Quoted from Jing Xia's report, there is a potential deadlock may happen
between kworker and checkpoint as below:
[T:writeback] [T:checkpoint]
- wb_writeback
- blk_start_plug
bio contains NodeA was plugged in writeback threads
- do_writepages -- sync write inodeB, inc wb_sync_req[DATA]
- f2fs_write_data_pages
- f2fs_write_single_data_page -- write last dirty page
- f2fs_do_write_data_page
- set_page_writeback -- clear page dirty flag and
PAGECACHE_TAG_DIRTY tag in radix tree
- f2fs_outplace_write_data
- f2fs_update_data_blkaddr
- f2fs_wait_on_page_writeback -- wait NodeA to writeback here
- inode_dec_dirty_pages
- writeback_sb_inodes
- writeback_single_inode
- do_writepages
- f2fs_write_data_pages -- skip writepages due to wb_sync_req[DATA]
- wbc->pages_skipped += get_dirty_pages() -- PAGECACHE_TAG_DIRTY is not set but get_dirty_pages() returns one
- requeue_inode -- requeue inode to wb->b_dirty queue due to non-zero.pages_skipped
- blk_finish_plug
Let's try to avoid deadlock condition by forcing unplugging previous bio via
blk_finish_plug(current->plug) once we'v skipped writeback in writepages()
due to valid sbi->wb_sync_req[DATA/NODE].
Fixes: 687de7f101 ("f2fs: avoid IO split due to mixed WB_SYNC_ALL and WB_SYNC_NONE")
Signed-off-by: Zhiguo Niu <zhiguo.niu@unisoc.com>
Signed-off-by: Jing Xia <jing.xia@unisoc.com>
Signed-off-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Lockdep uses lock class keys in its analysis. init_rwsem() instantiates
one lock class key with each init_rwsem() user as follows:
#define init_rwsem(sem) \
do { \
static struct lock_class_key __key; \
\
__init_rwsem((sem), #sem, &__key); \
} while (0)
Commit e4544b63a7 ("f2fs: move f2fs to use reader-unfair rwsems") reduced
the number of lock class keys from one per init_rwsem() user to one per
file in which init_f2fs_rwsem() is used. This causes the same lock class key
to be associated with multiple f2fs rwsems and also triggers a number of
false positive lockdep deadlock reports. Fix this by again instantiating one
lock class key with each init_f2fs_rwsem() caller.
Cc: Tim Murray <timmurray@google.com>
Reported-by: syzbot+0b9cadf5fc45a98a5083@syzkaller.appspotmail.com
Fixes: e4544b63a7 ("f2fs: move f2fs to use reader-unfair rwsems")
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch enables idmapped mounts for f2fs, since all dedicated helpers
for this functionality existsm, so, in this patch we just pass down the
user_namespace argument from the VFS methods to the relevant helpers.
Simple idmap example on f2fs image:
1. truncate -s 128M f2fs.img
2. mkfs.f2fs f2fs.img
3. mount f2fs.img /mnt/f2fs/
4. touch /mnt/f2fs/file
5. ls -ln /mnt/f2fs/
total 0
-rw-r--r-- 1 0 0 0 2月 4 13:17 file
6. ./mount-idmapped --map-mount b:0:1001:1 /mnt/f2fs/ /mnt/scratch_f2fs/
7. ls -ln /mnt/scratch_f2fs/
total 0
-rw-r--r-- 1 1001 1001 0 2月 4 13:17 file
Signed-off-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This adds a sysfs entry to call checkpoint during fsync() in order to avoid
long elapsed time to run roll-forward recovery when booting the device.
Default value doesn't enforce the limitation which is same as before.
Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Once F2FS_IPU_FORCE policy is enabled in some cases:
a) f2fs forces to use F2FS_IPU_FORCE in a small-sized volume
b) user sets F2FS_IPU_FORCE policy via sysfs
Then we may fail to defragment file due to IPU policy check, it doesn't
make sense, let's introduce a new IPU policy to allow OPU during file
defragmentation.
In small-sized volume, let's enable F2FS_IPU_HONOR_OPU_WRITE policy
by default.
Signed-off-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
In a fragmented image, entries in dnode block list may locate in
incontiguous physical block address space, however, in recovery flow,
we will always readahead BIO_MAX_VECS size blocks, so in such case,
current readahead policy is low efficient, let's adjust readahead
window size dynamically based on consecutiveness of dnode blocks.
Signed-off-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch exposes max_discard_request, min_discard_issue_time,
mid_discard_issue_time, and max_discard_issue_time in sysfs. This will
allow the user to fine tune discard operations.
Signed-off-by: Konstantin Vyshetsky <vkon@google.com>
Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch unifies parameters related to how often discard is issued and
how many requests go out at the same time by placing them in
discard_cmd_control. The move will allow the parameters to be modified
in the future without relying on hard-coded values.
Signed-off-by: Konstantin Vyshetsky <vkon@google.com>
Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
When calling setattr_prepare() to determine the validity of the attributes the
ia_{g,u}id fields contain the value that will be written to inode->i_{g,u}id.
When the {g,u}id attribute of the file isn't altered and the caller's fs{g,u}id
matches the current {g,u}id attribute the attribute change is allowed.
The value in ia_{g,u}id does already account for idmapped mounts and will have
taken the relevant idmapping into account. So in order to verify that the
{g,u}id attribute isn't changed we simple need to compare the ia_{g,u}id value
against the inode's i_{g,u}id value.
This only has any meaning for idmapped mounts as idmapping helpers are
idempotent without them. And for idmapped mounts this really only has a meaning
when circular idmappings are used, i.e. mappings where e.g. id 1000 is mapped
to id 1001 and id 1001 is mapped to id 1000. Such ciruclar mappings can e.g. be
useful when sharing the same home directory between multiple users at the same
time.
As an example consider a directory with two files: /source/file1 owned by
{g,u}id 1000 and /source/file2 owned by {g,u}id 1001. Assume we create an
idmapped mount at /target with an idmapping that maps files owned by {g,u}id
1000 to being owned by {g,u}id 1001 and files owned by {g,u}id 1001 to being
owned by {g,u}id 1000. In effect, the idmapped mount at /target switches the
ownership of /source/file1 and source/file2, i.e. /target/file1 will be owned
by {g,u}id 1001 and /target/file2 will be owned by {g,u}id 1000.
This means that a user with fs{g,u}id 1000 must be allowed to setattr
/target/file2 from {g,u}id 1000 to {g,u}id 1000. Similar, a user with fs{g,u}id
1001 must be allowed to setattr /target/file1 from {g,u}id 1001 to {g,u}id
1001. Conversely, a user with fs{g,u}id 1000 must fail to setattr /target/file1
from {g,u}id 1001 to {g,u}id 1000. And a user with fs{g,u}id 1001 must fail to
setattr /target/file2 from {g,u}id 1000 to {g,u}id 1000. Both cases must fail
with EPERM for non-capable callers.
Before this patch we could end up denying legitimate attribute changes and
allowing invalid attribute changes when circular mappings are used. To even get
into this situation the caller must've been privileged both to create that
mapping and to create that idmapped mount.
This hasn't been seen in the wild anywhere but came up when expanding the
testsuite during work on a series of hardening patches. All idmapped fstests
pass without any regressions and we add new tests to verify the behavior of
circular mappings.
Link: https://lore.kernel.org/r/20211109145713.1868404-1-brauner@kernel.org
Fixes: 2f221d6f7b ("attr: handle idmapped mounts")
Cc: Seth Forshee <seth.forshee@digitalocean.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: stable@vger.kernel.org
CC: linux-fsdevel@vger.kernel.org
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Seth Forshee <sforshee@digitalocean.com>
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
It needs to assign sbi->gc_mode with GC_IDLE_AT rather than GC_AT when
user tries to enable ATGC via gc_idle sysfs interface, fix it.
Fixes: 093749e296 ("f2fs: support age threshold based garbage collection")
Cc: Zhipeng Tan <tanzhipeng@hust.edu.cn>
Signed-off-by: Jicheng Shao <shaojicheng@hust.edu.cn>
Signed-off-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
f2fs rw_semaphores work better if writers can starve readers,
especially for the checkpoint thread, because writers are strictly
more important than reader threads. This prevents significant priority
inversion between low-priority readers that blocked while trying to
acquire the read lock and a second acquisition of the write lock that
might be blocking high priority work.
Signed-off-by: Tim Murray <timmurray@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
If the pinned file has a hole by partial truncation, application that has
the block map will be broken.
Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Compress page will invalidate in truncate block process too, so remove
redunant invalidate compress pages in f2fs_evict_inode.
Signed-off-by: Fengnan Chang <changfengnan@vivo.com>
Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
For compressed inode, in .{invalidate,release}page, we will call
f2fs_invalidate_compress_pages() to drop all compressed page cache of
current inode.
But we don't need to drop compressed page cache synchronously in
.invalidatepage, because, all trancation paths of compressed physical
block has been covered with f2fs_invalidate_compress_page().
And also we don't need to drop compressed page cache synchronously
in .releasepage, because, if there is out-of-memory, we can count
on page cache reclaim on sbi->compress_inode.
BTW, this patch may fix the issue reported below:
https://lore.kernel.org/linux-f2fs-devel/20211202092812.197647-1-changfengnan@vivo.com/T/#u
Signed-off-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
https://bugzilla.kernel.org/show_bug.cgi?id=204137
With below script, we will hit panic during new segment allocation:
DISK=bingo.img
MOUNT_DIR=/mnt/f2fs
dd if=/dev/zero of=$DISK bs=1M count=105
mkfs.f2fe -a 1 -o 19 -t 1 -z 1 -f -q $DISK
mount -t f2fs $DISK $MOUNT_DIR -o "noinline_dentry,flush_merge,noextent_cache,mode=lfs,io_bits=7,fsync_mode=strict"
for (( i = 0; i < 4096; i++ )); do
name=`head /dev/urandom | tr -dc A-Za-z0-9 | head -c 10`
mkdir $MOUNT_DIR/$name
done
umount $MOUNT_DIR
rm $DISK
--- Core dump ---
Call Trace:
allocate_segment_by_default+0x9d/0x100 [f2fs]
f2fs_allocate_data_block+0x3c0/0x5c0 [f2fs]
do_write_page+0x62/0x110 [f2fs]
f2fs_outplace_write_data+0x43/0xc0 [f2fs]
f2fs_do_write_data_page+0x386/0x560 [f2fs]
__write_data_page+0x706/0x850 [f2fs]
f2fs_write_cache_pages+0x267/0x6a0 [f2fs]
f2fs_write_data_pages+0x19c/0x2e0 [f2fs]
do_writepages+0x1c/0x70
__filemap_fdatawrite_range+0xaa/0xe0
filemap_fdatawrite+0x1f/0x30
f2fs_sync_dirty_inodes+0x74/0x1f0 [f2fs]
block_operations+0xdc/0x350 [f2fs]
f2fs_write_checkpoint+0x104/0x1150 [f2fs]
f2fs_sync_fs+0xa2/0x120 [f2fs]
f2fs_balance_fs_bg+0x33c/0x390 [f2fs]
f2fs_write_node_pages+0x4c/0x1f0 [f2fs]
do_writepages+0x1c/0x70
__writeback_single_inode+0x45/0x320
writeback_sb_inodes+0x273/0x5c0
wb_writeback+0xff/0x2e0
wb_workfn+0xa1/0x370
process_one_work+0x138/0x350
worker_thread+0x4d/0x3d0
kthread+0x109/0x140
ret_from_fork+0x25/0x30
The root cause here is, with IO alignment feature enables, in worst
case, we need F2FS_IO_SIZE() free blocks space for single one 4k write
due to IO alignment feature will fill dummy pages to make IO being
aligned.
So we will easily run out of free segments during non-inline directory's
data writeback, even in process of foreground GC.
In order to fix this issue, I just propose to reserve additional free
space for IO alignment feature to handle worst case of free space usage
ratio during FGGC.
Fixes: 0a595ebaaa ("f2fs: support IO alignment for DATA and NODE writes")
Signed-off-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Otherwise, nat_bit area may be persisted across boundary of CP area during
nat_bit rebuilding.
Fixes: 94c821fb28 ("f2fs: rebuild nat_bits during umount")
Signed-off-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
f2fs: support fault injection for f2fs_trylock_op()
This patch supports to inject fault into f2fs_trylock_op().
Usage:
a) echo 65536 > /sys/fs/f2fs/<dev>/inject_type or
b) mount -o fault_type=65536 <dev> <mountpoint>
Signed-off-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
As Wenqing Liu reported in bugzilla:
https://bugzilla.kernel.org/show_bug.cgi?id=215235
- Overview
page fault in f2fs_setxattr() when mount and operate on corrupted image
- Reproduce
tested on kernel 5.16-rc3, 5.15.X under root
1. unzip tmp7.zip
2. ./single.sh f2fs 7
Sometimes need to run the script several times
- Kernel dump
loop0: detected capacity change from 0 to 131072
F2FS-fs (loop0): Found nat_bits in checkpoint
F2FS-fs (loop0): Mounted with checkpoint version = 7548c2ee
BUG: unable to handle page fault for address: ffffe47bc7123f48
RIP: 0010:kfree+0x66/0x320
Call Trace:
__f2fs_setxattr+0x2aa/0xc00 [f2fs]
f2fs_setxattr+0xfa/0x480 [f2fs]
__f2fs_set_acl+0x19b/0x330 [f2fs]
__vfs_removexattr+0x52/0x70
__vfs_removexattr_locked+0xb1/0x140
vfs_removexattr+0x56/0x100
removexattr+0x57/0x80
path_removexattr+0xa3/0xc0
__x64_sys_removexattr+0x17/0x20
do_syscall_64+0x37/0xb0
entry_SYSCALL_64_after_hwframe+0x44/0xae
The root cause is in __f2fs_setxattr(), we missed to do sanity check on
last xattr entry, result in out-of-bound memory access during updating
inconsistent xattr data of target inode.
After the fix, it can detect such xattr inconsistency as below:
F2FS-fs (loop11): inode (7) has invalid last xattr entry, entry_size: 60676
F2FS-fs (loop11): inode (8) has corrupted xattr
F2FS-fs (loop11): inode (8) has corrupted xattr
F2FS-fs (loop11): inode (8) has invalid last xattr entry, entry_size: 47736
Cc: stable@vger.kernel.org
Reported-by: Wenqing Liu <wenqingliu0120@gmail.com>
Signed-off-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch tries to mitigate lock contention between f2fs_write_checkpoint and
f2fs_get_node_info along with nat_tree_lock.
The idea is, if checkpoint is currently running, other threads that try to grab
nat_tree_lock would be better to wait for checkpoint.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
There is a potential deadlock between writeback process and a process
performing write_begin() or write_cache_pages() while trying to write
same compress file, but not compressable, as below:
[Process A] - doing checkpoint
[Process B] [Process C]
f2fs_write_cache_pages()
- lock_page() [all pages in cluster, 0-31]
- f2fs_write_multi_pages()
- f2fs_write_raw_pages()
- f2fs_write_single_data_page()
- f2fs_do_write_data_page()
- return -EAGAIN [f2fs_trylock_op() failed]
- unlock_page(page) [e.g., page 0]
- generic_perform_write()
- f2fs_write_begin()
- f2fs_prepare_compress_overwrite()
- prepare_compress_overwrite()
- lock_page() [e.g., page 0]
- lock_page() [e.g., page 1]
- lock_page(page) [e.g., page 0]
Since there is no compress process, it is no longer necessary to hold
locks on every pages in cluster within f2fs_write_raw_pages().
This patch changes f2fs_write_raw_pages() to release all locks first
and then perform write same as the non-compress file in
f2fs_write_cache_pages().
Fixes: 4c8ff7095b ("f2fs: support data compression")
Signed-off-by: Hyeong-Jun Kim <hj514.kim@samsung.com>
Signed-off-by: Sungjong Seo <sj1557.seo@samsung.com>
Signed-off-by: Youngjin Gil <youngjin.gil@samsung.com>
Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Android OTA failed due to SBI_NEED_FSCK flag when pinning the file. Let's avoid
it since we can do in-place-updates.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Change-Id: I3fd33c984417c10b38e23de6cec017b03d588945
Added a new sysfs node called gc_urgent_high_remaining. The user can
set the trial count limit for GC urgent high mode with this value. If
GC thread gets to the limit, the mode will turn back to GC normal mode.
By default, the value is zero, which means there is no limit like before.
Signed-off-by: Daeho Jeong <daehojeong@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
In fuzzed image, SSA table may indicate that a data block belongs to
invalid node, which node ID is out-of-range (0, 1, 2 or max_nid), in
order to avoid migrating inconsistent data in such corrupted image,
let's do sanity check anyway before data block migration.
Cc: stable@vger.kernel.org
Signed-off-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
As report by Wenqing Liu in bugzilla:
https://bugzilla.kernel.org/show_bug.cgi?id=215231
If we enable CONFIG_F2FS_CHECK_FS config, and with fuzzed image attached
in above link, we will encounter panic when executing below script:
1. mkdir mnt
2. mount -t f2fs tmp1.img mnt
3. touch tmp
F2FS-fs (loop11): mismatched blkaddr 5765 (source_blkaddr 1) in seg 3
kernel BUG at fs/f2fs/gc.c:1042!
do_garbage_collect+0x90f/0xa80 [f2fs]
f2fs_gc+0x294/0x12a0 [f2fs]
f2fs_balance_fs+0x2c5/0x7d0 [f2fs]
f2fs_create+0x239/0xd90 [f2fs]
lookup_open+0x45e/0xa90
open_last_lookups+0x203/0x670
path_openat+0xae/0x490
do_filp_open+0xbc/0x160
do_sys_openat2+0x2f1/0x500
do_sys_open+0x5e/0xa0
__x64_sys_openat+0x28/0x40
Previously, f2fs tries to catch data inconcistency exception in between
SSA and SIT table during GC, however once the exception is caught, it will
call f2fs_bug_on to hang kernel, it's not needed, instead, let's set
SBI_NEED_FSCK flag and skip migrating current block.
Fixes: bbf9f7d90f ("f2fs: Fix indefinite loop in f2fs_gc()")
Signed-off-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
As report by Wenqing Liu in bugzilla:
https://bugzilla.kernel.org/show_bug.cgi?id=215231
- Overview
kernel NULL pointer dereference triggered in folio_mark_dirty() when mount and operate on a crafted f2fs image
- Reproduce
tested on kernel 5.16-rc3, 5.15.X under root
1. mkdir mnt
2. mount -t f2fs tmp1.img mnt
3. touch tmp
4. cp tmp mnt
F2FS-fs (loop0): sanity_check_inode: inode (ino=49) extent info [5942, 4294180864, 4] is incorrect, run fsck to fix
F2FS-fs (loop0): f2fs_check_nid_range: out-of-range nid=31340049, run fsck to fix.
BUG: kernel NULL pointer dereference, address: 0000000000000000
folio_mark_dirty+0x33/0x50
move_data_page+0x2dd/0x460 [f2fs]
do_garbage_collect+0xc18/0x16a0 [f2fs]
f2fs_gc+0x1d3/0xd90 [f2fs]
f2fs_balance_fs+0x13a/0x570 [f2fs]
f2fs_create+0x285/0x840 [f2fs]
path_openat+0xe6d/0x1040
do_filp_open+0xc5/0x140
do_sys_openat2+0x23a/0x310
do_sys_open+0x57/0x80
The root cause is for special file: e.g. character, block, fifo or socket file,
f2fs doesn't assign address space operations pointer array for mapping->a_ops field,
so, in a fuzzed image, SSA table indicates a data block belong to special file, when
f2fs tries to migrate that block, it causes NULL pointer access once move_data_page()
calls a_ops->set_dirty_page().
Cc: stable@vger.kernel.org
Reported-by: Wenqing Liu <wenqingliu0120@gmail.com>
Signed-off-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Previously, compressed page cache drop when clean page cache, but
POSIX_FADV_DONTNEED can't clean compressed page cache because raw page
don't have private data, and won't call f2fs_invalidate_compress_pages.
This commit call f2fs_invalidate_compress_pages() directly in
f2fs_file_fadvise() for POSIX_FADV_DONTNEED case.
Signed-off-by: Fengnan Chang <changfengnan@vivo.com>
Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Since compress inode not a regular file, generic_error_remove_page in
f2fs_invalidate_compress_pages will always be failed, set compress
inode as a regular file to fix it.
Fixes: 6ce19aff0b ("f2fs: compress: add compress_inode to cache compressed blocks")
Signed-off-by: Fengnan Chang <changfengnan@vivo.com>
Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Make f2fs_file_read_iter() and f2fs_file_write_iter() use the iomap
direct I/O implementation instead of the fs/direct-io.c one.
The iomap implementation is more efficient, and it also avoids the need
to add new features and optimizations to the old implementation.
This new implementation also eliminates the need for f2fs to hook bio
submission and completion and to allocate memory per-bio. This is
because it's possible to correctly update f2fs's in-flight DIO counters
using __iomap_dio_rw() in combination with an implementation of
iomap_dio_ops::end_io() (as suggested by Christoph Hellwig).
When possible, this new implementation preserves existing f2fs behavior
such as the conditions for falling back to buffered I/O.
This patch has been tested with xfstests by running 'gce-xfstests -c
f2fs -g auto -X generic/017' with and without this patch; no regressions
were seen. (Some tests fail both before and after. generic/017 hangs
both before and after, so it had to be excluded.)
Signed-off-by: Eric Biggers <ebiggers@google.com>
[Jaegeuk Kim: use spin_lock_bh for f2fs_update_iostat in softirq]
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Implement 'struct iomap_ops' for f2fs, in preparation for making f2fs
use iomap for direct I/O.
Note that this may be used for other things besides direct I/O in the
future; however, for now I've only tested it for direct I/O.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Pass in the original position and count rather than the position and
count that were updated by the write. Also use the correct types for
all arguments, in particular the file offset which was being truncated
to 32 bits on 32-bit platforms.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
DIO preallocates physical blocks before writing data, but if an error occurrs
or power-cut happens, we can see block contents from the disk. This patch tries
to fix it by 1) turning to buffered writes for DIO into holes, 2) truncating
unwritten blocks from error or power-cut.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
f2fs_write_begin() assumes that all blocks were preallocated by
default unless FI_NO_PREALLOC is explicitly set. This invites data
corruption, as there are cases in which not all blocks are preallocated.
Commit 47501f87c6 ("f2fs: preallocate DIO blocks when forcing
buffered_io") fixed one case, but there are others remaining.
Fix up this logic by replacing this flag with FI_PREALLOCATED_ALL, which
only gets set if all blocks for the current write were preallocated.
Also clean up f2fs_preallocate_blocks(), move it to file.c, and make it
handle some of the logic that was previously in write_iter() directly.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Don't alloc new page pointers array to replace old, just use old, introduce
valid_nr_cpages to indicate valid number of page pointers in array, try to
reduce one page array alloc and free when write compress page.
Signed-off-by: Fengnan Chang <changfengnan@vivo.com>
Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Add a done_before argument to iomap_dio_rw that indicates how much of
the request has already been transferred. When the request succeeds, we
report that done_before additional bytes were tranferred. This is
useful for finishing a request asynchronously when part of the request
has already been completed synchronously.
We'll use that to allow iomap_dio_rw to be used with page faults
disabled: when a page fault occurs while submitting a request, we
synchronously complete the part of the request that has already been
submitted. The caller can then take care of the page fault and call
iomap_dio_rw again for the rest of the request, passing in the number of
bytes already tranferred.
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
This information can be used to check how much time we need to give to issue
all the discard commands.
Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
We observed the following deadlock in the stress test under low
memory scenario:
Thread A Thread B
- erofs_shrink_scan
- erofs_try_to_release_workgroup
- erofs_workgroup_try_to_freeze -- A
- z_erofs_do_read_page
- z_erofs_collection_begin
- z_erofs_register_collection
- erofs_insert_workgroup
- xa_lock(&sbi->managed_pslots) -- B
- erofs_workgroup_get
- erofs_wait_on_workgroup_freezed -- A
- xa_erase
- xa_lock(&sbi->managed_pslots) -- B
To fix this, it needs to hold xa_lock before freezing the workgroup
since xarray will be touched then. So let's hold the lock before
accessing each workgroup, just like what we did with the radix tree
before.
[ Gao Xiang: Jianhua Hao also reports this issue at
https://lore.kernel.org/r/b10b85df30694bac8aadfe43537c897a@xiaomi.com ]
Link: https://lore.kernel.org/r/20211118135844.3559-1-huangjianan@oppo.com
Fixes: 64094a0441 ("erofs: convert workstn to XArray")
Reviewed-by: Chao Yu <chao@kernel.org>
Reviewed-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Signed-off-by: Huang Jianan <huangjianan@oppo.com>
Reported-by: Jianhua Hao <haojianhua1@xiaomi.com>
Signed-off-by: Gao Xiang <xiang@kernel.org>
There are pclusters in runtime marked with Z_EROFS_PCLUSTER_TAIL
before actual I/O submission. Thus, the decompression chain can be
extended if the following pcluster chain hooks such tail pcluster.
As the related comment mentioned, if some page is made of a hooked
pcluster and another followed pcluster, it can be reused for in-place
I/O (since I/O should be submitted anyway):
_______________________________________________________________
| tail (partial) page | head (partial) page |
|_____PRIMARY_HOOKED___|____________PRIMARY_FOLLOWED____________|
However, it's by no means safe to reuse as pagevec since if such
PRIMARY_HOOKED pclusters finally move into bypass chain without I/O
submission. It's somewhat hard to reproduce with LZ4 and I just found
it (general protection fault) by ro_fsstressing a LZMA image for long
time.
I'm going to actively clean up related code together with multi-page
folio adaption in the next few months. Let's address it directly for
easier backporting for now.
Call trace for reference:
z_erofs_decompress_pcluster+0x10a/0x8a0 [erofs]
z_erofs_decompress_queue.isra.36+0x3c/0x60 [erofs]
z_erofs_runqueue+0x5f3/0x840 [erofs]
z_erofs_readahead+0x1e8/0x320 [erofs]
read_pages+0x91/0x270
page_cache_ra_unbounded+0x18b/0x240
filemap_get_pages+0x10a/0x5f0
filemap_read+0xa9/0x330
new_sync_read+0x11b/0x1a0
vfs_read+0xf1/0x190
Link: https://lore.kernel.org/r/20211103182006.4040-1-xiang@kernel.org
Fixes: 3883a79abd ("staging: erofs: introduce VLE decompression support")
Cc: <stable@vger.kernel.org> # 4.19+
Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Currently, ->lru is a way to arrange non-LRU pages and has some
in-kernel users. In order to minimize noticable issues of page
reclaim and cache thrashing under high memory presure, limited
temporary pages were all chained with ->lru and can be reused
during the request. However, it seems that ->lru could be removed
when folio is landing.
Let's use page->private to chain temporary pages for now instead
and transform EROFS formally after the topic of the folio / file
page design is finalized.
Link: https://lore.kernel.org/r/20211022090120.14675-1-hsiangkao@linux.alibaba.com
Cc: Matthew Wilcox <willy@infradead.org>
Reviewed-by: Kent Overstreet <kent.overstreet@gmail.com>
Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Add MicroLZMA support in order to maximize compression ratios for
specific scenarios. For example, it's useful for low-end embedded
boards and as a secondary algorithm in a file for specific access
patterns.
MicroLZMA is a new container format for raw LZMA1, which was created
by Lasse Collin aiming to minimize old LZMA headers and get rid of
unnecessary EOPM (end of payload marker) as well as to enable
fixed-sized output compression, especially for 4KiB pclusters.
Similar to LZ4, inplace I/O approach is used to minimize runtime
memory footprint when dealing with I/O. Overlapped decompression is
handled with 1) bounced buffer for data under processing or 2) extra
short-lived pages from the on-stack pagepool which will be shared in
the same read request (128KiB for example).
Link: https://lore.kernel.org/r/20211010213145.17462-8-xiang@kernel.org
Acked-by: Chao Yu <chao@kernel.org>
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
