On Fri, 1 Jul 2011, James Bottomley wrote: > On Fri, 2011-07-01 at 17:04 -0400, Alan Stern wrote: > > On Fri, 1 Jul 2011, James Bottomley wrote: > > > > > On Fri, 2011-07-01 at 14:14 -0400, Dave Jones wrote: > > > > On Fri, Jul 01, 2011 at 10:05:31AM -0700, Andi Kleen wrote: > > > > > > > > > I found I can reliably crash a 3.0 system by pulling the > > > > > USB cable of a mounted USB cdrom (or rather a USB device which > > > > > has a builtin fake CD-ROM) > > > > > > > > > > I suspect it's a regression too. > > > > > > > > We've been seeing a lot of similar bugs in Fedora since we pushed > > > > a 2.6.38.8 update. Some of the traces are different, but some > > > > look to be the same as yours. (here's one for eg: https://bugzilla.redhat.com/show_bug.cgi?id=712830) > > > > > > > > The common cause seems to be 'device went away'. So USB CD drives, > > > > USB memory sticks, and for some reason virtualbox shutdown. > > > > > > I think it's something specific in the USB path. I can't reproduce on > > > 3.0-rc5 with a SATA DVD hot unplug. USB cc's added. > > > > I just took a look at the Red Hat bugzilla entry mentioned above. It > > seems to be quite different from the issue addressed by the patch I > > just posted -- a crash with invalid memory access rather than a lockdep > > violation and hang of the khubd thread. > > > > It's also notable that the stack dump in the bugzilla report doesn't > > contain any functions in the USB subsystem. Of course this doesn't > > prove anything, but it is suggestive. > > > > Evidently the sdev argument to scsi_prep_state_check() was NULL. This > > looks like the problem that cropped up before, where q->queuedata was > > getting set to NULL while the queue was still in use. I can't imagine > > how anything in usb-storage could have caused that. > > Right ... the device release function has already been called, so it's > some type of refcounting cockup. The fact that I can't reproduce with a > SATA unplug is what makes me think it might be USB specific ... of > course, that's not definitive. I'm not able to reproduce it on a vanilla 3.0-rc5 system. Can anybody give the exact sequence of steps you went through to trigger the bug? Alan Stern -- 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
