[Pkg-samba-maint] Bug#843153: Bug#843153: Provide a means to wake-up (reconnect) an existing share

[Mathieu Parent]

Control: reassign -1 src:linux 4.8.5-1
Control: affects -1 cifs-utils
Control: tags -1 + upstream

2016-11-04 12:13 GMT+01:00 martin f krafft <madduck at debian.org>:
> Package: cifs-utils
> Version: 2:6.6-1
> Severity: wishlist

hello Martin,

> After a laptop suspend, SMB sessions are usually disconnected on the
> server, and even the client will have a hard time just resuming.

This is most probably kernel-side.

> This will either lead to soft-errors or hard-blocks, until on the
> client, eventually, the kernel states

eventually? Meaning not always?

>   kernel: [27763.247021] CIFS VFS: Server samba.example.org has not
>   responded in 120 seconds. Reconnecting…
>
> and reconnects.
>
> One way around this is to lazy-umount the shares, and to remount
> them, but that's not easily done with e.g. libpam-mount, and it also
> doesn't solve the issue with open inodes.

Does mount -o remount $path works?

>
> I don't really want to shorten the 120 seconds (and I wouldn't know
> how, there seems to be no mount option),

120 comes from echo_interval [1] which looks undocumented.

[1] http://sources.debian.net/src/linux/4.8.5-1/fs/cifs/connect.c/?hl=496#L496

> but what would be great
> would be a way to prod the share and make it wake-up (i.e.
> reconnect), e.g.
>
>   mount.cifs --resume -a

Again, have you tried mount -o remount?

> which would immediately cause a reconnection, not only after 120
> seconds. This could then be executed by systemd for the resume
> target…

Looking at the NFS kernel code, I don't see any resume handling. I
don't know if it's affected too.

Regards

-- 
Mathieu