[pvrusb2] Looking for kernel oops evidence...

[Mike Isely]

On Sat, 10 Sep 2005, Mike Isely wrote:

> I just spent most of last night chasing this.  So far I've been unable to
> reproduce the problem.
>
> I in fact implemented something similar to what you described.  I added a
> signature word to the structure which is being checked now in front of a
> number of functions.  If the check fails, the code will print the file
> name and line number of the failure along with the contents of the buffer
> and then it should force an oops.  I've been running mplayer against this
> version of the driver now for the past 6 hours and it's still working
> perfectly.

I've been running an mplayer instance continuously now for 24 hours.  Not
even a hiccup.

I'm beginning to think that the driver is fine in this regard and
something is scribbling memory on this person's system.  I just wanted to
share that thought to alleviate any concerns here about a new driver
instability.  Aside from the memory leak below I think things right now
are still in pretty good shape (well, so long as you don't start tweaking
the encoder's parameters).

>
> BTW, while inspecting the code last night for this problem, I found a
> memory leak.  I'll have a new snapshot with that fixed before the weekend
> is over (no time this morning but hopefully tonight).  The leak happens
> when streaming is stopped - the control buffers are not being freed but
> their pointers are being tossed.  This is obviously a bad problem.

There will probably be a new driver snapshot tomorrow which fixes this
leak, adds the sanity testing code described above, and implements the
user & group overrides previously discussed.

  -Mike


-- 
                        |         Mike Isely          |     PGP fingerprint
     Spammers Die!!     |                             | 03 54 43 4D 75 E5 CC 92
                        |   isely @ pobox (dot) com   | 71 16 01 E2 B5 F5 C1 E8
                        |                             |