On Mon, Jan 20, 2020 at 08:46:50PM +0100, David Sterba wrote: > On Sat, Dec 28, 2019 at 03:03:26PM -0500, Zygo Blaxell wrote: > > > > [...etc...it dumps the same stacks again in the BUG] > > > > The deepest level of fs/btrfs code is the list_del_init in > > clean_dirty_subvols. > > > > After a reboot the next mount (and all subsequent mounts so far) > > was successful. I don't have a reproducer. > > For the record, this and other KASAN reports regarding reloc root, is > going to be fixed in 5.4.14, "btrfs: relocation: fix reloc_root lifespan > and access" (6282675e6708ec78). Thanks for the reports and debugging > help. Confirmed: I have not seen this bug since applying the first versions of the patch to my test kernels in 5.4.8. I've been testing continuously since then, replacing the early version with the final upstream version in 5.4.14, and now running 5.4.16. After applying "Btrfs: fix race between adding and putting tree mod seq elements and nodes" I have seen no crashing bugs _at all_ (no deadlock, BUG_ON or KASAN). Thank you for all the fixes! Much appreciated!
Attachment:
signature.asc
Description: PGP signature
