On Tue, Apr 14, 2009 at 12:41 PM, Dorrian, William M <william.m.dorr...@usace.army.mil> wrote: > Hi Darrel. > > The first thing I would suggest is to do a "modinfo cifs" and see which > version you're using. (It should be on the second row). I had a similar > problem a few months ago-I have a script which mounts a CIFS share every few > minutes, then does an rsync to a distant end Linux server. > > I was using version 1.50RH, if I'm not mistaken. I actually had to reboot my > Linux server every time I had the problem. > > I solved it (with the help of Steve French) by upgrading my kernel. A newer > CIFS module might solve your issue; Steve told me that 1.56 is more robust. > I'm not sure about 1.54 and 1.55, but they may be better also. > > -----Original Message----- > From: samba-bounces+william.m.dorrian=usace.army....@lists.samba.org > [mailto:samba-bounces+william.m.dorrian=usace.army....@lists.samba.org] On > Behalf Of Darrel Barton > Sent: Tuesday, April 14, 2009 1:45 PM > To: samba@lists.samba.org > Subject: [Samba] CIFS hang > > > I have a Proliant running Red Hat with the 2.6.9-5.ELsmp kernel. > > We have a Windows 2000 Server who's secondary disc drive is mounted on the > Linux box via CIFS using the following command: > > /sbin/mount.cifs //w2000/docs /var/www/documents -o > dir_mode=0777,file_mode=0444,gid=100,uid=501,username=winuser,password=winpas > s > > > From time to time, for whatever reason, the Windows 2000 Server requires > rebooting and if they reboot it without first allowing us to umount the > cifs, all requests to the cfis directory will hang. Requests, cd > commands, df commands ... all hang ... for as long as three HOURS and they > cannot be killed (kill -1, -15, -9) and the smb cannot be > restarted. Presumably this is all a result of the calls being hung within > the kernel. > > It has been suggested that we upgrade the kernel to 2.6.9-78.0.13. > > I'm not generally opposed to upgrades, but I'm hesitant to "upgrade to the > latest version" as a substitute for understanding the root of any problem > and engineering a proper fix. Maybe it's the years of Microsoft software, > but I believe that upgrading to the latest version has become tech support > slang for "we have NO IDEA what your problem is, so maybe the latest > version will fix it by accident -- but for very sure you'll be so deep in > the woods with NEW problems that you'll likely forget about this one > entirely." > > > Would anyone have any first hand knowledge of this problem and experience > that a later kernel is a solution? > Regards > Darrel > > -- > To unsubscribe from this list go to the following URL and read the > instructions: https://lists.samba.org/mailman/options/samba > -- > To unsubscribe from this list go to the following URL and read the > instructions: https://lists.samba.org/mailman/options/samba >
I think 1.56 is more robust because the sockets within cifs are now blocking instead of non-blockig which solved lot of data integrity issues due to lost responses etc.. But as soon as you notice hang, you can start a tcpdump on the cifs client box and issue a command like ls or cd (which hangs) and stop tcpdump after a while (2/3 minutes) and post tcpdump output file to look at. -- To unsubscribe from this list go to the following URL and read the instructions: https://lists.samba.org/mailman/options/samba