8.0-RC1: AMD CS5536 (Geode) USB 2.0 controller strange behavour

[Hans Petter Selasky]

On Monday 28 September 2009 05:42:08 Eugene Grosbein wrote:
> >> I've repeated the test with my USB flash drive formatted as FAT32.
> >> It gives the same result: for ICH7-based system I get 26 megabyte/s
> >> file writing speed and for AMD CS5536 it suffers from long periods of
> >> inactivity while writing data, hence very low awarage writing speed.
> >>
> >> It seems the root of the problem is in the interaction
> >> between USB stack and AMD CS5536 (Geode) USB 2.0 controller
> >> and not in media devices themselves.
> >>
> >> Any more hints please?
> >
> > sysctl hw.usb.ehci.debug=15
>
> For USB HDD:
>
> %sysctl hw.usb.ehci.debug=15
> hw.usb.ehci.debug: 0 -> 15
> %date; dd if=/dev/zero bs=64k of=/dev/da0 count=100; date
> Mon Sep 28 11:25:26 KRAST 2009
> 100+0 records in
> 100+0 records out
> 6553600 bytes transferred in 14.912198 secs (439479 bytes/sec)
> Mon Sep 28 11:25:43 KRAST 2009
>
> The log is half-megabyte and is available here (13KB compressed):
>
> http://www.grosbein.pp.ru/ehci.log.gz

I don't see any timeouts in there. You need to run the test until you see 
something about "timeout".

>
> > If the USB stack puts the new job into the schedule and the USB
> > controller does not pick it up, it is not an USB stack problem ...
>
> If so, any workarounds possible?

None which I know about.

--HPS