SMBMOUNT: process freezes on file copy

[Noel Kelly]

Thanks Urban.  I am now convinced that this problem is being caused by the
reshare of the Novell drive.  I have ended up copying the data (around 6Gb,
400,000 files) to the Windows 2000 server from the Novell server.  I then
mounted the Windows 2000 share using smbmount.  Copying this has been a
dream with no sign of any problems at all.

I think we will have to put this down to the MS Gateway Services for Netware
not keeping their end up and something being lost in the smbmount control as
a result.

I'll try and capture the network traffic...

Cheers,
Noel


-----Original Message-----
From: Urban Widmark [mailto:urban at teststation.com]
Sent: 05 December 2001 14:03
To: Noel Kelly
Cc: samba at lists.samba.org
Subject: smbfs: process freezes on file copy


On Tue, 4 Dec 2001, Noel Kelly wrote:

> Ducking in onto this thread seems like a better idea than starting a new
> one:  I also have a continuing problem with SMBMOUNT (at least I think it
is

Better to start a new thread for a different problem. Changing topic ...

> smbmount).  I had thought it was due to my inadvertently using the one
> supplied with RH6.2 rather than the compiled 2.2.2 version.  However I
still
> get hangs whilst copying files.

Which kernel version? If it is something that came with RH6.2 I suggest
you upgrade that first (to 2.2.19 or 2.4.16) before doing anything else
suggested here.

> I guess my problem is different in that the Windows client is NOT
> terminating the share (at least I don't think it is - the share I have
> mounted is from a Windows 2000 server SP2 running Netware Gateway Services
> and it has in turn mounted a Netware share to offer out).  The copying
> process seems to hit a particular file (not necessarily big - one it kept
> hanging on was a 5K GIF) and then the process freezes.  A "kill -9" is
then
> required and the share becomes useable again after a load of SMB messages
> are spewed onto the console.  Try the copy again and it stops at the same
> file again.  Delete/move that file and it will copy quite happily
gigabytes
> before fouling again.

The disconnects happen when the server thinks you have been idle for too
long. Not the case here.

Please redo the tcpdump but capture the full packets and write them to
file. (tcpdump -s 1600 -w /tmp/capturefile). I'd like to see what traffic
there is. If you could mail it to me (not the list, up to a few MBs) or
put it somewhere for me to download.

Or just look at them yourself (ethereal is nice). From the capture you
posted I'm thinking that the client is looping, and keeps fetching the
same page of the file over and over again. If it is you should be able to 
see that from the SMBread requests in the capture (ethereal will decode 
parts of SMB for you).

If you are copying gigabytes of files then start the capture after it
appears to have hung (or even better, just before you think it will hang).

Between runs is it always the same file it stops on? It would be possible
to patch smbfs to enable some extra logging for certain files ...


> Note that I don't actually kill the smbmount process but the copying
process
> - once the copying process is dead then the smbmount share becomes
available
> again.

The copying process has hung inside smbfs and is blocking everyone else.
When you kill it the others get access again.

/Urban