Re: em28xx: commit aab3125c43d8fecc7134e5f1e729fabf4dd196da broke HVR 900

[Mauro Carvalho Chehab <mchehab@redhat.com>]

Em Thu, 21 Mar 2013 16:34:13 +0100
Hans Verkuil <hverkuil@xs4all.nl> escreveu:

> On Thu March 21 2013 11:03:27 Mauro Carvalho Chehab wrote:
> > Em Thu, 21 Mar 2013 09:33:41 +0100
> > Hans Verkuil <hverkuil@xs4all.nl> escreveu:
> > 
> > > I tried to use my HVR 900 stick today and discovered that it no longer worked.
> > > I traced it to commit aab3125c43d8fecc7134e5f1e729fabf4dd196da: "em28xx: add
> > > support for registering multiple i2c buses".
> > > 
> > > The kernel messages for when it fails are:
> > ...
> > > Mar 21 09:26:57 telek kernel: [ 1396.542517] xc2028 12-0061: attaching existing instance
> > > Mar 21 09:26:57 telek kernel: [ 1396.542521] xc2028 12-0061: type set to XCeive xc2028/xc3028 tuner
> > > Mar 21 09:26:57 telek kernel: [ 1396.542523] em2882/3 #0: em2882/3 #0/2: xc3028 attached
> > ...
> > > Mar 21 09:26:57 telek kernel: [ 1396.547833] xc2028 12-0061: Error on line 1293: -19
> > 
> > Probably, the I2C speed is wrong. I noticed a small bug on this patch.
> > The following patch should fix it. Could you please test?
> 
> No luck, it didn't help.

On a first glance, I've no idea what else is different for devices with
just one I2C bus, like HVR-900.

Could you send me an USB sniff dump with the kernel that works and with
the broken kernel, with this patch applied?

The parsing tools are under v4l-utils contrib/ dir.

The first step is to check the usbmon interface for capture, with:

	$ ./parse_tcpdump_log.pl --list-devices
	usbmon4 ==> 020f (level 2)
	usbmon2 ==> USB2.0 Hub (level 1)
	usbmon1 ==> WinTV HVR-930C (level 7)

Then, you can run the parser to capture the data:

	# ./parse_tcpdump_log.pl --device usbmon1 | ./em28xx/parse_em28xx.pl 

Please load the em28xx driver only after starting the parser, as we want
to see what the driver is doing during the initialization.

Thanks!
Mauro