should i expect my SD card slot to be supported by a USB driver?

[Robert P. J. Day]

On Wed, 25 May 2011, J wrote:

> On Wed, May 25, 2011 at 07:04, Robert P. J. Day <rpjday at crashcourse.ca> wrote:
> > which doesn't do me a great deal of good if the device looks like a
> > SCSI USB storage device.  or am i misunderstanding something?  i was
> > anticipating running some of the MMC test routines on a blank card
> > but, under the circumstances, i'm not sure how i would do that.
> > thoughts?
>
> I would say try external card readers, but I'm almost positive those
> will also show up as USB devices since that's how they're connected.
>
> But this is all a matter of what card reader a manufacturer uses in
> a given system (and that can vary greatly from model to model or
> even between production runs on the same model), whether the reader
> presents as a USB device or perhaps a PCI device... so
> realistically, the only option unless you can find a way to fake it
> would be to buy another computer that has a card reader that doesn't
> resent as a usb device.  And I have no idea which ones those are at
> the moment... I have seen them before, but didn't pay enough
> attention to be able to list them.

  i'll do one more followup post, and leave it at that.  to recap,
i've got a hand-rolled, 2.6.39+ kernel running on my ubuntu 10.10
system, and i want to experiment with running the mmc_test module,
which lets you run a number of stress tests on a MMC/SD card.

  the most promising explanation is here:

http://www.spinics.net/lists/linux-mmc/msg05835.html

where you can see that, as far as that script is concerned, in order
to use the mmc_test module, you need to locate the driver that
*currently* has the MMC card and unbind it, at which point you bind
the mmc_test module and you're off and running.  and that's where the
problem lies.

  it's clear from that script and the explanation that you find here:

http://comments.gmane.org/gmane.linux.kernel.mmc/2692

that what the script *expects* to see is an existing entry under
/sys/bus/mmc/drivers, probably the standard "mmcblk" driver, which
explains what you see at that second link -- that under that existing
directory, there should be a symlink back to /sys/devices, from which
you extract the device name, in this case, "mmc0:0001", then you
use that to unbind from the MMC block driver, and bind to the mmc_test
driver.  i get all that.

  except on my system, there is no /sys/bus/mmc/drivers/mmcblk -- the
MMC card is handled by some combination of USB and SCSI drivers.  the
best i can find is that, since it shows up as /dev/sdb, i have this
under /sys/devices:

  /sys/bus/scsi/drivers/sd/21:0:0:0/block/sdb

i just don't know what to do with it.  if i could get by that
roadblock and figure out the device identifier to use in the bind and
unbind operations, i suspect the rest would be simple.

rday

-- 

========================================================================
Robert P. J. Day                                 Ottawa, Ontario, CANADA
                        http://crashcourse.ca

Twitter:                                       http://twitter.com/rpjday
LinkedIn:                               http://ca.linkedin.com/in/rpjday
========================================================================