ts2020: call get_rf_strength from frontend

PATCH In review Malcolm Priestley 2 participants 5 messages Jan 6, 2013

Subsystems (2)

Media (V4L/DVB) M88RS2000 media driver

lore

Share & tools

lore

Malcolm Priestley

author maintainer

2013-01-06 12:40 UTC #0 lore raw

PATCH 1/1 ts2020: call get_rf_strength from frontend

Restore ds3000.c read_signal_strength.

Call tuner get_rf_strength from frontend read_signal_strength.

We are able to do a NULL check and doesn't limit the tuner attach to the frontend attach area.

At the moment the lmedm04 tuner attach is stuck in frontend attach area.

Signed-off-by Malcolm Priestley <[email protected]>

---
 drivers/media/dvb-frontends/ds3000.c    | 10 ++++++++++
 drivers/media/dvb-frontends/m88rs2000.c |  4 +++-
 drivers/media/dvb-frontends/ts2020.c    |  1 -
 3 files changed, 13 insertions(+), 2 deletions(-)

drivers/media/dvb-frontends/ds3000.c +10 -0

index d128f85..1e344b0 100644
@@ -533,6 +533,15 @@ static int ds3000_read_ber(struct dvb_frontend *fe, u32* ber)
 	return 0;
 }
 
+static int ds3000_read_signal_strength(struct dvb_frontend *fe,
+						u16 *signal_strength)
+{
+	if (fe->ops.tuner_ops.get_rf_strength)
+		fe->ops.tuner_ops.get_rf_strength(fe, signal_strength);
+
+	return 0;
+}
+
 /* calculate DS3000 snr value in dB */
 static int ds3000_read_snr(struct dvb_frontend *fe, u16 *snr)
 {

@@ -1102,6 +1111,7 @@ static struct dvb_frontend_ops ds3000_ops = {
 	.i2c_gate_ctrl = ds3000_i2c_gate_ctrl,
 	.read_status = ds3000_read_status,
 	.read_ber = ds3000_read_ber,
+	.read_signal_strength = ds3000_read_signal_strength,
 	.read_snr = ds3000_read_snr,
 	.read_ucblocks = ds3000_read_ucblocks,
 	.set_voltage = ds3000_set_voltage,

drivers/media/dvb-frontends/m88rs2000.c +3 -1

index 283c90f..4da5272 100644
@@ -446,7 +446,9 @@ static int m88rs2000_read_ber(struct dvb_frontend *fe, u32 *ber)
 static int m88rs2000_read_signal_strength(struct dvb_frontend *fe,
 	u16 *strength)
 {
-	*strength = 0;
+	if (fe->ops.tuner_ops.get_rf_strength)
+		fe->ops.tuner_ops.get_rf_strength(fe, strength);
+
 	return 0;
 }

drivers/media/dvb-frontends/ts2020.c +0 -1

index f50e237..ad7ad85 100644
@@ -363,7 +363,6 @@ struct dvb_frontend *ts2020_attach(struct dvb_frontend *fe,
 
 	memcpy(&fe->ops.tuner_ops, &ts2020_tuner_ops,
 				sizeof(struct dvb_tuner_ops));
-	fe->ops.read_signal_strength = fe->ops.tuner_ops.get_rf_strength;
 
 	return fe;
 }

Signature · 1 line

1.8.0

Antti Palosaari

2013-01-06 13:37 UTC #1 lore raw

Quoted 9 lines · Malcolm Priestley

Restore ds3000.c read_signal_strength.
Call tuner get_rf_strength from frontend read_signal_strength.
We are able to do a NULL check and doesn't limit the tuner attach to the frontend attach area.
At the moment the lmedm04 tuner attach is stuck in frontend attach area.

I would like to nack that, as I see some problems:
1) it changes deviation against normal procedures
2) interface driver (usb/pci) should have full control to make decision
3) you shoot to our own leg easily in power management

* actually bug 3) already happened some drivers, like rtl28xxu. Tuner is behind demod and demod is put sleep => no access to tuner. FE callback is overridden (just like you are trying to do as default) which means user-space could still make queries => I/O errors.

Antti

Quoted 64 lines

Signed-off-by Malcolm Priestley <[email protected]>

---
  drivers/media/dvb-frontends/ds3000.c    | 10 ++++++++++
  drivers/media/dvb-frontends/m88rs2000.c |  4 +++-
  drivers/media/dvb-frontends/ts2020.c    |  1 -
  3 files changed, 13 insertions(+), 2 deletions(-)

drivers/media/dvb-frontends/ds3000.c +10 -0

index d128f85..1e344b0 100644
@@ -533,6 +533,15 @@ static int ds3000_read_ber(struct dvb_frontend *fe, u32* ber)
  	return 0;
  }
 
+static int ds3000_read_signal_strength(struct dvb_frontend *fe,
+						u16 *signal_strength)
+{
+	if (fe->ops.tuner_ops.get_rf_strength)
+		fe->ops.tuner_ops.get_rf_strength(fe, signal_strength);
+
+	return 0;
+}
+
  /* calculate DS3000 snr value in dB */
  static int ds3000_read_snr(struct dvb_frontend *fe, u16 *snr)
  {

@@ -1102,6 +1111,7 @@ static struct dvb_frontend_ops ds3000_ops = {
  	.i2c_gate_ctrl = ds3000_i2c_gate_ctrl,
  	.read_status = ds3000_read_status,
  	.read_ber = ds3000_read_ber,
+	.read_signal_strength = ds3000_read_signal_strength,
  	.read_snr = ds3000_read_snr,
  	.read_ucblocks = ds3000_read_ucblocks,
  	.set_voltage = ds3000_set_voltage,

drivers/media/dvb-frontends/m88rs2000.c +3 -1

index 283c90f..4da5272 100644
@@ -446,7 +446,9 @@ static int m88rs2000_read_ber(struct dvb_frontend *fe, u32 *ber)
  static int m88rs2000_read_signal_strength(struct dvb_frontend *fe,
  	u16 *strength)
  {
-	*strength = 0;
+	if (fe->ops.tuner_ops.get_rf_strength)
+		fe->ops.tuner_ops.get_rf_strength(fe, strength);
+
  	return 0;
  }

drivers/media/dvb-frontends/ts2020.c +0 -1

index f50e237..ad7ad85 100644
@@ -363,7 +363,6 @@ struct dvb_frontend *ts2020_attach(struct dvb_frontend *fe,
 
  	memcpy(&fe->ops.tuner_ops, &ts2020_tuner_ops,
  				sizeof(struct dvb_tuner_ops));
-	fe->ops.read_signal_strength = fe->ops.tuner_ops.get_rf_strength;
 
  	return fe;
  }

Signature · 1 line

http://palosaari.fi/

Malcolm Priestley

author maintainer

2013-01-06 18:14 UTC #2 lore raw

Quoted 8 lines · Antti Palosaari

Quoted 7 lines · Malcolm Priestley
Restore ds3000.c read_signal_strength.
Call tuner get_rf_strength from frontend read_signal_strength.
We are able to do a NULL check and doesn't limit the tuner attach to the frontend attach area.

Quoted 8 lines

Quoted 2 lines
At the moment the lmedm04 tuner attach is stuck in frontend attach area.
I would like to nack that, as I see some problems:
1) it changes deviation against normal procedures
2) interface driver (usb/pci) should have full control to make decision
3) you shoot to our own leg easily in power management

This patch does not do any operational changes, and is a proper way to call to another module with a run time NULL check. The same way as another tuner function from demodulator is called.

Quoted 4 lines

* actually bug 3) already happened some drivers, like rtl28xxu. Tuner is behind demod and demod is put sleep => no access to tuner. FE callback is overridden (just like you are trying to do as default) which means user-space could still make queries => I/O errors.

In such cases, the tuner init/sleep should also be called.

Regards

Malcolm

Quoted 69 lines

Antti

Quoted 64 lines

Signed-off-by Malcolm Priestley <[email protected]>

---
  drivers/media/dvb-frontends/ds3000.c    | 10 ++++++++++
  drivers/media/dvb-frontends/m88rs2000.c |  4 +++-
  drivers/media/dvb-frontends/ts2020.c    |  1 -
  3 files changed, 13 insertions(+), 2 deletions(-)

drivers/media/dvb-frontends/ds3000.c +10 -0

index d128f85..1e344b0 100644
@@ -533,6 +533,15 @@ static int ds3000_read_ber(struct dvb_frontend *fe, u32* ber)
  	return 0;
  }
 
+static int ds3000_read_signal_strength(struct dvb_frontend *fe,
+						u16 *signal_strength)
+{
+	if (fe->ops.tuner_ops.get_rf_strength)
+		fe->ops.tuner_ops.get_rf_strength(fe, signal_strength);
+
+	return 0;
+}
+
  /* calculate DS3000 snr value in dB */
  static int ds3000_read_snr(struct dvb_frontend *fe, u16 *snr)
  {

@@ -1102,6 +1111,7 @@ static struct dvb_frontend_ops ds3000_ops = {
  	.i2c_gate_ctrl = ds3000_i2c_gate_ctrl,
  	.read_status = ds3000_read_status,
  	.read_ber = ds3000_read_ber,
+	.read_signal_strength = ds3000_read_signal_strength,
  	.read_snr = ds3000_read_snr,
  	.read_ucblocks = ds3000_read_ucblocks,
  	.set_voltage = ds3000_set_voltage,

drivers/media/dvb-frontends/m88rs2000.c +3 -1

index 283c90f..4da5272 100644
@@ -446,7 +446,9 @@ static int m88rs2000_read_ber(struct dvb_frontend *fe, u32 *ber)
  static int m88rs2000_read_signal_strength(struct dvb_frontend *fe,
  	u16 *strength)
  {
-	*strength = 0;
+	if (fe->ops.tuner_ops.get_rf_strength)
+		fe->ops.tuner_ops.get_rf_strength(fe, strength);
+
  	return 0;
  }

drivers/media/dvb-frontends/ts2020.c +0 -1

index f50e237..ad7ad85 100644
@@ -363,7 +363,6 @@ struct dvb_frontend *ts2020_attach(struct dvb_frontend *fe,
 
  	memcpy(&fe->ops.tuner_ops, &ts2020_tuner_ops,
  				sizeof(struct dvb_tuner_ops));
-	fe->ops.read_signal_strength = fe->ops.tuner_ops.get_rf_strength;
 
  	return fe;
  }

Antti Palosaari

2013-01-06 18:46 UTC #3 lore raw

Quoted 12 lines · Malcolm Priestley

Quoted 11 lines · Antti Palosaari
Quoted 9 lines · Malcolm Priestley
Restore ds3000.c read_signal_strength.
Call tuner get_rf_strength from frontend read_signal_strength.
We are able to do a NULL check and doesn't limit the tuner attach to the frontend attach area.
At the moment the lmedm04 tuner attach is stuck in frontend attach area.

Quoted 8 lines

Quoted 5 lines
I would like to nack that, as I see some problems:
1) it changes deviation against normal procedures
2) interface driver (usb/pci) should have full control to make decision
3) you shoot to our own leg easily in power management
This patch does not do any operational changes, and is a proper way to call to another module with a run time NULL check. The same way as another tuner function from demodulator is called.

uh, certainly I understand it does not change functionality in that case! I tried to point out it is logically insane and error proof. Ee should make this kind of direct calls between drivers as less as possible - just to make life easier in future. It could work for you, but for some other it could cause problems as hardware is designed differently.

Quoted 6 lines

Quoted 4 lines
* actually bug 3) already happened some drivers, like rtl28xxu. Tuner is behind demod and demod is put sleep => no access to tuner. FE callback is overridden (just like you are trying to do as default) which means user-space could still make queries => I/O errors.
In such cases, the tuner init/sleep should also be called.

???????
I think you don't understand functionality at all!

See that bug report, I was referring
http://permalink.gmane.org/gmane.linux.drivers.video-input-infrastructure/48216

It is not long time I added checks to DVB-core to avoid it calling statistics IOCTLs if frontend is not active.

But simply, it is situation your demod driver should not make decision whether statistics are asked from the tuner or not. It is decision that should be left to interface driver. It is single line of code to override that logic in interface driver - or add even some other logic how to perform measurement and in which condition. Interface driver is driver which knows best power-management etc. related stuff.

Antti

Quoted 79 lines

Regards

Malcolm

Quoted 69 lines

Antti

Quoted 64 lines

Signed-off-by Malcolm Priestley <[email protected]>

---
   drivers/media/dvb-frontends/ds3000.c    | 10 ++++++++++
   drivers/media/dvb-frontends/m88rs2000.c |  4 +++-
   drivers/media/dvb-frontends/ts2020.c    |  1 -
   3 files changed, 13 insertions(+), 2 deletions(-)

drivers/media/dvb-frontends/ds3000.c +10 -0

index d128f85..1e344b0 100644
@@ -533,6 +533,15 @@ static int ds3000_read_ber(struct dvb_frontend *fe, u32* ber)
   	return 0;
   }
 
+static int ds3000_read_signal_strength(struct dvb_frontend *fe,
+						u16 *signal_strength)
+{
+	if (fe->ops.tuner_ops.get_rf_strength)
+		fe->ops.tuner_ops.get_rf_strength(fe, signal_strength);
+
+	return 0;
+}
+
   /* calculate DS3000 snr value in dB */
   static int ds3000_read_snr(struct dvb_frontend *fe, u16 *snr)
   {

@@ -1102,6 +1111,7 @@ static struct dvb_frontend_ops ds3000_ops = {
   	.i2c_gate_ctrl = ds3000_i2c_gate_ctrl,
   	.read_status = ds3000_read_status,
   	.read_ber = ds3000_read_ber,
+	.read_signal_strength = ds3000_read_signal_strength,
   	.read_snr = ds3000_read_snr,
   	.read_ucblocks = ds3000_read_ucblocks,
   	.set_voltage = ds3000_set_voltage,

drivers/media/dvb-frontends/m88rs2000.c +3 -1

index 283c90f..4da5272 100644
@@ -446,7 +446,9 @@ static int m88rs2000_read_ber(struct dvb_frontend *fe, u32 *ber)
   static int m88rs2000_read_signal_strength(struct dvb_frontend *fe,
   	u16 *strength)
   {
-	*strength = 0;
+	if (fe->ops.tuner_ops.get_rf_strength)
+		fe->ops.tuner_ops.get_rf_strength(fe, strength);
+
   	return 0;
   }

drivers/media/dvb-frontends/ts2020.c +0 -1

index f50e237..ad7ad85 100644
@@ -363,7 +363,6 @@ struct dvb_frontend *ts2020_attach(struct dvb_frontend *fe,
 
   	memcpy(&fe->ops.tuner_ops, &ts2020_tuner_ops,
   				sizeof(struct dvb_tuner_ops));
-	fe->ops.read_signal_strength = fe->ops.tuner_ops.get_rf_strength;
 
   	return fe;
   }

Signature · 1 line

http://palosaari.fi/

Malcolm Priestley

author maintainer

2013-01-06 20:35 UTC #4 lore raw

Quoted 34 lines · Antti Palosaari

Quoted 12 lines · Malcolm Priestley
Quoted 11 lines · Antti Palosaari
Quoted 9 lines · Malcolm Priestley
Restore ds3000.c read_signal_strength.
Call tuner get_rf_strength from frontend read_signal_strength.
We are able to do a NULL check and doesn't limit the tuner attach to the frontend attach area.
At the moment the lmedm04 tuner attach is stuck in frontend attach area.
Quoted 8 lines
Quoted 5 lines
I would like to nack that, as I see some problems:
1) it changes deviation against normal procedures
2) interface driver (usb/pci) should have full control to make decision
3) you shoot to our own leg easily in power management
This patch does not do any operational changes, and is a proper way to call to another module with a run time NULL check. The same way as another tuner function from demodulator is called.
uh, certainly I understand it does not change functionality in that case! I tried to point out it is logically insane and error proof. Ee should make this kind of direct calls between drivers as less as possible - just to make life easier in future. It could work for you, but for some other it could cause problems as hardware is designed differently.
Quoted 5 lines
Quoted 4 lines
* actually bug 3) already happened some drivers, like rtl28xxu. Tuner is behind demod and demod is put sleep => no access to tuner. FE callback is overridden (just like you are trying to do as default) which means user-space could still make queries => I/O errors.

Quoted 5 lines

Quoted 1 line
In such cases, the tuner init/sleep should also be called.
???????
I think you don't understand functionality at all!

Please, I have been working with the I2C bus in the electronics field for the last 20 years.

If you are really worried about the the tuner being a sleep. You add fe variable check this in it's own init/sleep and define the function something like this.

static int fe_foo_read_signal_strength(struct dvb_frontend *fe,

	u16 *strength)
{
	struct fe_foo_state *state = fe->demodulator_priv;

	if (state->fe_inactive) {
... any extra commands to restore tuner power
		if (fe->ops.tuner_ops.init)
			fe->ops.tuner_ops.init(fe);
			
	}		

	if (fe->ops.tuner_ops.get_rf_strength)
		fe->ops.tuner_ops.get_rf_strength(fe, strength);

	if (state->fe_inactive) {
		if (fe->ops.tuner_ops.sleep)
			fe->ops.tuner_ops.sleep(fe);
... any extra commands to remove tuner power
	}

	return 0;
}

However, I think this is unnecessary in the rs2000 and ds3000.

Regards

Malcolm