On Mon, Oct 8, 2012 at 6:22 PM, Ben Hutchings <ben@xxxxxxxxxxxxxxx> wrote: > On Mon, 2012-10-08 at 10:37 -0600, Bjorn Helgaas wrote: >> On Mon, Oct 8, 2012 at 10:28 AM, Greg KH <greg@xxxxxxxxx> wrote: >> > On Sun, Oct 07, 2012 at 09:57:47PM +0100, Ben Hutchings wrote: >> >> Based on <https://bugzilla.kernel.org/show_bug.cgi?id=46481>, the >> >> following commit should be cherry-picked for stable series < 3.6: >> >> >> >> commit dfb117b3e50c52c7b3416db4a4569224b8db80bb >> >> Author: Bjorn Helgaas <bhelgaas@xxxxxxxxxx> >> >> Date: Wed Jun 20 16:18:29 2012 -0600 >> >> >> >> PCI: acpiphp: check whether _ADR evaluation succeeded >> >> >> >> It applies cleanly to 3.2.y and presumably also to later stable >> >> branches. I've attached a version that applies to 2.6.32.y and 3.0.y, >> >> and presumably also to the intermediate stable branches. Only 3.2.y has >> >> actually been tested with this, though. >> > >> > Thanks, now applied. >> >> Has anybody looked deeper to see exactly *how* this patch fixes the >> problem? The original reports were oopses caused by dereferencing bad >> pointers in DRM, and it's not obvious to me how this patch is >> connected to those. > > Oh, I completely missed the oops log. Yes, it does look like there's > another more serious bug still to be fixed in 3.2.y (maybe in DRM, maybe > not). With a lot of help from Jeroen Nijhof <jeroen@xxxxxxxxxxxxxxxxxx> and Stefan Nagy <public@xxxxxxxxxxxxxx>, I analyzed this and convinced myself that the DRM oopses were in fact caused by the bogus slot we registered when we ignored the _ADR evaluation failures. More details in https://bugzilla.kernel.org/show_bug.cgi?id=46481 So I believe this _ADR problem was the root cause and we don't need to look for more DRM issues. Bjorn -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
