From: sashiko-bot@kernel.org
To: "Marcelo Schmitt" <marcelo.schmitt@analog.com>,
devicetree@vger.kernel.org
Cc: robh@kernel.org, conor+dt@kernel.org
Subject: Re: [PATCH v1 5/5] iio: adc: Add support for ADA4355 and similar ADCs
Date: Wed, 18 Nov 2026 13:27:15 +0000 [thread overview]
Message-ID: <sashiko-outbox-221011@kernel.org> (raw)
In-Reply-To: <2f5e83aa17aea08bbff69ca6ef7d32b11368f518.1795003864.git.marcelo.schmitt@analog.com>
Thank you for your contribution! Sashiko AI review found 6 potential
issues to consider:
- [High] iio: adc: ada4355: backend channel 0 left disabled after
calibration
- [High] iio: adc: ada4355: 1000x scale error in ada4355_fill_scale_tb()
- [Medium] iio: adc: ada4355: negative sampling freq wraps in
ada4355_write_raw()
- [Medium] iio: adc: ada4355: missing cleanup on error in
ada4355_calibrate()
- [Medium] iio: adc: ada4355: power supply regulators not enabled in
ada4355_probe()
- [Low] iio: adc: ada4355: inconsistent st->lock usage in read and
debugfs paths
--
commit 3d8753d370ebe3f8cd0cc6198719fc1f9fb5fb1d
Author: Marcelo Schmitt <marcelo.schmitt@analog.com>
iio: adc: Add support for ADA4355 and similar ADCs
This commit adds an IIO ADC driver for the Analog Devices ADA4355 and
ADA4356 current-input ADCs, using SPI for configuration and an IIO backend
for LVDS data capture.
> diff --git a/drivers/iio/adc/ada4355.c b/drivers/iio/adc/ada4355.c
> new file mode 100644
> index 0000000000000..7f9f58a143856
> --- /dev/null
> +++ b/drivers/iio/adc/ada4355.c
[ ... ]
> +/* ADC differential input span, ±1 V with the 1 V internal reference. */
> +#define ADA4355_ADC_SPAN_mV 2000
> +/* ADC differential input with no input current flowing. */
> +#define ADA4355_ADC_BIAS_mV 825
> +
> +#define ADA4355_NUM_GSEL_GPIOS 2
> +
> +enum ada4355_tz_options {
> + ADA4355_TZ0,
> + ADA4355_TZ1,
> + ADA4355_TZ2,
> +};
> +
> +static const unsigned int ada4355_tz_options_ohms[] = {
> + [ADA4355_TZ1] = 2000, /* -80 to +800 uA input range */
> + [ADA4355_TZ2] = 20000, /* -8.0 to +80 uA input range */
> + [ADA4355_TZ0] = 200000, /* -0.8 to +8.0 uA input range */
> +};
[ ... ]
> +static int ada4355_debugfs_reg_access(struct iio_dev *indio_dev, unsigned int reg,
> + unsigned int writeval, unsigned int *readval)
> +{
> + struct ada4355_state *st = iio_priv(indio_dev);
> +
> + if (readval)
> + return regmap_read(st->regmap, reg, readval);
> +
> + return regmap_write(st->regmap, reg, writeval);
> +}
> +
> +/*
> + * The ADC output code decreases as the input current increases so the scale to
> + * milliamps is sign reversed.
> + */
> +static void ada4355_fill_scale_tb(struct ada4355_state *st,
> + struct ada4355_chip_info const *chip_info)
> +{
> + u64 numerator, denominator;
> + u32 rem;
> +
> + st->scale_tb_size = chip_info->gain_options_size;
> + for (unsigned int i = 0; i < st->scale_tb_size; i++) {
> + numerator = (u64)ADA4355_ADC_SPAN_mV * NANO * MILLI;
> + denominator = BIT_ULL(st->chan->scan_type.realbits) * chip_info->gain_options[i];
> + numerator = div64_u64(numerator, denominator);
> + st->scale_tb[i][0] = div_u64_rem(numerator, NANO, &rem);
> + st->scale_tb[i][1] = -(int)rem;
> + }
> +}
[Severity: High]
Does multiplying ADA4355_ADC_SPAN_mV by MILLI here make st->scale_tb[][]
1000 times too large?
ADA4355_ADC_SPAN_mV is already in millivolts (2000 mV) and
chip_info->gain_options[i] is in Ohms, so dividing ADA4355_ADC_SPAN_mV by
gain_options[i] directly gives the full-scale current in milliamps. The
extra MILLI factor reports microamps per LSB instead of milliamps per LSB
(for example, -0.061035156 mA/LSB instead of -0.000061035 mA/LSB at 2000
Ohms).
> +
> +static int ada4355_get_scale(struct ada4355_state *st, int *val, int *val2)
> +{
> + *val = st->scale_tb[st->scale_idx][0];
> + *val2 = st->scale_tb[st->scale_idx][1];
> +
> + return IIO_VAL_INT_PLUS_NANO;
> +}
> +
> +static int ada4355_set_scale(struct ada4355_state *st, int val, int val2)
> +{
> + unsigned long gsel;
> + int ret;
> +
> + guard(mutex)(&st->lock);
[Severity: Low]
Should ada4355_get_scale() and ada4355_debugfs_reg_access() also take
st->lock?
ada4355_set_scale() and ada4355_update_clock() both run under
IIO_DEV_ACQUIRE_DIRECT_MODE() in ada4355_write_raw() and take st->lock,
whereas ada4355_get_scale() reads st->scale_idx twice without st->lock and
ada4355_debugfs_reg_access() accesses st->regmap without serializing
against the multi-register reset and calibration sequences in
ada4355_update_clock().
[ ... ]
> +static int ada4355_calibrate_prepare(struct ada4355_state *st)
> +{
[ ... ]
> + return iio_backend_chan_enable(st->back, 0);
> +}
> +
> +static int ada4355_calibrate_stop(struct ada4355_state *st)
> +{
> + int ret;
> +
> + ret = iio_backend_calibration_source_set(st->back,
> + IIO_BACKEND_CALIB_SOURCE_NONE,
> + 0);
> + if (ret)
> + return ret;
> +
> + ret = iio_backend_chan_disable(st->back, 0);
> + if (ret)
> + return ret;
> +
> + return ada4355_testmode_set(st, ADA4355_TEST_MODE_OFF);
> +}
[Severity: High]
Does disabling backend channel 0 in ada4355_calibrate_stop() leave the
channel disabled during buffered captures?
Nothing else in the driver calls iio_backend_chan_enable(st->back, 0):
ada4355_iio_info does not implement update_scan_mode and no buffer setup
ops are registered. Once ada4355_calibrate_stop() disables channel 0 (or
if the backend does not set IIO_BACKEND_CAP_CALIBRATION and skips
ada4355_calibrate() altogether), backend channel 0 stays disabled when
userspace starts a buffered capture.
[ ... ]
> +static int ada4355_calibrate(struct ada4355_state *st)
> +{
> + unsigned int lane;
> + int ret;
> +
> + /* all points invalid */
> + bitmap_fill(st->calib_map, st->calib_map_size);
> +
> + ret = ada4355_calibrate_prepare(st);
> + if (ret)
> + return ret;
> +
> + /*
> + * The frame signal has to be aligned first: the pattern match block
> + * only reports meaningful per-lane errors once the word boundary is
> + * known.
> + */
> + ret = ada4355_calibrate_lane(st, IIO_BACKEND_CALIB_SOURCE_FRAME,
> + ADA4355_FRAME_LANE,
> + ADA4355_MAX_DATA_LANES * ADA4355_MAX_TEST_POINTS);
> + if (ret)
> + return ret;
> +
> + for (lane = 0; lane < ADA4355_MAX_DATA_LANES; lane++) {\
> + ret = ada4355_calibrate_lane(st,
> + IIO_BACKEND_CALIB_SOURCE_DATA_LANE,
> + lane,
> + lane * ADA4355_MAX_TEST_POINTS);
> + if (ret)
> + return ret;
> + }
> +
> + return ada4355_calibrate_stop(st);
> +}
[Severity: Medium]
If ada4355_calibrate_lane() fails during a runtime clock update in
ada4355_update_clock(), does returning early here leave the ADC in
ADA4355_TEST_MODE_USER and the backend calibration source enabled?
[ ... ]
> +static int ada4355_write_raw(struct iio_dev *indio_dev,
> + struct iio_chan_spec const *chan,
> + int val, int val2, long mask)
> +{
> + struct ada4355_state *st = iio_priv(indio_dev);
> + long r_clk;
> +
> + IIO_DEV_ACQUIRE_DIRECT_MODE(indio_dev, claim);
> + if (IIO_DEV_ACQUIRE_FAILED(claim))
> + return -EBUSY;
> +
> + switch (mask) {
> + case IIO_CHAN_INFO_SAMP_FREQ:
> + r_clk = clk_round_rate(st->clk, val);
> + if (!in_range(r_clk, 1, ADA4355_MAX_SAMP_RATE_SPS))
> + return -ERANGE;
> +
> + return ada4355_update_clock(st, r_clk);
[Severity: Medium]
Can a negative or zero val in ada4355_write_raw() reprogram the clock
instead of returning an error?
clk_round_rate() takes an unsigned long rate, so a negative val sign-extends
to a large unsigned value and rounds down to the clock provider's maximum
rate (and 0 rounds up to its minimum rate). If that rounded rate falls
within [1, ADA4355_MAX_SAMP_RATE_SPS], ada4355_update_clock() applies it.
[ ... ]
> + st->clk = devm_clk_get_enabled(dev, NULL);
> + if (IS_ERR(st->clk))
> + return PTR_ERR(st->clk);
> +
> + indio_dev->name = chip_info->name;
> + indio_dev->channels = &ada4355_channel;
> + indio_dev->num_channels = 1;
> + indio_dev->info = &ada4355_iio_info;
> + st->chan = &ada4355_channel;
> +
> + ret = ada4355_setup(st, chip_info);
[Severity: Medium]
Should ada4355_probe() enable the vcc-supply, vea-supply, and ved-supply
regulators before accessing the device in ada4355_setup()?
The DT binding in Documentation/devicetree/bindings/iio/adc/adi,ada4355.yaml
requires vcc-supply and lists vea-supply and ved-supply as optional
supplies, but the driver never requests or enables them, leaving the chip
unpowered on boards where those regulators are not always-on.
--
Sashiko AI review · https://sashiko.dev/#/patchset/cover.1795003864.git.marcelo.schmitt@analog.com?part=5
prev parent reply other threads:[~2026-11-18 13:27 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-18 13:02 [PATCH v1 0/5] Add support for ADA4355 Marcelo Schmitt
2026-11-18 13:03 ` [PATCH v1 1/5] iio: backend: Add calibration source interface Marcelo Schmitt
2026-11-18 13:03 ` [PATCH v1 2/5] dt-bindings: iio: adc: Add AXI ADA4355 IP CORE Marcelo Schmitt
2026-11-18 13:03 ` [PATCH v1 3/5] iio: adc: adi-axi-adc: Support AXI ADA4355 IP core Marcelo Schmitt
2026-11-18 13:04 ` [PATCH v1 4/5] dt-bindings: iio: adc: Add ADA4355 Marcelo Schmitt
2026-11-18 13:04 ` [PATCH v1 5/5] iio: adc: Add support for ADA4355 and similar ADCs Marcelo Schmitt
2026-11-18 13:27 ` sashiko-bot [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=sashiko-outbox-221011@kernel.org \
--to=sashiko-bot@kernel.org \
--cc=conor+dt@kernel.org \
--cc=devicetree@vger.kernel.org \
--cc=marcelo.schmitt@analog.com \
--cc=robh@kernel.org \
--cc=sashiko-reviews@lists.linux.dev \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox