Currently, if we call fsync after updating the xattr date belongs to the
file, f2fs needs to trigger checkpoint to keep xattr data consistent. But,
this policy cause low performance as checkpoint will block most foreground
operations and cause unneeded and unrelated IOs around checkpoint.
This patch will reuse regular file recovery policy for xattr node block,
so, we change to write xattr node block tagged with fsync flag to warm
area instead of cold area, and during recovery, we search warm node chain
for fsynced xattr block, and do the recovery.
So, for below application IO pattern, performance can be improved
obviously:
- touch file
- create/update/delete xattr entry in file
- fsync file
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
We use following method to calculate size with current page index:
size = index << PAGE_SHIFT
If type of index has only 32-bits size, left shifting will incur overflow,
which makes result incorrect.
So let's cast index with 64-bits type to avoid such issue.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch remove unused values in function recover_fsync_data
Signed-off-by: Yunlei He <heyunlei@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
If a file needs to keep its i_size by fallocate, we need to turn off auto
recovery during roll-forward recovery.
This will resolve the below scenario.
1. xfs_io -f /mnt/f2fs/file -c "pwrite 0 4096" -c "fsync"
2. xfs_io -f /mnt/f2fs/file -c "falloc -k 4096 4096" -c "fsync"
3. md5sum /mnt/f2fs/file;
4. godown /mnt/f2fs/
5. umount /mnt/f2fs/
6. mount -t f2fs /dev/sdx /mnt/f2fs
7. md5sum /mnt/f2fs/file
Reported-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
If i_size is already valid during roll_forward recovery, we should not update
it according to the block alignment.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
i_times of inode will be set with current system time which can be
configured through 'date', so it's not safe to judge dnode block as
garbage data or unchanged inode depend on i_times.
Now, we have used enhanced 'cp_ver + cp' crc method to verify valid
dnode block, so I expect recoverying invalid dnode is almost not
possible.
This reverts commit 807b1e1c8e.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Shouldn't update in-memory i_atime with on-disk i_mtime of inode when
recovering inode.
Shuoran found this bug which is hidden for a long time, honour is belong
to him.
Signed-off-by: Shuoran Liu <liushuoran@huawei.com>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch introduces spinlock to protect updating process of ckpt_flags
field in struct f2fs_checkpoint, it avoids incorrectly updating in race
condition.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
[Jaegeuk Kim: add __is_set_ckpt_flags likewise __set_ckpt_flags]
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Previously, we used cp_version only to detect recoverable dnodes.
In order to avoid same garbage cp_version, we needed to truncate the next
dnode during checkpoint, resulting in additional discard or data write.
If we can distinguish this by using crc in addition to cp_version, we can
remove this overhead.
There is backward compatibility concern where it changes node_footer layout.
So, this patch introduces a new checkpoint flag, CP_CRC_RECOVERY_FLAG, to
detect new layout. New layout will be activated only when this flag is set.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Add roll-forward recovery process for encrypted dentry, so the first fsync
issued to an encrypted file does not need writing checkpoint.
This improves the performance of the following test at thousands of small
files: open -> write -> fsync -> close
Signed-off-by: Shuoran Liu <liushuoran@huawei.com>
Acked-by: Chao Yu <yuchao0@huawei.com>
[Jaegeuk Kim: modify kernel message to show encrypted names]
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Like most filesystems, f2fs will issue discard command synchronously, so
when user trigger fstrim through ioctl, multiple discard commands will be
issued serially with sync mode, which makes poor performance.
In this patch we try to support async discard, so that all discard
commands can be issued and be waited for endio in batch to improve
performance.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
It's enough to show BUG or WARN by f2fs_bug_on for error case.
Then, we don't need to remain corrupted filesystem.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch fixes to report the right error number of f2fs_find_entry to
its caller.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This mount option is to enable original log-structured filesystem forcefully.
So, there should be no random writes for main area.
Especially, this supports host-managed SMR device.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch reduces to call them across the whole tree.
- sync_inode_page()
- update_inode_page()
- update_inode()
- f2fs_write_inode()
Instead, checkpoint will flush all the dirty inode metadata before syncing
node pages.
Note that, this is doable, since we call mark_inode_dirty_sync() for all
inode's field change which needs to update on-disk inode as well.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
We don't need to use f2fs_bug_on() to treat with any error case when allocating
a block during recovery.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
When testing f2fs with inline_dentry option, generic/342 reports:
VFS: Busy inodes after unmount of dm-0. Self-destruct in 5 seconds. Have a nice day...
After rmmod f2fs module, kenrel shows following dmesg:
=============================================================================
BUG f2fs_inode_cache (Tainted: G O ): Objects remaining in f2fs_inode_cache on __kmem_cache_shutdown()
-----------------------------------------------------------------------------
Disabling lock debugging due to kernel taint
INFO: Slab 0xf51ca0e0 objects=22 used=1 fp=0xd1e6fc60 flags=0x40004080
CPU: 3 PID: 7455 Comm: rmmod Tainted: G B O 4.6.0-rc4+ #16
Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006
00000086 00000086 d062fe18 c13a83a0 f51ca0e0 d062fe38 d062fea4 c11c7276
c1981040 f51ca0e0 00000016 00000001 d1e6fc60 40004080 656a624f 20737463
616d6572 6e696e69 6e692067 66326620 6e695f73 5f65646f 68636163 6e6f2065
Call Trace:
[<c13a83a0>] dump_stack+0x5f/0x8f
[<c11c7276>] slab_err+0x76/0x80
[<c11cbfc0>] ? __kmem_cache_shutdown+0x100/0x2f0
[<c11cbfc0>] ? __kmem_cache_shutdown+0x100/0x2f0
[<c11cbfe5>] __kmem_cache_shutdown+0x125/0x2f0
[<c1198a38>] kmem_cache_destroy+0x158/0x1f0
[<c176b43d>] ? mutex_unlock+0xd/0x10
[<f8f15aa3>] exit_f2fs_fs+0x4b/0x5a8 [f2fs]
[<c10f596c>] SyS_delete_module+0x16c/0x1d0
[<c1001b10>] ? do_fast_syscall_32+0x30/0x1c0
[<c13c59bf>] ? __this_cpu_preempt_check+0xf/0x20
[<c10afa7d>] ? trace_hardirqs_on_caller+0xdd/0x210
[<c10ad50b>] ? trace_hardirqs_off+0xb/0x10
[<c1001b81>] do_fast_syscall_32+0xa1/0x1c0
[<c176d888>] sysenter_past_esp+0x45/0x74
INFO: Object 0xd1e6d9e0 @offset=6624
kmem_cache_destroy f2fs_inode_cache: Slab cache still has objects
CPU: 3 PID: 7455 Comm: rmmod Tainted: G B O 4.6.0-rc4+ #16
Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006
00000286 00000286 d062fef4 c13a83a0 f174b000 d062ff14 d062ff28 c1198ac7
c197fe18 f3c5b980 d062ff20 000d04f2 d062ff0c d062ff0c d062ff14 d062ff14
f8f20dc0 fffffff5 d062e000 d062ff30 f8f15aa3 d062ff7c c10f596c 73663266
Call Trace:
[<c13a83a0>] dump_stack+0x5f/0x8f
[<c1198ac7>] kmem_cache_destroy+0x1e7/0x1f0
[<f8f15aa3>] exit_f2fs_fs+0x4b/0x5a8 [f2fs]
[<c10f596c>] SyS_delete_module+0x16c/0x1d0
[<c1001b10>] ? do_fast_syscall_32+0x30/0x1c0
[<c13c59bf>] ? __this_cpu_preempt_check+0xf/0x20
[<c10afa7d>] ? trace_hardirqs_on_caller+0xdd/0x210
[<c10ad50b>] ? trace_hardirqs_off+0xb/0x10
[<c1001b81>] do_fast_syscall_32+0xa1/0x1c0
[<c176d888>] sysenter_past_esp+0x45/0x74
The reason is: in recovery flow, we use delayed iput mechanism for directory
which has recovered dentry block. It means the reference of inode will be
held until last dirty dentry page being writebacked.
But when we mount f2fs with inline_dentry option, during recovery, dirent
may only be recovered into dir inode page rather than dentry page, so there
are no chance for us to release inode reference in ->writepage when
writebacking last dentry page.
We can call paired iget/iput explicityly for inline_dentry case, but for
non-inline_dentry case, iput will call writeback_single_inode to write all
data pages synchronously, but during recovery, ->writepages of f2fs skips
writing all pages, result in losing dirent.
This patch fixes this issue by obsoleting old mechanism, and introduce a
new dir_list to hold all directory inodes which has recovered datas until
finishing recovery.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
In find_fsync_dnodes, get_tmp_page will read dnode page synchronously,
previously, ra_meta_page did the same work, which is redundant, remove
it.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
In order to give atomic writes, we should consider power failure during
sync_node_pages in fsync.
So, this patch marks fsync flag only in the last dnode block.
Acked-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time
ago with promise that one day it will be possible to implement page
cache with bigger chunks than PAGE_SIZE.
This promise never materialized. And unlikely will.
We have many places where PAGE_CACHE_SIZE assumed to be equal to
PAGE_SIZE. And it's constant source of confusion on whether
PAGE_CACHE_* or PAGE_* constant should be used in a particular case,
especially on the border between fs and mm.
Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much
breakage to be doable.
Let's stop pretending that pages in page cache are special. They are
not.
The changes are pretty straight-forward:
- <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN};
- page_cache_get() -> get_page();
- page_cache_release() -> put_page();
This patch contains automated changes generated with coccinelle using
script below. For some reason, coccinelle doesn't patch header files.
I've called spatch for them manually.
The only adjustment after coccinelle is revert of changes to
PAGE_CAHCE_ALIGN definition: we are going to drop it later.
There are few places in the code where coccinelle didn't reach. I'll
fix them manually in a separate patch. Comments and documentation also
will be addressed with the separate patch.
virtual patch
@@
expression E;
@@
- E << (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
expression E;
@@
- E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
@@
- PAGE_CACHE_SHIFT
+ PAGE_SHIFT
@@
@@
- PAGE_CACHE_SIZE
+ PAGE_SIZE
@@
@@
- PAGE_CACHE_MASK
+ PAGE_MASK
@@
expression E;
@@
- PAGE_CACHE_ALIGN(E)
+ PAGE_ALIGN(E)
@@
expression E;
@@
- page_cache_get(E)
+ get_page(E)
@@
expression E;
@@
- page_cache_release(E)
+ put_page(E)
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
f2fs support atomic write with following semantics:
1. open db file
2. ioctl start atomic write
3. (write db file) * n
4. ioctl commit atomic write
5. close db file
With this flow we can avoid file becoming corrupted when abnormal power
cut, because we hold data of transaction in referenced pages linked in
inmem_pages list of inode, but without setting them dirty, so these data
won't be persisted unless we commit them in step 4.
But we should still hold journal db file in memory by using volatile
write, because our semantics of 'atomic write support' is incomplete, in
step 4, we could fail to submit all dirty data of transaction, once
partial dirty data was committed in storage, then after a checkpoint &
abnormal power-cut, db file will be corrupted forever.
So this patch tries to improve atomic write flow by adding a revoking flow,
once inner error occurs in committing, this gives another chance to try to
revoke these partial submitted data of current transaction, it makes
committing operation more like aotmical one.
If we're not lucky, once revoking operation was failed, EAGAIN will be
reported to user for suggesting doing the recovery with held journal file,
or retrying current transaction again.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
There are redundant pointer conversion in following call stack:
- at position a, inode was been converted to f2fs_file_info.
- at position b, f2fs_file_info was been converted to inode again.
- truncate_blocks(inode,..)
- fi = F2FS_I(inode) ---a
- ADDRS_PER_PAGE(node_page, fi)
- addrs_per_inode(fi)
- inode = &fi->vfs_inode ---b
- f2fs_has_inline_xattr(inode)
- fi = F2FS_I(inode)
- is_inode_flag_set(fi,..)
In order to avoid unneeded conversion, alter ADDRS_PER_PAGE and
addrs_per_inode to acept parameter with type of inode pointer.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
In write_begin, if storage supports stable_page, we don't need to wait for
writeback to update its contents.
This patch introduces to use wait_for_stable_page instead of
wait_on_page_writeback.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
do_checkpoint and write_checkpoint can fail due to reasons like triggering
in a readonly fs or encountering IO error of storage device.
So it's better to report such error info to user, let user be aware of
failure of doing checkpoint.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
In recover_data, value of argument 'type' will be CURSEG_WARM_NODE all
the time, remove it for cleanup.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
If device does not support discard, some obsolete dnodes can be recovered
by roll-forward. This patch enhances the recovery flow.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Now, we use ra_meta_pages to reads continuous physical blocks as much as
possible to improve performance of following reads. However, ra_meta_pages
uses a synchronous readahead approach by submitting bio with READ, as READ
is with high priority, it can not be used in the case of preloading blocks,
and it's not sure when these RAed pages will be used.
This patch supports asynchronous readahead in ra_meta_pages by tagging bio
with READA flag in order to allow preloading.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
In recovery or checkpoint flow, we grab pages temperarily in meta inode's
mapping for caching temperary data, actually, datas in these pages were
not meta data of f2fs, but still we tag them with REQ_META flag. However,
lower device like eMMC may do some optimization for data of such type.
So in order to avoid wrong optimization, we'd better remove such flag
for temperary non-meta pages.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Protecting recovery flow by using cp_rwsem is not needed, since we have
prevent triggering any checkpoint by locking cp_mutex previously.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
We have potential overflow issue when calculating size of object, when
we left shift index with PAGE_CACHE_SHIFT bits, if type of index has only
32-bits space in 32-bit architecture, left shifting will incur overflow,
i.e:
pgoff_t index = 0xFFFFFFFF;
loff_t size = index << PAGE_CACHE_SHIFT;
size: 0xFFFFF000
So we should cast index with 64-bits type to avoid this issue.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
We should not write node pages when deleting orphan inodes.
In order to do that, we can eaisly set POR_DOING flag earlier before entering
orphan inode routine.
Reviewed-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
When testing with generic/101 in xfstests, error message outputed as below:
--- tests/generic/101.out
+++ results//generic/101.out.bad
@@ -10,10 +10,14 @@
File foo content after log replay:
0000000 aa aa aa aa aa aa aa aa aa aa aa aa aa aa aa aa
*
-0200000 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
+0200000 bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb
*
0372000
...
(Run 'diff -u tests/generic/101.out results/generic/101.out.bad' to see the entire diff)
The test flow is like below:
1. pwrite foo -S 0xaa 0 64K
2. pwrite foo -S 0xbb 64K 61K
3. sync
4. truncate foo 64K
5. truncate foo 125K
6. fsync foo
7. flakey drop writes
8. umount
After this test, we expect the data of recovered file will have the first
64k of data filling with value 0xaa and the next 61k of data filling with
value 0x00 because we have fsynced it before dropping writes in dm.
In f2fs, during recovering, we will only recover the valid block address
in direct node page if it is marked as a fsynced dnode, but block address
which means invalid/reserved (with value NULL_ADDR/NEW_ADDR) will not be
recovered. So, the file recovered shows its incorrect data 0xbb in range of
[61k, 125k].
In this patch, we fix to recover invalid/reserved block during recover flow.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
To avoid meeting garbage data in next free node block at the end of warm
node chain when doing recovery, we will try to zero out that invalid block.
If the device is not support discard, our way for zeroing out block is:
grabbing a temporary zeroed page in meta inode, then, issue write request
with this page.
But, we forget to release that temporary page, so our memory usage will
increase without gaining any hit ratio benefit, so it's better to free it
for saving memory.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch clean up codes through:
1.rename f2fs_replace_block to __f2fs_replace_block().
2.introduce new f2fs_replace_block() to include __f2fs_replace_block()
and some common related codes around __f2fs_replace_block().
Then, newly introduced function f2fs_replace_block can be used by
following patch.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch adds a bit flag to indicate whether or not i_name in the inode
is encrypted.
If this name is encrypted, we can't do recover_dentry during roll-forward.
So, f2fs_sync_file() needs to do checkpoint, if this will be needed in future.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Introduce a generic function replace_block base on recover_data_page,
and export it. So with it we can operate file's meta data which is in
CP/SSA area when we invoke fallocate with FALLOC_FL_COLLAPSE_RANGE
flag.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Export is_valid_blkaddr() and use it to replace some codes for readability.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch adds CP_RECOVERY to remain recovery information for checkpoint.
And, it makes sure writing checkpoint in this case.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
During the roll-forward recovery, if we found a new data index written fsync
lastly, we need to recover new block address.
But, if that address was corrupted, we should not recover that.
Otherwise, f2fs gets kernel panic from:
In check_index_in_prev_nodes(),
sentry = get_seg_entry(sbi, segno);
--------------------------> out-of-range segno.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
If there are multiple fsynced dnodes having a dent flag, roll-forward routine
sets FI_INC_LINK for their inode, and recovery_dentry increases its link count
accordingly.
That results in normal file having a link count as 2, so we can't unlink those
files.
This was added to handle several inode blocks having same inode number with
different directory paths.
But, current f2fs doesn't replay all of path changes and only recover its dentry
for the last fsynced inode block.
So, there is no reason to do this.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Jaegeuk Kim