Date: Thu, 10 Jun 2004 13:18:24 -0400 From: John Baldwin <jhb@FreeBSD.org> To: Anish Mistry <mistry.7@osu.edu> Cc: freebsd-hardware@FreeBSD.org Subject: Re: Fix for Logitech DiNovo cordless mouse [PATCH] Message-ID: <200406101318.24419.jhb@FreeBSD.org> In-Reply-To: <200406101215.24347.mistry.7@osu.edu> References: <200406091838.i59Icugc063061@smsgw.vianetworks.ch> <200406101122.35454.jhb@FreeBSD.org> <200406101215.24347.mistry.7@osu.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thursday 10 June 2004 12:15 pm, Anish Mistry wrote:
> On Thursday 10 June 2004 11:22 am, John Baldwin wrote:
> > On Wednesday 09 June 2004 08:16 pm, Anish Mistry wrote:
> > > On Wednesday 09 June 2004 02:38 pm, Markus Wild wrote:
> > > > Since yesterday I'm a happy owner of a Logitech dinovo
> > > > cordless USB keyboard/mouse. The keyboard works fine, however
> > > > the mouse didn't move a bit. I saw that other people had similar
> > > > "luck", so I enabled a bit of debugging. This is with
> > > > FreeBSD-current, btw.
> > > >
> > > > The result of the quest was: the hid.c:hid_report_size() function
> > > > returns a bogus iid value:
> > > >
> > > > Jun 9 19:37:06 mothra kernel: ums0: Logitech USB Receiver, rev
> > > > 1.10/24.04, addr 3, iclass 3/1
> > > > Jun 9 19:37:06 mothra kernel: ums_attach: bLength=7
> > > > bDescriptorType=5 bEndpoint Address=2-in bmAttributes=3
> > > > wMaxPacketSize=8 bInterval=10 Jun 9 19:37:06 mothra kernel: ums0: 7
> > > > buttons and Z dir.
> > > > Jun 9 19:37:06 mothra kernel: ums_attach: sc=0xc23a1800
> > > > Jun 9 19:37:06 mothra kernel: ums_attach: X 8/12
> > > > Jun 9 19:37:06 mothra kernel: ums_attach: Y 20/12
> > > > Jun 9 19:37:06 mothra kernel: ums_attach: Z 32/8
> > > > Jun 9 19:37:06 mothra kernel: ums_attach: B1 0/1
> > > > Jun 9 19:37:06 mothra kernel: ums_attach: B2 1/1
> > > > Jun 9 19:37:06 mothra kernel: ums_attach: B3 2/1
> > > > Jun 9 19:37:06 mothra kernel: ums_attach: B4 3/1
> > > > Jun 9 19:37:06 mothra kernel: ums_attach: B5 4/1
> > > > Jun 9 19:37:06 mothra kernel: ums_attach: B6 5/1
> > > > Jun 9 19:37:06 mothra kernel: ums_attach: B7 6/1
> > > > Jun 9 19:37:06 mothra kernel: ums_attach: size=36, id=17
> > > >
> > > > Since actual interrupt reports are issed with id 2:
> > > > Jun 9 18:42:10 mothra kernel: ums_intr: sc=0xc23a1800 status=0
> > > > Jun 9 18:42:10 mothra kernel: ums_intr: data = 02 00 fa
> > > >
> > > > So I added a bit of debugging to the id setting for-loop. It
> > > > looks like the ID cycles thru the following values at attach() time:
> > > > Jun 9 19:40:57 mothra kernel: hid_report_size: 00 -> 02
> > > > Jun 9 19:40:57 mothra kernel: hid_report_size: 02 -> 03
> > > > Jun 9 19:40:57 mothra kernel: hid_report_size: 03 -> 04
> > > > Jun 9 19:40:57 mothra kernel: hid_report_size: 04 -> 10
> > > > Jun 9 19:40:57 mothra kernel: hid_report_size: 10 -> 11
> > > > (numbers are hex here)
> > > >
> > > > With this, my current fix is simple: only set id if it's not
> > > > set already:
> > > > diff -u -r1.23 hid.c
> > > > --- hid.c 24 Aug 2003 17:55:54 -0000 1.23
> > > > +++ hid.c 9 Jun 2004 18:34:23 -0000
> > > > @@ -374,9 +374,10 @@
> > > > int size, id;
> > > >
> > > > id = 0;
> > > > + bzero (&h, sizeof (h));
> > > > for (d = hid_start_parse(buf, len, 1<<k); hid_get_item(d,
> > > > &h); ) - if (h.report_ID != 0)
> > > > - id = h.report_ID;
> > > > + if (h.report_ID != 0 && !id)
> > > > + id = h.report_ID;
> > > > hid_end_parse(d);
> > > > size = h.loc.pos;
> > > > if (id != 0) {
> > >
> > > I've attached at big patch that should fix the problem as well as a
> > > bunch of updates from the NetBSD sources. This is a patch against
> > > -CURRENT, so you may have to massage it a bit if you are on -STABLE. I
> > > won't have an offending device to test for at least a week so let me
> > > know of any problems.
> >
> > - M_ZERO is preferred to malloc() + memset(), so please don't make that
> > change.
> > - Lots of the changes add style bugs by adding spaces to the end of
> > lines. Please remove any trailing whitespace from your files.
> >
> > Other than that the patch looks cool.
>
> Ok, I think the style changes are done and I've attached the updated patch.
> I've got a question though, in style it says the enums should be in all
> caps, but the previous code wasn't and the NetBSD import code wasn't.
> Should I change it to what style says, or is it better to stay in sync with
> NetBSD?
Just keep it the way NetBSD has it. I also missed another minor nit (sorry!)
in that you want to leave the earlier sys/cdefs.h and FBSDID() lines as that
is a FreeBSD-specific change. Other than that it looks good.
--
John Baldwin <jhb@FreeBSD.org> <>< http://www.FreeBSD.org/~jhb/
"Power Users Use the Power to Serve" = http://www.FreeBSD.org
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200406101318.24419.jhb>