Closed Connor22 closed 3 months ago
It seems that more than just bch2_nocow_write is hanging, but the hung_task messages always start and are mostly comprised of it
After attempting to close smb during a hang, the following (interleaved?) dump was output
Superblock info (All device superblocks are consistent, no split-brain):
After some fsck attempts and a lot of journal_reclaim_would_deadlock
, I got the following page fault oops:
I think at this point I'll wait for 6.9 and see if any of the recent patches can help.
Tried doing a fsck -R
on Linux 6.9.0-rc1-1 from AUR linux-mainline
fsck segfaulted after snapshots_read... done
unsure if this is bcache-specific as there's not much to go off of in the log
Thanks to advice from py1hon on the IRC, I waited until the most recent commit of bcachefs-tools, then tried reconstruct_alloc again.
Mounting with bcache mount -o fsck,reconstruct_alloc,fix_errors
took a while, but seems to have fully restored functionality to my drives!
Closing for now, will amend if anything else pops up.
Environment
Arch Linux Server Kernel 6.8.1 1 SSD and 1 HDD
Situation
SQLite database being accessed over Samba, originally
compression=zstd
background_compression=zstd
nocow=0
Cause of issue
Setting
compression=none
background_compression=none
nocow=1
to database files usingbcache setattr
Issue
No disk i/o activity after flags are set, even after remounting or rebooting Every write (over the network) involving database files hangs indefinitely
Steps attempted
Example call trace