| Message ID | cover.1719366258.git.wqu@suse.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | btrfs: detect and fix the ram_bytes mismatch | expand |
On Wed, Jun 26, 2024 at 2:48 AM Qu Wenruo <wqu@suse.com> wrote: > > [CHANGELOG] > v2: > - Add the missing patch fixing ram_bytes > Now the 2nd patch would ignore the incorrect value and use a correct > one from btrfs_file_extent_item::disk_num_bytes. > > - Update the commit messages to fix my usual "would" and other grammar > errors > > There is a long existing mismatch between ram_bytes and disk_num_bytes > for regular non-compressed data extents. > > It turns out to be caused by truncated ordered extents, which modified > ram_bytes unnecessarily. > > Thankfully this is not going to cause any data corruption or whatever, > kernel can handle it correctly without any extra problem. > It's only a small violation on the on-disk format. > > This series would fix by: > > - Cleanup the @bytenr usage inside btrfs_extent_item_to_extent_map() > - Override the ram_bytes when reading file extent items from disk > So that we always get correct extent maps even if the on-disk one is > incorrect. > - Add an extra check on extent_map members > - Add the proper fix for the ram_bytes mismatch > - Add a tree-checker for the ram_bytes mismatch > Since we can have on-disk ram_bytes incorrect already, this check is > only for DEBUG and ASSERT builds, and it won't report error but only > does a kernel warning for us to catch. > > Qu Wenruo (5): > btrfs: cleanup the bytenr usage inside > btrfs_extent_item_to_extent_map() > btrfs: ignore incorrect btrfs_file_extent_item::ram_bytes > btrfs: make validate_extent_map() to catch ram_bytes mismatch > btrfs: fix the ram_bytes assignment for truncated ordered extents > btrfs: tree-checker: add extra ram_bytes and disk_num_bytes check > > fs/btrfs/extent_map.c | 5 +++++ > fs/btrfs/file-item.c | 16 +++++++++++----- > fs/btrfs/inode.c | 4 +--- > fs/btrfs/tree-checker.c | 18 ++++++++++++++++++ > 4 files changed, 35 insertions(+), 8 deletions(-) Reviewed-by: Filipe Manana <fdmanana@suse.com> Looks good, thanks. > > -- > 2.45.2 > >
[CHANGELOG] v2: - Add the missing patch fixing ram_bytes Now the 2nd patch would ignore the incorrect value and use a correct one from btrfs_file_extent_item::disk_num_bytes. - Update the commit messages to fix my usual "would" and other grammar errors There is a long existing mismatch between ram_bytes and disk_num_bytes for regular non-compressed data extents. It turns out to be caused by truncated ordered extents, which modified ram_bytes unnecessarily. Thankfully this is not going to cause any data corruption or whatever, kernel can handle it correctly without any extra problem. It's only a small violation on the on-disk format. This series would fix by: - Cleanup the @bytenr usage inside btrfs_extent_item_to_extent_map() - Override the ram_bytes when reading file extent items from disk So that we always get correct extent maps even if the on-disk one is incorrect. - Add an extra check on extent_map members - Add the proper fix for the ram_bytes mismatch - Add a tree-checker for the ram_bytes mismatch Since we can have on-disk ram_bytes incorrect already, this check is only for DEBUG and ASSERT builds, and it won't report error but only does a kernel warning for us to catch. Qu Wenruo (5): btrfs: cleanup the bytenr usage inside btrfs_extent_item_to_extent_map() btrfs: ignore incorrect btrfs_file_extent_item::ram_bytes btrfs: make validate_extent_map() to catch ram_bytes mismatch btrfs: fix the ram_bytes assignment for truncated ordered extents btrfs: tree-checker: add extra ram_bytes and disk_num_bytes check fs/btrfs/extent_map.c | 5 +++++ fs/btrfs/file-item.c | 16 +++++++++++----- fs/btrfs/inode.c | 4 +--- fs/btrfs/tree-checker.c | 18 ++++++++++++++++++ 4 files changed, 35 insertions(+), 8 deletions(-)