Date:	Fri, 21 Sep 2012 18:46:05 +0200
From:	Hans Verkuil <hverkuil@...all.nl>
To:	"andrey.smirnov@...vergeddevices.net" 
	<andrey.smirnov@...vergeddevices.net>
Cc:	linux-media@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 1/3] Add a core driver for SI476x MFD

On Fri September 21 2012 18:33:45 andrey.smirnov@...vergeddevices.net wrote:
> On 09/21/2012 12:31 AM, Hans Verkuil wrote:
> > On Fri September 21 2012 03:05:41 andrey.smirnov@...vergeddevices.net wrote:
> >> On 09/13/2012 11:44 PM, Hans Verkuil wrote:
> >>> Hi Andrey!
> >>>
> >>> Thanks for posting this driver. One request for the future: please split this
> >>> patch up in smaller pieces: one for each c source for example. That makes it
> >>> easier to review.
> >> Will do for next version.
> >>
> >>> +
> >>> +/**
> >>> + * __core_send_command() - sends a command to si476x and waits its
> >>> + * response
> >>> + * @core:    si476x_device structure for the device we are
> >>> + *            communicating with
> >>> + * @command:  command id
> >>> + * @args:     command arguments we are sending
> >>> + * @argn:     actual size of @args
> >>> + * @response: buffer to place the expected response from the device
> >>> + * @respn:    actual size of @response
> >>> + * @usecs:    amount of time to wait before reading the response (in
> >>> + *            usecs)
> >>> + *
> >>> + * Function returns 0 on succsess and negative error code on
> >>> + * failure
> >>> + */
> >>> +static int __core_send_command(struct si476x_core *core,
> >>> +				    const u8 command,
> >>> +				    const u8 args[],
> >>> +				    const int argn,
> >>> +				    u8 resp[],
> >>> +				    const int respn,
> >>> +				    const int usecs)
> >>> +{
> >>> +	struct i2c_client *client = core->client;
> >>> +	int err;
> >>> +	u8  data[CMD_MAX_ARGS_COUNT + 1];
> >>> +
> >>> +	if (argn > CMD_MAX_ARGS_COUNT) {
> >>> +		err = -ENOMEM;
> >>> +		goto exit;
> >>> Why goto exit? There is no clean up after the exit label, so just return
> >>> immediately. Ditto for all the other goto exit's in this function.
> >> To have only just on point of exit from the function that's just
> >> personal coding style preference.
> >> There are no technical reasons behind that, I can change that.
> >>
> >>>> +	}
> >>>> +
> >>>> +	if (!client->adapter) {
> >>>> +		err = -ENODEV;
> >>>> +		goto exit;
> >>>> +	}
> >>>> +
> >>>> +	/* First send the command and its arguments */
> >>>> +	data[0] = command;
> >>>> +	memcpy(&data[1], args, argn);
> >>>> +	DBG_BUFFER(&client->dev, "Command:\n", data, argn + 1);
> >>>> +
> >>>> +	err = si476x_i2c_xfer(core, SI476X_I2C_SEND, (char *) data, argn + 1);
> >>>> +	if (err != argn + 1) {
> >>>> +		dev_err(&core->client->dev,
> >>>> +			"Error while sending command 0x%02x\n",
> >>>> +			command);
> >>>> +		err = (err >= 0) ? -EIO : err;
> >>>> +		goto exit;
> >>>> +	}
> >>>> +	/* Set CTS to zero only after the command is send to avoid
> >>>> +	 * possible racing conditions when working in polling mode */
> >>>> +	atomic_set(&core->cts, 0);
> >>>> +
> >>>> +	if (!wait_event_timeout(core->command,
> >>>> +				atomic_read(&core->cts),
> >>>> +				usecs_to_jiffies(usecs) + 1))
> >>>> +		dev_warn(&core->client->dev,
> >>>> +			 "(%s) [CMD 0x%02x] Device took too much time to answer.\n",
> >>>> +			 __func__, command);
> >>>> +
> >>>> +	/*
> >>>> +	  When working in polling mode, for some reason the tuner will
> >>>> +	  report CTS bit as being set in the first status byte read,
> >>>> +	  but all the consequtive ones will return zros until the
> >>>> +	  tuner is actually completed the POWER_UP command. To
> >>>> +	  workaround that we wait for second CTS to be reported
> >>>> +	 */
> >>>> +	if (unlikely(!core->client->irq && command == CMD_POWER_UP)) {
> >>>> +		if (!wait_event_timeout(core->command,
> >>>> +					atomic_read(&core->cts),
> >>>> +					usecs_to_jiffies(usecs) + 1))
> >>>> +			dev_warn(&core->client->dev,
> >>>> +				 "(%s) Power up took too much time.\n",
> >>>> +				 __func__);
> >>>> +	}
> >>>> +
> >>>> +	/* Then get the response */
> >>>> +	err = si476x_i2c_xfer(core, SI476X_I2C_RECV, resp, respn);
> >>>> +	if (err != respn) {
> >>>> +		dev_err(&core->client->dev,
> >>>> +			"Error while reading response for command 0x%02x\n",
> >>>> +			command);
> >>>> +		err = (err >= 0) ? -EIO : err;
> >>>> +		goto exit;
> >>>> +	}
> >>>> +	DBG_BUFFER(&client->dev, "Response:\n", resp, respn);
> >>>> +
> >>>> +	err = 0;
> >>>> +
> >>>> +	if (resp[0] & SI476X_ERR) {
> >>>> +		dev_err(&core->client->dev, "Chip set error flag\n");
> >>>> +		err = si476x_core_parse_and_nag_about_error(core);
> >>>> +		goto exit;
> >>>> +	}
> >>>> +
> >>>> +	if (!(resp[0] & SI476X_CTS))
> >>>> +		err = -EBUSY;
> >>>> +exit:
> >>>> +	return err;
> >>>> +}
> >>>> +
> >>>> +#define CORE_SEND_COMMAND(core, cmd, args, resp, timeout)		\
> >>>> +	__core_send_command(core, cmd, args,				\
> >>>> +			    ARRAY_SIZE(args),				\
> >>>> +			    resp, ARRAY_SIZE(resp),			\
> >>>> +			    timeout)
> >>>> +
> >>>> +
> >>>> +static int __cmd_tune_seek_freq(struct si476x_core *core,
> >>>> +				uint8_t cmd,
> >>>> +				const uint8_t args[], size_t argn,
> >>>> +				uint8_t *resp, size_t respn,
> >>>> +				int (*clear_stcint) (struct si476x_core *core))
> >>>> +{
> >>>> +	int err;
> >>>> +
> >>>> +	atomic_set(&core->stc, 0);
> >>>> +	err = __core_send_command(core, cmd, args, argn,
> >>>> +				  resp, respn,
> >>>> +				  atomic_read(&core->timeouts.command));
> >>>> +	if (!err) {
> >>> Invert the test to simplify indentation.
> >>>
> >>>> +		if (!wait_event_timeout(core->tuning,
> >>>> +		atomic_read(&core->stc),
> >>>> +		usecs_to_jiffies(atomic_read(&core->timeouts.tune)) + 1)) {
> >>> Weird indentation above. Indent the arguments more to the right.
> >> 80 symbol line length limit is the reason for that indentation.
> > It's not a limit, it's a warning only. Usually readability improves if such
> > long lines are split up or otherwise shortened, but if readability becomes
> > worse because of that, then just leave in the long line.
> >
> >>> Andrey, you should look at the drivers/media/radio/si4713-i2c.c source.
> >>> It is for the same chip family and is much, much smaller.
> >>>
> >>> See if you can use some of the code that's there.
> >> I did when I started writing the driver, that driver and driver for
> >> wl1273 were my two examples. In my initial version of the driver I tried
> >> to blend both si4713 and si476x into one generic driver, but the problem
> >> is: si4713 is a transmitter and si476x are receiver chips, the
> >> "impedance mismatch" in functionality of the two, IMHO, was too much to
> >> justify the unification.
> > But the way the commands are handled, etc. should be the same or very similar.
> > That's the main area where I suspect you can reuse code from those other
> > drivers.
> 
> To reuse the si4713_send_command function from si4713 driver I would
> have to modify the way IRQs are handled by that driver and basically
> replace its code for mine. And the reason for that is because si4761 is
> a receiver, I cannot just rely on timeouts in the case of driver working
> in "no-IRQ" mode(current implementation in si4713 driver). I need a
> polling loop that would allow me to receive RDS and monitor the status
> of the chip. Plus having a polling loop allows me to have almost
> identical implementation of chip events handling for both cases(IRQ and
> no-IRQ). I don't feel comfortable making such drastic changes to si4713
> driver without having access to the actual hardware and being ability to
> test it.
> 
> I really did try to amalgamate two drivers into a single one(and even
> received an earful from my, at that time, supervisor for wasting my time
> on it). But as I mentioned earlier to unite both drivers I would have to
> modify si4713's code and without any actual hardware I cannot do it.

OK, good to know. Thank you for the explanation, and thanks for looking into
this. I'm OK with your approach, although I will take a closer look at the
code handling commands when you post the next version. In a first review I
generally only look at obvious things, in a second review I look more closely
at the code itself. My impression during the first review was that the code
that dealt with the commands could probably be simplified or shortened, but
I didn't look all that closely at it.

Regards,

	Hans
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists