Btrfs objectid validation problem with kernel 6.10.3

[ileoo]

With kernel 6.10.3 seems that btrfs does objectid validation, for me it causes btrfs mount to jump to read-only, but no idea how to fix the validity of the object. With kernel 6.10.2 mount works just fine and scrub and everything passes without problems.

This is the journal output for the issue that makes

Sep 15 11:00:41 ilkan-poytakone kernel: page: refcount:3 mapcount:0 mapping:00000000995878cf index:0x1ae04c349 pfn:0x1c32c0
Sep 15 11:00:41 ilkan-poytakone kernel: memcg:ffff8b3f800f4000
Sep 15 11:00:41 ilkan-poytakone kernel: aops:btree_aops [btrfs] ino:1
Sep 15 11:00:41 ilkan-poytakone kernel: flags: 0x17ffffc0004020(lru|private|node=0|zone=2|lastcpupid=0x1fffff)
Sep 15 11:00:41 ilkan-poytakone kernel: raw: 0017ffffc0004020 ffffd4914933d148 ffffd491470c7688 ffff8b3f835f9458
Sep 15 11:00:41 ilkan-poytakone kernel: raw: 00000001ae04c349 ffff8b403f840870 00000003ffffffff ffff8b3f800f4000
Sep 15 11:00:41 ilkan-poytakone kernel: page dumped because: eb page dump
Sep 15 11:00:41 ilkan-poytakone kernel: BTRFS critical (device nvme1n1p1): corrupt leaf: block=29550653509632 slot=13 extent bytenr=29010740482048 len=643072 invalid data ref objectid value 5
Sep 15 11:00:41 ilkan-poytakone kernel: BTRFS error (device nvme1n1p1): read time tree block corruption detected on logical 29550653509632 mirror 2
Sep 15 11:00:41 ilkan-poytakone kernel: page: refcount:3 mapcount:0 mapping:00000000995878cf index:0x1ae04c349 pfn:0x1c32c0
Sep 15 11:00:41 ilkan-poytakone kernel: memcg:ffff8b3f800f4000
Sep 15 11:00:41 ilkan-poytakone kernel: aops:btree_aops [btrfs] ino:1
Sep 15 11:00:41 ilkan-poytakone kernel: flags: 0x17ffffc0004020(lru|private|node=0|zone=2|lastcpupid=0x1fffff)
Sep 15 11:00:41 ilkan-poytakone kernel: raw: 0017ffffc0004020 ffffd4914933d148 ffffd491470c7688 ffff8b3f835f9458
Sep 15 11:00:41 ilkan-poytakone kernel: raw: 00000001ae04c349 ffff8b403f840870 00000003ffffffff ffff8b3f800f4000
Sep 15 11:00:41 ilkan-poytakone kernel: page dumped because: eb page dump
Sep 15 11:00:41 ilkan-poytakone kernel: BTRFS critical (device nvme1n1p1): corrupt leaf: block=29550653509632 slot=13 extent bytenr=29010740482048 len=643072 invalid data ref objectid value 5
Sep 15 11:00:41 ilkan-poytakone kernel: BTRFS error (device nvme1n1p1): read time tree block corruption detected on logical 29550653509632 mirror 1
Sep 15 11:00:41 ilkan-poytakone kernel: BTRFS error (device nvme1n1p1): failed to run delayed ref for logical 29010740482048 num_bytes 643072 type 178 action 1 ref_mod 1: -5
Sep 15 11:00:41 ilkan-poytakone kernel: BTRFS error (device nvme1n1p1 state A): Transaction aborted (error -5)
Sep 15 11:00:41 ilkan-poytakone kernel: BTRFS: error (device nvme1n1p1 state A) in btrfs_run_delayed_refs:2207: errno=-5 IO failure
Sep 15 11:00:41 ilkan-poytakone kernel: BTRFS info (device nvme1n1p1 state EA): forced readonly

kernel-package where issue appears is kernel-default-6.10.9-1.1.x86_64

Any idea what approach I should take to resolve that invalid data reference issue?

[hui]

There are some hints…

How to recover from BTRFS errors | Support | SUSE

[arvidjaar]

There were several similar threads recently on btrfs mailing list. You may consider reviewing them and follow guidelines there or posting your case to the list. This usually needs developer eyes to evaluate.

[ileoo]

Thanks for the pointer!

I managed to find 'Re: Tree corruption' - MARC and after compiling the linked btrfs-prog tree (GitHub - adam900710/btrfs-progs at ino_clear_fix) and doing the btrfs rescue clear-ino-cache for the disks in the mountpoint, seems that issue is resolved.