Re: When will the lunacy end? (Was Re: [PATCH] uswsusp: add pmops->{prepare,enter,finish} support (aka "platform mode"))

[Pavel Machek <pavel@ucw.cz>]

Hi!

On Mon 2006-09-25 16:06:48, Andrew Morton wrote:
> On Tue, 26 Sep 2006 00:45:00 +0200
> Pavel Machek <pavel@ucw.cz> wrote:
> 
> > Anyways this boils down to "find which drivers are delaying suspend
> > and fix them".
> 
> The first step would be "find some way of identifying where all the time is
> being spent".
> 
> Right now, netconsole gets disabled (or makes the machine hang) and most of
> these machines don't have serial ports and the printk buffer gets lost
> during resume.
> 
> The net result is that the machine takes a long time to suspend and resume,
> and you don't have a clue *why*.
> 
> And this is a significant issue, IMO.  In terms of
> niceness-of-user-interface, being able to suspend in twelve seconds instead
> of twenty seven rates fairly highly...

Your machines spend 15 seconds in drivers? Ouch, I did not realize
_that_. 

(My machine suspends in 7 seconds, perhaps 2-3 of that are playing
with drivers, so I just failed to see where the problem is).

Are these your big SMP servers? Any SCSI involved?

Rafael has "fakesuspend" patches somewhere, but you can probably just
swapoff -a, then echo disk > /sys/power/state. If you are lucky, that
should be slow, too... fortunately you'll have useful dmesg buffer
when you are done. CONFIG_PRINTK_TIMING or something, and you should
have enough clues...?

15 seconds spend within drivers is definitely _not_ okay.
								Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html