[linux-dvb] [PATCH] ir_rc5_timer_end decoder lockup fix - was - Problem with remote on ASUSTek P7131 Dual

hermann pitton hermann-pitton at arcor.de
Fri Jan 26 22:07:12 CET 2007


Hi!

Am Freitag, den 26.01.2007, 03:51 +0100 schrieb Vincent Penne:
> hermann pitton a écrit :
> > Hi Vincent,
> >
> > after having some time to look at it, it turns out my memory was
> > corrupted.
> >
> > You are definitely right.
> > A single return < 28000 will lock up the whole decoding needlessly.
> >
> > We should keep it, since I can produce conditions returning the other
> > dprintk(s), but not this one without changing the code to force it.
> >
> > Please provide a patch for your finding and sign-off.
> > Seems to move it below
> >
> > 	ir->active = 0;
> >
> > is sufficient.
> >
> >
> > Here the links how the code came in.
> >
> > Marks patch submission and followups with Nickolay.
> > http://marc.theaimsgroup.com/?l=linux-video&m=112841901208867&w=2
> >
> > Mark must have missed Nickolay's modifications. Last I find is here from
> > him then with comments.
> > http://marc.theaimsgroup.com/?l=linux-video&m=113735813926708&w=2
> >
> > Thanks,
> > Hermann
> >
> >   
> 
> Here is the patch as attachement. By the way, all my problems with the 
> remote are now fixed : I'm now using the stock kernel 2.6.19.2 (so now 
> ACPI works fine with my motherboard and I have usb working without the 
> "irqpoll" option). After two days testing with the machine doing various 
> things (compiling, playing games, crunching prime numbers, recording and 
> watching TV, receiving emails , etc ...) and a lot of use of the remote, 
> it's still working all fine.
> 
> This patch is very simple, it moves the resetting of ir->active in the 
> function ir_rc5_timer_end before
> the test gap < 28000 so that the remote continues working even after 
> this rare case arises.
> 
> Signed-off-by: Vincent Penne <ziggy at sashipa.com>
Signed-off-by: Hermann Pitton <hermann-pitton at arcor.de>


Vincent,

thanks a lot for doing stress tests on your critical hardware and for
discovering this bug.

Please, Mauro, Hartmut or Michael, can one of you pick it up ?
We don't want to see it on 2.6.21 again.

Thanks,
Hermann


> Cheers,
> 
> Vincent Penne
> 
> einfaches Textdokument-Anlage (my-lockup-fix.patch)
> diff -r cf2c23e0f950 linux/drivers/media/common/ir-functions.c
> --- a/linux/drivers/media/common/ir-functions.c	Thu Jan 25 06:00:01 2007 -0200
> +++ b/linux/drivers/media/common/ir-functions.c	Fri Jan 26 03:37:34 2007 +0100
> @@ -318,13 +318,15 @@ void ir_rc5_timer_end(unsigned long data
>  		    tv.tv_usec - ir->base_time.tv_usec;
>  	}
>  
> -	/* Allow some timmer jitter (RC5 is ~24ms anyway so this is ok) */
> +    /* signal we're ready to start a new code */
> +	ir->active = 0;
> +
> +	/* Allow some timer jitter (RC5 is ~24ms anyway so this is ok) */
>  	if (gap < 28000) {
>  		dprintk(1, "ir-common: spurious timer_end\n");
>  		return;
>  	}
>  
> -	ir->active = 0;
>  	if (ir->last_bit < 20) {
>  		/* ignore spurious codes (caused by light/other remotes) */
>  		dprintk(1, "ir-common: short code: %x\n", ir->code);
> --



-------------- next part --------------
A non-text attachment was scrubbed...
Name: ir_rc5_timer_end-decoder-lockup-fix.patch
Type: text/x-patch
Size: 736 bytes
Desc: not available
Url : http://www.linuxtv.org/pipermail/linux-dvb/attachments/20070126/f11c2c1b/ir_rc5_timer_end-decoder-lockup-fix.bin


More information about the linux-dvb mailing list