[linux-dvb] Re: Bug in software demux still in 2.6.10

[Manu Abraham <manu@kromtek.com>]

On Tue January 4 2005 5:10 pm, emard@softhome.net wrote:
> On Tue, Jan 04, 2005 at 01:43:08PM +0100, Valsecchi Patrick wrote:
> > Hi,
> >
> > Back in august 2004, I wrote an Email to this mailing list about a bug in
> > dvb/dvb-core/dvb_demux.c that sometimes cause the first section received
> > after a tunning to be bad. I didn't find that thread in the archive (?!?!
> > what happened? I've got censured? I've been rude without knowing it? If
> > yes, I'm sorry) so I've put my original email near the end of this one.
>
> I wrote that part of secttion code :), but I wasn't active on
> list at that time.
> My idea was to capture anything in order to fight certain packet
> loss from some boundary conditions and I didn't know I would break
> anything if I feed some initial garbage from the just tuned stream
>
> > I replied with a patch that he put in CVS. But then some people
> > complained that it was causing buffer overflows so it was removed. I
> > didn't have such problem with my setup (haupaugge NOVA-S and a home soft)
> > and I can understand that my patch may have some side effects unknown to
> > me. But the bug is here and it is still not patched.
>
> I don't think this has anything to do with buffer overflows?!
> those could be from av7110. Aren't they using different
> demux in firmware?
No it was not alone with the AV7110, other cards also had similar problems.. 
There were couple of people who complained, i also had the problem..
There were overflows on The Skystar as well as the Twinhan.. I had sent in 
some logs also ..

Not many people were quite comfortable with that area, and probably that was 
why nobody touched that area ..

I made a suggestion also that since you (emard) could certainly help out on 
that part ..

quoting inline old mail

-----------------------------------------------------------------------------
> 
> On Friday 27 August 2004 10:48, C.Y.M. wrote:
> > After updating cvs yesterday, I am getting a large amount 
> of these errors
> > in my syslog.  I am currently using the Nexus-S drivers 
> (dvb-ttpci and
> > stv0299).  Has anyone else noticed this?
> >
> > Aug 27 01:02:11 nofear vdr[10555]: ERROR: 2858 ring buffer overflows
> > (537205 bytes dropped)
> 
> Works fine here, with Nexus-S r2.1 and current CVS + MythTV, could be 
> something that only VDR triggers tho, anyone else seeing this?
> 
> Kenneth
> 

The only recent change that I can think of that may have caused this is the
change in the demux code from the other day.  But, if nobody else is seeing
these errors, I may have to wait and see if I can reproduce it.  The buffer
seems to get maxed out on the Nexus but I cant tell why yet. Also, it was
not something that happened right away (maybe after about 12-24 hours).

Thanks.



On Friday 27 Aug 2004 6:40 pm, C.Y.M. wrote:
> > On Friday 27 August 2004 10:48, C.Y.M. wrote:
> > > After updating cvs yesterday, I am getting a large amount
> >
> > of these errors
> >
> > > in my syslog.  I am currently using the Nexus-S drivers
> >
> > (dvb-ttpci and
> >
> > > stv0299).  Has anyone else noticed this?
> > >
> > > Aug 27 01:02:11 nofear vdr[10555]: ERROR: 2858 ring buffer overflows
> > > (537205 bytes dropped)
> >
> > Works fine here, with Nexus-S r2.1 and current CVS + MythTV, could be
> > something that only VDR triggers tho, anyone else seeing this?
> >
> > Kenneth
>
> The only recent change that I can think of that may have caused this is the
> change in the demux code from the other day.  But, if nobody else is seeing
> these errors, I may have to wait and see if I can reproduce it.  The buffer
> seems to get maxed out on the Nexus but I cant tell why yet. Also, it was
> not something that happened right away (maybe after about 12-24 hours).
>
> Thanks.


                Yes the overflow is due to the change in the demux code. For 
me it was 
instantaeneous. 30 ~ 40 seconds.

Regards,
Manu

On Friday 27 Aug 2004 7:31 pm, Valsecchi Patrick wrote:
> > -----Original Message-----
> > From: Manu Abraham
> > Hi,
> >             I tried the patch, since i also faced the same
> > problems. The result was
> > buffer overflows
>
> Hi,
>
> I get no such problem. Are you sure it's the only thing you patched? Maybe
> I fucked up... Can somebody who knows this part of the code double check?


Hi,
                I could not apply the newer CVS since i was using the 
dvb-bt8xx driver and 
the newer CVS needed a patch from bytesex.org.

                Because of the name "bytesex.org", my ISP had blocked it 
altogether.

I was getting CRC failures and that was the same reason as you had. I think 
Emard had earlier worked in that area ? Maybe he could help ?

Regards,
Manu


On Friday 27 August 2004 17:31, Valsecchi Patrick wrote:
> > -----Original Message-----
> > From: Manu Abraham
> > Hi,
> >   I tried the patch, since i also faced the same
> > problems. The result was
> > buffer overflows
>
> Hi,
>
> I get no such problem. Are you sure it's the only thing you patched? Maybe
> I fucked up... Can somebody who knows this part of the code double check?

Well, I'm not very familiar with this part, but it looks like you need to move 
sec->secbuf(p) to skip over the section that is incomplete. I've reversed the  
patch untill a better method is found.

Kenneth

-----------------------------------------------------------------------------
	End of quote

Manu

>
> > So can somebody, with a good knowledge of the dvb_demux.c innards, have a
> > look a it and make a patch? My patch proposal for kernel 2.6.10 is at the
> > end of the email, but it has greate chances of raising the same problems
> > than the one for 2.6.7.
>
> Should be no problem I think it should be applied to CVS.
> I can try this code myself but from first looking
> it should be good
>
> Emard