Replies: 1 comment 1 reply
|
There is a useful clue in the repeated A clean Btrfs scrub does not contradict this. Scrub checks the stored blocks against Btrfs' own checksums; fs-verity checks file contents against the Merkle root recorded when verity was enabled. The filesystem can therefore be internally checksum-consistent while fs-verity still rejects the file. I would first map every reported inode before changing anything: findmnt -T /sysroot -o TARGET,SOURCE,FSTYPE,OPTIONS
sudo find /sysroot -xdev \( -inum 23391777 -o -inum 16147539 -o -inum 16147546 -o -inum 24282025 \) -print
sudo btrfs inspect-internal inode-resolve 23391777 /sysrootWith composefs the visible path may be an overlay while the inode belongs to an OSTree object, so also check sudo fsverity measure /path/to/file
sudo dd if=/path/to/file of=/dev/null bs=1M status=progress
sudo ostree fsck --repo=/sysroot/ostree/repoThe full read should fail with |
Uh oh!
There was an error while loading. Please reload this page.
Hi,
I had a look at my system logs today and noticed that something seems to be not quite right with some of the fs-verity information on my
/sysrootfilesystem:This is an Arch Linux setup with kernel 7.2.2, native composefs with UKI booted via systemd-boot. It doesn't seem to effect the running system as far as I can see (yet?).
/sysrootis on a btrfs filesystem andbtrfs scrubdoesn't show any errors. Neither do the logs close to the fs-verity corruption warnings, so the FS seems to be fine.Any idea on how to debug which files are affected and what could be the root cause here?
Thanks!
All reactions