From Bugzilla Helper: User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.6) Gecko/20040206 Firefox/0.8 Description of problem: When trying to write large file to SCSI disk controlled by aic7xxx driver, it resets/aborts several times, then dumps state and offlines the device. Filesystem module (reiserfs) then complains about not being able to write journal. Version-Release number of selected component (if applicable): kernel-2.6.3-1.100 How reproducible: Didn't try second time Steps to Reproduce: 1. mv kernel-source-2.6.3-1.100-i386.rpm <filesystem-on-SCSI> 2. watch /var/log/messages Actual Results: Feb 24 22:50:32 pavelp kernel: scsi0:0:2:0: Attempting to queue an ABORT messageFeb 24 22:50:32 pavelp kernel: CDB: 0x2a 0x0 0x0 0x1 0xd4 0x8d 0x0 0x0 0x20 0x0 Feb 24 22:50:32 pavelp kernel: scsi0:0:2:0: Command already completed Feb 24 22:50:32 pavelp kernel: aic7xxx_abort returns 0x2002 Feb 24 22:50:42 pavelp kernel: scsi0:0:2:0: Attempting to queue an ABORT messageFeb 24 22:50:42 pavelp kernel: CDB: 0x0 0x0 0x0 0x0 0x0 0x0 Feb 24 22:50:42 pavelp kernel: scsi0:0:2:0: Command already completed Feb 24 22:50:42 pavelp kernel: aic7xxx_abort returns 0x2002 will attach the complete log Expected Results: no failures Additional info:
Created attachment 98040 [details] Excerpt from /var/log/messages
this might be related to bug 115703
Same happens with kernel-2.6.3-1.100 This time no heavy activity was done - the failure happened when trying to umount the filesystem as part of shutdown sequence. Messages were almost identical to those reported before.
Re-confirmed for 2.6.3-2.1.253. No heavy activity is needed for the driver to fail, only an attempt to read directory. aic7xxx_old works fine with the same kernel - could be something as simple as compile-time mis-configuration of aic7xxx (new) module.
This is a known problem, the patch exists in -mm tree: http://www.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.4/2.6.4-mm2/broken-out/aic7xxx-deadlock-fix.patch Please apply to RPM build pending inclusion to Linus tree. The patch completely fixes this problem.
fixed in the current errata kernel ?
Fedora Core 2 has now reached end of life, and no further updates will be provided by Red Hat. The Fedora legacy project will be producing further kernel updates for security problems only. If this bug has not been fixed in the latest Fedora Core 2 update kernel, please try to reproduce it under Fedora Core 3, and reopen if necessary, changing the product version accordingly. Thank you.