USB/HID UPS issue (was Re: USB scanner issue)

USB/HID UPS issue (was Re: USB scanner issue)

[arnaud.quette]

Hi folks,

I take the opportunity of the previous discussion (USB
scanner obsolescence) to relaunch the one about hidddev.


On Mon, Dec 01, 2003 at 11:21:58AM -0800, Greg KH wrote:
> ...
> Can't you use xsane without the scanner kernel driver?  I thought the
> latest versions used libusb/usbfs to talk directly to the hardware.
> Because of this, the USB scanner driver is marked to be removed from the
> kernel sometime in the near future.

I'm thinking about doing the same with hiddev
as I'm facing new problems each times we solve
one! Currently, MGE devices won't work fine with
latest kernels (2.4 and 2.6). I've identified that
config/report descriptors are queried another time
at the end of HID enumeration, but the device doesn't
have time to answer (about 1,5 sec, instead of the
standard "5 sec" timeout). I've not found the cause
after digging, so I've directly gone to the next
step.

So I've setup the libHID project [1] to allow this
HID support, only used by UPSs at the moment (I've
seen passing some comments about USB radio and
other gadgets). But others complex HID devices are
welcomed.

My last problem is that I can't succeed in deactivating
hiddev at runtime (I've tryed various form of
"alias hiddev off" in modules.conf without success).
This still cause nasty side effect to the device.

So my question is: If we have userland HID support
through libHID (in the same way as libusb), is hiddev
still needed? I obviously don't talk about hid (core),
joy/keyb/mouse which are still needed...

Your feedback about all of that is welcomed.

Arnaud

[1] http://savannah.nongnu.org/projects/libhid
http://www.ghz.cc/~clepple/libHID/doc/html/index.html

Re: USB/HID UPS issue (was Re: USB scanner issue)

[reg]

I havent anything to add to your comments, except to make sure you saw
my comments about USB/HID problems that I have encountered with a UPS.

Since the message was short, I reproduce it here:

---

(previous post to linux-kernel)

I am posting this here since I dont seem to be able to post to
either of the USB lists.

In working with the code apcupsd, I have found two problems that appeare
in the 2.6.0-testX kernels that did not appear in the 2.4.x series of kernels.

    (1) When doing a read to get hiddev_event structures, 2.4 only
	gave the 'real' events from the device that one expected.
	Under 2.6.0-testx there are several ZERO event structures/sec
	where the entire structure is ZERO, both hid and value.

	For the current code there may be a 'real' event every few
	seconds, and 5-10 of these zero events/sec.  I have no
	idea where they are coming from.

    (2) In one thread the code does a select, followed by a read if
	data is available.  If one just 'falls thru' to the read with
	the few lines of code it takes to do the checking, one gets
	up to 45000 messages/minute (750/sec) reading:

	    kernel: drivers/usb/input/hid-core.c: control queue full

	If one puts a 1/10sec sleep between these two commands, the
	error messages go away. 

Anyone know anything about either of these errors?
Or how to report them to the USB people if you cant post to the USB lists?
 
-- 
                                        Reg.Clemens
                                        reg@dwf.com