On Fri, 21 May 2004, David Brownell wrote:

> Guennadi Liakhovetski wrote:
> >
> > Hence the main question - what could this BUG mean?
>
> Usually it means that the HCD freed a TD before the HC was
> quite done with it ... so the HC put it on the donelist,
> and donelist processing tried to find the TD given its
> DMA address (using that routine), then BUG().
>
> There are some fairly non-obvious races that can cause that,
> and even the 2.6 code had a bug in that area that was only
> fixed in 2.6.5 or so ... which bug will certainly be in the
> 2.4 usb-ohci code, it came from there.  That one involved
> the "unlink empty ED" case taking an inappropriate shortcut;
> it really needs to wait a frame like all the other cases.
> (However, the usb-ohci "wait a frame" logic is broken too,
> in ways that showed up mostly on slower systems.  I think a
> PXA would hit those bugs often enough.)

Ok, thanks. Then, 2 other questions - can these races or some other bugs
in ohci also explain why the driver works fine on the PCI-controller
internal OHCI chip and doesn't work so well on the cardbus USB2.0 card? I
also tried a PCI USB card with 4 ports (OHCI too). And that one produced
quite funny results: on the first power-on the kernel boots as far as up
to "freeing init memory". After a reboot it gets a couple lines further. 2
more reboots and you get a prompt, which works for some time (a couple of
seconds / minutes) - if I have ohci hard linked in the kernel... Then the
system just hangs. If I compile USB as modules, the system hangs as soon
as I load usb-ohci-pci. A number on network cards in the same PCI slot
work fine.

Besides, I tried one more CardBus card - a Xircom ethernet card, which
also has problems - after some time / amount of data transferred, the
network starts to lose packets.

So, it looks like the problem is not (or, at least, not only) in the ohci
driver. Rather - from the USB-OHCI point of view, which PCI / DMA /
CardBus / ... problems can lead to this BUG in dma_to_ed_td?

Thanks
Guennadi
---------------------------------
Guennadi Liakhovetski, Ph.D.
DSA Daten- und Systemtechnik GmbH
Pascalstr. 28
D-52076 Aachen
Germany





-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. 
Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click
_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel

Reply via email to