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

["Rafael J. Wysocki" <rjw@sisk.pl>]

Hi,

On Monday 27 of September 2004 00:33, Pavel Machek wrote:
[-- snip --]
> > I've got two logs (attached), one of which is taken from the system with 
all 
> > modules loaded (swsusp.log), and the other comes from the system with no 
> > modules except for ipv6 (swsusp-nomod.log).  As you can see from the first 
> > log, the system with all modules loaded slows down significantly after 
> > pci_device_resume() is called for the device having vendor id = 0x10de 
> > (NVidia) and device id = 0x00d7 (no idea).  The system without modules is 
> 
> lspci, and take a look?

albercik:~ # lspci -n
[ ... ]
0000:00:02.0 Class 0c03: 10de:00d7 (rev a5)
0000:00:02.1 Class 0c03: 10de:00d7 (rev a5)
0000:00:02.2 Class 0c03: 10de:00d8 (rev a2)
[ ... ]

albercik:~ # lspci
[ ... ]
0000:00:02.0 USB Controller: nVidia Corporation nForce3 USB 1.1 (rev a5)
0000:00:02.1 USB Controller: nVidia Corporation nForce3 USB 1.1 (rev a5)
0000:00:02.2 USB Controller: nVidia Corporation nForce3 USB 2.0 (rev a2)
[ ... ]

So, it's the USB 1.1 controller (OHCI).

> > capable of writing 80-83% of pages to the swap _before_ slows down too and 
I 
> > have to wait for 1/2 h for the remaining ~20%.
> 
> Strange, *very* strange.

Yes, it is.

> > I'm afraid I can't get any more info until I sort out the sysrq
> > problem.
> 
> This should remap magic key to both-shifts-both-alts-key. Worked for
> me once...

Thanks a lot, but it turned out to be much simpler: sysrq was disabled by an 
initscript (shame on me, shame, shame).  I'll get the traces as soon as I can 
get something to attach to the serial console.  In the meantime, I'll try to 
look at the USB stuff.

Greets,
RJW

-- 
- Would you tell me, please, which way I ought to go from here?
- That depends a good deal on where you want to get to.
		-- Lewis Carroll "Alice's Adventures in Wonderland"