Red Hat Bugzilla – Bug 73656
nfs unreliable on 2.4.18-14
Last modified: 2007-04-18 12:46:28 EDT
From Bugzilla Helper: User-Agent: Mozilla/5.0 Galeon/1.2.5 (X11; Linux i686; U;) Gecko/20020809 Description of problem: After updating to kernel 2.4.18-14, athlon arch, on an NFS client, NFS became unreliable. I was playing ogg files, and recording a CD to ogg format using grip, when it told me that "access was denied writing mp3 file". I looked at the tracks that had been recorded, and noticed that 1, 2, and 3 were complete. 5 was being written. None of the disks were anywhere near full. I looked at the output of 'dmesg', and saw this repeated: RPC: sendmsg returned error 101 nfs: RPC call returned error 101 I've gone back to kernel 2.4.18-12.5, and don't see the issue any longer. The output of dmesg before I shut down 2.4.18-14 will follow as an attachment. Version-Release number of selected component (if applicable): How reproducible: Didn't try
Created attachment 75427 [details] dmesg output from machine in question
FWIW, I haven't seen this problem (between 2.4.18-14 client and 2.4.19-rc5aa1 server, NFS v3, TCP or UDP).
there are no NFS changes between -12.5 and -14.... :(
This is a long shot... but perhaps this may be remotely related to Bug 74315 where -14 crashes and -12.5 is 100% stable on my Athlon laptop. Perhaps (certain) Athlon related? yinyang, could you please post your "lspci -vn". barryn, what is your arch? Could you please post your "lspci -vn" too?
Warren~ Your experience certainly isn't mine. -12.5 was awful. Every time I boot it up, it runs for a couple of hours, and then programs start segfaulting when they exit. -14 seems to be doing just random corruption things. This morning, I booted the system and it reported that the root filesystem contained errors (and forced fsck). I haven't run memtest86 yet (again, that is, the system was good a year ago). The system was totally stable prior to the installation of Limbo.
Created attachment 76764 [details] output of lspci -vn
Arch for me is i686 (750MHz mobile PIII that has SpeedStep). I'll post lspci output if I get a chance.
Warren, yinyang: Have either of you tried an iX86 (like i686) kernel instead of an Athlon kernel, and does that make a difference?
Created attachment 76784 [details] output of "lspci -vn" on my system with no NFS problems under 2.4.18-14
Are you using dhcp to assign the nfs client's IP address? This could be related to bug 90174 which in my case can cause processes that frequently access NFS files to fail
I am indeed using dhcp on the NFS client. I'll try to "tail -f" a file until the lease expires and see if it dies at that time.
yinyang, are you still experiencing this problem with a supported release (i.e. Red Hat Enterprise Linux or Fedora Core)? If so, you should update the Product and Version of this bug. If not, this bug should be closed because Red Hat 8.0 has reached end-of-life.
I don't believe I've seen this problem anytime recently.