- 16 Jan, 2018 8 commits
-
-
Yufen Yu authored
Cgroup writeback requires explicit support from the filesystem. f2fs's data and node writeback IOs go through __write_data_page, which sets fio for submiting IOs. So, we add io_wbc for fio, associate bios with blkcg by invoking wbc_init_bio() and account IOs issuing by wbc_account_io(). In addtion, f2fs_fill_super() is updated to set SB_I_CGROUPWB. Meta writeback IOs is left alone by this patch and will always be attributed to the root cgroup. The results show that f2fs can throttle writeback nicely for data writing and file creating. Reviewed-by:
Chao Yu <yuchao0@huawei.com> Signed-off-by:
Yufen Yu <yuyufen@huawei.com> Signed-off-by:
Jaegeuk Kim <jaegeuk@kernel.org>
-
Chao Yu authored
In commit 78997b56 ("f2fs: split discard policy"), we have get rid of using pend_list_tag field in struct discard_cmd_control, but forgot to remove it, now do it. Signed-off-by:
-
Chao Yu authored
We take very long time to finish generic/476, this is because we will check consistence of all discard entries in global rb tree while traversing all different granularity pending lists, even when the list is empty, in order to avoid that unneeded overhead, we have to skip the check when coming up an empty list. generic/476 time consumption: cost Before patch & w/o consistence check 57s Before patch & w/ consistence check 1426s After patch 78s Signed-off-by:
-
Wei Yongjun authored
Fixes the following sparse warnings: fs/f2fs/segment.c:887:6: warning: symbol '__check_sit_bitmap' was not declared. Should it be static? fs/f2fs/segment.c:1327:6: warning: symbol 'f2fs_wait_discard_bio' was not declared. Should it be static? fs/f2fs/super.c:1661:5: warning: symbol 'f2fs_get_projid' was not declared. Should it be static? Signed-off-by:
Wei Yongjun <weiyongjun1@huawei.com> Reviewed-by:
-
Jaegeuk Kim authored
This patch allows root to reserve some blocks via mount option. "-o reserve_root=N" means N x 4KB-sized blocks for root only. Reviewed-by:
-
Yunlong Song authored
In some case, the node blocks has wrong blkaddr whose segment type is NODE, e.g., recover inode has missing xattr flag and the blkaddr is in the xattr range. Since fsck.f2fs does not check the recovery nodes, this will cause __f2fs_replace_block change the curseg of node and do the update_sit_entry(sbi, new_blkaddr, 1) with no next_blkoff refresh, as a result, when recovery process write checkpoint and sync nodes, the next_blkoff of curseg is used in the segment bit map, then it will cause f2fs_bug_on. So let's check segment type in __f2fs_replace_block. Signed-off-by:
Yunlong Song <yunlong.song@huawei.com> Reviewed-by:
-
Yunlei He authored
After checkpoint, 1. creat a new file A ,(with dirty inode && dirty inode page && xattr info) 2. backgroud wb write back file A inode page (without update from inode cache) 3. fsync file A, write back inode page of file A with inode cache info 4. sudden power off before new checkpoint In this case, recovery process will try to recover a zero inode page. Inline xattr flag of file A will be miss and xattr info will be taken as blkaddr index. Signed-off-by:
Yunlei He <heyunlei@huawei.com> Reviewed-by:
-
Jaegeuk Kim authored
Let's show precise # of blocks that user/root can use through bavail and bfree respectively. Reviewed-by:
-
- 04 Jan, 2018 3 commits
-
-
Chao Yu authored
Commit 6afc662e ("f2fs: support flexible inline xattr size") declared f2fs_sb_has_flexible_inline_xattr in f2fs.h for latter being used in get_inline_xattr_addrs, but in latter version, related code has been changed, leave f2fs_sb_has_flexible_inline_xattr w/o any users. Let's remove it for cleanup. Signed-off-by:
-
Chao Yu authored
While doing direct IO, if we run out-of-space when we preallocate blocks, we should not return ENOSPC error directly, instead, we should continue to do following direct IO, which will keep directIO of f2fs acting like other filesystems. Signed-off-by:
-
Jaegeuk Kim authored
We have to enable quota only when remounting from read to write. Otherwise, we'll get remount failure. (e.g., write to write case) Reviewed-by:
-
- 03 Jan, 2018 29 commits
-
-
Jaegeuk Kim authored
We can give another chance to write user data, which can resolve generic/441. Reviewed-by:
-
Jaegeuk Kim authored
This fixes generic/449 hang problem caused by no ENOSPC forever which should be returned by setxattr under disk full scenario. Reviewed-by:
-
Jaegeuk Kim authored
This fixes generic/342 which doesn't recover renamed file which was fsynced before. It will be done via another fsync on newly created file. Reviewed-by:
-
Jaegeuk Kim authored
Let's avoid BUG_ON during fill_super, when on-disk was totall corrupted. Reviewed-by:
-
Yunlei He authored
-Thread A Thread B -write_checkpoint -block_operations -f2fs_unlock_all -f2fs_sync_file -f2fs_write_inode -f2fs_inode_synced -f2fs_sync_inode_meta -sync_node_pages -set_page_drity In this case, if sudden power off without next new checkpoint, the last inode page update will lost. wb_writeback is same with fsync. Yunlei also reproduced the bug by: @@ -366,7 +366,7 @@ int update_inode(struct inode *inode, struct page *node_page) struct extent_tree *et = F2FS_I(inode)->extent_tree; f2fs_inode_synced(inode); - + msleep(10000); f2fs_wait_on_page_writeback(node_page, NODE, true); shell 1: shell2: dd if=/dev/zero of=./test bs=1M count=10 sync echo "hello" >> ./test fsync test // sleep 10s sync //return quickly echo c > /proc/sysrq-trigger Signed-off-by: -
Chao Yu authored
As Jia-Ju Bai reported: "According to fs/f2fs/trace.c, the kernel module may sleep under a spinlock. The function call path is: f2fs_trace_pid (acquire the spinlock) f2fs_radix_tree_insert cond_resched --> may sleep I do not find a good way to fix it, so I only report. This possible bug is found by my static analysis tool (DSAC) and my code review." Obviously, it's problemetic to schedule in critical region of spinlock, which will cause uninterruptable sleep if there is no waker. This patch changes to use mutex lock intead of spinlock to avoid this condition. Reported-by:Jia-Ju Bai <baijiaju1990@gmail.com> Signed-off-by:
-
Yunlei He authored
No need return value in restore summary process Signed-off-by:
-
LiFan authored
Since the variable release is only nonzero when another unlikely case occurs, use unlikely() on it seems logical. Signed-off-by:
Fan li <fanofcode.li@samsung.com> Reviewed-by:
-
Chao Yu authored
There is no caller cares about return value of truncate_data_blocks_range, remove it. Signed-off-by:
-
Chao Yu authored
f2fs_map_blocks(): if (blkaddr == NEW_ADDR || blkaddr == NULL_ADDR) { if (create) { ... } else { ... if (flag == F2FS_GET_BLOCK_FIEMAP && blkaddr == NULL_ADDR) { ... } if (flag != F2FS_GET_BLOCK_FIEMAP || blkaddr != NEW_ADDR) goto sync_out; } It means we can break the loop in cases of: a) flag != F2FS_GET_BLOCK_FIEMAP or b) flag == F2FS_GET_BLOCK_FIEMAP && blkaddr == NULL_ADDR Condition b) is the same as previous one, so merge operations of them for readability. Signed-off-by: -
Chao Yu authored
f2fs_chksum and f2fs_crc32 use the same 'crc32' crypto engine, also their implementation are almost the same, except with different shash description context. Introduce __f2fs_crc32 to wrap the common codes, and reuse it in f2fs_chksum and f2fs_crc32. Signed-off-by:
-
Chao Yu authored
In fill_super, if we fail to call f2fs_build_stats(), it needs to detach from global f2fs shrink list, otherwise once system starts to shrink slab cache, we will encounter below panic: BUG: unable to handle kernel paging request at 00007d35 Oops: 0002 [#1] PREEMPT SMP EIP: __lock_acquire+0x70/0x12c0 Call Trace: lock_acquire+0xae/0x220 mutex_trylock+0xc5/0xf0 f2fs_shrink_count+0x32/0xb0 [f2fs] shrink_slab+0xf1/0x5b0 drop_slab_node+0x35/0x60 drop_slab+0xf/0x20 drop_caches_sysctl_handler+0x79/0xc0 proc_sys_call_handler+0xa4/0xc0 proc_sys_write+0x1f/0x30 __vfs_write+0x24/0x150 SyS_write+0x44/0x90 do_fast_syscall_32+0xa1/0x1ca entry_SYSENTER_32+0x4c/0x7b In addition, this patch relocates f2fs_join_shrinker in fill_super to avoid unneeded error handling of it. Signed-off-by:
-
Chao Yu authored
Use f2fs_k{m,z}alloc as much as possible to increase fault injection points. Signed-off-by: -
Chao Yu authored
This patch supports to inject fault into kvmalloc/kvzalloc. Signed-off-by:
-
Chao Yu authored
This patch introduces f2fs_kzalloc based on f2fs_kmalloc in order to support error injection for kzalloc(). Signed-off-by:
-
LiFan authored
Avoid checking is_inode repeatedly, and make the logic a little bit clearer. Signed-off-by:
-
Hyunchul Lee authored
When blocks are allocated for direct write, select the type of segment using the kiocb hint. But if an inode has FI_NO_ALLOC, use the inode hint. Signed-off-by:
Hyunchul Lee <cheol.lee@lge.com> Reviewed-by:
-
Eric Biggers authored
Reviewed-by:
Eric Biggers <ebiggers@google.com> Signed-off-by:
-
Eric Biggers authored
Reviewed-by:
-
Eric Biggers authored
Reviewed-by:
-
Eric Biggers authored
Reviewed-by:
-
Eric Biggers authored
Reviewed-by:
-
Elena Reshetova authored
atomic_t variables are currently used to implement reference counters with the following properties: - counter is initialized to 1 using atomic_set() - a resource is freed upon counter reaching zero - once counter reaches zero, its further increments aren't allowed - counter schema uses basic atomic operations (set, inc, inc_not_zero, dec_and_test, etc.) Such atomic variables should be converted to a newly provided refcount_t type and API that prevents accidental counter overflows and underflows. This is important since overflows and underflows can lead to use-after-free situation and be exploitable. The variable posix_acl.a_refcount is used as pure reference counter. Convert it to refcount_t and fix up the operations. **Important note for maintainers: Some functions from refcount_t API defined in lib/refcount.c have different memory ordering guarantees than their atomic counterparts. The full comparison can be seen in https://lkml.org/lkml/2017/11/15/57 and it is hopefully soon in state to be merged to the documentation tree. Normally the differences should not matter since refcount_t provides enough guarantees to satisfy the refcounting use cases, but in some rare cases it might matter. Please double check that you don't have some undocumented memory guarantees for this variable usage. For the posix_acl.a_refcount it might make a difference in following places: - get_cached_acl(): increment in refcount_inc_not_zero() only guarantees control dependency on success vs. fully ordered atomic counterpart. However this operation is performed under rcu_read_lock(), so this should be fine. - posix_acl_release(): decrement in refcount_dec_and_test() only provides RELEASE ordering and control dependency on success vs. fully ordered atomic counterpart Suggested-by:
Kees Cook <keescook@chromium.org> Reviewed-by:
David Windsor <dwindsor@gmail.com> Reviewed-by:
Hans Liljestrand <ishkamiel@gmail.com> Signed-off-by:
Elena Reshetova <elena.reshetova@intel.com> Signed-off-by:
-
Zhikang Zhang authored
f2fs: remove repeated f2fs_bug_on which has already existed in function invalidate_blocks. Signed-off-by:Zhikang Zhang <zhangzhikang1@huawei.com> Reviewed-by:
-
LiFan authored
Remove the variable page_idx which no one would miss. Signed-off-by:
-
Chao Yu authored
test/generic/208 reports a potential deadlock as below: Chain exists of: &mm->mmap_sem --> &fi->i_mmap_sem --> &fi->dio_rwsem[WRITE] Possible unsafe locking scenario: CPU0 CPU1 ---- ---- lock(&fi->dio_rwsem[WRITE]); lock(&fi->i_mmap_sem); lock(&fi->dio_rwsem[WRITE]); lock(&mm->mmap_sem); This patch changes the lock dependency as below in fallocate() to fix this issue: - dio_rwsem - i_mmap_sem Fixes: bb06664a ("f2fs: avoid race in between GC and block exchange") Signed-off-by: -
Sheng Yong authored
Commit d260081c ("f2fs: change recovery policy of xattr node block") removes the use of blkaddr, which is no longer used. So remove the parameter. Signed-off-by:
Sheng Yong <shengyong1@huawei.com> Reviewed-by:
-
Sheng Yong authored
If there is not enough space left, f2fs_preallocate_blocks may only preallocte partial blocks. As a result, the write operation fails but i_blocks is not 0. To avoid this, f2fs should write data in non-preallocation way and write as many data as the size of i_blocks. Signed-off-by:
-
Sheng Yong authored
This patch introduces a sysfs interface readdir_ra to enable/disable readaheading inode block in f2fs_readdir. When readdir_ra is enabled, it improves the performance of "readdir + stat". For 300,000 files: time find /data/test > /dev/null disable readdir_ra: 1m25.69s real 0m01.94s user 0m50.80s system enable readdir_ra: 0m18.55s real 0m00.44s user 0m15.39s system Signed-off-by:
-
