Tejun Heo wrote:
This oops happened because get_request() was invoked with NULL @q. It seems like SCSI midlayer refcounting mixup. I'll dig deeper and report again as soon as I can find something concrete.
Hello, all & James.
I've bisected the patch series and the winner is #221 git-scsi-misc.patch which seems to contain eight commits. I think SCSI people can hunt this down from now on. The bug happens whenever sr block device is accessed - mount, cat, whatever, and now it seems like some kind of data overrun.