Re: [linux-usb-devel] USB Storage oops in 2.5.69-bk8

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

On Sat, May 17, 2003 at 11:29:30AM -0400, Alan Stern wrote:
> On Fri, 16 May 2003, Mike Anderson wrote:

> What happens is that the device crashes because of the 3-byte 
> MODE-SENSE(6) request.  The scsi error handler tries various levels of 
> fault recovery, all the way up to a bus reset.  Usb-storage implements the 
> bus reset as a USB port reset, which requires usbcore to re-enumerate 
> the device.  The line
> 
>    drivers/usb/core/message.c: usb_control/bulk_msg: timeout
> 
> refers to a failure during the re-enumeration.  Apparently even the port
> reset was not enough to revive the device.  I don't know the reason for
> the line
> 
>    Error handler scsi_eh_1 waking up

The above is the first thing output when the eh wakes up, so the
re-enumeration could not have started yet (at least it could not have been
initiated via scsi eh calling the bus reset or other abort/reset related
functions), and the USB timeout message could not be part of the scsi eh,
though we did not see the scsi_times_out log message (since it's wrapped
in SCSI_LOG_TIMEOUT, not SCSI_LOG_ERROR_RECOVERY).

Turning on timeout (or even all) logging might give a bit more data:

	echo "scsi log timeout 4" > /proc/scsi/scsi

Or (warning - logs the logging messages if logging to a scsi disk):

	echo "scsi log all" > /proc/scsi/scsi

-- Patrick Mansfield