|
|
|
{
|
2016-11-10 12:42:03 +00:00
|
|
|
struct cap_msg_args arg;
|
|
|
|
|
|
|
|
arg.session = session;
|
|
|
|
arg.ino = ceph_vino(inode).ino;
|
|
|
|
arg.cid = 0;
|
|
|
|
arg.follows = capsnap->follows;
|
|
|
|
arg.flush_tid = capsnap->cap_flush.tid;
|
|
|
|
arg.oldest_flush_tid = oldest_flush_tid;
|
|
|
|
|
|
|
|
arg.size = capsnap->size;
|
|
|
|
arg.max_size = 0;
|
|
|
|
arg.xattr_version = capsnap->xattr_version;
|
|
|
|
arg.xattr_buf = capsnap->xattr_blob;
|
|
|
|
|
|
|
|
arg.atime = capsnap->atime;
|
|
|
|
arg.mtime = capsnap->mtime;
|
|
|
|
arg.ctime = capsnap->ctime;
|
|
|
|
|
|
|
|
arg.op = CEPH_CAP_OP_FLUSHSNAP;
|
|
|
|
arg.caps = capsnap->issued;
|
|
|
|
arg.wanted = 0;
|
|
|
|
arg.dirty = capsnap->dirty;
|
|
|
|
|
|
|
|
arg.seq = 0;
|
|
|
|
arg.issue_seq = 0;
|
|
|
|
arg.mseq = mseq;
|
|
|
|
arg.time_warp_seq = capsnap->time_warp_seq;
|
|
|
|
|
|
|
|
arg.uid = capsnap->uid;
|
|
|
|
arg.gid = capsnap->gid;
|
|
|
|
arg.mode = capsnap->mode;
|
|
|
|
|
|
|
|
arg.inline_data = capsnap->inline_data;
|
2016-11-10 12:42:06 +00:00
|
|
|
arg.flags = 0;
|
2016-11-10 12:42:03 +00:00
|
|
|
|
|
|
|
return send_cap_msg(&arg);
|
2016-07-04 10:06:41 +00:00
|
|
|
}
|
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
/*
|
|
|
|
* When a snapshot is taken, clients accumulate dirty metadata on
|
|
|
|
* inodes with capabilities in ceph_cap_snaps to describe the file
|
|
|
|
* state at the time the snapshot was taken. This must be flushed
|
|
|
|
* asynchronously back to the MDS once sync writes complete and dirty
|
|
|
|
* data is written out.
|
|
|
|
*
|
2011-11-30 17:47:09 +00:00
|
|
|
* Called under i_ceph_lock. Takes s_mutex as needed.
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
*/
|
2016-07-05 13:08:07 +00:00
|
|
|
static void __ceph_flush_snaps(struct ceph_inode_info *ci,
|
|
|
|
struct ceph_mds_session *session)
|
2011-11-30 17:47:09 +00:00
|
|
|
__releases(ci->i_ceph_lock)
|
|
|
|
__acquires(ci->i_ceph_lock)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
|
|
|
struct inode *inode = &ci->vfs_inode;
|
2016-07-05 13:08:07 +00:00
|
|
|
struct ceph_mds_client *mdsc = session->s_mdsc;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
struct ceph_cap_snap *capsnap;
|
2016-07-05 13:08:07 +00:00
|
|
|
u64 oldest_flush_tid = 0;
|
|
|
|
u64 first_tid = 1, last_tid = 0;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2016-07-05 13:08:07 +00:00
|
|
|
dout("__flush_snaps %p session %p\n", inode, session);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
list_for_each_entry(capsnap, &ci->i_cap_snaps, ci_item) {
|
|
|
|
/*
|
|
|
|
* we need to wait for sync writes to complete and for dirty
|
|
|
|
* pages to be written out.
|
|
|
|
*/
|
|
|
|
if (capsnap->dirty_pages || capsnap->writing)
|
2010-09-14 22:50:59 +00:00
|
|
|
break;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2015-05-01 08:57:16 +00:00
|
|
|
/* should be removed by ceph_try_drop_cap_snap() */
|
|
|
|
BUG_ON(!capsnap->need_flush);
|
2010-04-01 16:33:46 +00:00
|
|
|
|
2010-09-17 15:03:08 +00:00
|
|
|
/* only flush each capsnap once */
|
2016-07-04 10:06:41 +00:00
|
|
|
if (capsnap->cap_flush.tid > 0) {
|
2016-07-05 13:08:07 +00:00
|
|
|
dout(" already flushed %p, skipping\n", capsnap);
|
2010-09-17 15:03:08 +00:00
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
2015-06-09 07:48:57 +00:00
|
|
|
spin_lock(&mdsc->cap_dirty_lock);
|
2016-07-04 10:06:41 +00:00
|
|
|
capsnap->cap_flush.tid = ++mdsc->last_cap_flush_tid;
|
|
|
|
list_add_tail(&capsnap->cap_flush.g_list,
|
|
|
|
&mdsc->cap_flush_list);
|
2016-07-05 13:08:07 +00:00
|
|
|
if (oldest_flush_tid == 0)
|
|
|
|
oldest_flush_tid = __get_oldest_flush_tid(mdsc);
|
2016-07-04 10:06:41 +00:00
|
|
|
if (list_empty(&ci->i_flushing_item)) {
|
|
|
|
list_add_tail(&ci->i_flushing_item,
|
|
|
|
&session->s_cap_flushing);
|
|
|
|
}
|
2015-06-09 07:48:57 +00:00
|
|
|
spin_unlock(&mdsc->cap_dirty_lock);
|
|
|
|
|
2016-07-04 10:06:41 +00:00
|
|
|
list_add_tail(&capsnap->cap_flush.i_list,
|
|
|
|
&ci->i_cap_flush_list);
|
|
|
|
|
2016-07-05 13:08:07 +00:00
|
|
|
if (first_tid == 1)
|
|
|
|
first_tid = capsnap->cap_flush.tid;
|
|
|
|
last_tid = capsnap->cap_flush.tid;
|
|
|
|
}
|
|
|
|
|
|
|
|
ci->i_ceph_flags &= ~CEPH_I_FLUSH_SNAPS;
|
|
|
|
|
|
|
|
while (first_tid <= last_tid) {
|
|
|
|
struct ceph_cap *cap = ci->i_auth_cap;
|
|
|
|
struct ceph_cap_flush *cf;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
if (!(cap && cap->session == session)) {
|
|
|
|
dout("__flush_snaps %p auth cap %p not mds%d, "
|
|
|
|
"stop\n", inode, cap, session->s_mds);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = -ENOENT;
|
|
|
|
list_for_each_entry(cf, &ci->i_cap_flush_list, i_list) {
|
|
|
|
if (cf->tid >= first_tid) {
|
|
|
|
ret = 0;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (ret < 0)
|
|
|
|
break;
|
|
|
|
|
|
|
|
first_tid = cf->tid + 1;
|
|
|
|
|
|
|
|
capsnap = container_of(cf, struct ceph_cap_snap, cap_flush);
|
2017-03-03 09:15:07 +00:00
|
|
|
refcount_inc(&capsnap->nref);
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2016-07-05 13:08:07 +00:00
|
|
|
dout("__flush_snaps %p capsnap %p tid %llu %s\n",
|
|
|
|
inode, capsnap, cf->tid, ceph_cap_string(capsnap->dirty));
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2016-07-05 13:08:07 +00:00
|
|
|
ret = __send_flush_snap(inode, session, capsnap, cap->mseq,
|
|
|
|
oldest_flush_tid);
|
|
|
|
if (ret < 0) {
|
|
|
|
pr_err("__flush_snaps: error sending cap flushsnap, "
|
|
|
|
"ino (%llx.%llx) tid %llu follows %llu\n",
|
|
|
|
ceph_vinop(inode), cf->tid, capsnap->follows);
|
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2016-07-05 13:08:07 +00:00
|
|
|
ceph_put_cap_snap(capsnap);
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_lock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
2016-07-05 13:08:07 +00:00
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2016-07-05 13:08:07 +00:00
|
|
|
void ceph_flush_snaps(struct ceph_inode_info *ci,
|
|
|
|
struct ceph_mds_session **psession)
|
|
|
|
{
|
|
|
|
struct inode *inode = &ci->vfs_inode;
|
|
|
|
struct ceph_mds_client *mdsc = ceph_inode_to_client(inode)->mdsc;
|
2016-08-04 00:43:33 +00:00
|
|
|
struct ceph_mds_session *session = NULL;
|
2016-07-05 13:08:07 +00:00
|
|
|
int mds;
|
2016-08-04 00:43:33 +00:00
|
|
|
|
2016-07-05 13:08:07 +00:00
|
|
|
dout("ceph_flush_snaps %p\n", inode);
|
2016-08-04 00:43:33 +00:00
|
|
|
if (psession)
|
|
|
|
session = *psession;
|
2016-07-05 13:08:07 +00:00
|
|
|
retry:
|
|
|
|
spin_lock(&ci->i_ceph_lock);
|
|
|
|
if (!(ci->i_ceph_flags & CEPH_I_FLUSH_SNAPS)) {
|
|
|
|
dout(" no capsnap needs flush, doing nothing\n");
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
if (!ci->i_auth_cap) {
|
|
|
|
dout(" no auth cap (migrating?), doing nothing\n");
|
|
|
|
goto out;
|
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2016-07-05 13:08:07 +00:00
|
|
|
mds = ci->i_auth_cap->session->s_mds;
|
|
|
|
if (session && session->s_mds != mds) {
|
|
|
|
dout(" oops, wrong session %p mutex\n", session);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
mutex_unlock(&session->s_mutex);
|
|
|
|
ceph_put_mds_session(session);
|
2016-07-05 13:08:07 +00:00
|
|
|
session = NULL;
|
|
|
|
}
|
|
|
|
if (!session) {
|
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
|
|
|
mutex_lock(&mdsc->mutex);
|
|
|
|
session = __ceph_lookup_mds_session(mdsc, mds);
|
|
|
|
mutex_unlock(&mdsc->mutex);
|
|
|
|
if (session) {
|
|
|
|
dout(" inverting session/ino locks on %p\n", session);
|
|
|
|
mutex_lock(&session->s_mutex);
|
|
|
|
}
|
|
|
|
goto retry;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
2016-07-05 13:08:07 +00:00
|
|
|
__ceph_flush_snaps(ci, session);
|
|
|
|
out:
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
2016-07-05 13:08:07 +00:00
|
|
|
|
|
|
|
if (psession) {
|
|
|
|
*psession = session;
|
|
|
|
} else {
|
|
|
|
mutex_unlock(&session->s_mutex);
|
|
|
|
ceph_put_mds_session(session);
|
|
|
|
}
|
|
|
|
/* we flushed them all; remove this inode from the queue */
|
|
|
|
spin_lock(&mdsc->snap_flush_lock);
|
|
|
|
list_del_init(&ci->i_snap_flush_item);
|
|
|
|
spin_unlock(&mdsc->snap_flush_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
2009-10-16 01:13:53 +00:00
|
|
|
/*
|
2011-05-04 18:33:47 +00:00
|
|
|
* Mark caps dirty. If inode is newly dirty, return the dirty flags.
|
|
|
|
* Caller is then responsible for calling __mark_inode_dirty with the
|
|
|
|
* returned flags value.
|
2009-10-16 01:13:53 +00:00
|
|
|
*/
|
2015-06-10 09:26:13 +00:00
|
|
|
int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
|
|
|
|
struct ceph_cap_flush **pcf)
|
2009-10-16 01:13:53 +00:00
|
|
|
{
|
2010-03-26 09:40:33 +00:00
|
|
|
struct ceph_mds_client *mdsc =
|
2010-04-06 22:14:15 +00:00
|
|
|
ceph_sb_to_client(ci->vfs_inode.i_sb)->mdsc;
|
2009-10-16 01:13:53 +00:00
|
|
|
struct inode *inode = &ci->vfs_inode;
|
|
|
|
int was = ci->i_dirty_caps;
|
|
|
|
int dirty = 0;
|
|
|
|
|
2015-03-24 03:36:08 +00:00
|
|
|
if (!ci->i_auth_cap) {
|
|
|
|
pr_warn("__mark_dirty_caps %p %llx mask %s, "
|
|
|
|
"but no auth cap (session was closed?)\n",
|
|
|
|
inode, ceph_ino(inode), ceph_cap_string(mask));
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2009-10-16 01:13:53 +00:00
|
|
|
dout("__mark_dirty_caps %p %s dirty %s -> %s\n", &ci->vfs_inode,
|
|
|
|
ceph_cap_string(mask), ceph_cap_string(was),
|
|
|
|
ceph_cap_string(was | mask));
|
|
|
|
ci->i_dirty_caps |= mask;
|
|
|
|
if (was == 0) {
|
2015-06-10 09:26:13 +00:00
|
|
|
WARN_ON_ONCE(ci->i_prealloc_cap_flush);
|
|
|
|
swap(ci->i_prealloc_cap_flush, *pcf);
|
|
|
|
|
2015-05-01 09:49:16 +00:00
|
|
|
if (!ci->i_head_snapc) {
|
|
|
|
WARN_ON_ONCE(!rwsem_is_locked(&mdsc->snap_rwsem));
|
2010-08-24 15:44:16 +00:00
|
|
|
ci->i_head_snapc = ceph_get_snap_context(
|
|
|
|
ci->i_snap_realm->cached_context);
|
2015-05-01 09:49:16 +00:00
|
|
|
}
|
2012-11-19 02:49:07 +00:00
|
|
|
dout(" inode %p now dirty snapc %p auth cap %p\n",
|
|
|
|
&ci->vfs_inode, ci->i_head_snapc, ci->i_auth_cap);
|
2009-10-16 01:13:53 +00:00
|
|
|
BUG_ON(!list_empty(&ci->i_dirty_item));
|
|
|
|
spin_lock(&mdsc->cap_dirty_lock);
|
2013-11-24 06:44:38 +00:00
|
|
|
list_add(&ci->i_dirty_item, &mdsc->cap_dirty);
|
2009-10-16 01:13:53 +00:00
|
|
|
spin_unlock(&mdsc->cap_dirty_lock);
|
|
|
|
if (ci->i_flushing_caps == 0) {
|
2011-05-03 16:28:08 +00:00
|
|
|
ihold(inode);
|
2009-10-16 01:13:53 +00:00
|
|
|
dirty |= I_DIRTY_SYNC;
|
|
|
|
}
|
2015-06-10 09:26:13 +00:00
|
|
|
} else {
|
|
|
|
WARN_ON_ONCE(!ci->i_prealloc_cap_flush);
|
2009-10-16 01:13:53 +00:00
|
|
|
}
|
|
|
|
BUG_ON(list_empty(&ci->i_dirty_item));
|
|
|
|
if (((was | ci->i_flushing_caps) & CEPH_CAP_FILE_BUFFER) &&
|
|
|
|
(mask & CEPH_CAP_FILE_BUFFER))
|
|
|
|
dirty |= I_DIRTY_DATASYNC;
|
|
|
|
__cap_delay_requeue(mdsc, ci);
|
2011-05-04 18:33:47 +00:00
|
|
|
return dirty;
|
2009-10-16 01:13:53 +00:00
|
|
|
}
|
|
|
|
|
2015-06-10 09:26:13 +00:00
|
|
|
struct ceph_cap_flush *ceph_alloc_cap_flush(void)
|
|
|
|
{
|
|
|
|
return kmem_cache_alloc(ceph_cap_flush_cachep, GFP_KERNEL);
|
|
|
|
}
|
|
|
|
|
|
|
|
void ceph_free_cap_flush(struct ceph_cap_flush *cf)
|
|
|
|
{
|
|
|
|
if (cf)
|
|
|
|
kmem_cache_free(ceph_cap_flush_cachep, cf);
|
|
|
|
}
|
|
|
|
|
2015-06-10 03:09:32 +00:00
|
|
|
static u64 __get_oldest_flush_tid(struct ceph_mds_client *mdsc)
|
|
|
|
{
|
2016-07-06 03:12:56 +00:00
|
|
|
if (!list_empty(&mdsc->cap_flush_list)) {
|
2015-06-10 03:09:32 +00:00
|
|
|
struct ceph_cap_flush *cf =
|
2016-07-06 03:12:56 +00:00
|
|
|
list_first_entry(&mdsc->cap_flush_list,
|
|
|
|
struct ceph_cap_flush, g_list);
|
2015-06-10 03:09:32 +00:00
|
|
|
return cf->tid;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2016-07-07 07:22:38 +00:00
|
|
|
/*
|
|
|
|
* Remove cap_flush from the mdsc's or inode's flushing cap list.
|
|
|
|
* Return true if caller needs to wake up flush waiters.
|
|
|
|
*/
|
|
|
|
static bool __finish_cap_flush(struct ceph_mds_client *mdsc,
|
|
|
|
struct ceph_inode_info *ci,
|
|
|
|
struct ceph_cap_flush *cf)
|
|
|
|
{
|
|
|
|
struct ceph_cap_flush *prev;
|
|
|
|
bool wake = cf->wake;
|
|
|
|
if (mdsc) {
|
|
|
|
/* are there older pending cap flushes? */
|
|
|
|
if (wake && cf->g_list.prev != &mdsc->cap_flush_list) {
|
|
|
|
prev = list_prev_entry(cf, g_list);
|
|
|
|
prev->wake = true;
|
|
|
|
wake = false;
|
|
|
|
}
|
|
|
|
list_del(&cf->g_list);
|
|
|
|
} else if (ci) {
|
|
|
|
if (wake && cf->i_list.prev != &ci->i_cap_flush_list) {
|
|
|
|
prev = list_prev_entry(cf, i_list);
|
|
|
|
prev->wake = true;
|
|
|
|
wake = false;
|
|
|
|
}
|
|
|
|
list_del(&cf->i_list);
|
|
|
|
} else {
|
|
|
|
BUG_ON(1);
|
|
|
|
}
|
|
|
|
return wake;
|
|
|
|
}
|
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
/*
|
|
|
|
* Add dirty inode to the flushing list. Assigned a seq number so we
|
|
|
|
* can wait for caps to flush without starving.
|
2009-10-14 21:24:19 +00:00
|
|
|
*
|
2011-11-30 17:47:09 +00:00
|
|
|
* Called under i_ceph_lock.
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
*/
|
2009-10-14 21:24:19 +00:00
|
|
|
static int __mark_caps_flushing(struct inode *inode,
|
2016-07-07 07:22:38 +00:00
|
|
|
struct ceph_mds_session *session, bool wake,
|
2015-06-10 03:09:32 +00:00
|
|
|
u64 *flush_tid, u64 *oldest_flush_tid)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
2010-04-06 22:14:15 +00:00
|
|
|
struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
struct ceph_inode_info *ci = ceph_inode(inode);
|
2015-06-10 09:26:13 +00:00
|
|
|
struct ceph_cap_flush *cf = NULL;
|
2009-10-14 21:24:19 +00:00
|
|
|
int flushing;
|
2009-12-01 22:12:07 +00:00
|
|
|
|
2009-10-14 21:24:19 +00:00
|
|
|
BUG_ON(ci->i_dirty_caps == 0);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
BUG_ON(list_empty(&ci->i_dirty_item));
|
2015-06-10 09:26:13 +00:00
|
|
|
BUG_ON(!ci->i_prealloc_cap_flush);
|
2009-10-14 21:24:19 +00:00
|
|
|
|
|
|
|
flushing = ci->i_dirty_caps;
|
|
|
|
dout("__mark_caps_flushing flushing %s, flushing_caps %s -> %s\n",
|
|
|
|
ceph_cap_string(flushing),
|
|
|
|
ceph_cap_string(ci->i_flushing_caps),
|
|
|
|
ceph_cap_string(ci->i_flushing_caps | flushing));
|
|
|
|
ci->i_flushing_caps |= flushing;
|
|
|
|
ci->i_dirty_caps = 0;
|
2009-10-14 21:27:38 +00:00
|
|
|
dout(" inode %p now !dirty\n", inode);
|
2009-10-14 21:24:19 +00:00
|
|
|
|
2015-06-10 09:26:13 +00:00
|
|
|
swap(cf, ci->i_prealloc_cap_flush);
|
2015-06-09 07:48:57 +00:00
|
|
|
cf->caps = flushing;
|
2016-07-07 07:22:38 +00:00
|
|
|
cf->wake = wake;
|
2015-06-09 07:48:57 +00:00
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
spin_lock(&mdsc->cap_dirty_lock);
|
2009-10-14 21:27:38 +00:00
|
|
|
list_del_init(&ci->i_dirty_item);
|
|
|
|
|
2015-06-09 07:48:57 +00:00
|
|
|
cf->tid = ++mdsc->last_cap_flush_tid;
|
2016-07-06 03:12:56 +00:00
|
|
|
list_add_tail(&cf->g_list, &mdsc->cap_flush_list);
|
2015-06-10 03:09:32 +00:00
|
|
|
*oldest_flush_tid = __get_oldest_flush_tid(mdsc);
|
2015-06-09 07:48:57 +00:00
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
if (list_empty(&ci->i_flushing_item)) {
|
|
|
|
list_add_tail(&ci->i_flushing_item, &session->s_cap_flushing);
|
|
|
|
mdsc->num_cap_flushing++;
|
|
|
|
}
|
|
|
|
spin_unlock(&mdsc->cap_dirty_lock);
|
2009-10-14 21:24:19 +00:00
|
|
|
|
2016-07-06 03:12:56 +00:00
|
|
|
list_add_tail(&cf->i_list, &ci->i_cap_flush_list);
|
2015-06-09 07:48:57 +00:00
|
|
|
|
|
|
|
*flush_tid = cf->tid;
|
2009-10-14 21:24:19 +00:00
|
|
|
return flushing;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
2010-02-17 18:43:37 +00:00
|
|
|
/*
|
|
|
|
* try to invalidate mapping pages without blocking.
|
|
|
|
*/
|
|
|
|
static int try_nonblocking_invalidate(struct inode *inode)
|
|
|
|
{
|
|
|
|
struct ceph_inode_info *ci = ceph_inode(inode);
|
|
|
|
u32 invalidating_gen = ci->i_rdcache_gen;
|
|
|
|
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
2010-02-17 18:43:37 +00:00
|
|
|
invalidate_mapping_pages(&inode->i_data, 0, -1);
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_lock(&ci->i_ceph_lock);
|
2010-02-17 18:43:37 +00:00
|
|
|
|
2010-09-17 17:46:44 +00:00
|
|
|
if (inode->i_data.nrpages == 0 &&
|
2010-02-17 18:43:37 +00:00
|
|
|
invalidating_gen == ci->i_rdcache_gen) {
|
|
|
|
/* success. */
|
|
|
|
dout("try_nonblocking_invalidate %p success\n", inode);
|
2010-11-04 18:05:05 +00:00
|
|
|
/* save any racing async invalidate some trouble */
|
|
|
|
ci->i_rdcache_revoking = ci->i_rdcache_gen - 1;
|
2010-02-17 18:43:37 +00:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
dout("try_nonblocking_invalidate %p failed\n", inode);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
2017-05-22 04:03:32 +00:00
|
|
|
bool __ceph_should_report_size(struct ceph_inode_info *ci)
|
|
|
|
{
|
|
|
|
loff_t size = ci->vfs_inode.i_size;
|
|
|
|
/* mds will adjust max size according to the reported size */
|
|
|
|
if (ci->i_flushing_caps & CEPH_CAP_FILE_WR)
|
|
|
|
return false;
|
|
|
|
if (size >= ci->i_max_size)
|
|
|
|
return true;
|
|
|
|
/* half of previous max_size increment has been used */
|
|
|
|
if (ci->i_max_size > ci->i_reported_size &&
|
|
|
|
(size << 1) >= ci->i_max_size + ci->i_reported_size)
|
|
|
|
return true;
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
/*
|
|
|
|
* Swiss army knife function to examine currently used and wanted
|
|
|
|
* versus held caps. Release, flush, ack revoked caps to mds as
|
|
|
|
* appropriate.
|
|
|
|
*
|
|
|
|
* CHECK_CAPS_NODELAY - caller is delayed work and we should not delay
|
|
|
|
* cap release further.
|
|
|
|
* CHECK_CAPS_AUTHONLY - we should only check the auth cap
|
|
|
|
* CHECK_CAPS_FLUSH - we should flush any dirty caps immediately, without
|
|
|
|
* further delay.
|
|
|
|
*/
|
|
|
|
void ceph_check_caps(struct ceph_inode_info *ci, int flags,
|
|
|
|
struct ceph_mds_session *session)
|
|
|
|
{
|
2010-04-06 22:14:15 +00:00
|
|
|
struct ceph_fs_client *fsc = ceph_inode_to_client(&ci->vfs_inode);
|
|
|
|
struct ceph_mds_client *mdsc = fsc->mdsc;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
struct inode *inode = &ci->vfs_inode;
|
|
|
|
struct ceph_cap *cap;
|
2015-06-10 03:09:32 +00:00
|
|
|
u64 flush_tid, oldest_flush_tid;
|
2013-01-04 06:37:57 +00:00
|
|
|
int file_wanted, used, cap_used;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
int took_snap_rwsem = 0; /* true if mdsc->snap_rwsem held */
|
2010-02-09 21:41:18 +00:00
|
|
|
int issued, implemented, want, retain, revoking, flushing = 0;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
int mds = -1; /* keep track of how far we've gone through i_caps list
|
|
|
|
to avoid an infinite loop on retry */
|
|
|
|
struct rb_node *p;
|
2016-07-06 07:37:42 +00:00
|
|
|
int delayed = 0, sent = 0, num;
|
|
|
|
bool is_delayed = flags & CHECK_CAPS_NODELAY;
|
|
|
|
bool queue_invalidate = false;
|
|
|
|
bool force_requeue = false;
|
|
|
|
bool tried_invalidate = false;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
/* if we are unmounting, flush any unused caps immediately. */
|
|
|
|
if (mdsc->stopping)
|
|
|
|
is_delayed = 1;
|
|
|
|
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_lock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
if (ci->i_ceph_flags & CEPH_I_FLUSH)
|
|
|
|
flags |= CHECK_CAPS_FLUSH;
|
|
|
|
|
|
|
|
goto retry_locked;
|
|
|
|
retry:
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_lock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
retry_locked:
|
|
|
|
file_wanted = __ceph_caps_file_wanted(ci);
|
|
|
|
used = __ceph_caps_used(ci);
|
2010-02-09 21:41:18 +00:00
|
|
|
issued = __ceph_caps_issued(ci, &implemented);
|
|
|
|
revoking = implemented & ~issued;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2015-05-25 09:36:42 +00:00
|
|
|
want = file_wanted;
|
|
|
|
retain = file_wanted | used | CEPH_CAP_PIN;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
if (!mdsc->stopping && inode->i_nlink > 0) {
|
2015-05-25 09:36:42 +00:00
|
|
|
if (file_wanted) {
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
retain |= CEPH_CAP_ANY; /* be greedy */
|
2015-03-26 11:06:00 +00:00
|
|
|
} else if (S_ISDIR(inode->i_mode) &&
|
|
|
|
(issued & CEPH_CAP_FILE_SHARED) &&
|
|
|
|
__ceph_dir_is_complete(ci)) {
|
|
|
|
/*
|
|
|
|
* If a directory is complete, we want to keep
|
|
|
|
* the exclusive cap. So that MDS does not end up
|
|
|
|
* revoking the shared cap on every create/unlink
|
|
|
|
* operation.
|
|
|
|
*/
|
|
|
|
want = CEPH_CAP_ANY_SHARED | CEPH_CAP_FILE_EXCL;
|
|
|
|
retain |= want;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
} else {
|
2015-03-26 11:06:00 +00:00
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
retain |= CEPH_CAP_ANY_SHARED;
|
|
|
|
/*
|
|
|
|
* keep RD only if we didn't have the file open RW,
|
|
|
|
* because then the mds would revoke it anyway to
|
|
|
|
* journal max_size=0.
|
|
|
|
*/
|
|
|
|
if (ci->i_max_size == 0)
|
|
|
|
retain |= CEPH_CAP_ANY_RD;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
dout("check_caps %p file_want %s used %s dirty %s flushing %s"
|
2010-02-09 21:41:18 +00:00
|
|
|
" issued %s revoking %s retain %s %s%s%s\n", inode,
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
ceph_cap_string(file_wanted),
|
|
|
|
ceph_cap_string(used), ceph_cap_string(ci->i_dirty_caps),
|
|
|
|
ceph_cap_string(ci->i_flushing_caps),
|
2010-02-09 21:41:18 +00:00
|
|
|
ceph_cap_string(issued), ceph_cap_string(revoking),
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
ceph_cap_string(retain),
|
|
|
|
(flags & CHECK_CAPS_AUTHONLY) ? " AUTHONLY" : "",
|
|
|
|
(flags & CHECK_CAPS_NODELAY) ? " NODELAY" : "",
|
|
|
|
(flags & CHECK_CAPS_FLUSH) ? " FLUSH" : "");
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If we no longer need to hold onto old our caps, and we may
|
|
|
|
* have cached pages, but don't want them, then try to invalidate.
|
|
|
|
* If we fail, it's because pages are locked.... try again later.
|
|
|
|
*/
|
|
|
|
if ((!is_delayed || mdsc->stopping) &&
|
2015-06-16 12:48:56 +00:00
|
|
|
!S_ISDIR(inode->i_mode) && /* ignore readdir cache */
|
2016-05-18 12:58:26 +00:00
|
|
|
!(ci->i_wb_ref || ci->i_wrbuffer_ref) && /* no dirty pages... */
|
2015-06-16 12:48:56 +00:00
|
|
|
inode->i_data.nrpages && /* have cached pages */
|
2015-10-26 08:08:43 +00:00
|
|
|
(revoking & (CEPH_CAP_FILE_CACHE|
|
|
|
|
CEPH_CAP_FILE_LAZYIO)) && /* or revoking cache */
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
!tried_invalidate) {
|
|
|
|
dout("check_caps trying to invalidate on %p\n", inode);
|
2010-02-17 18:43:37 +00:00
|
|
|
if (try_nonblocking_invalidate(inode) < 0) {
|
2010-05-27 17:40:43 +00:00
|
|
|
if (revoking & (CEPH_CAP_FILE_CACHE|
|
|
|
|
CEPH_CAP_FILE_LAZYIO)) {
|
2010-02-17 18:43:37 +00:00
|
|
|
dout("check_caps queuing invalidate\n");
|
2016-07-06 07:37:42 +00:00
|
|
|
queue_invalidate = true;
|
2010-02-17 18:43:37 +00:00
|
|
|
ci->i_rdcache_revoking = ci->i_rdcache_gen;
|
|
|
|
} else {
|
|
|
|
dout("check_caps failed to invalidate pages\n");
|
|
|
|
/* we failed to invalidate pages. check these
|
|
|
|
caps again later. */
|
2016-07-06 07:37:42 +00:00
|
|
|
force_requeue = true;
|
2010-02-17 18:43:37 +00:00
|
|
|
__cap_set_timeouts(mdsc, ci);
|
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
2016-07-06 07:37:42 +00:00
|
|
|
tried_invalidate = true;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
goto retry_locked;
|
|
|
|
}
|
|
|
|
|
|
|
|
num = 0;
|
|
|
|
for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) {
|
|
|
|
cap = rb_entry(p, struct ceph_cap, ci_node);
|
|
|
|
num++;
|
|
|
|
|
|
|
|
/* avoid looping forever */
|
|
|
|
if (mds >= cap->mds ||
|
|
|
|
((flags & CHECK_CAPS_AUTHONLY) && cap != ci->i_auth_cap))
|
|
|
|
continue;
|
|
|
|
|
|
|
|
/* NOTE: no side-effects allowed, until we take s_mutex */
|
|
|
|
|
2013-01-04 06:37:57 +00:00
|
|
|
cap_used = used;
|
|
|
|
if (ci->i_auth_cap && cap != ci->i_auth_cap)
|
|
|
|
cap_used &= ~ci->i_auth_cap->issued;
|
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
revoking = cap->implemented & ~cap->issued;
|
2013-01-04 06:37:57 +00:00
|
|
|
dout(" mds%d cap %p used %s issued %s implemented %s revoking %s\n",
|
2016-05-18 12:58:26 +00:00
|
|
|
cap->mds, cap, ceph_cap_string(cap_used),
|
|
|
|
ceph_cap_string(cap->issued),
|
2011-01-18 16:56:01 +00:00
|
|
|
ceph_cap_string(cap->implemented),
|
|
|
|
ceph_cap_string(revoking));
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
if (cap == ci->i_auth_cap &&
|
|
|
|
(cap->issued & CEPH_CAP_FILE_WR)) {
|
|
|
|
/* request larger max_size from MDS? */
|
|
|
|
if (ci->i_wanted_max_size > ci->i_max_size &&
|
|
|
|
ci->i_wanted_max_size > ci->i_requested_max_size) {
|
|
|
|
dout("requesting new max_size\n");
|
|
|
|
goto ack;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* approaching file_max? */
|
2017-05-22 04:03:32 +00:00
|
|
|
if (__ceph_should_report_size(ci)) {
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
dout("i_size approaching max_size\n");
|
|
|
|
goto ack;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
/* flush anything dirty? */
|
2016-07-07 10:34:45 +00:00
|
|
|
if (cap == ci->i_auth_cap) {
|
|
|
|
if ((flags & CHECK_CAPS_FLUSH) && ci->i_dirty_caps) {
|
|
|
|
dout("flushing dirty caps\n");
|
|
|
|
goto ack;
|
|
|
|
}
|
|
|
|
if (ci->i_ceph_flags & CEPH_I_FLUSH_SNAPS) {
|
|
|
|
dout("flushing snap caps\n");
|
|
|
|
goto ack;
|
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/* completed revocation? going down and there are no caps? */
|
2013-01-04 06:37:57 +00:00
|
|
|
if (revoking && (revoking & cap_used) == 0) {
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
dout("completed revocation of %s\n",
|
|
|
|
ceph_cap_string(cap->implemented & ~cap->issued));
|
|
|
|
goto ack;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* want more caps from mds? */
|
|
|
|
if (want & ~(cap->mds_wanted | cap->issued))
|
|
|
|
goto ack;
|
|
|
|
|
|
|
|
/* things we might delay */
|
|
|
|
if ((cap->issued & ~retain) == 0 &&
|
|
|
|
cap->mds_wanted == want)
|
|
|
|
continue; /* nope, all good */
|
|
|
|
|
|
|
|
if (is_delayed)
|
|
|
|
goto ack;
|
|
|
|
|
|
|
|
/* delay? */
|
|
|
|
if ((ci->i_ceph_flags & CEPH_I_NODELAY) == 0 &&
|
|
|
|
time_before(jiffies, ci->i_hold_caps_max)) {
|
|
|
|
dout(" delaying issued %s -> %s, wanted %s -> %s\n",
|
|
|
|
ceph_cap_string(cap->issued),
|
|
|
|
ceph_cap_string(cap->issued & retain),
|
|
|
|
ceph_cap_string(cap->mds_wanted),
|
|
|
|
ceph_cap_string(want));
|
|
|
|
delayed++;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
ack:
|
2010-03-01 23:16:56 +00:00
|
|
|
if (ci->i_ceph_flags & CEPH_I_NOFLUSH) {
|
|
|
|
dout(" skipping %p I_NOFLUSH set\n", inode);
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
if (session && session != cap->session) {
|
|
|
|
dout("oops, wrong session %p mutex\n", session);
|
|
|
|
mutex_unlock(&session->s_mutex);
|
|
|
|
session = NULL;
|
|
|
|
}
|
|
|
|
if (!session) {
|
|
|
|
session = cap->session;
|
|
|
|
if (mutex_trylock(&session->s_mutex) == 0) {
|
|
|
|
dout("inverting session/ino locks on %p\n",
|
|
|
|
session);
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
if (took_snap_rwsem) {
|
|
|
|
up_read(&mdsc->snap_rwsem);
|
|
|
|
took_snap_rwsem = 0;
|
|
|
|
}
|
|
|
|
mutex_lock(&session->s_mutex);
|
|
|
|
goto retry;
|
|
|
|
}
|
|
|
|
}
|
2016-07-07 10:34:45 +00:00
|
|
|
|
|
|
|
/* kick flushing and flush snaps before sending normal
|
|
|
|
* cap message */
|
|
|
|
if (cap == ci->i_auth_cap &&
|
|
|
|
(ci->i_ceph_flags &
|
|
|
|
(CEPH_I_KICK_FLUSH | CEPH_I_FLUSH_SNAPS))) {
|
|
|
|
if (ci->i_ceph_flags & CEPH_I_KICK_FLUSH) {
|
|
|
|
spin_lock(&mdsc->cap_dirty_lock);
|
|
|
|
oldest_flush_tid = __get_oldest_flush_tid(mdsc);
|
|
|
|
spin_unlock(&mdsc->cap_dirty_lock);
|
|
|
|
__kick_flushing_caps(mdsc, session, ci,
|
|
|
|
oldest_flush_tid);
|
|
|
|
ci->i_ceph_flags &= ~CEPH_I_KICK_FLUSH;
|
|
|
|
}
|
2016-07-05 13:08:07 +00:00
|
|
|
if (ci->i_ceph_flags & CEPH_I_FLUSH_SNAPS)
|
|
|
|
__ceph_flush_snaps(ci, session);
|
|
|
|
|
2016-07-07 10:34:45 +00:00
|
|
|
goto retry_locked;
|
|
|
|
}
|
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
/* take snap_rwsem after session mutex */
|
|
|
|
if (!took_snap_rwsem) {
|
|
|
|
if (down_read_trylock(&mdsc->snap_rwsem) == 0) {
|
|
|
|
dout("inverting snap/in locks on %p\n",
|
|
|
|
inode);
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
down_read(&mdsc->snap_rwsem);
|
|
|
|
took_snap_rwsem = 1;
|
|
|
|
goto retry;
|
|
|
|
}
|
|
|
|
took_snap_rwsem = 1;
|
|
|
|
}
|
|
|
|
|
2015-06-09 07:48:57 +00:00
|
|
|
if (cap == ci->i_auth_cap && ci->i_dirty_caps) {
|
2016-07-07 07:22:38 +00:00
|
|
|
flushing = __mark_caps_flushing(inode, session, false,
|
2015-06-10 03:09:32 +00:00
|
|
|
&flush_tid,
|
|
|
|
&oldest_flush_tid);
|
2015-06-09 07:48:57 +00:00
|
|
|
} else {
|
2011-01-18 16:48:06 +00:00
|
|
|
flushing = 0;
|
2015-06-09 07:48:57 +00:00
|
|
|
flush_tid = 0;
|
2015-06-10 03:09:32 +00:00
|
|
|
spin_lock(&mdsc->cap_dirty_lock);
|
|
|
|
oldest_flush_tid = __get_oldest_flush_tid(mdsc);
|
|
|
|
spin_unlock(&mdsc->cap_dirty_lock);
|
2015-06-09 07:48:57 +00:00
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
mds = cap->mds; /* remember mds, so we don't repeat */
|
|
|
|
sent++;
|
|
|
|
|
2011-11-30 17:47:09 +00:00
|
|
|
/* __send_cap drops i_ceph_lock */
|
2016-11-10 12:42:06 +00:00
|
|
|
delayed += __send_cap(mdsc, cap, CEPH_CAP_OP_UPDATE, false,
|
|
|
|
cap_used, want, retain, flushing,
|
|
|
|
flush_tid, oldest_flush_tid);
|
2011-11-30 17:47:09 +00:00
|
|
|
goto retry; /* retake i_ceph_lock and restart our cap scan. */
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Reschedule delayed caps release if we delayed anything,
|
|
|
|
* otherwise cancel.
|
|
|
|
*/
|
|
|
|
if (delayed && is_delayed)
|
2016-07-06 07:37:42 +00:00
|
|
|
force_requeue = true; /* __send_cap delayed release; requeue */
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
if (!delayed && !is_delayed)
|
|
|
|
__cap_delay_cancel(mdsc, ci);
|
|
|
|
else if (!is_delayed || force_requeue)
|
|
|
|
__cap_delay_requeue(mdsc, ci);
|
|
|
|
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2010-02-09 21:41:18 +00:00
|
|
|
if (queue_invalidate)
|
2010-02-09 23:24:44 +00:00
|
|
|
ceph_queue_invalidate(inode);
|
2010-02-09 21:41:18 +00:00
|
|
|
|
2010-03-16 20:39:28 +00:00
|
|
|
if (session)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
mutex_unlock(&session->s_mutex);
|
|
|
|
if (took_snap_rwsem)
|
|
|
|
up_read(&mdsc->snap_rwsem);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Try to flush dirty caps back to the auth mds.
|
|
|
|
*/
|
2015-06-09 07:48:57 +00:00
|
|
|
static int try_flush_caps(struct inode *inode, u64 *ptid)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
2010-04-06 22:14:15 +00:00
|
|
|
struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
struct ceph_inode_info *ci = ceph_inode(inode);
|
2013-10-31 08:44:14 +00:00
|
|
|
struct ceph_mds_session *session = NULL;
|
2015-05-27 01:59:48 +00:00
|
|
|
int flushing = 0;
|
2015-06-10 03:09:32 +00:00
|
|
|
u64 flush_tid = 0, oldest_flush_tid = 0;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
retry:
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_lock(&ci->i_ceph_lock);
|
2010-03-01 23:16:56 +00:00
|
|
|
if (ci->i_ceph_flags & CEPH_I_NOFLUSH) {
|
|
|
|
dout("try_flush_caps skipping %p I_NOFLUSH set\n", inode);
|
|
|
|
goto out;
|
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
if (ci->i_dirty_caps && ci->i_auth_cap) {
|
|
|
|
struct ceph_cap *cap = ci->i_auth_cap;
|
|
|
|
int used = __ceph_caps_used(ci);
|
|
|
|
int want = __ceph_caps_wanted(ci);
|
|
|
|
int delayed;
|
|
|
|
|
2013-10-31 08:44:14 +00:00
|
|
|
if (!session || session != cap->session) {
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
2013-10-31 08:44:14 +00:00
|
|
|
if (session)
|
|
|
|
mutex_unlock(&session->s_mutex);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
session = cap->session;
|
|
|
|
mutex_lock(&session->s_mutex);
|
|
|
|
goto retry;
|
|
|
|
}
|
|
|
|
if (cap->session->s_state < CEPH_MDS_SESSION_OPEN)
|
|
|
|
goto out;
|
|
|
|
|
2016-07-07 07:22:38 +00:00
|
|
|
flushing = __mark_caps_flushing(inode, session, true,
|
|
|
|
&flush_tid, &oldest_flush_tid);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2011-11-30 17:47:09 +00:00
|
|
|
/* __send_cap drops i_ceph_lock */
|
2016-11-10 12:42:06 +00:00
|
|
|
delayed = __send_cap(mdsc, cap, CEPH_CAP_OP_FLUSH, true,
|
|
|
|
used, want, (cap->issued | cap->implemented),
|
|
|
|
flushing, flush_tid, oldest_flush_tid);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2015-06-09 07:48:57 +00:00
|
|
|
if (delayed) {
|
|
|
|
spin_lock(&ci->i_ceph_lock);
|
2015-05-27 01:59:48 +00:00
|
|
|
__cap_delay_requeue(mdsc, ci);
|
2015-06-09 07:48:57 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
|
|
|
}
|
|
|
|
} else {
|
2016-07-06 03:12:56 +00:00
|
|
|
if (!list_empty(&ci->i_cap_flush_list)) {
|
2015-06-09 07:48:57 +00:00
|
|
|
struct ceph_cap_flush *cf =
|
2016-07-06 03:12:56 +00:00
|
|
|
list_last_entry(&ci->i_cap_flush_list,
|
2016-07-07 07:22:38 +00:00
|
|
|
struct ceph_cap_flush, i_list);
|
|
|
|
cf->wake = true;
|
2015-06-09 07:48:57 +00:00
|
|
|
flush_tid = cf->tid;
|
|
|
|
}
|
|
|
|
flushing = ci->i_flushing_caps;
|
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
out:
|
2013-10-31 08:44:14 +00:00
|
|
|
if (session)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
mutex_unlock(&session->s_mutex);
|
2015-06-09 07:48:57 +00:00
|
|
|
|
|
|
|
*ptid = flush_tid;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
return flushing;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Return true if we've flushed caps through the given flush_tid.
|
|
|
|
*/
|
2015-06-09 07:48:57 +00:00
|
|
|
static int caps_are_flushed(struct inode *inode, u64 flush_tid)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
|
|
|
struct ceph_inode_info *ci = ceph_inode(inode);
|
2015-06-09 07:48:57 +00:00
|
|
|
int ret = 1;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_lock(&ci->i_ceph_lock);
|
2016-07-06 03:12:56 +00:00
|
|
|
if (!list_empty(&ci->i_cap_flush_list)) {
|
|
|
|
struct ceph_cap_flush * cf =
|
|
|
|
list_first_entry(&ci->i_cap_flush_list,
|
|
|
|
struct ceph_cap_flush, i_list);
|
2015-06-09 07:48:57 +00:00
|
|
|
if (cf->tid <= flush_tid)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
ret = 0;
|
2015-05-27 01:59:48 +00:00
|
|
|
}
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2015-05-27 03:19:34 +00:00
|
|
|
/*
|
2015-10-27 10:36:06 +00:00
|
|
|
* wait for any unsafe requests to complete.
|
2015-05-27 03:19:34 +00:00
|
|
|
*/
|
2015-10-27 10:36:06 +00:00
|
|
|
static int unsafe_request_wait(struct inode *inode)
|
2015-05-27 03:19:34 +00:00
|
|
|
{
|
|
|
|
struct ceph_inode_info *ci = ceph_inode(inode);
|
2015-10-27 10:36:06 +00:00
|
|
|
struct ceph_mds_request *req1 = NULL, *req2 = NULL;
|
|
|
|
int ret, err = 0;
|
2015-05-27 03:19:34 +00:00
|
|
|
|
|
|
|
spin_lock(&ci->i_unsafe_lock);
|
2015-10-27 10:36:06 +00:00
|
|
|
if (S_ISDIR(inode->i_mode) && !list_empty(&ci->i_unsafe_dirops)) {
|
|
|
|
req1 = list_last_entry(&ci->i_unsafe_dirops,
|
|
|
|
struct ceph_mds_request,
|
|
|
|
r_unsafe_dir_item);
|
|
|
|
ceph_mdsc_get_request(req1);
|
|
|
|
}
|
|
|
|
if (!list_empty(&ci->i_unsafe_iops)) {
|
|
|
|
req2 = list_last_entry(&ci->i_unsafe_iops,
|
|
|
|
struct ceph_mds_request,
|
|
|
|
r_unsafe_target_item);
|
|
|
|
ceph_mdsc_get_request(req2);
|
|
|
|
}
|
|
|
|
spin_unlock(&ci->i_unsafe_lock);
|
2015-05-27 03:19:34 +00:00
|
|
|
|
2016-11-16 14:45:22 +00:00
|
|
|
dout("unsafe_request_wait %p wait on tid %llu %llu\n",
|
2015-10-27 10:36:06 +00:00
|
|
|
inode, req1 ? req1->r_tid : 0ULL, req2 ? req2->r_tid : 0ULL);
|
|
|
|
if (req1) {
|
|
|
|
ret = !wait_for_completion_timeout(&req1->r_safe_completion,
|
|
|
|
ceph_timeout_jiffies(req1->r_timeout));
|
2015-05-27 03:19:34 +00:00
|
|
|
if (ret)
|
2015-10-27 10:36:06 +00:00
|
|
|
err = -EIO;
|
|
|
|
ceph_mdsc_put_request(req1);
|
|
|
|
}
|
|
|
|
if (req2) {
|
|
|
|
ret = !wait_for_completion_timeout(&req2->r_safe_completion,
|
|
|
|
ceph_timeout_jiffies(req2->r_timeout));
|
|
|
|
if (ret)
|
|
|
|
err = -EIO;
|
|
|
|
ceph_mdsc_put_request(req2);
|
|
|
|
}
|
|
|
|
return err;
|
2015-05-27 03:19:34 +00:00
|
|
|
}
|
|
|
|
|
2011-07-17 00:44:56 +00:00
|
|
|
int ceph_fsync(struct file *file, loff_t start, loff_t end, int datasync)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
2010-05-26 15:53:25 +00:00
|
|
|
struct inode *inode = file->f_mapping->host;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
struct ceph_inode_info *ci = ceph_inode(inode);
|
2015-06-09 07:48:57 +00:00
|
|
|
u64 flush_tid;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
int ret;
|
|
|
|
int dirty;
|
|
|
|
|
|
|
|
dout("fsync %p%s\n", inode, datasync ? " datasync" : "");
|
2016-06-15 08:29:18 +00:00
|
|
|
|
2011-07-17 00:44:56 +00:00
|
|
|
ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
if (ret < 0)
|
2015-05-27 03:19:34 +00:00
|
|
|
goto out;
|
|
|
|
|
|
|
|
if (datasync)
|
|
|
|
goto out;
|
|
|
|
|
2016-01-22 20:40:57 +00:00
|
|
|
inode_lock(inode);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2015-06-09 07:48:57 +00:00
|
|
|
dirty = try_flush_caps(inode, &flush_tid);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
dout("fsync dirty caps are %s\n", ceph_cap_string(dirty));
|
|
|
|
|
2015-10-27 10:36:06 +00:00
|
|
|
ret = unsafe_request_wait(inode);
|
2015-05-27 03:19:34 +00:00
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
/*
|
|
|
|
* only wait on non-file metadata writeback (the mds
|
|
|
|
* can recover size and mtime, so we don't need to
|
|
|
|
* wait for that)
|
|
|
|
*/
|
2015-05-27 03:19:34 +00:00
|
|
|
if (!ret && (dirty & ~CEPH_CAP_ANY_FILE_WR)) {
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
ret = wait_event_interruptible(ci->i_cap_wq,
|
2015-05-27 03:19:34 +00:00
|
|
|
caps_are_flushed(inode, flush_tid));
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
2016-01-22 20:40:57 +00:00
|
|
|
inode_unlock(inode);
|
2015-05-27 03:19:34 +00:00
|
|
|
out:
|
|
|
|
dout("fsync %p%s result=%d\n", inode, datasync ? " datasync" : "", ret);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Flush any dirty caps back to the mds. If we aren't asked to wait,
|
|
|
|
* queue inode for flush but don't do so immediately, because we can
|
|
|
|
* get by with fewer MDS messages if we wait for data writeback to
|
|
|
|
* complete first.
|
|
|
|
*/
|
2010-01-18 00:53:08 +00:00
|
|
|
int ceph_write_inode(struct inode *inode, struct writeback_control *wbc)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
|
|
|
struct ceph_inode_info *ci = ceph_inode(inode);
|
2015-06-09 07:48:57 +00:00
|
|
|
u64 flush_tid;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
int err = 0;
|
|
|
|
int dirty;
|
2010-01-18 00:53:08 +00:00
|
|
|
int wait = wbc->sync_mode == WB_SYNC_ALL;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
dout("write_inode %p wait=%d\n", inode, wait);
|
|
|
|
if (wait) {
|
2015-06-09 07:48:57 +00:00
|
|
|
dirty = try_flush_caps(inode, &flush_tid);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
if (dirty)
|
|
|
|
err = wait_event_interruptible(ci->i_cap_wq,
|
|
|
|
caps_are_flushed(inode, flush_tid));
|
|
|
|
} else {
|
2010-03-26 09:40:33 +00:00
|
|
|
struct ceph_mds_client *mdsc =
|
2010-04-06 22:14:15 +00:00
|
|
|
ceph_sb_to_client(inode->i_sb)->mdsc;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_lock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
if (__ceph_caps_dirty(ci))
|
|
|
|
__cap_delay_requeue_front(mdsc, ci);
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2016-07-04 10:06:41 +00:00
|
|
|
static void __kick_flushing_caps(struct ceph_mds_client *mdsc,
|
|
|
|
struct ceph_mds_session *session,
|
|
|
|
struct ceph_inode_info *ci,
|
|
|
|
u64 oldest_flush_tid)
|
|
|
|
__releases(ci->i_ceph_lock)
|
|
|
|
__acquires(ci->i_ceph_lock)
|
2015-06-09 07:48:57 +00:00
|
|
|
{
|
|
|
|
struct inode *inode = &ci->vfs_inode;
|
|
|
|
struct ceph_cap *cap;
|
|
|
|
struct ceph_cap_flush *cf;
|
2016-07-04 10:06:41 +00:00
|
|
|
int ret;
|
2015-06-09 07:48:57 +00:00
|
|
|
u64 first_tid = 0;
|
|
|
|
|
2016-07-06 03:12:56 +00:00
|
|
|
list_for_each_entry(cf, &ci->i_cap_flush_list, i_list) {
|
|
|
|
if (cf->tid < first_tid)
|
|
|
|
continue;
|
|
|
|
|
2015-06-09 07:48:57 +00:00
|
|
|
cap = ci->i_auth_cap;
|
|
|
|
if (!(cap && cap->session == session)) {
|
2016-07-04 10:06:41 +00:00
|
|
|
pr_err("%p auth cap %p not mds%d ???\n",
|
|
|
|
inode, cap, session->s_mds);
|
2015-06-09 07:48:57 +00:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
first_tid = cf->tid + 1;
|
|
|
|
|
2016-07-04 10:06:41 +00:00
|
|
|
if (cf->caps) {
|
|
|
|
dout("kick_flushing_caps %p cap %p tid %llu %s\n",
|
|
|
|
inode, cap, cf->tid, ceph_cap_string(cf->caps));
|
|
|
|
ci->i_ceph_flags |= CEPH_I_NODELAY;
|
|
|
|
ret = __send_cap(mdsc, cap, CEPH_CAP_OP_FLUSH,
|
2016-11-10 12:42:06 +00:00
|
|
|
false, __ceph_caps_used(ci),
|
2016-07-04 10:06:41 +00:00
|
|
|
__ceph_caps_wanted(ci),
|
|
|
|
cap->issued | cap->implemented,
|
|
|
|
cf->caps, cf->tid, oldest_flush_tid);
|
|
|
|
if (ret) {
|
|
|
|
pr_err("kick_flushing_caps: error sending "
|
|
|
|
"cap flush, ino (%llx.%llx) "
|
|
|
|
"tid %llu flushing %s\n",
|
|
|
|
ceph_vinop(inode), cf->tid,
|
|
|
|
ceph_cap_string(cf->caps));
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
struct ceph_cap_snap *capsnap =
|
|
|
|
container_of(cf, struct ceph_cap_snap,
|
|
|
|
cap_flush);
|
|
|
|
dout("kick_flushing_caps %p capsnap %p tid %llu %s\n",
|
|
|
|
inode, capsnap, cf->tid,
|
|
|
|
ceph_cap_string(capsnap->dirty));
|
|
|
|
|
2017-03-03 09:15:07 +00:00
|
|
|
refcount_inc(&capsnap->nref);
|
2016-07-04 10:06:41 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
|
|
|
|
|
|
|
ret = __send_flush_snap(inode, session, capsnap, cap->mseq,
|
|
|
|
oldest_flush_tid);
|
|
|
|
if (ret < 0) {
|
|
|
|
pr_err("kick_flushing_caps: error sending "
|
|
|
|
"cap flushsnap, ino (%llx.%llx) "
|
|
|
|
"tid %llu follows %llu\n",
|
|
|
|
ceph_vinop(inode), cf->tid,
|
|
|
|
capsnap->follows);
|
|
|
|
}
|
|
|
|
|
|
|
|
ceph_put_cap_snap(capsnap);
|
|
|
|
}
|
2016-07-06 03:12:56 +00:00
|
|
|
|
|
|
|
spin_lock(&ci->i_ceph_lock);
|
2015-06-09 07:48:57 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-06-10 07:17:56 +00:00
|
|
|
void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
|
|
|
|
struct ceph_mds_session *session)
|
|
|
|
{
|
|
|
|
struct ceph_inode_info *ci;
|
|
|
|
struct ceph_cap *cap;
|
2016-07-04 10:06:41 +00:00
|
|
|
u64 oldest_flush_tid;
|
2015-06-10 07:17:56 +00:00
|
|
|
|
|
|
|
dout("early_kick_flushing_caps mds%d\n", session->s_mds);
|
2016-07-04 10:06:41 +00:00
|
|
|
|
|
|
|
spin_lock(&mdsc->cap_dirty_lock);
|
|
|
|
oldest_flush_tid = __get_oldest_flush_tid(mdsc);
|
|
|
|
spin_unlock(&mdsc->cap_dirty_lock);
|
|
|
|
|
2015-06-10 07:17:56 +00:00
|
|
|
list_for_each_entry(ci, &session->s_cap_flushing, i_flushing_item) {
|
|
|
|
spin_lock(&ci->i_ceph_lock);
|
|
|
|
cap = ci->i_auth_cap;
|
|
|
|
if (!(cap && cap->session == session)) {
|
|
|
|
pr_err("%p auth cap %p not mds%d ???\n",
|
|
|
|
&ci->vfs_inode, cap, session->s_mds);
|
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* if flushing caps were revoked, we re-send the cap flush
|
|
|
|
* in client reconnect stage. This guarantees MDS * processes
|
|
|
|
* the cap flush message before issuing the flushing caps to
|
|
|
|
* other client.
|
|
|
|
*/
|
|
|
|
if ((cap->issued & ci->i_flushing_caps) !=
|
|
|
|
ci->i_flushing_caps) {
|
2016-07-05 08:45:21 +00:00
|
|
|
ci->i_ceph_flags &= ~CEPH_I_KICK_FLUSH;
|
2016-07-04 10:06:41 +00:00
|
|
|
__kick_flushing_caps(mdsc, session, ci,
|
|
|
|
oldest_flush_tid);
|
2016-07-05 08:45:21 +00:00
|
|
|
} else {
|
|
|
|
ci->i_ceph_flags |= CEPH_I_KICK_FLUSH;
|
2015-06-10 07:17:56 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
|
|
|
|
struct ceph_mds_session *session)
|
|
|
|
{
|
|
|
|
struct ceph_inode_info *ci;
|
2016-07-05 08:45:21 +00:00
|
|
|
struct ceph_cap *cap;
|
2016-07-04 10:06:41 +00:00
|
|
|
u64 oldest_flush_tid;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
dout("kick_flushing_caps mds%d\n", session->s_mds);
|
2016-07-04 10:06:41 +00:00
|
|
|
|
|
|
|
spin_lock(&mdsc->cap_dirty_lock);
|
|
|
|
oldest_flush_tid = __get_oldest_flush_tid(mdsc);
|
|
|
|
spin_unlock(&mdsc->cap_dirty_lock);
|
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
list_for_each_entry(ci, &session->s_cap_flushing, i_flushing_item) {
|
2016-07-04 10:06:41 +00:00
|
|
|
spin_lock(&ci->i_ceph_lock);
|
2016-07-05 08:45:21 +00:00
|
|
|
cap = ci->i_auth_cap;
|
|
|
|
if (!(cap && cap->session == session)) {
|
|
|
|
pr_err("%p auth cap %p not mds%d ???\n",
|
|
|
|
&ci->vfs_inode, cap, session->s_mds);
|
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
if (ci->i_ceph_flags & CEPH_I_KICK_FLUSH) {
|
|
|
|
ci->i_ceph_flags &= ~CEPH_I_KICK_FLUSH;
|
|
|
|
__kick_flushing_caps(mdsc, session, ci,
|
|
|
|
oldest_flush_tid);
|
|
|
|
}
|
2016-07-04 10:06:41 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2011-01-18 16:56:01 +00:00
|
|
|
static void kick_flushing_inode_caps(struct ceph_mds_client *mdsc,
|
|
|
|
struct ceph_mds_session *session,
|
|
|
|
struct inode *inode)
|
2016-07-04 10:06:41 +00:00
|
|
|
__releases(ci->i_ceph_lock)
|
2011-01-18 16:56:01 +00:00
|
|
|
{
|
|
|
|
struct ceph_inode_info *ci = ceph_inode(inode);
|
|
|
|
struct ceph_cap *cap;
|
|
|
|
|
|
|
|
cap = ci->i_auth_cap;
|
2015-06-09 09:20:12 +00:00
|
|
|
dout("kick_flushing_inode_caps %p flushing %s\n", inode,
|
|
|
|
ceph_cap_string(ci->i_flushing_caps));
|
2013-05-31 08:40:24 +00:00
|
|
|
|
2016-07-04 10:06:41 +00:00
|
|
|
if (!list_empty(&ci->i_cap_flush_list)) {
|
|
|
|
u64 oldest_flush_tid;
|
2013-05-31 08:40:24 +00:00
|
|
|
spin_lock(&mdsc->cap_dirty_lock);
|
|
|
|
list_move_tail(&ci->i_flushing_item,
|
|
|
|
&cap->session->s_cap_flushing);
|
2016-07-04 10:06:41 +00:00
|
|
|
oldest_flush_tid = __get_oldest_flush_tid(mdsc);
|
2013-05-31 08:40:24 +00:00
|
|
|
spin_unlock(&mdsc->cap_dirty_lock);
|
|
|
|
|
2016-07-05 08:45:21 +00:00
|
|
|
ci->i_ceph_flags &= ~CEPH_I_KICK_FLUSH;
|
2016-07-04 10:06:41 +00:00
|
|
|
__kick_flushing_caps(mdsc, session, ci, oldest_flush_tid);
|
2015-06-09 07:48:57 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
2011-01-18 16:56:01 +00:00
|
|
|
} else {
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
2011-01-18 16:56:01 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Take references to capabilities we hold, so that we don't release
|
|
|
|
* them to the MDS prematurely.
|
|
|
|
*
|
2011-11-30 17:47:09 +00:00
|
|
|
* Protected by i_ceph_lock.
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
*/
|
2015-04-30 06:40:54 +00:00
|
|
|
static void __take_cap_refs(struct ceph_inode_info *ci, int got,
|
|
|
|
bool snap_rwsem_locked)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
|
|
|
if (got & CEPH_CAP_PIN)
|
|
|
|
ci->i_pin_ref++;
|
|
|
|
if (got & CEPH_CAP_FILE_RD)
|
|
|
|
ci->i_rd_ref++;
|
|
|
|
if (got & CEPH_CAP_FILE_CACHE)
|
|
|
|
ci->i_rdcache_ref++;
|
2015-04-30 06:40:54 +00:00
|
|
|
if (got & CEPH_CAP_FILE_WR) {
|
|
|
|
if (ci->i_wr_ref == 0 && !ci->i_head_snapc) {
|
|
|
|
BUG_ON(!snap_rwsem_locked);
|
|
|
|
ci->i_head_snapc = ceph_get_snap_context(
|
|
|
|
ci->i_snap_realm->cached_context);
|
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
ci->i_wr_ref++;
|
2015-04-30 06:40:54 +00:00
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
if (got & CEPH_CAP_FILE_BUFFER) {
|
2011-05-11 10:29:54 +00:00
|
|
|
if (ci->i_wb_ref == 0)
|
2011-05-03 16:28:08 +00:00
|
|
|
ihold(&ci->vfs_inode);
|
2011-05-11 10:29:54 +00:00
|
|
|
ci->i_wb_ref++;
|
|
|
|
dout("__take_cap_refs %p wb %d -> %d (?)\n",
|
|
|
|
&ci->vfs_inode, ci->i_wb_ref-1, ci->i_wb_ref);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Try to grab cap references. Specify those refs we @want, and the
|
|
|
|
* minimal set we @need. Also include the larger offset we are writing
|
|
|
|
* to (when applicable), and check against max_size here as well.
|
|
|
|
* Note that caller is responsible for ensuring max_size increases are
|
|
|
|
* requested from the MDS.
|
|
|
|
*/
|
|
|
|
static int try_get_cap_refs(struct ceph_inode_info *ci, int need, int want,
|
2015-04-30 06:40:54 +00:00
|
|
|
loff_t endoff, bool nonblock, int *got, int *err)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
|
|
|
struct inode *inode = &ci->vfs_inode;
|
2015-04-30 06:40:54 +00:00
|
|
|
struct ceph_mds_client *mdsc = ceph_inode_to_client(inode)->mdsc;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
int ret = 0;
|
2015-01-09 07:56:18 +00:00
|
|
|
int have, implemented;
|
2010-03-01 17:57:54 +00:00
|
|
|
int file_wanted;
|
2015-04-30 06:40:54 +00:00
|
|
|
bool snap_rwsem_locked = false;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
dout("get_cap_refs %p need %s want %s\n", inode,
|
|
|
|
ceph_cap_string(need), ceph_cap_string(want));
|
2015-01-09 07:56:18 +00:00
|
|
|
|
2015-04-30 06:40:54 +00:00
|
|
|
again:
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_lock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2010-03-01 17:57:54 +00:00
|
|
|
/* make sure file is actually open */
|
|
|
|
file_wanted = __ceph_caps_file_wanted(ci);
|
2016-04-08 07:27:16 +00:00
|
|
|
if ((file_wanted & need) != need) {
|
2010-03-01 17:57:54 +00:00
|
|
|
dout("try_get_cap_refs need %s file_wanted %s, EBADF\n",
|
|
|
|
ceph_cap_string(need), ceph_cap_string(file_wanted));
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
*err = -EBADF;
|
|
|
|
ret = 1;
|
2014-11-14 14:10:07 +00:00
|
|
|
goto out_unlock;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
2013-04-12 08:11:10 +00:00
|
|
|
/* finish pending truncate */
|
|
|
|
while (ci->i_truncate_pending) {
|
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
2015-04-30 06:40:54 +00:00
|
|
|
if (snap_rwsem_locked) {
|
|
|
|
up_read(&mdsc->snap_rwsem);
|
|
|
|
snap_rwsem_locked = false;
|
|
|
|
}
|
2013-07-02 04:40:19 +00:00
|
|
|
__ceph_do_pending_vmtruncate(inode);
|
2013-04-12 08:11:10 +00:00
|
|
|
spin_lock(&ci->i_ceph_lock);
|
|
|
|
}
|
|
|
|
|
2013-08-05 06:10:29 +00:00
|
|
|
have = __ceph_caps_issued(ci, &implemented);
|
|
|
|
|
|
|
|
if (have & need & CEPH_CAP_FILE_WR) {
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
if (endoff >= 0 && endoff > (loff_t)ci->i_max_size) {
|
|
|
|
dout("get_cap_refs %p endoff %llu > maxsize %llu\n",
|
|
|
|
inode, endoff, ci->i_max_size);
|
2013-08-05 06:10:29 +00:00
|
|
|
if (endoff > ci->i_requested_max_size) {
|
2015-04-30 06:40:54 +00:00
|
|
|
*err = -EAGAIN;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
ret = 1;
|
|
|
|
}
|
2014-11-14 14:10:07 +00:00
|
|
|
goto out_unlock;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
/*
|
|
|
|
* If a sync write is in progress, we must wait, so that we
|
|
|
|
* can get a final snapshot value for size+mtime.
|
|
|
|
*/
|
|
|
|
if (__ceph_have_pending_cap_snap(ci)) {
|
|
|
|
dout("get_cap_refs %p cap_snap_pending\n", inode);
|
2014-11-14 14:10:07 +00:00
|
|
|
goto out_unlock;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if ((have & need) == need) {
|
|
|
|
/*
|
|
|
|
* Look at (implemented & ~have & not) so that we keep waiting
|
|
|
|
* on transition from wanted -> needed caps. This is needed
|
|
|
|
* for WRBUFFER|WR -> WR to avoid a new WR sync write from
|
|
|
|
* going before a prior buffered writeback happens.
|
|
|
|
*/
|
|
|
|
int not = want & ~(have & need);
|
|
|
|
int revoking = implemented & ~have;
|
|
|
|
dout("get_cap_refs %p have %s but not %s (revoking %s)\n",
|
|
|
|
inode, ceph_cap_string(have), ceph_cap_string(not),
|
|
|
|
ceph_cap_string(revoking));
|
|
|
|
if ((revoking & not) == 0) {
|
2015-04-30 06:40:54 +00:00
|
|
|
if (!snap_rwsem_locked &&
|
|
|
|
!ci->i_head_snapc &&
|
|
|
|
(need & CEPH_CAP_FILE_WR)) {
|
|
|
|
if (!down_read_trylock(&mdsc->snap_rwsem)) {
|
|
|
|
/*
|
|
|
|
* we can not call down_read() when
|
|
|
|
* task isn't in TASK_RUNNING state
|
|
|
|
*/
|
|
|
|
if (nonblock) {
|
|
|
|
*err = -EAGAIN;
|
|
|
|
ret = 1;
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
|
|
|
|
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
|
|
|
down_read(&mdsc->snap_rwsem);
|
|
|
|
snap_rwsem_locked = true;
|
|
|
|
goto again;
|
|
|
|
}
|
|
|
|
snap_rwsem_locked = true;
|
|
|
|
}
|
2015-01-09 07:56:18 +00:00
|
|
|
*got = need | (have & want);
|
2016-05-18 12:31:55 +00:00
|
|
|
if ((need & CEPH_CAP_FILE_RD) &&
|
|
|
|
!(*got & CEPH_CAP_FILE_CACHE))
|
|
|
|
ceph_disable_fscache_readpage(ci);
|
2015-04-30 06:40:54 +00:00
|
|
|
__take_cap_refs(ci, *got, true);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
ret = 1;
|
|
|
|
}
|
|
|
|
} else {
|
2015-01-05 03:04:04 +00:00
|
|
|
int session_readonly = false;
|
|
|
|
if ((need & CEPH_CAP_FILE_WR) && ci->i_auth_cap) {
|
|
|
|
struct ceph_mds_session *s = ci->i_auth_cap->session;
|
|
|
|
spin_lock(&s->s_cap_lock);
|
|
|
|
session_readonly = s->s_readonly;
|
|
|
|
spin_unlock(&s->s_cap_lock);
|
|
|
|
}
|
|
|
|
if (session_readonly) {
|
|
|
|
dout("get_cap_refs %p needed %s but mds%d readonly\n",
|
|
|
|
inode, ceph_cap_string(need), ci->i_auth_cap->mds);
|
|
|
|
*err = -EROFS;
|
|
|
|
ret = 1;
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
|
|
|
|
2016-04-08 07:27:16 +00:00
|
|
|
if (ci->i_ceph_flags & CEPH_I_CAP_DROPPED) {
|
|
|
|
int mds_wanted;
|
2016-12-26 09:26:34 +00:00
|
|
|
if (READ_ONCE(mdsc->fsc->mount_state) ==
|
2016-04-08 07:27:16 +00:00
|
|
|
CEPH_MOUNT_SHUTDOWN) {
|
|
|
|
dout("get_cap_refs %p forced umount\n", inode);
|
|
|
|
*err = -EIO;
|
|
|
|
ret = 1;
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
2017-01-29 14:15:47 +00:00
|
|
|
mds_wanted = __ceph_caps_mds_wanted(ci, false);
|
2017-01-12 09:18:00 +00:00
|
|
|
if (need & ~(mds_wanted & need)) {
|
2016-04-08 07:27:16 +00:00
|
|
|
dout("get_cap_refs %p caps were dropped"
|
|
|
|
" (session killed?)\n", inode);
|
|
|
|
*err = -ESTALE;
|
|
|
|
ret = 1;
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
2017-01-12 09:18:00 +00:00
|
|
|
if (!(file_wanted & ~mds_wanted))
|
2016-04-08 07:27:16 +00:00
|
|
|
ci->i_ceph_flags &= ~CEPH_I_CAP_DROPPED;
|
2015-07-01 08:27:46 +00:00
|
|
|
}
|
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
dout("get_cap_refs %p have %s needed %s\n", inode,
|
|
|
|
ceph_cap_string(have), ceph_cap_string(need));
|
|
|
|
}
|
2014-11-14 14:10:07 +00:00
|
|
|
out_unlock:
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
2015-04-30 06:40:54 +00:00
|
|
|
if (snap_rwsem_locked)
|
|
|
|
up_read(&mdsc->snap_rwsem);
|
2014-11-14 14:10:07 +00:00
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
dout("get_cap_refs %p ret %d got %s\n", inode,
|
2015-01-09 07:56:18 +00:00
|
|
|
ret, ceph_cap_string(*got));
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check the offset we are writing up to against our current
|
|
|
|
* max_size. If necessary, tell the MDS we want to write to
|
|
|
|
* a larger offset.
|
|
|
|
*/
|
|
|
|
static void check_max_size(struct inode *inode, loff_t endoff)
|
|
|
|
{
|
|
|
|
struct ceph_inode_info *ci = ceph_inode(inode);
|
|
|
|
int check = 0;
|
|
|
|
|
|
|
|
/* do we need to explicitly request a larger max_size? */
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_lock(&ci->i_ceph_lock);
|
2013-08-05 06:10:29 +00:00
|
|
|
if (endoff >= ci->i_max_size && endoff > ci->i_wanted_max_size) {
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
dout("write %p at large endoff %llu, req max_size\n",
|
|
|
|
inode, endoff);
|
|
|
|
ci->i_wanted_max_size = endoff;
|
|
|
|
}
|
2013-08-05 06:10:29 +00:00
|
|
|
/* duplicate ceph_check_caps()'s logic */
|
|
|
|
if (ci->i_auth_cap &&
|
|
|
|
(ci->i_auth_cap->issued & CEPH_CAP_FILE_WR) &&
|
|
|
|
ci->i_wanted_max_size > ci->i_max_size &&
|
|
|
|
ci->i_wanted_max_size > ci->i_requested_max_size)
|
|
|
|
check = 1;
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
if (check)
|
|
|
|
ceph_check_caps(ci, CHECK_CAPS_AUTHONLY, NULL);
|
|
|
|
}
|
|
|
|
|
2016-10-25 02:51:55 +00:00
|
|
|
int ceph_try_get_caps(struct ceph_inode_info *ci, int need, int want, int *got)
|
|
|
|
{
|
|
|
|
int ret, err = 0;
|
|
|
|
|
|
|
|
BUG_ON(need & ~CEPH_CAP_FILE_RD);
|
|
|
|
BUG_ON(want & ~(CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO));
|
|
|
|
ret = ceph_pool_perm_check(ci, need);
|
|
|
|
if (ret < 0)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
ret = try_get_cap_refs(ci, need, want, 0, true, got, &err);
|
|
|
|
if (ret) {
|
|
|
|
if (err == -EAGAIN) {
|
|
|
|
ret = 0;
|
|
|
|
} else if (err < 0) {
|
|
|
|
ret = err;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
/*
|
|
|
|
* Wait for caps, and take cap references. If we can't get a WR cap
|
|
|
|
* due to a small max_size, make sure we check_max_size (and possibly
|
|
|
|
* ask the mds) so we don't get hung up indefinitely.
|
|
|
|
*/
|
2014-11-14 14:10:07 +00:00
|
|
|
int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
|
|
|
|
loff_t endoff, int *got, struct page **pinned_page)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
2015-04-30 06:40:54 +00:00
|
|
|
int _got, ret, err = 0;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2015-04-27 07:33:28 +00:00
|
|
|
ret = ceph_pool_perm_check(ci, need);
|
|
|
|
if (ret < 0)
|
|
|
|
return ret;
|
|
|
|
|
2015-04-30 06:40:54 +00:00
|
|
|
while (true) {
|
|
|
|
if (endoff > 0)
|
|
|
|
check_max_size(&ci->vfs_inode, endoff);
|
2015-01-09 07:56:18 +00:00
|
|
|
|
2015-04-30 06:40:54 +00:00
|
|
|
err = 0;
|
|
|
|
_got = 0;
|
|
|
|
ret = try_get_cap_refs(ci, need, want, endoff,
|
|
|
|
false, &_got, &err);
|
|
|
|
if (ret) {
|
|
|
|
if (err == -EAGAIN)
|
|
|
|
continue;
|
|
|
|
if (err < 0)
|
2016-04-08 07:27:16 +00:00
|
|
|
ret = err;
|
2015-04-30 06:40:54 +00:00
|
|
|
} else {
|
2016-10-11 09:04:09 +00:00
|
|
|
DEFINE_WAIT_FUNC(wait, woken_wake_function);
|
|
|
|
add_wait_queue(&ci->i_cap_wq, &wait);
|
|
|
|
|
|
|
|
while (!try_get_cap_refs(ci, need, want, endoff,
|
2016-12-22 08:05:43 +00:00
|
|
|
true, &_got, &err)) {
|
|
|
|
if (signal_pending(current)) {
|
|
|
|
ret = -ERESTARTSYS;
|
|
|
|
break;
|
|
|
|
}
|
2016-10-11 09:04:09 +00:00
|
|
|
wait_woken(&wait, TASK_INTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
|
2016-12-22 08:05:43 +00:00
|
|
|
}
|
2016-10-11 09:04:09 +00:00
|
|
|
|
|
|
|
remove_wait_queue(&ci->i_cap_wq, &wait);
|
|
|
|
|
2015-04-30 06:40:54 +00:00
|
|
|
if (err == -EAGAIN)
|
|
|
|
continue;
|
|
|
|
if (err < 0)
|
|
|
|
ret = err;
|
2016-04-08 07:27:16 +00:00
|
|
|
}
|
|
|
|
if (ret < 0) {
|
|
|
|
if (err == -ESTALE) {
|
|
|
|
/* session was killed, try renew caps */
|
|
|
|
ret = ceph_renew_caps(&ci->vfs_inode);
|
|
|
|
if (ret == 0)
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
return ret;
|
2015-04-30 06:40:54 +00:00
|
|
|
}
|
2015-01-09 07:56:18 +00:00
|
|
|
|
2015-04-30 06:40:54 +00:00
|
|
|
if (ci->i_inline_version != CEPH_INLINE_NONE &&
|
|
|
|
(_got & (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO)) &&
|
|
|
|
i_size_read(&ci->vfs_inode) > 0) {
|
|
|
|
struct page *page =
|
|
|
|
find_get_page(ci->vfs_inode.i_mapping, 0);
|
|
|
|
if (page) {
|
|
|
|
if (PageUptodate(page)) {
|
|
|
|
*pinned_page = page;
|
|
|
|
break;
|
|
|
|
}
|
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros
PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time
ago with promise that one day it will be possible to implement page
cache with bigger chunks than PAGE_SIZE.
This promise never materialized. And unlikely will.
We have many places where PAGE_CACHE_SIZE assumed to be equal to
PAGE_SIZE. And it's constant source of confusion on whether
PAGE_CACHE_* or PAGE_* constant should be used in a particular case,
especially on the border between fs and mm.
Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much
breakage to be doable.
Let's stop pretending that pages in page cache are special. They are
not.
The changes are pretty straight-forward:
- <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN};
- page_cache_get() -> get_page();
- page_cache_release() -> put_page();
This patch contains automated changes generated with coccinelle using
script below. For some reason, coccinelle doesn't patch header files.
I've called spatch for them manually.
The only adjustment after coccinelle is revert of changes to
PAGE_CAHCE_ALIGN definition: we are going to drop it later.
There are few places in the code where coccinelle didn't reach. I'll
fix them manually in a separate patch. Comments and documentation also
will be addressed with the separate patch.
virtual patch
@@
expression E;
@@
- E << (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
expression E;
@@
- E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
@@
- PAGE_CACHE_SHIFT
+ PAGE_SHIFT
@@
@@
- PAGE_CACHE_SIZE
+ PAGE_SIZE
@@
@@
- PAGE_CACHE_MASK
+ PAGE_MASK
@@
expression E;
@@
- PAGE_CACHE_ALIGN(E)
+ PAGE_ALIGN(E)
@@
expression E;
@@
- page_cache_get(E)
+ get_page(E)
@@
expression E;
@@
- page_cache_release(E)
+ put_page(E)
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 12:29:47 +00:00
|
|
|
put_page(page);
|
2015-01-09 07:56:18 +00:00
|
|
|
}
|
2015-04-30 06:40:54 +00:00
|
|
|
/*
|
|
|
|
* drop cap refs first because getattr while
|
|
|
|
* holding * caps refs can cause deadlock.
|
|
|
|
*/
|
|
|
|
ceph_put_cap_refs(ci, _got);
|
|
|
|
_got = 0;
|
2015-01-09 07:56:18 +00:00
|
|
|
|
2015-04-30 06:40:54 +00:00
|
|
|
/*
|
|
|
|
* getattr request will bring inline data into
|
|
|
|
* page cache
|
|
|
|
*/
|
|
|
|
ret = __ceph_do_getattr(&ci->vfs_inode, NULL,
|
|
|
|
CEPH_STAT_CAP_INLINE_DATA,
|
|
|
|
true);
|
|
|
|
if (ret < 0)
|
|
|
|
return ret;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
break;
|
2015-01-09 07:56:18 +00:00
|
|
|
}
|
2015-04-30 06:40:54 +00:00
|
|
|
|
2016-05-18 12:31:55 +00:00
|
|
|
if ((_got & CEPH_CAP_FILE_RD) && (_got & CEPH_CAP_FILE_CACHE))
|
|
|
|
ceph_fscache_revalidate_cookie(ci);
|
|
|
|
|
2015-01-09 07:56:18 +00:00
|
|
|
*got = _got;
|
|
|
|
return 0;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Take cap refs. Caller must already know we hold at least one ref
|
|
|
|
* on the caps in question or we don't know this is safe.
|
|
|
|
*/
|
|
|
|
void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps)
|
|
|
|
{
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_lock(&ci->i_ceph_lock);
|
2015-04-30 06:40:54 +00:00
|
|
|
__take_cap_refs(ci, caps, false);
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
2015-05-01 08:57:16 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* drop cap_snap that is not associated with any snapshot.
|
|
|
|
* we don't need to send FLUSHSNAP message for it.
|
|
|
|
*/
|
2016-07-06 08:21:30 +00:00
|
|
|
static int ceph_try_drop_cap_snap(struct ceph_inode_info *ci,
|
|
|
|
struct ceph_cap_snap *capsnap)
|
2015-05-01 08:57:16 +00:00
|
|
|
{
|
|
|
|
if (!capsnap->need_flush &&
|
|
|
|
!capsnap->writing && !capsnap->dirty_pages) {
|
|
|
|
dout("dropping cap_snap %p follows %llu\n",
|
|
|
|
capsnap, capsnap->follows);
|
2016-07-04 10:06:41 +00:00
|
|
|
BUG_ON(capsnap->cap_flush.tid > 0);
|
2015-05-01 08:57:16 +00:00
|
|
|
ceph_put_snap_context(capsnap->context);
|
2016-07-06 08:21:30 +00:00
|
|
|
if (!list_is_last(&capsnap->ci_item, &ci->i_cap_snaps))
|
|
|
|
ci->i_ceph_flags |= CEPH_I_FLUSH_SNAPS;
|
|
|
|
|
2015-05-01 08:57:16 +00:00
|
|
|
list_del(&capsnap->ci_item);
|
|
|
|
ceph_put_cap_snap(capsnap);
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
/*
|
|
|
|
* Release cap refs.
|
|
|
|
*
|
|
|
|
* If we released the last ref on any given cap, call ceph_check_caps
|
|
|
|
* to release (or schedule a release).
|
|
|
|
*
|
|
|
|
* If we are releasing a WR cap (from a sync write), finalize any affected
|
|
|
|
* cap_snap, and wake up any waiters.
|
|
|
|
*/
|
|
|
|
void ceph_put_cap_refs(struct ceph_inode_info *ci, int had)
|
|
|
|
{
|
|
|
|
struct inode *inode = &ci->vfs_inode;
|
|
|
|
int last = 0, put = 0, flushsnaps = 0, wake = 0;
|
|
|
|
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_lock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
if (had & CEPH_CAP_PIN)
|
|
|
|
--ci->i_pin_ref;
|
|
|
|
if (had & CEPH_CAP_FILE_RD)
|
|
|
|
if (--ci->i_rd_ref == 0)
|
|
|
|
last++;
|
|
|
|
if (had & CEPH_CAP_FILE_CACHE)
|
|
|
|
if (--ci->i_rdcache_ref == 0)
|
|
|
|
last++;
|
|
|
|
if (had & CEPH_CAP_FILE_BUFFER) {
|
2011-05-11 10:29:54 +00:00
|
|
|
if (--ci->i_wb_ref == 0) {
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
last++;
|
|
|
|
put++;
|
|
|
|
}
|
2011-05-11 10:29:54 +00:00
|
|
|
dout("put_cap_refs %p wb %d -> %d (?)\n",
|
|
|
|
inode, ci->i_wb_ref+1, ci->i_wb_ref);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
if (had & CEPH_CAP_FILE_WR)
|
|
|
|
if (--ci->i_wr_ref == 0) {
|
|
|
|
last++;
|
2015-05-01 08:57:16 +00:00
|
|
|
if (__ceph_have_pending_cap_snap(ci)) {
|
|
|
|
struct ceph_cap_snap *capsnap =
|
|
|
|
list_last_entry(&ci->i_cap_snaps,
|
|
|
|
struct ceph_cap_snap,
|
|
|
|
ci_item);
|
|
|
|
capsnap->writing = 0;
|
2016-07-06 08:21:30 +00:00
|
|
|
if (ceph_try_drop_cap_snap(ci, capsnap))
|
2015-05-01 08:57:16 +00:00
|
|
|
put++;
|
|
|
|
else if (__ceph_finish_cap_snap(ci, capsnap))
|
|
|
|
flushsnaps = 1;
|
|
|
|
wake = 1;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
2015-04-30 06:40:54 +00:00
|
|
|
if (ci->i_wrbuffer_ref_head == 0 &&
|
|
|
|
ci->i_dirty_caps == 0 &&
|
|
|
|
ci->i_flushing_caps == 0) {
|
|
|
|
BUG_ON(!ci->i_head_snapc);
|
|
|
|
ceph_put_snap_context(ci->i_head_snapc);
|
|
|
|
ci->i_head_snapc = NULL;
|
|
|
|
}
|
2015-03-23 12:12:20 +00:00
|
|
|
/* see comment in __ceph_remove_cap() */
|
|
|
|
if (!__ceph_is_any_caps(ci) && ci->i_snap_realm)
|
|
|
|
drop_inode_snap_realm(ci);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2010-04-01 16:33:46 +00:00
|
|
|
dout("put_cap_refs %p had %s%s%s\n", inode, ceph_cap_string(had),
|
|
|
|
last ? " last" : "", put ? " put" : "");
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
if (last && !flushsnaps)
|
|
|
|
ceph_check_caps(ci, 0, NULL);
|
|
|
|
else if (flushsnaps)
|
2016-07-05 13:08:07 +00:00
|
|
|
ceph_flush_snaps(ci, NULL);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
if (wake)
|
2010-07-27 20:11:08 +00:00
|
|
|
wake_up_all(&ci->i_cap_wq);
|
2015-05-01 08:57:16 +00:00
|
|
|
while (put-- > 0)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
iput(inode);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Release @nr WRBUFFER refs on dirty pages for the given @snapc snap
|
|
|
|
* context. Adjust per-snap dirty page accounting as appropriate.
|
|
|
|
* Once all dirty data for a cap_snap is flushed, flush snapped file
|
|
|
|
* metadata back to the MDS. If we dropped the last ref, call
|
|
|
|
* ceph_check_caps.
|
|
|
|
*/
|
|
|
|
void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
|
|
|
|
struct ceph_snap_context *snapc)
|
|
|
|
{
|
|
|
|
struct inode *inode = &ci->vfs_inode;
|
|
|
|
struct ceph_cap_snap *capsnap = NULL;
|
2016-07-06 08:21:30 +00:00
|
|
|
int put = 0;
|
|
|
|
bool last = false;
|
|
|
|
bool found = false;
|
|
|
|
bool flush_snaps = false;
|
|
|
|
bool complete_capsnap = false;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_lock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
ci->i_wrbuffer_ref -= nr;
|
2016-07-06 08:21:30 +00:00
|
|
|
if (ci->i_wrbuffer_ref == 0) {
|
|
|
|
last = true;
|
|
|
|
put++;
|
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
if (ci->i_head_snapc == snapc) {
|
|
|
|
ci->i_wrbuffer_ref_head -= nr;
|
2010-08-24 15:44:16 +00:00
|
|
|
if (ci->i_wrbuffer_ref_head == 0 &&
|
2015-04-30 06:40:54 +00:00
|
|
|
ci->i_wr_ref == 0 &&
|
|
|
|
ci->i_dirty_caps == 0 &&
|
|
|
|
ci->i_flushing_caps == 0) {
|
2010-08-24 15:44:16 +00:00
|
|
|
BUG_ON(!ci->i_head_snapc);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
ceph_put_snap_context(ci->i_head_snapc);
|
|
|
|
ci->i_head_snapc = NULL;
|
|
|
|
}
|
|
|
|
dout("put_wrbuffer_cap_refs on %p head %d/%d -> %d/%d %s\n",
|
|
|
|
inode,
|
|
|
|
ci->i_wrbuffer_ref+nr, ci->i_wrbuffer_ref_head+nr,
|
|
|
|
ci->i_wrbuffer_ref, ci->i_wrbuffer_ref_head,
|
|
|
|
last ? " LAST" : "");
|
|
|
|
} else {
|
|
|
|
list_for_each_entry(capsnap, &ci->i_cap_snaps, ci_item) {
|
|
|
|
if (capsnap->context == snapc) {
|
2016-07-06 08:21:30 +00:00
|
|
|
found = true;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
BUG_ON(!found);
|
2010-04-01 16:33:46 +00:00
|
|
|
capsnap->dirty_pages -= nr;
|
|
|
|
if (capsnap->dirty_pages == 0) {
|
2016-07-06 08:21:30 +00:00
|
|
|
complete_capsnap = true;
|
|
|
|
if (!capsnap->writing) {
|
|
|
|
if (ceph_try_drop_cap_snap(ci, capsnap)) {
|
|
|
|
put++;
|
|
|
|
} else {
|
|
|
|
ci->i_ceph_flags |= CEPH_I_FLUSH_SNAPS;
|
|
|
|
flush_snaps = true;
|
|
|
|
}
|
|
|
|
}
|
2010-04-01 16:33:46 +00:00
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
dout("put_wrbuffer_cap_refs on %p cap_snap %p "
|
2015-05-01 08:57:16 +00:00
|
|
|
" snap %lld %d/%d -> %d/%d %s%s\n",
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
inode, capsnap, capsnap->context->seq,
|
|
|
|
ci->i_wrbuffer_ref+nr, capsnap->dirty_pages + nr,
|
|
|
|
ci->i_wrbuffer_ref, capsnap->dirty_pages,
|
|
|
|
last ? " (wrbuffer last)" : "",
|
2015-05-01 08:57:16 +00:00
|
|
|
complete_capsnap ? " (complete capsnap)" : "");
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
if (last) {
|
|
|
|
ceph_check_caps(ci, CHECK_CAPS_AUTHONLY, NULL);
|
2016-07-06 08:21:30 +00:00
|
|
|
} else if (flush_snaps) {
|
2016-07-05 13:08:07 +00:00
|
|
|
ceph_flush_snaps(ci, NULL);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
2016-07-06 08:21:30 +00:00
|
|
|
if (complete_capsnap)
|
|
|
|
wake_up_all(&ci->i_cap_wq);
|
|
|
|
while (put-- > 0)
|
2010-04-01 16:33:46 +00:00
|
|
|
iput(inode);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
2013-07-21 02:07:51 +00:00
|
|
|
/*
|
|
|
|
* Invalidate unlinked inode's aliases, so we can drop the inode ASAP.
|
|
|
|
*/
|
|
|
|
static void invalidate_aliases(struct inode *inode)
|
|
|
|
{
|
|
|
|
struct dentry *dn, *prev = NULL;
|
|
|
|
|
|
|
|
dout("invalidate_aliases inode %p\n", inode);
|
|
|
|
d_prune_aliases(inode);
|
|
|
|
/*
|
|
|
|
* For non-directory inode, d_find_alias() only returns
|
2014-01-16 22:42:53 +00:00
|
|
|
* hashed dentry. After calling d_invalidate(), the
|
|
|
|
* dentry becomes unhashed.
|
2013-07-21 02:07:51 +00:00
|
|
|
*
|
2013-09-02 07:19:54 +00:00
|
|
|
* For directory inode, d_find_alias() can return
|
2014-01-16 22:42:53 +00:00
|
|
|
* unhashed dentry. But directory inode should have
|
2013-07-21 02:07:51 +00:00
|
|
|
* one alias at most.
|
|
|
|
*/
|
|
|
|
while ((dn = d_find_alias(inode))) {
|
|
|
|
if (dn == prev) {
|
|
|
|
dput(dn);
|
|
|
|
break;
|
|
|
|
}
|
2013-09-02 07:19:54 +00:00
|
|
|
d_invalidate(dn);
|
2013-07-21 02:07:51 +00:00
|
|
|
if (prev)
|
|
|
|
dput(prev);
|
|
|
|
prev = dn;
|
|
|
|
}
|
|
|
|
if (prev)
|
|
|
|
dput(prev);
|
|
|
|
}
|
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
/*
|
|
|
|
* Handle a cap GRANT message from the MDS. (Note that a GRANT may
|
|
|
|
* actually be a revocation if it specifies a smaller cap set.)
|
|
|
|
*
|
2011-11-30 17:47:09 +00:00
|
|
|
* caller holds s_mutex and i_ceph_lock, we drop both.
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
*/
|
2014-04-18 05:20:27 +00:00
|
|
|
static void handle_cap_grant(struct ceph_mds_client *mdsc,
|
|
|
|
struct inode *inode, struct ceph_mds_caps *grant,
|
2016-03-07 01:35:06 +00:00
|
|
|
struct ceph_string **pns, u64 inline_version,
|
|
|
|
void *inline_data, u32 inline_len,
|
2014-04-18 05:20:27 +00:00
|
|
|
struct ceph_buffer *xattr_buf,
|
2010-03-16 20:42:00 +00:00
|
|
|
struct ceph_mds_session *session,
|
2016-03-07 01:35:06 +00:00
|
|
|
struct ceph_cap *cap, int issued)
|
2014-04-18 05:20:27 +00:00
|
|
|
__releases(ci->i_ceph_lock)
|
2014-12-23 07:30:54 +00:00
|
|
|
__releases(mdsc->snap_rwsem)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
|
|
|
struct ceph_inode_info *ci = ceph_inode(inode);
|
|
|
|
int mds = session->s_mds;
|
2010-10-28 03:59:49 +00:00
|
|
|
int seq = le32_to_cpu(grant->seq);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
int newcaps = le32_to_cpu(grant->caps);
|
2014-04-18 05:20:27 +00:00
|
|
|
int used, wanted, dirty;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
u64 size = le64_to_cpu(grant->size);
|
|
|
|
u64 max_size = le64_to_cpu(grant->max_size);
|
|
|
|
struct timespec mtime, atime, ctime;
|
2010-03-16 20:42:00 +00:00
|
|
|
int check_caps = 0;
|
2014-10-09 21:16:35 +00:00
|
|
|
bool wake = false;
|
|
|
|
bool writeback = false;
|
|
|
|
bool queue_trunc = false;
|
|
|
|
bool queue_invalidate = false;
|
|
|
|
bool deleted_inode = false;
|
2014-11-14 13:41:55 +00:00
|
|
|
bool fill_inline = false;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2010-10-28 03:59:49 +00:00
|
|
|
dout("handle_cap_grant inode %p cap %p mds%d seq %d %s\n",
|
|
|
|
inode, cap, mds, seq, ceph_cap_string(newcaps));
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
dout(" size %llu max_size %llu, i_size %llu\n", size, max_size,
|
|
|
|
inode->i_size);
|
|
|
|
|
2013-11-24 06:44:38 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* auth mds of the inode changed. we received the cap export message,
|
|
|
|
* but still haven't received the cap import message. handle_cap_export
|
|
|
|
* updated the new auth MDS' cap.
|
|
|
|
*
|
|
|
|
* "ceph_seq_cmp(seq, cap->seq) <= 0" means we are processing a message
|
|
|
|
* that was sent before the cap import message. So don't remove caps.
|
|
|
|
*/
|
|
|
|
if (ceph_seq_cmp(seq, cap->seq) <= 0) {
|
|
|
|
WARN_ON(cap != ci->i_auth_cap);
|
|
|
|
WARN_ON(cap->cap_id != le64_to_cpu(grant->cap_id));
|
|
|
|
seq = cap->seq;
|
|
|
|
newcaps |= cap->issued;
|
|
|
|
}
|
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
/*
|
|
|
|
* If CACHE is being revoked, and we have no dirty buffers,
|
|
|
|
* try to invalidate (once). (If there are dirty buffers, we
|
|
|
|
* will invalidate _after_ writeback.)
|
|
|
|
*/
|
2015-06-16 12:48:56 +00:00
|
|
|
if (!S_ISDIR(inode->i_mode) && /* don't invalidate readdir cache */
|
|
|
|
((cap->issued & ~newcaps) & CEPH_CAP_FILE_CACHE) &&
|
2010-06-10 20:20:33 +00:00
|
|
|
(newcaps & CEPH_CAP_FILE_LAZYIO) == 0 &&
|
2016-05-18 12:58:26 +00:00
|
|
|
!(ci->i_wrbuffer_ref || ci->i_wb_ref)) {
|
2013-08-16 05:00:25 +00:00
|
|
|
if (try_nonblocking_invalidate(inode)) {
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
/* there were locked pages.. invalidate later
|
|
|
|
in a separate thread. */
|
|
|
|
if (ci->i_rdcache_revoking != ci->i_rdcache_gen) {
|
2014-10-09 21:16:35 +00:00
|
|
|
queue_invalidate = true;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
ci->i_rdcache_revoking = ci->i_rdcache_gen;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* side effects now are allowed */
|
2009-11-09 20:05:48 +00:00
|
|
|
cap->cap_gen = session->s_cap_gen;
|
2013-11-24 06:44:38 +00:00
|
|
|
cap->seq = seq;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
__check_cap_issue(ci, cap, newcaps);
|
|
|
|
|
2014-04-17 00:55:50 +00:00
|
|
|
if ((newcaps & CEPH_CAP_AUTH_SHARED) &&
|
|
|
|
(issued & CEPH_CAP_AUTH_EXCL) == 0) {
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
inode->i_mode = le32_to_cpu(grant->mode);
|
2013-01-31 10:56:19 +00:00
|
|
|
inode->i_uid = make_kuid(&init_user_ns, le32_to_cpu(grant->uid));
|
|
|
|
inode->i_gid = make_kgid(&init_user_ns, le32_to_cpu(grant->gid));
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
dout("%p mode 0%o uid.gid %d.%d\n", inode, inode->i_mode,
|
2013-01-31 12:05:39 +00:00
|
|
|
from_kuid(&init_user_ns, inode->i_uid),
|
|
|
|
from_kgid(&init_user_ns, inode->i_gid));
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
2014-04-17 00:55:50 +00:00
|
|
|
if ((newcaps & CEPH_CAP_AUTH_SHARED) &&
|
|
|
|
(issued & CEPH_CAP_LINK_EXCL) == 0) {
|
2011-10-28 12:13:29 +00:00
|
|
|
set_nlink(inode, le32_to_cpu(grant->nlink));
|
2013-07-21 02:07:51 +00:00
|
|
|
if (inode->i_nlink == 0 &&
|
|
|
|
(newcaps & (CEPH_CAP_LINK_SHARED | CEPH_CAP_LINK_EXCL)))
|
2014-10-09 21:16:35 +00:00
|
|
|
deleted_inode = true;
|
2013-07-21 02:07:51 +00:00
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
if ((issued & CEPH_CAP_XATTR_EXCL) == 0 && grant->xattr_len) {
|
|
|
|
int len = le32_to_cpu(grant->xattr_len);
|
|
|
|
u64 version = le64_to_cpu(grant->xattr_version);
|
|
|
|
|
|
|
|
if (version > ci->i_xattrs.version) {
|
|
|
|
dout(" got new xattrs v%llu on %p len %d\n",
|
|
|
|
version, inode, len);
|
|
|
|
if (ci->i_xattrs.blob)
|
|
|
|
ceph_buffer_put(ci->i_xattrs.blob);
|
|
|
|
ci->i_xattrs.blob = ceph_buffer_get(xattr_buf);
|
|
|
|
ci->i_xattrs.version = version;
|
2013-11-11 07:18:03 +00:00
|
|
|
ceph_forget_all_cached_acls(inode);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2014-04-17 00:55:50 +00:00
|
|
|
if (newcaps & CEPH_CAP_ANY_RD) {
|
|
|
|
/* ctime/mtime/atime? */
|
|
|
|
ceph_decode_timespec(&mtime, &grant->mtime);
|
|
|
|
ceph_decode_timespec(&atime, &grant->atime);
|
|
|
|
ceph_decode_timespec(&ctime, &grant->ctime);
|
|
|
|
ceph_fill_file_time(inode, issued,
|
|
|
|
le32_to_cpu(grant->time_warp_seq),
|
|
|
|
&ctime, &mtime, &atime);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (newcaps & (CEPH_CAP_ANY_FILE_RD | CEPH_CAP_ANY_FILE_WR)) {
|
|
|
|
/* file layout may have changed */
|
2016-02-03 13:24:49 +00:00
|
|
|
s64 old_pool = ci->i_layout.pool_id;
|
2016-03-07 01:35:06 +00:00
|
|
|
struct ceph_string *old_ns;
|
|
|
|
|
2016-02-03 13:24:49 +00:00
|
|
|
ceph_file_layout_from_legacy(&ci->i_layout, &grant->layout);
|
2016-03-07 01:35:06 +00:00
|
|
|
old_ns = rcu_dereference_protected(ci->i_layout.pool_ns,
|
|
|
|
lockdep_is_held(&ci->i_ceph_lock));
|
|
|
|
rcu_assign_pointer(ci->i_layout.pool_ns, *pns);
|
|
|
|
|
|
|
|
if (ci->i_layout.pool_id != old_pool || *pns != old_ns)
|
2016-02-03 13:24:49 +00:00
|
|
|
ci->i_ceph_flags &= ~CEPH_I_POOL_PERM;
|
2016-02-14 10:06:41 +00:00
|
|
|
|
2016-03-07 01:35:06 +00:00
|
|
|
*pns = old_ns;
|
|
|
|
|
2014-04-17 00:55:50 +00:00
|
|
|
/* size/truncate_seq? */
|
|
|
|
queue_trunc = ceph_fill_file_size(inode, issued,
|
|
|
|
le32_to_cpu(grant->truncate_seq),
|
|
|
|
le64_to_cpu(grant->truncate_size),
|
|
|
|
size);
|
2017-05-16 00:55:34 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
if (ci->i_auth_cap == cap && (newcaps & CEPH_CAP_ANY_FILE_WR)) {
|
|
|
|
if (max_size != ci->i_max_size) {
|
2014-04-17 00:55:50 +00:00
|
|
|
dout("max_size %lld -> %llu\n",
|
|
|
|
ci->i_max_size, max_size);
|
|
|
|
ci->i_max_size = max_size;
|
|
|
|
if (max_size >= ci->i_wanted_max_size) {
|
|
|
|
ci->i_wanted_max_size = 0; /* reset */
|
|
|
|
ci->i_requested_max_size = 0;
|
|
|
|
}
|
2014-10-09 21:16:35 +00:00
|
|
|
wake = true;
|
2017-05-16 00:55:34 +00:00
|
|
|
} else if (ci->i_wanted_max_size > ci->i_max_size &&
|
|
|
|
ci->i_wanted_max_size > ci->i_requested_max_size) {
|
|
|
|
/* CEPH_CAP_OP_IMPORT */
|
|
|
|
wake = true;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* check cap bits */
|
|
|
|
wanted = __ceph_caps_wanted(ci);
|
|
|
|
used = __ceph_caps_used(ci);
|
|
|
|
dirty = __ceph_caps_dirty(ci);
|
|
|
|
dout(" my wanted = %s, used = %s, dirty %s\n",
|
|
|
|
ceph_cap_string(wanted),
|
|
|
|
ceph_cap_string(used),
|
|
|
|
ceph_cap_string(dirty));
|
|
|
|
if (wanted != le32_to_cpu(grant->wanted)) {
|
|
|
|
dout("mds wanted %s -> %s\n",
|
|
|
|
ceph_cap_string(le32_to_cpu(grant->wanted)),
|
|
|
|
ceph_cap_string(wanted));
|
2013-01-04 07:30:10 +00:00
|
|
|
/* imported cap may not have correct mds_wanted */
|
|
|
|
if (le32_to_cpu(grant->op) == CEPH_CAP_OP_IMPORT)
|
|
|
|
check_caps = 1;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/* revocation, grant, or no-op? */
|
|
|
|
if (cap->issued & ~newcaps) {
|
2010-06-10 20:20:33 +00:00
|
|
|
int revoking = cap->issued & ~newcaps;
|
|
|
|
|
|
|
|
dout("revocation: %s -> %s (revoking %s)\n",
|
|
|
|
ceph_cap_string(cap->issued),
|
|
|
|
ceph_cap_string(newcaps),
|
|
|
|
ceph_cap_string(revoking));
|
2010-08-05 20:53:18 +00:00
|
|
|
if (revoking & used & CEPH_CAP_FILE_BUFFER)
|
2014-10-09 21:16:35 +00:00
|
|
|
writeback = true; /* initiate writeback; will delay ack */
|
2010-06-10 20:20:33 +00:00
|
|
|
else if (revoking == CEPH_CAP_FILE_CACHE &&
|
|
|
|
(newcaps & CEPH_CAP_FILE_LAZYIO) == 0 &&
|
|
|
|
queue_invalidate)
|
|
|
|
; /* do nothing yet, invalidation will be queued */
|
|
|
|
else if (cap == ci->i_auth_cap)
|
|
|
|
check_caps = 1; /* check auth cap only */
|
|
|
|
else
|
|
|
|
check_caps = 2; /* check all caps */
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
cap->issued = newcaps;
|
2010-03-08 23:27:53 +00:00
|
|
|
cap->implemented |= newcaps;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
} else if (cap->issued == newcaps) {
|
|
|
|
dout("caps unchanged: %s -> %s\n",
|
|
|
|
ceph_cap_string(cap->issued), ceph_cap_string(newcaps));
|
|
|
|
} else {
|
|
|
|
dout("grant: %s -> %s\n", ceph_cap_string(cap->issued),
|
|
|
|
ceph_cap_string(newcaps));
|
2013-07-02 04:40:21 +00:00
|
|
|
/* non-auth MDS is revoking the newly grant caps ? */
|
|
|
|
if (cap == ci->i_auth_cap &&
|
|
|
|
__ceph_caps_revoking_other(ci, cap, newcaps))
|
|
|
|
check_caps = 2;
|
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
cap->issued = newcaps;
|
|
|
|
cap->implemented |= newcaps; /* add bits only, to
|
|
|
|
* avoid stepping on a
|
|
|
|
* pending revocation */
|
2014-10-09 21:16:35 +00:00
|
|
|
wake = true;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
2010-03-08 23:27:53 +00:00
|
|
|
BUG_ON(cap->issued & ~cap->implemented);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2014-11-14 13:41:55 +00:00
|
|
|
if (inline_version > 0 && inline_version >= ci->i_inline_version) {
|
|
|
|
ci->i_inline_version = inline_version;
|
|
|
|
if (ci->i_inline_version != CEPH_INLINE_NONE &&
|
|
|
|
(newcaps & (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO)))
|
|
|
|
fill_inline = true;
|
|
|
|
}
|
|
|
|
|
2014-04-18 05:20:27 +00:00
|
|
|
if (le32_to_cpu(grant->op) == CEPH_CAP_OP_IMPORT) {
|
|
|
|
if (newcaps & ~issued)
|
2014-10-09 21:16:35 +00:00
|
|
|
wake = true;
|
2016-07-04 10:06:41 +00:00
|
|
|
kick_flushing_inode_caps(mdsc, session, inode);
|
|
|
|
up_read(&mdsc->snap_rwsem);
|
|
|
|
} else {
|
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
2014-04-18 05:20:27 +00:00
|
|
|
}
|
|
|
|
|
2014-11-14 13:41:55 +00:00
|
|
|
if (fill_inline)
|
|
|
|
ceph_fill_inline_data(inode, NULL, inline_data, inline_len);
|
|
|
|
|
2016-05-20 07:41:20 +00:00
|
|
|
if (queue_trunc)
|
2014-04-11 02:18:07 +00:00
|
|
|
ceph_queue_vmtruncate(inode);
|
|
|
|
|
2010-02-09 23:24:44 +00:00
|
|
|
if (writeback)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
/*
|
|
|
|
* queue inode for writeback: we can't actually call
|
|
|
|
* filemap_write_and_wait, etc. from message handler
|
|
|
|
* context.
|
|
|
|
*/
|
2010-02-09 23:24:44 +00:00
|
|
|
ceph_queue_writeback(inode);
|
|
|
|
if (queue_invalidate)
|
|
|
|
ceph_queue_invalidate(inode);
|
2013-07-21 02:07:51 +00:00
|
|
|
if (deleted_inode)
|
|
|
|
invalidate_aliases(inode);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
if (wake)
|
2010-07-27 20:11:08 +00:00
|
|
|
wake_up_all(&ci->i_cap_wq);
|
2010-03-16 20:42:00 +00:00
|
|
|
|
|
|
|
if (check_caps == 1)
|
|
|
|
ceph_check_caps(ci, CHECK_CAPS_NODELAY|CHECK_CAPS_AUTHONLY,
|
|
|
|
session);
|
|
|
|
else if (check_caps == 2)
|
|
|
|
ceph_check_caps(ci, CHECK_CAPS_NODELAY, session);
|
|
|
|
else
|
|
|
|
mutex_unlock(&session->s_mutex);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Handle FLUSH_ACK from MDS, indicating that metadata we sent to the
|
|
|
|
* MDS has been safely committed.
|
|
|
|
*/
|
2009-12-22 19:24:33 +00:00
|
|
|
static void handle_cap_flush_ack(struct inode *inode, u64 flush_tid,
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
struct ceph_mds_caps *m,
|
|
|
|
struct ceph_mds_session *session,
|
|
|
|
struct ceph_cap *cap)
|
2011-11-30 17:47:09 +00:00
|
|
|
__releases(ci->i_ceph_lock)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
|
|
|
struct ceph_inode_info *ci = ceph_inode(inode);
|
2010-04-06 22:14:15 +00:00
|
|
|
struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
|
2016-07-06 03:12:56 +00:00
|
|
|
struct ceph_cap_flush *cf, *tmp_cf;
|
2015-06-09 07:48:57 +00:00
|
|
|
LIST_HEAD(to_remove);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
unsigned seq = le32_to_cpu(m->seq);
|
|
|
|
int dirty = le32_to_cpu(m->dirty);
|
|
|
|
int cleaned = 0;
|
2016-07-07 07:22:38 +00:00
|
|
|
bool drop = false;
|
|
|
|
bool wake_ci = 0;
|
|
|
|
bool wake_mdsc = 0;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2016-07-06 03:12:56 +00:00
|
|
|
list_for_each_entry_safe(cf, tmp_cf, &ci->i_cap_flush_list, i_list) {
|
2015-06-09 07:48:57 +00:00
|
|
|
if (cf->tid == flush_tid)
|
|
|
|
cleaned = cf->caps;
|
2016-07-04 10:06:41 +00:00
|
|
|
if (cf->caps == 0) /* capsnap */
|
|
|
|
continue;
|
2015-06-09 07:48:57 +00:00
|
|
|
if (cf->tid <= flush_tid) {
|
2016-07-07 07:22:38 +00:00
|
|
|
if (__finish_cap_flush(NULL, ci, cf))
|
|
|
|
wake_ci = true;
|
2016-07-06 03:12:56 +00:00
|
|
|
list_add_tail(&cf->i_list, &to_remove);
|
2015-06-09 07:48:57 +00:00
|
|
|
} else {
|
|
|
|
cleaned &= ~cf->caps;
|
|
|
|
if (!cleaned)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
dout("handle_cap_flush_ack inode %p mds%d seq %d on %s cleaned %s,"
|
|
|
|
" flushing %s -> %s\n",
|
|
|
|
inode, session->s_mds, seq, ceph_cap_string(dirty),
|
|
|
|
ceph_cap_string(cleaned), ceph_cap_string(ci->i_flushing_caps),
|
|
|
|
ceph_cap_string(ci->i_flushing_caps & ~cleaned));
|
|
|
|
|
2015-06-09 09:20:12 +00:00
|
|
|
if (list_empty(&to_remove) && !cleaned)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
goto out;
|
|
|
|
|
|
|
|
ci->i_flushing_caps &= ~cleaned;
|
|
|
|
|
|
|
|
spin_lock(&mdsc->cap_dirty_lock);
|
2015-06-09 09:20:12 +00:00
|
|
|
|
2016-07-07 07:22:38 +00:00
|
|
|
list_for_each_entry(cf, &to_remove, i_list) {
|
|
|
|
if (__finish_cap_flush(mdsc, NULL, cf))
|
|
|
|
wake_mdsc = true;
|
2015-06-09 09:20:12 +00:00
|
|
|
}
|
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
if (ci->i_flushing_caps == 0) {
|
2016-07-04 10:06:41 +00:00
|
|
|
if (list_empty(&ci->i_cap_flush_list)) {
|
|
|
|
list_del_init(&ci->i_flushing_item);
|
|
|
|
if (!list_empty(&session->s_cap_flushing)) {
|
|
|
|
dout(" mds%d still flushing cap on %p\n",
|
|
|
|
session->s_mds,
|
|
|
|
&list_first_entry(&session->s_cap_flushing,
|
|
|
|
struct ceph_inode_info,
|
|
|
|
i_flushing_item)->vfs_inode);
|
|
|
|
}
|
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
mdsc->num_cap_flushing--;
|
|
|
|
dout(" inode %p now !flushing\n", inode);
|
2009-10-14 21:27:38 +00:00
|
|
|
|
|
|
|
if (ci->i_dirty_caps == 0) {
|
|
|
|
dout(" inode %p now clean\n", inode);
|
|
|
|
BUG_ON(!list_empty(&ci->i_dirty_item));
|
2016-07-07 07:22:38 +00:00
|
|
|
drop = true;
|
2015-04-30 06:40:54 +00:00
|
|
|
if (ci->i_wr_ref == 0 &&
|
|
|
|
ci->i_wrbuffer_ref_head == 0) {
|
2010-08-24 15:44:16 +00:00
|
|
|
BUG_ON(!ci->i_head_snapc);
|
|
|
|
ceph_put_snap_context(ci->i_head_snapc);
|
|
|
|
ci->i_head_snapc = NULL;
|
|
|
|
}
|
2009-10-16 01:13:53 +00:00
|
|
|
} else {
|
|
|
|
BUG_ON(list_empty(&ci->i_dirty_item));
|
2009-10-14 21:27:38 +00:00
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
spin_unlock(&mdsc->cap_dirty_lock);
|
|
|
|
|
|
|
|
out:
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
2015-06-09 07:48:57 +00:00
|
|
|
|
|
|
|
while (!list_empty(&to_remove)) {
|
|
|
|
cf = list_first_entry(&to_remove,
|
2016-07-06 03:12:56 +00:00
|
|
|
struct ceph_cap_flush, i_list);
|
|
|
|
list_del(&cf->i_list);
|
2015-06-10 09:26:13 +00:00
|
|
|
ceph_free_cap_flush(cf);
|
2015-06-09 07:48:57 +00:00
|
|
|
}
|
2016-07-07 07:22:38 +00:00
|
|
|
|
|
|
|
if (wake_ci)
|
|
|
|
wake_up_all(&ci->i_cap_wq);
|
|
|
|
if (wake_mdsc)
|
|
|
|
wake_up_all(&mdsc->cap_flushing_wq);
|
2009-10-14 21:27:38 +00:00
|
|
|
if (drop)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
iput(inode);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Handle FLUSHSNAP_ACK. MDS has flushed snap data to disk and we can
|
|
|
|
* throw away our cap_snap.
|
|
|
|
*
|
|
|
|
* Caller hold s_mutex.
|
|
|
|
*/
|
2009-12-22 19:24:33 +00:00
|
|
|
static void handle_cap_flushsnap_ack(struct inode *inode, u64 flush_tid,
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
struct ceph_mds_caps *m,
|
|
|
|
struct ceph_mds_session *session)
|
|
|
|
{
|
|
|
|
struct ceph_inode_info *ci = ceph_inode(inode);
|
2015-05-05 13:22:13 +00:00
|
|
|
struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
u64 follows = le64_to_cpu(m->snap_follows);
|
|
|
|
struct ceph_cap_snap *capsnap;
|
2016-07-07 07:22:38 +00:00
|
|
|
bool flushed = false;
|
|
|
|
bool wake_ci = false;
|
|
|
|
bool wake_mdsc = false;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
dout("handle_cap_flushsnap_ack inode %p ci %p mds%d follows %lld\n",
|
|
|
|
inode, ci, session->s_mds, follows);
|
|
|
|
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_lock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
list_for_each_entry(capsnap, &ci->i_cap_snaps, ci_item) {
|
|
|
|
if (capsnap->follows == follows) {
|
2016-07-04 10:06:41 +00:00
|
|
|
if (capsnap->cap_flush.tid != flush_tid) {
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
dout(" cap_snap %p follows %lld tid %lld !="
|
|
|
|
" %lld\n", capsnap, follows,
|
2016-07-04 10:06:41 +00:00
|
|
|
flush_tid, capsnap->cap_flush.tid);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
break;
|
|
|
|
}
|
2016-07-07 07:22:38 +00:00
|
|
|
flushed = true;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
break;
|
|
|
|
} else {
|
|
|
|
dout(" skipping cap_snap %p follows %lld\n",
|
|
|
|
capsnap, capsnap->follows);
|
|
|
|
}
|
|
|
|
}
|
2016-07-04 10:06:41 +00:00
|
|
|
if (flushed) {
|
|
|
|
WARN_ON(capsnap->dirty_pages || capsnap->writing);
|
|
|
|
dout(" removing %p cap_snap %p follows %lld\n",
|
|
|
|
inode, capsnap, follows);
|
|
|
|
list_del(&capsnap->ci_item);
|
2016-07-07 07:22:38 +00:00
|
|
|
if (__finish_cap_flush(NULL, ci, &capsnap->cap_flush))
|
|
|
|
wake_ci = true;
|
2016-07-04 10:06:41 +00:00
|
|
|
|
|
|
|
spin_lock(&mdsc->cap_dirty_lock);
|
|
|
|
|
|
|
|
if (list_empty(&ci->i_cap_flush_list))
|
|
|
|
list_del_init(&ci->i_flushing_item);
|
|
|
|
|
2016-07-07 07:22:38 +00:00
|
|
|
if (__finish_cap_flush(mdsc, NULL, &capsnap->cap_flush))
|
|
|
|
wake_mdsc = true;
|
2016-07-04 10:06:41 +00:00
|
|
|
|
|
|
|
spin_unlock(&mdsc->cap_dirty_lock);
|
|
|
|
}
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
2016-07-04 10:06:41 +00:00
|
|
|
if (flushed) {
|
|
|
|
ceph_put_snap_context(capsnap->context);
|
|
|
|
ceph_put_cap_snap(capsnap);
|
2016-07-07 07:22:38 +00:00
|
|
|
if (wake_ci)
|
|
|
|
wake_up_all(&ci->i_cap_wq);
|
|
|
|
if (wake_mdsc)
|
|
|
|
wake_up_all(&mdsc->cap_flushing_wq);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
iput(inode);
|
2016-07-04 10:06:41 +00:00
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Handle TRUNC from MDS, indicating file truncation.
|
|
|
|
*
|
|
|
|
* caller hold s_mutex.
|
|
|
|
*/
|
|
|
|
static void handle_cap_trunc(struct inode *inode,
|
|
|
|
struct ceph_mds_caps *trunc,
|
|
|
|
struct ceph_mds_session *session)
|
2011-11-30 17:47:09 +00:00
|
|
|
__releases(ci->i_ceph_lock)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
|
|
|
struct ceph_inode_info *ci = ceph_inode(inode);
|
|
|
|
int mds = session->s_mds;
|
|
|
|
int seq = le32_to_cpu(trunc->seq);
|
|
|
|
u32 truncate_seq = le32_to_cpu(trunc->truncate_seq);
|
|
|
|
u64 truncate_size = le64_to_cpu(trunc->truncate_size);
|
|
|
|
u64 size = le64_to_cpu(trunc->size);
|
|
|
|
int implemented = 0;
|
|
|
|
int dirty = __ceph_caps_dirty(ci);
|
|
|
|
int issued = __ceph_caps_issued(ceph_inode(inode), &implemented);
|
|
|
|
int queue_trunc = 0;
|
|
|
|
|
|
|
|
issued |= implemented | dirty;
|
|
|
|
|
|
|
|
dout("handle_cap_trunc inode %p mds%d seq %d to %lld seq %d\n",
|
|
|
|
inode, mds, seq, truncate_size, truncate_seq);
|
|
|
|
queue_trunc = ceph_fill_file_size(inode, issued,
|
|
|
|
truncate_seq, truncate_size, size);
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2016-05-20 07:41:20 +00:00
|
|
|
if (queue_trunc)
|
2010-02-09 23:24:44 +00:00
|
|
|
ceph_queue_vmtruncate(inode);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Handle EXPORT from MDS. Cap is being migrated _from_ this mds to a
|
|
|
|
* different one. If we are the most recent migration we've seen (as
|
|
|
|
* indicated by mseq), make note of the migrating cap bits for the
|
|
|
|
* duration (until we see the corresponding IMPORT).
|
|
|
|
*
|
|
|
|
* caller holds s_mutex
|
|
|
|
*/
|
|
|
|
static void handle_cap_export(struct inode *inode, struct ceph_mds_caps *ex,
|
2013-11-24 06:44:38 +00:00
|
|
|
struct ceph_mds_cap_peer *ph,
|
|
|
|
struct ceph_mds_session *session)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
2011-05-24 18:46:31 +00:00
|
|
|
struct ceph_mds_client *mdsc = ceph_inode_to_client(inode)->mdsc;
|
2013-11-24 06:44:38 +00:00
|
|
|
struct ceph_mds_session *tsession = NULL;
|
2014-04-18 01:57:11 +00:00
|
|
|
struct ceph_cap *cap, *tcap, *new_cap = NULL;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
struct ceph_inode_info *ci = ceph_inode(inode);
|
2013-11-24 06:44:38 +00:00
|
|
|
u64 t_cap_id;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
unsigned mseq = le32_to_cpu(ex->migrate_seq);
|
2013-11-24 06:44:38 +00:00
|
|
|
unsigned t_seq, t_mseq;
|
|
|
|
int target, issued;
|
|
|
|
int mds = session->s_mds;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2013-11-24 06:44:38 +00:00
|
|
|
if (ph) {
|
|
|
|
t_cap_id = le64_to_cpu(ph->cap_id);
|
|
|
|
t_seq = le32_to_cpu(ph->seq);
|
|
|
|
t_mseq = le32_to_cpu(ph->mseq);
|
|
|
|
target = le32_to_cpu(ph->mds);
|
|
|
|
} else {
|
|
|
|
t_cap_id = t_seq = t_mseq = 0;
|
|
|
|
target = -1;
|
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2013-11-24 06:44:38 +00:00
|
|
|
dout("handle_cap_export inode %p ci %p mds%d mseq %d target %d\n",
|
|
|
|
inode, ci, mds, mseq, target);
|
|
|
|
retry:
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_lock(&ci->i_ceph_lock);
|
2013-11-24 06:44:38 +00:00
|
|
|
cap = __get_cap_for_mds(ci, mds);
|
2014-04-21 07:46:37 +00:00
|
|
|
if (!cap || cap->cap_id != le64_to_cpu(ex->cap_id))
|
2013-11-24 06:44:38 +00:00
|
|
|
goto out_unlock;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2013-11-24 06:44:38 +00:00
|
|
|
if (target < 0) {
|
|
|
|
__ceph_remove_cap(cap, false);
|
2016-04-08 07:27:16 +00:00
|
|
|
if (!ci->i_auth_cap)
|
|
|
|
ci->i_ceph_flags |= CEPH_I_CAP_DROPPED;
|
2013-11-24 06:44:38 +00:00
|
|
|
goto out_unlock;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
2013-11-24 06:44:38 +00:00
|
|
|
/*
|
|
|
|
* now we know we haven't received the cap import message yet
|
|
|
|
* because the exported cap still exist.
|
|
|
|
*/
|
2011-05-24 18:46:31 +00:00
|
|
|
|
2013-11-24 06:44:38 +00:00
|
|
|
issued = cap->issued;
|
|
|
|
WARN_ON(issued != cap->implemented);
|
|
|
|
|
|
|
|
tcap = __get_cap_for_mds(ci, target);
|
|
|
|
if (tcap) {
|
|
|
|
/* already have caps from the target */
|
|
|
|
if (tcap->cap_id != t_cap_id ||
|
|
|
|
ceph_seq_cmp(tcap->seq, t_seq) < 0) {
|
|
|
|
dout(" updating import cap %p mds%d\n", tcap, target);
|
|
|
|
tcap->cap_id = t_cap_id;
|
|
|
|
tcap->seq = t_seq - 1;
|
|
|
|
tcap->issue_seq = t_seq - 1;
|
|
|
|
tcap->mseq = t_mseq;
|
|
|
|
tcap->issued |= issued;
|
|
|
|
tcap->implemented |= issued;
|
|
|
|
if (cap == ci->i_auth_cap)
|
|
|
|
ci->i_auth_cap = tcap;
|
2017-01-24 02:02:32 +00:00
|
|
|
|
2016-07-04 10:06:41 +00:00
|
|
|
if (!list_empty(&ci->i_cap_flush_list) &&
|
|
|
|
ci->i_auth_cap == tcap) {
|
2013-11-24 06:44:38 +00:00
|
|
|
spin_lock(&mdsc->cap_dirty_lock);
|
|
|
|
list_move_tail(&ci->i_flushing_item,
|
|
|
|
&tcap->session->s_cap_flushing);
|
|
|
|
spin_unlock(&mdsc->cap_dirty_lock);
|
2011-05-24 18:46:31 +00:00
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
2013-09-22 02:15:58 +00:00
|
|
|
__ceph_remove_cap(cap, false);
|
2013-11-24 06:44:38 +00:00
|
|
|
goto out_unlock;
|
2014-04-18 01:57:11 +00:00
|
|
|
} else if (tsession) {
|
2013-11-24 06:44:38 +00:00
|
|
|
/* add placeholder for the export tagert */
|
2014-04-18 01:57:11 +00:00
|
|
|
int flag = (cap == ci->i_auth_cap) ? CEPH_CAP_FLAG_AUTH : 0;
|
2017-01-24 02:02:32 +00:00
|
|
|
tcap = new_cap;
|
2013-11-24 06:44:38 +00:00
|
|
|
ceph_add_cap(inode, tsession, t_cap_id, -1, issued, 0,
|
2014-04-18 01:57:11 +00:00
|
|
|
t_seq - 1, t_mseq, (u64)-1, flag, &new_cap);
|
|
|
|
|
2017-01-24 02:02:32 +00:00
|
|
|
if (!list_empty(&ci->i_cap_flush_list) &&
|
|
|
|
ci->i_auth_cap == tcap) {
|
|
|
|
spin_lock(&mdsc->cap_dirty_lock);
|
|
|
|
list_move_tail(&ci->i_flushing_item,
|
|
|
|
&tcap->session->s_cap_flushing);
|
|
|
|
spin_unlock(&mdsc->cap_dirty_lock);
|
|
|
|
}
|
|
|
|
|
2014-04-18 01:57:11 +00:00
|
|
|
__ceph_remove_cap(cap, false);
|
|
|
|
goto out_unlock;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
2013-11-24 06:44:38 +00:00
|
|
|
mutex_unlock(&session->s_mutex);
|
|
|
|
|
|
|
|
/* open target session */
|
|
|
|
tsession = ceph_mdsc_open_export_target_session(mdsc, target);
|
|
|
|
if (!IS_ERR(tsession)) {
|
|
|
|
if (mds > target) {
|
|
|
|
mutex_lock(&session->s_mutex);
|
|
|
|
mutex_lock_nested(&tsession->s_mutex,
|
|
|
|
SINGLE_DEPTH_NESTING);
|
|
|
|
} else {
|
|
|
|
mutex_lock(&tsession->s_mutex);
|
|
|
|
mutex_lock_nested(&session->s_mutex,
|
|
|
|
SINGLE_DEPTH_NESTING);
|
|
|
|
}
|
2014-04-18 01:57:11 +00:00
|
|
|
new_cap = ceph_get_cap(mdsc, NULL);
|
2013-11-24 06:44:38 +00:00
|
|
|
} else {
|
|
|
|
WARN_ON(1);
|
|
|
|
tsession = NULL;
|
|
|
|
target = -1;
|
|
|
|
}
|
|
|
|
goto retry;
|
|
|
|
|
|
|
|
out_unlock:
|
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
|
|
|
mutex_unlock(&session->s_mutex);
|
|
|
|
if (tsession) {
|
|
|
|
mutex_unlock(&tsession->s_mutex);
|
|
|
|
ceph_put_mds_session(tsession);
|
|
|
|
}
|
2014-04-18 01:57:11 +00:00
|
|
|
if (new_cap)
|
|
|
|
ceph_put_cap(mdsc, new_cap);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2014-04-18 05:20:27 +00:00
|
|
|
* Handle cap IMPORT.
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
*
|
2014-04-18 05:20:27 +00:00
|
|
|
* caller holds s_mutex. acquires i_ceph_lock
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
*/
|
|
|
|
static void handle_cap_import(struct ceph_mds_client *mdsc,
|
|
|
|
struct inode *inode, struct ceph_mds_caps *im,
|
2013-11-24 06:43:46 +00:00
|
|
|
struct ceph_mds_cap_peer *ph,
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
struct ceph_mds_session *session,
|
2014-04-18 05:20:27 +00:00
|
|
|
struct ceph_cap **target_cap, int *old_issued)
|
|
|
|
__acquires(ci->i_ceph_lock)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
|
|
|
struct ceph_inode_info *ci = ceph_inode(inode);
|
2014-04-18 05:20:27 +00:00
|
|
|
struct ceph_cap *cap, *ocap, *new_cap = NULL;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
int mds = session->s_mds;
|
2014-04-18 05:20:27 +00:00
|
|
|
int issued;
|
|
|
|
unsigned caps = le32_to_cpu(im->caps);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
unsigned wanted = le32_to_cpu(im->wanted);
|
|
|
|
unsigned seq = le32_to_cpu(im->seq);
|
|
|
|
unsigned mseq = le32_to_cpu(im->migrate_seq);
|
|
|
|
u64 realmino = le64_to_cpu(im->realm);
|
|
|
|
u64 cap_id = le64_to_cpu(im->cap_id);
|
2013-11-24 06:43:46 +00:00
|
|
|
u64 p_cap_id;
|
|
|
|
int peer;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2013-11-24 06:43:46 +00:00
|
|
|
if (ph) {
|
|
|
|
p_cap_id = le64_to_cpu(ph->cap_id);
|
|
|
|
peer = le32_to_cpu(ph->mds);
|
|
|
|
} else {
|
|
|
|
p_cap_id = 0;
|
|
|
|
peer = -1;
|
|
|
|
}
|
2011-05-24 18:46:31 +00:00
|
|
|
|
2013-11-24 06:43:46 +00:00
|
|
|
dout("handle_cap_import inode %p ci %p mds%d mseq %d peer %d\n",
|
|
|
|
inode, ci, mds, mseq, peer);
|
|
|
|
|
2014-04-18 01:57:11 +00:00
|
|
|
retry:
|
2013-11-24 06:43:46 +00:00
|
|
|
spin_lock(&ci->i_ceph_lock);
|
2014-04-18 01:57:11 +00:00
|
|
|
cap = __get_cap_for_mds(ci, mds);
|
|
|
|
if (!cap) {
|
|
|
|
if (!new_cap) {
|
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
|
|
|
new_cap = ceph_get_cap(mdsc, NULL);
|
|
|
|
goto retry;
|
|
|
|
}
|
2014-04-18 05:20:27 +00:00
|
|
|
cap = new_cap;
|
|
|
|
} else {
|
|
|
|
if (new_cap) {
|
|
|
|
ceph_put_cap(mdsc, new_cap);
|
|
|
|
new_cap = NULL;
|
|
|
|
}
|
2014-04-18 01:57:11 +00:00
|
|
|
}
|
|
|
|
|
2014-04-18 05:20:27 +00:00
|
|
|
__ceph_caps_issued(ci, &issued);
|
|
|
|
issued |= __ceph_caps_dirty(ci);
|
|
|
|
|
|
|
|
ceph_add_cap(inode, session, cap_id, -1, caps, wanted, seq, mseq,
|
2014-04-18 01:57:11 +00:00
|
|
|
realmino, CEPH_CAP_FLAG_AUTH, &new_cap);
|
|
|
|
|
2014-04-18 05:20:27 +00:00
|
|
|
ocap = peer >= 0 ? __get_cap_for_mds(ci, peer) : NULL;
|
|
|
|
if (ocap && ocap->cap_id == p_cap_id) {
|
2013-11-24 06:43:46 +00:00
|
|
|
dout(" remove export cap %p mds%d flags %d\n",
|
2014-04-18 05:20:27 +00:00
|
|
|
ocap, peer, ph->flags);
|
2013-11-24 06:43:46 +00:00
|
|
|
if ((ph->flags & CEPH_CAP_FLAG_AUTH) &&
|
2014-04-18 05:20:27 +00:00
|
|
|
(ocap->seq != le32_to_cpu(ph->seq) ||
|
|
|
|
ocap->mseq != le32_to_cpu(ph->mseq))) {
|
2013-11-24 06:43:46 +00:00
|
|
|
pr_err("handle_cap_import: mismatched seq/mseq: "
|
|
|
|
"ino (%llx.%llx) mds%d seq %d mseq %d "
|
|
|
|
"importer mds%d has peer seq %d mseq %d\n",
|
2014-04-18 05:20:27 +00:00
|
|
|
ceph_vinop(inode), peer, ocap->seq,
|
|
|
|
ocap->mseq, mds, le32_to_cpu(ph->seq),
|
2013-11-24 06:43:46 +00:00
|
|
|
le32_to_cpu(ph->mseq));
|
2011-05-24 18:46:31 +00:00
|
|
|
}
|
2014-04-18 05:20:27 +00:00
|
|
|
__ceph_remove_cap(ocap, (ph->flags & CEPH_CAP_FLAG_RELEASE));
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
2013-11-24 06:43:46 +00:00
|
|
|
/* make sure we re-request max_size, if necessary */
|
|
|
|
ci->i_requested_max_size = 0;
|
2014-04-18 01:57:11 +00:00
|
|
|
|
2014-04-18 05:20:27 +00:00
|
|
|
*old_issued = issued;
|
|
|
|
*target_cap = cap;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Handle a caps message from the MDS.
|
|
|
|
*
|
|
|
|
* Identify the appropriate session, inode, and call the right handler
|
|
|
|
* based on the cap op.
|
|
|
|
*/
|
|
|
|
void ceph_handle_caps(struct ceph_mds_session *session,
|
|
|
|
struct ceph_msg *msg)
|
|
|
|
{
|
|
|
|
struct ceph_mds_client *mdsc = session->s_mdsc;
|
2010-04-06 22:14:15 +00:00
|
|
|
struct super_block *sb = mdsc->fsc->sb;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
struct inode *inode;
|
2011-11-30 17:47:09 +00:00
|
|
|
struct ceph_inode_info *ci;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
struct ceph_cap *cap;
|
|
|
|
struct ceph_mds_caps *h;
|
2013-11-24 06:43:46 +00:00
|
|
|
struct ceph_mds_cap_peer *peer = NULL;
|
2016-03-07 01:35:06 +00:00
|
|
|
struct ceph_snap_realm *realm = NULL;
|
|
|
|
struct ceph_string *pool_ns = NULL;
|
2010-02-22 23:12:16 +00:00
|
|
|
int mds = session->s_mds;
|
2014-04-18 05:20:27 +00:00
|
|
|
int op, issued;
|
2010-06-09 23:47:10 +00:00
|
|
|
u32 seq, mseq;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
struct ceph_vino vino;
|
2009-12-22 19:24:33 +00:00
|
|
|
u64 tid;
|
2014-11-14 13:29:55 +00:00
|
|
|
u64 inline_version = 0;
|
|
|
|
void *inline_data = NULL;
|
|
|
|
u32 inline_len = 0;
|
2010-03-01 21:20:50 +00:00
|
|
|
void *snaptrace;
|
2010-08-02 22:09:39 +00:00
|
|
|
size_t snaptrace_len;
|
2014-11-14 13:29:55 +00:00
|
|
|
void *p, *end;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
dout("handle_caps from mds%d\n", mds);
|
|
|
|
|
|
|
|
/* decode */
|
2013-11-24 06:43:46 +00:00
|
|
|
end = msg->front.iov_base + msg->front.iov_len;
|
2009-12-22 19:24:33 +00:00
|
|
|
tid = le64_to_cpu(msg->hdr.tid);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
if (msg->front.iov_len < sizeof(*h))
|
|
|
|
goto bad;
|
|
|
|
h = msg->front.iov_base;
|
|
|
|
op = le32_to_cpu(h->op);
|
|
|
|
vino.ino = le64_to_cpu(h->ino);
|
|
|
|
vino.snap = CEPH_NOSNAP;
|
|
|
|
seq = le32_to_cpu(h->seq);
|
2010-06-09 23:47:10 +00:00
|
|
|
mseq = le32_to_cpu(h->migrate_seq);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
2010-08-02 22:09:39 +00:00
|
|
|
snaptrace = h + 1;
|
|
|
|
snaptrace_len = le32_to_cpu(h->snap_trace_len);
|
2014-11-14 13:29:55 +00:00
|
|
|
p = snaptrace + snaptrace_len;
|
2010-08-02 22:09:39 +00:00
|
|
|
|
|
|
|
if (le16_to_cpu(msg->hdr.version) >= 2) {
|
2014-11-14 13:29:55 +00:00
|
|
|
u32 flock_len;
|
2010-08-02 22:09:39 +00:00
|
|
|
ceph_decode_32_safe(&p, end, flock_len, bad);
|
2013-11-24 06:43:46 +00:00
|
|
|
if (p + flock_len > end)
|
|
|
|
goto bad;
|
2014-11-14 13:29:55 +00:00
|
|
|
p += flock_len;
|
2010-08-02 22:09:39 +00:00
|
|
|
}
|
|
|
|
|
2013-11-24 06:43:46 +00:00
|
|
|
if (le16_to_cpu(msg->hdr.version) >= 3) {
|
|
|
|
if (op == CEPH_CAP_OP_IMPORT) {
|
|
|
|
if (p + sizeof(*peer) > end)
|
|
|
|
goto bad;
|
|
|
|
peer = p;
|
2014-11-14 13:29:55 +00:00
|
|
|
p += sizeof(*peer);
|
2013-11-24 06:44:38 +00:00
|
|
|
} else if (op == CEPH_CAP_OP_EXPORT) {
|
|
|
|
/* recorded in unused fields */
|
|
|
|
peer = (void *)&h->size;
|
2013-11-24 06:43:46 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2014-11-14 13:29:55 +00:00
|
|
|
if (le16_to_cpu(msg->hdr.version) >= 4) {
|
|
|
|
ceph_decode_64_safe(&p, end, inline_version, bad);
|
|
|
|
ceph_decode_32_safe(&p, end, inline_len, bad);
|
|
|
|
if (p + inline_len > end)
|
|
|
|
goto bad;
|
|
|
|
inline_data = p;
|
|
|
|
p += inline_len;
|
|
|
|
}
|
|
|
|
|
2017-04-13 15:07:04 +00:00
|
|
|
if (le16_to_cpu(msg->hdr.version) >= 5) {
|
|
|
|
struct ceph_osd_client *osdc = &mdsc->fsc->client->osdc;
|
|
|
|
u32 epoch_barrier;
|
|
|
|
|
|
|
|
ceph_decode_32_safe(&p, end, epoch_barrier, bad);
|
|
|
|
ceph_osdc_update_epoch_barrier(osdc, epoch_barrier);
|
|
|
|
}
|
|
|
|
|
2016-02-14 10:06:41 +00:00
|
|
|
if (le16_to_cpu(msg->hdr.version) >= 8) {
|
|
|
|
u64 flush_tid;
|
|
|
|
u32 caller_uid, caller_gid;
|
2016-03-07 01:35:06 +00:00
|
|
|
u32 pool_ns_len;
|
2017-04-13 15:07:04 +00:00
|
|
|
|
2016-02-14 10:06:41 +00:00
|
|
|
/* version >= 6 */
|
|
|
|
ceph_decode_64_safe(&p, end, flush_tid, bad);
|
|
|
|
/* version >= 7 */
|
|
|
|
ceph_decode_32_safe(&p, end, caller_uid, bad);
|
|
|
|
ceph_decode_32_safe(&p, end, caller_gid, bad);
|
|
|
|
/* version >= 8 */
|
|
|
|
ceph_decode_32_safe(&p, end, pool_ns_len, bad);
|
2016-03-07 01:35:06 +00:00
|
|
|
if (pool_ns_len > 0) {
|
|
|
|
ceph_decode_need(&p, end, pool_ns_len, bad);
|
|
|
|
pool_ns = ceph_find_or_create_string(p, pool_ns_len);
|
|
|
|
p += pool_ns_len;
|
|
|
|
}
|
2016-02-14 10:06:41 +00:00
|
|
|
}
|
|
|
|
|
2014-09-16 23:45:12 +00:00
|
|
|
/* lookup ino */
|
|
|
|
inode = ceph_find_inode(sb, vino);
|
|
|
|
ci = ceph_inode(inode);
|
|
|
|
dout(" op %s ino %llx.%llx inode %p\n", ceph_cap_op_name(op), vino.ino,
|
|
|
|
vino.snap, inode);
|
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
mutex_lock(&session->s_mutex);
|
|
|
|
session->s_seq++;
|
|
|
|
dout(" mds%d seq %lld cap seq %u\n", session->s_mds, session->s_seq,
|
|
|
|
(unsigned)seq);
|
|
|
|
|
|
|
|
if (!inode) {
|
|
|
|
dout(" i don't have ino %llx\n", vino.ino);
|
2010-06-09 23:47:10 +00:00
|
|
|
|
2013-09-22 02:15:58 +00:00
|
|
|
if (op == CEPH_CAP_OP_IMPORT) {
|
2015-05-14 09:22:42 +00:00
|
|
|
cap = ceph_get_cap(mdsc, NULL);
|
|
|
|
cap->cap_ino = vino.ino;
|
|
|
|
cap->queue_release = 1;
|
2016-03-07 01:35:06 +00:00
|
|
|
cap->cap_id = le64_to_cpu(h->cap_id);
|
2015-05-14 09:22:42 +00:00
|
|
|
cap->mseq = mseq;
|
|
|
|
cap->seq = seq;
|
2016-11-17 11:55:30 +00:00
|
|
|
cap->issue_seq = seq;
|
2013-09-22 02:15:58 +00:00
|
|
|
spin_lock(&session->s_cap_lock);
|
2015-05-14 09:22:42 +00:00
|
|
|
list_add_tail(&cap->session_caps,
|
|
|
|
&session->s_cap_releases);
|
|
|
|
session->s_num_cap_releases++;
|
2013-09-22 02:15:58 +00:00
|
|
|
spin_unlock(&session->s_cap_lock);
|
|
|
|
}
|
2010-10-06 22:46:30 +00:00
|
|
|
goto flush_cap_releases;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/* these will work even if we don't have a cap yet */
|
|
|
|
switch (op) {
|
|
|
|
case CEPH_CAP_OP_FLUSHSNAP_ACK:
|
2009-12-22 19:24:33 +00:00
|
|
|
handle_cap_flushsnap_ack(inode, tid, h, session);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
goto done;
|
|
|
|
|
|
|
|
case CEPH_CAP_OP_EXPORT:
|
2013-11-24 06:44:38 +00:00
|
|
|
handle_cap_export(inode, h, peer, session);
|
|
|
|
goto done_unlocked;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
case CEPH_CAP_OP_IMPORT:
|
2014-12-23 07:30:54 +00:00
|
|
|
realm = NULL;
|
|
|
|
if (snaptrace_len) {
|
|
|
|
down_write(&mdsc->snap_rwsem);
|
|
|
|
ceph_update_snap_trace(mdsc, snaptrace,
|
|
|
|
snaptrace + snaptrace_len,
|
|
|
|
false, &realm);
|
|
|
|
downgrade_write(&mdsc->snap_rwsem);
|
|
|
|
} else {
|
|
|
|
down_read(&mdsc->snap_rwsem);
|
|
|
|
}
|
2013-11-24 06:43:46 +00:00
|
|
|
handle_cap_import(mdsc, inode, h, peer, session,
|
2014-04-18 05:20:27 +00:00
|
|
|
&cap, &issued);
|
2016-03-07 01:35:06 +00:00
|
|
|
handle_cap_grant(mdsc, inode, h, &pool_ns,
|
2014-11-14 13:29:55 +00:00
|
|
|
inline_version, inline_data, inline_len,
|
2016-03-07 01:35:06 +00:00
|
|
|
msg->middle, session, cap, issued);
|
2014-12-23 07:30:54 +00:00
|
|
|
if (realm)
|
|
|
|
ceph_put_snap_realm(mdsc, realm);
|
2014-04-18 05:20:27 +00:00
|
|
|
goto done_unlocked;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/* the rest require a cap */
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_lock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
cap = __get_cap_for_mds(ceph_inode(inode), mds);
|
|
|
|
if (!cap) {
|
2010-06-10 20:21:20 +00:00
|
|
|
dout(" no cap on %p ino %llx.%llx from mds%d\n",
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
inode, ceph_ino(inode), ceph_snap(inode), mds);
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
2010-10-06 22:46:30 +00:00
|
|
|
goto flush_cap_releases;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
|
2011-11-30 17:47:09 +00:00
|
|
|
/* note that each of these drops i_ceph_lock for us */
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
switch (op) {
|
|
|
|
case CEPH_CAP_OP_REVOKE:
|
|
|
|
case CEPH_CAP_OP_GRANT:
|
2014-04-18 05:20:27 +00:00
|
|
|
__ceph_caps_issued(ci, &issued);
|
|
|
|
issued |= __ceph_caps_dirty(ci);
|
2016-03-07 01:35:06 +00:00
|
|
|
handle_cap_grant(mdsc, inode, h, &pool_ns,
|
2014-11-14 13:29:55 +00:00
|
|
|
inline_version, inline_data, inline_len,
|
2016-03-07 01:35:06 +00:00
|
|
|
msg->middle, session, cap, issued);
|
2010-03-16 20:42:00 +00:00
|
|
|
goto done_unlocked;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
case CEPH_CAP_OP_FLUSH_ACK:
|
2009-12-22 19:24:33 +00:00
|
|
|
handle_cap_flush_ack(inode, tid, h, session, cap);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
break;
|
|
|
|
|
|
|
|
case CEPH_CAP_OP_TRUNC:
|
|
|
|
handle_cap_trunc(inode, h, session);
|
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
pr_err("ceph_handle_caps: unknown cap op %d %s\n", op,
|
|
|
|
ceph_cap_op_name(op));
|
|
|
|
}
|
|
|
|
|
2010-10-06 22:46:30 +00:00
|
|
|
goto done;
|
|
|
|
|
|
|
|
flush_cap_releases:
|
|
|
|
/*
|
2015-05-14 09:22:42 +00:00
|
|
|
* send any cap release message to try to move things
|
2010-10-06 22:46:30 +00:00
|
|
|
* along for the mds (who clearly thinks we still have this
|
|
|
|
* cap).
|
|
|
|
*/
|
|
|
|
ceph_send_cap_releases(mdsc, session);
|
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
done:
|
2010-03-16 20:42:00 +00:00
|
|
|
mutex_unlock(&session->s_mutex);
|
|
|
|
done_unlocked:
|
2014-11-02 14:20:59 +00:00
|
|
|
iput(inode);
|
2016-03-07 01:35:06 +00:00
|
|
|
ceph_put_string(pool_ns);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
return;
|
|
|
|
|
|
|
|
bad:
|
|
|
|
pr_err("ceph_handle_caps: corrupt message\n");
|
2009-12-14 23:13:47 +00:00
|
|
|
ceph_msg_dump(msg);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Delayed work handler to process end of delayed cap release LRU list.
|
|
|
|
*/
|
2009-10-14 21:27:38 +00:00
|
|
|
void ceph_check_delayed_caps(struct ceph_mds_client *mdsc)
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
{
|
2017-06-27 09:17:24 +00:00
|
|
|
struct inode *inode;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
struct ceph_inode_info *ci;
|
|
|
|
int flags = CHECK_CAPS_NODELAY;
|
|
|
|
|
|
|
|
dout("check_delayed_caps\n");
|
|
|
|
while (1) {
|
|
|
|
spin_lock(&mdsc->cap_delay_lock);
|
|
|
|
if (list_empty(&mdsc->cap_delay_list))
|
|
|
|
break;
|
|
|
|
ci = list_first_entry(&mdsc->cap_delay_list,
|
|
|
|
struct ceph_inode_info,
|
|
|
|
i_cap_delay_list);
|
|
|
|
if ((ci->i_ceph_flags & CEPH_I_FLUSH) == 0 &&
|
|
|
|
time_before(jiffies, ci->i_hold_caps_max))
|
|
|
|
break;
|
|
|
|
list_del_init(&ci->i_cap_delay_list);
|
2017-06-27 09:17:24 +00:00
|
|
|
|
|
|
|
inode = igrab(&ci->vfs_inode);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
spin_unlock(&mdsc->cap_delay_lock);
|
2017-06-27 09:17:24 +00:00
|
|
|
|
|
|
|
if (inode) {
|
|
|
|
dout("check_delayed_caps on %p\n", inode);
|
|
|
|
ceph_check_caps(ci, flags, NULL);
|
|
|
|
iput(inode);
|
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
spin_unlock(&mdsc->cap_delay_lock);
|
|
|
|
}
|
|
|
|
|
2009-10-14 21:27:38 +00:00
|
|
|
/*
|
|
|
|
* Flush all dirty caps to the mds
|
|
|
|
*/
|
|
|
|
void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc)
|
|
|
|
{
|
2011-05-24 18:46:31 +00:00
|
|
|
struct ceph_inode_info *ci;
|
|
|
|
struct inode *inode;
|
2009-10-14 21:27:38 +00:00
|
|
|
|
|
|
|
dout("flush_dirty_caps\n");
|
|
|
|
spin_lock(&mdsc->cap_dirty_lock);
|
2011-05-24 18:46:31 +00:00
|
|
|
while (!list_empty(&mdsc->cap_dirty)) {
|
|
|
|
ci = list_first_entry(&mdsc->cap_dirty, struct ceph_inode_info,
|
|
|
|
i_dirty_item);
|
2011-05-27 16:24:26 +00:00
|
|
|
inode = &ci->vfs_inode;
|
|
|
|
ihold(inode);
|
2011-05-24 18:46:31 +00:00
|
|
|
dout("flush_dirty_caps %p\n", inode);
|
2009-10-14 21:27:38 +00:00
|
|
|
spin_unlock(&mdsc->cap_dirty_lock);
|
2011-05-27 16:24:26 +00:00
|
|
|
ceph_check_caps(ci, CHECK_CAPS_NODELAY|CHECK_CAPS_FLUSH, NULL);
|
|
|
|
iput(inode);
|
2009-10-14 21:27:38 +00:00
|
|
|
spin_lock(&mdsc->cap_dirty_lock);
|
|
|
|
}
|
|
|
|
spin_unlock(&mdsc->cap_dirty_lock);
|
2011-05-24 18:46:31 +00:00
|
|
|
dout("flush_dirty_caps done\n");
|
2009-10-14 21:27:38 +00:00
|
|
|
}
|
|
|
|
|
2016-06-06 08:01:39 +00:00
|
|
|
void __ceph_get_fmode(struct ceph_inode_info *ci, int fmode)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
int bits = (fmode << 1) | 1;
|
|
|
|
for (i = 0; i < CEPH_FILE_MODE_BITS; i++) {
|
|
|
|
if (bits & (1 << i))
|
|
|
|
ci->i_nr_by_mode[i]++;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
/*
|
|
|
|
* Drop open file reference. If we were the last open file,
|
|
|
|
* we may need to release capabilities to the MDS (or schedule
|
|
|
|
* their delayed release).
|
|
|
|
*/
|
|
|
|
void ceph_put_fmode(struct ceph_inode_info *ci, int fmode)
|
|
|
|
{
|
2016-06-06 08:01:39 +00:00
|
|
|
int i, last = 0;
|
|
|
|
int bits = (fmode << 1) | 1;
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_lock(&ci->i_ceph_lock);
|
2016-06-06 08:01:39 +00:00
|
|
|
for (i = 0; i < CEPH_FILE_MODE_BITS; i++) {
|
|
|
|
if (bits & (1 << i)) {
|
|
|
|
BUG_ON(ci->i_nr_by_mode[i] == 0);
|
|
|
|
if (--ci->i_nr_by_mode[i] == 0)
|
|
|
|
last++;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
dout("put_fmode %p fmode %d {%d,%d,%d,%d}\n",
|
|
|
|
&ci->vfs_inode, fmode,
|
|
|
|
ci->i_nr_by_mode[0], ci->i_nr_by_mode[1],
|
|
|
|
ci->i_nr_by_mode[2], ci->i_nr_by_mode[3]);
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
if (last && ci->i_vino.snap == CEPH_NOSNAP)
|
|
|
|
ceph_check_caps(ci, 0, NULL);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Helpers for embedding cap and dentry lease releases into mds
|
|
|
|
* requests.
|
|
|
|
*
|
|
|
|
* @force is used by dentry_release (below) to force inclusion of a
|
|
|
|
* record for the directory inode, even when there aren't any caps to
|
|
|
|
* drop.
|
|
|
|
*/
|
|
|
|
int ceph_encode_inode_release(void **p, struct inode *inode,
|
|
|
|
int mds, int drop, int unless, int force)
|
|
|
|
{
|
|
|
|
struct ceph_inode_info *ci = ceph_inode(inode);
|
|
|
|
struct ceph_cap *cap;
|
|
|
|
struct ceph_mds_request_release *rel = *p;
|
2010-06-24 22:12:37 +00:00
|
|
|
int used, dirty;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
int ret = 0;
|
|
|
|
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_lock(&ci->i_ceph_lock);
|
2010-03-16 22:01:07 +00:00
|
|
|
used = __ceph_caps_used(ci);
|
2010-06-24 22:12:37 +00:00
|
|
|
dirty = __ceph_caps_dirty(ci);
|
2010-03-16 22:01:07 +00:00
|
|
|
|
2010-06-24 22:12:37 +00:00
|
|
|
dout("encode_inode_release %p mds%d used|dirty %s drop %s unless %s\n",
|
|
|
|
inode, mds, ceph_cap_string(used|dirty), ceph_cap_string(drop),
|
2010-03-16 22:01:07 +00:00
|
|
|
ceph_cap_string(unless));
|
|
|
|
|
2010-06-24 22:12:37 +00:00
|
|
|
/* only drop unused, clean caps */
|
|
|
|
drop &= ~(used | dirty);
|
2010-03-16 22:01:07 +00:00
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
cap = __get_cap_for_mds(ci, mds);
|
|
|
|
if (cap && __cap_is_valid(cap)) {
|
|
|
|
if (force ||
|
|
|
|
((cap->issued & drop) &&
|
|
|
|
(cap->issued & unless) == 0)) {
|
|
|
|
if ((cap->issued & drop) &&
|
|
|
|
(cap->issued & unless) == 0) {
|
2013-06-03 10:22:17 +00:00
|
|
|
int wanted = __ceph_caps_wanted(ci);
|
|
|
|
if ((ci->i_ceph_flags & CEPH_I_NODELAY) == 0)
|
|
|
|
wanted |= cap->mds_wanted;
|
|
|
|
dout("encode_inode_release %p cap %p "
|
|
|
|
"%s -> %s, wanted %s -> %s\n", inode, cap,
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
ceph_cap_string(cap->issued),
|
2013-06-03 10:22:17 +00:00
|
|
|
ceph_cap_string(cap->issued & ~drop),
|
|
|
|
ceph_cap_string(cap->mds_wanted),
|
|
|
|
ceph_cap_string(wanted));
|
|
|
|
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
cap->issued &= ~drop;
|
|
|
|
cap->implemented &= ~drop;
|
2013-06-03 10:22:17 +00:00
|
|
|
cap->mds_wanted = wanted;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
} else {
|
|
|
|
dout("encode_inode_release %p cap %p %s"
|
|
|
|
" (force)\n", inode, cap,
|
|
|
|
ceph_cap_string(cap->issued));
|
|
|
|
}
|
|
|
|
|
|
|
|
rel->ino = cpu_to_le64(ceph_ino(inode));
|
|
|
|
rel->cap_id = cpu_to_le64(cap->cap_id);
|
|
|
|
rel->seq = cpu_to_le32(cap->seq);
|
2014-07-23 14:41:11 +00:00
|
|
|
rel->issue_seq = cpu_to_le32(cap->issue_seq);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
rel->mseq = cpu_to_le32(cap->mseq);
|
2014-04-17 00:02:02 +00:00
|
|
|
rel->caps = cpu_to_le32(cap->implemented);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
rel->wanted = cpu_to_le32(cap->mds_wanted);
|
|
|
|
rel->dname_len = 0;
|
|
|
|
rel->dname_seq = 0;
|
|
|
|
*p += sizeof(*rel);
|
|
|
|
ret = 1;
|
|
|
|
} else {
|
|
|
|
dout("encode_inode_release %p cap %p %s\n",
|
|
|
|
inode, cap, ceph_cap_string(cap->issued));
|
|
|
|
}
|
|
|
|
}
|
2011-11-30 17:47:09 +00:00
|
|
|
spin_unlock(&ci->i_ceph_lock);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
int ceph_encode_dentry_release(void **p, struct dentry *dentry,
|
2016-12-15 13:37:59 +00:00
|
|
|
struct inode *dir,
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
int mds, int drop, int unless)
|
|
|
|
{
|
2016-12-15 13:37:59 +00:00
|
|
|
struct dentry *parent = NULL;
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
struct ceph_mds_request_release *rel = *p;
|
|
|
|
struct ceph_dentry_info *di = ceph_dentry(dentry);
|
|
|
|
int force = 0;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* force an record for the directory caps if we have a dentry lease.
|
2011-11-30 17:47:09 +00:00
|
|
|
* this is racy (can't take i_ceph_lock and d_lock together), but it
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
* doesn't have to be perfect; the mds will revoke anything we don't
|
|
|
|
* release.
|
|
|
|
*/
|
|
|
|
spin_lock(&dentry->d_lock);
|
|
|
|
if (di->lease_session && di->lease_session->s_mds == mds)
|
|
|
|
force = 1;
|
2016-12-15 13:37:59 +00:00
|
|
|
if (!dir) {
|
|
|
|
parent = dget(dentry->d_parent);
|
|
|
|
dir = d_inode(parent);
|
|
|
|
}
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
spin_unlock(&dentry->d_lock);
|
|
|
|
|
2016-12-15 13:37:59 +00:00
|
|
|
ret = ceph_encode_inode_release(p, dir, mds, drop, unless, force);
|
2016-12-15 13:37:58 +00:00
|
|
|
dput(parent);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
|
|
|
|
spin_lock(&dentry->d_lock);
|
|
|
|
if (ret && di->lease_session && di->lease_session->s_mds == mds) {
|
|
|
|
dout("encode_dentry_release %p mds%d seq %d\n",
|
|
|
|
dentry, mds, (int)di->lease_seq);
|
|
|
|
rel->dname_len = cpu_to_le32(dentry->d_name.len);
|
|
|
|
memcpy(*p, dentry->d_name.name, dentry->d_name.len);
|
|
|
|
*p += dentry->d_name.len;
|
|
|
|
rel->dname_seq = cpu_to_le32(di->lease_seq);
|
2010-07-23 20:54:21 +00:00
|
|
|
__ceph_mdsc_drop_dentry_lease(dentry);
|
ceph: capability management
The Ceph metadata servers control client access to inode metadata and
file data by issuing capabilities, granting clients permission to read
and/or write both inode field and file data to OSDs (storage nodes).
Each capability consists of a set of bits indicating which operations
are allowed.
If the client holds a *_SHARED cap, the client has a coherent value
that can be safely read from the cached inode.
In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the client
is allowed to change inode attributes (e.g., file size, mtime), note
its dirty state in the ceph_cap, and asynchronously flush that
metadata change to the MDS.
In the event of a conflicting operation (perhaps by another client),
the MDS will revoke the conflicting client capabilities.
In order for a client to cache an inode, it must hold a capability
with at least one MDS server. When inodes are released, release
notifications are batched and periodically sent en masse to the MDS
cluster to release server state.
Signed-off-by: Sage Weil <sage@newdream.net>
2009-10-06 18:31:12 +00:00
|
|
|
}
|
|
|
|
spin_unlock(&dentry->d_lock);
|
|
|
|
return ret;
|
|
|
|
}
|