2.6.32 regression: can't tune DVB with firedtv

2.6.32 regression: can't tune DVB with firedtv

[Stefan Richter]

Hi list,

I just switched from kernel 2.6.31 to 2.6.32-rc5.  Using kaffeine, I
can't tune FireDTV-C and FireDTV-T boxes via the firedtv driver anymore.
Electronic program guide data is still displayed though.

Under 2.6.31, I used firedtv at the same patchlevel as present in
2.6.32-rc5, hence I guess that it is a DVB core problem rather than a
driver problem.

Any suggestions where to look for the cause?

(I am not subscribed to the list.)
-- 
Stefan Richter
-=====-==--= =-=- =---=
http://arcgraph.de/sr/

Re: 2.6.32 regression: can't tune DVB with firedtv

[Stefan Richter]

Stefan Richter wrote:
> I just switched from kernel 2.6.31 to 2.6.32-rc5.  Using kaffeine, I
> can't tune FireDTV-C and FireDTV-T boxes via the firedtv driver anymore.
> Electronic program guide data is still displayed though.
> 
> Under 2.6.31, I used firedtv at the same patchlevel as present in
> 2.6.32-rc5, hence I guess that it is a DVB core problem rather than a
> driver problem.
> 
> Any suggestions where to look for the cause?

>From looking at the git history of dvb-core, I think the cause is
"V4L/DVB (12685): dvb-core: check fe->ops.set_frontend return value",
but the actual bug is in the driver in
firedtv-fe.c::fdtv_set_frontend().  Let me boot back into 2.6.32-rc5 and
check my theory...

> (I am not subscribed to the list.)
-- 
Stefan Richter
-=====-==--= =-=- =---=
http://arcgraph.de/sr/

[PATCH] firedtv: fix regression: tuning fails due to bogus error return

[Stefan Richter]

Since 2.6.32(-rc1), DVB core checks the return value of
dvb_frontend_ops.set_frontend.  Now it becomes apparent that firedtv
always returned a bogus value from its set_frontend method.

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
---
 drivers/media/dvb/firewire/firedtv-avc.c |    7 +++++--
 drivers/media/dvb/firewire/firedtv-fe.c  |    8 +-------
 2 files changed, 6 insertions(+), 9 deletions(-)

Index: linux-2.6.32-rc5/drivers/media/dvb/firewire/firedtv-avc.c
===================================================================
--- linux-2.6.32-rc5.orig/drivers/media/dvb/firewire/firedtv-avc.c
+++ linux-2.6.32-rc5/drivers/media/dvb/firewire/firedtv-avc.c
@@ -573,8 +573,11 @@ int avc_tuner_dsd(struct firedtv *fdtv,
 
 	msleep(500);
 #if 0
-	/* FIXME: */
-	/* u8 *status was an out-parameter of avc_tuner_dsd, unused by caller */
+	/*
+	 * FIXME:
+	 * u8 *status was an out-parameter of avc_tuner_dsd, unused by caller
+	 * Check for AVC_RESPONSE_ACCEPTED here instead?
+	 */
 	if (status)
 		*status = r->operand[2];
 #endif
Index: linux-2.6.32-rc5/drivers/media/dvb/firewire/firedtv-fe.c
===================================================================
--- linux-2.6.32-rc5.orig/drivers/media/dvb/firewire/firedtv-fe.c
+++ linux-2.6.32-rc5/drivers/media/dvb/firewire/firedtv-fe.c
@@ -141,18 +141,12 @@ static int fdtv_read_uncorrected_blocks(
 	return -EOPNOTSUPP;
 }
 
-#define ACCEPTED 0x9
-
 static int fdtv_set_frontend(struct dvb_frontend *fe,
 			     struct dvb_frontend_parameters *params)
 {
 	struct firedtv *fdtv = fe->sec_priv;
 
-	/* FIXME: avc_tuner_dsd never returns ACCEPTED. Check status? */
-	if (avc_tuner_dsd(fdtv, params) != ACCEPTED)
-		return -EINVAL;
-	else
-		return 0; /* not sure of this... */
+	return avc_tuner_dsd(fdtv, params);
 }
 
 static int fdtv_get_frontend(struct dvb_frontend *fe,

-- 
Stefan Richter
-=====-==--= =-=- =---=
http://arcgraph.de/sr/

Re: [PATCH] firedtv: fix regression: tuning fails due to bogus error return

[Stefan Richter]

Stefan Richter wrote on 2009-10-17:
> Since 2.6.32(-rc1), DVB core checks the return value of
> dvb_frontend_ops.set_frontend.  Now it becomes apparent that firedtv
> always returned a bogus value from its set_frontend method.
> 
> Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>

Hi Mauro,

when you committed this, you added "CC: stable@kernel.org".  This is not
necessary, 2.6.32-rc1 is the first kernel version with the regression
(see changelog).

Well, OK, the patch would become necessary in -stable if somebody was to
put the new set_frontend return code check in dvb core into -stable. But
I hope nobody is going to do that; it'd be a bad idea.

Furthermore...

> ---
>  drivers/media/dvb/firewire/firedtv-avc.c |    7 +++++--
>  drivers/media/dvb/firewire/firedtv-fe.c  |    8 +-------
>  2 files changed, 6 insertions(+), 9 deletions(-)
> 
> Index: linux-2.6.32-rc5/drivers/media/dvb/firewire/firedtv-avc.c
> ===================================================================
> --- linux-2.6.32-rc5.orig/drivers/media/dvb/firewire/firedtv-avc.c
> +++ linux-2.6.32-rc5/drivers/media/dvb/firewire/firedtv-avc.c
> @@ -573,8 +573,11 @@ int avc_tuner_dsd(struct firedtv *fdtv,
>  
>  	msleep(500);
>  #if 0
> -	/* FIXME: */
> -	/* u8 *status was an out-parameter of avc_tuner_dsd, unused by caller */
> +	/*
> +	 * FIXME:
> +	 * u8 *status was an out-parameter of avc_tuner_dsd, unused by caller
> +	 * Check for AVC_RESPONSE_ACCEPTED here instead?
> +	 */
>  	if (status)
>  		*status = r->operand[2];
>  #endif

...the firedtv-avc.c part of the patch vanished when you committed it. I
guess this was accident, not deliberate --- otherwise a note in the
changelog below of my Signed-off-by would have been appreciated.

I will resubmit this missing comment change or a variation of it or even
a fix of the FIXME sometime in the future.

> Index: linux-2.6.32-rc5/drivers/media/dvb/firewire/firedtv-fe.c
> ===================================================================
> --- linux-2.6.32-rc5.orig/drivers/media/dvb/firewire/firedtv-fe.c
> +++ linux-2.6.32-rc5/drivers/media/dvb/firewire/firedtv-fe.c
> @@ -141,18 +141,12 @@ static int fdtv_read_uncorrected_blocks(
>  	return -EOPNOTSUPP;
>  }
>  
> -#define ACCEPTED 0x9
> -
>  static int fdtv_set_frontend(struct dvb_frontend *fe,
>  			     struct dvb_frontend_parameters *params)
>  {
>  	struct firedtv *fdtv = fe->sec_priv;
>  
> -	/* FIXME: avc_tuner_dsd never returns ACCEPTED. Check status? */
> -	if (avc_tuner_dsd(fdtv, params) != ACCEPTED)
> -		return -EINVAL;
> -	else
> -		return 0; /* not sure of this... */
> +	return avc_tuner_dsd(fdtv, params);
>  }
>  
>  static int fdtv_get_frontend(struct dvb_frontend *fe,
> 

-- 
Stefan Richter
-=====-==--= =-== -==-=
http://arcgraph.de/sr/

Re: [PATCH] firedtv: fix regression: tuning fails due to bogus error return

[Stefan Richter]

Stefan Richter wrote:
> ...the firedtv-avc.c part of the patch vanished when you committed it.

It was still there in the linuxtv-commits message of the respective
Mercurial commit, BTW.  Something wrong with the hg to git workflow?
-- 
Stefan Richter
-=====-==--= =-== -==-=
http://arcgraph.de/sr/

Re: 2.6.32 regression: can't tune DVB with firedtv

[Mario Bachmann]

Am Sat, 17 Oct 2009 21:01:50 +0200
schrieb Stefan Richter <stefanr@s5r6.in-berlin.de>:

> Hi list,
> 
> I just switched from kernel 2.6.31 to 2.6.32-rc5.  Using kaffeine, I
> can't tune FireDTV-C and FireDTV-T boxes via the firedtv driver
> anymore. Electronic program guide data is still displayed though.
> 
> Under 2.6.31, I used firedtv at the same patchlevel as present in
> 2.6.32-rc5, hence I guess that it is a DVB core problem rather than a
> driver problem.
> 
> Any suggestions where to look for the cause?
> 
> (I am not subscribed to the list.)

Hi there, 

perhaps it is related:
with kernel 2.6.30.8 my "TwinhanDTV USB-Ter USB1.1 / Magic Box I"
worked. 

Now with kernel 2.6.32-rc3 (and 2.6.31.1) the modules seems to be
loaded fine, but tzap/kaffeine/mplayer can not tune to a channel. 

i think the cause must be here:
/usr/src/linux-2.6.32-rc3/drivers/media/dvb/dvb-usb/dibusb-common.c
line 136 to line 146

i changed this hole section to the version of 2.6.30.8 and it works
again.

Mario