Re: [PATCH] intel-agp.c: Fix crash when accessing nonexistent GTT entries in i915

[Miguel Ojeda <miguel.ojeda.sandonis@gmail.com>]

On Tue, Mar 23, 2010 at 1:57 AM, Andrew Morton
<akpm@linux-foundation.org> wrote:
> On Sun, 21 Mar 2010 16:30:20 +0100 Miguel Ojeda <miguel.ojeda.sandonis@gmail.com> wrote:
>
>> On Sun, Mar 21, 2010 at 2:58 PM, Zhenyu Wang <zhenyuw@linux.intel.com> wrote:
>> > On 2010.03.20 14:04:56 +0100, Miguel Ojeda wrote:
>> >> On Fri, Mar 19, 2010 at 9:27 PM, Andrew Morton
>> >> <akpm@linux-foundation.org> wrote:
>> >> > On Thu, 11 Mar 2010 16:54:26 +0100
>> >> > Miguel Ojeda <miguel.ojeda.sandonis@gmail.com> wrote:
>> >> >
>> >> >> Attached dmesg, lspci -vv, config and xorg.
>> >> >>
>> >> >> When the X server crashes, the kernel does not report anything
>> >> >
>> >> > This seems to have gone all quiet?
>> >>
>> >> It seems so, I was about to ping Zhenyu.
>> >
>> > Could you try recent X.org intel driver release? Your failure X log
>> > showed some pretty old UMS driver's render flush time out message,
>> > I think whose version never has been well tested with new kms/gem stuff.
>> >
>>
>> Sure. Debian Stable got version 2.3.
>>
>> >>
>> >> >
>> >> > As this was a 2.6.32->2.6.32.4 regression, I assume that it's also a
>> >> > 2.6.32->2.6.33 regression?
>> >>
>> >> Yep. All kernels I tested since 2.6.32.4 crash, including 2.6.32.10,
>> >> 2.6.33 and 2.6.34-rc1. See my original message for more details about
>> >> the crash.
>> >>
>> >
>> > Could you bisect? I doubt it's caused by David's patch, as if it is, you
>> > will have trouble at early agp init time, instead of current problem looks
>> > like something causing rendering hang..and I can't think of how mapping GTT
>> > to a scratch page could cause problem in case GTT bar is truely 256K.
>>
>> I bisected in order to find the commit 5877960869333e42ebeb733e8d9d5630ff96d350.
>
> I have no 5877960869333e42ebeb733e8d9d5630ff96d350 here.  This is why
> we ask that people always identify patches by both the hash and the
> full title..

Sorry, I thought that saying 2.6.32.x and the commit would be enough:.
This is the one I meant:

http://git.kernel.org/?p=linux/kernel/git/stable/linux-2.6.32.y.git;a=commit;h=5877960869333e42ebeb733e8d9d5630ff96d350

>
>
>> I don't know if such commit is really the true problem (does not seem
>> so if you are right); however, it is the commit that breaks the X
>> server as I stated in the original message.
>
>