This round introduces several interesting features such as on-disk NAT bitmaps,
IO alignment, and a discard thread. And it includes a couple of major bug fixes
as below.
== Enhancement ==
- introduce on-disk bitmaps to avoid scanning NAT blocks when getting free nids
- support IO alignment to prepare open-channel SSD integration in future
- introduce a discard thread to avoid long latency during checkpoint and fstrim
- use SSR for warm node and enable inline_xattr by default
- introduce in-memory bitmaps to check FS consistency for debugging
- improve write_begin by avoiding needless read IO
== Bug fix ==
- fix broken zone_reset behavior for SMR drive
- fix wrong victim selection policy during GC
- fix missing behavior when preparing discard commands
- fix bugs in atomic write support and fiemap
- workaround to handle multiple f2fs_add_link calls having same name
And it includes a bunch of clean-up patches as well.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2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=tw8c
-----END PGP SIGNATURE-----
Merge tag 'for-f2fs-4.11' of git://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs
Pull f2fs updates from Jaegeuk Kim:
"This round introduces several interesting features such as on-disk NAT
bitmaps, IO alignment, and a discard thread. And it includes a couple
of major bug fixes as below.
Enhancements:
- introduce on-disk bitmaps to avoid scanning NAT blocks when getting
free nids
- support IO alignment to prepare open-channel SSD integration in
future
- introduce a discard thread to avoid long latency during checkpoint
and fstrim
- use SSR for warm node and enable inline_xattr by default
- introduce in-memory bitmaps to check FS consistency for debugging
- improve write_begin by avoiding needless read IO
Bug fixes:
- fix broken zone_reset behavior for SMR drive
- fix wrong victim selection policy during GC
- fix missing behavior when preparing discard commands
- fix bugs in atomic write support and fiemap
- workaround to handle multiple f2fs_add_link calls having same name
... and it includes a bunch of clean-up patches as well"
* tag 'for-f2fs-4.11' of git://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs: (97 commits)
f2fs: avoid to flush nat journal entries
f2fs: avoid to issue redundant discard commands
f2fs: fix a plint compile warning
f2fs: add f2fs_drop_inode tracepoint
f2fs: Fix zoned block device support
f2fs: remove redundant set_page_dirty()
f2fs: fix to enlarge size of write_io_dummy mempool
f2fs: fix memory leak of write_io_dummy mempool during umount
f2fs: fix to update F2FS_{CP_}WB_DATA count correctly
f2fs: use MAX_FREE_NIDS for the free nids target
f2fs: introduce free nid bitmap
f2fs: new helper cur_cp_crc() getting crc in f2fs_checkpoint
f2fs: update the comment of default nr_pages to skipping
f2fs: drop the duplicate pval in f2fs_getxattr
f2fs: Don't update the xattr data that same as the exist
f2fs: kill __is_extent_same
f2fs: avoid bggc->fggc when enough free segments are avaliable after cp
f2fs: select target segment with closer temperature in SSR mode
f2fs: show simple call stack in fault injection message
f2fs: no need lock_op in f2fs_write_inline_data
...
This patch adds a missing condition which flushes nat journal entries
unnecessarily introduced by:
f2fs: add bitmaps for empty or full NAT blocks
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
If segs_per_sec is over 1 like under SMR, previously f2fs issues discard
commands redundantly on the same section, since we didn't move end position
for the previous discard command.
E.g.,
start end
| |
prefree_bitmap = [01111100111100]
And, after issue discard for this section,
end start
| |
prefree_bitmap = [01111100111100]
Select this section again by searching from (end + 1),
start end
| |
prefree_bitmap = [01111100111100]
Fixes: 36abef4e796d38 ("f2fs: introduce mode=lfs mount option")
Cc: <stable@vger.kernel.org>
Cc: Damien Le Moal <damien.lemoal@wdc.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
fix such pclint warning:
...
Loss of precision (arg. no. 2) (unsigned long long to unsigned int))
Signed-off-by: Hou Pengyang <houpengyang@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
The introduction of the multi-device feature partially broke the support
for zoned block devices. In the function f2fs_scan_devices, sbi->devs
allocation and initialization is skipped in the case of a single device
mount. This result in no device information structure being allocated
for the device. This is fine if the device is a regular device, but in
the case of a zoned block device, the device zone type array is not
initialized, which causes the function __f2fs_issue_discard_zone to fail
as get_blkz_type is unable to determine the zone type of a section.
Fix this by always allocating and initializing the sbi->devs device
information array even in the case of a single device if that device is
zoned. For this particular case, make sure to obtain a reference on the
single device so that the call to blkdev_put() in destroy_device_list
operates as expected.
Fixes: 3c62be17d4f562f4 ("f2fs: support multiple devices")
Cc: <stable@vger.kernel.org> # v4.10
Signed-off-by: Masato Suzuki <masato.suzuki@wdc.com>
Acked-by: Damien Le Moal <damien.lemoal@wdc.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch remove redundant set_page_dirty in truncate_blocks
Signed-off-by: Yunlei He <heyunlei@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
It needs to double cache size of write_io_dummy mempool, otherwise we may
run out of cache in scenraio of Data/Node IOs were issued concurrently.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
We should only account F2FS_{CP_}WB_DATA IOs for write path, fix it.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
F2FS has define MAX_FREE_NIDS for maximum of cached free nids target.
Signed-off-by: Kinglong Mee <kinglongmee@gmail.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
In scenario of intensively node allocation, free nids will be ran out
soon, then it needs to stop to load free nids by traversing NAT blocks,
in worse case, if NAT blocks does not be cached in memory, it generates
IOs which slows down our foreground operations.
In order to speed up node allocation, in this patch we introduce a new
free_nid_bitmap array, so there is an bitmap table for each NAT block,
Once the NAT block is loaded, related bitmap cache will be switched on,
and bitmap will be set during traversing nat entries in NAT block, later
we can query and update nid usage status in memory completely.
With such implementation, I expect performance of node allocation can be
improved in the long-term after filesystem image is mounted.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
There are four places that getting the crc value in f2fs_checkpoint,
just add a new helper cur_cp_crc for them.
Signed-off-by: Kinglong Mee <kinglongmee@gmail.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
f2fs removes the old xattr data and appends the new data although
the new data is same as the exist.
Signed-off-by: Kinglong Mee <kinglongmee@gmail.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Since commit ee6d182f2a19 ("f2fs: remove syncing inode page in all the
cases") delayed inode element updating from inode cache to node page
cache, so once largest cached extent is updated, we can make inode dirty
immediately instead of checking and updating it in the end of extent
cache update.
The above commit didn't clean up unneeded codes in extent_cache.c, let's
finish the job in this patch.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
We use has_not_enough_free_secs to check if there are enough free segments,
(free_sections(sbi) + freed) <=
(node_secs + 2 * dent_secs + imeta_secs +
reserved_sections(sbi) + needed);
Under scenario with large number of dirty nodes, these nodes would be flushed
during cp, as a result, right side of the inequality would be decreased, while
left side stays unchanged if these nodes are flushed in SSR way, which means
there are enough free segments after this cp.
For this case, we just do a bggc instead of fggc.
Signed-off-by: Hou Pengyang <houpengyang@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
In SSR mode, we can allocate target segment which has different
temperature type from the type of current block, in order to avoid
mixing coldest and hottest data/node as much as possible, change
SSR allocation policy to select closer temperature for current
block prior.
Signed-off-by: Yunlong Song <yunlong.song@huawei.com>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Previously kernel message can show that in which function we do the
injection, but unfortunately, most of the caller are the same, for
tracking more information of injection path, it needs to show upper
caller's name. This patch supports that ability.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Similar as f2fs_write_inode, f2fs_write_inline_data just
mark inode page dirty, so it's no need to write inline data
under read lock of cp_rwsem.
Signed-off-by: Yunlei He <heyunlei@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patches adds bitmaps to represent empty or full NAT blocks containing
free nid entries.
If we can find valid crc|cp_ver in the last block of checkpoint pack, we'll
use these bitmaps when building free nids. In order to avoid checkpointing
burden, up-to-date bitmaps will be flushed only during umount time. So,
normally we can get this gain, but when power-cut happens, we rely on fsck.f2fs
which recovers this bitmap again.
After this patch, we build free nids from nid #0 at mount time to make more
full NAT blocks, but in runtime, we check empty NAT blocks to load free nids
without loading any NAT pages from disk.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch replace rw semaphore extent_tree_lock with mutex lock
for no read cases with this lock.
Signed-off-by: Yunlei He <heyunlei@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
When more than one data blocks are allocated, the F2FS_MAP_UNWRITTEN/MAPPED
flags will be overlapped by F2FS_MAP_NEW at the later times.
Signed-off-by: Kinglong Mee <kinglongmee@gmail.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
proc A: proc B:
- writeback_sb_inodes
- __writeback_single_inode
- do_writepages
- f2fs_write_node_pages
- f2fs_balance_fs_bg - write_checkpoint
- build_free_nids - flush_nat_entries
- __build_free_nids - __flush_nat_entry_set
- ra_meta_pages - get_next_nat_page
- current_nat_addr - set_to_next_nat
[do nat_bitmap checking] - f2fs_change_bit
For proc A, nat_bitmap and nat_bitmap_mir would be compared without lock_op and
nm_i->nat_tree_lock, while proc B is changing nat_bitmap/nat_bitmap_ver in cp.
So it is normal for nat_bitmap/nat_bitmap diffrence under such scenario.
This patch fix this by removing the monitoring point.
[Fix: 599a09b f2fs: check in-memory nat version bitmap]
Signed-off-by: Hou Pengyang <houpengyang@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Since has_not_enough_free_secs(sbi, 0, 0) must be true if has_not_enough_
free_secs(sbi, sec_freed, 0) is true, write_checkpoint is sure to execute in
both conditions.
Signed-off-by: Yunlong Song <yunlong.song@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
For converntional zones, we don't need to align discard commands to exact zone
size.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
We don't need to wait for each discard commands when unmounting the image.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
We have a kernel thread to issue discard commands, so we can increase the
number of batched discard sections. By default, now it becomes 4GB range.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch adds MAX_DISCARD_BLOCKS() to avoid issuing too much large single
discard command.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
->fault(), ->page_mkwrite(), and ->pfn_mkwrite() calls do not need to
take a vma and vmf parameter when the vma already resides in vmf.
Remove the vma parameter to simplify things.
[arnd@arndb.de: fix ARM build]
Link: http://lkml.kernel.org/r/20170125223558.1451224-1-arnd@arndb.de
Link: http://lkml.kernel.org/r/148521301778.19116.10840599906674778980.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Cc: Theodore Ts'o <tytso@mit.edu>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Matthew Wilcox <mawilcox@microsoft.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Jan Kara <jack@suse.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Previously, if type is CURSEG_HOT_DATA, we only check CURSEG_HOT_DATA only.
This patch fixes to search all the different types for SSR.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
In allocate_segment_by_default(), need_SSR() already detected it's time to do
SSR. So, let's try to find victims for data segments more aggressively in time.
Signed-off-by: Yunlong Song <yunlong.song@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
As data segment gc may lead dnode dirty, so the greedy cost for data segment
should be valid blocks * 2, that is data segment is prior to node segment.
Signed-off-by: Hou Pengyang <houpengyang@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
SIT information should be updated before segment allocation, since SSR needs
latest valid block information. Current code does not update the old_blkaddr
info in sit_entry, so adjust the allocate_segment to its proper location. Commit
5e443818fa0b2a2845561ee25bec181424fb2889 ("f2fs: handle dirty segments inside
refresh_sit_entry") puts it into wrong location.
Signed-off-by: Yunlong Song <yunlong.song@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
For foreground gc, greedy algorithm should be adapted, which makes
this formula work well:
(2 * (100 / config.overprovision + 1) + 6)
But currently, we fg_gc have a prior to select bg_gc victim segments to gc
first, these victims are selected by cost-benefit algorithm, we can't guarantee
such segments have the small valid blocks, which may destroy the f2fs rule, on
the worstest case, would consume all the free segments.
This patch fix this by add a filter in check_bg_victims, if segment's has # of
valid blocks over overprovision ratio, skip such segments.
Cc: <stable@vger.kernel.org>
Signed-off-by: Hou Pengyang <houpengyang@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
It turns out a stakable filesystem like sdcardfs in AOSP can trigger multiple
vfs_create() to lower filesystem. In that case, f2fs will add multiple dentries
having same name which breaks filesystem consistency.
Until upper layer fixes, let's work around by f2fs, which shows actually not
much performance regression.
Cc: <stable@vger.kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
We have had node chains, but haven't used it so far due to stale node blocks.
Now, we have crc|cp_ver in node footer and give random cp_ver at format time,
we can start to use it again.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
In android, since SElinux is enable, security policy will be appliedd for
each file, it stores in inode as an xattr entry, so it will take one 4k
size node block additionally for each file.
Let's enable inline_xattr by default in order to save storage space.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch introduces new mount option 'noinline_xattr', so we can disable
inline xattr functionality which is already set as a default mount option.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
We've not seen this buggy case for a long time, so it's time to avoid this
unnecessary get_node_info() call which reading NAT page to cache nat entry.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
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>
Declare fscrypt_operations structure as const as it is only stored in
the s_cop field of a super_block structure. This field is of type const,
so fscrypt_operations structure having this property can be made const
too.
File size before: fs/f2fs/super.o
text data bss dec hex filename
54131 31355 184 85670 14ea6 fs/f2fs/super.o
File size after: fs/f2fs/super.o
text data bss dec hex filename
54227 31259 184 85670 14ea6 fs/f2fs/super.o
Signed-off-by: Bhumika Goyal <bhumirks@gmail.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
