Re: PID/MAC filtering

[Ralph Metzler <rjkm@netcologne.de>]

Thomas 'Dent' Mirlacher writes:
 > hi list,
 > 
 > ... i know this reply coms a little bit late, but i still have some
 > problems ... :(
 > 
 > > The lower 8 bits are the filter mask, i.e. it specifies which bits are
 > > checked, the higher 8 bits specify which value those bits should have. 
 > > So, in the example from CyberLogin.c:
 > > 
 > > struct bitfilter filt = {
 > >   0x200,
 > >   { 0x3eff, 0x00ff, 0x00ff, 0x0000, 0x0000, 0x0000, 0x00ff, 0x00ff,
 > >     0x00ff, 0x00ff, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000},
 > >   0x1c,0
 > > };
 > 
 > so if the hi-byte is the value and the lo-byte is the mask, there has
 > to be a problem somewhere (in my thoughts or in cyberlogin)
 > 
 > according to the standard the above would filter:
 > 
 > table_id
 > section_syntax_indicator
 > private_indicator
 > section_length
 > section_number
 > last_section_number
 > MACaddr[3]
 > MACaddr[2]

Sorry, I should have explained this earlier.
The second and third bytes (section_syntax_indicator ... section_length)
are not filtered.
I just left them out of the filter definition. So, the second byte 
is MACaddr[6] ...

One would have to cut them out anyway before giving the filter to the
firmware.
A general DVB filter API for Linux which also works for other cards will have to
be implemented later.

 > would appreciate any comments on this. another question is, if we can separate
 > pids from the bitmask - or can someone give me a good reason to have one
 > bitmask per pid - and not multiple pids and multiple bitmasks (applying to
 > the sum of filtered pids)

The hardware (TMX320AV7110) has a table of 32 PIDs and each of the
max. 32 filters a pointer to the table entry.
So, while more than one filter can point to the same PID, several PIDs
cannot point to the same filter. Itīs a hardware limitation.
Also consider that the max. hardware filter length is 12, software
(in firmware) filters go up to 16 bytes but take processing time from the 
ARM core.


Ralph