Re: S3 test tool (was : Re: Bizarre oops after suspend to RAM (was: Re: [ACPI] Resume from Suspend to RAM))

[dagit@codersbase.com]

Pavel Machek <pavel@ucw.cz> writes:

> Hi!
>
>> If you've looked at this bug you will know that myself at and atleast
>> one other person experience a reboot on resume at a specific line in
>> the wakeup code:
>> http://bugme.osdl.org/show_bug.cgi?id=3586
>> 
>> One note about the code in the bug, my code for detecting PM is
>> backwards, so ignore it, what I say in this email is still valid.
>> 
>> Specifically, if I get rid of the pushl;popl then the computer does
>> not reboot.  See the attached diff.  The question is 1) is this
>> pushl;popl the final nail in the coffin? 2) Does windows not clear the
>> flags completely, but instead sets them to some "special value"?
>> 
>> The reason for (1) is because as I understand it, when a certain
>> number of illegal operations (3 iirc) are issued at certain times
>> (real mode iirc) the machine automatically reboots.  That could be
>> what we are seeing here.
>
> You got this wrong. It is three illegal instructions but
> *nested*. Like error, error in fault handler, error in doublefault
> handler.

Ah.  Yeah, this isn't an area I know much about :)  Thanks for the
correction. 

> Try replacing flags manipulation with any stack manipulation to see
> what is wrong.

Do you mean try something like this? Replace the push 0 with push
0x1234 ; push 0x1234 ; pop ; pop and try to figure out which line
causes the reboot?

> Like perhaps processor docs?

Is that what I want to look at?  I was the one asking the question. ;)

Thanks,
Jason