On Thu, Nov 12, 2015 at 4:03 AM, Sagi Grimberg <sagig@xxxxxxxxxxxxxxxxxx> wrote: > >> The bug is caused by this patch: >> >> 659743b02c411075b26601725947b21df0bb29c8 >> >> which allowed the task lists to be manipulated under different locks >> in the xmit and completion path. >> >> To fix the oops this patch just reverts that patch. It also reverts >> these 2 patches for regressions that were also a result of that patch: > > > Whoa now Mike, this is a major change. Can we take a step back and think > about this for a second? > > My understanding is that the kfifo circular buffer design allows a > writer (e.g. the producer) and a reader (e.g. the consumer) to avoid > extra locking when accessing the kfifo buffer. I don't think the problem is with the kfifo cmdpool, but rather the connection mgmtqueue/cmdqueue linked lists. iscsi_task structs are linked in using the task->running list_head. - Chris -- To unsubscribe from this list: 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
