Re: [PATCH v2] smb/client: fix chan_max and UNC state corruption in smb3_reconfigure

[Henrique Carvalho]

Isn't this a policy change? If multichannel fails all the other changes
up to that point are persisted (i.e. password change). Despite that, we
will be returning an error to the user, but have a state that we have no
idea what actually happened. Did the password change fail? Did the
multichannel change fail?

Before we had a best-effort remount and after this change we will be
having a partial success reported as failure.

So I would say that, for this change, if this is indeed a bug,
no need for mchan_rc, roll back all the changes that happened during
smb3_reconfigure.

Futhermore, we need to handle mchan update ses channels just like we do
during mount, because remount could end up being delayed the same way as
mount, AFAICS. I have a patch for this with the series of that
serialization smb3_update_ses_channels patch.

-- 
Henrique
SUSE Labs