- 01 Oct, 2022 18 commits
-
-
Jason Yan authored
Factor out ext4_fast_commit_init(). No functional change. Signed-off-by:
Jason Yan <yanaijie@huawei.com> Reviewed-by:
Jan Kara <jack@suse.cz> Reviewed-by:
Ritesh Harjani (IBM) <ritesh.list@gmail.com> Link: https://lore.kernel.org/r/20220916141527.1012715-6-yanaijie@huawei.comSigned-off-by:
Theodore Ts'o <tytso@mit.edu>
-
Jason Yan authored
Factor out ext4_handle_clustersize(). No functional change. Signed-off-by:
-
Jason Yan authored
Factor out ext4_set_def_opts(). No functional change. Signed-off-by:
-
Jason Yan authored
The 'cantfind_ext4' error handler is just a error msg print and then goto failed_mount. This two level goto makes the code complex and not easy to read. The only benefit is that is saves a little bit code. However some branches can merge and some branches dot not even need it. So do some refactor and remove it. Signed-off-by:
-
Jason Yan authored
Before these two branches neither loaded the journal nor created the xattr cache. So the right label to goto is 'failed_mount3a'. Although this did not cause any issues because the error handler validated if the pointer is null. However this still made me confused when reading the code. So it's still worth to modify to goto the right label. Signed-off-by:
-
Ye Bin authored
If fastcommit is already disabled, there isn't need to mark inode ineligible. So move 'ext4_fc_disabled()' judgement bofore 'ext4_should_journal_data(inode)' judgement which can avoid to do meaningless judgement. Signed-off-by:
Ye Bin <yebin10@huawei.com> Reviewed-by:
-
Ye Bin authored
Factor out ext4_fc_disabled(). No functional change. Signed-off-by:
-
Ye Bin authored
In 'ext4_fc_write_inode' function first call 'ext4_get_inode_loc' get 'iloc', after use it miss release 'iloc.bh'. So just release 'iloc.bh' before 'ext4_fc_write_inode' return. Cc: stable@kernel.org Signed-off-by:
-
Ye Bin authored
In 'jbd2_fc_wait_bufs' use 'bh' after put buffer head reference count which may lead to use-after-free. So judge buffer if uptodate before put buffer head reference count. Cc: stable@kernel.org Signed-off-by:
-
Ye Bin authored
As in 'jbd2_fc_wait_bufs' if buffer isn't uptodate, will return -EIO without update 'journal->j_fc_off'. But 'jbd2_fc_release_bufs' will release buffer head from ‘j_fc_off - 1’ if 'bh' is NULL will terminal release which will lead to buffer head buffer head reference count leak. To solve above issue, update 'journal->j_fc_off' before return -EIO. Cc: stable@kernel.org Signed-off-by:
-
Zhihao Cheng authored
Following process may lead to fs corruption: 1. ext4_create(dir/foo) ext4_add_nondir ext4_add_entry ext4_dx_add_entry a. add_dirent_to_buf ext4_mark_inode_dirty ext4_handle_dirty_metadata // dir inode bh is recorded into journal b. ext4_append // dx_get_count(entries) == dx_get_limit(entries) ext4_bread(EXT4_GET_BLOCKS_CREATE) ext4_getblk ext4_map_blocks ext4_ext_map_blocks ext4_mb_new_blocks dquot_alloc_block dquot_alloc_space_nodirty inode_add_bytes // update dir's i_blocks ext4_ext_insert_extent ext4_ext_dirty // record extent bh into journal ext4_handle_dirty_metadata(bh) // record new block into journal inode->i_size += inode->i_sb->s_blocksize // new size(in mem) c. ext4_handle_dirty_dx_node(bh2) // record dir's new block(dx_node) into journal d. ext4_handle_dirty_dx_node((frame - 1)->bh) e. ext4_handle_dirty_dx_node(frame->bh) f. do_split // ret err! g. add_dirent_to_buf ext4_mark_inode_dirty(dir) // update raw_inode on disk(skipped) 2. fsck -a /dev/sdb drop last block(dx_node) which beyonds dir's i_size. /dev/sdb: recovering journal /dev/sdb contains a file system with errors, check forced. /dev/sdb: Inode 12, end of extent exceeds allowed value (logical block 128, physical block 3938, len 1) 3. fsck -fn /dev/sdb dx_node->entry[i].blk > dir->i_size Pass 2: Checking directory structure Problem in HTREE directory inode 12 (/dir): bad block number 128. Clear HTree index? no Problem in HTREE directory inode 12: block #3 has invalid depth (2) Problem in HTREE directory inode 12: block #3 has bad max hash Problem in HTREE directory inode 12: block #3 not referenced Fix it by marking inode dirty directly inside ext4_append(). Fetch a reproducer in [Link]. Link: https://bugzilla.kernel.org/show_bug.cgi?id=216466 Cc: stable@vger.kernel.org Signed-off-by:Zhihao Cheng <chengzhihao1@huawei.com> Reviewed-by:
-
Gaosheng Cui authored
ext4_inline_data_fiemap() has been removed since commit d3b6f23f ("ext4: move ext4_fiemap to use iomap framework"), so remove it. Signed-off-by:
Gaosheng Cui <cuigaosheng1@huawei.com> Reviewed-by:
-
Jeff Layton authored
ext4 currently updates the i_version counter when the atime is updated during a read. This is less than ideal as it can cause unnecessary cache invalidations with NFSv4 and unnecessary remeasurements for IMA. The increment in ext4_mark_iloc_dirty is also problematic since it can corrupt the i_version counter for ea_inodes. We aren't bumping the file times in ext4_mark_iloc_dirty, so changing the i_version there seems wrong, and is the cause of both problems. Remove that callsite and add increments to the setattr, setxattr and ioctl codepaths, at the same times that we update the ctime. The i_version bump that already happens during timestamp updates should take care of the rest. In ext4_move_extents, increment the i_version on both inodes, and also add in missing ctime updates. [ Some minor updates since we've already enabled the i_version counter unconditionally already via another patch series. -- TYT ] Cc: stable@kernel.org Cc: Lukas Czerner <lczerner@redhat.com> Reviewed-by:
Christian Brauner (Microsoft) <brauner@kernel.org> Signed-off-by:
Jeff Layton <jlayton@kernel.org> Link: https://lore.kernel.org/r/20220908172448.208585-3-jlayton@kernel.orgSigned-off-by:
-
Jan Kara authored
Commit 307af6c8 ("mbcache: automatically delete entries from cache on freeing") started nesting cache->c_list_lock under the bit locks protecting hash buckets of the mbcache hash table in mb_cache_entry_create(). This causes problems for real-time kernels because there spinlocks are sleeping locks while bitlocks stay atomic. Luckily the nesting is easy to avoid by holding entry reference until the entry is added to the LRU list. This makes sure we cannot race with entry deletion. Cc: stable@kernel.org Fixes: 307af6c8 ("mbcache: automatically delete entries from cache on freeing") Reported-by:
Mike Galbraith <efault@gmx.de> Signed-off-by:
-
Andrew Perepechko authored
LIFO wakeup order is unfair and sometimes leads to a journal user not being able to get a journal handle for hundreds of transactions in a row. FIFO wakeup can make things more fair. Cc: stable@kernel.org Signed-off-by:
Alexey Lyashkov <alexey.lyashkov@gmail.com> Reviewed-by:
-
Jinke Han authored
In our product environment, we encounter some jbd hung waiting handles to stop while several writters were doing memory reclaim for buffer head allocation in delay alloc write path. Ext4 do buffer head allocation with holding transaction handle which may be blocked too long if the reclaim works not so smooth. According to our bcc trace, the reclaim time in buffer head allocation can reach 258s and the jbd transaction commit also take almost the same time meanwhile. Except for these extreme cases, we often see several seconds delays for cgroup memory reclaim on our servers. This is more likely to happen considering docker environment. One thing to note, the allocation of buffer heads is as often as page allocation or more often when blocksize less than page size. Just like page cache allocation, we should also place the buffer head allocation before startting the handle. Cc: stable@kernel.org Signed-off-by:
Jinke Han <hanjinke.666@bytedance.com> Link: https://lore.kernel.org/r/20220903012429.22555-1-hanjinke.666@bytedance.comSigned-off-by:
-
Zhang Yi authored
Recently we notice that ext4 filesystem would occasionally fail to read metadata from disk and report error message, but the disk and block layer looks fine. After analyse, we lockon commit 88dbcbb3 ("blkdev: avoid migration stalls for blkdev pages"). It provide a migration method for the bdev, we could move page that has buffers without extra users now, but it lock the buffers on the page, which breaks the fragile metadata read operation on ext4 filesystem, ext4_read_bh_lock() was copied from ll_rw_block(), it depends on the assumption of that locked buffer means it is under IO. So it just trylock the buffer and skip submit IO if it lock failed, after wait_on_buffer() we conclude IO error because the buffer is not uptodate. This issue could be easily reproduced by add some delay just after buffer_migrate_lock_buffers() in __buffer_migrate_folio() and do fsstress on ext4 filesystem. EXT4-fs error (device pmem1): __ext4_find_entry:1658: inode #73193: comm fsstress: reading directory lblock 0 EXT4-fs error (device pmem1): __ext4_find_entry:1658: inode #75334: comm fsstress: reading directory lblock 0 Fix it by removing the trylock logic in ext4_read_bh_lock(), just lock the buffer and submit IO if it's not uptodate, and also leave over readahead helper. Cc: stable@kernel.org Signed-off-by:
Zhang Yi <yi.zhang@huawei.com> Reviewed-by:
-
Jeff Layton authored
The original i_version implementation was pretty expensive, requiring a log flush on every change. Because of this, it was gated behind a mount option (implemented via the MS_I_VERSION mountoption flag). Commit ae5e165d (fs: new API for handling inode->i_version) made the i_version flag much less expensive, so there is no longer a performance penalty from enabling it. xfs and btrfs already enable it unconditionally when the on-disk format can support it. Have ext4 ignore the SB_I_VERSION flag, and just enable it unconditionally. While we're in here, mark the i_version mount option Opt_removed. [ Removed leftover bits of i_version from ext4_apply_options() since it now can't ever be set in ctx->mask_s_flags -- lczerner ] Cc: stable@kernel.org Cc: Dave Chinner <david@fromorbit.com> Cc: Benjamin Coddington <bcodding@redhat.com> Cc: Christoph Hellwig <hch@infradead.org> Cc: Darrick J. Wong <djwong@kernel.org> Signed-off-by:
Lukas Czerner <lczerner@redhat.com> Reviewed-by:
-
- 30 Sep, 2022 8 commits
-
-
Lukas Czerner authored
Currently the I_DIRTY_TIME will never get set if the inode already has I_DIRTY_INODE with assumption that it supersedes I_DIRTY_TIME. That's true, however ext4 will only update the on-disk inode in ->dirty_inode(), not on actual writeback. As a result if the inode already has I_DIRTY_INODE state by the time we get to __mark_inode_dirty() only with I_DIRTY_TIME, the time was already filled into on-disk inode and will not get updated until the next I_DIRTY_INODE update, which might never come if we crash or get a power failure. The problem can be reproduced on ext4 by running xfstest generic/622 with -o iversion mount option. Fix it by allowing I_DIRTY_TIME to be set even if the inode already has I_DIRTY_INODE. Also make sure that the case is properly handled in writeback_single_inode() as well. Additionally changes in xfs_fs_dirty_inode() was made to accommodate for I_DIRTY_TIME in flag. Thanks Jan Kara for suggestions on how to make this work properly. Cc: Dave Chinner <david@fromorbit.com> Cc: Christoph Hellwig <hch@infradead.org> Cc: stable@kernel.org Signed-off-by:
-
Lukas Czerner authored
ea_inodes are using i_version for storing part of the reference count so we really need to leave it alone. The problem can be reproduced by xfstest ext4/026 when iversion is enabled. Fix it by not calling inode_inc_iversion() for EXT4_EA_INODE_FL inodes in ext4_mark_iloc_dirty(). Cc: stable@kernel.org Signed-off-by:
-
Jan Kara authored
The check in __ext4_read_dirblock() for block being outside of directory size was wrong because it compared block number against directory size in bytes. Fix it. Fixes: 65f8ea4c ("ext4: check if directory block is within i_size") CVE: CVE-2022-1184 CC: stable@vger.kernel.org Signed-off-by:
-
Ritesh Harjani (IBM) authored
submit_bh/submit_bh_wbc are non-blocking functions which just submit the bio and return. The caller of submit_bh/submit_bh_wbc needs to wait on buffer till I/O completion and then check buffer head's b_state field to know if there was any I/O error. Hence there is no need for these functions to have any return type. Even now they always returns 0. Hence drop the return value and make their return type as void to avoid any confusion. Reviewed-by:
Christoph Hellwig <hch@lst.de> Reviewed-by:
-
Ritesh Harjani (IBM) authored
submit_bh always returns 0. This patch drops the useless return value of submit_bh from __sync_dirty_buffer(). Once all of submit_bh callers are cleaned up, we can make it's return type as void. Signed-off-by:
-
Ritesh Harjani (IBM) authored
submit_bh always returns 0. This patch drops the useless return value of submit_bh from ntfs_submit_bh_for_read(). Once all of submit_bh callers are cleaned up, we can make it's return type as void. Reviewed-by:
kernel test robot <lkp@intel.com> Signed-off-by:
-
Ritesh Harjani (IBM) authored
submit_bh always returns 0. This patch cleans up 2 of it's caller in jbd2 to drop submit_bh's useless return value. Once all submit_bh callers are cleaned up, we can make it's return type as void. Reviewed-by:
-
Lalith Rajendran authored
ext4_lazyinit_thread is not set freezable. Hence when the thread calls try_to_freeze it doesn't freeze during suspend and continues to send requests to the storage during suspend, resulting in suspend failures. Cc: stable@kernel.org Signed-off-by:
Lalith Rajendran <lalithkraj@google.com> Link: https://lore.kernel.org/r/20220818214049.1519544-1-lalithkraj@google.comSigned-off-by:
-
- 29 Sep, 2022 4 commits
-
-
Baokun Li authored
I caught a null-ptr-deref bug as follows: ================================================================== KASAN: null-ptr-deref in range [0x0000000000000068-0x000000000000006f] CPU: 1 PID: 1589 Comm: umount Not tainted 5.10.0-02219-dirty #339 RIP: 0010:ext4_write_info+0x53/0x1b0 [...] Call Trace: dquot_writeback_dquots+0x341/0x9a0 ext4_sync_fs+0x19e/0x800 __sync_filesystem+0x83/0x100 sync_filesystem+0x89/0xf0 generic_shutdown_super+0x79/0x3e0 kill_block_super+0xa1/0x110 deactivate_locked_super+0xac/0x130 deactivate_super+0xb6/0xd0 cleanup_mnt+0x289/0x400 __cleanup_mnt+0x16/0x20 task_work_run+0x11c/0x1c0 exit_to_user_mode_prepare+0x203/0x210 syscall_exit_to_user_mode+0x5b/0x3a0 do_syscall_64+0x59/0x70 entry_SYSCALL_64_after_hwframe+0x44/0xa9 ================================================================== Above issue may happen as follows: ------------------------------------- exit_to_user_mode_prepare task_work_run __cleanup_mnt cleanup_mnt deactivate_super deactivate_locked_super kill_block_super generic_shutdown_super shrink_dcache_for_umount dentry = sb->s_root sb->s_root = NULL <--- Here set NULL sync_filesystem __sync_filesystem sb->s_op->sync_fs > ext4_sync_fs dquot_writeback_dquots sb->dq_op->write_info > ext4_write_info ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2) d_inode(sb->s_root) s_root->d_inode <--- Null pointer dereference To solve this problem, we use ext4_journal_start_sb directly to avoid s_root being used. Cc: stable@kernel.org Signed-off-by:Baokun Li <libaokun1@huawei.com> Reviewed-by:
-
Josh Triplett authored
On a read-only filesystem, we won't invoke the block allocator, so we don't need to prefetch the block bitmaps. This avoids starting and running the ext4lazyinit thread at all on a system with no read-write ext4 filesystems (for instance, a container VM with read-only filesystems underneath an overlayfs). Fixes: 21175ca4 ("ext4: make prefetch_block_bitmaps default") Signed-off-by:
Josh Triplett <josh@joshtriplett.org> Reviewed-by:
-
Yang Xu authored
These two options should have been removed since 3.5, but none notices it. Recently, I and Darrick found this. Also, have some discussion for this[1][2][3]. So now, let's remove them. Link: https://lore.kernel.org/linux-ext4/6258F7BB.8010104@fujitsu.com/T/#u[1] Link: https://lore.kernel.org/linux-ext4/20220602110421.ymoug3rwfspmryqg@fedora/T/#t[2] Link: https://lore.kernel.org/linux-ext4/08e2ca4c8f6344bdcd76d75b821116c6147fd57a.camel@kernel.org/T/#t[3]Signed-off-by:
Yang Xu <xuyang2018.jy@fujitsu.com> Reviewed-by:
-
Jan Kara authored
When inode is created and written to using direct IO, there is nothing to clear the EXT4_STATE_MAY_INLINE_DATA flag. Thus when inode gets truncated later to say 1 byte and written using normal write, we will try to store the data as inline data. This confuses the code later because the inode now has both normal block and inline data allocated and the confusion manifests for example as: kernel BUG at fs/ext4/inode.c:2721! invalid opcode: 0000 [#1] PREEMPT SMP KASAN CPU: 0 PID: 359 Comm: repro Not tainted 5.19.0-rc8-00001-g31ba1e3b8305-dirty #15 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.0-1.fc36 04/01/2014 RIP: 0010:ext4_writepages+0x363d/0x3660 RSP: 0018:ffffc90000ccf260 EFLAGS: 00010293 RAX: ffffffff81e1abcd RBX: 0000008000000000 RCX: ffff88810842a180 RDX: 0000000000000000 RSI: 0000008000000000 RDI: 0000000000000000 RBP: ffffc90000ccf650 R08: ffffffff81e17d58 R09: ffffed10222c680b R10: dfffe910222c680c R11: 1ffff110222c680a R12: ffff888111634128 R13: ffffc90000ccf880 R14: 0000008410000000 R15: 0000000000000001 FS: 00007f72635d2640(0000) GS:ffff88811b000000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000565243379180 CR3: 000000010aa74000 CR4: 0000000000150eb0 Call Trace: <TASK> do_writepages+0x397/0x640 filemap_fdatawrite_wbc+0x151/0x1b0 file_write_and_wait_range+0x1c9/0x2b0 ext4_sync_file+0x19e/0xa00 vfs_fsync_range+0x17b/0x190 ext4_buffered_write_iter+0x488/0x530 ext4_file_write_iter+0x449/0x1b90 vfs_write+0xbcd/0xf40 ksys_write+0x198/0x2c0 __x64_sys_write+0x7b/0x90 do_syscall_64+0x3d/0x90 entry_SYSCALL_64_after_hwframe+0x63/0xcd </TASK> Fix the problem by clearing EXT4_STATE_MAY_INLINE_DATA when we are doing direct IO write to a file. Cc: stable@kernel.org Reported-by:
Tadeusz Struk <tadeusz.struk@linaro.org> Reported-by: syzbot+bd13648a53ed6933ca49@syzkaller.appspotmail.com Link: https://syzkaller.appspot.com/bug?id=a1e89d09bbbcbd5c4cb45db230ee28c822953984Signed-off-by:
-
- 27 Sep, 2022 2 commits
-
-
Eric Whitney authored
Modify the error returns for two file types that can't be defragged to more clearly communicate those restrictions to a caller. When the defrag code is applied to swap files, return -ETXTBSY, and when applied to quota files, return -EOPNOTSUPP. Move an extent tree search whose results are only occasionally required to the site always requiring them for improved efficiency. Address a few typos. Signed-off-by:
Eric Whitney <enwlinux@gmail.com> Link: https://lore.kernel.org/r/20220722163910.268564-1-enwlinux@gmail.comSigned-off-by:
-
Jerry Lee 李修賢 authored
When expanding a file system from (16TiB-2MiB) to 18TiB, the operation exits early which leads to result inconsistency between resize2fs and Ext4 kernel driver. === before === ○ → resize2fs /dev/mapper/thin resize2fs 1.45.5 (07-Jan-2020) Filesystem at /dev/mapper/thin is mounted on /mnt/test; on-line resizing required old_desc_blocks = 2048, new_desc_blocks = 2304 The filesystem on /dev/mapper/thin is now 4831837696 (4k) blocks long. [ 865.186308] EXT4-fs (dm-5): mounted filesystem with ordered data mode. Opts: (null). Quota mode: none. [ 912.091502] dm-4: detected capacity change from 34359738368 to 38654705664 [ 970.030550] dm-5: detected capacity change from 34359734272 to 38654701568 [ 1000.012751] EXT4-fs (dm-5): resizing filesystem from 4294966784 to 4831837696 blocks [ 1000.012878] EXT4-fs (dm-5): resized filesystem to 4294967296 === after === [ 129.104898] EXT4-fs (dm-5): mounted filesystem with ordered data mode. Opts: (null). Quota mode: none. [ 143.773630] dm-4: detected capacity change from 34359738368 to 38654705664 [ 198.203246] dm-5: detected capacity change from 34359734272 to 38654701568 [ 207.918603] EXT4-fs (dm-5): resizing filesystem from 4294966784 to 4831837696 blocks [ 207.918754] EXT4-fs (dm-5): resizing filesystem from 4294967296 to 4831837696 blocks [ 207.918758] EXT4-fs (dm-5): Converting file system to meta_bg [ 207.918790] EXT4-fs (dm-5): resizing filesystem from 4294967296 to 4831837696 blocks [ 221.454050] EXT4-fs (dm-5): resized to 4658298880 blocks [ 227.634613] EXT4-fs (dm-5): resized filesystem to 4831837696 Signed-off-by:
Jerry Lee <jerrylee@qnap.com> Link: https://lore.kernel.org/r/PU1PR04MB22635E739BD21150DC182AC6A18C9@PU1PR04MB2263.apcprd04.prod.outlook.comSigned-off-by:
-
- 26 Sep, 2022 1 commit
-
-
Jan Kara authored
Variable 'grp' may be left uninitialized if there's no group with suitable average fragment size (or larger). Fix the problem by initializing it earlier. Link: https://lore.kernel.org/r/20220922091542.pkhedytey7wzp5fi@quack3 Fixes: 83e80a6e ("ext4: use buckets for cr 1 block scan instead of rbtree") Cc: stable@kernel.org Reported-by:
Dan Carpenter <dan.carpenter@oracle.com> Signed-off-by:
-
- 22 Sep, 2022 7 commits
-
-
Theodore Ts'o authored
This patch avoids threads live-locking for hours when a large number threads are competing over the last few free extents as they blocks getting added and removed from preallocation pools. From our bug reporter: A reliable way for triggering this has multiple writers continuously write() to files when the filesystem is full, while small amounts of space are freed (e.g. by truncating a large file -1MiB at a time). In the local filesystem, this can be done by simply not checking the return code of write (0) and/or the error (ENOSPACE) that is set. Over NFS with an async mount, even clients with proper error checking will behave this way since the linux NFS client implementation will not propagate the server errors [the write syscalls immediately return success] until the file handle is closed. This leads to a situation where NFS clients send a continuous stream of WRITE rpcs which result in ERRNOSPACE -- but since the client isn't seeing this, the stream of writes continues at maximum network speed. When some space does appear, multiple writers will all attempt to claim it for their current write. For NFS, we may see dozens to hundreds of threads that do this. The real-world scenario of this is database backup tooling (in particular, github.com/mdkent/percona-xtrabackup) which may write large files (>1TiB) to NFS for safe keeping. Some temporary files are written, rewound, and read back -- all before closing the file handle (the temp file is actually unlinked, to trigger automatic deletion on close/crash.) An application like this operating on an async NFS mount will not see an error code until TiB have been written/read. The lockup was observed when running this database backup on large filesystems (64 TiB in this case) with a high number of block groups and no free space. Fragmentation is generally not a factor in this filesystem (~thousands of large files, mostly contiguous except for the parts written while the filesystem is at capacity.) Signed-off-by: -
Luís Henriques authored
When walking through an inode extents, the ext4_ext_binsearch_idx() function assumes that the extent header has been previously validated. However, there are no checks that verify that the number of entries (eh->eh_entries) is non-zero when depth is > 0. And this will lead to problems because the EXT_FIRST_INDEX() and EXT_LAST_INDEX() will return garbage and result in this: [ 135.245946] ------------[ cut here ]------------ [ 135.247579] kernel BUG at fs/ext4/extents.c:2258! [ 135.249045] invalid opcode: 0000 [#1] PREEMPT SMP [ 135.250320] CPU: 2 PID: 238 Comm: tmp118 Not tainted 5.19.0-rc8+ #4 [ 135.252067] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.15.0-0-g2dd4b9b-rebuilt.opensuse.org 04/01/2014 [ 135.255065] RIP: 0010:ext4_ext_map_blocks+0xc20/0xcb0 [ 135.256475] Code: [ 135.261433] RSP: 0018:ffffc900005939f8 EFLAGS: 00010246 [ 135.262847] RAX: 0000000000000024 RBX: ffffc90000593b70 RCX: 0000000000000023 [ 135.264765] RDX: ffff8880038e5f10 RSI: 0000000000000003 RDI: ffff8880046e922c [ 135.266670] RBP: ffff8880046e9348 R08: 0000000000000001 R09: ffff888002ca580c [ 135.268576] R10: 0000000000002602 R11: 0000000000000000 R12: 0000000000000024 [ 135.270477] R13: 0000000000000000 R14: 0000000000000024 R15: 0000000000000000 [ 135.272394] FS: 00007fdabdc56740(0000) GS:ffff88807dd00000(0000) knlGS:0000000000000000 [ 135.274510] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 135.276075] CR2: 00007ffc26bd4f00 CR3: 0000000006261004 CR4: 0000000000170ea0 [ 135.277952] Call Trace: [ 135.278635] <TASK> [ 135.279247] ? preempt_count_add+0x6d/0xa0 [ 135.280358] ? percpu_counter_add_batch+0x55/0xb0 [ 135.281612] ? _raw_read_unlock+0x18/0x30 [ 135.282704] ext4_map_blocks+0x294/0x5a0 [ 135.283745] ? xa_load+0x6f/0xa0 [ 135.284562] ext4_mpage_readpages+0x3d6/0x770 [ 135.285646] read_pages+0x67/0x1d0 [ 135.286492] ? folio_add_lru+0x51/0x80 [ 135.287441] page_cache_ra_unbounded+0x124/0x170 [ 135.288510] filemap_get_pages+0x23d/0x5a0 [ 135.289457] ? path_openat+0xa72/0xdd0 [ 135.290332] filemap_read+0xbf/0x300 [ 135.291158] ? _raw_spin_lock_irqsave+0x17/0x40 [ 135.292192] new_sync_read+0x103/0x170 [ 135.293014] vfs_read+0x15d/0x180 [ 135.293745] ksys_read+0xa1/0xe0 [ 135.294461] do_syscall_64+0x3c/0x80 [ 135.295284] entry_SYSCALL_64_after_hwframe+0x46/0xb0 This patch simply adds an extra check in __ext4_ext_check(), verifying that eh_entries is not 0 when eh_depth is > 0. Link: https://bugzilla.kernel.org/show_bug.cgi?id=215941 Link: https://bugzilla.kernel.org/show_bug.cgi?id=216283 Cc: Baokun Li <libaokun1@huawei.com> Cc: stable@kernel.org Signed-off-by:
Luís Henriques <lhenriques@suse.de> Reviewed-by:
-
Jan Kara authored
Using rbtree for sorting groups by average fragment size is relatively expensive (needs rbtree update on every block freeing or allocation) and leads to wide spreading of allocations because selection of block group is very sentitive both to changes in free space and amount of blocks allocated. Furthermore selecting group with the best matching average fragment size is not necessary anyway, even more so because the variability of fragment sizes within a group is likely large so average is not telling much. We just need a group with large enough average fragment size so that we have high probability of finding large enough free extent and we don't want average fragment size to be too big so that we are likely to find free extent only somewhat larger than what we need. So instead of maintaing rbtree of groups sorted by fragment size keep bins (lists) or groups where average fragment size is in the interval [2^i, 2^(i+1)). This structure requires less updates on block allocation / freeing, generally avoids chaotic spreading of allocations into block groups, and still is able to quickly (even faster that the rbtree) provide a block group which is likely to have a suitably sized free space extent. This patch reduces number of block groups used when untarring archive with medium sized files (size somewhat above 64k which is default mballoc limit for avoiding locality group preallocation) to about half and thus improves write speeds for eMMC flash significantly. Fixes: 196e402a ("ext4: improve cr 0 / cr 1 group scanning") CC: stable@kernel.org Reported-and-tested-by:
Stefan Wahren <stefan.wahren@i2se.com> Tested-by:
Ojaswin Mujoo <ojaswin@linux.ibm.com> Signed-off-by:
-
Jan Kara authored
Curently we don't use any preallocation when a file is already closed when allocating blocks (from writeback code when converting delayed allocation). However for small files, using locality group preallocation is actually desirable as that is not specific to a particular file. Rather it is a method to pack small files together to reduce fragmentation and for that the fact the file is closed is actually even stronger hint the file would benefit from packing. So change the logic to allow locality group preallocation in this case. Fixes: 196e402a ("ext4: improve cr 0 / cr 1 group scanning") CC: stable@kernel.org Reported-and-tested-by:
-
Jan Kara authored
Currently the Orlov inode allocator searches for free inodes for a directory only in flex block groups with at most inodes_per_group/16 more directory inodes than average per flex block group. However with growing size of flex block group this becomes unnecessarily strict. Scale allowed difference from average directory count per flex block group with flex block group size as we do with other metrics. Tested-by:
-
Jan Kara authored
mb_set_largest_free_order() updates lists containing groups with largest chunk of free space of given order. The way it updates it leads to always moving the group to the tail of the list. Thus allocations looking for free space of given order effectively end up cycling through all groups (and due to initialization in last to first order). This spreads allocations among block groups which reduces performance for rotating disks or low-end flash media. Change mb_set_largest_free_order() to only update lists if the order of the largest free chunk in the group changed. Fixes: 196e402a ("ext4: improve cr 0 / cr 1 group scanning") CC: stable@kernel.org Reported-and-tested-by:
-
Jan Kara authored
One of the side-effects of mb_optimize_scan was that the optimized functions to select next group to try were called even before we tried the goal group. As a result we no longer allocate files close to corresponding inodes as well as we don't try to expand currently allocated extent in the same group. This results in reaim regression with workfile.disk workload of upto 8% with many clients on my test machine: baseline mb_optimize_scan Hmean disk-1 2114.16 ( 0.00%) 2099.37 ( -0.70%) Hmean disk-41 87794.43 ( 0.00%) 83787.47 * -4.56%* Hmean disk-81 148170.73 ( 0.00%) 135527.05 * -8.53%* Hmean disk-121 177506.11 ( 0.00%) 166284.93 * -6.32%* Hmean disk-161 220951.51 ( 0.00%) 207563.39 * -6.06%* Hmean disk-201 208722.74 ( 0.00%) 203235.59 ( -2.63%) Hmean disk-241 222051.60 ( 0.00%) 217705.51 ( -1.96%) Hmean disk-281 252244.17 ( 0.00%) 241132.72 * -4.41%* Hmean disk-321 255844.84 ( 0.00%) 245412.84 * -4.08%* Also this is causing huge regression (time increased by a factor of 5 or so) when untarring archive with lots of small files on some eMMC storage cards. Fix the problem by making sure we try goal group first. Fixes: 196e402a ("ext4: improve cr 0 / cr 1 group scanning") CC: stable@kernel.org Reported-and-tested-by:
-
