[media] dvb-frontends/mxl5xx: fix lock check order

Em Sun, 20 Aug 2017 12:45:45 +0200
Daniel Scheller <d.scheller.oss@gmail.com> escreveu:

> From: Daniel Scheller <d.scheller@gmx.net>
> 

Always add a description at the patch.


> Signed-off-by: Daniel Scheller <d.scheller@gmx.net>
> ---
> When the mxl5xx driver together with the ddbridge glue gets merged ([1]),
> this one should go in aswell - this fix is part of the dddvb-0.9.31
> release.
> 
>  drivers/media/dvb-frontends/mxl5xx.c | 5 +++--
>  1 file changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/media/dvb-frontends/mxl5xx.c b/drivers/media/dvb-frontends/mxl5xx.c
> index 676c96c216c3..d9136d67f5d4 100644
> --- a/drivers/media/dvb-frontends/mxl5xx.c
> +++ b/drivers/media/dvb-frontends/mxl5xx.c
> @@ -638,13 +638,14 @@ static int tune(struct dvb_frontend *fe, bool re_tune,
>  		state->tune_time = jiffies;
>  		return 0;
>  	}
> -	if (*status & FE_HAS_LOCK)
> -		return 0;
>  
>  	r = read_status(fe, status);
>  	if (r)
>  		return r;
>  
> +	if (*status & FE_HAS_LOCK)
> +		return 0;
> +
>  	return 0;

That's stupid! it will return 0 on all situations, no matter if FE_HAS_LOCK
or not. So, no need for the if.

Anyway, IMHO, either the original code is right and it needs to
use a previously cached value (with sounds weird to me, although
it is possible), or the logic is utterly broken, and we should,
instead, apply the enclosed patch.

>  	if (r)
>  		return r;

>  }
>  

Thanks,
Mauro

[PATCH RFC] media: mxl5xx: fix tuning logic

The tuning logic is broken with regards to status report:
it relies on a previously-cached value that may not be valid
if retuned.

Change the logic to always read the status.

Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
  

Am Sun, 27 Aug 2017 09:18:07 -0300
schrieb Mauro Carvalho Chehab <mchehab@s-opensource.com>:

Thanks for looking at this.

> Em Sun, 20 Aug 2017 12:45:45 +0200
> Daniel Scheller <d.scheller.oss@gmail.com> escreveu:
> 
> > From: Daniel Scheller <d.scheller@gmx.net>
> >   
> 
> Always add a description at the patch.

Sorry. Will remember to do so even for these rather self-explanatory
changes.

> > Signed-off-by: Daniel Scheller <d.scheller@gmx.net>
> > ---
> > When the mxl5xx driver together with the ddbridge glue gets merged
> > ([1]), this one should go in aswell - this fix is part of the
> > dddvb-0.9.31 release.
> > 
> >  drivers/media/dvb-frontends/mxl5xx.c | 5 +++--
> >  1 file changed, 3 insertions(+), 2 deletions(-)
> > 
> > diff --git a/drivers/media/dvb-frontends/mxl5xx.c
> > b/drivers/media/dvb-frontends/mxl5xx.c index
> > 676c96c216c3..d9136d67f5d4 100644 ---
> > a/drivers/media/dvb-frontends/mxl5xx.c +++
> > b/drivers/media/dvb-frontends/mxl5xx.c @@ -638,13 +638,14 @@ static
> > int tune(struct dvb_frontend *fe, bool re_tune, state->tune_time =
> > jiffies; return 0;
> >  	}
> > -	if (*status & FE_HAS_LOCK)
> > -		return 0;
> >  
> >  	r = read_status(fe, status);
> >  	if (r)
> >  		return r;
> >  
> > +	if (*status & FE_HAS_LOCK)
> > +		return 0;
> > +
> >  	return 0;  
> 
> That's stupid! it will return 0 on all situations, no matter if
> FE_HAS_LOCK or not. So, no need for the if.
> 
> Anyway, IMHO, either the original code is right and it needs to
> use a previously cached value (with sounds weird to me, although
> it is possible), or the logic is utterly broken, and we should,
> instead, apply the enclosed patch.

Well, in fact, this change was straight picked from upstream ([1]), but
I honestly didn't take into account that during cleanup the
#if-0/#endif was removed (which will in turn result in the same
behaviour though - return 0 in all cases; some leftovers?).

[1]
https://github.com/DigitalDevices/dddvb/commit/a44dbc889b7030a100599d2bd2c93c976c011a03

> 
> >  	if (r)
> >  		return r;  
> 
> >  }
> >    
> 
> Thanks,
> Mauro
> 
> [PATCH RFC] media: mxl5xx: fix tuning logic
> 
> The tuning logic is broken with regards to status report:
> it relies on a previously-cached value that may not be valid
> if retuned.
> 
> Change the logic to always read the status.
> 
> Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
> 
> 
> diff --git a/drivers/media/dvb-frontends/mxl5xx.c
> b/drivers/media/dvb-frontends/mxl5xx.c index
> 676c96c216c3..4b449a6943c5 100644 ---
> a/drivers/media/dvb-frontends/mxl5xx.c +++
> b/drivers/media/dvb-frontends/mxl5xx.c @@ -636,16 +636,9 @@ static
> int tune(struct dvb_frontend *fe, bool re_tune, if (r)
>  			return r;
>  		state->tune_time = jiffies;
> -		return 0;
>  	}
> -	if (*status & FE_HAS_LOCK)
> -		return 0;
>  
> -	r = read_status(fe, status);
> -	if (r)
> -		return r;
> -
> -	return 0;
> +	return read_status(fe, status);
>  }
>  
>  static enum fe_code_rate conv_fec(enum MXL_HYDRA_FEC_E fec)
> 

Indeed that makes more sense in this context (compared to dddvb
upstream), albeit on re_tune=1 read_status() will now also be called,
which I guess is totally fine. Let's do it this way. So:

Acked-by: Daniel Scheller <d.scheller@gmx.net>

Thanks & best regards,
Daniel Scheller