RFC: can we improve the mount program's error messages?

[David Collier-Brown]

On 06/02/2012 02:56 PM, David Collier-Brown wrote:
>   Fred Weigel, one of my Smarter Colleagues at work made an smb mount of
> a drive on our internal network, and some hours later connected to a VPN
> to a customer site.  This all was fine until he wanted to do an ls on
> his local machine, and it */hung/*....
>
> Approximately three minutes late it unstuck, and said
> ---
> [fred at dejah ~]$ ls
> ls: cannot access UserHome: Host is down
> ls: cannot access Solutions: Host is down
> 2ax3                  gambc-stuff                scheme
> a                        gfortran_switches       scm.2
> a.c                     gsc-config                   setfpucw.txt
> ambeg               gscript                        show-passes.py
> ---
>
> We tried a GUI to do the ls (using Thunar) but it hung too, and didn't
> even bring up it's initial window...
>
> Eventually we got the ls error message, then Thunar started and
> displayed the same message.  UserHome and Solutions were the names of
> shares, and while "Host is down" wasn't actually correct, it did allow
> us to see it was a software problem. This was a great relief, as we'd
> both suspected a hardware problem.
>
> Looking at syslog gave us a bit more information
> ---
> [fred at dejah ~]$ sudo tail /var/log/messages
> May 31 14:46:51 dejah kernel: [63915.904107] CIFS VFS: Unexpected lookup
> error -112
> May 31 14:46:54 dejah pptp[13117]: nm-pptp-service-13107
> log[logecho:pptp_ctrl.c:692]: Echo Reply received.
> May 31 14:46:54 dejah pptp[13117]: nm-pptp-service-13107
> log[logecho:pptp_ctrl.c:694]: no more Echo Reply/Request packets will be
> reported.
> May 31 14:47:11 dejah kernel: [63935.906107] CIFS VFS: Unexpected lookup
> error -112
> May 31 14:47:31 dejah kernel: [63955.908132] CIFS VFS: Unexpected lookup
> error -112
> May 31 14:47:51 dejah kernel: [63975.910114] CIFS VFS: Unexpected lookup
> error -112
> May 31 14:48:11 dejah kernel: [63995.912131] CIFS VFS: Unexpected lookup
> error -112
> May 31 14:48:31 dejah kernel: [64015.913287] CIFS VFS: Unexpected lookup
> error -112
> May 31 14:48:45 dejah wall[13654]: wall: user fred broadcasted 1 lines
> (14 chars)
> May 31 14:48:51 dejah kernel: [64035.915128] CIFS VFS: Unexpected lookup
> error -112
> --
>
> After a few moments, we figured out what had happened.  The VPN had
> changed the default route, and we had tried to do an SMB request to a
> machine on the other end of the VPN, so of course it failed, and
> eventually timed out.
>
> This therefore was NOT a samba/mount.cifs problem, but rather a problem
> with the VPN, merely exacerbated by the longish timeout and the
> minimalist error messages. 
> We can help a bit by improving the messages,
> and I specifically would like your opinion about the famous
>
>         NFS server foo nor responding, still trying
>
> message from the early days of network filesystems. 
>
> That message existed because the failure was at the kernel level, and
> the nfs program didn't know what user or users were affected by the
> problem. So it used wall(1) to notify everyone.  If you didn't like wall
> messages, you set "mesg n" and shut it up.
>
> We still suffer the same problem as early Unix, only now it's worse. 
> It's hard to tell who is suffering from a slow, unreachable or failed
> host, and the GUIs don't help.  What might help are
> 1) better messages
> 2) a notifier like wall that works with the major GUIs (is there one???)
> 3) a way to make the timeouts shorter without introducing new problems.
>
> Today, I'd like to propose two variants of (1)
> i) use wall for the first failed response in a specified time period
> ii) make the message "SMB server //<host>/<share> not responding, still
> trying"
>
> The ls messages only supply the share, and they say "Host is down",
> which will make non-techies go looking for a host named <share>. The
> syslog messages are precise but not meaningful, so I'd append the "not
> responding" message to .
>
>
> --dave
> Sigh: I hardly get to code any more, just diagnose stuff. Maybe we
> should make the messages *more* obscure, so I'd get asked interesting
> questions more often...

Fred mentioned that the new (gnome) notifier was
     notify-send [-opts] summary [body]
where the opts are --urgency, --expire-time, --icon, --category and --hint.

--dave

-- 
David Collier-Brown,         | Always do right. This will gratify
System Programmer and Author | some people and astonish the rest
davecb at spamcop.net           |                      -- Mark Twain
(416) 223-8968