[Samba] Change in mount behavior of the CIFS module from Debian kernel 6.12 to 6.16

[Peter Milesson]

On 19.11.2025 15:40, Rowland Penny via samba wrote:
> On Wed, 19 Nov 2025 14:42:27 +0100
> Peter Milesson via samba <samba at lists.samba.org> wrote:
>
>> Hi folks,
>>
>> I got a seriously unpleasant problem, when LXDE stopped working in a
>> bunch of terminals after kernel upgrade from 6.12.48 to 6.16.3. There
>> were no indications about errors in the log files, so tracking down
>> the bug was very hard. The only find was a couple of errors in the
>> ~/.xsession-errors file. The error was basically that ssh-agent could
>> not bind to sockets in the ~/.ssh/agent directory, as they were
>> missing.
>>
>> I'm mounting the user's folder under /home from a CIFS server with
>> pam-mount after successful authentication. The mount parameters were:
>>
>> dir_mode=0700,nosuid,nodev,sec=krb5i,cruid=<userid>,mfsymlinks,nobrl,vers=3.0
>>
>> Using kernel 6.16.3, the sshd-agent could not create sockets in the
>> ~/.ssh/agent directory. That worked with kernel 6.12.48, and LXDE was
>> happy. I read through the CIFS documentation and found that the
>> parameter sfu had any relevance. I replaced mfsymlinks with sfu so
>> the parameters sent to cifs mount were:
>>
>> dir_mode=0700,nosuid,nodev,sec=krb5i,cruid=<userid>,sfu,nobrl,vers=3.0
>>
>> After this change, sockets were created, and LXDE started on the
>> terminal.
>>
>> Please, could somebody who knows explain to me, if the change in
>> mfsymlinks behavior was intentional, or if it is a bug? Is the sfu
>> parameter expected to stay for the foreseeable future, or will that
>> also be changed or removed? Are there any alternatives to mfsymlinks
>> and sfu?
>>
>> Best regards,
>>
>> Peter
> I have no idea about this, but feel you stand more chance of getting a
> response on this if you post to the samba-technical list, that may
> attract the attention of Steve French (the 'f' in mfsymlinks).
>
> Rowland
>
Hi Rowland,

Thanks for the information, I'll sign up there.

I think the changed behavior is important enough for a discussion. At 
least, I was chased out of bed at 6 o'clock in the morning to solve a 
bunch of non working terminals.

Best regards,

Peter