Attached is the full kernel debug log file.
The hard hang occurs with the following log message:
May 27 10:55:31 draal kernel: uhci: host controller process error. something bad
happened
May 27 10:55:31 draal kernel: uhci: host controller halted. very bad
May 27 10:55:54 draal kernel: usb-storage: command_abort() called
The writing process is then unkillable, the usb modules are
un-rmmod-able, and I must reboot the machine to return to normal.
The above is using the uhci driver. This also happens with the usb-uhci
driver. It happens on both my machines, one is an athlon, Asus A7V
mobo, the other is an alpha ev56 (LX164) with a PCI USB adapter. (both
UHCI)
Matthew Dharm [[EMAIL PROTECTED]] wrote:
> Could you turn on USB Mass Storage verbose debugging and send a log of the
> simplest case tht does something strange?
>
> Matt
>
> On Mon, May 14, 2001 at 05:18:12PM -0500, Bob McElrath wrote:
> > I'm using the usb-storage driver to read/write an MMC card using the
> > SanDisk ImageMate (SDDR12). Under kernel 2.4.4, and usb-uhci I'm having
> > the following problems:
> >
> > * The application hangs when it closes the filehandle. Currently I
> > fh = open("/dev/sdc", O_RDWR) the device, write a bunch of stuff, and
> > then close(fh). Upon close the program hangs in the "D" state as
> > reported by ps (forever...). The program worked with older kernels...
> > The program also works fine if no data is written.
> >
> > * When it did work, the close(fh) would also hang for several seconds,
> > but eventually return. My program transfers files to the MMC, and
> > when transferring the file, the orange light on the device blinks
> > rapidly. When the transferring is done, and the filehandle is closed,
> > the light on the device turns of for several seconds. I've tried
> > inserting ioctl(fh, SCSI_IOCTL_SYNC) in several places, but this
> > didn't seem to have any affect. The amount of time it hung was
> > independent of how much data was transferred, unless no data was
> > written, then there was no hang.
> >
> > So in either case it seems like something funny is happening when the
> > device is closed, and data was written. Does anyone have any ideas off
> > the top of their heads? Suggestions as to how I should go about
> > tracking this down?
> >
> > Does anyone else actually use the SDDR12 under linux, or am I the only
> > one? ;)
> >
> > The software in question is mjmmc, part of mj-tools
> > (http://www.sourceforge.net/projects/mj-tools)
Cheers,
-- Bob
Bob McElrath ([EMAIL PROTECTED])
Univ. of Wisconsin at Madison, Department of Physics
kern.debug.gz
PGP signature
