DVB V5 API: Event Model

DVB V5 API: Event Model

[Martin Rudge]

Hello All,

When using the V5 API (DVB-S/S2) for the DVB frontend device (with the
now merged SEC functionality), setting properties DTV_VOLTAGE and/or
DTV_TONE generates extra (unwanted?) events.  This is due to utilising
the legacy FE_SET_FRONTEND IOCTL in their respective implementations.

Depending on their placement in one "atomic" FE_SET_PROPERTY call,
they can cause an "incorrect" (premature) SYNC/LOCK event to be
generated.  For example, when looping issuing tune requests in
succession during a scan operation. This was with a fairly recent
media build (pulled Saturday).

Conversly using DTV_CLEAR clears the cached values, but doesn't affect
the frontend (LNB).  This is probably desirable behaviour.

Any thoughts, working as designed/intended?

Thanks
Martin

Re: DVB V5 API: Event Model

[Mauro Carvalho Chehab]

Em 12-11-2012 11:04, Martin Rudge escreveu:
> Hello All,
>
> When using the V5 API (DVB-S/S2) for the DVB frontend device (with the
> now merged SEC functionality), setting properties DTV_VOLTAGE and/or
> DTV_TONE generates extra (unwanted?) events.  This is due to utilising
> the legacy FE_SET_FRONTEND IOCTL in their respective implementations.
>
> Depending on their placement in one "atomic" FE_SET_PROPERTY call,
> they can cause an "incorrect" (premature) SYNC/LOCK event to be
> generated.  For example, when looping issuing tune requests in
> succession during a scan operation. This was with a fairly recent
> media build (pulled Saturday).

I suspect that this is an undesirable behavior, likely there since the initial
version of DVB-S2 API. It could be too late to fix, as userspace apps may be
trusting on this behavior.

Maybe you could propose a patch and ask app developers what they thing about
it.

> Conversly using DTV_CLEAR clears the cached values, but doesn't affect
> the frontend (LNB).  This is probably desirable behaviour.

This is desirable.

> Any thoughts, working as designed/intended?

Regards,
Mauro

Re: DVB V5 API: Event Model

[Martin Rudge]

Thanks Mauro.

I'm not familiar with the code, however I'll review the design/code and see what could be done.

A quick look suggests the events are originating from the tuning loop, so it's down to the timing of copying the parms for voltage/tone (not currently instigated by a DTV_TUNE).

Thanks
Martin

On 13 Nov 2012, at 09:24, Mauro Carvalho Chehab <mchehab@redhat.com> wrote:

> Em 12-11-2012 11:04, Martin Rudge escreveu:
>> Hello All,
>> 
>> When using the V5 API (DVB-S/S2) for the DVB frontend device (with the
>> now merged SEC functionality), setting properties DTV_VOLTAGE and/or
>> DTV_TONE generates extra (unwanted?) events.  This is due to utilising
>> the legacy FE_SET_FRONTEND IOCTL in their respective implementations.
>> 
>> Depending on their placement in one "atomic" FE_SET_PROPERTY call,
>> they can cause an "incorrect" (premature) SYNC/LOCK event to be
>> generated.  For example, when looping issuing tune requests in
>> succession during a scan operation. This was with a fairly recent
>> media build (pulled Saturday).
> 
> I suspect that this is an undesirable behavior, likely there since the initial
> version of DVB-S2 API. It could be too late to fix, as userspace apps may be
> trusting on this behavior.
> 
> Maybe you could propose a patch and ask app developers what they thing about
> it.
> 
>> Conversly using DTV_CLEAR clears the cached values, but doesn't affect
>> the frontend (LNB).  This is probably desirable behaviour.
> 
> This is desirable.
> 
>> Any thoughts, working as designed/intended?
> 
> Regards,
> Mauro