> > > hello, > > > > > > Since we no longuer do internal queuing, we push the lock > > > inside ahc_linux_run_command and completly remove the bursted > > > internal queue freezing machinery. > > > Based on Hannes Reinecke patch for aic79xxx and Christoph > > > Hellwig comments. > > > > > > Signed-off-by: Emmanuel Fuste <emmanuel.fuste@xxxxxxxxxxx> > > > > > > --- > > > > > > aic7xxx_osm.c | 82 > > > ++++++++++++---------------------------------------------- > > > aic7xxx_osm.h | 1 > > > 2 files changed, 17 insertions(+), 66 deletions(-) > > > > > Ok, > > > > Kernel booted and works. > > cdrwtool -d /dev/sr0 -q still freeze the kernel after some time. > > Will try with "scsi_deactivate_tcq(sdev, 2);". > > It seems that there is some adherence between this and the > > recovery code. > > The recovery code success, but nothing was recovered in > > reality -> freeze. > > With a queue depth of 2, my hard disk continue to be serviced. > > It only my view of the pblm, and I could be (surely) wrong. > > It is only part of the problem since the driver receive a > > command that crash the driver/the sequencer but one thing at > > time... it could be view as a good stress test for the > > recovery code ;-). > > > > If you want traces/logs/etc... tell me. > Ok, more on this: > > I was wrong, the freeze is not caused by > "scsi_deactivate_tcq(sdev, 1);". > In fact, I've got the freeze since the second version of your > "semaphore to completion conversion" patch. I should not do > more than one thing at time ggrrrrrrhhhh. > > I will retry with your previous patch. > > It will be: > 2.6.15 > + James Bottomley "aic7xxx timer handling bug" fix > + "Turn off ordered flush barriers for SCSI driver" > + "semaphore to completion conversion V1" from your > + "scsi_deactivate_tcq(sdev, 1);" > + "Remove the platform_data->qfrozen logic" > > Results monday evening. Hello, Same result. First card dump, second card dump (aic7xxx_abort return 0x2002), freeze. So, the best result was my last try loggued on buzilla: http://bugzilla.kernel.org/show_bug.cgi?id=5921 It was: 2.6.15 + James Bottomley "aic7xxx timer handling bug" fix + "Turn off ordered flush barriers for SCSI driver" + "semaphore to completion conversion V1" from your There is some underlying assumptions which invalidate the result of additional obvious cleanups. I give up, it is out of my competence. It's a a pity that one of the most sell scsi chipset in the world (part of, because of good linux support) is in such a bad state. Feel free to send me any aic patch for debug/alpha/beta tests. It is the only way I could help for now. Regards, Emmanuel. Accédez au courrier électronique de La Poste : www.laposte.net ; 3615 LAPOSTENET (0,34 ?/mn) ; tél : 08 92 68 13 50 (0,34?/mn) - : send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
