The only difference between how we handle data=ordered and
data=writeback is a single call to ext4_jbd2_file_inode(). Eliminate
code duplication by factoring out redundant the code paths.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Reviewed-by: Lukas Czerner <lczerner@redhat.com>
Commit 84c17543ab (ext4: move work from io_end to inode) triggered a
regression when running xfstest #270 when the file system is mounted
with dioread_nolock.
The problem is that after ext4_evict_inode() calls ext4_ioend_wait(),
this guarantees that last io_end structure has been freed, but it does
not guarantee that the workqueue structure, which was moved into the
inode by commit 84c17543ab, is actually finished. Once
ext4_flush_completed_IO() calls ext4_free_io_end() on CPU #1, this
will allow ext4_ioend_wait() to return on CPU #2, at which point the
evict_inode() codepath can race against the workqueue code on CPU #1
accessing EXT4_I(inode)->i_unwritten_work to find the next item of
work to do.
Fix this by calling cancel_work_sync() in ext4_ioend_wait(), which
will be renamed ext4_ioend_shutdown(), since it is only used by
ext4_evict_inode(). Also, move the call to ext4_ioend_shutdown()
until after truncate_inode_pages() and filemap_write_and_wait() are
called, to make sure all dirty pages have been written back and
flushed from the page cache first.
BUG: unable to handle kernel NULL pointer dereference at (null)
IP: [<c01dda6a>] cwq_activate_delayed_work+0x3b/0x7e
*pdpt = 0000000030bc3001 *pde = 0000000000000000
Oops: 0000 [#1] SMP DEBUG_PAGEALLOC
Modules linked in:
Pid: 6, comm: kworker/u:0 Not tainted 3.8.0-rc3-00013-g84c1754-dirty #91 Bochs Bochs
EIP: 0060:[<c01dda6a>] EFLAGS: 00010046 CPU: 0
EIP is at cwq_activate_delayed_work+0x3b/0x7e
EAX: 00000000 EBX: 00000000 ECX: f505fe54 EDX: 00000000
ESI: ed5b697c EDI: 00000006 EBP: f64b7e8c ESP: f64b7e84
DS: 007b ES: 007b FS: 00d8 GS: 0000 SS: 0068
CR0: 8005003b CR2: 00000000 CR3: 30bc2000 CR4: 000006f0
DR0: 00000000 DR1: 00000000 DR2: 00000000 DR3: 00000000
DR6: ffff0ff0 DR7: 00000400
Process kworker/u:0 (pid: 6, ti=f64b6000 task=f64b4160 task.ti=f64b6000)
Stack:
f505fe00 00000006 f64b7e9c c01de3d7 f6435540 00000003 f64b7efc c01def1d
f6435540 00000002 00000000 0000008a c16d0808 c040a10b c16d07d8 c16d08b0
f505fe00 c16d0780 00000000 00000000 ee153df4 c1ce4a30 c17d0e30 00000000
Call Trace:
[<c01de3d7>] cwq_dec_nr_in_flight+0x71/0xfb
[<c01def1d>] process_one_work+0x5d8/0x637
[<c040a10b>] ? ext4_end_bio+0x300/0x300
[<c01e3105>] worker_thread+0x249/0x3ef
[<c01ea317>] kthread+0xd8/0xeb
[<c01e2ebc>] ? manage_workers+0x4bb/0x4bb
[<c023a370>] ? trace_hardirqs_on+0x27/0x37
[<c0f1b4b7>] ret_from_kernel_thread+0x1b/0x28
[<c01ea23f>] ? __init_kthread_worker+0x71/0x71
Code: 01 83 15 ac ff 6c c1 00 31 db 89 c6 8b 00 a8 04 74 12 89 c3 30 db 83 05 b0 ff 6c c1 01 83 15 b4 ff 6c c1 00 89 f0 e8 42 ff ff ff <8b> 13 89 f0 83 05 b8 ff 6c c1
6c c1 00 31 c9 83
EIP: [<c01dda6a>] cwq_activate_delayed_work+0x3b/0x7e SS:ESP 0068:f64b7e84
CR2: 0000000000000000
---[ end trace a1923229da53d8a4 ]---
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: Jan Kara <jack@suse.cz>
A user who was using a 8TB+ file system and with a very large flexbg
size (> 65536) could cause the atomic_t used in the struct flex_groups
to overflow. This was detected by PaX security patchset:
http://forums.grsecurity.net/viewtopic.php?f=3&t=3289&p=12551#p12551
This bug was introduced in commit 9f24e4208f, so it's been around
since 2.6.30. :-(
Fix this by using an atomic64_t for struct orlav_stats's
free_clusters.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Reviewed-by: Lukas Czerner <lczerner@redhat.com>
Cc: stable@vger.kernel.org
Use a percpu counter rather than atomic types for shrinker accounting.
There's no need for ultimate accuracy in the shrinker, so this
should come a little more cheaply. The percpu struct is somewhat
large, but there was a big gap before the cache-aligned
s_es_lru_lock anyway, and it fits nicely in there.
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
When the system is under memory pressure, ext4_es_srhink() will get
called very often. So optimize returning the number of items in the
file system's extent status cache by keeping a per-filesystem count,
instead of calculating it each time by scanning all of the inodes in
the extent status cache.
Also rename the slab used for the extent status cache to be
"ext4_extent_status" so it's obviousl the slab in question is created
by ext4.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: Zheng Liu <gnehzuil.liu@gmail.com>
Although extent status is loaded on-demand, we also need to reclaim
extent from the tree when we are under a heavy memory pressure because
in some cases fragmented extent tree causes status tree costs too much
memory.
Here we maintain a lru list in super_block. When the extent status of
an inode is accessed and changed, this inode will be move to the tail
of the list. The inode will be dropped from this list when it is
cleared. In the inode, a counter is added to count the number of
cached objects in extent status tree. Here only written/unwritten/hole
extent is counted because delayed extent doesn't be reclaimed due to
fiemap, bigalloc and seek_data/hole need it. The counter will be
increased as a new extent is allocated, and it will be decreased as a
extent is freed.
In this commit we use normal shrinker framework to reclaim memory from
the status tree. ext4_es_reclaim_extents_count() traverses the lru list
to count the number of reclaimable extents. ext4_es_shrink() tries to
reclaim written/unwritten/hole extents from extent status tree. The
inode that has been shrunk is moved to the tail of lru list.
Signed-off-by: Zheng Liu <wenqing.lz@taobao.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: Jan kara <jack@suse.cz>
Single extent cache could be removed because we have extent status tree
as a extent cache, and it would be better.
Signed-off-by: Zheng Liu <wenqing.lz@taobao.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: Jan kara <jack@suse.cz>
After tracking all extent status, we already have a extent cache in
memory. Every time we want to lookup a block mapping, we can first
try to lookup it in extent status tree to avoid a potential disk I/O.
A new function called ext4_es_lookup_extent is defined to finish this
work. When we try to lookup a block mapping, we always call
ext4_map_blocks and/or ext4_da_map_blocks. So in these functions we
first try to lookup a block mapping in extent status tree.
A new flag EXT4_GET_BLOCKS_NO_PUT_HOLE is used in ext4_da_map_blocks
in order not to put a hole into extent status tree because this hole
will be converted to delayed extent in the tree immediately.
Signed-off-by: Zheng Liu <wenqing.lz@taobao.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: Jan kara <jack@suse.cz>
By recording the phycisal block and status, extent status tree is able
to track the status of every extents. When we call _map_blocks
functions to lookup an extent or create a new written/unwritten/delayed
extent, this extent will be inserted into extent status tree.
We don't load all extents from disk in alloc_inode() because it costs
too much memory, and if a file is opened and closed frequently it will
takes too much time to load all extent information. So currently when
we create/lookup an extent, this extent will be inserted into extent
status tree. Hence, the extent status tree may not comprehensively
contain all of the extents found in the file.
Here a condition we need to take care is that an extent might contains
unwritten and delayed status simultaneously because an extent is delayed
allocated and could be allocated by fallocate. At this time we need to
keep delayed status because later we need to update delayed reservation
space using it.
Signed-off-by: Zheng Liu <wenqing.lz@taobao.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: Jan kara <jack@suse.cz>
In ext4_{create,mknod,mkdir,symlink}(), don't start the journal handle
until the inode has been succesfully allocated. In order to do this,
we need to start the handle in the ext4_new_inode(). So create a new
variant of this function, ext4_new_inode_start_handle(), so the handle
can be created at the last possible minute, before we need to modify
the inode allocation bitmap block.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Operations which modify extended attributes may need extra journal
credits if inline data is used, since there is a chance that some
extended attributes may need to get pushed to an external attribute
block.
Changes to reflect this was made in xattr.c, but they were missed in
fs/ext4/acl.c. To fix this, abstract the calculation of the number of
credits needed for xattr operations to an inline function defined in
ext4_jbd2.h, and use it in acl.c and xattr.c.
Also move the function declarations used in inline.c from xattr.h
(where they are non-obviously hidden, and caused problems since
ext4_jbd2.h needs to use the function ext4_has_inline_data), and move
them to ext4.h.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Reviewed-by: Tao Ma <boyu.mt@taobao.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Move the jbd2 wrapper functions which start and stop handles out of
super.c, where they don't really logically belong, and into
ext4_jbd2.c.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
It does not make much sense to have struct work in ext4_io_end_t
because we always use it for only one ext4_io_end_t per inode (the
first one in the i_completed_io list). So just move the structure to
inode itself. This also allows for a small simplification in
processing io_end structures.
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Currently we sometimes used block_write_full_page() and sometimes
ext4_bio_write_page() for writeback (depending on mount options and call
path). Let's always use ext4_bio_write_page() to simplify things a bit.
Reviewed-by: Zheng Liu <wenqing.lz@taobao.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
This patch add supports for indirect file support punching hole. It
is almost the same as ext4_ext_punch_hole. First, we invalidate all
pages between this hole, and then we try to deallocate all blocks of
this hole.
A recursive function is used to handle deallocation of blocks. In
this function, it iterates over the entries in inode's i_blocks or
indirect blocks, and try to free the block for each one of them.
After applying this patch, xfstest #255 will not pass w/o extent because
indirect-based file doesn't support unwritten extents.
Signed-off-by: Zheng Liu <wenqing.lz@taobao.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
After we have finished extending the file system, we need to trigger a
the lazy inode table thread to zero out the inode tables.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
inline data, which allows small files or directories to be stored in
the in-inode extended attribute area. (This requires that the file
system use inodes which are at least 256 bytes or larger; 128 byte
inodes do not have any room for in-inode xattrs.)
The second new feature is SEEK_HOLE/SEEK_DATA support. This is
enabled by the extent status tree patches, and this infrastructure
will be used to further optimize ext4 in the future.
Beyond that, we have the usual collection of code cleanups and bug
fixes.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)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=aNOf
-----END PGP SIGNATURE-----
Merge tag 'ext4_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4
Pull ext4 update from Ted Ts'o:
"There are two major features for this merge window. The first is
inline data, which allows small files or directories to be stored in
the in-inode extended attribute area. (This requires that the file
system use inodes which are at least 256 bytes or larger; 128 byte
inodes do not have any room for in-inode xattrs.)
The second new feature is SEEK_HOLE/SEEK_DATA support. This is
enabled by the extent status tree patches, and this infrastructure
will be used to further optimize ext4 in the future.
Beyond that, we have the usual collection of code cleanups and bug
fixes."
* tag 'ext4_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4: (63 commits)
ext4: zero out inline data using memset() instead of empty_zero_page
ext4: ensure Inode flags consistency are checked at build time
ext4: Remove CONFIG_EXT4_FS_XATTR
ext4: remove unused variable from ext4_ext_in_cache()
ext4: remove redundant initialization in ext4_fill_super()
ext4: remove redundant code in ext4_alloc_inode()
ext4: use sync_inode_metadata() when syncing inode metadata
ext4: enable ext4 inline support
ext4: let fallocate handle inline data correctly
ext4: let ext4_truncate handle inline data correctly
ext4: evict inline data out if we need to strore xattr in inode
ext4: let fiemap work with inline data
ext4: let ext4_rename handle inline dir
ext4: let empty_dir handle inline dir
ext4: let ext4_delete_entry() handle inline data
ext4: make ext4_delete_entry generic
ext4: let ext4_find_entry handle inline data
ext4: create a new function search_dir
ext4: let ext4_readdir handle inline data
ext4: let add_dir_entry handle inline data properly
...
Flags being used by atomic operations in inode flags (e.g.
ext4_test_inode_flag(), should be consistent with that actually stored
in inodes, i.e.: EXT4_XXX_FL.
It ensures that this consistency is checked at build-time, not at
run-time.
Currently, the flags consistency are being checked at run-time, but,
there is no real reason to not do a build-time check instead of a
run-time check. The code is comparing macro defined values with enum
type variables, where both are constants, so, there is no problem in
comparing constants at build-time.
enum variables are treated as constants by the C compiler, according
to the C99 specs (see www.open-std.org/jtc1/sc22/wg14/www/docs/n1124.pdf
sec. 6.2.5, item 16), so, there is no real problem in comparing an
enumeration type at build time
Signed-off-by: Carlos Maiolino <cmaiolino@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Ted has sent out a RFC about removing this feature. Eric and Jan
confirmed that both RedHat and SUSE enable this feature in all their
product. David also said that "As far as I know, it's enabled in all
Android kernels that use ext4." So it seems OK for us.
And what's more, as inline data depends its implementation on xattr,
and to be frank, I don't run any test again inline data enabled while
xattr disabled. So I think we should add inline data and remove this
config option in the same release.
[ The savings if you disable CONFIG_EXT4_FS_XATTR is only 27k, which
isn't much in the grand scheme of things. Since no one seems to be
testing this configuration except for some automated compile farms, on
balance we are better removing this config option, and so that it is
effectively always enabled. -- tytso ]
Cc: David Brown <davidb@codeaurora.org>
Cc: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Tao Ma <boyu.mt@taobao.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Currently ext4_delete_entry() is used only for dir entry removing from
a dir block. So let us create a new function
ext4_generic_delete_entry and this function takes a entry_buf and a
buf_size so that it can be used for inline data.
Signed-off-by: Tao Ma <boyu.mt@taobao.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
search_dirblock is used to search a dir block, but the code is almost
the same for searching an inline dir.
So create a new fuction search_dir and let search_dirblock call it.
Signed-off-by: Tao Ma <boyu.mt@taobao.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
For "." and "..", we just call filldir by ourselves
instead of iterating the real dir entry.
Signed-off-by: Tao Ma <boyu.mt@taobao.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
This patch let add_dir_entry handle the inline data case. So the
dir is initialized as inline dir first and then we can try to add
some files to it, when the inline space can't hold all the entries,
a dir block will be created and the dir entry will be moved to it.
Also for an inlined dir, "." and ".." are removed and we only use
4 bytes to store the parent inode number. These 2 entries will be
added when we convert an inline dir to a block-based one.
[ Folded in patch from Dan Carpenter to remove an unused variable. ]
Signed-off-by: Tao Ma <boyu.mt@taobao.com>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
The old add_dirent_to_buf handles all the work related to the
work of adding dir entry to a dir block. Now we have inline data,
so create 2 new function __ext4_find_dest_de and __ext4_insert_dentry
that do the real work and let add_dirent_to_buf call them.
Signed-off-by: Tao Ma <boyu.mt@taobao.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
The __ext4_check_dir_entry() function() is used to check whether the
de is over the block boundary. Now with inline data, it could be
within the block boundary while exceeds the inode size. So check this
function to check the overflow more precisely.
Signed-off-by: Tao Ma <boyu.mt@taobao.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Currently, the initialization of dot and dotdot are encapsulated in
ext4_mkdir and also bond with dir_block. So create a new function
named ext4_init_new_dir and the initialization is moved to
ext4_init_dot_dotdot. Now it will called either in the normal non-inline
case(rec_len of ".." will cover the whole block) or when we converting an
inline dir to a block(rec len of ".." will be the real length). The start
of the next entry is also returned for inline dir usage.
Signed-off-by: Tao Ma <boyu.mt@taobao.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
For delayed allocation mode, we write to inline data if the file
is small enough. And in case of we write to some offset larger
than the inline size, the 1st page is dirtied, so that
ext4_da_writepages can handle the conversion. When the 1st page
is initialized with blocks, the inline part is removed.
Signed-off-by: Tao Ma <boyu.mt@taobao.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
For a normal write case (not journalled write, not delayed
allocation), we write to the inline if the file is small and convert
it to an extent based file when the write is larger than the max
inline size.
Signed-off-by: Tao Ma <boyu.mt@taobao.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Implement inline data with xattr.
Now we use "system.data" to store xattr, and the xattr will
be extended if the i_size is increased while we don't release
the space during truncate.
Signed-off-by: Tao Ma <boyu.mt@taobao.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Previously, ext4_extents.h was being included at the end of ext4.h,
which was bad for a number of reasons: (a) it was not being included
in the expected place, and (b) it caused the header to be included
multiple times. There were #ifdef's to prevent this from causing any
problems, but it still was unnecessary.
By moving the function declarations that were in ext4_extents.h to
ext4.h, which is standard practice for where the function declarations
for the rest of ext4.h can be found, we can remove ext4_extents.h from
being included in ext4.h at all, and then we can only include
ext4_extents.h where it is needed in ext4's source files.
It should be possible to move a few more things into ext4.h, and
further reduce the number of source files that need to #include
ext4_extents.h, but that's a cleanup for another day.
Reported-by: Sachin Kamat <sachin.kamat@linaro.org>
Reported-by: Wei Yongjun <weiyj.lk@gmail.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
"Whether" is misspelled in various comments across the tree; this
fixes them. No code changes.
Signed-off-by: Adam Buchbinder <adam.buchbinder@gmail.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
This patch adds two structures that supports extent status tree, extent_status
and ext4_es_tree. Currently extent_status is used to track a delay extent for
an inode, which record the start block and the length of the delay extent.
ext4_es_tree is used to store all extent_status for an inode in memory.
Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com>
Signed-off-by: Allison Henderson <achender@linux.vnet.ibm.com>
Signed-off-by: Zheng Liu <wenqing.lz@taobao.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
In mke2fs, we only checksum the whole bitmap block and it is right.
While in the kernel, we use EXT4_BLOCKS_PER_GROUP to indicate the
size of the checksumed bitmap which is wrong when we enable bigalloc.
The right size should be EXT4_CLUSTERS_PER_GROUP and this patch fixes
it.
Also as every caller of ext4_block_bitmap_csum_set and
ext4_block_bitmap_csum_verify pass in EXT4_BLOCKS_PER_GROUP(sb)/8,
we'd better removes this parameter and sets it in the function itself.
Signed-off-by: Tao Ma <boyu.mt@taobao.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Reviewed-by: Lukas Czerner <lczerner@redhat.com>
Cc: stable@vger.kernel.org
The function ext4_handle_dirty_super() was calculating the superblock
on the wrong block data. As a result, when the superblock is modified
while it is mounted (most commonly, when inodes are added or removed
from the orphan list), the superblock checksum would be wrong. We
didn't notice because the superblock *was* being correctly calculated
in ext4_commit_super(), and this would get called when the file system
was unmounted. So the problem only became obvious if the system
crashed while the file system was mounted.
Fix this by removing the poorly designed function signature for
ext4_superblock_csum_set(); if it only took a single argument, the
pointer to a struct superblock, the ambiguity which caused this
mistake would have been impossible.
Reported-by: George Spelvin <linux@horizon.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: stable@vger.kernel.org
BUG #1) All places where we call ext4_flush_completed_IO are broken
because buffered io and DIO/AIO goes through three stages
1) submitted io,
2) completed io (in i_completed_io_list) conversion pended
3) finished io (conversion done)
And by calling ext4_flush_completed_IO we will flush only
requests which were in (2) stage, which is wrong because:
1) punch_hole and truncate _must_ wait for all outstanding unwritten io
regardless to it's state.
2) fsync and nolock_dio_read should also wait because there is
a time window between end_page_writeback() and ext4_add_complete_io()
As result integrity fsync is broken in case of buffered write
to fallocated region:
fsync blkdev_completion
->filemap_write_and_wait_range
->ext4_end_bio
->end_page_writeback
<-- filemap_write_and_wait_range return
->ext4_flush_completed_IO
sees empty i_completed_io_list but pended
conversion still exist
->ext4_add_complete_io
BUG #2) Race window becomes wider due to the 'ext4: completed_io
locking cleanup V4' patch series
This patch make following changes:
1) ext4_flush_completed_io() now first try to flush completed io and when
wait for any outstanding unwritten io via ext4_unwritten_wait()
2) Rename function to more appropriate name.
3) Assert that all callers of ext4_flush_unwritten_io should hold i_mutex to
prevent endless wait
Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Reviewed-by: Jan Kara <jack@suse.cz>
Inode's block defrag and ext4_change_inode_journal_flag() may
affect nonlocked DIO reads result, so proper synchronization
required.
- Add missed inode_dio_wait() calls where appropriate
- Check inode state under extra i_dio_count reference.
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Current unwritten extent conversion state-machine is very fuzzy.
- For unknown reason it performs conversion under i_mutex. What for?
My diagnosis:
We already protect extent tree with i_data_sem, truncate and punch_hole
should wait for DIO, so the only data we have to protect is end_io->flags
modification, but only flush_completed_IO and end_io_work modified this
flags and we can serialize them via i_completed_io_lock.
Currently all these games with mutex_trylock result in the following deadlock
truncate: kworker:
ext4_setattr ext4_end_io_work
mutex_lock(i_mutex)
inode_dio_wait(inode) ->BLOCK
DEADLOCK<- mutex_trylock()
inode_dio_done()
#TEST_CASE1_BEGIN
MNT=/mnt_scrach
unlink $MNT/file
fallocate -l $((1024*1024*1024)) $MNT/file
aio-stress -I 100000 -O -s 100m -n -t 1 -c 10 -o 2 -o 3 $MNT/file
sleep 2
truncate -s 0 $MNT/file
#TEST_CASE1_END
Or use 286's xfstests https://github.com/dmonakhov/xfstests/blob/devel/286
This patch makes state machine simple and clean:
(1) xxx_end_io schedule final extent conversion simply by calling
ext4_add_complete_io(), which append it to ei->i_completed_io_list
NOTE1: because of (2A) work should be queued only if
->i_completed_io_list was empty, otherwise the work is scheduled already.
(2) ext4_flush_completed_IO is responsible for handling all pending
end_io from ei->i_completed_io_list
Flushing sequence consists of following stages:
A) LOCKED: Atomically drain completed_io_list to local_list
B) Perform extents conversion
C) LOCKED: move converted io's to to_free list for final deletion
This logic depends on context which we was called from.
D) Final end_io context destruction
NOTE1: i_mutex is no longer required because end_io->flags modification
is protected by ei->ext4_complete_io_lock
Full list of changes:
- Move all completion end_io related routines to page-io.c in order to improve
logic locality
- Move open coded logic from various xx_end_xx routines to ext4_add_complete_io()
- remove EXT4_IO_END_FSYNC
- Improve SMP scalability by removing useless i_mutex which does not
protect io->flags anymore.
- Reduce lock contention on i_completed_io_lock by optimizing list walk.
- Rename ext4_end_io_nolock to end4_end_io and make it static
- Check flush completion status to ext4_ext_punch_hole(). Because it is
not good idea to punch blocks from corrupted inode.
Changes since V3 (in request to Jan's comments):
Fall back to active flush_completed_IO() approach in order to prevent
performance issues with nolocked DIO reads.
Changes since V2:
Fix use-after-free caused by race truncate vs end_io_work
Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
AIO/DIO prefix is wrong because it account unwritten extents which
also may be scheduled from buffered write endio
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Generic inode has unused i_private pointer which may be used as cur_aio_dio
storage.
TODO: If cur_aio_dio will be passed as an argument to get_block_t this allow
to have concurent AIO_DIO requests.
Reviewed-by: Zheng Liu <wenqing.lz@taobao.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Previously we allocated the s_group_info array with enough space for
any future possible growth of the file system via online resize. This
is unfortunate because it wastes memory, and it doesn't work for the
meta_bg scheme, since there is no limit based on the number of
reserved gdt blocks. So add the code to grow the s_group_info array
as needed.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Previously, we allocated the s_flex_groups array to the maximum size
that the file system could be resized. There was two problems with
this approach. First, it wasted memory in the common case where the
file system was not resized. Secondly, once we start allowing online
resizing using the meta_bg scheme, there is no maximum size that the
file system can be resized. So instead, we need to grow the
s_flex_groups at inline resize time.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Currently in ext4 the length of zero-out chunk is set to 7 file system
blocks. But if an inode has uninitailized extents from using
fallocate to preallocate space, and the workload issues many random
writes, this can cause a fragmented extent tree that will
unnecessarily grow the extent tree.
So create a new sysfs tunable, extent_max_zeroout_kb, which controls
the maximum size where blocks will be zeroed out instead of creating a
new uninitialized extent. The default of this has been sent to 32kb.
CC: Zach Brown <zab@zabbo.net>
CC: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: Zheng Liu <wenqing.lz@taobao.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Very large directories can cause significant performance problems, or
perhaps even invoke the OOM killer, if the process is running in a
highly constrained memory environment (whether it is VM's with a small
amount of memory or in a small memory cgroup).
So it is useful, in cloud server/data center environments, to be able
to set a filesystem-wide cap on the maximum size of a directory, to
ensure that directories never get larger than a sane size. We do this
via a new mount option, max_dir_size_kb. If there is an attempt to
grow the directory larger than max_dir_size_kb, the system call will
return ENOSPC instead.
Google-Bug-Id: 6863013
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
The last user of ext4_mark_super_dirty() in ext4_file_open() is so
rare it can well be modifying the superblock properly by journalling
the change. Change it and get rid of ext4_mark_super_dirty() as it's
not needed anymore.
Artem: small amendments.
Artem: tested using xfstests for both journalled and non-journalled ext4.
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Tested-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
The ext4_checksum() inline function was using a dynamic array size,
which is not legal C. (It is a gcc extension).
Remove it.
Cc: "Darrick J. Wong" <djwong@us.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
This patch adds support for quotas as a first class feature in ext4;
which is to say, the quota files are stored in hidden inodes as file
system metadata, instead of as separate files visible in the file system
directory hierarchy.
It is based on the proposal at:
https://ext4.wiki.kernel.org/index.php/Design_For_1st_Class_Quota_in_Ext4
This patch introduces a new feature - EXT4_FEATURE_RO_COMPAT_QUOTA
which, when turned on, enables quota accounting at mount time
iteself. Also, the quota inodes are stored in two additional superblock
fields. Some changes introduced by this patch that should be pointed
out are:
1) Two new ext4-superblock fields - s_usr_quota_inum and
s_grp_quota_inum for storing the quota inodes in use.
2) Default quota inodes are: inode#3 for tracking userquota and inode#4
for tracking group quota. The superblock fields can be set to use
other inodes as well.
3) If the QUOTA feature and corresponding quota inodes are set in
superblock, the quota usage tracking is turned on at mount time. On
'quotaon' ioctl, the quota limits enforcement is turned
on. 'quotaoff' ioctl turns off only the limits enforcement in this
case.
4) When QUOTA feature is in use, the quota mount options 'quota',
'usrquota', 'grpquota' are ignored by the kernel.
5) mke2fs or tune2fs can be used to set the QUOTA feature and initialize
quota inodes. The default reserved inodes will not be visible to user
as regular files.
6) The quota-tools will need to be modified to support hidden quota
files on ext4. E2fsprogs will also include support for creating and
fixing quota files.
7) Support is only for the new V2 quota file format.
Tested-by: Jan Kara <jack@suse.cz>
Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: Johann Lombardi <johann@whamcloud.com>
Signed-off-by: Aditya Kali <adityakali@google.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
EXT4_GET_BLOCKS_NO_LOCK flag is added to indicate that we don't need
to acquire i_data_sem lock in ext4_map_blocks. Meanwhile, it changes
ext4_get_block() to not start a new journal because when we do a
overwrite dio, there is no any metadata that needs to be modified.
We define a new function called ext4_get_block_write_nolock, which is
used in dio overwrite nolock. In this function, it doesn't try to
acquire i_data_sem lock and doesn't start a new journal as it does a
lookup.
CC: Tao Ma <tm@tao.ma>
CC: Eric Sandeen <sandeen@redhat.com>
CC: Robin Dong <hao.bigrat@gmail.com>
Signed-off-by: Zheng Liu <wenqing.lz@taobao.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Commit f975d6bcc7 introduced bug which caused ext4_statfs() to
miscalculate the number of file system overhead blocks. This causes
the f_blocks field in the statfs structure to be larger than it should
be. This would in turn cause the "df" output to show the number of
data blocks in the file system and the number of data blocks used to
be larger than they should be.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: stable@kernel.org