2006-01-16 16:50:04 +00:00
|
|
|
/*
|
|
|
|
* Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
|
2008-01-31 16:31:39 +00:00
|
|
|
* Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
|
2006-01-16 16:50:04 +00:00
|
|
|
*
|
|
|
|
* This copyrighted material is made available to anyone wishing to use,
|
|
|
|
* modify, copy, or redistribute it subject to the terms and conditions
|
2006-09-01 15:05:15 +00:00
|
|
|
* of the GNU General Public License version 2.
|
2006-01-16 16:50:04 +00:00
|
|
|
*/
|
|
|
|
|
2014-03-06 20:10:45 +00:00
|
|
|
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
|
|
|
|
|
2006-01-16 16:50:04 +00:00
|
|
|
#include <linux/sched.h>
|
|
|
|
#include <linux/slab.h>
|
|
|
|
#include <linux/spinlock.h>
|
|
|
|
#include <linux/completion.h>
|
|
|
|
#include <linux/buffer_head.h>
|
|
|
|
#include <linux/blkdev.h>
|
|
|
|
#include <linux/kthread.h>
|
2011-05-26 20:00:52 +00:00
|
|
|
#include <linux/export.h>
|
2006-08-25 16:13:37 +00:00
|
|
|
#include <linux/namei.h>
|
|
|
|
#include <linux/mount.h>
|
2006-02-27 22:23:27 +00:00
|
|
|
#include <linux/gfs2_ondisk.h>
|
2009-09-15 08:59:02 +00:00
|
|
|
#include <linux/quotaops.h>
|
2012-09-05 20:55:11 +00:00
|
|
|
#include <linux/lockdep.h>
|
2013-03-03 03:39:14 +00:00
|
|
|
#include <linux/module.h>
|
2006-01-16 16:50:04 +00:00
|
|
|
|
|
|
|
#include "gfs2.h"
|
2006-02-27 22:23:27 +00:00
|
|
|
#include "incore.h"
|
2007-12-12 00:49:21 +00:00
|
|
|
#include "bmap.h"
|
2006-01-16 16:50:04 +00:00
|
|
|
#include "glock.h"
|
|
|
|
#include "glops.h"
|
|
|
|
#include "inode.h"
|
|
|
|
#include "recovery.h"
|
|
|
|
#include "rgrp.h"
|
|
|
|
#include "super.h"
|
|
|
|
#include "sys.h"
|
2006-02-27 22:23:27 +00:00
|
|
|
#include "util.h"
|
2007-08-16 15:03:57 +00:00
|
|
|
#include "log.h"
|
2008-11-19 10:08:22 +00:00
|
|
|
#include "quota.h"
|
2008-11-26 12:49:26 +00:00
|
|
|
#include "dir.h"
|
2013-12-06 16:19:54 +00:00
|
|
|
#include "meta_io.h"
|
2009-06-12 07:49:20 +00:00
|
|
|
#include "trace_gfs2.h"
|
2006-01-16 16:50:04 +00:00
|
|
|
|
|
|
|
#define DO 0
|
|
|
|
#define UNDO 1
|
|
|
|
|
2008-08-08 12:45:13 +00:00
|
|
|
/**
|
|
|
|
* gfs2_tune_init - Fill a gfs2_tune structure with default values
|
|
|
|
* @gt: tune
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
|
|
|
|
static void gfs2_tune_init(struct gfs2_tune *gt)
|
|
|
|
{
|
|
|
|
spin_lock_init(>->gt_spin);
|
|
|
|
|
|
|
|
gt->gt_quota_warn_period = 10;
|
|
|
|
gt->gt_quota_scale_num = 1;
|
|
|
|
gt->gt_quota_scale_den = 1;
|
|
|
|
gt->gt_new_files_jdata = 0;
|
|
|
|
gt->gt_max_readahead = 1 << 18;
|
|
|
|
gt->gt_complain_secs = 10;
|
|
|
|
}
|
|
|
|
|
2006-01-16 16:50:04 +00:00
|
|
|
static struct gfs2_sbd *init_sbd(struct super_block *sb)
|
|
|
|
{
|
|
|
|
struct gfs2_sbd *sdp;
|
2013-12-06 16:19:54 +00:00
|
|
|
struct address_space *mapping;
|
2006-01-16 16:50:04 +00:00
|
|
|
|
2006-09-07 18:40:21 +00:00
|
|
|
sdp = kzalloc(sizeof(struct gfs2_sbd), GFP_KERNEL);
|
2006-01-16 16:50:04 +00:00
|
|
|
if (!sdp)
|
|
|
|
return NULL;
|
|
|
|
|
2006-02-27 22:23:27 +00:00
|
|
|
sb->s_fs_info = sdp;
|
2006-01-16 16:50:04 +00:00
|
|
|
sdp->sd_vfs = sb;
|
GFS2: glock statistics gathering
The stats are divided into two sets: those relating to the
super block and those relating to an individual glock. The
super block stats are done on a per cpu basis in order to
try and reduce the overhead of gathering them. They are also
further divided by glock type.
In the case of both the super block and glock statistics,
the same information is gathered in each case. The super
block statistics are used to provide default values for
most of the glock statistics, so that newly created glocks
should have, as far as possible, a sensible starting point.
The statistics are divided into three pairs of mean and
variance, plus two counters. The mean/variance pairs are
smoothed exponential estimates and the algorithm used is
one which will be very familiar to those used to calculation
of round trip times in network code.
The three pairs of mean/variance measure the following
things:
1. DLM lock time (non-blocking requests)
2. DLM lock time (blocking requests)
3. Inter-request time (again to the DLM)
A non-blocking request is one which will complete right
away, whatever the state of the DLM lock in question. That
currently means any requests when (a) the current state of
the lock is exclusive (b) the requested state is either null
or unlocked or (c) the "try lock" flag is set. A blocking
request covers all the other lock requests.
There are two counters. The first is there primarily to show
how many lock requests have been made, and thus how much data
has gone into the mean/variance calculations. The other counter
is counting queueing of holders at the top layer of the glock
code. Hopefully that number will be a lot larger than the number
of dlm lock requests issued.
So why gather these statistics? There are several reasons
we'd like to get a better idea of these timings:
1. To be able to better set the glock "min hold time"
2. To spot performance issues more easily
3. To improve the algorithm for selecting resource groups for
allocation (to base it on lock wait time, rather than blindly
using a "try lock")
Due to the smoothing action of the updates, a step change in
some input quantity being sampled will only fully be taken
into account after 8 samples (or 4 for the variance) and this
needs to be carefully considered when interpreting the
results.
Knowing both the time it takes a lock request to complete and
the average time between lock requests for a glock means we
can compute the total percentage of the time for which the
node is able to use a glock vs. time that the rest of the
cluster has its share. That will be very useful when setting
the lock min hold time.
The other point to remember is that all times are in
nanoseconds. Great care has been taken to ensure that we
measure exactly the quantities that we want, as accurately
as possible. There are always inaccuracies in any
measuring system, but I hope this is as accurate as we
can reasonably make it.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2012-01-20 10:38:36 +00:00
|
|
|
sdp->sd_lkstats = alloc_percpu(struct gfs2_pcpu_lkstats);
|
|
|
|
if (!sdp->sd_lkstats) {
|
|
|
|
kfree(sdp);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2010-06-14 09:01:30 +00:00
|
|
|
set_bit(SDF_NOJOURNALID, &sdp->sd_flags);
|
2006-01-16 16:50:04 +00:00
|
|
|
gfs2_tune_init(&sdp->sd_tune);
|
|
|
|
|
2010-01-25 11:20:19 +00:00
|
|
|
init_waitqueue_head(&sdp->sd_glock_wait);
|
|
|
|
atomic_set(&sdp->sd_glock_disposal, 0);
|
2011-07-11 07:53:30 +00:00
|
|
|
init_completion(&sdp->sd_locking_init);
|
2013-02-13 12:21:40 +00:00
|
|
|
init_completion(&sdp->sd_wdack);
|
2006-01-16 16:50:04 +00:00
|
|
|
spin_lock_init(&sdp->sd_statfs_spin);
|
|
|
|
|
|
|
|
spin_lock_init(&sdp->sd_rindex_spin);
|
GFS2: Use rbtree for resource groups and clean up bitmap buffer ref count scheme
Here is an update of Bob's original rbtree patch which, in addition, also
resolves the rather strange ref counting that was being done relating to
the bitmap blocks.
Originally we had a dual system for journaling resource groups. The metadata
blocks were journaled and also the rgrp itself was added to a list. The reason
for adding the rgrp to the list in the journal was so that the "repolish
clones" code could be run to update the free space, and potentially send any
discard requests when the log was flushed. This was done by comparing the
"cloned" bitmap with what had been written back on disk during the transaction
commit.
Due to this, there was a requirement to hang on to the rgrps' bitmap buffers
until the journal had been flushed. For that reason, there was a rather
complicated set up in the ->go_lock ->go_unlock functions for rgrps involving
both a mutex and a spinlock (the ->sd_rindex_spin) to maintain a reference
count on the buffers.
However, the journal maintains a reference count on the buffers anyway, since
they are being journaled as metadata buffers. So by moving the code which deals
with the post-journal accounting for bitmap blocks to the metadata journaling
code, we can entirely dispense with the rather strange buffer ref counting
scheme and also the requirement to journal the rgrps.
The net result of all this is that the ->sd_rindex_spin is left to do exactly
one job, and that is to look after the rbtree or rgrps.
This patch is designed to be a stepping stone towards using RCU for the rbtree
of resource groups, however the reduction in the number of uses of the
->sd_rindex_spin is likely to have benefits for multi-threaded workloads,
anyway.
The patch retains ->go_lock and ->go_unlock for rgrps, however these maybe also
be removed in future in favour of calling the functions directly where required
in the code. That will allow locking of resource groups without needing to
actually read them in - something that could be useful in speeding up statfs.
In the mean time though it is valid to dereference ->bi_bh only when the rgrp
is locked. This is basically the same rule as before, modulo the references not
being valid until the following journal flush.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Cc: Benjamin Marzinski <bmarzins@redhat.com>
2011-08-31 08:53:19 +00:00
|
|
|
sdp->sd_rindex_tree.rb_node = NULL;
|
2006-01-16 16:50:04 +00:00
|
|
|
|
|
|
|
INIT_LIST_HEAD(&sdp->sd_jindex_list);
|
|
|
|
spin_lock_init(&sdp->sd_jindex_spin);
|
2006-02-21 12:51:39 +00:00
|
|
|
mutex_init(&sdp->sd_jindex_mutex);
|
2014-06-02 13:40:25 +00:00
|
|
|
init_completion(&sdp->sd_journal_ready);
|
2006-01-16 16:50:04 +00:00
|
|
|
|
|
|
|
INIT_LIST_HEAD(&sdp->sd_quota_list);
|
2006-02-21 12:51:39 +00:00
|
|
|
mutex_init(&sdp->sd_quota_mutex);
|
2013-10-04 11:29:34 +00:00
|
|
|
mutex_init(&sdp->sd_quota_sync_mutex);
|
2008-11-17 14:25:37 +00:00
|
|
|
init_waitqueue_head(&sdp->sd_quota_wait);
|
2008-11-18 13:38:48 +00:00
|
|
|
INIT_LIST_HEAD(&sdp->sd_trunc_list);
|
|
|
|
spin_lock_init(&sdp->sd_trunc_lock);
|
2013-12-13 11:46:28 +00:00
|
|
|
spin_lock_init(&sdp->sd_bitmap_lock);
|
2006-01-16 16:50:04 +00:00
|
|
|
|
2013-12-06 16:19:54 +00:00
|
|
|
mapping = &sdp->sd_aspace;
|
|
|
|
|
2014-01-09 16:34:04 +00:00
|
|
|
address_space_init_once(mapping);
|
2014-03-31 16:48:27 +00:00
|
|
|
mapping->a_ops = &gfs2_rgrp_aops;
|
2013-12-06 16:19:54 +00:00
|
|
|
mapping->host = sb->s_bdev->bd_inode;
|
|
|
|
mapping->flags = 0;
|
|
|
|
mapping_set_gfp_mask(mapping, GFP_NOFS);
|
|
|
|
mapping->private_data = NULL;
|
|
|
|
mapping->backing_dev_info = sb->s_bdi;
|
|
|
|
mapping->writeback_index = 0;
|
|
|
|
|
2006-01-16 16:50:04 +00:00
|
|
|
spin_lock_init(&sdp->sd_log_lock);
|
GFS2: Various gfs2_logd improvements
This patch contains various tweaks to how log flushes and active item writeback
work. gfs2_logd is now managed by a waitqueue, and gfs2_log_reseve now waits
for gfs2_logd to do the log flushing. Multiple functions were rewritten to
remove the need to call gfs2_log_lock(). Instead of using one test to see if
gfs2_logd had work to do, there are now seperate tests to check if there
are two many buffers in the incore log or if there are two many items on the
active items list.
This patch is a port of a patch Steve Whitehouse wrote about a year ago, with
some minor changes. Since gfs2_ail1_start always submits all the active items,
it no longer needs to keep track of the first ai submitted, so this has been
removed. In gfs2_log_reserve(), the order of the calls to
prepare_to_wait_exclusive() and wake_up() when firing off the logd thread has
been switched. If it called wake_up first there was a small window for a race,
where logd could run and return before gfs2_log_reserve was ready to get woken
up. If gfs2_logd ran, but did not free up enough blocks, gfs2_log_reserve()
would be left waiting for gfs2_logd to eventualy run because it timed out.
Finally, gt_logd_secs, which controls how long to wait before gfs2_logd times
out, and flushes the log, can now be set on mount with ar_commit.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2010-05-04 19:29:16 +00:00
|
|
|
atomic_set(&sdp->sd_log_pinned, 0);
|
2006-01-16 16:50:04 +00:00
|
|
|
INIT_LIST_HEAD(&sdp->sd_log_le_revoke);
|
2007-09-02 09:48:13 +00:00
|
|
|
INIT_LIST_HEAD(&sdp->sd_log_le_ordered);
|
2013-01-28 09:30:07 +00:00
|
|
|
spin_lock_init(&sdp->sd_ordered_lock);
|
2006-01-16 16:50:04 +00:00
|
|
|
|
GFS2: Various gfs2_logd improvements
This patch contains various tweaks to how log flushes and active item writeback
work. gfs2_logd is now managed by a waitqueue, and gfs2_log_reseve now waits
for gfs2_logd to do the log flushing. Multiple functions were rewritten to
remove the need to call gfs2_log_lock(). Instead of using one test to see if
gfs2_logd had work to do, there are now seperate tests to check if there
are two many buffers in the incore log or if there are two many items on the
active items list.
This patch is a port of a patch Steve Whitehouse wrote about a year ago, with
some minor changes. Since gfs2_ail1_start always submits all the active items,
it no longer needs to keep track of the first ai submitted, so this has been
removed. In gfs2_log_reserve(), the order of the calls to
prepare_to_wait_exclusive() and wake_up() when firing off the logd thread has
been switched. If it called wake_up first there was a small window for a race,
where logd could run and return before gfs2_log_reserve was ready to get woken
up. If gfs2_logd ran, but did not free up enough blocks, gfs2_log_reserve()
would be left waiting for gfs2_logd to eventualy run because it timed out.
Finally, gt_logd_secs, which controls how long to wait before gfs2_logd times
out, and flushes the log, can now be set on mount with ar_commit.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2010-05-04 19:29:16 +00:00
|
|
|
init_waitqueue_head(&sdp->sd_log_waitq);
|
|
|
|
init_waitqueue_head(&sdp->sd_logd_waitq);
|
2011-03-11 11:52:25 +00:00
|
|
|
spin_lock_init(&sdp->sd_ail_lock);
|
2006-01-16 16:50:04 +00:00
|
|
|
INIT_LIST_HEAD(&sdp->sd_ail1_list);
|
|
|
|
INIT_LIST_HEAD(&sdp->sd_ail2_list);
|
|
|
|
|
2006-03-29 14:12:12 +00:00
|
|
|
init_rwsem(&sdp->sd_log_flush_lock);
|
2007-09-17 09:59:52 +00:00
|
|
|
atomic_set(&sdp->sd_log_in_flight, 0);
|
|
|
|
init_waitqueue_head(&sdp->sd_log_flush_wait);
|
GFS2: remove transaction glock
GFS2 has a transaction glock, which must be grabbed for every
transaction, whose purpose is to deal with freezing the filesystem.
Aside from this involving a large amount of locking, it is very easy to
make the current fsfreeze code hang on unfreezing.
This patch rewrites how gfs2 handles freezing the filesystem. The
transaction glock is removed. In it's place is a freeze glock, which is
cached (but not held) in a shared state by every node in the cluster
when the filesystem is mounted. This lock only needs to be grabbed on
freezing, and actions which need to be safe from freezing, like
recovery.
When a node wants to freeze the filesystem, it grabs this glock
exclusively. When the freeze glock state changes on the nodes (either
from shared to unlocked, or shared to exclusive), the filesystem does a
special log flush. gfs2_log_flush() does all the work for flushing out
the and shutting down the incore log, and then it tries to grab the
freeze glock in a shared state again. Since the filesystem is stuck in
gfs2_log_flush, no new transaction can start, and nothing can be written
to disk. Unfreezing the filesytem simply involes dropping the freeze
glock, allowing gfs2_log_flush() to grab and then release the shared
lock, so it is cached for next time.
However, in order for the unfreezing ioctl to occur, gfs2 needs to get a
shared lock on the filesystem root directory inode to check permissions.
If that glock has already been grabbed exclusively, fsfreeze will be
unable to get the shared lock and unfreeze the filesystem.
In order to allow the unfreeze, this patch makes gfs2 grab a shared lock
on the filesystem root directory during the freeze, and hold it until it
unfreezes the filesystem. The functions which need to grab a shared
lock in order to allow the unfreeze ioctl to be issued now use the lock
grabbed by the freeze code instead.
The freeze and unfreeze code take care to make sure that this shared
lock will not be dropped while another process is using it.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2014-05-02 03:26:55 +00:00
|
|
|
init_waitqueue_head(&sdp->sd_log_frozen_wait);
|
|
|
|
atomic_set(&sdp->sd_log_freeze, 0);
|
|
|
|
atomic_set(&sdp->sd_frozen_root, 0);
|
|
|
|
init_waitqueue_head(&sdp->sd_frozen_root_wait);
|
2006-01-16 16:50:04 +00:00
|
|
|
|
|
|
|
return sdp;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2008-08-08 12:45:13 +00:00
|
|
|
/**
|
|
|
|
* gfs2_check_sb - Check superblock
|
|
|
|
* @sdp: the filesystem
|
|
|
|
* @sb: The superblock
|
|
|
|
* @silent: Don't print a message if the check fails
|
|
|
|
*
|
|
|
|
* Checks the version code of the FS is one that we understand how to
|
|
|
|
* read and that the sizes of the various on-disk structures have not
|
|
|
|
* changed.
|
|
|
|
*/
|
|
|
|
|
2011-05-10 14:01:59 +00:00
|
|
|
static int gfs2_check_sb(struct gfs2_sbd *sdp, int silent)
|
2008-08-08 12:45:13 +00:00
|
|
|
{
|
2011-05-10 14:01:59 +00:00
|
|
|
struct gfs2_sb_host *sb = &sdp->sd_sb;
|
|
|
|
|
2008-08-08 12:45:13 +00:00
|
|
|
if (sb->sb_magic != GFS2_MAGIC ||
|
|
|
|
sb->sb_type != GFS2_METATYPE_SB) {
|
|
|
|
if (!silent)
|
2014-03-06 20:10:45 +00:00
|
|
|
pr_warn("not a GFS2 filesystem\n");
|
2008-08-08 12:45:13 +00:00
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* If format numbers match exactly, we're done. */
|
|
|
|
|
|
|
|
if (sb->sb_fs_format == GFS2_FORMAT_FS &&
|
|
|
|
sb->sb_multihost_format == GFS2_FORMAT_MULTI)
|
|
|
|
return 0;
|
|
|
|
|
2010-09-24 08:55:07 +00:00
|
|
|
fs_warn(sdp, "Unknown on-disk format, unable to mount\n");
|
2008-08-08 12:45:13 +00:00
|
|
|
|
2010-09-24 08:55:07 +00:00
|
|
|
return -EINVAL;
|
2008-08-08 12:45:13 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static void end_bio_io_page(struct bio *bio, int error)
|
|
|
|
{
|
|
|
|
struct page *page = bio->bi_private;
|
|
|
|
|
|
|
|
if (!error)
|
|
|
|
SetPageUptodate(page);
|
|
|
|
else
|
2014-03-06 20:10:45 +00:00
|
|
|
pr_warn("error %d reading superblock\n", error);
|
2008-08-08 12:45:13 +00:00
|
|
|
unlock_page(page);
|
|
|
|
}
|
|
|
|
|
2011-05-10 14:01:59 +00:00
|
|
|
static void gfs2_sb_in(struct gfs2_sbd *sdp, const void *buf)
|
2008-08-08 12:45:13 +00:00
|
|
|
{
|
2011-05-10 14:01:59 +00:00
|
|
|
struct gfs2_sb_host *sb = &sdp->sd_sb;
|
|
|
|
struct super_block *s = sdp->sd_vfs;
|
2008-08-08 12:45:13 +00:00
|
|
|
const struct gfs2_sb *str = buf;
|
|
|
|
|
|
|
|
sb->sb_magic = be32_to_cpu(str->sb_header.mh_magic);
|
|
|
|
sb->sb_type = be32_to_cpu(str->sb_header.mh_type);
|
|
|
|
sb->sb_format = be32_to_cpu(str->sb_header.mh_format);
|
|
|
|
sb->sb_fs_format = be32_to_cpu(str->sb_fs_format);
|
|
|
|
sb->sb_multihost_format = be32_to_cpu(str->sb_multihost_format);
|
|
|
|
sb->sb_bsize = be32_to_cpu(str->sb_bsize);
|
|
|
|
sb->sb_bsize_shift = be32_to_cpu(str->sb_bsize_shift);
|
|
|
|
sb->sb_master_dir.no_addr = be64_to_cpu(str->sb_master_dir.no_addr);
|
|
|
|
sb->sb_master_dir.no_formal_ino = be64_to_cpu(str->sb_master_dir.no_formal_ino);
|
|
|
|
sb->sb_root_dir.no_addr = be64_to_cpu(str->sb_root_dir.no_addr);
|
|
|
|
sb->sb_root_dir.no_formal_ino = be64_to_cpu(str->sb_root_dir.no_formal_ino);
|
|
|
|
|
|
|
|
memcpy(sb->sb_lockproto, str->sb_lockproto, GFS2_LOCKNAME_LEN);
|
|
|
|
memcpy(sb->sb_locktable, str->sb_locktable, GFS2_LOCKNAME_LEN);
|
2011-05-10 14:01:59 +00:00
|
|
|
memcpy(s->s_uuid, str->sb_uuid, 16);
|
2008-08-08 12:45:13 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* gfs2_read_super - Read the gfs2 super block from disk
|
|
|
|
* @sdp: The GFS2 super block
|
|
|
|
* @sector: The location of the super block
|
|
|
|
* @error: The error code to return
|
|
|
|
*
|
|
|
|
* This uses the bio functions to read the super block from disk
|
|
|
|
* because we want to be 100% sure that we never read cached data.
|
|
|
|
* A super block is read twice only during each GFS2 mount and is
|
|
|
|
* never written to by the filesystem. The first time its read no
|
|
|
|
* locks are held, and the only details which are looked at are those
|
|
|
|
* relating to the locking protocol. Once locking is up and working,
|
|
|
|
* the sb is read again under the lock to establish the location of
|
|
|
|
* the master directory (contains pointers to journals etc) and the
|
|
|
|
* root directory.
|
|
|
|
*
|
|
|
|
* Returns: 0 on success or error
|
|
|
|
*/
|
|
|
|
|
2011-05-10 14:01:59 +00:00
|
|
|
static int gfs2_read_super(struct gfs2_sbd *sdp, sector_t sector, int silent)
|
2008-08-08 12:45:13 +00:00
|
|
|
{
|
|
|
|
struct super_block *sb = sdp->sd_vfs;
|
|
|
|
struct gfs2_sb *p;
|
|
|
|
struct page *page;
|
|
|
|
struct bio *bio;
|
|
|
|
|
|
|
|
page = alloc_page(GFP_NOFS);
|
|
|
|
if (unlikely(!page))
|
2014-01-16 10:31:13 +00:00
|
|
|
return -ENOMEM;
|
2008-08-08 12:45:13 +00:00
|
|
|
|
|
|
|
ClearPageUptodate(page);
|
|
|
|
ClearPageDirty(page);
|
|
|
|
lock_page(page);
|
|
|
|
|
|
|
|
bio = bio_alloc(GFP_NOFS, 1);
|
2013-10-11 22:44:27 +00:00
|
|
|
bio->bi_iter.bi_sector = sector * (sb->s_blocksize >> 9);
|
2008-08-08 12:45:13 +00:00
|
|
|
bio->bi_bdev = sb->s_bdev;
|
|
|
|
bio_add_page(bio, page, PAGE_SIZE, 0);
|
|
|
|
|
|
|
|
bio->bi_end_io = end_bio_io_page;
|
|
|
|
bio->bi_private = page;
|
2011-10-31 09:52:02 +00:00
|
|
|
submit_bio(READ_SYNC | REQ_META, bio);
|
2008-08-08 12:45:13 +00:00
|
|
|
wait_on_page_locked(page);
|
|
|
|
bio_put(bio);
|
|
|
|
if (!PageUptodate(page)) {
|
|
|
|
__free_page(page);
|
|
|
|
return -EIO;
|
|
|
|
}
|
|
|
|
p = kmap(page);
|
2011-05-10 14:01:59 +00:00
|
|
|
gfs2_sb_in(sdp, p);
|
2008-08-08 12:45:13 +00:00
|
|
|
kunmap(page);
|
|
|
|
__free_page(page);
|
2011-05-10 14:01:59 +00:00
|
|
|
return gfs2_check_sb(sdp, silent);
|
2008-08-08 12:45:13 +00:00
|
|
|
}
|
2009-03-09 09:03:51 +00:00
|
|
|
|
2008-08-08 12:45:13 +00:00
|
|
|
/**
|
|
|
|
* gfs2_read_sb - Read super block
|
|
|
|
* @sdp: The GFS2 superblock
|
|
|
|
* @silent: Don't print message if mount fails
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
|
2009-03-09 09:03:51 +00:00
|
|
|
static int gfs2_read_sb(struct gfs2_sbd *sdp, int silent)
|
2008-08-08 12:45:13 +00:00
|
|
|
{
|
|
|
|
u32 hash_blocks, ind_blocks, leaf_blocks;
|
|
|
|
u32 tmp_blocks;
|
|
|
|
unsigned int x;
|
|
|
|
int error;
|
|
|
|
|
2011-05-10 14:01:59 +00:00
|
|
|
error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift, silent);
|
2008-08-08 12:45:13 +00:00
|
|
|
if (error) {
|
|
|
|
if (!silent)
|
|
|
|
fs_err(sdp, "can't read superblock\n");
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
|
|
|
|
GFS2_BASIC_BLOCK_SHIFT;
|
|
|
|
sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
|
|
|
|
sdp->sd_diptrs = (sdp->sd_sb.sb_bsize -
|
|
|
|
sizeof(struct gfs2_dinode)) / sizeof(u64);
|
|
|
|
sdp->sd_inptrs = (sdp->sd_sb.sb_bsize -
|
|
|
|
sizeof(struct gfs2_meta_header)) / sizeof(u64);
|
|
|
|
sdp->sd_jbsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header);
|
|
|
|
sdp->sd_hash_bsize = sdp->sd_sb.sb_bsize / 2;
|
|
|
|
sdp->sd_hash_bsize_shift = sdp->sd_sb.sb_bsize_shift - 1;
|
|
|
|
sdp->sd_hash_ptrs = sdp->sd_hash_bsize / sizeof(u64);
|
|
|
|
sdp->sd_qc_per_block = (sdp->sd_sb.sb_bsize -
|
|
|
|
sizeof(struct gfs2_meta_header)) /
|
|
|
|
sizeof(struct gfs2_quota_change);
|
2012-10-19 12:32:51 +00:00
|
|
|
sdp->sd_blocks_per_bitmap = (sdp->sd_sb.sb_bsize -
|
|
|
|
sizeof(struct gfs2_meta_header))
|
|
|
|
* GFS2_NBBY; /* not the rgrp bitmap, subsequent bitmaps only */
|
2008-08-08 12:45:13 +00:00
|
|
|
|
|
|
|
/* Compute maximum reservation required to add a entry to a directory */
|
|
|
|
|
|
|
|
hash_blocks = DIV_ROUND_UP(sizeof(u64) * (1 << GFS2_DIR_MAX_DEPTH),
|
|
|
|
sdp->sd_jbsize);
|
|
|
|
|
|
|
|
ind_blocks = 0;
|
|
|
|
for (tmp_blocks = hash_blocks; tmp_blocks > sdp->sd_diptrs;) {
|
|
|
|
tmp_blocks = DIV_ROUND_UP(tmp_blocks, sdp->sd_inptrs);
|
|
|
|
ind_blocks += tmp_blocks;
|
|
|
|
}
|
|
|
|
|
|
|
|
leaf_blocks = 2 + GFS2_DIR_MAX_DEPTH;
|
|
|
|
|
|
|
|
sdp->sd_max_dirres = hash_blocks + ind_blocks + leaf_blocks;
|
|
|
|
|
|
|
|
sdp->sd_heightsize[0] = sdp->sd_sb.sb_bsize -
|
|
|
|
sizeof(struct gfs2_dinode);
|
|
|
|
sdp->sd_heightsize[1] = sdp->sd_sb.sb_bsize * sdp->sd_diptrs;
|
|
|
|
for (x = 2;; x++) {
|
|
|
|
u64 space, d;
|
|
|
|
u32 m;
|
|
|
|
|
|
|
|
space = sdp->sd_heightsize[x - 1] * sdp->sd_inptrs;
|
|
|
|
d = space;
|
|
|
|
m = do_div(d, sdp->sd_inptrs);
|
|
|
|
|
|
|
|
if (d != sdp->sd_heightsize[x - 1] || m)
|
|
|
|
break;
|
|
|
|
sdp->sd_heightsize[x] = space;
|
|
|
|
}
|
|
|
|
sdp->sd_max_height = x;
|
|
|
|
sdp->sd_heightsize[x] = ~0;
|
|
|
|
gfs2_assert(sdp, sdp->sd_max_height <= GFS2_MAX_META_HEIGHT);
|
|
|
|
|
|
|
|
sdp->sd_jheightsize[0] = sdp->sd_sb.sb_bsize -
|
|
|
|
sizeof(struct gfs2_dinode);
|
|
|
|
sdp->sd_jheightsize[1] = sdp->sd_jbsize * sdp->sd_diptrs;
|
|
|
|
for (x = 2;; x++) {
|
|
|
|
u64 space, d;
|
|
|
|
u32 m;
|
|
|
|
|
|
|
|
space = sdp->sd_jheightsize[x - 1] * sdp->sd_inptrs;
|
|
|
|
d = space;
|
|
|
|
m = do_div(d, sdp->sd_inptrs);
|
|
|
|
|
|
|
|
if (d != sdp->sd_jheightsize[x - 1] || m)
|
|
|
|
break;
|
|
|
|
sdp->sd_jheightsize[x] = space;
|
|
|
|
}
|
|
|
|
sdp->sd_max_jheight = x;
|
|
|
|
sdp->sd_jheightsize[x] = ~0;
|
|
|
|
gfs2_assert(sdp, sdp->sd_max_jheight <= GFS2_MAX_META_HEIGHT);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2006-01-16 16:50:04 +00:00
|
|
|
static int init_names(struct gfs2_sbd *sdp, int silent)
|
|
|
|
{
|
|
|
|
char *proto, *table;
|
|
|
|
int error = 0;
|
|
|
|
|
|
|
|
proto = sdp->sd_args.ar_lockproto;
|
|
|
|
table = sdp->sd_args.ar_locktable;
|
|
|
|
|
|
|
|
/* Try to autodetect */
|
|
|
|
|
|
|
|
if (!proto[0] || !table[0]) {
|
2011-05-10 14:01:59 +00:00
|
|
|
error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift, silent);
|
2007-06-01 13:11:58 +00:00
|
|
|
if (error)
|
|
|
|
return error;
|
2006-10-02 15:49:41 +00:00
|
|
|
|
2006-01-16 16:50:04 +00:00
|
|
|
if (!proto[0])
|
2006-10-02 15:49:41 +00:00
|
|
|
proto = sdp->sd_sb.sb_lockproto;
|
2006-01-16 16:50:04 +00:00
|
|
|
if (!table[0])
|
2006-10-02 15:49:41 +00:00
|
|
|
table = sdp->sd_sb.sb_locktable;
|
2006-01-16 16:50:04 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
if (!table[0])
|
|
|
|
table = sdp->sd_vfs->s_id;
|
|
|
|
|
2008-05-09 15:59:51 +00:00
|
|
|
strlcpy(sdp->sd_proto_name, proto, GFS2_FSNAME_LEN);
|
|
|
|
strlcpy(sdp->sd_table_name, table, GFS2_FSNAME_LEN);
|
2006-01-16 16:50:04 +00:00
|
|
|
|
2007-08-13 03:01:58 +00:00
|
|
|
table = sdp->sd_table_name;
|
|
|
|
while ((table = strchr(table, '/')))
|
2007-06-07 14:10:01 +00:00
|
|
|
*table = '_';
|
|
|
|
|
2006-01-16 16:50:04 +00:00
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
|
|
|
|
int undo)
|
|
|
|
{
|
|
|
|
int error = 0;
|
|
|
|
|
|
|
|
if (undo)
|
|
|
|
goto fail_trans;
|
|
|
|
|
|
|
|
error = gfs2_glock_nq_num(sdp,
|
|
|
|
GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
|
|
|
|
LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
|
|
|
|
mount_gh);
|
|
|
|
if (error) {
|
|
|
|
fs_err(sdp, "can't acquire mount glock: %d\n", error);
|
|
|
|
goto fail;
|
|
|
|
}
|
|
|
|
|
|
|
|
error = gfs2_glock_nq_num(sdp,
|
|
|
|
GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
|
|
|
|
LM_ST_SHARED,
|
2006-04-26 18:58:26 +00:00
|
|
|
LM_FLAG_NOEXP | GL_EXACT,
|
2006-01-16 16:50:04 +00:00
|
|
|
&sdp->sd_live_gh);
|
|
|
|
if (error) {
|
|
|
|
fs_err(sdp, "can't acquire live glock: %d\n", error);
|
|
|
|
goto fail_mount;
|
|
|
|
}
|
|
|
|
|
|
|
|
error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
|
|
|
|
CREATE, &sdp->sd_rename_gl);
|
|
|
|
if (error) {
|
|
|
|
fs_err(sdp, "can't create rename glock: %d\n", error);
|
|
|
|
goto fail_live;
|
|
|
|
}
|
|
|
|
|
GFS2: remove transaction glock
GFS2 has a transaction glock, which must be grabbed for every
transaction, whose purpose is to deal with freezing the filesystem.
Aside from this involving a large amount of locking, it is very easy to
make the current fsfreeze code hang on unfreezing.
This patch rewrites how gfs2 handles freezing the filesystem. The
transaction glock is removed. In it's place is a freeze glock, which is
cached (but not held) in a shared state by every node in the cluster
when the filesystem is mounted. This lock only needs to be grabbed on
freezing, and actions which need to be safe from freezing, like
recovery.
When a node wants to freeze the filesystem, it grabs this glock
exclusively. When the freeze glock state changes on the nodes (either
from shared to unlocked, or shared to exclusive), the filesystem does a
special log flush. gfs2_log_flush() does all the work for flushing out
the and shutting down the incore log, and then it tries to grab the
freeze glock in a shared state again. Since the filesystem is stuck in
gfs2_log_flush, no new transaction can start, and nothing can be written
to disk. Unfreezing the filesytem simply involes dropping the freeze
glock, allowing gfs2_log_flush() to grab and then release the shared
lock, so it is cached for next time.
However, in order for the unfreezing ioctl to occur, gfs2 needs to get a
shared lock on the filesystem root directory inode to check permissions.
If that glock has already been grabbed exclusively, fsfreeze will be
unable to get the shared lock and unfreeze the filesystem.
In order to allow the unfreeze, this patch makes gfs2 grab a shared lock
on the filesystem root directory during the freeze, and hold it until it
unfreezes the filesystem. The functions which need to grab a shared
lock in order to allow the unfreeze ioctl to be issued now use the lock
grabbed by the freeze code instead.
The freeze and unfreeze code take care to make sure that this shared
lock will not be dropped while another process is using it.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2014-05-02 03:26:55 +00:00
|
|
|
error = gfs2_glock_get(sdp, GFS2_FREEZE_LOCK, &gfs2_freeze_glops,
|
|
|
|
CREATE, &sdp->sd_freeze_gl);
|
2006-01-16 16:50:04 +00:00
|
|
|
if (error) {
|
|
|
|
fs_err(sdp, "can't create transaction glock: %d\n", error);
|
|
|
|
goto fail_rename;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
2006-06-14 19:32:57 +00:00
|
|
|
fail_trans:
|
GFS2: remove transaction glock
GFS2 has a transaction glock, which must be grabbed for every
transaction, whose purpose is to deal with freezing the filesystem.
Aside from this involving a large amount of locking, it is very easy to
make the current fsfreeze code hang on unfreezing.
This patch rewrites how gfs2 handles freezing the filesystem. The
transaction glock is removed. In it's place is a freeze glock, which is
cached (but not held) in a shared state by every node in the cluster
when the filesystem is mounted. This lock only needs to be grabbed on
freezing, and actions which need to be safe from freezing, like
recovery.
When a node wants to freeze the filesystem, it grabs this glock
exclusively. When the freeze glock state changes on the nodes (either
from shared to unlocked, or shared to exclusive), the filesystem does a
special log flush. gfs2_log_flush() does all the work for flushing out
the and shutting down the incore log, and then it tries to grab the
freeze glock in a shared state again. Since the filesystem is stuck in
gfs2_log_flush, no new transaction can start, and nothing can be written
to disk. Unfreezing the filesytem simply involes dropping the freeze
glock, allowing gfs2_log_flush() to grab and then release the shared
lock, so it is cached for next time.
However, in order for the unfreezing ioctl to occur, gfs2 needs to get a
shared lock on the filesystem root directory inode to check permissions.
If that glock has already been grabbed exclusively, fsfreeze will be
unable to get the shared lock and unfreeze the filesystem.
In order to allow the unfreeze, this patch makes gfs2 grab a shared lock
on the filesystem root directory during the freeze, and hold it until it
unfreezes the filesystem. The functions which need to grab a shared
lock in order to allow the unfreeze ioctl to be issued now use the lock
grabbed by the freeze code instead.
The freeze and unfreeze code take care to make sure that this shared
lock will not be dropped while another process is using it.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2014-05-02 03:26:55 +00:00
|
|
|
gfs2_glock_put(sdp->sd_freeze_gl);
|
2006-06-14 19:32:57 +00:00
|
|
|
fail_rename:
|
2006-01-16 16:50:04 +00:00
|
|
|
gfs2_glock_put(sdp->sd_rename_gl);
|
2006-06-14 19:32:57 +00:00
|
|
|
fail_live:
|
2006-01-16 16:50:04 +00:00
|
|
|
gfs2_glock_dq_uninit(&sdp->sd_live_gh);
|
2006-06-14 19:32:57 +00:00
|
|
|
fail_mount:
|
2006-01-16 16:50:04 +00:00
|
|
|
gfs2_glock_dq_uninit(mount_gh);
|
2006-06-14 19:32:57 +00:00
|
|
|
fail:
|
2006-01-16 16:50:04 +00:00
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2008-08-08 12:45:13 +00:00
|
|
|
static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr,
|
|
|
|
u64 no_addr, const char *name)
|
2006-01-30 18:34:10 +00:00
|
|
|
{
|
2008-08-08 12:45:13 +00:00
|
|
|
struct gfs2_sbd *sdp = sb->s_fs_info;
|
|
|
|
struct dentry *dentry;
|
|
|
|
struct inode *inode;
|
|
|
|
|
2011-03-17 20:19:58 +00:00
|
|
|
inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0, 0);
|
2008-08-08 12:45:13 +00:00
|
|
|
if (IS_ERR(inode)) {
|
|
|
|
fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode));
|
|
|
|
return PTR_ERR(inode);
|
|
|
|
}
|
2012-01-09 03:15:13 +00:00
|
|
|
dentry = d_make_root(inode);
|
2008-08-08 12:45:13 +00:00
|
|
|
if (!dentry) {
|
|
|
|
fs_err(sdp, "can't alloc %s dentry\n", name);
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
*dptr = dentry;
|
|
|
|
return 0;
|
2006-01-30 18:34:10 +00:00
|
|
|
}
|
|
|
|
|
2008-08-08 12:45:13 +00:00
|
|
|
static int init_sb(struct gfs2_sbd *sdp, int silent)
|
2006-01-16 16:50:04 +00:00
|
|
|
{
|
|
|
|
struct super_block *sb = sdp->sd_vfs;
|
|
|
|
struct gfs2_holder sb_gh;
|
2007-05-15 14:37:50 +00:00
|
|
|
u64 no_addr;
|
2008-08-08 12:45:13 +00:00
|
|
|
int ret;
|
2006-01-16 16:50:04 +00:00
|
|
|
|
2008-08-08 12:45:13 +00:00
|
|
|
ret = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops,
|
|
|
|
LM_ST_SHARED, 0, &sb_gh);
|
|
|
|
if (ret) {
|
|
|
|
fs_err(sdp, "can't acquire superblock glock: %d\n", ret);
|
|
|
|
return ret;
|
2006-01-16 16:50:04 +00:00
|
|
|
}
|
2006-09-25 13:26:04 +00:00
|
|
|
|
2009-03-09 09:03:51 +00:00
|
|
|
ret = gfs2_read_sb(sdp, silent);
|
2008-08-08 12:45:13 +00:00
|
|
|
if (ret) {
|
|
|
|
fs_err(sdp, "can't read superblock: %d\n", ret);
|
2006-01-16 16:50:04 +00:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Set up the buffer cache and SB for real */
|
2009-05-22 21:17:49 +00:00
|
|
|
if (sdp->sd_sb.sb_bsize < bdev_logical_block_size(sb->s_bdev)) {
|
2008-08-08 12:45:13 +00:00
|
|
|
ret = -EINVAL;
|
2006-01-16 16:50:04 +00:00
|
|
|
fs_err(sdp, "FS block size (%u) is too small for device "
|
|
|
|
"block size (%u)\n",
|
2009-05-22 21:17:49 +00:00
|
|
|
sdp->sd_sb.sb_bsize, bdev_logical_block_size(sb->s_bdev));
|
2006-01-16 16:50:04 +00:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
|
2008-08-08 12:45:13 +00:00
|
|
|
ret = -EINVAL;
|
2006-01-16 16:50:04 +00:00
|
|
|
fs_err(sdp, "FS block size (%u) is too big for machine "
|
|
|
|
"page size (%u)\n",
|
|
|
|
sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
|
|
|
|
|
2006-01-30 18:34:10 +00:00
|
|
|
/* Get the root inode */
|
2007-05-15 14:37:50 +00:00
|
|
|
no_addr = sdp->sd_sb.sb_root_dir.no_addr;
|
2008-08-08 12:45:13 +00:00
|
|
|
ret = gfs2_lookup_root(sb, &sdp->sd_root_dir, no_addr, "root");
|
|
|
|
if (ret)
|
2006-01-30 18:34:10 +00:00
|
|
|
goto out;
|
2006-01-16 16:50:04 +00:00
|
|
|
|
2008-08-08 12:45:13 +00:00
|
|
|
/* Get the master inode */
|
|
|
|
no_addr = sdp->sd_sb.sb_master_dir.no_addr;
|
|
|
|
ret = gfs2_lookup_root(sb, &sdp->sd_master_dir, no_addr, "master");
|
|
|
|
if (ret) {
|
|
|
|
dput(sdp->sd_root_dir);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
sb->s_root = dget(sdp->sd_args.ar_meta ? sdp->sd_master_dir : sdp->sd_root_dir);
|
2006-01-30 18:34:10 +00:00
|
|
|
out:
|
2006-01-16 16:50:04 +00:00
|
|
|
gfs2_glock_dq_uninit(&sb_gh);
|
2008-08-08 12:45:13 +00:00
|
|
|
return ret;
|
2006-01-16 16:50:04 +00:00
|
|
|
}
|
|
|
|
|
2009-01-12 10:43:39 +00:00
|
|
|
static void gfs2_others_may_mount(struct gfs2_sbd *sdp)
|
2008-01-30 15:34:04 +00:00
|
|
|
{
|
2009-01-12 10:43:39 +00:00
|
|
|
char *message = "FIRSTMOUNT=Done";
|
|
|
|
char *envp[] = { message, NULL };
|
2012-01-09 22:18:05 +00:00
|
|
|
|
|
|
|
fs_info(sdp, "first mount done, others may mount\n");
|
|
|
|
|
|
|
|
if (sdp->sd_lockstruct.ls_ops->lm_first_done)
|
|
|
|
sdp->sd_lockstruct.ls_ops->lm_first_done(sdp);
|
|
|
|
|
2009-01-12 10:43:39 +00:00
|
|
|
kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp);
|
2008-01-30 15:34:04 +00:00
|
|
|
}
|
|
|
|
|
2008-11-26 12:49:26 +00:00
|
|
|
/**
|
|
|
|
* gfs2_jindex_hold - Grab a lock on the jindex
|
|
|
|
* @sdp: The GFS2 superblock
|
|
|
|
* @ji_gh: the holder for the jindex glock
|
|
|
|
*
|
|
|
|
* Returns: errno
|
|
|
|
*/
|
|
|
|
|
|
|
|
static int gfs2_jindex_hold(struct gfs2_sbd *sdp, struct gfs2_holder *ji_gh)
|
|
|
|
{
|
|
|
|
struct gfs2_inode *dip = GFS2_I(sdp->sd_jindex);
|
|
|
|
struct qstr name;
|
|
|
|
char buf[20];
|
|
|
|
struct gfs2_jdesc *jd;
|
|
|
|
int error;
|
|
|
|
|
|
|
|
name.name = buf;
|
|
|
|
|
|
|
|
mutex_lock(&sdp->sd_jindex_mutex);
|
|
|
|
|
|
|
|
for (;;) {
|
|
|
|
error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, ji_gh);
|
|
|
|
if (error)
|
|
|
|
break;
|
|
|
|
|
|
|
|
name.len = sprintf(buf, "journal%u", sdp->sd_journals);
|
|
|
|
name.hash = gfs2_disk_hash(name.name, name.len);
|
|
|
|
|
|
|
|
error = gfs2_dir_check(sdp->sd_jindex, &name, NULL);
|
|
|
|
if (error == -ENOENT) {
|
|
|
|
error = 0;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
gfs2_glock_dq_uninit(ji_gh);
|
|
|
|
|
|
|
|
if (error)
|
|
|
|
break;
|
|
|
|
|
|
|
|
error = -ENOMEM;
|
|
|
|
jd = kzalloc(sizeof(struct gfs2_jdesc), GFP_KERNEL);
|
|
|
|
if (!jd)
|
|
|
|
break;
|
|
|
|
|
|
|
|
INIT_LIST_HEAD(&jd->extent_list);
|
2014-03-06 22:19:15 +00:00
|
|
|
INIT_LIST_HEAD(&jd->jd_revoke_list);
|
|
|
|
|
2010-07-20 20:09:02 +00:00
|
|
|
INIT_WORK(&jd->jd_work, gfs2_recover_func);
|
2008-11-26 12:49:26 +00:00
|
|
|
jd->jd_inode = gfs2_lookupi(sdp->sd_jindex, &name, 1);
|
|
|
|
if (!jd->jd_inode || IS_ERR(jd->jd_inode)) {
|
|
|
|
if (!jd->jd_inode)
|
|
|
|
error = -ENOENT;
|
|
|
|
else
|
|
|
|
error = PTR_ERR(jd->jd_inode);
|
|
|
|
kfree(jd);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
spin_lock(&sdp->sd_jindex_spin);
|
|
|
|
jd->jd_jid = sdp->sd_journals++;
|
|
|
|
list_add_tail(&jd->jd_list, &sdp->sd_jindex_list);
|
|
|
|
spin_unlock(&sdp->sd_jindex_spin);
|
|
|
|
}
|
|
|
|
|
|
|
|
mutex_unlock(&sdp->sd_jindex_mutex);
|
|
|
|
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2013-09-04 16:08:02 +00:00
|
|
|
/**
|
|
|
|
* check_journal_clean - Make sure a journal is clean for a spectator mount
|
|
|
|
* @sdp: The GFS2 superblock
|
|
|
|
* @jd: The journal descriptor
|
|
|
|
*
|
|
|
|
* Returns: 0 if the journal is clean or locked, else an error
|
|
|
|
*/
|
|
|
|
static int check_journal_clean(struct gfs2_sbd *sdp, struct gfs2_jdesc *jd)
|
|
|
|
{
|
|
|
|
int error;
|
|
|
|
struct gfs2_holder j_gh;
|
|
|
|
struct gfs2_log_header_host head;
|
|
|
|
struct gfs2_inode *ip;
|
|
|
|
|
|
|
|
ip = GFS2_I(jd->jd_inode);
|
|
|
|
error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_NOEXP |
|
|
|
|
GL_EXACT | GL_NOCACHE, &j_gh);
|
|
|
|
if (error) {
|
|
|
|
fs_err(sdp, "Error locking journal for spectator mount.\n");
|
|
|
|
return -EPERM;
|
|
|
|
}
|
|
|
|
error = gfs2_jdesc_check(jd);
|
|
|
|
if (error) {
|
|
|
|
fs_err(sdp, "Error checking journal for spectator mount.\n");
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
|
|
|
error = gfs2_find_jhead(jd, &head);
|
|
|
|
if (error) {
|
|
|
|
fs_err(sdp, "Error parsing journal for spectator mount.\n");
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
|
|
|
if (!(head.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) {
|
|
|
|
error = -EPERM;
|
|
|
|
fs_err(sdp, "jid=%u: Journal is dirty, so the first mounter "
|
|
|
|
"must not be a spectator.\n", jd->jd_jid);
|
|
|
|
}
|
|
|
|
|
|
|
|
out_unlock:
|
|
|
|
gfs2_glock_dq_uninit(&j_gh);
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2006-01-16 16:50:04 +00:00
|
|
|
static int init_journal(struct gfs2_sbd *sdp, int undo)
|
|
|
|
{
|
2008-08-08 12:45:13 +00:00
|
|
|
struct inode *master = sdp->sd_master_dir->d_inode;
|
2006-01-16 16:50:04 +00:00
|
|
|
struct gfs2_holder ji_gh;
|
2006-02-27 22:23:27 +00:00
|
|
|
struct gfs2_inode *ip;
|
2006-01-16 16:50:04 +00:00
|
|
|
int jindex = 1;
|
|
|
|
int error = 0;
|
|
|
|
|
|
|
|
if (undo) {
|
|
|
|
jindex = 0;
|
2009-05-19 09:01:18 +00:00
|
|
|
goto fail_jinode_gh;
|
2006-01-16 16:50:04 +00:00
|
|
|
}
|
|
|
|
|
2008-08-08 12:45:13 +00:00
|
|
|
sdp->sd_jindex = gfs2_lookup_simple(master, "jindex");
|
2006-03-20 17:30:04 +00:00
|
|
|
if (IS_ERR(sdp->sd_jindex)) {
|
2006-01-16 16:50:04 +00:00
|
|
|
fs_err(sdp, "can't lookup journal index: %d\n", error);
|
2006-03-20 17:30:04 +00:00
|
|
|
return PTR_ERR(sdp->sd_jindex);
|
2006-01-16 16:50:04 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Load in the journal index special file */
|
|
|
|
|
|
|
|
error = gfs2_jindex_hold(sdp, &ji_gh);
|
|
|
|
if (error) {
|
|
|
|
fs_err(sdp, "can't read journal index: %d\n", error);
|
|
|
|
goto fail;
|
|
|
|
}
|
|
|
|
|
2010-02-05 23:25:41 +00:00
|
|
|
error = -EUSERS;
|
2006-01-16 16:50:04 +00:00
|
|
|
if (!gfs2_jindex_size(sdp)) {
|
|
|
|
fs_err(sdp, "no journals!\n");
|
2006-09-25 13:26:04 +00:00
|
|
|
goto fail_jindex;
|
2006-01-16 16:50:04 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
if (sdp->sd_args.ar_spectator) {
|
|
|
|
sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
|
2007-11-08 14:55:03 +00:00
|
|
|
atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
|
GFS2: Various gfs2_logd improvements
This patch contains various tweaks to how log flushes and active item writeback
work. gfs2_logd is now managed by a waitqueue, and gfs2_log_reseve now waits
for gfs2_logd to do the log flushing. Multiple functions were rewritten to
remove the need to call gfs2_log_lock(). Instead of using one test to see if
gfs2_logd had work to do, there are now seperate tests to check if there
are two many buffers in the incore log or if there are two many items on the
active items list.
This patch is a port of a patch Steve Whitehouse wrote about a year ago, with
some minor changes. Since gfs2_ail1_start always submits all the active items,
it no longer needs to keep track of the first ai submitted, so this has been
removed. In gfs2_log_reserve(), the order of the calls to
prepare_to_wait_exclusive() and wake_up() when firing off the logd thread has
been switched. If it called wake_up first there was a small window for a race,
where logd could run and return before gfs2_log_reserve was ready to get woken
up. If gfs2_logd ran, but did not free up enough blocks, gfs2_log_reserve()
would be left waiting for gfs2_logd to eventualy run because it timed out.
Finally, gt_logd_secs, which controls how long to wait before gfs2_logd times
out, and flushes the log, can now be set on mount with ar_commit.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2010-05-04 19:29:16 +00:00
|
|
|
atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
|
|
|
|
atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
|
2006-01-16 16:50:04 +00:00
|
|
|
} else {
|
|
|
|
if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) {
|
|
|
|
fs_err(sdp, "can't mount journal #%u\n",
|
|
|
|
sdp->sd_lockstruct.ls_jid);
|
|
|
|
fs_err(sdp, "there are only %u journals (0 - %u)\n",
|
|
|
|
gfs2_jindex_size(sdp),
|
|
|
|
gfs2_jindex_size(sdp) - 1);
|
|
|
|
goto fail_jindex;
|
|
|
|
}
|
|
|
|
sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid);
|
|
|
|
|
2006-06-14 19:32:57 +00:00
|
|
|
error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid,
|
2006-01-16 16:50:04 +00:00
|
|
|
&gfs2_journal_glops,
|
|
|
|
LM_ST_EXCLUSIVE, LM_FLAG_NOEXP,
|
|
|
|
&sdp->sd_journal_gh);
|
|
|
|
if (error) {
|
|
|
|
fs_err(sdp, "can't acquire journal glock: %d\n", error);
|
|
|
|
goto fail_jindex;
|
|
|
|
}
|
|
|
|
|
2006-06-14 19:32:57 +00:00
|
|
|
ip = GFS2_I(sdp->sd_jdesc->jd_inode);
|
|
|
|
error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
|
2007-08-08 22:08:14 +00:00
|
|
|
LM_FLAG_NOEXP | GL_EXACT | GL_NOCACHE,
|
2006-01-16 16:50:04 +00:00
|
|
|
&sdp->sd_jinode_gh);
|
|
|
|
if (error) {
|
|
|
|
fs_err(sdp, "can't acquire journal inode glock: %d\n",
|
|
|
|
error);
|
|
|
|
goto fail_journal_gh;
|
|
|
|
}
|
|
|
|
|
|
|
|
error = gfs2_jdesc_check(sdp->sd_jdesc);
|
|
|
|
if (error) {
|
|
|
|
fs_err(sdp, "my journal (%u) is bad: %d\n",
|
|
|
|
sdp->sd_jdesc->jd_jid, error);
|
|
|
|
goto fail_jinode_gh;
|
|
|
|
}
|
2007-11-08 14:55:03 +00:00
|
|
|
atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
|
GFS2: Various gfs2_logd improvements
This patch contains various tweaks to how log flushes and active item writeback
work. gfs2_logd is now managed by a waitqueue, and gfs2_log_reseve now waits
for gfs2_logd to do the log flushing. Multiple functions were rewritten to
remove the need to call gfs2_log_lock(). Instead of using one test to see if
gfs2_logd had work to do, there are now seperate tests to check if there
are two many buffers in the incore log or if there are two many items on the
active items list.
This patch is a port of a patch Steve Whitehouse wrote about a year ago, with
some minor changes. Since gfs2_ail1_start always submits all the active items,
it no longer needs to keep track of the first ai submitted, so this has been
removed. In gfs2_log_reserve(), the order of the calls to
prepare_to_wait_exclusive() and wake_up() when firing off the logd thread has
been switched. If it called wake_up first there was a small window for a race,
where logd could run and return before gfs2_log_reserve was ready to get woken
up. If gfs2_logd ran, but did not free up enough blocks, gfs2_log_reserve()
would be left waiting for gfs2_logd to eventualy run because it timed out.
Finally, gt_logd_secs, which controls how long to wait before gfs2_logd times
out, and flushes the log, can now be set on mount with ar_commit.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2010-05-04 19:29:16 +00:00
|
|
|
atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
|
|
|
|
atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
|
2007-12-12 00:49:21 +00:00
|
|
|
|
|
|
|
/* Map the extents for this journal's blocks */
|
2014-03-03 13:35:57 +00:00
|
|
|
gfs2_map_journal_extents(sdp, sdp->sd_jdesc);
|
2006-01-16 16:50:04 +00:00
|
|
|
}
|
2009-06-12 07:49:20 +00:00
|
|
|
trace_gfs2_log_blocks(sdp, atomic_read(&sdp->sd_log_blks_free));
|
2006-01-16 16:50:04 +00:00
|
|
|
|
|
|
|
if (sdp->sd_lockstruct.ls_first) {
|
|
|
|
unsigned int x;
|
|
|
|
for (x = 0; x < sdp->sd_journals; x++) {
|
2013-09-04 16:08:02 +00:00
|
|
|
struct gfs2_jdesc *jd = gfs2_jdesc_find(sdp, x);
|
|
|
|
|
|
|
|
if (sdp->sd_args.ar_spectator) {
|
|
|
|
error = check_journal_clean(sdp, jd);
|
|
|
|
if (error)
|
|
|
|
goto fail_jinode_gh;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
error = gfs2_recover_journal(jd, true);
|
2006-01-16 16:50:04 +00:00
|
|
|
if (error) {
|
|
|
|
fs_err(sdp, "error recovering journal %u: %d\n",
|
|
|
|
x, error);
|
|
|
|
goto fail_jinode_gh;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2009-01-12 10:43:39 +00:00
|
|
|
gfs2_others_may_mount(sdp);
|
2006-01-16 16:50:04 +00:00
|
|
|
} else if (!sdp->sd_args.ar_spectator) {
|
2010-07-20 20:09:02 +00:00
|
|
|
error = gfs2_recover_journal(sdp->sd_jdesc, true);
|
2006-01-16 16:50:04 +00:00
|
|
|
if (error) {
|
|
|
|
fs_err(sdp, "error recovering my journal: %d\n", error);
|
|
|
|
goto fail_jinode_gh;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
|
|
|
|
gfs2_glock_dq_uninit(&ji_gh);
|
|
|
|
jindex = 0;
|
GFS2: remove transaction glock
GFS2 has a transaction glock, which must be grabbed for every
transaction, whose purpose is to deal with freezing the filesystem.
Aside from this involving a large amount of locking, it is very easy to
make the current fsfreeze code hang on unfreezing.
This patch rewrites how gfs2 handles freezing the filesystem. The
transaction glock is removed. In it's place is a freeze glock, which is
cached (but not held) in a shared state by every node in the cluster
when the filesystem is mounted. This lock only needs to be grabbed on
freezing, and actions which need to be safe from freezing, like
recovery.
When a node wants to freeze the filesystem, it grabs this glock
exclusively. When the freeze glock state changes on the nodes (either
from shared to unlocked, or shared to exclusive), the filesystem does a
special log flush. gfs2_log_flush() does all the work for flushing out
the and shutting down the incore log, and then it tries to grab the
freeze glock in a shared state again. Since the filesystem is stuck in
gfs2_log_flush, no new transaction can start, and nothing can be written
to disk. Unfreezing the filesytem simply involes dropping the freeze
glock, allowing gfs2_log_flush() to grab and then release the shared
lock, so it is cached for next time.
However, in order for the unfreezing ioctl to occur, gfs2 needs to get a
shared lock on the filesystem root directory inode to check permissions.
If that glock has already been grabbed exclusively, fsfreeze will be
unable to get the shared lock and unfreeze the filesystem.
In order to allow the unfreeze, this patch makes gfs2 grab a shared lock
on the filesystem root directory during the freeze, and hold it until it
unfreezes the filesystem. The functions which need to grab a shared
lock in order to allow the unfreeze ioctl to be issued now use the lock
grabbed by the freeze code instead.
The freeze and unfreeze code take care to make sure that this shared
lock will not be dropped while another process is using it.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2014-05-02 03:26:55 +00:00
|
|
|
if (!sdp->sd_args.ar_spectator) {
|
|
|
|
error = gfs2_glock_nq_init(sdp->sd_freeze_gl, LM_ST_SHARED, 0,
|
|
|
|
&sdp->sd_thaw_gh);
|
|
|
|
if (error) {
|
|
|
|
fs_err(sdp, "can't acquire freeze glock: %d\n", error);
|
|
|
|
goto fail_jinode_gh;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
gfs2_glock_dq_uninit(&sdp->sd_thaw_gh);
|
2006-01-16 16:50:04 +00:00
|
|
|
return 0;
|
|
|
|
|
2006-09-04 16:04:26 +00:00
|
|
|
fail_jinode_gh:
|
2006-01-16 16:50:04 +00:00
|
|
|
if (!sdp->sd_args.ar_spectator)
|
|
|
|
gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
|
2006-09-04 16:04:26 +00:00
|
|
|
fail_journal_gh:
|
2006-01-16 16:50:04 +00:00
|
|
|
if (!sdp->sd_args.ar_spectator)
|
|
|
|
gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
|
2006-09-04 16:04:26 +00:00
|
|
|
fail_jindex:
|
2006-01-16 16:50:04 +00:00
|
|
|
gfs2_jindex_free(sdp);
|
|
|
|
if (jindex)
|
|
|
|
gfs2_glock_dq_uninit(&ji_gh);
|
2006-09-04 16:04:26 +00:00
|
|
|
fail:
|
2006-01-30 18:34:10 +00:00
|
|
|
iput(sdp->sd_jindex);
|
2006-01-16 16:50:04 +00:00
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2012-09-05 20:55:11 +00:00
|
|
|
static struct lock_class_key gfs2_quota_imutex_key;
|
2006-01-16 16:50:04 +00:00
|
|
|
|
|
|
|
static int init_inodes(struct gfs2_sbd *sdp, int undo)
|
|
|
|
{
|
|
|
|
int error = 0;
|
2008-08-08 12:45:13 +00:00
|
|
|
struct inode *master = sdp->sd_master_dir->d_inode;
|
2006-01-16 16:50:04 +00:00
|
|
|
|
|
|
|
if (undo)
|
2006-01-30 18:34:10 +00:00
|
|
|
goto fail_qinode;
|
|
|
|
|
|
|
|
error = init_journal(sdp, undo);
|
2014-06-02 13:40:25 +00:00
|
|
|
complete_all(&sdp->sd_journal_ready);
|
2006-01-30 18:34:10 +00:00
|
|
|
if (error)
|
2008-08-08 12:45:13 +00:00
|
|
|
goto fail;
|
2006-01-16 16:50:04 +00:00
|
|
|
|
|
|
|
/* Read in the master statfs inode */
|
2008-08-08 12:45:13 +00:00
|
|
|
sdp->sd_statfs_inode = gfs2_lookup_simple(master, "statfs");
|
2006-03-20 17:30:04 +00:00
|
|
|
if (IS_ERR(sdp->sd_statfs_inode)) {
|
|
|
|
error = PTR_ERR(sdp->sd_statfs_inode);
|
2006-01-16 16:50:04 +00:00
|
|
|
fs_err(sdp, "can't read in statfs inode: %d\n", error);
|
2009-08-27 14:51:07 +00:00
|
|
|
goto fail_journal;
|
2006-01-16 16:50:04 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Read in the resource index inode */
|
2008-08-08 12:45:13 +00:00
|
|
|
sdp->sd_rindex = gfs2_lookup_simple(master, "rindex");
|
2006-03-20 17:30:04 +00:00
|
|
|
if (IS_ERR(sdp->sd_rindex)) {
|
|
|
|
error = PTR_ERR(sdp->sd_rindex);
|
2006-01-16 16:50:04 +00:00
|
|
|
fs_err(sdp, "can't get resource index inode: %d\n", error);
|
|
|
|
goto fail_statfs;
|
|
|
|
}
|
2008-01-31 16:31:39 +00:00
|
|
|
sdp->sd_rindex_uptodate = 0;
|
2006-01-16 16:50:04 +00:00
|
|
|
|
|
|
|
/* Read in the quota inode */
|
2008-08-08 12:45:13 +00:00
|
|
|
sdp->sd_quota_inode = gfs2_lookup_simple(master, "quota");
|
2006-03-20 17:30:04 +00:00
|
|
|
if (IS_ERR(sdp->sd_quota_inode)) {
|
|
|
|
error = PTR_ERR(sdp->sd_quota_inode);
|
2006-01-16 16:50:04 +00:00
|
|
|
fs_err(sdp, "can't get quota file inode: %d\n", error);
|
|
|
|
goto fail_rindex;
|
|
|
|
}
|
2012-09-05 20:55:11 +00:00
|
|
|
/*
|
|
|
|
* i_mutex on quota files is special. Since this inode is hidden system
|
|
|
|
* file, we are safe to define locking ourselves.
|
|
|
|
*/
|
|
|
|
lockdep_set_class(&sdp->sd_quota_inode->i_mutex,
|
|
|
|
&gfs2_quota_imutex_key);
|
2012-02-24 15:09:14 +00:00
|
|
|
|
|
|
|
error = gfs2_rindex_update(sdp);
|
|
|
|
if (error)
|
|
|
|
goto fail_qinode;
|
|
|
|
|
2006-01-16 16:50:04 +00:00
|
|
|
return 0;
|
|
|
|
|
2006-01-30 18:34:10 +00:00
|
|
|
fail_qinode:
|
|
|
|
iput(sdp->sd_quota_inode);
|
|
|
|
fail_rindex:
|
2006-01-16 16:50:04 +00:00
|
|
|
gfs2_clear_rgrpd(sdp);
|
2006-01-30 18:34:10 +00:00
|
|
|
iput(sdp->sd_rindex);
|
|
|
|
fail_statfs:
|
|
|
|
iput(sdp->sd_statfs_inode);
|
|
|
|
fail_journal:
|
|
|
|
init_journal(sdp, UNDO);
|
|
|
|
fail:
|
2006-01-16 16:50:04 +00:00
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int init_per_node(struct gfs2_sbd *sdp, int undo)
|
|
|
|
{
|
2006-01-30 18:34:10 +00:00
|
|
|
struct inode *pn = NULL;
|
2006-01-16 16:50:04 +00:00
|
|
|
char buf[30];
|
|
|
|
int error = 0;
|
2006-02-27 22:23:27 +00:00
|
|
|
struct gfs2_inode *ip;
|
2008-08-08 12:45:13 +00:00
|
|
|
struct inode *master = sdp->sd_master_dir->d_inode;
|
2006-01-16 16:50:04 +00:00
|
|
|
|
|
|
|
if (sdp->sd_args.ar_spectator)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
if (undo)
|
|
|
|
goto fail_qc_gh;
|
|
|
|
|
2008-08-08 12:45:13 +00:00
|
|
|
pn = gfs2_lookup_simple(master, "per_node");
|
2006-03-20 17:30:04 +00:00
|
|
|
if (IS_ERR(pn)) {
|
|
|
|
error = PTR_ERR(pn);
|
2006-01-16 16:50:04 +00:00
|
|
|
fs_err(sdp, "can't find per_node directory: %d\n", error);
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
|
2006-03-20 17:30:04 +00:00
|
|
|
sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf);
|
|
|
|
if (IS_ERR(sdp->sd_sc_inode)) {
|
|
|
|
error = PTR_ERR(sdp->sd_sc_inode);
|
2006-01-16 16:50:04 +00:00
|
|
|
fs_err(sdp, "can't find local \"sc\" file: %d\n", error);
|
2009-08-27 14:51:07 +00:00
|
|
|
goto fail;
|
2006-01-16 16:50:04 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
|
2006-03-20 17:30:04 +00:00
|
|
|
sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf);
|
|
|
|
if (IS_ERR(sdp->sd_qc_inode)) {
|
|
|
|
error = PTR_ERR(sdp->sd_qc_inode);
|
2006-01-16 16:50:04 +00:00
|
|
|
fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
|
|
|
|
goto fail_ut_i;
|
|
|
|
}
|
|
|
|
|
2006-01-30 18:34:10 +00:00
|
|
|
iput(pn);
|
2006-01-16 16:50:04 +00:00
|
|
|
pn = NULL;
|
|
|
|
|
2006-06-14 19:32:57 +00:00
|
|
|
ip = GFS2_I(sdp->sd_sc_inode);
|
2009-08-27 14:51:07 +00:00
|
|
|
error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
|
2006-01-16 16:50:04 +00:00
|
|
|
&sdp->sd_sc_gh);
|
|
|
|
if (error) {
|
|
|
|
fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
|
2009-08-27 14:51:07 +00:00
|
|
|
goto fail_qc_i;
|
2006-01-16 16:50:04 +00:00
|
|
|
}
|
|
|
|
|
2006-06-14 19:32:57 +00:00
|
|
|
ip = GFS2_I(sdp->sd_qc_inode);
|
2009-08-27 14:51:07 +00:00
|
|
|
error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
|
2006-01-16 16:50:04 +00:00
|
|
|
&sdp->sd_qc_gh);
|
|
|
|
if (error) {
|
|
|
|
fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);
|
|
|
|
goto fail_ut_gh;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
2006-09-04 16:04:26 +00:00
|
|
|
fail_qc_gh:
|
2006-01-16 16:50:04 +00:00
|
|
|
gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
|
2006-09-04 16:04:26 +00:00
|
|
|
fail_ut_gh:
|
2006-01-16 16:50:04 +00:00
|
|
|
gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
|
2006-09-04 16:04:26 +00:00
|
|
|
fail_qc_i:
|
2006-01-30 18:34:10 +00:00
|
|
|
iput(sdp->sd_qc_inode);
|
2006-09-04 16:04:26 +00:00
|
|
|
fail_ut_i:
|
2006-01-30 18:34:10 +00:00
|
|
|
iput(sdp->sd_sc_inode);
|
2006-09-04 16:04:26 +00:00
|
|
|
fail:
|
2006-01-16 16:50:04 +00:00
|
|
|
if (pn)
|
2006-01-30 18:34:10 +00:00
|
|
|
iput(pn);
|
2006-01-16 16:50:04 +00:00
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2009-01-12 10:43:39 +00:00
|
|
|
static const match_table_t nolock_tokens = {
|
|
|
|
{ Opt_jid, "jid=%d\n", },
|
|
|
|
{ Opt_err, NULL },
|
|
|
|
};
|
|
|
|
|
|
|
|
static const struct lm_lockops nolock_ops = {
|
|
|
|
.lm_proto_name = "lock_nolock",
|
2011-03-09 10:58:04 +00:00
|
|
|
.lm_put_lock = gfs2_glock_free,
|
2009-01-12 10:43:39 +00:00
|
|
|
.lm_tokens = &nolock_tokens,
|
|
|
|
};
|
|
|
|
|
2008-01-30 15:34:04 +00:00
|
|
|
/**
|
|
|
|
* gfs2_lm_mount - mount a locking protocol
|
|
|
|
* @sdp: the filesystem
|
tree-wide: Assorted spelling fixes
In particular, several occurances of funny versions of 'success',
'unknown', 'therefore', 'acknowledge', 'argument', 'achieve', 'address',
'beginning', 'desirable', 'separate' and 'necessary' are fixed.
Signed-off-by: Daniel Mack <daniel@caiaq.de>
Cc: Joe Perches <joe@perches.com>
Cc: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2010-02-03 00:01:28 +00:00
|
|
|
* @args: mount arguments
|
2008-01-30 15:34:04 +00:00
|
|
|
* @silent: if 1, don't complain if the FS isn't a GFS2 fs
|
|
|
|
*
|
|
|
|
* Returns: errno
|
|
|
|
*/
|
|
|
|
|
|
|
|
static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent)
|
|
|
|
{
|
2009-01-12 10:43:39 +00:00
|
|
|
const struct lm_lockops *lm;
|
|
|
|
struct lm_lockstruct *ls = &sdp->sd_lockstruct;
|
|
|
|
struct gfs2_args *args = &sdp->sd_args;
|
|
|
|
const char *proto = sdp->sd_proto_name;
|
|
|
|
const char *table = sdp->sd_table_name;
|
|
|
|
char *o, *options;
|
|
|
|
int ret;
|
2008-01-30 15:34:04 +00:00
|
|
|
|
2009-01-12 10:43:39 +00:00
|
|
|
if (!strcmp("lock_nolock", proto)) {
|
|
|
|
lm = &nolock_ops;
|
|
|
|
sdp->sd_args.ar_localflocks = 1;
|
|
|
|
#ifdef CONFIG_GFS2_FS_LOCKING_DLM
|
|
|
|
} else if (!strcmp("lock_dlm", proto)) {
|
|
|
|
lm = &gfs2_dlm_ops;
|
|
|
|
#endif
|
|
|
|
} else {
|
2014-03-06 20:10:45 +00:00
|
|
|
pr_info("can't find protocol %s\n", proto);
|
2009-01-12 10:43:39 +00:00
|
|
|
return -ENOENT;
|
|
|
|
}
|
2008-01-30 15:34:04 +00:00
|
|
|
|
|
|
|
fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table);
|
|
|
|
|
2009-01-12 10:43:39 +00:00
|
|
|
ls->ls_ops = lm;
|
|
|
|
ls->ls_first = 1;
|
2008-01-30 15:34:04 +00:00
|
|
|
|
2009-01-12 10:43:39 +00:00
|
|
|
for (options = args->ar_hostdata; (o = strsep(&options, ":")); ) {
|
|
|
|
substring_t tmp[MAX_OPT_ARGS];
|
|
|
|
int token, option;
|
|
|
|
|
|
|
|
if (!o || !*o)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
token = match_token(o, *lm->lm_tokens, tmp);
|
|
|
|
switch (token) {
|
|
|
|
case Opt_jid:
|
|
|
|
ret = match_int(&tmp[0], &option);
|
|
|
|
if (ret || option < 0)
|
|
|
|
goto hostdata_error;
|
2010-06-14 09:01:30 +00:00
|
|
|
if (test_and_clear_bit(SDF_NOJOURNALID, &sdp->sd_flags))
|
|
|
|
ls->ls_jid = option;
|
2009-01-12 10:43:39 +00:00
|
|
|
break;
|
|
|
|
case Opt_id:
|
dlm: fixes for nodir mode
The "nodir" mode (statically assign master nodes instead
of using the resource directory) has always been highly
experimental, and never seriously used. This commit
fixes a number of problems, making nodir much more usable.
- Major change to recovery: recover all locks and restart
all in-progress operations after recovery. In some
cases it's not possible to know which in-progess locks
to recover, so recover all. (Most require recovery
in nodir mode anyway since rehashing changes most
master nodes.)
- Change the way nodir mode is enabled, from a command
line mount arg passed through gfs2, into a sysfs
file managed by dlm_controld, consistent with the
other config settings.
- Allow recovering MSTCPY locks on an rsb that has not
yet been turned into a master copy.
- Ignore RCOM_LOCK and RCOM_LOCK_REPLY recovery messages
from a previous, aborted recovery cycle. Base this
on the local recovery status not being in the state
where any nodes should be sending LOCK messages for the
current recovery cycle.
- Hold rsb lock around dlm_purge_mstcpy_locks() because it
may run concurrently with dlm_recover_master_copy().
- Maintain highbast on process-copy lkb's (in addition to
the master as is usual), because the lkb can switch
back and forth between being a master and being a
process copy as the master node changes in recovery.
- When recovering MSTCPY locks, flag rsb's that have
non-empty convert or waiting queues for granting
at the end of recovery. (Rename flag from LOCKS_PURGED
to RECOVER_GRANT and similar for the recovery function,
because it's not only resources with purged locks
that need grant a grant attempt.)
- Replace a couple of unnecessary assertion panics with
error messages.
Signed-off-by: David Teigland <teigland@redhat.com>
2012-04-26 20:54:29 +00:00
|
|
|
case Opt_nodir:
|
2009-09-09 14:59:35 +00:00
|
|
|
/* Obsolete, but left for backward compat purposes */
|
2009-01-12 10:43:39 +00:00
|
|
|
break;
|
|
|
|
case Opt_first:
|
|
|
|
ret = match_int(&tmp[0], &option);
|
|
|
|
if (ret || (option != 0 && option != 1))
|
|
|
|
goto hostdata_error;
|
|
|
|
ls->ls_first = option;
|
|
|
|
break;
|
|
|
|
case Opt_err:
|
|
|
|
default:
|
|
|
|
hostdata_error:
|
|
|
|
fs_info(sdp, "unknown hostdata (%s)\n", o);
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
2008-01-30 15:34:04 +00:00
|
|
|
}
|
|
|
|
|
2009-01-12 10:43:39 +00:00
|
|
|
if (lm->lm_mount == NULL) {
|
|
|
|
fs_info(sdp, "Now mounting FS...\n");
|
2011-07-26 08:15:45 +00:00
|
|
|
complete_all(&sdp->sd_locking_init);
|
2009-01-12 10:43:39 +00:00
|
|
|
return 0;
|
2008-01-30 15:34:04 +00:00
|
|
|
}
|
2012-01-09 22:18:05 +00:00
|
|
|
ret = lm->lm_mount(sdp, table);
|
2009-01-12 10:43:39 +00:00
|
|
|
if (ret == 0)
|
|
|
|
fs_info(sdp, "Joined cluster. Now mounting FS...\n");
|
2011-07-26 08:15:45 +00:00
|
|
|
complete_all(&sdp->sd_locking_init);
|
2009-01-12 10:43:39 +00:00
|
|
|
return ret;
|
2008-01-30 15:34:04 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
void gfs2_lm_unmount(struct gfs2_sbd *sdp)
|
|
|
|
{
|
2009-01-12 10:43:39 +00:00
|
|
|
const struct lm_lockops *lm = sdp->sd_lockstruct.ls_ops;
|
|
|
|
if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) &&
|
|
|
|
lm->lm_unmount)
|
|
|
|
lm->lm_unmount(sdp);
|
2008-01-30 15:34:04 +00:00
|
|
|
}
|
2006-01-16 16:50:04 +00:00
|
|
|
|
2010-06-14 09:01:30 +00:00
|
|
|
static int wait_on_journal(struct gfs2_sbd *sdp)
|
|
|
|
{
|
|
|
|
if (sdp->sd_lockstruct.ls_ops->lm_mount == NULL)
|
|
|
|
return 0;
|
|
|
|
|
sched: Remove proliferation of wait_on_bit() action functions
The current "wait_on_bit" interface requires an 'action'
function to be provided which does the actual waiting.
There are over 20 such functions, many of them identical.
Most cases can be satisfied by one of just two functions, one
which uses io_schedule() and one which just uses schedule().
So:
Rename wait_on_bit and wait_on_bit_lock to
wait_on_bit_action and wait_on_bit_lock_action
to make it explicit that they need an action function.
Introduce new wait_on_bit{,_lock} and wait_on_bit{,_lock}_io
which are *not* given an action function but implicitly use
a standard one.
The decision to error-out if a signal is pending is now made
based on the 'mode' argument rather than being encoded in the action
function.
All instances of the old wait_on_bit and wait_on_bit_lock which
can use the new version have been changed accordingly and their
action functions have been discarded.
wait_on_bit{_lock} does not return any specific error code in the
event of a signal so the caller must check for non-zero and
interpolate their own error code as appropriate.
The wait_on_bit() call in __fscache_wait_on_invalidate() was
ambiguous as it specified TASK_UNINTERRUPTIBLE but used
fscache_wait_bit_interruptible as an action function.
David Howells confirms this should be uniformly
"uninterruptible"
The main remaining user of wait_on_bit{,_lock}_action is NFS
which needs to use a freezer-aware schedule() call.
A comment in fs/gfs2/glock.c notes that having multiple 'action'
functions is useful as they display differently in the 'wchan'
field of 'ps'. (and /proc/$PID/wchan).
As the new bit_wait{,_io} functions are tagged "__sched", they
will not show up at all, but something higher in the stack. So
the distinction will still be visible, only with different
function names (gds2_glock_wait versus gfs2_glock_dq_wait in the
gfs2/glock.c case).
Since first version of this patch (against 3.15) two new action
functions appeared, on in NFS and one in CIFS. CIFS also now
uses an action function that makes the same freezer aware
schedule call as NFS.
Signed-off-by: NeilBrown <neilb@suse.de>
Acked-by: David Howells <dhowells@redhat.com> (fscache, keys)
Acked-by: Steven Whitehouse <swhiteho@redhat.com> (gfs2)
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Steve French <sfrench@samba.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/20140707051603.28027.72349.stgit@notabene.brown
Signed-off-by: Ingo Molnar <mingo@kernel.org>
2014-07-07 05:16:04 +00:00
|
|
|
return wait_on_bit(&sdp->sd_flags, SDF_NOJOURNALID, TASK_INTERRUPTIBLE)
|
|
|
|
? -EINTR : 0;
|
2010-06-14 09:01:30 +00:00
|
|
|
}
|
|
|
|
|
2009-07-31 10:07:29 +00:00
|
|
|
void gfs2_online_uevent(struct gfs2_sbd *sdp)
|
|
|
|
{
|
|
|
|
struct super_block *sb = sdp->sd_vfs;
|
|
|
|
char ro[20];
|
|
|
|
char spectator[20];
|
|
|
|
char *envp[] = { ro, spectator, NULL };
|
|
|
|
sprintf(ro, "RDONLY=%d", (sb->s_flags & MS_RDONLY) ? 1 : 0);
|
|
|
|
sprintf(spectator, "SPECTATOR=%d", sdp->sd_args.ar_spectator ? 1 : 0);
|
|
|
|
kobject_uevent_env(&sdp->sd_kobj, KOBJ_ONLINE, envp);
|
|
|
|
}
|
|
|
|
|
2006-01-16 16:50:04 +00:00
|
|
|
/**
|
|
|
|
* fill_super - Read in superblock
|
|
|
|
* @sb: The VFS superblock
|
|
|
|
* @data: Mount options
|
|
|
|
* @silent: Don't complain if it's not a GFS2 filesystem
|
|
|
|
*
|
|
|
|
* Returns: errno
|
|
|
|
*/
|
|
|
|
|
2009-09-28 09:30:49 +00:00
|
|
|
static int fill_super(struct super_block *sb, struct gfs2_args *args, int silent)
|
2006-01-16 16:50:04 +00:00
|
|
|
{
|
|
|
|
struct gfs2_sbd *sdp;
|
|
|
|
struct gfs2_holder mount_gh;
|
|
|
|
int error;
|
|
|
|
|
|
|
|
sdp = init_sbd(sb);
|
|
|
|
if (!sdp) {
|
2014-03-06 20:10:45 +00:00
|
|
|
pr_warn("can't alloc struct gfs2_sbd\n");
|
2006-01-16 16:50:04 +00:00
|
|
|
return -ENOMEM;
|
|
|
|
}
|
2009-09-28 09:30:49 +00:00
|
|
|
sdp->sd_args = *args;
|
2006-01-16 16:50:04 +00:00
|
|
|
|
2009-05-19 09:01:18 +00:00
|
|
|
if (sdp->sd_args.ar_spectator) {
|
2009-01-06 11:52:25 +00:00
|
|
|
sb->s_flags |= MS_RDONLY;
|
2012-01-09 19:40:06 +00:00
|
|
|
set_bit(SDF_RORECOVERY, &sdp->sd_flags);
|
2009-05-19 09:01:18 +00:00
|
|
|
}
|
2009-01-06 11:52:25 +00:00
|
|
|
if (sdp->sd_args.ar_posix_acl)
|
|
|
|
sb->s_flags |= MS_POSIXACL;
|
2009-10-30 07:03:27 +00:00
|
|
|
if (sdp->sd_args.ar_nobarrier)
|
|
|
|
set_bit(SDF_NOBARRIERS, &sdp->sd_flags);
|
2009-01-06 11:52:25 +00:00
|
|
|
|
2011-06-16 13:06:55 +00:00
|
|
|
sb->s_flags |= MS_NOSEC;
|
2008-09-18 12:53:59 +00:00
|
|
|
sb->s_magic = GFS2_MAGIC;
|
|
|
|
sb->s_op = &gfs2_super_ops;
|
2010-12-18 17:06:56 +00:00
|
|
|
sb->s_d_op = &gfs2_dops;
|
2008-09-18 12:53:59 +00:00
|
|
|
sb->s_export_op = &gfs2_export_ops;
|
2009-08-26 17:41:32 +00:00
|
|
|
sb->s_xattr = gfs2_xattr_handlers;
|
2009-09-15 08:59:02 +00:00
|
|
|
sb->s_qcop = &gfs2_quotactl_ops;
|
|
|
|
sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
|
2008-09-18 12:53:59 +00:00
|
|
|
sb->s_time_gran = 1;
|
|
|
|
sb->s_maxbytes = MAX_LFS_FILESIZE;
|
2006-03-02 21:33:41 +00:00
|
|
|
|
|
|
|
/* Set up the buffer cache and fill in some fake block size values
|
|
|
|
to allow us to read-in the on-disk superblock. */
|
|
|
|
sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
|
|
|
|
sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
|
|
|
|
sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
|
|
|
|
GFS2_BASIC_BLOCK_SHIFT;
|
|
|
|
sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
|
2006-01-16 16:50:04 +00:00
|
|
|
|
GFS2: Various gfs2_logd improvements
This patch contains various tweaks to how log flushes and active item writeback
work. gfs2_logd is now managed by a waitqueue, and gfs2_log_reseve now waits
for gfs2_logd to do the log flushing. Multiple functions were rewritten to
remove the need to call gfs2_log_lock(). Instead of using one test to see if
gfs2_logd had work to do, there are now seperate tests to check if there
are two many buffers in the incore log or if there are two many items on the
active items list.
This patch is a port of a patch Steve Whitehouse wrote about a year ago, with
some minor changes. Since gfs2_ail1_start always submits all the active items,
it no longer needs to keep track of the first ai submitted, so this has been
removed. In gfs2_log_reserve(), the order of the calls to
prepare_to_wait_exclusive() and wake_up() when firing off the logd thread has
been switched. If it called wake_up first there was a small window for a race,
where logd could run and return before gfs2_log_reserve was ready to get woken
up. If gfs2_logd ran, but did not free up enough blocks, gfs2_log_reserve()
would be left waiting for gfs2_logd to eventualy run because it timed out.
Finally, gt_logd_secs, which controls how long to wait before gfs2_logd times
out, and flushes the log, can now be set on mount with ar_commit.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2010-05-04 19:29:16 +00:00
|
|
|
sdp->sd_tune.gt_logd_secs = sdp->sd_args.ar_commit;
|
2009-10-20 07:39:44 +00:00
|
|
|
sdp->sd_tune.gt_quota_quantum = sdp->sd_args.ar_quota_quantum;
|
|
|
|
if (sdp->sd_args.ar_statfs_quantum) {
|
|
|
|
sdp->sd_tune.gt_statfs_slow = 0;
|
|
|
|
sdp->sd_tune.gt_statfs_quantum = sdp->sd_args.ar_statfs_quantum;
|
2011-05-10 14:01:59 +00:00
|
|
|
} else {
|
2009-10-20 07:39:44 +00:00
|
|
|
sdp->sd_tune.gt_statfs_slow = 1;
|
|
|
|
sdp->sd_tune.gt_statfs_quantum = 30;
|
|
|
|
}
|
2009-05-13 13:49:48 +00:00
|
|
|
|
2006-01-16 16:50:04 +00:00
|
|
|
error = init_names(sdp, silent);
|
2012-06-13 14:27:41 +00:00
|
|
|
if (error) {
|
|
|
|
/* In this case, we haven't initialized sysfs, so we have to
|
|
|
|
manually free the sdp. */
|
|
|
|
free_percpu(sdp->sd_lkstats);
|
|
|
|
kfree(sdp);
|
|
|
|
sb->s_fs_info = NULL;
|
|
|
|
return error;
|
|
|
|
}
|
2006-01-16 16:50:04 +00:00
|
|
|
|
2012-01-09 22:18:05 +00:00
|
|
|
snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s", sdp->sd_table_name);
|
|
|
|
|
2006-01-16 16:50:04 +00:00
|
|
|
error = gfs2_sys_fs_add(sdp);
|
2012-06-13 14:27:41 +00:00
|
|
|
/*
|
|
|
|
* If we hit an error here, gfs2_sys_fs_add will have called function
|
|
|
|
* kobject_put which causes the sysfs usage count to go to zero, which
|
|
|
|
* causes sysfs to call function gfs2_sbd_release, which frees sdp.
|
|
|
|
* Subsequent error paths here will call gfs2_sys_fs_del, which also
|
|
|
|
* kobject_put to free sdp.
|
|
|
|
*/
|
2006-01-16 16:50:04 +00:00
|
|
|
if (error)
|
2012-06-13 14:27:41 +00:00
|
|
|
return error;
|
|
|
|
|
|
|
|
gfs2_create_debugfs_file(sdp);
|
2006-01-16 16:50:04 +00:00
|
|
|
|
|
|
|
error = gfs2_lm_mount(sdp, silent);
|
|
|
|
if (error)
|
2012-06-13 14:27:41 +00:00
|
|
|
goto fail_debug;
|
2006-01-16 16:50:04 +00:00
|
|
|
|
|
|
|
error = init_locking(sdp, &mount_gh, DO);
|
|
|
|
if (error)
|
|
|
|
goto fail_lm;
|
|
|
|
|
2008-08-08 12:45:13 +00:00
|
|
|
error = init_sb(sdp, silent);
|
2006-01-16 16:50:04 +00:00
|
|
|
if (error)
|
|
|
|
goto fail_locking;
|
|
|
|
|
2010-06-14 09:01:30 +00:00
|
|
|
error = wait_on_journal(sdp);
|
|
|
|
if (error)
|
|
|
|
goto fail_sb;
|
|
|
|
|
2010-09-29 14:04:18 +00:00
|
|
|
/*
|
|
|
|
* If user space has failed to join the cluster or some similar
|
|
|
|
* failure has occurred, then the journal id will contain a
|
|
|
|
* negative (error) number. This will then be returned to the
|
|
|
|
* caller (of the mount syscall). We do this even for spectator
|
|
|
|
* mounts (which just write a jid of 0 to indicate "ok" even though
|
|
|
|
* the jid is unused in the spectator case)
|
|
|
|
*/
|
|
|
|
if (sdp->sd_lockstruct.ls_jid < 0) {
|
|
|
|
error = sdp->sd_lockstruct.ls_jid;
|
|
|
|
sdp->sd_lockstruct.ls_jid = 0;
|
|
|
|
goto fail_sb;
|
|
|
|
}
|
|
|
|
|
2012-01-09 22:18:05 +00:00
|
|
|
if (sdp->sd_args.ar_spectator)
|
|
|
|
snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s",
|
|
|
|
sdp->sd_table_name);
|
|
|
|
else
|
|
|
|
snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u",
|
|
|
|
sdp->sd_table_name, sdp->sd_lockstruct.ls_jid);
|
|
|
|
|
2006-01-16 16:50:04 +00:00
|
|
|
error = init_inodes(sdp, DO);
|
|
|
|
if (error)
|
2006-01-30 18:34:10 +00:00
|
|
|
goto fail_sb;
|
2006-01-16 16:50:04 +00:00
|
|
|
|
|
|
|
error = init_per_node(sdp, DO);
|
|
|
|
if (error)
|
|
|
|
goto fail_inodes;
|
|
|
|
|
|
|
|
error = gfs2_statfs_init(sdp);
|
|
|
|
if (error) {
|
|
|
|
fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
|
|
|
|
goto fail_per_node;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!(sb->s_flags & MS_RDONLY)) {
|
|
|
|
error = gfs2_make_fs_rw(sdp);
|
|
|
|
if (error) {
|
|
|
|
fs_err(sdp, "can't make FS RW: %d\n", error);
|
2013-12-12 11:34:09 +00:00
|
|
|
goto fail_per_node;
|
2006-01-16 16:50:04 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
gfs2_glock_dq_uninit(&mount_gh);
|
2009-07-31 10:07:29 +00:00
|
|
|
gfs2_online_uevent(sdp);
|
2006-01-16 16:50:04 +00:00
|
|
|
return 0;
|
|
|
|
|
2006-09-04 16:04:26 +00:00
|
|
|
fail_per_node:
|
2006-01-16 16:50:04 +00:00
|
|
|
init_per_node(sdp, UNDO);
|
2006-09-04 16:04:26 +00:00
|
|
|
fail_inodes:
|
2006-01-16 16:50:04 +00:00
|
|
|
init_inodes(sdp, UNDO);
|
2006-09-04 16:04:26 +00:00
|
|
|
fail_sb:
|
2008-08-08 12:45:13 +00:00
|
|
|
if (sdp->sd_root_dir)
|
|
|
|
dput(sdp->sd_root_dir);
|
|
|
|
if (sdp->sd_master_dir)
|
|
|
|
dput(sdp->sd_master_dir);
|
2009-01-20 16:39:23 +00:00
|
|
|
if (sb->s_root)
|
|
|
|
dput(sb->s_root);
|
2008-08-08 12:45:13 +00:00
|
|
|
sb->s_root = NULL;
|
2006-09-04 16:04:26 +00:00
|
|
|
fail_locking:
|
2006-01-16 16:50:04 +00:00
|
|
|
init_locking(sdp, &mount_gh, UNDO);
|
2006-09-04 16:04:26 +00:00
|
|
|
fail_lm:
|
2014-06-02 13:40:25 +00:00
|
|
|
complete_all(&sdp->sd_journal_ready);
|
2008-12-19 15:32:06 +00:00
|
|
|
gfs2_gl_hash_clear(sdp);
|
2006-01-16 16:50:04 +00:00
|
|
|
gfs2_lm_unmount(sdp);
|
2012-06-13 14:27:41 +00:00
|
|
|
fail_debug:
|
2007-04-18 16:41:11 +00:00
|
|
|
gfs2_delete_debugfs_file(sdp);
|
GFS2: glock statistics gathering
The stats are divided into two sets: those relating to the
super block and those relating to an individual glock. The
super block stats are done on a per cpu basis in order to
try and reduce the overhead of gathering them. They are also
further divided by glock type.
In the case of both the super block and glock statistics,
the same information is gathered in each case. The super
block statistics are used to provide default values for
most of the glock statistics, so that newly created glocks
should have, as far as possible, a sensible starting point.
The statistics are divided into three pairs of mean and
variance, plus two counters. The mean/variance pairs are
smoothed exponential estimates and the algorithm used is
one which will be very familiar to those used to calculation
of round trip times in network code.
The three pairs of mean/variance measure the following
things:
1. DLM lock time (non-blocking requests)
2. DLM lock time (blocking requests)
3. Inter-request time (again to the DLM)
A non-blocking request is one which will complete right
away, whatever the state of the DLM lock in question. That
currently means any requests when (a) the current state of
the lock is exclusive (b) the requested state is either null
or unlocked or (c) the "try lock" flag is set. A blocking
request covers all the other lock requests.
There are two counters. The first is there primarily to show
how many lock requests have been made, and thus how much data
has gone into the mean/variance calculations. The other counter
is counting queueing of holders at the top layer of the glock
code. Hopefully that number will be a lot larger than the number
of dlm lock requests issued.
So why gather these statistics? There are several reasons
we'd like to get a better idea of these timings:
1. To be able to better set the glock "min hold time"
2. To spot performance issues more easily
3. To improve the algorithm for selecting resource groups for
allocation (to base it on lock wait time, rather than blindly
using a "try lock")
Due to the smoothing action of the updates, a step change in
some input quantity being sampled will only fully be taken
into account after 8 samples (or 4 for the variance) and this
needs to be carefully considered when interpreting the
results.
Knowing both the time it takes a lock request to complete and
the average time between lock requests for a glock means we
can compute the total percentage of the time for which the
node is able to use a glock vs. time that the rest of the
cluster has its share. That will be very useful when setting
the lock min hold time.
The other point to remember is that all times are in
nanoseconds. Great care has been taken to ensure that we
measure exactly the quantities that we want, as accurately
as possible. There are always inaccuracies in any
measuring system, but I hope this is as accurate as we
can reasonably make it.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2012-01-20 10:38:36 +00:00
|
|
|
free_percpu(sdp->sd_lkstats);
|
2012-06-13 14:27:41 +00:00
|
|
|
/* gfs2_sys_fs_del must be the last thing we do, since it causes
|
|
|
|
* sysfs to call function gfs2_sbd_release, which frees sdp. */
|
|
|
|
gfs2_sys_fs_del(sdp);
|
2006-02-27 22:23:27 +00:00
|
|
|
sb->s_fs_info = NULL;
|
2006-01-16 16:50:04 +00:00
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2009-09-28 09:30:49 +00:00
|
|
|
static int set_gfs2_super(struct super_block *s, void *data)
|
2006-01-16 16:50:04 +00:00
|
|
|
{
|
2009-09-28 09:30:49 +00:00
|
|
|
s->s_bdev = data;
|
|
|
|
s->s_dev = s->s_bdev->bd_dev;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We set the bdi here to the queue backing, file systems can
|
|
|
|
* overwrite this in ->fill_super()
|
|
|
|
*/
|
|
|
|
s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info;
|
|
|
|
return 0;
|
2006-08-25 16:13:37 +00:00
|
|
|
}
|
|
|
|
|
2009-09-28 09:30:49 +00:00
|
|
|
static int test_gfs2_super(struct super_block *s, void *ptr)
|
2009-06-05 06:18:57 +00:00
|
|
|
{
|
|
|
|
struct block_device *bdev = ptr;
|
|
|
|
return (bdev == s->s_bdev);
|
|
|
|
}
|
|
|
|
|
2009-09-28 09:30:49 +00:00
|
|
|
/**
|
2010-07-26 12:03:58 +00:00
|
|
|
* gfs2_mount - Get the GFS2 superblock
|
2009-09-28 09:30:49 +00:00
|
|
|
* @fs_type: The GFS2 filesystem type
|
|
|
|
* @flags: Mount flags
|
|
|
|
* @dev_name: The name of the device
|
|
|
|
* @data: The mount arguments
|
|
|
|
*
|
|
|
|
* Q. Why not use get_sb_bdev() ?
|
|
|
|
* A. We need to select one of two root directories to mount, independent
|
|
|
|
* of whether this is the initial, or subsequent, mount of this sb
|
|
|
|
*
|
|
|
|
* Returns: 0 or -ve on error
|
|
|
|
*/
|
|
|
|
|
2010-07-26 12:03:58 +00:00
|
|
|
static struct dentry *gfs2_mount(struct file_system_type *fs_type, int flags,
|
|
|
|
const char *dev_name, void *data)
|
2009-09-28 09:30:49 +00:00
|
|
|
{
|
|
|
|
struct block_device *bdev;
|
|
|
|
struct super_block *s;
|
2010-11-13 10:55:18 +00:00
|
|
|
fmode_t mode = FMODE_READ | FMODE_EXCL;
|
2009-09-28 09:30:49 +00:00
|
|
|
int error;
|
|
|
|
struct gfs2_args args;
|
|
|
|
struct gfs2_sbd *sdp;
|
|
|
|
|
|
|
|
if (!(flags & MS_RDONLY))
|
|
|
|
mode |= FMODE_WRITE;
|
|
|
|
|
2010-11-13 10:55:18 +00:00
|
|
|
bdev = blkdev_get_by_path(dev_name, mode, fs_type);
|
2009-09-28 09:30:49 +00:00
|
|
|
if (IS_ERR(bdev))
|
2010-07-26 12:03:58 +00:00
|
|
|
return ERR_CAST(bdev);
|
2009-09-28 09:30:49 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* once the super is inserted into the list by sget, s_umount
|
|
|
|
* will protect the lockfs code from trying to start a snapshot
|
|
|
|
* while we are mounting
|
|
|
|
*/
|
|
|
|
mutex_lock(&bdev->bd_fsfreeze_mutex);
|
|
|
|
if (bdev->bd_fsfreeze_count > 0) {
|
|
|
|
mutex_unlock(&bdev->bd_fsfreeze_mutex);
|
|
|
|
error = -EBUSY;
|
|
|
|
goto error_bdev;
|
|
|
|
}
|
2012-06-25 11:55:37 +00:00
|
|
|
s = sget(fs_type, test_gfs2_super, set_gfs2_super, flags, bdev);
|
2009-09-28 09:30:49 +00:00
|
|
|
mutex_unlock(&bdev->bd_fsfreeze_mutex);
|
|
|
|
error = PTR_ERR(s);
|
|
|
|
if (IS_ERR(s))
|
|
|
|
goto error_bdev;
|
|
|
|
|
2013-12-06 11:52:34 +00:00
|
|
|
if (s->s_root) {
|
|
|
|
/*
|
|
|
|
* s_umount nests inside bd_mutex during
|
|
|
|
* __invalidate_device(). blkdev_put() acquires
|
|
|
|
* bd_mutex and can't be called under s_umount. Drop
|
|
|
|
* s_umount temporarily. This is safe as we're
|
|
|
|
* holding an active reference.
|
|
|
|
*/
|
|
|
|
up_write(&s->s_umount);
|
2010-11-13 10:55:18 +00:00
|
|
|
blkdev_put(bdev, mode);
|
2013-12-06 11:52:34 +00:00
|
|
|
down_write(&s->s_umount);
|
|
|
|
}
|
2010-07-26 12:03:58 +00:00
|
|
|
|
2009-09-28 09:30:49 +00:00
|
|
|
memset(&args, 0, sizeof(args));
|
|
|
|
args.ar_quota = GFS2_QUOTA_DEFAULT;
|
|
|
|
args.ar_data = GFS2_DATA_DEFAULT;
|
GFS2: Various gfs2_logd improvements
This patch contains various tweaks to how log flushes and active item writeback
work. gfs2_logd is now managed by a waitqueue, and gfs2_log_reseve now waits
for gfs2_logd to do the log flushing. Multiple functions were rewritten to
remove the need to call gfs2_log_lock(). Instead of using one test to see if
gfs2_logd had work to do, there are now seperate tests to check if there
are two many buffers in the incore log or if there are two many items on the
active items list.
This patch is a port of a patch Steve Whitehouse wrote about a year ago, with
some minor changes. Since gfs2_ail1_start always submits all the active items,
it no longer needs to keep track of the first ai submitted, so this has been
removed. In gfs2_log_reserve(), the order of the calls to
prepare_to_wait_exclusive() and wake_up() when firing off the logd thread has
been switched. If it called wake_up first there was a small window for a race,
where logd could run and return before gfs2_log_reserve was ready to get woken
up. If gfs2_logd ran, but did not free up enough blocks, gfs2_log_reserve()
would be left waiting for gfs2_logd to eventualy run because it timed out.
Finally, gt_logd_secs, which controls how long to wait before gfs2_logd times
out, and flushes the log, can now be set on mount with ar_commit.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2010-05-04 19:29:16 +00:00
|
|
|
args.ar_commit = 30;
|
2009-10-20 07:39:44 +00:00
|
|
|
args.ar_statfs_quantum = 30;
|
|
|
|
args.ar_quota_quantum = 60;
|
2009-09-28 09:30:49 +00:00
|
|
|
args.ar_errors = GFS2_ERRORS_DEFAULT;
|
|
|
|
|
|
|
|
error = gfs2_mount_args(&args, data);
|
|
|
|
if (error) {
|
2014-03-06 20:10:45 +00:00
|
|
|
pr_warn("can't parse mount arguments\n");
|
2010-07-26 12:03:58 +00:00
|
|
|
goto error_super;
|
2009-09-28 09:30:49 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
if (s->s_root) {
|
|
|
|
error = -EBUSY;
|
|
|
|
if ((flags ^ s->s_flags) & MS_RDONLY)
|
|
|
|
goto error_super;
|
|
|
|
} else {
|
|
|
|
char b[BDEVNAME_SIZE];
|
|
|
|
|
|
|
|
s->s_mode = mode;
|
|
|
|
strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
|
|
|
|
sb_set_blocksize(s, block_size(bdev));
|
|
|
|
error = fill_super(s, &args, flags & MS_SILENT ? 1 : 0);
|
2010-07-26 12:03:58 +00:00
|
|
|
if (error)
|
|
|
|
goto error_super;
|
2009-09-28 09:30:49 +00:00
|
|
|
s->s_flags |= MS_ACTIVE;
|
|
|
|
bdev->bd_super = s;
|
|
|
|
}
|
|
|
|
|
|
|
|
sdp = s->s_fs_info;
|
|
|
|
if (args.ar_meta)
|
2010-07-26 12:03:58 +00:00
|
|
|
return dget(sdp->sd_master_dir);
|
2009-09-28 09:30:49 +00:00
|
|
|
else
|
2010-07-26 12:03:58 +00:00
|
|
|
return dget(sdp->sd_root_dir);
|
2009-09-28 09:30:49 +00:00
|
|
|
|
|
|
|
error_super:
|
|
|
|
deactivate_locked_super(s);
|
2010-07-26 12:03:58 +00:00
|
|
|
return ERR_PTR(error);
|
2009-09-28 09:30:49 +00:00
|
|
|
error_bdev:
|
2010-11-13 10:55:18 +00:00
|
|
|
blkdev_put(bdev, mode);
|
2010-07-26 12:03:58 +00:00
|
|
|
return ERR_PTR(error);
|
2009-09-28 09:30:49 +00:00
|
|
|
}
|
|
|
|
|
2009-06-05 06:18:57 +00:00
|
|
|
static int set_meta_super(struct super_block *s, void *ptr)
|
|
|
|
{
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
2010-07-26 12:03:58 +00:00
|
|
|
static struct dentry *gfs2_mount_meta(struct file_system_type *fs_type,
|
|
|
|
int flags, const char *dev_name, void *data)
|
2006-08-25 16:13:37 +00:00
|
|
|
{
|
2009-06-05 06:18:57 +00:00
|
|
|
struct super_block *s;
|
2009-06-10 09:31:45 +00:00
|
|
|
struct gfs2_sbd *sdp;
|
2009-04-03 01:17:03 +00:00
|
|
|
struct path path;
|
2006-08-25 16:13:37 +00:00
|
|
|
int error;
|
2006-09-25 13:26:04 +00:00
|
|
|
|
2009-04-03 01:17:03 +00:00
|
|
|
error = kern_path(dev_name, LOOKUP_FOLLOW, &path);
|
2006-08-25 16:13:37 +00:00
|
|
|
if (error) {
|
2014-03-06 20:10:45 +00:00
|
|
|
pr_warn("path_lookup on %s returned error %d\n",
|
|
|
|
dev_name, error);
|
2010-07-26 12:03:58 +00:00
|
|
|
return ERR_PTR(error);
|
2006-08-25 16:13:37 +00:00
|
|
|
}
|
2012-06-25 11:55:37 +00:00
|
|
|
s = sget(&gfs2_fs_type, test_gfs2_super, set_meta_super, flags,
|
2009-06-05 06:18:57 +00:00
|
|
|
path.dentry->d_inode->i_sb->s_bdev);
|
2009-04-03 01:17:03 +00:00
|
|
|
path_put(&path);
|
2009-06-05 06:18:57 +00:00
|
|
|
if (IS_ERR(s)) {
|
2014-03-06 20:10:45 +00:00
|
|
|
pr_warn("gfs2 mount does not exist\n");
|
2010-07-26 12:03:58 +00:00
|
|
|
return ERR_CAST(s);
|
2006-08-25 16:13:37 +00:00
|
|
|
}
|
2009-09-28 09:30:49 +00:00
|
|
|
if ((flags ^ s->s_flags) & MS_RDONLY) {
|
|
|
|
deactivate_locked_super(s);
|
2010-07-26 12:03:58 +00:00
|
|
|
return ERR_PTR(-EBUSY);
|
2009-09-28 09:30:49 +00:00
|
|
|
}
|
2009-06-05 06:18:57 +00:00
|
|
|
sdp = s->s_fs_info;
|
2010-07-26 12:03:58 +00:00
|
|
|
return dget(sdp->sd_master_dir);
|
2006-01-16 16:50:04 +00:00
|
|
|
}
|
|
|
|
|
2006-03-02 21:33:41 +00:00
|
|
|
static void gfs2_kill_sb(struct super_block *sb)
|
2006-08-25 16:13:37 +00:00
|
|
|
{
|
|
|
|
struct gfs2_sbd *sdp = sb->s_fs_info;
|
2008-12-19 15:43:05 +00:00
|
|
|
|
|
|
|
if (sdp == NULL) {
|
|
|
|
kill_block_super(sb);
|
|
|
|
return;
|
2008-11-27 08:27:28 +00:00
|
|
|
}
|
2008-12-19 15:43:05 +00:00
|
|
|
|
GFS2: remove transaction glock
GFS2 has a transaction glock, which must be grabbed for every
transaction, whose purpose is to deal with freezing the filesystem.
Aside from this involving a large amount of locking, it is very easy to
make the current fsfreeze code hang on unfreezing.
This patch rewrites how gfs2 handles freezing the filesystem. The
transaction glock is removed. In it's place is a freeze glock, which is
cached (but not held) in a shared state by every node in the cluster
when the filesystem is mounted. This lock only needs to be grabbed on
freezing, and actions which need to be safe from freezing, like
recovery.
When a node wants to freeze the filesystem, it grabs this glock
exclusively. When the freeze glock state changes on the nodes (either
from shared to unlocked, or shared to exclusive), the filesystem does a
special log flush. gfs2_log_flush() does all the work for flushing out
the and shutting down the incore log, and then it tries to grab the
freeze glock in a shared state again. Since the filesystem is stuck in
gfs2_log_flush, no new transaction can start, and nothing can be written
to disk. Unfreezing the filesytem simply involes dropping the freeze
glock, allowing gfs2_log_flush() to grab and then release the shared
lock, so it is cached for next time.
However, in order for the unfreezing ioctl to occur, gfs2 needs to get a
shared lock on the filesystem root directory inode to check permissions.
If that glock has already been grabbed exclusively, fsfreeze will be
unable to get the shared lock and unfreeze the filesystem.
In order to allow the unfreeze, this patch makes gfs2 grab a shared lock
on the filesystem root directory during the freeze, and hold it until it
unfreezes the filesystem. The functions which need to grab a shared
lock in order to allow the unfreeze ioctl to be issued now use the lock
grabbed by the freeze code instead.
The freeze and unfreeze code take care to make sure that this shared
lock will not be dropped while another process is using it.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2014-05-02 03:26:55 +00:00
|
|
|
gfs2_log_flush(sdp, NULL, SYNC_FLUSH);
|
2008-12-19 15:43:05 +00:00
|
|
|
dput(sdp->sd_root_dir);
|
|
|
|
dput(sdp->sd_master_dir);
|
|
|
|
sdp->sd_root_dir = NULL;
|
|
|
|
sdp->sd_master_dir = NULL;
|
2008-12-19 15:32:06 +00:00
|
|
|
shrink_dcache_sb(sb);
|
2008-12-19 15:43:05 +00:00
|
|
|
gfs2_delete_debugfs_file(sdp);
|
GFS2: glock statistics gathering
The stats are divided into two sets: those relating to the
super block and those relating to an individual glock. The
super block stats are done on a per cpu basis in order to
try and reduce the overhead of gathering them. They are also
further divided by glock type.
In the case of both the super block and glock statistics,
the same information is gathered in each case. The super
block statistics are used to provide default values for
most of the glock statistics, so that newly created glocks
should have, as far as possible, a sensible starting point.
The statistics are divided into three pairs of mean and
variance, plus two counters. The mean/variance pairs are
smoothed exponential estimates and the algorithm used is
one which will be very familiar to those used to calculation
of round trip times in network code.
The three pairs of mean/variance measure the following
things:
1. DLM lock time (non-blocking requests)
2. DLM lock time (blocking requests)
3. Inter-request time (again to the DLM)
A non-blocking request is one which will complete right
away, whatever the state of the DLM lock in question. That
currently means any requests when (a) the current state of
the lock is exclusive (b) the requested state is either null
or unlocked or (c) the "try lock" flag is set. A blocking
request covers all the other lock requests.
There are two counters. The first is there primarily to show
how many lock requests have been made, and thus how much data
has gone into the mean/variance calculations. The other counter
is counting queueing of holders at the top layer of the glock
code. Hopefully that number will be a lot larger than the number
of dlm lock requests issued.
So why gather these statistics? There are several reasons
we'd like to get a better idea of these timings:
1. To be able to better set the glock "min hold time"
2. To spot performance issues more easily
3. To improve the algorithm for selecting resource groups for
allocation (to base it on lock wait time, rather than blindly
using a "try lock")
Due to the smoothing action of the updates, a step change in
some input quantity being sampled will only fully be taken
into account after 8 samples (or 4 for the variance) and this
needs to be carefully considered when interpreting the
results.
Knowing both the time it takes a lock request to complete and
the average time between lock requests for a glock means we
can compute the total percentage of the time for which the
node is able to use a glock vs. time that the rest of the
cluster has its share. That will be very useful when setting
the lock min hold time.
The other point to remember is that all times are in
nanoseconds. Great care has been taken to ensure that we
measure exactly the quantities that we want, as accurately
as possible. There are always inaccuracies in any
measuring system, but I hope this is as accurate as we
can reasonably make it.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2012-01-20 10:38:36 +00:00
|
|
|
free_percpu(sdp->sd_lkstats);
|
2012-06-13 14:27:41 +00:00
|
|
|
kill_block_super(sb);
|
2006-08-25 16:13:37 +00:00
|
|
|
}
|
|
|
|
|
2006-01-16 16:50:04 +00:00
|
|
|
struct file_system_type gfs2_fs_type = {
|
|
|
|
.name = "gfs2",
|
|
|
|
.fs_flags = FS_REQUIRES_DEV,
|
2010-07-26 12:03:58 +00:00
|
|
|
.mount = gfs2_mount,
|
2006-03-02 21:33:41 +00:00
|
|
|
.kill_sb = gfs2_kill_sb,
|
|
|
|
.owner = THIS_MODULE,
|
|
|
|
};
|
2013-03-03 03:39:14 +00:00
|
|
|
MODULE_ALIAS_FS("gfs2");
|
2006-03-02 21:33:41 +00:00
|
|
|
|
|
|
|
struct file_system_type gfs2meta_fs_type = {
|
|
|
|
.name = "gfs2meta",
|
|
|
|
.fs_flags = FS_REQUIRES_DEV,
|
2010-07-26 12:03:58 +00:00
|
|
|
.mount = gfs2_mount_meta,
|
2006-01-16 16:50:04 +00:00
|
|
|
.owner = THIS_MODULE,
|
|
|
|
};
|
2013-03-03 03:39:14 +00:00
|
|
|
MODULE_ALIAS_FS("gfs2meta");
|