From 290237fde9491ca26cf4020bbf5a2b330452e7db Mon Sep 17 00:00:00 2001 From: David Disseldorp Date: Mon, 10 Feb 2025 22:17:29 +1100 Subject: [PATCH 1/6] btrfs: selftests: fix btrfs_test_delayed_refs() leak of transaction The btrfs_transaction struct leaks, which can cause sporadic fstests failures when kmemleak checking is enabled: kmemleak: 5 new suspected memory leaks (see /sys/kernel/debug/kmemleak) > cat /sys/kernel/debug/kmemleak unreferenced object 0xffff88810fdc6c00 (size 512): comm "modprobe", pid 203, jiffies 4294892552 hex dump (first 32 bytes): 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ backtrace (crc 6736050f): __kmalloc_cache_noprof+0x133/0x2c0 btrfs_test_delayed_refs+0x6f/0xbb0 [btrfs] btrfs_run_sanity_tests.cold+0x91/0xf9 [btrfs] 0xffffffffa02fd055 do_one_initcall+0x49/0x1c0 do_init_module+0x5b/0x1f0 init_module_from_file+0x70/0x90 idempotent_init_module+0xe8/0x2c0 __x64_sys_finit_module+0x6b/0xd0 do_syscall_64+0x54/0x110 entry_SYSCALL_64_after_hwframe+0x76/0x7e The transaction struct was initially stack-allocated but switched to heap following frame size compiler warnings. Fixes: 2b34879d97e27 ("btrfs: selftests: add delayed ref self test cases") Signed-off-by: David Disseldorp Reviewed-by: David Sterba Signed-off-by: David Sterba --- fs/btrfs/tests/delayed-refs-tests.c | 1 + 1 file changed, 1 insertion(+) diff --git a/fs/btrfs/tests/delayed-refs-tests.c b/fs/btrfs/tests/delayed-refs-tests.c index 6558508c2ddf..265370e79a54 100644 --- a/fs/btrfs/tests/delayed-refs-tests.c +++ b/fs/btrfs/tests/delayed-refs-tests.c @@ -1009,6 +1009,7 @@ int btrfs_test_delayed_refs(u32 sectorsize, u32 nodesize) if (!ret) ret = select_delayed_refs_test(&trans); + kfree(transaction); out_free_fs_info: btrfs_free_dummy_fs_info(fs_info); return ret; From 59f37036bb7ab3d554c24abc856aabca01126414 Mon Sep 17 00:00:00 2001 From: Filipe Manana Date: Sat, 15 Feb 2025 11:36:15 +0000 Subject: [PATCH 2/6] btrfs: fix use-after-free on inode when scanning root during em shrinking At btrfs_scan_root() we are accessing the inode's root (and fs_info) in a call to btrfs_fs_closing() after we have scheduled the inode for a delayed iput, and that can result in a use-after-free on the inode in case the cleaner kthread does the iput before we dereference the inode in the call to btrfs_fs_closing(). Fix this by using the fs_info stored already in a local variable instead of doing inode->root->fs_info. Fixes: 102044384056 ("btrfs: make the extent map shrinker run asynchronously as a work queue job") CC: stable@vger.kernel.org # 6.13+ Tested-by: Ivan Shapovalov Link: https://lore.kernel.org/linux-btrfs/0414d690ac5680d0d77dfc930606cdc36e42e12f.camel@intelfx.name/ Reviewed-by: Johannes Thumshirn Reviewed-by: Qu Wenruo Signed-off-by: Filipe Manana Signed-off-by: David Sterba --- fs/btrfs/extent_map.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index 67ce85ff0ae2..bee1b94a1049 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -1222,8 +1222,7 @@ static long btrfs_scan_root(struct btrfs_root *root, struct btrfs_em_shrink_ctx fs_info->em_shrinker_last_ino = btrfs_ino(inode); btrfs_add_delayed_iput(inode); - if (ctx->scanned >= ctx->nr_to_scan || - btrfs_fs_closing(inode->root->fs_info)) + if (ctx->scanned >= ctx->nr_to_scan || btrfs_fs_closing(fs_info)) break; cond_resched(); From c6c9c4d56483d941f567eb921434c25fc6086dfa Mon Sep 17 00:00:00 2001 From: Filipe Manana Date: Sat, 15 Feb 2025 11:04:15 +0000 Subject: [PATCH 3/6] btrfs: skip inodes without loaded extent maps when shrinking extent maps If there are inodes that don't have any loaded extent maps, we end up grabbing a reference on them and later adding a delayed iput, which wakes up the cleaner and makes it do unnecessary work. This is common when for example the inodes were open only to run stat(2) or all their extent maps were already released through the folio release callback (btrfs_release_folio()) or released by a previous run of the shrinker, or directories which never have extent maps. Reported-by: Ivan Shapovalov Tested-by: Ivan Shapovalov Link: https://lore.kernel.org/linux-btrfs/0414d690ac5680d0d77dfc930606cdc36e42e12f.camel@intelfx.name/ CC: stable@vger.kernel.org # 6.13+ Reviewed-by: Johannes Thumshirn Reviewed-by: Qu Wenruo Signed-off-by: Filipe Manana Signed-off-by: David Sterba --- fs/btrfs/extent_map.c | 78 +++++++++++++++++++++++++++++++------------ 1 file changed, 57 insertions(+), 21 deletions(-) diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index bee1b94a1049..8c6b85ffd18f 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -1128,6 +1128,8 @@ static long btrfs_scan_inode(struct btrfs_inode *inode, struct btrfs_em_shrink_c long nr_dropped = 0; struct rb_node *node; + lockdep_assert_held_write(&tree->lock); + /* * Take the mmap lock so that we serialize with the inode logging phase * of fsync because we may need to set the full sync flag on the inode, @@ -1139,28 +1141,12 @@ static long btrfs_scan_inode(struct btrfs_inode *inode, struct btrfs_em_shrink_c * to find new extents, which may not be there yet because ordered * extents haven't completed yet. * - * We also do a try lock because otherwise we could deadlock. This is - * because the shrinker for this filesystem may be invoked while we are - * in a path that is holding the mmap lock in write mode. For example in - * a reflink operation while COWing an extent buffer, when allocating - * pages for a new extent buffer and under memory pressure, the shrinker - * may be invoked, and therefore we would deadlock by attempting to read - * lock the mmap lock while we are holding already a write lock on it. + * We also do a try lock because we don't want to block for too long and + * we are holding the extent map tree's lock in write mode. */ if (!down_read_trylock(&inode->i_mmap_lock)) return 0; - /* - * We want to be fast so if the lock is busy we don't want to spend time - * waiting for it - either some task is about to do IO for the inode or - * we may have another task shrinking extent maps, here in this code, so - * skip this inode. - */ - if (!write_trylock(&tree->lock)) { - up_read(&inode->i_mmap_lock); - return 0; - } - node = rb_first(&tree->root); while (node) { struct rb_node *next = rb_next(node); @@ -1201,12 +1187,61 @@ static long btrfs_scan_inode(struct btrfs_inode *inode, struct btrfs_em_shrink_c break; node = next; } - write_unlock(&tree->lock); up_read(&inode->i_mmap_lock); return nr_dropped; } +static struct btrfs_inode *find_first_inode_to_shrink(struct btrfs_root *root, + u64 min_ino) +{ + struct btrfs_inode *inode; + unsigned long from = min_ino; + + xa_lock(&root->inodes); + while (true) { + struct extent_map_tree *tree; + + inode = xa_find(&root->inodes, &from, ULONG_MAX, XA_PRESENT); + if (!inode) + break; + + tree = &inode->extent_tree; + + /* + * We want to be fast so if the lock is busy we don't want to + * spend time waiting for it (some task is about to do IO for + * the inode). + */ + if (!write_trylock(&tree->lock)) + goto next; + + /* + * Skip inode if it doesn't have loaded extent maps, so we avoid + * getting a reference and doing an iput later. This includes + * cases like files that were opened for things like stat(2), or + * files with all extent maps previously released through the + * release folio callback (btrfs_release_folio()) or released in + * a previous run, or directories which never have extent maps. + */ + if (RB_EMPTY_ROOT(&tree->root)) { + write_unlock(&tree->lock); + goto next; + } + + if (igrab(&inode->vfs_inode)) + break; + + write_unlock(&tree->lock); +next: + from = btrfs_ino(inode) + 1; + cond_resched_lock(&root->inodes.xa_lock); + } + xa_unlock(&root->inodes); + + return inode; +} + static long btrfs_scan_root(struct btrfs_root *root, struct btrfs_em_shrink_ctx *ctx) { struct btrfs_fs_info *fs_info = root->fs_info; @@ -1214,9 +1249,10 @@ static long btrfs_scan_root(struct btrfs_root *root, struct btrfs_em_shrink_ctx long nr_dropped = 0; u64 min_ino = fs_info->em_shrinker_last_ino + 1; - inode = btrfs_find_first_inode(root, min_ino); + inode = find_first_inode_to_shrink(root, min_ino); while (inode) { nr_dropped += btrfs_scan_inode(inode, ctx); + write_unlock(&inode->extent_tree.lock); min_ino = btrfs_ino(inode) + 1; fs_info->em_shrinker_last_ino = btrfs_ino(inode); @@ -1227,7 +1263,7 @@ static long btrfs_scan_root(struct btrfs_root *root, struct btrfs_em_shrink_ctx cond_resched(); - inode = btrfs_find_first_inode(root, min_ino); + inode = find_first_inode_to_shrink(root, min_ino); } if (inode) { From 15b3b3254d1453a8db038b7d44b311a2d6c71f98 Mon Sep 17 00:00:00 2001 From: Filipe Manana Date: Sat, 15 Feb 2025 11:11:29 +0000 Subject: [PATCH 4/6] btrfs: do regular iput instead of delayed iput during extent map shrinking The extent map shrinker now runs in the system unbound workqueue and no longer in kswapd context so it can directly do an iput() on inodes even if that blocks or needs to acquire any lock (we aren't holding any locks when requesting the delayed iput from the shrinker). So we don't need to add a delayed iput, wake up the cleaner and delegate the iput() to the cleaner, which also adds extra contention on the spinlock that protects the delayed iputs list. Reported-by: Ivan Shapovalov Tested-by: Ivan Shapovalov Link: https://lore.kernel.org/linux-btrfs/0414d690ac5680d0d77dfc930606cdc36e42e12f.camel@intelfx.name/ CC: stable@vger.kernel.org # 6.12+ Reviewed-by: Johannes Thumshirn Reviewed-by: Qu Wenruo Signed-off-by: Filipe Manana Signed-off-by: David Sterba --- fs/btrfs/extent_map.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index 8c6b85ffd18f..7f46abbd6311 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -1256,7 +1256,7 @@ static long btrfs_scan_root(struct btrfs_root *root, struct btrfs_em_shrink_ctx min_ino = btrfs_ino(inode) + 1; fs_info->em_shrinker_last_ino = btrfs_ino(inode); - btrfs_add_delayed_iput(inode); + iput(&inode->vfs_inode); if (ctx->scanned >= ctx->nr_to_scan || btrfs_fs_closing(fs_info)) break; From b1bf18223a8340cf5d52162d320badcfe07b905d Mon Sep 17 00:00:00 2001 From: Qu Wenruo Date: Mon, 17 Feb 2025 20:16:39 +1030 Subject: [PATCH 5/6] btrfs: output an error message if btrfs failed to find the seed fsid [BUG] If btrfs failed to locate the seed device for whatever reason, mounting the sprouted device will fail without any meaning error message: # mkfs.btrfs -f /dev/test/scratch1 # btrfstune -S1 /dev/test/scratch1 # mount /dev/test/scratch1 /mnt/btrfs # btrfs dev add -f /dev/test/scratch2 /mnt/btrfs # umount /mnt/btrfs # btrfs dev scan -u # btrfs mount /dev/test/scratch2 /mnt/btrfs mount: /mnt/btrfs: fsconfig system call failed: No such file or directory. dmesg(1) may have more information after failed mount system call. # dmesg -t | tail -n6 BTRFS info (device dm-5): first mount of filesystem 64252ded-5953-4868-b962-cea48f7ac4ea BTRFS info (device dm-5): using crc32c (crc32c-generic) checksum algorithm BTRFS info (device dm-5): using free-space-tree BTRFS error (device dm-5): failed to read chunk tree: -2 BTRFS error (device dm-5): open_ctree failed: -2 [CAUSE] The failure to mount is pretty straight forward, just unable to find the seed device and its fsid, caused by `btrfs dev scan -u`. But the lack of any useful info is a problem. [FIX] Just add an extra error message in open_seed_devices() to indicate the error. Now the error message would look like this: BTRFS info (device dm-4): first mount of filesystem 7769223d-4db1-4e4c-ac29-0a96f53576ab BTRFS info (device dm-4): using crc32c (crc32c-generic) checksum algorithm BTRFS info (device dm-4): using free-space-tree BTRFS error (device dm-4): failed to find fsid e87c12e6-584b-4e98-8b88-962c33a619ff when attempting to open seed devices BTRFS error (device dm-4): failed to read chunk tree: -2 BTRFS error (device dm-4): open_ctree failed: -2 Link: https://github.com/kdave/btrfs-progs/issues/959 Reviewed-by: Anand Jain Reviewed-by: Filipe Manana Signed-off-by: Qu Wenruo Reviewed-by: David Sterba Signed-off-by: David Sterba --- fs/btrfs/volumes.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index a594f66daedf..f6ae76815e4b 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -7196,8 +7196,12 @@ static struct btrfs_fs_devices *open_seed_devices(struct btrfs_fs_info *fs_info, fs_devices = find_fsid(fsid, NULL); if (!fs_devices) { - if (!btrfs_test_opt(fs_info, DEGRADED)) + if (!btrfs_test_opt(fs_info, DEGRADED)) { + btrfs_err(fs_info, + "failed to find fsid %pU when attempting to open seed devices", + fsid); return ERR_PTR(-ENOENT); + } fs_devices = alloc_fs_devices(fsid); if (IS_ERR(fs_devices)) From efa11fd269c139e29b71ec21bc9c9c0063fde40d Mon Sep 17 00:00:00 2001 From: Qu Wenruo Date: Wed, 19 Feb 2025 09:06:33 +1030 Subject: [PATCH 6/6] btrfs: fix data overwriting bug during buffered write when block size < page size [BUG] When running generic/418 with a btrfs whose block size < page size (subpage cases), it always fails. And the following minimal reproducer is more than enough to trigger it reliably: workload() { mkfs.btrfs -s 4k -f $dev > /dev/null dmesg -C mount $dev $mnt $fsstree_dir/src/dio-invalidate-cache -r -b 4096 -n 3 -i 1 -f $mnt/diotest ret=$? umount $mnt stop_trace if [ $ret -ne 0 ]; then fail fi } for (( i = 0; i < 1024; i++)); do echo "=== $i/$runtime ===" workload done [CAUSE] With extra trace printk added to the following functions: - btrfs_buffered_write() * Which folio is touched * The file offset (start) where the buffered write is at * How many bytes are copied * The content of the write (the first 2 bytes) - submit_one_sector() * Which folio is touched * The position inside the folio * The content of the page cache (the first 2 bytes) - pagecache_isize_extended() * The parameters of the function itself * The parameters of the folio_zero_range() Which are enough to show the problem: 22.158114: btrfs_buffered_write: folio pos=0 start=0 copied=4096 content=0x0101 22.158161: submit_one_sector: r/i=5/257 folio=0 pos=0 content=0x0101 22.158609: btrfs_buffered_write: folio pos=0 start=4096 copied=4096 content=0x0101 22.158634: btrfs_buffered_write: folio pos=0 start=8192 copied=4096 content=0x0101 22.158650: pagecache_isize_extended: folio=0 from=4096 to=8192 bsize=4096 zero off=4096 len=8192 22.158682: submit_one_sector: r/i=5/257 folio=0 pos=4096 content=0x0000 22.158686: submit_one_sector: r/i=5/257 folio=0 pos=8192 content=0x0101 The tool dio-invalidate-cache will start 3 threads, each doing a buffered write with 0x01 at offset 0, 4096 and 8192, do a fsync, then do a direct read, and compare the read buffer with the write buffer. Note that all 3 btrfs_buffered_write() are writing the correct 0x01 into the page cache. But at submit_one_sector(), at file offset 4096, the content is zeroed out, by pagecache_isize_extended(). The race happens like this: Thread A is writing into range [4K, 8K). Thread B is writing into range [8K, 12k). Thread A | Thread B -------------------------------------+------------------------------------ btrfs_buffered_write() | btrfs_buffered_write() |- old_isize = 4K; | |- old_isize = 4096; |- btrfs_inode_lock() | | |- write into folio range [4K, 8K) | | |- pagecache_isize_extended() | | | extend isize from 4096 to 8192 | | | no folio_zero_range() called | | |- btrfs_inode_lock() | | | |- btrfs_inode_lock() | |- write into folio range [8K, 12K) | |- pagecache_isize_extended() | | calling folio_zero_range(4K, 8K) | | This is caused by the old_isize is | | grabbed too early, without any | | inode lock. | |- btrfs_inode_unlock() The @old_isize is grabbed without inode lock, causing race between two buffered write threads and making pagecache_isize_extended() to zero range which is still containing cached data. And this is only affecting subpage btrfs, because for regular blocksize == page size case, the function pagecache_isize_extended() will do nothing if the block size >= page size. [FIX] Grab the old i_size while holding the inode lock. This means each buffered write thread will have a stable view of the old inode size, thus avoid the above race. CC: stable@vger.kernel.org # 5.15+ Fixes: 5e8b9ef30392 ("btrfs: move pos increment and pagecache extension to btrfs_buffered_write") Reviewed-by: Filipe Manana Signed-off-by: Qu Wenruo Signed-off-by: David Sterba --- fs/btrfs/file.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index ed3c0d6546c5..0b568c8d24cb 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -1090,7 +1090,7 @@ ssize_t btrfs_buffered_write(struct kiocb *iocb, struct iov_iter *i) u64 lockend; size_t num_written = 0; ssize_t ret; - loff_t old_isize = i_size_read(inode); + loff_t old_isize; unsigned int ilock_flags = 0; const bool nowait = (iocb->ki_flags & IOCB_NOWAIT); unsigned int bdp_flags = (nowait ? BDP_ASYNC : 0); @@ -1103,6 +1103,13 @@ ssize_t btrfs_buffered_write(struct kiocb *iocb, struct iov_iter *i) if (ret < 0) return ret; + /* + * We can only trust the isize with inode lock held, or it can race with + * other buffered writes and cause incorrect call of + * pagecache_isize_extended() to overwrite existing data. + */ + old_isize = i_size_read(inode); + ret = generic_write_checks(iocb, i); if (ret <= 0) goto out;