Re: 2.6.9-rc2-mm3: swsusp horribly slow on AMD64

[Stefan Seyfried <seife@suse.de>]

Rafael J. Wysocki wrote:
> On Sunday 26 of September 2004 12:09, Pavel Machek wrote:

>>>>We have seen something similar after hdparm was used on specific
>>>>machines. Are you using hdparm?

Pavel, i am pretty sure the issue with hdparm and 32-bit disk access was
just a symptom, not the cause. Rafael, please try the patch i posted in
the other mail, i believe this is the right thing to do.

>>>Not explicitly, but it's used by SuSE initscripts to set IDE DMA, AFAICS.  
>>>However, the problem did not occur on 2.6.9-rc2-mm1 with the same 
>>>initscripts.
>>
>>Okay, so try what happens without the initscripts
> 
> I turned the stuff off but of course it didn't change anything. :-)

That's what i expected.

>>and try to locate change that breaks it...

> Well, I'm a bit confused:
> 
> --- linux-2.6.9-rc2-mm1/kernel/power/swsusp.c   2004-09-16 14:06:56.000000000 
> +0200
> +++ linux-2.6.9-rc2-mm3/kernel/power/swsusp.c   2004-09-24 11:35:18.000000000 
> +0200
> @@ -862,8 +862,8 @@
>         error = swsusp_arch_suspend();
>         /* Restore control flow magically appears here */
>         restore_processor_state();
> -       local_irq_enable();
>         restore_highmem();
> +       local_irq_enable();
>         return error;
>  }

without this one is needed or highmem will break "sometimes". Was really
nasty. You did have highmem-resume problems, didn't you?

    Stefan