Red Hat Bugzilla – Bug 17820
Problems with smbfs/smbmount on Linux client
Last modified: 2008-05-01 11:37:58 EDT
A while after I use smbmount to mount on my Linux machine a SMB share on a Windows NT server, it stops working. Any subsequent attempts of accessing it result in a "Input/output error" message. Also, smbumount doesn't work, and I have to umount it as root. df shows the filesystem /mnt/blurb is mounted from //blurb/share, but the filesystem isn't mounted - everything ends up at / instead (because it can't write to the smbfs /mnt/blurb). Some errors from /var/log/messages: Sep 21 04:02:28 vin kernel: smb_retry: caught signal Sep 21 09:44:28 vin kernel: smb_retry: signal failed, error=-3 Sep 21 09:44:43 vin kernel: smb_retry: signal failed, error=-3 Sep 21 09:44:51 vin kernel: smb_retry: signal failed, error=-3 Sep 21 09:44:56 vin kernel: smb_retry: signal failed, error=-3 samba-client and samba-common, both 2.0.7-4, are installed. The machine is not a samba-server so that package is not installed. The kernel-version is 2.2.16-3 (installed from RPM). I am having this problem on several boxes with the same configuration. I have seen several people reporting similar problems different places, but I haven't yet seen a solution that actually works... (only workarounds like umount the fs regularly and then smbmount it again. Not really a _solution_ :))
If you keep smb shares mounted for a long time you should have a look here: http://www.hojdpunkten.ac.se/054/samba/ It's the site of the smbfs kernel maintainer and it has patches for smbmount related problems too. Don't sure those patch will solve yout problem, though. :-)
I had this problem. I belive it's the NT server's fault. I fixed this by puting in crontab a script which echoed some text into file on the mounted filesystem every five minutes. Then, it was okey. G.
Yes, I have the same problems with RH 7.0 (2.2.16-22) and samba 2.0.7-23 (latest Rawhide) Looks like this problem is somehow related to network geometry because when I have linux station and WindowsNT machine in one segment with 224 mask (32 addresses) switched into ordinary hub, samba mount on linux remained stable fo weeks without any problems. The problems began when we changed Network geometry. The segment was expanded to 128 addresses (128 netmask /25) and both machines was switched into 10/100 3Com SuperStack II switch. Do not know exactly is this a problem of "switch" that inteligently filter broadcasts or of the segment expand. In a period of inactivity smb mount die and start putting error messages to syslog (smb_retry: signal failed, error=-3) But anyway, solution suggested by gdh@simnet.is corrected the problem. :-) P.S. I've tried latest samba-client-2.0.7-23 (from RawHide), problem still exist.
This may be the wrong venue to bring this up as it appears to be an NT issue. I am unable to duplicate this with W95/98 clients. Has anyone reported this to the Samba Team and gotten a response?
*** This bug has been marked as a duplicate of 11448 ***