Hi, I have been maintaining a patch for a quirky USB device built-in to
a UMPC, the Raon Digital Everun. Without the patch, the builtin
keyboard, mouse, and touchscreen are useless. I have never submitted it
for inclusion in the kernel because there are less than 5 people who
have ever even attempted to install Linux on this device, and it's
discontinued. However, *I* use the thing daily, and its wifi device
(libertas) has continued to get better support in the recent kernels so
I port the patch forward.
Anyway the patch adds a quirk handler that does a mandatory
initialization step to turn the USB device on at powerup or resume. In
recent kernels (I moved from 2.6.28, which was reliable, at 2.6.29-rc5,
I believe) this has become unreliable; about half the time the device is
not enabled at boot, and about 1 time in 10 it is disabled when resuming
from suspend-to-RAM.
I noticed that the structure of the quirk handling code has changed
somewhat; I left my patch using the "old" way of adding an entry to
hid_blacklist in hid-quirks.c. Is there a document or email trail that
describes the new way of structuring this quirk code? Any pitfalls to
look out for?
Thanks,
Bill Gribble
On Fri, 13 Mar 2009, Bill Gribble wrote:
> Hi, I have been maintaining a patch for a quirky USB device built-in to
> a UMPC, the Raon Digital Everun. Without the patch, the builtin
> keyboard, mouse, and touchscreen are useless. I have never submitted it
> for inclusion in the kernel because there are less than 5 people who
> have ever even attempted to install Linux on this device, and it's
> discontinued. However, *I* use the thing daily, and its wifi device
> (libertas) has continued to get better support in the recent kernels so
> I port the patch forward.
Hi Bill,
we have drivers (and even whole architectures) supported by the kernel,
which have less than 5 users in the world, that's not a big deal.
It makes 5 people happy, and saves you the pain from having to keep the
out-of-tree code working with the Linux kernel API changing all the time.
> I noticed that the structure of the quirk handling code has changed
> somewhat; I left my patch using the "old" way of adding an entry to
> hid_blacklist in hid-quirks.c. Is there a document or email trail that
> describes the new way of structuring this quirk code? Any pitfalls to
> look out for?
Looking at your patch, the only thing you need is sending a special URB to
the device in order to make it operational, right?
Please look at hid-sony driver, which does exactly that for some of the
PS3 devices.
You then only need to add your device to hid_blacklist[], so that generic
driver doesn't get bound to it and it'd be driven by your special driver.
--
Jiri Kosina
SUSE Labs
On Wed, 2009-03-18 at 15:14 +0100, Jiri Kosina wrote:
> we have drivers (and even whole architectures) supported by the kernel,
> which have less than 5 users in the world, that's not a big deal.
Great to hear that! I will try to clean up my patches according to the
guidelines and send them in.
> Looking at your patch, the only thing you need is sending a special URB to
> the device in order to make it operational, right?
That's right.
> Please look at hid-sony driver, which does exactly that for some of the
> PS3 devices.
>
> You then only need to add your device to hid_blacklist[], so that generic
> driver doesn't get bound to it and it'd be driven by your special driver.
OK, thanks for the advice. I will do that.
Bill Gribble