From 1afc708dcae41bc538fda0f85d4f0ecf276fa505 Mon Sep 17 00:00:00 2001 From: Filipe Manana Date: Wed, 14 Oct 2020 10:10:36 +0100 Subject: btrfs: fix relocation failure due to race with fallocate When doing a fallocate() we have a short time window, after reserving an extent and before starting a transaction, where if relocation for the block group containing the reserved extent happens, we can end up missing the extent in the data relocation inode causing relocation to fail later. This only happens when we don't pass a transaction to the internal fallocate function __btrfs_prealloc_file_range(), which is for all the cases where fallocate() is called from user space (the internal use cases include space cache extent allocation and relocation). When the race triggers the relocation failure, it produces a trace like the following: [200611.995995] ------------[ cut here ]------------ [200611.997084] BTRFS: Transaction aborted (error -2) [200611.998208] WARNING: CPU: 3 PID: 235845 at fs/btrfs/ctree.c:1074 __btrfs_cow_block+0x3a0/0x5b0 [btrfs] [200611.999042] Modules linked in: dm_thin_pool dm_persistent_data (...) [200612.003287] CPU: 3 PID: 235845 Comm: btrfs Not tainted 5.9.0-rc6-btrfs-next-69 #1 [200612.004442] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014 [200612.006186] RIP: 0010:__btrfs_cow_block+0x3a0/0x5b0 [btrfs] [200612.007110] Code: 1b 00 00 02 72 2a 83 f8 fb 0f 84 b8 01 (...) [200612.007341] BTRFS warning (device sdb): Skipping commit of aborted transaction. [200612.008959] RSP: 0018:ffffaee38550f918 EFLAGS: 00010286 [200612.009672] BTRFS: error (device sdb) in cleanup_transaction:1901: errno=-30 Readonly filesystem [200612.010428] RAX: 0000000000000000 RBX: ffff9174d96f4000 RCX: 0000000000000000 [200612.011078] BTRFS info (device sdb): forced readonly [200612.011862] RDX: 0000000000000001 RSI: ffffffffa8161978 RDI: 00000000ffffffff [200612.013215] RBP: ffff9172569a0f80 R08: 0000000000000000 R09: 0000000000000000 [200612.014263] R10: 0000000000000000 R11: 0000000000000000 R12: ffff9174b8403b88 [200612.015203] R13: ffff9174b8400a88 R14: ffff9174c90f1000 R15: ffff9174a5a60e08 [200612.016182] FS: 00007fa55cf878c0(0000) GS:ffff9174ece00000(0000) knlGS:0000000000000000 [200612.017174] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [200612.018418] CR2: 00007f8fb8048148 CR3: 0000000428a46003 CR4: 00000000003706e0 [200612.019510] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [200612.020648] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 [200612.021520] Call Trace: [200612.022434] btrfs_cow_block+0x10b/0x250 [btrfs] [200612.023407] do_relocation+0x54e/0x7b0 [btrfs] [200612.024343] ? do_raw_spin_unlock+0x4b/0xc0 [200612.025280] ? _raw_spin_unlock+0x29/0x40 [200612.026200] relocate_tree_blocks+0x3bc/0x6d0 [btrfs] [200612.027088] relocate_block_group+0x2f3/0x600 [btrfs] [200612.027961] btrfs_relocate_block_group+0x15e/0x340 [btrfs] [200612.028896] btrfs_relocate_chunk+0x38/0x110 [btrfs] [200612.029772] btrfs_balance+0xb22/0x1790 [btrfs] [200612.030601] ? btrfs_ioctl_balance+0x253/0x380 [btrfs] [200612.031414] btrfs_ioctl_balance+0x2cf/0x380 [btrfs] [200612.032279] btrfs_ioctl+0x620/0x36f0 [btrfs] [200612.033077] ? _raw_spin_unlock+0x29/0x40 [200612.033948] ? handle_mm_fault+0x116d/0x1ca0 [200612.034749] ? up_read+0x18/0x240 [200612.035542] ? __x64_sys_ioctl+0x83/0xb0 [200612.036244] __x64_sys_ioctl+0x83/0xb0 [200612.037269] do_syscall_64+0x33/0x80 [200612.038190] entry_SYSCALL_64_after_hwframe+0x44/0xa9 [200612.038976] RIP: 0033:0x7fa55d07ed87 [200612.040127] Code: 00 00 00 48 8b 05 09 91 0c 00 64 c7 00 26 (...) [200612.041669] RSP: 002b:00007ffd5ebf03e8 EFLAGS: 00000206 ORIG_RAX: 0000000000000010 [200612.042437] RAX: ffffffffffffffda RBX: 0000000000000001 RCX: 00007fa55d07ed87 [200612.043511] RDX: 00007ffd5ebf0470 RSI: 00000000c4009420 RDI: 0000000000000003 [200612.044250] RBP: 0000000000000003 R08: 000055d8362642a0 R09: 00007fa55d148be0 [200612.044963] R10: fffffffffffff52e R11: 0000000000000206 R12: 00007ffd5ebf1614 [200612.045683] R13: 00007ffd5ebf0470 R14: 0000000000000002 R15: 00007ffd5ebf0470 [200612.046361] irq event stamp: 0 [200612.047040] hardirqs last enabled at (0): [<0000000000000000>] 0x0 [200612.047725] hardirqs last disabled at (0): [] copy_process+0x823/0x1bc0 [200612.048387] softirqs last enabled at (0): [] copy_process+0x823/0x1bc0 [200612.049024] softirqs last disabled at (0): [<0000000000000000>] 0x0 [200612.049722] ---[ end trace 49006c6876e65227 ]--- The race happens like this: 1) Task A starts an fallocate() (plain or zero range) and it calls __btrfs_prealloc_file_range() with the 'trans' parameter set to NULL; 2) Task A calls btrfs_reserve_extent() and gets an extent that belongs to block group X; 3) Before task A gets into btrfs_replace_file_extents(), through the call to insert_prealloc_file_extent(), task B starts relocation of block group X; 4) Task B enters btrfs_relocate_block_group() and it sets block group X to RO mode; 5) Task B enters relocate_block_group(), it calls prepare_to_relocate() whichs joins/starts a transaction and then commits the transaction; 6) Task B then starts scanning the extent tree looking for extents that belong to block group X - it does not find yet the extent reserved by task A, since that extent was not yet added to the extent tree, as its delayed reference was not even yet created at this point; 7) The data relocation inode ends up not having the extent reserved by task A associated to it; 8) Task A then starts a transaction through btrfs_replace_file_extents(), inserts a file extent item in the subvolume tree pointing to the reserved extent and creates a delayed reference for it; 9) Task A finishes and returns success to user space; 10) Later on, while relocation is still in progress, the leaf where task A inserted the new file extent item is COWed, so we end up at __btrfs_cow_block(), which calls btrfs_reloc_cow_block(), and that in turn calls relocation.c:replace_file_extents(); 11) At relocation.c:replace_file_extents() we iterate over all the items in the leaf and find the file extent item pointing to the extent that was allocated by task A, and then call relocation.c:get_new_location(), to find the new location for the extent; 12) However relocation.c:get_new_location() fails, returning -ENOENT, because it couldn't find a corresponding file extent item associated with the data relocation inode. This is because the extent was not seen in the extent tree at step 6). The -ENOENT error is propagated to __btrfs_cow_block(), which aborts the transaction. So fix this simply by decrementing the block group's number of reservations after calling insert_prealloc_file_extent(), as relocation waits for that counter to go down to zero before calling prepare_to_relocate() and start looking for extents in the extent tree. This issue only started to happen recently as of commit 8fccebfa534c79 ("btrfs: fix metadata reservation for fallocate that leads to transaction aborts"), because now we can reserve an extent before starting/joining a transaction, and previously we always did it after that, so relocation ended up waiting for a concurrent fallocate() to finish because before searching for the extents of the block group, it starts/joins a transaction and then commits it (at prepare_to_relocate()), which made it wait for the fallocate task to complete first. Fixes: 8fccebfa534c79 ("btrfs: fix metadata reservation for fallocate that leads to transaction aborts") Reviewed-by: Josef Bacik Signed-off-by: Filipe Manana Signed-off-by: David Sterba --- fs/btrfs/inode.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 936c3137c646..da58c58ef9aa 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -9672,10 +9672,16 @@ static int __btrfs_prealloc_file_range(struct inode *inode, int mode, * clear_offset by our extent size. */ clear_offset += ins.offset; - btrfs_dec_block_group_reservations(fs_info, ins.objectid); last_alloc = ins.offset; trans = insert_prealloc_file_extent(trans, inode, &ins, cur_offset); + /* + * Now that we inserted the prealloc extent we can finally + * decrement the number of reservations in the block group. + * If we did it before, we could race with relocation and have + * relocation miss the reserved extent, making it fail later. + */ + btrfs_dec_block_group_reservations(fs_info, ins.objectid); if (IS_ERR(trans)) { ret = PTR_ERR(trans); btrfs_free_reserved_extent(fs_info, ins.objectid, -- cgit v1.2.3 From c334730988ee07908ba4eb816ce78d3fe06fecaa Mon Sep 17 00:00:00 2001 From: Filipe Manana Date: Wed, 4 Nov 2020 11:07:31 +0000 Subject: btrfs: fix missing delalloc new bit for new delalloc ranges When doing a buffered write, through one of the write family syscalls, we look for ranges which currently don't have allocated extents and set the 'delalloc new' bit on them, so that we can report a correct number of used blocks to the stat(2) syscall until delalloc is flushed and ordered extents complete. However there are a few other places where we can do a buffered write against a range that is mapped to a hole (no extent allocated) and where we do not set the 'new delalloc' bit. Those places are: - Doing a memory mapped write against a hole; - Cloning an inline extent into a hole starting at file offset 0; - Calling btrfs_cont_expand() when the i_size of the file is not aligned to the sector size and is located in a hole. For example when cloning to a destination offset beyond EOF. So after such cases, until the corresponding delalloc range is flushed and the respective ordered extents complete, we can report an incorrect number of blocks used through the stat(2) syscall. In some cases we can end up reporting 0 used blocks to stat(2), which is a particular bad value to report as it may mislead tools to think a file is completely sparse when its i_size is not zero, making them skip reading any data, an undesired consequence for tools such as archivers and other backup tools, as reported a long time ago in the following thread (and other past threads): https://lists.gnu.org/archive/html/bug-tar/2016-07/msg00001.html Example reproducer: $ cat reproducer.sh #!/bin/bash MNT=/mnt/sdi DEV=/dev/sdi mkfs.btrfs -f $DEV > /dev/null # mkfs.xfs -f $DEV > /dev/null # mkfs.ext4 -F $DEV > /dev/null # mkfs.f2fs -f $DEV > /dev/null mount $DEV $MNT xfs_io -f -c "truncate 64K" \ -c "mmap -w 0 64K" \ -c "mwrite -S 0xab 0 64K" \ -c "munmap" \ $MNT/foo blocks_used=$(stat -c %b $MNT/foo) echo "blocks used: $blocks_used" if [ $blocks_used -eq 0 ]; then echo "ERROR: blocks used is 0" fi umount $DEV $ ./reproducer.sh blocks used: 0 ERROR: blocks used is 0 So move the logic that decides to set the 'delalloc bit' bit into the function btrfs_set_extent_delalloc(), since that is what we use for all those missing cases as well as for the cases that currently work well. This change is also preparatory work for an upcoming patch that fixes other problems related to tracking and reporting the number of bytes used by an inode. CC: stable@vger.kernel.org # 4.19+ Reviewed-by: Josef Bacik Signed-off-by: Filipe Manana Signed-off-by: David Sterba --- fs/btrfs/file.c | 57 ------------------------------------------- fs/btrfs/inode.c | 58 ++++++++++++++++++++++++++++++++++++++++++++ fs/btrfs/tests/inode-tests.c | 12 ++++++--- 3 files changed, 66 insertions(+), 61 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 87355a38a654..4373da7bcc0d 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -452,46 +452,6 @@ static void btrfs_drop_pages(struct page **pages, size_t num_pages) } } -static int btrfs_find_new_delalloc_bytes(struct btrfs_inode *inode, - const u64 start, - const u64 len, - struct extent_state **cached_state) -{ - u64 search_start = start; - const u64 end = start + len - 1; - - while (search_start < end) { - const u64 search_len = end - search_start + 1; - struct extent_map *em; - u64 em_len; - int ret = 0; - - em = btrfs_get_extent(inode, NULL, 0, search_start, search_len); - if (IS_ERR(em)) - return PTR_ERR(em); - - if (em->block_start != EXTENT_MAP_HOLE) - goto next; - - em_len = em->len; - if (em->start < search_start) - em_len -= search_start - em->start; - if (em_len > search_len) - em_len = search_len; - - ret = set_extent_bit(&inode->io_tree, search_start, - search_start + em_len - 1, - EXTENT_DELALLOC_NEW, - NULL, cached_state, GFP_NOFS); -next: - search_start = extent_map_end(em); - free_extent_map(em); - if (ret) - return ret; - } - return 0; -} - /* * after copy_from_user, pages need to be dirtied and we need to make * sure holes are created between the current EOF and the start of @@ -528,23 +488,6 @@ int btrfs_dirty_pages(struct btrfs_inode *inode, struct page **pages, EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0, cached); - if (!btrfs_is_free_space_inode(inode)) { - if (start_pos >= isize && - !(inode->flags & BTRFS_INODE_PREALLOC)) { - /* - * There can't be any extents following eof in this case - * so just set the delalloc new bit for the range - * directly. - */ - extra_bits |= EXTENT_DELALLOC_NEW; - } else { - err = btrfs_find_new_delalloc_bytes(inode, start_pos, - num_bytes, cached); - if (err) - return err; - } - } - err = btrfs_set_extent_delalloc(inode, start_pos, end_of_last_block, extra_bits, cached); if (err) diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index da58c58ef9aa..7e8d8169779d 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2253,11 +2253,69 @@ static int add_pending_csums(struct btrfs_trans_handle *trans, return 0; } +static int btrfs_find_new_delalloc_bytes(struct btrfs_inode *inode, + const u64 start, + const u64 len, + struct extent_state **cached_state) +{ + u64 search_start = start; + const u64 end = start + len - 1; + + while (search_start < end) { + const u64 search_len = end - search_start + 1; + struct extent_map *em; + u64 em_len; + int ret = 0; + + em = btrfs_get_extent(inode, NULL, 0, search_start, search_len); + if (IS_ERR(em)) + return PTR_ERR(em); + + if (em->block_start != EXTENT_MAP_HOLE) + goto next; + + em_len = em->len; + if (em->start < search_start) + em_len -= search_start - em->start; + if (em_len > search_len) + em_len = search_len; + + ret = set_extent_bit(&inode->io_tree, search_start, + search_start + em_len - 1, + EXTENT_DELALLOC_NEW, + NULL, cached_state, GFP_NOFS); +next: + search_start = extent_map_end(em); + free_extent_map(em); + if (ret) + return ret; + } + return 0; +} + int btrfs_set_extent_delalloc(struct btrfs_inode *inode, u64 start, u64 end, unsigned int extra_bits, struct extent_state **cached_state) { WARN_ON(PAGE_ALIGNED(end)); + + if (start >= i_size_read(&inode->vfs_inode) && + !(inode->flags & BTRFS_INODE_PREALLOC)) { + /* + * There can't be any extents following eof in this case so just + * set the delalloc new bit for the range directly. + */ + extra_bits |= EXTENT_DELALLOC_NEW; + } else { + int ret; + + ret = btrfs_find_new_delalloc_bytes(inode, start, + end + 1 - start, + cached_state); + if (ret) + return ret; + } + return set_extent_delalloc(&inode->io_tree, start, end, extra_bits, cached_state); } diff --git a/fs/btrfs/tests/inode-tests.c b/fs/btrfs/tests/inode-tests.c index e6719f7db386..04022069761d 100644 --- a/fs/btrfs/tests/inode-tests.c +++ b/fs/btrfs/tests/inode-tests.c @@ -983,7 +983,8 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize) ret = clear_extent_bit(&BTRFS_I(inode)->io_tree, BTRFS_MAX_EXTENT_SIZE >> 1, (BTRFS_MAX_EXTENT_SIZE >> 1) + sectorsize - 1, - EXTENT_DELALLOC | EXTENT_UPTODATE, 0, 0, NULL); + EXTENT_DELALLOC | EXTENT_DELALLOC_NEW | + EXTENT_UPTODATE, 0, 0, NULL); if (ret) { test_err("clear_extent_bit returned %d", ret); goto out; @@ -1050,7 +1051,8 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize) ret = clear_extent_bit(&BTRFS_I(inode)->io_tree, BTRFS_MAX_EXTENT_SIZE + sectorsize, BTRFS_MAX_EXTENT_SIZE + 2 * sectorsize - 1, - EXTENT_DELALLOC | EXTENT_UPTODATE, 0, 0, NULL); + EXTENT_DELALLOC | EXTENT_DELALLOC_NEW | + EXTENT_UPTODATE, 0, 0, NULL); if (ret) { test_err("clear_extent_bit returned %d", ret); goto out; @@ -1082,7 +1084,8 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize) /* Empty */ ret = clear_extent_bit(&BTRFS_I(inode)->io_tree, 0, (u64)-1, - EXTENT_DELALLOC | EXTENT_UPTODATE, 0, 0, NULL); + EXTENT_DELALLOC | EXTENT_DELALLOC_NEW | + EXTENT_UPTODATE, 0, 0, NULL); if (ret) { test_err("clear_extent_bit returned %d", ret); goto out; @@ -1097,7 +1100,8 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize) out: if (ret) clear_extent_bit(&BTRFS_I(inode)->io_tree, 0, (u64)-1, - EXTENT_DELALLOC | EXTENT_UPTODATE, 0, 0, NULL); + EXTENT_DELALLOC | EXTENT_DELALLOC_NEW | + EXTENT_UPTODATE, 0, 0, NULL); iput(inode); btrfs_free_dummy_root(root); btrfs_free_dummy_fs_info(fs_info); -- cgit v1.2.3