[EMAIL PROTECTED] wrote:
Hi,
can i use usbtest also on host pc with kernel 2.4.21 (i think kernel 2.6.
seems to be recommanded) ? What are the beste test conditions and which other
applications can i use to test the controller driver (also windows
applications) ?
You'd need to backport usbtest to 2.4, and bring the 2.4 kernel up
to the quality level of 2.6 ... I'm pretty sure the HCDs there, and
usbcore, won't pass all those tests even for the APIs that are in
both kernels. I'd not bother; it's simple enough to run 2.6 kernels,
where those tests are known to work. (And "correct host" is a key
part of such test configurations ...)
One of these days I should probably write up a test checklist, but
here's much of what I usually do:
- "usbtest" on Linux 2.6 host with "Gadget Zero" to cover all
basic control and bulk functionality in sanity and stress
test modes. If any test fails to run overnight, that needs
fixing. (Except note that test #14 needs to use a smaller
upper bound, gadget zero only supports up to 256 byte control
writes at the moment.)
- USBCV from a Windows host. If you passed "usbtest as above,
this will probably work with no trouble.
- g_file_storage ... most any OS should recognize a storage
class device, and there are lots of disk access/stress tests
you can use. Copy files, check them. Run "bonnie". Etc.
See if you can make it trigger the halt processing.
- g_ether ... this is probably the best all-around test other
than "g_zero", since it uses queues in both directions and
(unlike g_file_storage) doesn't synchronize the two streams.
Use it with a Linux host (probably CDC Ethernet) for now,
but at least sanity test the less-stable RNDIS support with
a Windows host. SSH and so on should work; "ttcp" can be
used for load testing.
- gadgetfs ... the gadgetfs test program makes sure you can
properly defer ep0 responses (for "get descriptor"), and
lets you use AIO to build arbitrary queues. So it's the
best way to test isochronous transfers for now. (2.6-only,
unless you're brave ... the 2.4 version isn't as robust,
and I'm certainly not maintaining that.) That involves
some "usbtest" modes beyond control/bulk tests.
There's also g_serial, which isn't part of my current battery
of tests but which many people really want. (Not all boards
have real serial lines!)
Once all that works it's ready for real-world testing. Yes,
that's a fair amount of testing ... but usually once g_zero
starts to be robust in overnight testing, the rest goes very
quickly.
- Dave
-------------------------------------------------------
This SF.Net email is sponsored by: GNOME Foundation
Hackers Unite! GUADEC: The world's #1 Open Source Desktop Event.
GNOME Users and Developers European Conference, 28-30th June in Norway
http://2004/guadec.org
_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
