Fix for the data reception problem and more

["Ralph Metzler <" <rjkm@netcologne.de>]

Michael Holzt writes:
 > Ok, i took some time to dig through the sources of driver 0.7 and found
 > the bug which occured now. This bug is existant in old drivers too(!),
 > but wasn't catched there (instead some internal data structures where
 > used using wrong pointers, ouch!)
 > 
 > The calculation of the bit filter in use in function debiirq, file
 > dvb.c is somewhat wrong. I get filters 128, 130 ... where the maximum
 > allowed value is 31 (the card can only handle 32 bitfilters). The new
 > driver catches this bug in line 575
 >                 if ((handle>=MAXFILT) || !dvb->hfilter[handle])
 >                         break;
 > where the old driver just accessed internal structures using this wrong
 > index
 >                 if (dvb->filter_flags[handle]&FILTER_MEM)
 > [...]
 > 
 > 
 > 
 > As the bitfilter can go up to 31 maximum as indicated above, i changed
 > line 516 in dvb.c to read
 >         int handle=(type>>8) & 0x1F;
 > instead of
 >         int handle=(type>>8);
 > 
 > 
 > This fixes the bug for me, although i'm not absolutely sure if this
 > really fixes the source of the problems or probably only the place
 > where it occures.
 > 
 > Our friends at convergence need to check this.


Yes, sorry for that.
The fix is correct. If the SECTION_EIT flag is set in the mode field
of the filter, bit 7 of the returned handle is set to indicate this.
dvbd does not set the SECTION_EIT  flag but the firmware implicitely
sets it for the SECTION_IPMPE mode.
Thatīs because SECTION_EIT is not just for EIT data but it just 
allocates more buffer space for a filter.


And, while we are at it, as David Toth pointed out to me,
there is another bug in SetFront:

In the new driver (as suggested by Klaus Schmidinger) I also
check if the sync was lost to determine if action for a channel change
is to be taken. But I forgot to update the frontend structure with
new values before the check.
It should work like this:

SetFront(struct dvb_struct *dvb, struct frontend *front)
{
        dprintk("setfront\n");
        decoder_command(dvb, DVB_GET_FRONTEND, &dvb->front);  /* FIX */
        if (dvb->dvbtype==DVB_S) {
...

This was not tragic if you always did a getfrontend call in the
application but could result in filters being shut down when calling 
a setfrontend twice in a row with the same values but with the 
sync entry cleared. 



Iīll put a new release on the web page later today.


Ralph