kern/177876: [mips] kernel stack overflow panic on mips64, EdgeRouter Lite

[Joe Holden]

Joe Holden wrote:
> Joe Holden wrote:
>> Warner Losh wrote:
>>> On Apr 24, 2013, at 11:27 AM, Joe Holden wrote:
>>>
>>>> Warner Losh wrote:
>>>>> On Apr 23, 2013, at 3:20 PM, Joe Holden wrote:
>>>>>> Warner Losh wrote:
>>>>>>> I just committed r249790 which makes everything match.
>>>>>>> On Apr 22, 2013, at 2:29 PM, Juli Mallett wrote:
>>>>>>>> On Mon, Apr 22, 2013 at 1:27 PM, Warner Losh <imp at bsdimp.com> 
>>>>>>>> wrote:
>>>>>>>>> It sure looks like you are right...  I have a full new tree 
>>>>>>>>> building just to make sure... :(
>>>>>>>>>
>>>>>>>>> Then again, it would be useful to document where these 
>>>>>>>>> dependencies lie to help prevent others from tripping over this 
>>>>>>>>> in the future :(
>>>>>>>> We should use genassym to generate trapframe offsets instead of 
>>>>>>>> having
>>>>>>>> hand-coded defines for these, perhaps?  I haven't thought it 
>>>>>>>> through
>>>>>>>> very fully, but that seems a lot better.
>>>>>>> I'll look into this...
>>>>>>> Warner
>>>>>> I rebuilt just the kernel but the same problem persists - started 
>>>>>> building world and I'm seeing:
>>>>>>
>>>>>> /usr/src/gnu/usr.bin/gdb/kgdb/trgt_mips.c:122: error: 'struct 
>>>>>> trapframe' has no member named 't4'
>>>>>>
>>>>>> and so on, have just commented them out for now so it completes, 
>>>>>> may just be left overs from previous build - may be getting 
>>>>>> confused :)
>>>>> You can just rebuild the kernel (I'll fix the kgdb stuff, I didn't 
>>>>> see that when I built the world for some reason).
>>>>> Please try this patch to see if this fixes it.
>>>>> Warner
>>>> So far so good, boots into multiuser - will give it a thrashing and 
>>>> see if I can make it break :)
>>>
>>> Thanks Joe.  I'll go ahead and commit it then and work on a longer 
>>> term solution.
>>>
>>> Warner
>>>
>> Excellent, still seeing trapframe panics under cpu load but it's at 
>> least usable!
>>
>> I'll let you get on with fiddling now :)
>>
>> Cheers
> 12 hours into a buildworld and it's still good, looks like whatever 
> buliding nano does to upset it hasn't happened yet!
> 
> Joe
Succcess! :)

Buildworld+kernel and a handful of ports succeeded (eventually), looks 
stable enough for day to day use now.

Thanks all!