On Sat, 14 Jun 2003, David Brownell wrote:

> Major A wrote:
> >>[ 2.5.70 + ehci-0613 patch, log output sent, "Subject:" wrong ]
> >>
> >>Just to summarize for the peanut gallery:  the logs showed no
> >>problems at all ... until the PCI files spontaneously vanished
> >>from sysfs.  That's not allowed to happen, something's deeply
> >>wrong outside USB on that system if PCI goes that bad.
> > 
> > 
> > Isn't it only in the last ehci-* directory that files are missing? 
> 
> Yes, only the PCI files -- but not ones added by the EHCI driver.
> Though the usb5/* directory was there, without its USB files; and
> without the subdirectory indicating any device plugged in.
> 
> 
> > That is probably because cp fails to obtain these files after the
> > actual crash/freeze, I guess, and also the reason why cp can't be
> > killed (it's stuck in kernel mode).
> 
> There was also the problem that there were only usb-storage messages,
> while a misbehavior of that magnitude should have been spewing *all
> sorts* of informative messages.  You can use Alt-SysRq-T to find
> out the kernel stack info, too.  ("echo t > /proc/sysrq*" maybe.)

To enable the magic sysrq key, "echo 1 >/proc/sys/kernel/sysrq".

It wouldn't be surprising if the file that cp hung on was the same one 
that generated those usb_bulk/control_msg timeouts in the previous set of 
logs.  Here's one way to find out which file is causing the problem.  
Andras, before you do your big md5sum, use find to create a list of all 
the regular files (not directories) in the sysfs subtree that you're 
copying.  Then after usb-storage hangs up, have a little script go through 
that list, writing out the name of each file before trying to copy it 
somewhere else.  The last name written out will be the one causing a 
problem.

At that point, you could try removing that filename from the list and 
running the script again, to see if any other files also cause problems.

Also, there's not much point in including the sysfs copies from before 
the hang occurs.  All they do is show everything working correctly.  It's 
the state of the system after everything stops that matters.

Alan Stern



-------------------------------------------------------
This SF.NET email is sponsored by: eBay
Great deals on office technology -- on eBay now! Click here:
http://adfarm.mediaplex.com/ad/ck/711-11697-6916-5
_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel

Reply via email to