On Fri, 26 Apr 2024 17:42:16 +0200
Nuno Sa via B4 Relay <devnull+nuno.sa.analog.com@xxxxxxxxxx> wrote:
> From: Nuno Sa <nuno.sa@xxxxxxxxxx>
>
> To make sure that we have the best timings on the serial data interface
> we should calibrate it. This means going through the device supported
> values and see for which ones we get a successful result. To do that, we
> use a prbs test pattern both in the IIO backend and in the frontend
> devices. Then for each of the test points we see if there are any
> errors. Note that the backend is responsible to look for those errors.
>
> As calibrating the interface also requires that the data format is disabled
> (the one thing being done in ad9467_setup()), ad9467_setup() was removed
> and configuring the data fomat is now part of the calibration process.
>
> Signed-off-by: Nuno Sa <nuno.sa@xxxxxxxxxx>
One trivial comment.
I'd have picked up the whole series, but it feels too big to do on a Sunday
when you only posted on Friday. Hence, lets let it sit for at least
a few more days to see if others have comments.
It might not make this cycle as a result. I've picked up the 2 fixes
today to increase the chances those make it.
Jonathan
> static int ad9467_read_raw(struct iio_dev *indio_dev,
> struct iio_chan_spec const *chan,
> int *val, int *val2, long m)
> @@ -345,7 +606,9 @@ static int ad9467_write_raw(struct iio_dev *indio_dev,
> {
> struct ad9467_state *st = iio_priv(indio_dev);
> const struct ad9467_chip_info *info = st->info;
> + unsigned long sample_rate;
> long r_clk;
> + int ret;
>
> switch (mask) {
> case IIO_CHAN_INFO_SCALE:
> @@ -358,7 +621,23 @@ static int ad9467_write_raw(struct iio_dev *indio_dev,
> return -EINVAL;
> }
>
> - return clk_set_rate(st->clk, r_clk);
> + sample_rate = clk_get_rate(st->clk);
> + /*
> + * clk_set_rate() would also do this but since we would still
> + * need it for avoiding an unnecessary calibration, do it now.
> + */
> + if (sample_rate == r_clk)
> + return 0;
> +
> + iio_device_claim_direct_scoped(return -EBUSY, indio_dev) {
> + ret = clk_set_rate(st->clk, r_clk);
> + if (ret)
> + return ret;
> +
> + guard(mutex)(&st->lock);
> + ret = ad9467_calibrate(st);
return ad9467_calibrate(st);
> + }
unreachable();
not totally elegant but I think the early return makes more sense and we should
just use an unreachable() to squash the resulting compiler warning.
> + return ret;
> default:
> return -EINVAL;
> }
