From 3324d0547861b16cf436d54abba7052e0c8aa9de Mon Sep 17 00:00:00 2001 From: Omar Sandoval Date: Thu, 4 Jan 2024 11:48:47 -0800 Subject: btrfs: avoid copying BTRFS_ROOT_SUBVOL_DEAD flag to snapshot of subvolume being deleted Sweet Tea spotted a race between subvolume deletion and snapshotting that can result in the root item for the snapshot having the BTRFS_ROOT_SUBVOL_DEAD flag set. The race is: Thread 1 | Thread 2 ----------------------------------------------|---------- btrfs_delete_subvolume | btrfs_set_root_flags(BTRFS_ROOT_SUBVOL_DEAD)| |btrfs_mksubvol | down_read(subvol_sem) | create_snapshot | ... | create_pending_snapshot | copy root item from source down_write(subvol_sem) | This flag is only checked in send and swap activate, which this would cause to fail mysteriously. create_snapshot() now checks the root refs to reject a deleted subvolume, so we can fix this by locking subvol_sem earlier so that the BTRFS_ROOT_SUBVOL_DEAD flag and the root refs are updated atomically. CC: stable@vger.kernel.org # 4.14+ Reported-by: Sweet Tea Dorminy Reviewed-by: Sweet Tea Dorminy Reviewed-by: Anand Jain Signed-off-by: Omar Sandoval Reviewed-by: David Sterba Signed-off-by: David Sterba --- fs/btrfs/inode.c | 22 +++++++++++++--------- 1 file changed, 13 insertions(+), 9 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index b3e39610cc95..7bcc1c03437a 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -4458,6 +4458,8 @@ int btrfs_delete_subvolume(struct btrfs_inode *dir, struct dentry *dentry) u64 root_flags; int ret; + down_write(&fs_info->subvol_sem); + /* * Don't allow to delete a subvolume with send in progress. This is * inside the inode lock so the error handling that has to drop the bit @@ -4469,25 +4471,25 @@ int btrfs_delete_subvolume(struct btrfs_inode *dir, struct dentry *dentry) btrfs_warn(fs_info, "attempt to delete subvolume %llu during send", dest->root_key.objectid); - return -EPERM; + ret = -EPERM; + goto out_up_write; } if (atomic_read(&dest->nr_swapfiles)) { spin_unlock(&dest->root_item_lock); btrfs_warn(fs_info, "attempt to delete subvolume %llu with active swapfile", root->root_key.objectid); - return -EPERM; + ret = -EPERM; + goto out_up_write; } root_flags = btrfs_root_flags(&dest->root_item); btrfs_set_root_flags(&dest->root_item, root_flags | BTRFS_ROOT_SUBVOL_DEAD); spin_unlock(&dest->root_item_lock); - down_write(&fs_info->subvol_sem); - ret = may_destroy_subvol(dest); if (ret) - goto out_up_write; + goto out_undead; btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); /* @@ -4497,7 +4499,7 @@ int btrfs_delete_subvolume(struct btrfs_inode *dir, struct dentry *dentry) */ ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 5, true); if (ret) - goto out_up_write; + goto out_undead; trans = btrfs_start_transaction(root, 0); if (IS_ERR(trans)) { @@ -4563,15 +4565,17 @@ out_end_trans: inode->i_flags |= S_DEAD; out_release: btrfs_subvolume_release_metadata(root, &block_rsv); -out_up_write: - up_write(&fs_info->subvol_sem); +out_undead: if (ret) { spin_lock(&dest->root_item_lock); root_flags = btrfs_root_flags(&dest->root_item); btrfs_set_root_flags(&dest->root_item, root_flags & ~BTRFS_ROOT_SUBVOL_DEAD); spin_unlock(&dest->root_item_lock); - } else { + } +out_up_write: + up_write(&fs_info->subvol_sem); + if (!ret) { d_invalidate(dentry); btrfs_prune_dentries(dest); ASSERT(dest->send_in_progress == 0); -- cgit v1.2.3 From 5571e41ec6e56e35f34ae9f5b3a335ef510e0ade Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Wed, 31 Jan 2024 14:27:25 -0500 Subject: btrfs: don't drop extent_map for free space inode on write error While running the CI for an unrelated change I hit the following panic with generic/648 on btrfs_holes_spacecache. assertion failed: block_start != EXTENT_MAP_HOLE, in fs/btrfs/extent_io.c:1385 ------------[ cut here ]------------ kernel BUG at fs/btrfs/extent_io.c:1385! invalid opcode: 0000 [#1] PREEMPT SMP NOPTI CPU: 1 PID: 2695096 Comm: fsstress Kdump: loaded Tainted: G W 6.8.0-rc2+ #1 RIP: 0010:__extent_writepage_io.constprop.0+0x4c1/0x5c0 Call Trace: extent_write_cache_pages+0x2ac/0x8f0 extent_writepages+0x87/0x110 do_writepages+0xd5/0x1f0 filemap_fdatawrite_wbc+0x63/0x90 __filemap_fdatawrite_range+0x5c/0x80 btrfs_fdatawrite_range+0x1f/0x50 btrfs_write_out_cache+0x507/0x560 btrfs_write_dirty_block_groups+0x32a/0x420 commit_cowonly_roots+0x21b/0x290 btrfs_commit_transaction+0x813/0x1360 btrfs_sync_file+0x51a/0x640 __x64_sys_fdatasync+0x52/0x90 do_syscall_64+0x9c/0x190 entry_SYSCALL_64_after_hwframe+0x6e/0x76 This happens because we fail to write out the free space cache in one instance, come back around and attempt to write it again. However on the second pass through we go to call btrfs_get_extent() on the inode to get the extent mapping. Because this is a new block group, and with the free space inode we always search the commit root to avoid deadlocking with the tree, we find nothing and return a EXTENT_MAP_HOLE for the requested range. This happens because the first time we try to write the space cache out we hit an error, and on an error we drop the extent mapping. This is normal for normal files, but the free space cache inode is special. We always expect the extent map to be correct. Thus the second time through we end up with a bogus extent map. Since we're deprecating this feature, the most straightforward way to fix this is to simply skip dropping the extent map range for this failed range. I shortened the test by using error injection to stress the area to make it easier to reproduce. With this patch in place we no longer panic with my error injection test. CC: stable@vger.kernel.org # 4.14+ Reviewed-by: Filipe Manana Signed-off-by: Josef Bacik Signed-off-by: David Sterba --- fs/btrfs/inode.c | 19 +++++++++++++++++-- 1 file changed, 17 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 7bcc1c03437a..d232eca1bbee 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -3184,8 +3184,23 @@ out: unwritten_start += logical_len; clear_extent_uptodate(io_tree, unwritten_start, end, NULL); - /* Drop extent maps for the part of the extent we didn't write. */ - btrfs_drop_extent_map_range(inode, unwritten_start, end, false); + /* + * Drop extent maps for the part of the extent we didn't write. + * + * We have an exception here for the free_space_inode, this is + * because when we do btrfs_get_extent() on the free space inode + * we will search the commit root. If this is a new block group + * we won't find anything, and we will trip over the assert in + * writepage where we do ASSERT(em->block_start != + * EXTENT_MAP_HOLE). + * + * Theoretically we could also skip this for any NOCOW extent as + * we don't mess with the extent map tree in the NOCOW case, but + * for now simply skip this if we are the free space inode. + */ + if (!btrfs_is_free_space_inode(inode)) + btrfs_drop_extent_map_range(inode, unwritten_start, + end, false); /* * If the ordered extent had an IOERR or something else went -- cgit v1.2.3 From 1bd96c92c6a0a4d43815eb685c15aa4b78879dc9 Mon Sep 17 00:00:00 2001 From: Filipe Manana Date: Fri, 2 Feb 2024 12:09:22 +0000 Subject: btrfs: reject encoded write if inode has nodatasum flag set Currently we allow an encoded write against inodes that have the NODATASUM flag set, either because they are NOCOW files or they were created while the filesystem was mounted with "-o nodatasum". This results in having compressed extents without corresponding checksums, which is a filesystem inconsistency reported by 'btrfs check'. For example, running btrfs/281 with MOUNT_OPTIONS="-o nodatacow" triggers this and 'btrfs check' errors out with: [1/7] checking root items [2/7] checking extents [3/7] checking free space tree [4/7] checking fs roots root 256 inode 257 errors 1040, bad file extent, some csum missing root 256 inode 258 errors 1040, bad file extent, some csum missing ERROR: errors found in fs roots (...) So reject encoded writes if the target inode has NODATASUM set. CC: stable@vger.kernel.org # 6.1+ Reviewed-by: Johannes Thumshirn Signed-off-by: Filipe Manana Reviewed-by: David Sterba Signed-off-by: David Sterba --- fs/btrfs/inode.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index d232eca1bbee..5f39462dceb0 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -10288,6 +10288,13 @@ ssize_t btrfs_do_encoded_write(struct kiocb *iocb, struct iov_iter *from, if (encoded->encryption != BTRFS_ENCODED_IO_ENCRYPTION_NONE) return -EINVAL; + /* + * Compressed extents should always have checksums, so error out if we + * have a NOCOW file or inode was created while mounted with NODATASUM. + */ + if (inode->flags & BTRFS_INODE_NODATASUM) + return -EINVAL; + orig_count = iov_iter_count(from); /* The extent size must be sane. */ -- cgit v1.2.3