Support of Nova S SE DVB card missing

Actual kernel does not support the Hauppauge WinTV Nova S SE PCI card anymore:

Quoted 3 lines · Eberhard Kaltenhaeuser

Actual kernel does not support the Hauppauge WinTV Nova S SE PCI card anymore:

I think it is a problem of this changeset http://linuxtv.org/hg/v4l-dvb/rev/358d281e6a3d
from Patrick Boettcher. The S5H1420 isn't able to understand repeated start conditions. The i2c-read code was changed from:

	if ((ret = i2c_transfer (state->i2c, &msg1, 1)) != 1)
		return ret;

	if ((ret = i2c_transfer (state->i2c, &msg2, 1)) != 1)
		return ret;

to:

	if (state->config->repeated_start_workaround) {
		ret = i2c_transfer(state->i2c, msg, 3);
		if (ret != 3)
			return ret;
	} else {
		ret = i2c_transfer(state->i2c, &msg[1], 2);
		if (ret != 2)
			return ret;
	}

-Hartmut

Hi Hartmut and Eberhard,

thanks for pointing that out, I almost overlooked the previous mail from Eberhard.

Eberhard, are you able to try patches and to compile your own drivers in order to help finding the best solution.

One option is to put back the original code in case the repeated-start-workaround is not set. But this one looks not very protected. I mean between the two i2c_transfer-calls something else could happen.

Is there no other mean to tell to the i2c-adapter to do a repeated start within one i2c_transfer-call?

Another option would be to try to set the "repeated_start_workaround" option also for the Nova SE card.

What do you think?

Patrick.

--
Mail: [email protected]
WWW: http://www.wi-bw.tfh-wildau.de/~pboettch/

Quoted 27 lines · On Sat, 23 Aug 2008, e9hack wrote:

Quoted 4 lines · Eberhard Kaltenhaeuser

Actual kernel does not support the Hauppauge WinTV Nova S SE PCI card anymore:

I think it is a problem of this changeset http://linuxtv.org/hg/v4l-dvb/rev/358d281e6a3d
from Patrick Boettcher. The S5H1420 isn't able to understand repeated start conditions. The i2c-read code was changed from:

	if ((ret = i2c_transfer (state->i2c, &msg1, 1)) != 1)
		return ret;

	if ((ret = i2c_transfer (state->i2c, &msg2, 1)) != 1)
		return ret;

to:

	if (state->config->repeated_start_workaround) {
		ret = i2c_transfer(state->i2c, msg, 3);
		if (ret != 3)
			return ret;
	} else {
		ret = i2c_transfer(state->i2c, &msg[1], 2);
		if (ret != 2)
			return ret;
	}

Quoted 3 lines

-Hartmut

_______________________________________________
linux-dvb mailing list
[email protected]
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb

Quoted 3 lines · Eberhard Kaltenhaeuser

Actual kernel does not support the Hauppauge WinTV Nova S SE PCI card anymore:

I think it is a problem of this changeset http://linuxtv.org/hg/v4l-dvb/rev/358d281e6a3d
from Patrick Boettcher. The S5H1420 isn't able to understand repeated start conditions. The i2c-read code was changed from:

	if ((ret = i2c_transfer (state->i2c, &msg1, 1)) != 1)
		return ret;

	if ((ret = i2c_transfer (state->i2c, &msg2, 1)) != 1)
		return ret;

to:

	if (state->config->repeated_start_workaround) {
		ret = i2c_transfer(state->i2c, msg, 3);
		if (ret != 3)
			return ret;
	} else {
		ret = i2c_transfer(state->i2c, &msg[1], 2);
		if (ret != 2)
			return ret;
	}

Btw, I don't understand Patrick's workaround.

Hi,

Quoted 1 line · Oliver Endriss

Btw, I don't understand Patrick's workaround.

The Flexcop i2c-interface is not very flexible. You cannot send just a single write request with an independent read request following.

The same problematic applies for several USB-I2C requests as we have it in dvb-usb at several places.

In addition (see my other mail in that thread), sending two independent i2c_transfers which actually belong together is not really safe.

(However
I understand that for most single-receiver boards it is no real problem, as long as no one is using this i2c-adapter from user-space at the same time.)

Quoted 2 lines

In addition (see my other mail in that thread), sending two independent i2c_transfers which actually belong together is not really safe.

The current code in the else path will *never* work, because the tuner does not support repeated start conditions. The problem is not the I2C master (saa7146/flexcop) but the I2C slave (s5h1420).

Quoted 3 lines · Oliver Endriss

Quoted 2 lines

In addition (see my other mail in that thread), sending two independent i2c_transfers which actually belong together is not really safe.

Quoted 3 lines

The current code in the else path will *never* work, because the tuner does not support repeated start conditions. The problem is not the I2C master (saa7146/flexcop) but the I2C slave (s5h1420).

Wouldn't it be more correct to have a flag signaling to the i2c_tranfer-function that a repeated start is not wanted even though it is two i2c-messages glued (which are interpreted today as a read with repeated start).

As there is a flag I2C_M_NOSTART in the I2C subsystem in recent kernels, we could pass this flag to the I2C driver and add a workaround to the i2c master_xfer function.

Quoted 8 lines · Patrick Boettcher

Quoted 6 lines · Oliver Endriss

Quoted 2 lines

In addition (see my other mail in that thread), sending two independent i2c_transfers which actually belong together is not really safe.

The current code in the else path will *never* work, because the tuner does not support repeated start conditions. The problem is not the I2C master (saa7146/flexcop) but the I2C slave (s5h1420).

Quoted 4 lines

Wouldn't it be more correct to have a flag signaling to the i2c_tranfer-function that a repeated start is not wanted even though it is two i2c-messages glued (which are interpreted today as a read with repeated start).

I remember that we had the same discussion for the stv0297 driver a long time ago.

For the stv0297 I have an experimental patch which intercepts the master_xfer routine, but this is not very nice either.

Quoted 14 lines · e9hack wrote:

Quoted 3 lines · Eberhard Kaltenhaeuser

Actual kernel does not support the Hauppauge WinTV Nova S SE PCI card anymore:

I think it is a problem of this changeset http://linuxtv.org/hg/v4l-dvb/rev/358d281e6a3d from Patrick Boettcher. The S5H1420 isn't able to understand repeated start conditions. The i2c-read code was changed from:

	if ((ret = i2c_transfer (state->i2c, &msg1, 1)) != 1)
		return ret;

	if ((ret = i2c_transfer (state->i2c, &msg2, 1)) != 1)
		return ret;

to:

Quoted 10 lines

	if (state->config->repeated_start_workaround) {
		ret = i2c_transfer(state->i2c, msg, 3);
		if (ret != 3)
			return ret;
	} else {
		ret = i2c_transfer(state->i2c, &msg[1], 2);
		if (ret != 2)
			return ret;
	}

I think you are right.

Btw, I don't understand Patrick's workaround.

As the tuner does not support repeated start conditions, the solution is to send two separate messages, as it was before.

Does the attached patch fix the problem?

CU
Oliver

------------------------------------------------------------------------

_______________________________________________
linux-dvb mailing list
[email protected]
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb

Quoted 16 lines · e9hack wrote:

Quoted 4 lines · Eberhard Kaltenhaeuser

Actual kernel does not support the Hauppauge WinTV Nova S SE PCI card anymore:

I think it is a problem of this changeset http://linuxtv.org/hg/v4l-dvb/rev/358d281e6a3d from Patrick Boettcher. The S5H1420 isn't able to understand repeated start conditions. The i2c-read code was changed from:

	if ((ret = i2c_transfer (state->i2c, &msg1, 1)) != 1)
		return ret;

	if ((ret = i2c_transfer (state->i2c, &msg2, 1)) != 1)
		return ret;

to:

Quoted 11 lines

	if (state->config->repeated_start_workaround) {
		ret = i2c_transfer(state->i2c, msg, 3);
		if (ret != 3)
			return ret;
	} else {
		ret = i2c_transfer(state->i2c, &msg[1], 2);
		if (ret != 2)
			return ret;
	}

I think you are right.

Btw, I don't understand Patrick's workaround.

As the tuner does not support repeated start conditions, the solution is to send two separate messages, as it was before.

Does the attached patch fix the problem?

CU
Oliver

------------------------------------------------------------------------

diff -r 1760a612cc98 linux/drivers/media/dvb/frontends/s5h1420.c

linux/drivers/media/dvb/frontends/s5h1420.c +5 -3

@@ -94,8 +94,11 @@ static u8 s5h1420_readreg(struct s5h1420
 		if (ret != 3)
 			return ret;
 	} else {
-		ret = i2c_transfer(state->i2c, &msg[1], 2);
-		if (ret != 2)
+		ret = i2c_transfer(state->i2c, &msg[1], 1);
+		if (ret != 1)
+			return ret;
+		ret = i2c_transfer(state->i2c, &msg[2], 1);
+		if (ret != 1)
 			return ret;
 	}
 
 
 
------------------------------------------------------------------------

_______________________________________________
linux-dvb mailing list
[email protected]
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb