[Skiboot] [PATCH v7 08/22] hdata: Add support for early OPAL crash

[Vasant Hegde]

On 05/16/2019 10:38 AM, Nicholas Piggin wrote:
> Vasant Hegde's on May 14, 2019 9:33 pm:
>> On 05/09/2019 10:59 AM, Nicholas Piggin wrote:
>>> Vasant Hegde's on April 13, 2019 7:15 pm:
>>>> Lets hardcode few entries in MDST and MDDT table.. so that we can capture
>>>> early OPAL crash. Note that we will have 1 to 1 mapping between MDST and
>>>> MDDT table entry.
>>>
>>> How useful is this? It could be useful if:
>>> - crash is not reproducable, and
>>> - customer site where we can't access the BMC, and
>>> - system eventually boots so crash dump can be saved, and
>>> - dump is sufficient to debug (i.e., debugging does not involve hardware
>>>     registers etc).
>>
>> Correct.
>>
>>>
>>> I'm a bit skeptical. I reckon we'd get more milage out of changing assert
>>> to trap so we get regs.
>>
>> We do capture registers.. but yes, its useful only when system boots eventually.
> 
> I'd like to consider holding off the OPAL crash feature (presumably
> the Linux dump is significantly more useful and has more moving parts
> to solve). We should be able to improve boot crashes a lot without
> this, and then reconsider if it's still wanted.

You mean early OPAL crashes? Sure. We can defer that.


-Vasant