| Message ID | 1417990203-758-2-git-send-email-benjamin@southpole.se (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 12/08/2014 12:10 AM, Benjamin Larsson wrote: > The firmware must be loaded one byte at a time via the 0xf6 register. I don't think so. Currently it downloads firmware in 22 byte chunks and it seems to work, at least for me, both mn88472 and mn88473. > Signed-off-by: Benjamin Larsson <benjamin@southpole.se> > --- > drivers/staging/media/mn88472/mn88472.c | 21 +++++++-------------- > 1 file changed, 7 insertions(+), 14 deletions(-) > > diff --git a/drivers/staging/media/mn88472/mn88472.c b/drivers/staging/media/mn88472/mn88472.c > index ffee187..ba1bc8d 100644 > --- a/drivers/staging/media/mn88472/mn88472.c > +++ b/drivers/staging/media/mn88472/mn88472.c > @@ -290,7 +290,7 @@ static int mn88472_init(struct dvb_frontend *fe) > { > struct i2c_client *client = fe->demodulator_priv; > struct mn88472_dev *dev = i2c_get_clientdata(client); > - int ret, len, remaining; > + int ret, i; > const struct firmware *fw = NULL; > u8 *fw_file = MN88472_FIRMWARE; > > @@ -330,19 +330,12 @@ static int mn88472_init(struct dvb_frontend *fe) > if (ret) > goto err; > > - for (remaining = fw->size; remaining > 0; > - remaining -= (dev->i2c_wr_max - 1)) { > - len = remaining; > - if (len > (dev->i2c_wr_max - 1)) > - len = (dev->i2c_wr_max - 1); > - > - ret = regmap_bulk_write(dev->regmap[0], 0xf6, > - &fw->data[fw->size - remaining], len); > - if (ret) { > - dev_err(&client->dev, > - "firmware download failed=%d\n", ret); > - goto err; > - } > + for (i = 0 ; i < fw->size ; i++) > + ret |= regmap_write(dev->regmap[0], 0xf6, fw->data[i]); > + if (ret) { > + dev_err(&client->dev, > + "firmware download failed=%d\n", ret); > + goto err; > } Not nice. 1) You mask status and you could not know if error code is valid after mask few thousand error codes. 2) Even worse, it is loop that runs thousand of times. Guess how much I/O errors there could happen. There is many times situation when first error occur then all the rest commands are failing too. And many cases failing I2C command could be failed USB message, which could take few seconds. Very typical USB timeout is 2 secs, this means 2k firmware, 2*2000=4000 sec => it blocks that routine over *one* hour. > > ret = regmap_write(dev->regmap[0], 0xf5, 0x00); > regards Antti
On 12/07/2014 11:36 PM, Antti Palosaari wrote: > On 12/08/2014 12:10 AM, Benjamin Larsson wrote: >> The firmware must be loaded one byte at a time via the 0xf6 register. > > I don't think so. Currently it downloads firmware in 22 byte chunks > and it seems to work, at least for me, both mn88472 and mn88473. With both these changes I get much better sensitivity. So something is better then before. I will track down the needed changes and respin the patches. MvH Benjamin Larsson -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 12/07/2014 11:36 PM, Antti Palosaari wrote: > On 12/08/2014 12:10 AM, Benjamin Larsson wrote: >> The firmware must be loaded one byte at a time via the 0xf6 register. > > I don't think so. Currently it downloads firmware in 22 byte chunks > and it seems to work, at least for me, both mn88472 and mn88473. Ok, I have now tried the driver with my defaults patch in and with your method of loading the firmware and my patch. I have my antenna placed in a bad location with bad reception. With my patch I am getting data from the device, without my patch I am not. So whatever my code does it makes the device more sensitive. And then there is this comment in the regmap code: regmap_bulk_write(): Write multiple registers to the device In this case we want to write multiple bytes to the same register. So I think that my patch is correct in principle. MvH Benjamin Larsson -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Moikka! On 12/08/2014 01:12 PM, Benjamin Larsson wrote: > On 12/07/2014 11:36 PM, Antti Palosaari wrote: >> On 12/08/2014 12:10 AM, Benjamin Larsson wrote: >>> The firmware must be loaded one byte at a time via the 0xf6 register. >> >> I don't think so. Currently it downloads firmware in 22 byte chunks >> and it seems to work, at least for me, both mn88472 and mn88473. > > With both these changes I get much better sensitivity. So something is > better then before. I will track down the needed changes and respin the > patches. I suspect it is that initialization of all registers which has something to do with sensitivity. I haven't tested if firmware uploading is critical, what happens when some byte is skipped or so... Did you saw there is config parameter i2c_wr_max? Setting it to '1' does same what that your patch did, but leaves amount of max I2C bytes configurable... Anyhow, good finding, which needs to be track down. regards Antti
Hello! On 12/08/2014 06:04 PM, Benjamin Larsson wrote: > On 12/07/2014 11:36 PM, Antti Palosaari wrote: >> On 12/08/2014 12:10 AM, Benjamin Larsson wrote: >>> The firmware must be loaded one byte at a time via the 0xf6 register. >> >> I don't think so. Currently it downloads firmware in 22 byte chunks >> and it seems to work, at least for me, both mn88472 and mn88473. > > Ok, I have now tried the driver with my defaults patch in and with your > method of loading the firmware and my patch. I have my antenna placed in > a bad location with bad reception. With my patch I am getting data from > the device, without my patch I am not. So whatever my code does it makes > the device more sensitive. > > And then there is this comment in the regmap code: > > regmap_bulk_write(): Write multiple registers to the device > > In this case we want to write multiple bytes to the same register. So I > think that my patch is correct in principle. You haven't make any test whether it is possible to write that firmware in a large chunks *or* writing one byte (smallest possible ~chuck) at the time? I think it does not matter. I suspect you could even download whole firmware as one go - but rtl2832p I2C adapter does support only 22 bytes on one xfer. Even those are written to one register, chip knows how many bytes one message has and could increase its internal address counter. That is usually called register address auto-increment. A) writing: f6 00 f6 01 f6 02 f6 03 f6 04 f6 05 f6 06 f6 07 f6 08 f6 09 B) writing: f6 00 01 02 03 04 f6 05 06 07 08 09 will likely end up same. B is better as only 2 xfers are done - much less IO. regards Antti
diff --git a/drivers/staging/media/mn88472/mn88472.c b/drivers/staging/media/mn88472/mn88472.c index ffee187..ba1bc8d 100644 --- a/drivers/staging/media/mn88472/mn88472.c +++ b/drivers/staging/media/mn88472/mn88472.c @@ -290,7 +290,7 @@ static int mn88472_init(struct dvb_frontend *fe) { struct i2c_client *client = fe->demodulator_priv; struct mn88472_dev *dev = i2c_get_clientdata(client); - int ret, len, remaining; + int ret, i; const struct firmware *fw = NULL; u8 *fw_file = MN88472_FIRMWARE; @@ -330,19 +330,12 @@ static int mn88472_init(struct dvb_frontend *fe) if (ret) goto err; - for (remaining = fw->size; remaining > 0; - remaining -= (dev->i2c_wr_max - 1)) { - len = remaining; - if (len > (dev->i2c_wr_max - 1)) - len = (dev->i2c_wr_max - 1); - - ret = regmap_bulk_write(dev->regmap[0], 0xf6, - &fw->data[fw->size - remaining], len); - if (ret) { - dev_err(&client->dev, - "firmware download failed=%d\n", ret); - goto err; - } + for (i = 0 ; i < fw->size ; i++) + ret |= regmap_write(dev->regmap[0], 0xf6, fw->data[i]); + if (ret) { + dev_err(&client->dev, + "firmware download failed=%d\n", ret); + goto err; } ret = regmap_write(dev->regmap[0], 0xf5, 0x00);
The firmware must be loaded one byte at a time via the 0xf6 register. Signed-off-by: Benjamin Larsson <benjamin@southpole.se> --- drivers/staging/media/mn88472/mn88472.c | 21 +++++++-------------- 1 file changed, 7 insertions(+), 14 deletions(-)