Re: I2C regression in post 2.6.14 gits.. -- get_user_pages() is real bug..

[Matti Aarnio]

On Fri, Nov 04, 2005 at 10:29:34AM +0100, Jean Delvare wrote:
> Hi Matti,
> 
> On 2005-11-04, Matti Aarnio wrote:
> > Specifically,  BTTV driver fails to control internal
> > I2C devices in Hauppauge WinTV card with Bt848 onboard.
> 
> Can you detail how you came up with this diagnosis?
> 
> > Baseline 2.6.14 works fine,  but 2.6.14-git2 fails.
> 
> Can you please try 2.6.14-git1 and 2.6.14-rc5-mm1? This should help us
> norrow the faulty patch. Also, please confirm that 2.6.14-git7 still has
> the problem.
> 
> > Any ideas ?  Possible fixes ?
> 
> Enable as many debugging as possible, in particular I2C Core debugging
> (I2C_DEBUG_CORE=y) in a working kernel and a non-working kernel, and
> send both logs for comparison.
> 
> Thanks a lot for testing, BTW.


Since then I have been tracking of what really fails in the bttv driver,
and running:   strace -o tvtime.log tvtime
Result is most revealing in failures.

The fault is not in I2C, but appears in   video_buf  module failing
calls of   get_user_pages()   with  -EFAULT.

There has been lots of discussion about it back in August, but apparently
only now in post 2.6.14 the changes made it into   mm/memory.c  which
faulted this thing for BTTV. 


Oh yes,  I do use Fedora Core development kernels, and don't usually
(anymore) compile things myself, but now I had a weekend off...


> --
> Jean Delvare

/Matti Aarnio
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/