Use sbi.stat_lock to protect sbi->unusable_block_count accesss/udpate, in
order to avoid potential race on it.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Previously, f2fs_is_valid_blkaddr(, blkaddr, DATA_GENERIC) will check
whether @blkaddr locates in main area or not.
That check is weak, since the block address in range of main area can
point to the address which is not valid in segment info table, and we
can not detect such condition, we may suffer worse corruption as system
continues running.
So this patch introduce DATA_GENERIC_ENHANCE to enhance the sanity check
which trigger SIT bitmap check rather than only range check.
This patch did below changes as wel:
- set SBI_NEED_FSCK in f2fs_is_valid_blkaddr().
- get rid of is_valid_data_blkaddr() to avoid panic if blkaddr is invalid.
- introduce verify_fio_blkaddr() to wrap fio {new,old}_blkaddr validation check.
- spread blkaddr check in:
* f2fs_get_node_info()
* __read_out_blkaddrs()
* f2fs_submit_page_read()
* ra_data_block()
* do_recover_data()
This patch can fix bug reported from bugzilla below:
https://bugzilla.kernel.org/show_bug.cgi?id=203215https://bugzilla.kernel.org/show_bug.cgi?id=203223https://bugzilla.kernel.org/show_bug.cgi?id=203231https://bugzilla.kernel.org/show_bug.cgi?id=203235https://bugzilla.kernel.org/show_bug.cgi?id=203241
= Update by Jaegeuk Kim =
DATA_GENERIC_ENHANCE enhanced to validate block addresses on read/write paths.
But, xfstest/generic/446 compalins some generated kernel messages saying invalid
bitmap was detected when reading a block. The reaons is, when we get the
block addresses from extent_cache, there is no lock to synchronize it from
truncating the blocks in parallel.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
As Park Ju Hyung reported in mailing list:
https://sourceforge.net/p/linux-f2fs/mailman/message/36639787/
generic_make_request: Trying to write to read-only block-device loop0 (partno 0)
WARNING: CPU: 0 PID: 23437 at block/blk-core.c:2174 generic_make_request_checks+0x594/0x630
generic_make_request+0x46/0x3d0
submit_bio+0x30/0x110
__submit_merged_bio+0x68/0x390
f2fs_submit_page_write+0x1bb/0x7f0
f2fs_do_write_meta_page+0x7f/0x160
__f2fs_write_meta_page+0x70/0x140
f2fs_sync_meta_pages+0x140/0x250
f2fs_write_checkpoint+0x5c5/0x17b0
f2fs_sync_fs+0x9c/0x110
sync_filesystem+0x66/0x80
f2fs_recover_fsync_data+0x790/0xa30
f2fs_fill_super+0xe4e/0x1980
mount_bdev+0x518/0x610
mount_fs+0x34/0x13f
vfs_kern_mount.part.11+0x4f/0x120
do_mount+0x2d1/0xe40
__x64_sys_mount+0xbf/0xe0
do_syscall_64+0x4a/0xf0
entry_SYSCALL_64_after_hwframe+0x44/0xa9
print_req_error: I/O error, dev loop0, sector 4096
If block device is readonly, we should never trigger write IO from
filesystem layer, but previously, orphan and journal recovery didn't
consider such condition, result in triggering above warning, fix it.
Reported-by: Park Ju Hyung <qkrwngud825@gmail.com>
Tested-by: Park Ju Hyung <qkrwngud825@gmail.com>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
For large_nat_bitmap feature, there is a design flaw:
Previous:
struct f2fs_checkpoint layout:
+--------------------------+ 0x0000
| checkpoint_ver |
| ...... |
| checksum_offset |------+
| ...... | |
| sit_nat_version_bitmap[] |<-----|-------+
| ...... | | |
| checksum_value |<-----+ |
+--------------------------+ 0x1000 |
| | nat_bitmap + sit_bitmap
| payload blocks | |
| | |
+--------------------------|<-------------+
Obviously, if nat_bitmap size + sit_bitmap size is larger than
MAX_BITMAP_SIZE_IN_CKPT, nat_bitmap or sit_bitmap may overlap
checkpoint checksum's position, once checkpoint() is triggered
from kernel, nat or sit bitmap will be damaged by checksum field.
In order to fix this, let's relocate checksum_value's position
to the head of sit_nat_version_bitmap as below, then nat/sit
bitmap and chksum value update will become safe.
After:
struct f2fs_checkpoint layout:
+--------------------------+ 0x0000
| checkpoint_ver |
| ...... |
| checksum_offset |------+
| ...... | |
| sit_nat_version_bitmap[] |<-----+
| ...... |<-------------+
| | |
+--------------------------+ 0x1000 |
| | nat_bitmap + sit_bitmap
| payload blocks | |
| | |
+--------------------------|<-------------+
Related report and discussion:
https://sourceforge.net/p/linux-f2fs/mailman/message/36642346/
Reported-by: Park Ju Hyung <qkrwngud825@gmail.com>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Previously, f2fs_checkpoint.checksum_offset points fixed position of
f2fs_checkpoint structure:
"#define CP_CHKSUM_OFFSET 4092"
It is unnecessary, and it breaks the consecutiveness of nat and sit
bitmap stored across checkpoint park block and payload blocks.
This patch allows f2fs to handle unfixed .checksum_offset.
In addition, for the case checksum value is stored in the middle of
checkpoint park, calculating checksum value with superposition method
like we did for inode_checksum.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch changes codes as below:
- don't use is_read_io() as a condition to judge the meta IO.
- use .is_por to replace .is_meta to indicate IO is from recovery explicitly.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
As Hagbard Celine reported:
Hi, this is a long standing bug that I've hit before on older kernels,
but I was not able to get the syslog saved because of the nature of
the bug. This time I had booted form a pen-drive, and was able to save
the log to it's efi-partition.
What i did to trigger it was to create a partition and format it f2fs,
then mount it with options:
"rw,relatime,lazytime,background_gc=on,disable_ext_identify,discard,heap,user_xattr,inline_xattr,acl,inline_data,inline_dentry,flush_merge,data_flush,extent_cache,mode=adaptive,active_logs=6,whint_mode=fs-based,alloc_mode=default,fsync_mode=strict".
Then I unpacked a big .tar.xz to the partition (I used a
gentoo-stage3-tarball as I was in process of installing Gentoo).
Same options just without data_flush gives no problems.
Mar 20 20:54:01 usbgentoo kernel: FAT-fs (nvme0n1p4): Volume was not
properly unmounted. Some data may be corrupt. Please run fsck.
Mar 20 21:05:23 usbgentoo kernel: kworker/dying (1588) used greatest
stack depth: 12064 bytes left
Mar 20 21:06:40 usbgentoo kernel: BUG: stack guard page was hit at
00000000a4b0733c (stack is 0000000056016422..0000000096e7463f)
Mar 20 21:06:40 usbgentoo kernel: kernel stack overflow
......
Mar 20 21:06:40 usbgentoo kernel: Call Trace:
Mar 20 21:06:40 usbgentoo kernel: read_node_page+0x71/0xf0
Mar 20 21:06:40 usbgentoo kernel: ? xas_load+0x8/0x50
Mar 20 21:06:40 usbgentoo kernel: __get_node_page+0x73/0x2a0
Mar 20 21:06:40 usbgentoo kernel: f2fs_get_dnode_of_data+0x34e/0x580
Mar 20 21:06:40 usbgentoo kernel: f2fs_write_inline_data+0x5e/0x2a0
Mar 20 21:06:40 usbgentoo kernel: __write_data_page+0x421/0x690
Mar 20 21:06:40 usbgentoo kernel: f2fs_write_cache_pages+0x1cf/0x460
Mar 20 21:06:40 usbgentoo kernel: f2fs_write_data_pages+0x2b3/0x2e0
Mar 20 21:06:40 usbgentoo kernel: ? f2fs_inode_chksum_verify+0x1d/0xc0
Mar 20 21:06:40 usbgentoo kernel: ? read_node_page+0x71/0xf0
Mar 20 21:06:40 usbgentoo kernel: do_writepages+0x3c/0xd0
Mar 20 21:06:40 usbgentoo kernel: __filemap_fdatawrite_range+0x7c/0xb0
Mar 20 21:06:40 usbgentoo kernel: f2fs_sync_dirty_inodes+0xf2/0x200
Mar 20 21:06:40 usbgentoo kernel: f2fs_balance_fs_bg+0x2a3/0x2c0
Mar 20 21:06:40 usbgentoo kernel: ? f2fs_inode_dirtied+0x21/0xc0
Mar 20 21:06:40 usbgentoo kernel: f2fs_balance_fs+0xd6/0x2b0
Mar 20 21:06:40 usbgentoo kernel: __write_data_page+0x4fb/0x690
......
Mar 20 21:06:40 usbgentoo kernel: __writeback_single_inode+0x2a1/0x340
Mar 20 21:06:40 usbgentoo kernel: ? soft_cursor+0x1b4/0x220
Mar 20 21:06:40 usbgentoo kernel: writeback_sb_inodes+0x1d5/0x3e0
Mar 20 21:06:40 usbgentoo kernel: __writeback_inodes_wb+0x58/0xa0
Mar 20 21:06:40 usbgentoo kernel: wb_writeback+0x250/0x2e0
Mar 20 21:06:40 usbgentoo kernel: ? 0xffffffff8c000000
Mar 20 21:06:40 usbgentoo kernel: ? cpumask_next+0x16/0x20
Mar 20 21:06:40 usbgentoo kernel: wb_workfn+0x2f6/0x3b0
Mar 20 21:06:40 usbgentoo kernel: ? __switch_to_asm+0x40/0x70
Mar 20 21:06:40 usbgentoo kernel: process_one_work+0x1f5/0x3f0
Mar 20 21:06:40 usbgentoo kernel: worker_thread+0x28/0x3c0
Mar 20 21:06:40 usbgentoo kernel: ? rescuer_thread+0x330/0x330
Mar 20 21:06:40 usbgentoo kernel: kthread+0x10e/0x130
Mar 20 21:06:40 usbgentoo kernel: ? kthread_create_on_node+0x60/0x60
Mar 20 21:06:40 usbgentoo kernel: ret_from_fork+0x35/0x40
The root cause is that we run into an infinite recursive calling in
between f2fs_balance_fs_bg and writepage() as described below:
- f2fs_write_data_pages --- A
- __write_data_page
- f2fs_balance_fs
- f2fs_balance_fs_bg --- B
- f2fs_sync_dirty_inodes
- filemap_fdatawrite
- f2fs_write_data_pages --- A
...
- f2fs_balance_fs_bg --- B
...
In order to fix this issue, let's detect such condition in __write_data_page()
and just skip calling f2fs_balance_fs() recursively.
Reported-by: Hagbard Celine <hagbardcelin@gmail.com>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
As Gao Xiang reported in bugzilla:
https://bugzilla.kernel.org/show_bug.cgi?id=202749
f2fs may skip pageout() due to incorrect page reference count.
The problem here is that MM defined the rule [1] very clearly that
once page was set with PG_private flag, we should increment the
refcount in that page, also main flows like pageout(), migrate_page()
will assume there is one additional page reference count if
page_has_private() returns true.
But currently, f2fs won't add/del refcount when changing PG_private
flag. Anyway, f2fs should follow MM's rule to make MM's related flows
running as expected.
[1] https://lore.kernel.org/lkml/2b19b3c4-2bc4-15fa-15cc-27a13e5c7af1@aol.com/
Reported-by: Gao Xiang <gaoxiang25@huawei.com>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
If we met this once, let fsck.f2fs clear this only.
Note that, this addresses all the subtle fault injection test.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Some works after roll-forward recovery can get an error which will release
all the data structures. Let's flush them in order to make it clean.
One possible corruption came from:
[ 90.400500] list_del corruption. prev->next should be ffffffed1f566208, but was (null)
[ 90.675349] Call trace:
[ 90.677869] __list_del_entry_valid+0x94/0xb4
[ 90.682351] remove_dirty_inode+0xac/0x114
[ 90.686563] __f2fs_write_data_pages+0x6a8/0x6c8
[ 90.691302] f2fs_write_data_pages+0x40/0x4c
[ 90.695695] do_writepages+0x80/0xf0
[ 90.699372] __writeback_single_inode+0xdc/0x4ac
[ 90.704113] writeback_sb_inodes+0x280/0x440
[ 90.708501] wb_writeback+0x1b8/0x3d0
[ 90.712267] wb_workfn+0x1a8/0x4d4
[ 90.715765] process_one_work+0x1c0/0x3d4
[ 90.719883] worker_thread+0x224/0x344
[ 90.723739] kthread+0x120/0x130
[ 90.727055] ret_from_fork+0x10/0x18
Reported-by: Sahitya Tummala <stummala@codeaurora.org>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This mode returns mount() quickly with EAGAIN. We can trigger this by
shutdown(F2FS_GOING_DOWN_NEED_FSCK).
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
For all ordered cases in f2fs_wait_on_page_writeback(), we need to
check PageWriteback status, so let's clean up to relocate the check
into f2fs_wait_on_page_writeback().
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch cleans up checkpoint flow a bit:
- remove unneeded circulation of flushing meta pages.
- don't flush nat_bits pages in prior to other checkpoint pages.
- add bug_on to check remained meta pages after flushing.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
One report says memalloc failure during mount.
(unwind_backtrace) from [<c010cd4c>] (show_stack+0x10/0x14)
(show_stack) from [<c049c6b8>] (dump_stack+0x8c/0xa0)
(dump_stack) from [<c024fcf0>] (warn_alloc+0xc4/0x160)
(warn_alloc) from [<c0250218>] (__alloc_pages_nodemask+0x3f4/0x10d0)
(__alloc_pages_nodemask) from [<c0270450>] (kmalloc_order_trace+0x2c/0x120)
(kmalloc_order_trace) from [<c03fa748>] (build_node_manager+0x35c/0x688)
(build_node_manager) from [<c03de494>] (f2fs_fill_super+0xf0c/0x16cc)
(f2fs_fill_super) from [<c02a5864>] (mount_bdev+0x15c/0x188)
(mount_bdev) from [<c03da624>] (f2fs_mount+0x18/0x20)
(f2fs_mount) from [<c02a68b8>] (mount_fs+0x158/0x19c)
(mount_fs) from [<c02c3c9c>] (vfs_kern_mount+0x78/0x134)
(vfs_kern_mount) from [<c02c76ac>] (do_mount+0x474/0xca4)
(do_mount) from [<c02c8264>] (SyS_mount+0x94/0xbc)
(SyS_mount) from [<c0108180>] (ret_fast_syscall+0x0/0x48)
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch reorders flow from
- update page
- set_page_dirty
- wait_on_page_writeback
to
- wait_on_page_writeback
- update page
- set_page_dirty
The reason is:
- set_page_dirty will increase reference of dirty page, the reference
should be cleared before wait_on_page_writeback to keep its consistency.
- some devices need stable page during page writebacking, so we
should not change page's data.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
In F2FS_HAS_FEATURE(), we will use F2FS_SB(sb) to get sbi pointer to
access .raw_super field, to avoid unneeded pointer conversion, this
patch changes to F2FS_HAS_FEATURE() accept sbi parameter directly.
Just do cleanup, no logic change.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
For journalled quota mode, let checkpoint to flush dquot dirty data
and quota file data to guarntee persistence of all quota sysfile in
last checkpoint, by this way, we can avoid corrupting quota sysfile
when encountering SPO.
The implementation is as below:
1. add a global state SBI_QUOTA_NEED_FLUSH to indicate that there is
cached dquot metadata changes in quota subsystem, and later checkpoint
should:
a) flush dquot metadata into quota file.
b) flush quota file to storage to keep file usage be consistent.
2. add a global state SBI_QUOTA_NEED_REPAIR to indicate that quota
operation failed due to -EIO or -ENOSPC, so later,
a) checkpoint will skip syncing dquot metadata.
b) CP_QUOTA_NEED_FSCK_FLAG will be set in last cp pack to give a
hint for fsck repairing.
3. add a global state SBI_QUOTA_SKIP_FLUSH, in checkpoint, if quota
data updating is very heavy, it may cause hungtask in block_operation().
To avoid this, if our retry time exceed threshold, let's just skip
flushing and retry in next checkpoint().
Signed-off-by: Weichao Guo <guoweichao@huawei.com>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
[Jaegeuk Kim: avoid warnings and set fsck flag]
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Note that, it requires "f2fs: return correct errno in f2fs_gc".
This adds a lightweight non-persistent snapshotting scheme to f2fs.
To use, mount with the option checkpoint=disable, and to return to
normal operation, remount with checkpoint=enable. If the filesystem
is shut down before remounting with checkpoint=enable, it will revert
back to its apparent state when it was first mounted with
checkpoint=disable. This is useful for situations where you wish to be
able to roll back the state of the disk in case of some critical
failure.
Signed-off-by: Daniel Rosenberg <drosen@google.com>
[Jaegeuk Kim: use SB_RDONLY instead of MS_RDONLY]
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
There is one case that we can leave bio in f2fs, result in hanging
page writeback waiter.
Thread A Thread B
- f2fs_write_cache_pages
- f2fs_submit_page_write
page #0 cached in bio #0 of cold log
- f2fs_submit_page_write
page #1 cached in bio #1 of warm log
- f2fs_write_cache_pages
- f2fs_submit_page_write
bio is full, submit bio #1 contain page #1
- f2fs_submit_merged_write_cond(, page #1)
fail to submit bio #0 due to page #1 is not in any cached bios.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch avoids BUG_ON when f2fs_get_meta_page_nofail got EIO during
xfstests/generic/475.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Remove the verbose license text from f2fs files and replace them with
SPDX tags. This does not change the license of any of the code.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
generic/417 reported as blow:
------------[ cut here ]------------
kernel BUG at /home/yuchao/git/devf2fs/inode.c:695!
invalid opcode: 0000 [#1] PREEMPT SMP
CPU: 1 PID: 21697 Comm: umount Tainted: G W O 4.18.0-rc2+ #39
Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006
EIP: f2fs_evict_inode+0x556/0x580 [f2fs]
Call Trace:
? _raw_spin_unlock+0x2c/0x50
evict+0xa8/0x170
dispose_list+0x34/0x40
evict_inodes+0x118/0x120
generic_shutdown_super+0x41/0x100
? rcu_read_lock_sched_held+0x97/0xa0
kill_block_super+0x22/0x50
kill_f2fs_super+0x6f/0x80 [f2fs]
deactivate_locked_super+0x3d/0x70
deactivate_super+0x40/0x60
cleanup_mnt+0x39/0x70
__cleanup_mnt+0x10/0x20
task_work_run+0x81/0xa0
exit_to_usermode_loop+0x59/0xa7
do_fast_syscall_32+0x1f5/0x22c
entry_SYSENTER_32+0x53/0x86
EIP: f2fs_evict_inode+0x556/0x580 [f2fs]
It can simply reproduced with scripts:
Enable quota feature during mkfs.
Testcase1:
1. mkfs.f2fs /dev/zram0
2. mount -t f2fs /dev/zram0 /mnt/f2fs
3. xfs_io -f /mnt/f2fs/file -c "pwrite 0 4k" -c "fsync"
4. godown /mnt/f2fs
5. umount /mnt/f2fs
6. mount -t f2fs -o ro /dev/zram0 /mnt/f2fs
7. umount /mnt/f2fs
Testcase2:
1. mkfs.f2fs /dev/zram0
2. mount -t f2fs /dev/zram0 /mnt/f2fs
3. touch /mnt/f2fs/file
4. create process[pid = x] do:
a) open /mnt/f2fs/file;
b) unlink /mnt/f2fs/file
5. godown -f /mnt/f2fs
6. kill process[pid = x]
7. umount /mnt/f2fs
8. mount -t f2fs -o ro /dev/zram0 /mnt/f2fs
9. umount /mnt/f2fs
The reason is: during recovery, i_{c,m}time of inode will be updated, then
the inode can be set dirty w/o being tracked in sbi->inode_list[DIRTY_META]
global list, so later write_checkpoint will not flush such dirty inode into
node page.
Once umount is called, sync_filesystem() in generic_shutdown_super() will
skip syncng dirty inodes due to sb_rdonly check, leaving dirty inodes
there.
To solve this issue, during umount, add remove SB_RDONLY flag in
sb->s_flags, to make sure sync_filesystem() will not be skipped.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
When CONFIG_F2FS_FAULT_INJECTION is disabled, we get a warning about an
unused label:
fs/f2fs/segment.c: In function '__submit_discard_cmd':
fs/f2fs/segment.c:1059:1: error: label 'submit' defined but not used [-Werror=unused-label]
This could be fixed by adding another #ifdef around it, but the more
reliable way of doing this seems to be to remove the other #ifdefs
where that is easily possible.
By defining time_to_inject() as a trivial stub, most of the checks for
CONFIG_F2FS_FAULT_INJECTION can go away. This also leads to nicer
formatting of the code.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Previously, once fault injection is on, by default, all kind of faults
will be injected to f2fs, if we want to trigger single or specified
combined type during the test, we need to configure sysfs entry, it will
be a little inconvenient to integrate sysfs configuring into testsuit,
such as xfstest.
So this patch introduces a new mount option 'fault_type' to assist old
option 'fault_injection', with these two mount options, we can specify
any fault rate/type at mount-time.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
f2fs recovery flow is relying on dnode block link list, it means fsynced
file recovery depends on previous dnode's persistence in the list, so
during fsync() we should wait on all regular inode's dnode writebacked
before issuing flush.
By this way, we can avoid dnode block list being broken by out-of-order
IO submission due to IO scheduler or driver.
Sheng Yong helps to do the test with this patch:
Target:/data (f2fs, -)
64MB / 32768KB / 4KB / 8
1 / PERSIST / Index
Base:
SEQ-RD(MB/s) SEQ-WR(MB/s) RND-RD(IOPS) RND-WR(IOPS) Insert(TPS) Update(TPS) Delete(TPS)
1 867.82 204.15 41440.03 41370.54 680.8 1025.94 1031.08
2 871.87 205.87 41370.3 40275.2 791.14 1065.84 1101.7
3 866.52 205.69 41795.67 40596.16 694.69 1037.16 1031.48
Avg 868.7366667 205.2366667 41535.33333 40747.3 722.21 1042.98 1054.753333
After:
SEQ-RD(MB/s) SEQ-WR(MB/s) RND-RD(IOPS) RND-WR(IOPS) Insert(TPS) Update(TPS) Delete(TPS)
1 798.81 202.5 41143 40613.87 602.71 838.08 913.83
2 805.79 206.47 40297.2 41291.46 604.44 840.75 924.27
3 814.83 206.17 41209.57 40453.62 602.85 834.66 927.91
Avg 806.4766667 205.0466667 40883.25667 40786.31667 603.3333333 837.83 922.0033333
Patched/Original:
0.928332713 0.999074239 0.984300676 1.000957528 0.835398753 0.803303994 0.874141189
It looks like atomic write will suffer performance regression.
I suspect that the criminal is that we forcing to wait all dnode being in
storage cache before we issue PREFLUSH+FUA.
BTW, will commit ("f2fs: don't need to wait for node writes for atomic write")
cause the problem: we will lose data of last transaction after SPO, even if
atomic write return no error:
- atomic_open();
- write() P1, P2, P3;
- atomic_commit();
- writeback data: P1, P2, P3;
- writeback node: N1, N2, N3; <--- If N1, N2 is not writebacked, N3 with fsync_mark is
writebacked, In SPOR, we won't find N3 since node chain is broken, turns out that losing
last transaction.
- preflush + fua;
- power-cut
If we don't wait dnode writeback for atomic_write:
SEQ-RD(MB/s) SEQ-WR(MB/s) RND-RD(IOPS) RND-WR(IOPS) Insert(TPS) Update(TPS) Delete(TPS)
1 779.91 206.03 41621.5 40333.16 716.9 1038.21 1034.85
2 848.51 204.35 40082.44 39486.17 791.83 1119.96 1083.77
3 772.12 206.27 41335.25 41599.65 723.29 1055.07 971.92
Avg 800.18 205.55 41013.06333 40472.99333 744.0066667 1071.08 1030.18
Patched/Original:
0.92108464 1.001526693 0.987425886 0.993268102 1.030180511 1.026942031 0.976702294
SQLite's performance recovers.
Jaegeuk:
"Practically, I don't see db corruption becase of this. We can excuse to lose
the last transaction."
Finally, we decide to keep original implementation of atomic write interface
sematics that we don't wait all dnode writeback before preflush+fua submission.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
For migration of encrypted inode's block, we load data of encrypted block
into meta inode's page cache, after checkpoint, those all intermediate
pages should be clean, and no one will read them again, so let's just
release them for more memory.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
If caller of __get_meta_page() can handle error, let's propagate error
from __get_meta_page().
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch introduces verify_blkaddr to check meta/data block address
with valid range to detect bug earlier.
In addition, once we encounter an invalid blkaddr, notice user to run
fsck to fix, and let the kernel panic.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch fix hungtask problem which can be reproduced as follow:
Thread 0~3:
while true
do
touch /xxx/test/file_xxx
done
Thread 4 write a new checkpoint every three seconds.
In the meantime, fio start 16 threads for randwrite.
With my debug info, cycles num will exceed 1000 in function
f2fs_sync_dirty_inodes, and most of cycle will be dropped
into congestion_wait() and sleep more than 20ms. Cycles num
reduced to 3 with this patch.
Signed-off-by: Yunlei He <heyunlei@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
During orphan inode recovery, checkpoint should never succeed due to
SBI_POR_DOING flag, so we don't need acquire orphan ino which only be
used by checkpoint.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Once we shutdown f2fs, we have to flush stale pages in order to unmount
the system. In order to make stable, we need to stop fault injection as well.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
It turns out losing meta pages in shutdown period makes f2fs very unstable
so that I could see many unexpected error conditions.
Let's keep meta pages for fault injection and sudden power-off tests.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Although mixed sync/async IOs can have continuous LBA, as they have
different IO priority, block IO scheduler will add them into different
queues and commit them separately, result in splited IOs which causes
wrose performance.
This patch gives high priority to synchronous IO of nodes, means that
once synchronous flow starts, it can interrupt asynchronous writeback
flow of system flusher, so more big IOs can be expected.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
If we change system time to the past, get_mtime() will return a
overflowed time, and SIT_I(sbi)->max_mtime will be udpated
incorrectly, this patch fixes the two issues.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
- rename is_valid_blkaddr() to is_valid_meta_blkaddr() for readability.
- introduce is_valid_blkaddr() for cleanup.
No logic change in this patch.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch removes duplicated dquot_initialize in recover_orphan_inode(),
and fix the error handling if dquot_initialize fails.
Signed-off-by: Sheng Yong <shengyong1@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Related to https://lkml.org/lkml/2018/4/8/661
Sometimes, we need to write meta data to new allocated block address,
then we will allocate a zeroed page in inner inode's address space, and
fill partial data in it, and leave other place with zero value which means
some fields are initial status.
There are two inner inodes (meta inode and node inode) setting __GFP_ZERO,
I have just checked them, for both of them, we can avoid using __GFP_ZERO,
and do initialization by ourselves to avoid unneeded/redundant zeroing
from mm.
Cc: <stable@vger.kernel.org>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch reverts copied f2fs_set_page_dirty_nobuffer to use generic function
for stability.
This reverts commit fe76b796fc.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch check blkaddr more accuratly before issue a
write or read bio.
Signed-off-by: Yunlei He <heyunlei@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Previously, we attempt to flush the whole cp pack in a single bio,
however, when suddenly powering off at this time, we could get into
an extreme scenario that cp pack 1 page and cp pack 2 page are updated
and latest, but payload or current summaries are still partially
outdated. (see reliable write in the UFS specification)
This patch submits the whole cp pack except cp pack 2 page at first,
and then writes the cp pack 2 page with an extra independent
bio with pre-io barrier.
Signed-off-by: Gao Xiang <gaoxiang25@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This is to detect dquot_initialize errors early from evict_inode
for orphan inodes.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Once CP_TRIMMED_FLAG is set, after a reboot, we will never issue discard
before LBA becomes invalid again, fix it by clearing the flag in
checkpoint without CP_TRIMMED reason.
Fixes: 1f43e2ad7b ("f2fs: introduce CP_TRIMMED_FLAG to avoid unneeded discard")
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
If fsck.f2fs changes crc, we have no way to recover some inode blocks by roll-
forward recovery. Let's relax the condition to recover them.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Chao Yu
Jaegeuk Kim
Daniel Rosenberg
Arnd Bergmann
Sheng Yong
Yunlei He
Kees Cook
Gao Xiang