Red Hat Bugzilla – Bug 123852
scsi tape: hung "mt tell"
Last modified: 2007-11-30 17:10:43 EST
From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.6) Gecko/20040116 Description of problem: After a SCSI error, we cannot access the SCSI tape drive: there is a hung "mt tell" process and any access to the tape drive reports "Device or resource busy". The only known way to recover from this is by rebooting the machine. Instead, "mt tell" should not hang forever, but should report an error and exit. The currently hung machine is a dual-opteron box running fedora-1, but we periodically see this problem all machines with tape drives (and they all happen to be dual-cpu boxes). Version-Release number of selected component (if applicable): kernel-smp-2.4.22-1.2174.nptl How reproducible: Sometimes Additional info: Attached is stack trace from alt-sysrq-t: the "mt" process is stuck in "st_do_scsi->wait_for_completion". It looks like the SCSI error recovery path failed to wakeup "mt" after the SCSI request failed. Attached are the scsi errors from the system log (grep -i scsi /var/log/messages). K.O.
Created attachment 100404 [details] scsi errors from machine with hung "mt tell"
Created attachment 100405 [details] the real sysrq-T trace. The previous file was the dump of scsi errors.
Thanks for the bug report. However, Red Hat no longer maintains this version of the product. Please upgrade to the latest version and open a new bug if the problem persists. The Fedora Legacy project (http://fedoralegacy.org/) maintains some older releases, and if you believe this bug is interesting to them, please report the problem in the bug tracker at: http://bugzilla.fedora.us/
This problem no longer exists after we replaced the AMD64 FC1 kernel (2.4 based) with the AMD64 FC2 kernel (2.6.8-1.521smp). Userland is still FC1. K.O.