Kernel panic in moea_pvo_to_pte

[Nathan Whitehorn]

On 12/08/10 16:08, Justin Hibbits wrote:
> On Dec 7, 2010, at 9:56 AM, Nathan Whitehorn wrote:
>
>>
>> On Dec 7, 2010, at 7:23 AM, Justin Hibbits wrote:
>>
>>> Since I started rebuilding ports yesterday, I've been hit by this panic
>>> (copied from a low-res camera):
>>>
>>> panic: moea_pvo_to_pte: pvo 0x17e6810 pte does not match pte 
>>> 0x26e0740 in
>>> moea_pteg_table
>>>
>>> 0xe52f68a0: at panic+0x13c
>>> 0xe52f68f0: at moea_pvo_to_pte+0x154
>>> 0xe52f6910: at moea_pvo_remove+0x20
>>> 0xe52f6930: at moea_remove+0x94
>>> 0xe52f6960: at pmap_remove+0x80
>>> 0xe52f6980: at vm_map_delete+0x1d0
>>> 0xe52f69e0: at vm_map_remove+0x74
>>> 0xe52f6a00: at vmspace_exit+0x108
>>> 0xe52f6a30: at exit1+0x48c
>>> 0xe52f6a70: at sys_exit+0x1c
>>> 0xe52f6a80: at syscallenter+0x258
>>> ...
>>>
>>> The svn version is r215752, I'm currently building r216254 hoping 
>>> that might
>>> fix something.  This panic seems to occur after about 1-2 hours of 
>>> ports
>>> building, so is reproducible to me.  I also have the crash dumps 
>>> available
>>> (2).
>>
>> This is a serious bug. Could you rebuild the kernel with the attached 
>> patch? That should provide some information on why the kernel has 
>> triggered this panic. Also, could you provide some more details on 
>> your system? I'm guessing it is some variety of G4, but knowing the 
>> specific CPU model and whether it is SMP would be very helpful.
>> -Nathan
>> <test.diff>
>
> Took a couple days, it seems to only crash while building packages 
> within the x11/xorg metapackage umbrella, which is odd.... I built all 
> of Qt and the rest of my desktop with no crash.
>
> Anyway, the details are:
>
> panic: moea_pvo_to_pte: pvo 0x413f5960 pte does not match pte 
> 0x2a33340 in moea_pteg_table (0x87399bbf.0x570a0180, 
> 0x87399bf.0x570a6012)
>
> And, it is fully reproducible for me just by backing up ports and 
> building x11/xorg.

Weird. These are mappings for different, but nearby, pages, and the 
second mapping looks totally bogus (all WIMG bits are zero). It looks 
like you missed a character transcribing the second PTE values. Could 
you check them?
-Nathan