Re: aborts in usb-storage in branch 2.6

[Patrick Mansfield <patmans@us.ibm.com>]

On Fri, Oct 24, 2003 at 05:15:13PM -0400, Alan Stern wrote:
> On Thu, 23 Oct 2003, Pedro Larroy wrote:

> > abort.log -> 2.6.0test8
> > devices -> 2.6.0test8
> > kernel.log -> 2.6.0testx  where x < 8
> > 
> > In abort.log, is the paste of the log with kernel 2.6.0test8 in kern.log
> > there is a log from 2.6.0test4 or 2.6.0test4-mm
> > 
> > On previous 2.6.0testx kernels the bug appeared to trigger when ACPI debug
> > statements were printed, notice there are aborts after ACPI printks.
> > 
> > On 2.6.0test8 it seems to be triggered when doing concurrent access to the
> > filesystem.
> > 
> > I would like to help in anyway I can, but ACPI and USB is far beyond my
> > kernel knowledge, so I'm not currently able to fix the bugs myself.

> I looked at your logs, but I can't add much to what you already know.  
> This certainly does appear to be some sort of interrupt sharing/routing
> problem -- the logs don't indicate anything specific to USB.  However
> there's no easy way to tell what the source of that problem is.

The logs look like a SCSI timeout - a 30 second gap, followed by abort,
test unit ready, and a resend of the timed out write. That does not help
much. It could certainly be a lost interrupt.

I did not see any timeout/aborts happening (immediately) after the acpi
messages.

> Maybe someone who knows more about ACPI and interrupt handling can help.

I don't know ACPI, but I thought interrupt problems it might cause
happened more consistently.

-- Patrick Mansfield