Open steelman opened 4 months ago
Your filesystem doesn't work correctly.
borg does a file "open" and the fs responds with:
OSError: [Errno 95] Operation not supported
Indeed it looks very similar. However, it needs to be noted that:
a) it happens only with borg check
and not with borg create
b) I can open()
files for writing on this file system (especially inside the borg repository) without any problems.
You can try to reproduce this on a local filesystem to see whether it only happens on cifs.
"not supported" as error response to "open()" is weird.
Oh, and try without that "gvfs" and just do a normal cifs mount and try with that also.
On kernel-based cifs volume borg check
works fine. On gvfs with strace(1)
attached to gvfs-fuse serving /run/user/2137/gvfs
borg works fine. Without strace(1)
it fails. From my experience this looks like some race condition in gvfs. But it is borg that is able to trigger it and that is interesting.
EDIT: borg check
running under strace(1)
also does not crash.
EVEN MORE EDIT: Even weirder is that the exception happens on the second attempt to open /run/user/2137/gvfs/smb-share:server=[snip],share=homefolder/steelman/Backup/Mail/lock.roster
after asking for the password. The first successful attempt takes place right after startup.
I played with touch(1)
and strace(1)
. The former randomly fails to create files in /run/user/2137/gvfs/smb-share:server=[snip],share=homefolder/steelman/Backup/Mail/
(with EOPNOTSUPP
) so it seems there is little chance this isn't a fault of gvfs. I will keep investigating.
EDIT: No, it's not at random. openat(2)
sets errno
to EOPNOTSUPP
when the file to be created exists. The first call succeeds:
openat(AT_FDCWD, "/run/user/2137/gvfs/smb-share:server=[snip],share=homefolder/steelman/Backup/Mail/__test__", O_WRONLY|O_CREAT|O_NOCTTY|O_NONBLOCK, 0666) = 3
consequent fail:
openat(AT_FDCWD, "/run/user/2137/gvfs/smb-share:server=[snip],share=homefolder/steelman/Backup/Mail/__test__", O_WRONLY|O_CREAT|O_NOCTTY|O_NONBLOCK, 0666) = -1 EOPNOTSUPP (Operation not supported)
OK, so guess this is a gvfs bug? Or is borg doing something wrong or unusual?
It looks like a gvfs bug to me, but this is just like, my opinion. I will find out.
Have you checked borgbackup docs, FAQ, and open GitHub issues?
Yes
Is this a BUG / ISSUE report or a QUESTION?
BUG
System information. For client/server mode post info for both machines.
Your borg version (borg -V).
borg 1.2.4
Operating system (distribution) and version.
Debian 12.5
Hardware / network configuration, and filesystems used.
Source: local ext4
Destination: CIFS mounted via FUSE (gvfs) with
gio mount smb://…
.How much data is handled by borg?
(The question isn't clear)
Full borg commandline that lead to the problem (leave away excludes and passwords)
borg check ::2024-W08-1
Describe the problem you're observing.
borg crashes
Can you reproduce the problem? If so, describe how. If not, describe troubleshooting steps you took before opening the issue.
Yes, it happens every time.
Include any warning/errors/backtraces from the system logs