The bad format is actually caught by btree_err_on(bch2_bkey_format_invalid(c, &bn->format, write, &buf1), ...) in validate_bset, but since the return from __btree_err matches BCH_err_fsck_fix, the code doesn't really do anything about it.
- [syzbot] [bcachefs?] UBSAN: shift-out-of-bounds in __bch2_b... syzbot
- Re: [syzbot] [bcachefs?] UBSAN: shift-out-of-bounds in... Bharadwaj Raju
