Bill Holler wrote:
> On 03/17/09 19:01, Li, Aubrey wrote:
>> Bill.Holler wrote:
>>
>>  
>>> Here is the webrev:
>>> http://cr.opensolaris.org/~bholler/6818652/
>>>
>>> The bug report should be here as soon as the web is refreshed:
>>> http://bugs.opensolaris.org/view_bug.do?bug_id=6818652
>>>
>>> Thank you,
>>> Bill
>>>     
>>
>> Hmm..., buggy BIOS, :(
>> Some changes in the patch are coding style fix, right?
>>   
>
> Yes. A number of lines wrapped around etc.
>
>> It looks good to me.
>>   
>
> Thank you!
>
>> Do we need to add ontrap protection to the other objects like _PSS, 
>> _TSS?
>>   
>
> I think so. ;-)
> However these other objects have been in use by Solaris long enough
> that they will not cause unexpected failures on systems during
> upgrades. I will see if _CST on_trap() protection can get into the
> next OpenSolaris build.
This strikes me as quite odd.  The returned package claims to have 2
elements, but access to Elements[0] results in a panic?  Perhaps there's
something else going on here; I don't think it should be necessary
to enclose every access to objects returned by ACPI CA in ontrap
protected sections.

In fact, I'd like to see the output of 'iasl -g' for the impacted system.
I'm thinking this may be an ACPI CA bug rather than a BIOS bug.
I haven't delved into the details of the code, but I suspect ACPI CA
constructs the returned object by parsing the BIOS AML, not by
accessing a table directly.

This fix may not be required at all - please hold off pushing it until we
do some more investigation.

Dana


Reply via email to