[pvrusb2] read attempt on video device leads to error messages

[Lars Schillingmann]

Am Freitag, 14. Oktober 2005 13:45 schrieb Mike Isely:
> I just found a key clue in the log you sent:
>
>    pvrusb2 i2c_attach [client=pcf8574 @ 0x21 id=0x3fe]
>
> That's the pvrusb2 driver reporting the attachment of an I2C client to
> itself at I2C address 0x21.  Only problem, that's the wrong client at that
> address.  Here's what we should see (looking at a log I have):
>
>    pvrusb2 i2c_attach [client=saa7115[50] @ 0x21 id=0xf1]
>    pvrusb2 attaching saa7115 I2C client
>
> The second line is the driver recognizing that it has recognized this
> particular client.  You don't see that second line because the pvrusb2
> driver didn't recognize this alien client.
>
> We need to understand why a pcf8574 driver (whatever the heck that is) has
> gotten in the way.  THIS is why we can't control the saa7115.  One thing
> you can try to do is identify which kernel configuration option controls
> the selecting of that I2C client and disable it.  Then let's see if things
> work better.  I have no idea at all why that I2C client is jumping in and
> spoiling the party.
>
> FYI, the way all this I2C stuff works is that the pvrusb2 driver
> implements an "I2C adapter" driver for its I2C bus inside the device, and
> then it plugs that into the I2C core.  When the new adapter driver appears
> in the I2C core, all client drivers will take notice and examine the new
> bus to see if there's a device there it should be controlling.  Under
> normal circumstances, msp3400.ko, tuner.ko and saa7115.ko will all see the
> bus, spot their chips and attach.  But in this case, something called
> pcf8574 is attaching in place of saa7115, occupying the bus address that
> saa7115 is supposed to get.
>
> I saw a problem similar to this back in March.  When you try to run the
> pvrusb2 driver on an x-box, there's an I2C client driver there which
> incorrectly sticks its nose into the pvrusb2's bus and screws things up.
> In that case, there's a patch that has to be applied to the xbox
> distribution to fix that problem.

I have compiled several i2c drivers in the kernel for reading out sensors and 
stuff. I will remove that stuff first. It's a shame i didn't get the idea on 
my own ;) Maybe this is the reason. If this does not work i will do the other 
things you mentioned. 
Taking of the cover and taking a photo will be no problem.
I have played around with the sysfs interface for the device some days ago and 
tried to set it to PAL. The results were the same without setting it.
If this works i owe you one. I will report back this evening local time.

I hope that surgery wasn't too hard.

 -Lars