[Samba] samba client error (smbfs?).

Urban Widmark urban at teststation.com
Sat Apr 13 02:21:02 GMT 2002


On Thu, 11 Apr 2002, Brian Spears wrote:

> weirdest problem, from a RH7.2 (base or + everything) samba client running 
> eclipse utilizing SMB mount.
> 
> Apr 11 13:43:54 testsmbbug kernel: smb_get_length: recv error = 512
> Apr 11 13:43:54 testsmbbug kernel: smb_request: result -512, setting invalid
> Apr 11 13:43:54 testsmbbug kernel: smb_retry: caught signal
> Apr 11 13:43:54 testsmbbug kernel: smb_retry: caught signal
> Apr 11 13:43:55 testsmbbug kernel: smb_retry: successful, new pid=18163, 
> generation=8

The process accessing smbfs has caught a signal while accessing smbfs.
smbfs doesn't know how to tell that from a network error so it retries.

It does try to mask all signals except SIGKILL and SIGSTOP. You could try
modifying fs/smbfs/sock.c so that it only allows SIGKILL or even nothing
(just replace them with a 0).


A "cleaner" solution is being worked on. It separates the network access
from the process accessing smbfs using a kernel thread. This is part of
the "async requests" patch. But in its current state you shouldn't be
using that patch if you are afraid of very experimental code.

/Urban





More information about the samba mailing list