Dear Dennis,
Without the bcache cache device, I was able to mount the file system (
root ) for read-only. I recreated it on the same LVM volume, and
rsync-ed the files. The files I lost (i/o error because checksum error)
was not important. Your case might be different. But without the cache,
additional corruption stopped happening, so I think it is safe to
recover your files with the gcc9 + 5.1 kernel (on bcache0) - if cache
device is not present. If you cannot mount the fs, you can try the
standard btrfs recovery methods (mount read only) or btrfsck (please be
aware you can make things worse with btrfsck, so if the data is
important backup the device as byte stream for new chance)
I want to mention I had another corruption (btrfs checksum error), but
it was simply not there after reboot without bcacha cache device. This
means the corruption that time was not written back to the backing
device, so I haven't lost data that time. I think I was lucky.
László Szalma
2019. 05. 28. 14:36 keltezéssel, Dennis Schridde írta:
Dear Szalma!
Thank you for the information.
On Dienstag, 28. Mai 2019 08:40:40 CEST Szalma László wrote:
I experienced the same, the problem is with bcache + gcc9. Immediately
remove the cache device from the bcache, as it prevents more damages to
your filesystem. After it downgrade gcc to 8 or avoid using bcache (with
cache device) until the problem is solved by upstream.
Please see here for more information:
https://bugzilla.kernel.org/show_bug.cgi?id=203573
This is a very serious issue I think, but in my case I could save all my
files after reboot without bcache cache device (except 1 insignificant
one), but I guess you might need backups (I hope you have).
I did the following so far:
readlink /sys/fs/bcache/*/cache0/set | sed 's,.*/,,' > /sys/block/bcache0/
bcache/detach
How should I proceed from here on the path that you took to recover your
system?
--Dennis
