I'm posting this as a series but this is really three independant patches. The first is a real bug that has made our board unable to boot as it was corrupting the kernel and image CRC validation failed (yay for checksums -- any plan of enabling validation of data checksums too? I see the code is there for metadata already, so it's probably not missing that much, but unfortunately as far as I understand it would only have seen that the extent data was correct and still corrupted the final buffer as it's just an offset error...)
The second patch is an opportunistic optimization as I noticed the code behaved differently than before, and there doesn't seem to be a reason the reads couldn't be coalesced.
The last patch is just something that looked odd and looks "obviously better" but I'll leave that up to you.
Thanks!
Link: https://lore.kernel.org/linux-btrfs/20200525063257.46757-1-wqu@suse.com/ [1] Signed-off-by: Dominique Martinet dominique.martinet@atmark-techno.com --- Dominique Martinet (3): btrfs: fix offset within btrfs_read_extent_reg() btrfs: btrfs_file_read: allow opportunistic read until the end btrfs: btfs_file_read: zero trailing data if no extent was found
fs/btrfs/inode.c | 32 ++++++++++++++++++-------------- 1 file changed, 18 insertions(+), 14 deletions(-) --- base-commit: 5db4972a5bbdbf9e3af48ffc9bc4fec73b7b6a79 change-id: 20230418-btrfs-extent-reads-e2df6e329ad4
Best regards,