Re: Remove HID ignore quirk for MGE UPS devices

On Mon, 2008-09-01 at 13:50 +0200, Arnaud Quette wrote:

first, note that I'm the originator of the below mentioned (2.4.25)
patch, and the one that has developed the USB UPS support code (helped
only for the kernel part, and done the userland one).

After reading al the kernel and NUT HID stuff, I'm really not convinced
any of the low level HID table parsing should be done in userspace at
all.

2008/8/15 Richard Hughes [off-list ref]:

quoted

The attached patch reverts the change made four years ago here:
http://www.vg.kernel.org/pub/linux/kernel/people/gregkh/usb/2.4/usb-hid-2.4.25-pre7.patch

UPS's made by MGE can be used with usbhid just fine, and by removing the
ignore quirk allows them to be used with HAL so they just work when plugged
in, without needing to be manually configured.

this is true for recent units, but hiddev (not itself but something in
the USB stack at that time) was breaking the communication with some
of MGE *UPS* small devices. Note that this might have been solved
since then.

Right, so it should be a tiny patch to the kernel HID parser, if it's
not been fixed already.

the assertion that is false is that without this, HAL support can't work.
the code to support USB UPSs that I've developed since this oldish
kernel since that allows this (Richard, you have missed to compile nut
with --with-hal...). This latter is more complete (support all USB HID
UPS + all USB devices, along with more features, more specific
handling, more data, more ... everything) and is *portable* (or
ported) on other USB enabled Unix (*bsd, solaris and darwin). So as to
the HAL support code using this exact same NUT code.

Sure, I've used --with-hal myself. The number of desktop users who
install NUT and use the hal-nut sub package is virtually nil in the real
world. It also excludes server machines that don't want or need HAL, and
doesn't really work well when DeviceKit starts being used instead. I've
got thousands of lshal outputs from desktop machines, and the majority
of USB UPS devices are not managed by NUT as the users probably think it
"doesn't work with Linux".

I appreciate some of the older more random serial cable units probably
belong in a userspace library, but when the new standard USB HID stuff
in the kernel makes this just work without any drivers, libraries or
daemons, then I really think we should use that.

quoted

With the ignore quirk in place a user would have to configure NUT before the
UPS could be used, as NUT uses it's own internal USB matching framework
to match against the USB devices, do low level control messages on the
device and then parse the HID tables all in userspace.

wrong, see above.

I don't see why a user should have to install a multi-megabyte library
and daemon just to use the power page device that the kernel already
supports. This stuff can _just work_, and does for all the other of
thousands of UPS devices that support the HID power page.

The only thing I can say, as in 2004 is that in kernel UPS handling is
too fat, incomplete and unuseful.

Right, so the proper way to do this is to work with the subsystem
maintainers, and push the quirk patches lower into the kernel as they
are probably useful for other devices.

the userland support is there, and for some time mature, propose more
features, best device handling (and there are many devices specifics
out there)

Right, there are a ton of HID quirks right now, I don't think adding a
couple more for obsolete devices there would be a problem.

, support all USB UPSs (not only HID ones), support HAL and
hotplug / 0 config things and is portable.

the PDC support is hiddev should simply be removed! that would give a
small diet to the kernel too...

I don't agree. After all the work I've done with userspace drivers the
conclusion I've come to is that they are more complex, slower and often
are not as stable as in kernel drivers. If the PDC page is removed, then
NUT has to be installed by default on every machine, just in case a USB
UPS is ever inserted. I don't think that's a good idea.

Richard.