Devicetree
 help / color / mirror / Atom feed
From: sashiko-bot@kernel.org
To: "Olivier Moysan" <olivier.moysan@foss.st.com>
Cc: devicetree@vger.kernel.org, dri-devel@lists.freedesktop.org,
	linux-sound@vger.kernel.org, robh@kernel.org,
	conor+dt@kernel.org, media-ci@linuxtv.org
Subject: Re: [PATCH v2 02/11] iio: adc: stm32-mdf: add core registration
Date: Wed, 18 Nov 2026 15:45:15 +0000	[thread overview]
Message-ID: <sashiko-outbox-221370@kernel.org> (raw)
In-Reply-To: <20261118-upstream_mdf_v2-v2-2-7939d3c3720a@foss.st.com>

Thank you for your contribution! Sashiko AI review found 12 potential
issues to consider:
- [High] iio: adc: stm32-mdf: reset in stm32_mdf_core_parse_of() breaks
  regcache/probe
- [Medium] iio: adc: stm32-mdf: unchecked read and missing rollback in
  start_mdf()
- [Medium] iio: adc: stm32-mdf: lockless n_active_ch races in
  start_mdf() and stop_mdf()
- [Medium] iio: adc: stm32-mdf: unlocked regcache_cache_bypass() races
  with trigger()
- [Medium] iio: adc: stm32-mdf: missing state rollback on suspend and
  resume failure
- [Medium] iio: adc: stm32-mdf: unmapped resource 1 read in
  stm32_mdf_core_probe()
- [Low] iio: adc: stm32-mdf: unused PM callbacks when CONFIG_PM_SLEEP or
  CONFIG_PM=n
- [Low] iio: adc: stm32-mdf: drvdata pointer type mismatch in
  stm32_mdf_core_probe()
- [Low] iio: adc: stm32-mdf: unchecked regmap_read() for MDF_VERR_REG
- [Low] iio: adc: stm32-mdf: guard() mixed with goto in
  stm32_mdf_core_trigger()
- [Low] iio: adc: stm32-mdf: invalid OPTION() macro in
  MDF_OR_OPTION_MASK
- [Low] iio: adc: stm32-mdf: missing header include guards in
  stm32-mdf.h
--

commit 016595fa648050f874751e37d47a5cfa33a78bde
Author: Olivier Moysan <olivier.moysan@foss.st.com>

iio: adc: stm32-mdf: add core registration

This commit adds the core driver for the STM32 multifunction digital
filter (MDF) to manage shared registers, clocks, and power resources on
STM32MP2.

> diff --git a/drivers/iio/adc/stm32-mdf-core.c b/drivers/iio/adc/stm32-mdf-core.c
> new file mode 100644
> index 0000000000000..ca50ce566a435
> --- /dev/null
> +++ b/drivers/iio/adc/stm32-mdf-core.c
[ ... ]
> +int stm32_mdf_core_start_mdf(struct stm32_mdf *mdf)
> +{
> +	struct stm32_mdf_priv *priv = to_stm32_mdf_priv(mdf);
> +	struct device *dev = &priv->pdev->dev;
> +	int ret;
> +	u32 val;
> +
> +	if (atomic_inc_return(&priv->n_active_ch) == 1) {
> +		ret = pm_runtime_resume_and_get(dev);
> +		if (ret < 0)
> +			goto err;
> +
> +		/* Enable PROCDIV and CCKDIV clock dividers */
> +		ret = regmap_set_bits(priv->regmap, MDF_CKGCR_REG, MDF_CKG_CKGDEN);
> +		if (ret < 0)
> +			goto pm_put;
> +
> +		/* Check clock status. Bypass cache to access volatile CKGACTIVE active bit */
> +		regcache_cache_bypass(priv->regmap, true);
> +		regmap_read(priv->regmap, MDF_CKGCR_REG, &val);
> +		regcache_cache_bypass(priv->regmap, false);
> +		if (!(val & MDF_CKG_ACTIVE)) {
> +			ret = -EINVAL;
> +			dev_err(dev, "MDF clock not active\n");
> +			goto pm_put;
> +		}

[Severity: Medium]
What happens if regmap_read() fails here, or if MDF_CKG_ACTIVE is not set?
The return value of regmap_read() is not checked, so a read failure tests
an uninitialized val on the stack.

In addition, jumping to pm_put when MDF_CKG_ACTIVE is not set leaves
MDF_CKG_CKGDEN set in hardware and in REGCACHE_FLAT, so subsequent
non-forced regmap_set_bits() calls will skip writing MDF_CKG_CKGDEN.

> +
> +		if (!priv->exclusive_flg) {
> +			clk_rate_exclusive_get(priv->kclk);
> +			priv->exclusive_flg = true;
> +		}
> +	}
> +
> +	return 0;
> +
> +pm_put:
> +	pm_runtime_put_sync(dev);
> +err:
> +	atomic_dec(&priv->n_active_ch);
> +
> +	return ret;
> +}
> +EXPORT_SYMBOL_GPL(stm32_mdf_core_start_mdf);
> +
> +/**
> + * stm32_mdf_core_trigger() - Trigger MDF filters
> + * @mdf: MDF instance
> + *
> + * Trigger MDF filters via TRGO trigger.
> + *
> + * Return: 0 on success, or a negative error code on failure.
> + */
> +int stm32_mdf_core_trigger(struct stm32_mdf *mdf)
> +{
> +	struct stm32_mdf_priv *priv = to_stm32_mdf_priv(mdf);
> +	int ret;
> +	u32 val;
> +
> +	guard(spinlock)(&priv->lock);
> +
> +	/* Bypass cache to access TRGO volatile bit */
> +	regcache_cache_bypass(priv->regmap, true);
> +
> +	ret = regmap_read(priv->regmap, MDF_GCR_REG, &val);
> +	if (ret < 0)
> +		goto err;
> +
> +	if (val & MDF_GCR_TRGO) {
> +		ret = -EBUSY;
> +		goto err;
> +	}
> +
> +	ret = regmap_set_bits(priv->regmap, MDF_GCR_REG, MDF_GCR_TRGO);
> +
> +err:
> +	regcache_cache_bypass(priv->regmap, false);
> +
> +	return ret;
> +}

[Severity: Medium]
Can regcache_cache_bypass() in stm32_mdf_core_start_mdf() race with
stm32_mdf_core_trigger()?

stm32_mdf_core_trigger() takes priv->lock before toggling map-wide
cache_bypass on priv->regmap, but stm32_mdf_core_start_mdf() and
stm32_mdf_core_stop_mdf() access priv->regmap and toggle cache_bypass
without holding priv->lock:

CPU0                                  CPU1
stm32_mdf_core_trigger()              stm32_mdf_core_start_mdf()
  guard(spinlock)(&priv->lock)
  regcache_cache_bypass(..., true)
  regmap_read(..., MDF_GCR_REG, ...)
                                        regcache_cache_bypass(..., true)
                                        regmap_read(..., MDF_CKGCR_REG, ...)
                                        regcache_cache_bypass(..., false)
  regmap_set_bits(..., MDF_GCR_TRGO)
    <-- cached in REGCACHE_FLAT

With MDF_GCR_TRGO cached in REGCACHE_FLAT, later regcache_sync() calls on
resume will re-trigger the filters.

[Severity: Low]
This isn't a bug, but stm32_mdf_core_trigger() mixes guard(spinlock) with
goto err error handling in the same function.

> +EXPORT_SYMBOL_GPL(stm32_mdf_core_trigger);
> +
> +/**
> + * stm32_mdf_core_stop_mdf() - Stop an MDF channel
> + * @mdf: MDF instance
> + *
> + * Drop one active-channel reference. When the last channel stops, release the
> + * exclusive kernel clock rate and disable the MDF clock.
> + *
> + * Return: 0 on success, or a negative error code on failure.
> + */
> +int stm32_mdf_core_stop_mdf(struct stm32_mdf *mdf)
> +{
> +	struct stm32_mdf_priv *priv = to_stm32_mdf_priv(mdf);
> +	int ret = 0;
> +
> +	if (atomic_dec_and_test(&priv->n_active_ch)) {
> +		if (priv->exclusive_flg) {
> +			clk_rate_exclusive_put(priv->kclk);
> +			priv->exclusive_flg = false;
> +		}
> +
> +		ret = regmap_clear_bits(priv->regmap, MDF_CKGCR_REG, MDF_CKG_CKGDEN);
> +
> +		pm_runtime_put_sync(&priv->pdev->dev);
> +	}
> +
> +	return ret;
> +}

[Severity: Medium]
Can concurrent calls to stm32_mdf_core_start_mdf() and
stm32_mdf_core_stop_mdf() race on priv->n_active_ch?

Only the atomic counter update is serialized, not the hardware startup and
teardown sequences gated by the 0->1 and 1->0 transitions:

CPU0                                  CPU1
stm32_mdf_core_start_mdf()            stm32_mdf_core_start_mdf()
  atomic_inc_return() == 1
                                        atomic_inc_return() == 2
                                        return 0  <-- clocks not enabled yet
  pm_runtime_resume_and_get()
  ...

A similar race occurs when one channel calls stm32_mdf_core_stop_mdf()
(decrementing n_active_ch from 1 to 0) concurrently with another channel
calling stm32_mdf_core_start_mdf() (incrementing n_active_ch from 0 to 1),
allowing stm32_mdf_core_stop_mdf() to clear MDF_CKG_CKGDEN and call
pm_runtime_put_sync() after the new channel has started.  Using a mutex to
serialize start and stop would avoid both races.

[ ... ]
> +static int stm32_mdf_core_parse_of(struct platform_device *pdev, struct stm32_mdf_priv *priv)
> +{
> +	struct device_node *node = pdev->dev.of_node;
> +	struct reset_control *rst;
> +	struct device *dev = &pdev->dev;
> +	u32 count;
> +	int ret;
> +
> +	if (!node)
> +		return -EINVAL;
> +
> +	rst = devm_reset_control_get_optional_exclusive(&pdev->dev, NULL);
> +	if (IS_ERR(rst))
> +		return dev_err_probe(dev, PTR_ERR(rst), "Failed to get reset controller\n");
> +
> +	ret = reset_control_reset(rst);
> +	if (ret)
> +		return dev_err_probe(dev, ret, "Failed to reset MDF\n");

[Severity: High]
Does reset_control_reset() fail with -ENOTSUPP on STM32MP25 when resets is
present in the device tree?

The STM32 RCC reset controller driver only implements .assert, .deassert,
and .status:

drivers/clk/stm32/reset-stm32.c:
static const struct reset_control_ops stm32_reset_ops = {
	.assert		= stm32_reset_assert,
	.deassert	= stm32_reset_deassert,
	.status		= stm32_reset_status,
};

Because .reset is NULL, reset_control_reset() returns -ENOTSUPP and aborts
probe.

In addition, stm32_mdf_core_probe() calls stm32_mdf_regmap_init() and
stm32_mdf_core_identification() before stm32_mdf_core_parse_of() resets
the hardware.  Reading identification registers and populating
REGCACHE_FLAT before deasserting reset reads registers while the block may
still be in reset, and resetting afterward clears hardware registers behind
the initialized cache.

[ ... ]
> +static int stm32_mdf_core_identification(struct platform_device *pdev, struct stm32_mdf_priv *priv)
> +{
> +	struct stm32_mdf *mdf = &priv->mdf;
> +	u32 val;
> +	int ret;
> +
> +	/* If filter number is explicitly defined, don't check identification registers */
> +	mdf->nbf = ((unsigned long)device_get_match_data(&pdev->dev));
> +	if (mdf->nbf)
> +		return 0;
> +
> +	ret = regmap_read(priv->regmap, MDF_IPIDR_REG, &val);
> +	if (ret)
> +		return ret;
> +
> +	if (val == STM32MP25_MDF_IPIDR_NUMBER) {
> +		ret = regmap_read(priv->regmap, MDF_HWCFGR_REG, &val);
> +		if (ret)
> +			return ret;
> +
> +		mdf->nbf = FIELD_GET(MDF_HWCFGR_NBF_MASK, val);
> +
> +		ret = regmap_read(priv->regmap, MDF_VERR_REG, &val);
> +
> +		dev_dbg(&pdev->dev, "MDF version: %lu.%lu\n", FIELD_GET(MDF_VERR_MAJREV_MASK, val),
> +			FIELD_GET(MDF_VERR_MINREV_MASK, val));

[Severity: Low]
Should ret be checked after reading MDF_VERR_REG?  Right now ret is
assigned and ignored, so a failed read logs stale bits from MDF_HWCFGR_REG
and returns 0.

[ ... ]
> +static int stm32_mdf_core_probe(struct platform_device *pdev)
> +{
> +	struct stm32_mdf_priv *priv;
> +	struct resource *res;
> +	int ret;
> +
> +	priv = devm_kzalloc(&pdev->dev, sizeof(*priv), GFP_KERNEL);
> +	if (!priv)
> +		return -ENOMEM;
> +	priv->pdev = pdev;
> +	spin_lock_init(&priv->lock);
> +
> +	priv->base = devm_platform_get_and_ioremap_resource(pdev, 0, &res);
> +	if (IS_ERR(priv->base))
> +		return PTR_ERR(priv->base);
> +	priv->phys_base = res->start;
> +
> +	priv->kclk = devm_clk_get(&pdev->dev, NULL);
> +	if (IS_ERR(priv->kclk))
> +		return dev_err_probe(&pdev->dev, PTR_ERR(priv->kclk),
> +				     "Failed to get kernel clock\n");
> +	priv->mdf.kclk = priv->kclk;
> +
> +	priv->regmap = stm32_mdf_regmap_init(&pdev->dev, priv->kclk, priv->base,
> +					     &stm32_mdf_regmap_cfg);

[Severity: Medium]
Does stm32_mdf_regmap_init() read past the end of the ioremapped resource
0 region?

In the DT binding (Documentation/devicetree/bindings/iio/adc/st,stm32mp25-mdf.yaml),
resource 0 is only 8 bytes (reg = <0x504d0000 0x8>, <0x504d0ff0 0x10>),
while the identification registers at 0xff0..0xffc are in resource 1.
stm32_mdf_core_probe() only maps resource 0, yet stm32_mdf_regmap_cfg sets
max_register = MDF_SIDR_REG (0xffc) with REGCACHE_FLAT, causing both
regcache_hw_init() and stm32_mdf_core_identification() to read offsets
0xff0..0xffc from priv->base without requesting or mapping resource 1.

[ ... ]
> +	INIT_LIST_HEAD(&priv->mdf.filter_list);
> +
> +	platform_set_drvdata(pdev, priv);

[Severity: Low]
Should platform_set_drvdata() pass &priv->mdf instead of priv?
stm32_mdf_core_suspend(), stm32_mdf_core_resume(), and child filter
drivers treat dev_get_drvdata() as a struct stm32_mdf * pointer and pass
it to to_stm32_mdf_priv(), which only works today because mdf happens to
be the first member of struct stm32_mdf_priv.

[ ... ]
> +static int stm32_mdf_core_suspend(struct device *dev)
> +{
> +	struct stm32_mdf *mdf = dev_get_drvdata(dev);
> +	struct stm32_mdf_priv *priv = to_stm32_mdf_priv(mdf);
> +	int ret;
> +
> +	ret = pm_runtime_force_suspend(dev);
> +	if (ret)
> +		return ret;
> +
> +	regcache_cache_only(priv->regmap, true);
> +	regcache_mark_dirty(priv->regmap);
> +
> +	/* Balance regmap_mmio_attach_clk() clk_prepare() */
> +	clk_unprepare(priv->kclk);
> +
> +	return pinctrl_pm_select_sleep_state(dev);
> +}
> +
> +static int stm32_mdf_core_resume(struct device *dev)
> +{
> +	struct stm32_mdf *mdf = dev_get_drvdata(dev);
> +	struct stm32_mdf_priv *priv = to_stm32_mdf_priv(mdf);
> +	int ret;
> +
> +	ret = pinctrl_pm_select_default_state(dev);
> +	if (ret) {
> +		dev_err(dev, "Failed to set pins default state: %d\n", ret);
> +		return ret;
> +	}
> +
> +	ret = clk_prepare(priv->kclk);
> +	if (ret) {
> +		dev_err(dev, "Failed to prepare kernel clock: %d\n", ret);
> +		goto err_clk;
> +	}
> +
> +	regcache_cache_only(priv->regmap, false);
> +	ret = regcache_sync(priv->regmap);
> +	if (ret) {
> +		dev_err(dev, "Failed to sync cache: %d\n", ret);
> +		goto err_cache;
> +	}
> +
> +	return pm_runtime_force_resume(dev);
> +
> +err_cache:
> +	clk_unprepare(priv->kclk);
> +err_clk:
> +	pinctrl_pm_select_sleep_state(dev);
> +
> +	return ret;
> +}

[Severity: Medium]
If pinctrl_pm_select_sleep_state() fails in stm32_mdf_core_suspend(), does
the device stay in cache-only mode with priv->kclk unprepared?  When
.suspend returns an error, .resume is not called, so failing without
re-preparing priv->kclk and restoring regcache_cache_only(false) and
pm_runtime_force_resume() leaves the clock prepare count unbalanced on
teardown.

Similarly, in stm32_mdf_core_resume(), if regcache_sync() fails, err_cache
unprepares priv->kclk while leaving regcache_cache_only(priv->regmap,
false) set.

> +
> +static int stm32_mdf_core_runtime_suspend(struct device *dev)
> +{
> +	return 0;
> +}
> +
> +static int stm32_mdf_core_runtime_resume(struct device *dev)
> +{
> +	return 0;
> +}
> +
> +static const struct dev_pm_ops stm32_mdf_core_pm_ops = {
> +	SET_SYSTEM_SLEEP_PM_OPS(stm32_mdf_core_suspend, stm32_mdf_core_resume)
> +	SET_RUNTIME_PM_OPS(stm32_mdf_core_runtime_suspend, stm32_mdf_core_runtime_resume, NULL)
> +};
> +
> +static struct platform_driver stm32_mdf_driver = {
> +	.probe = stm32_mdf_core_probe,
> +	.remove = stm32_mdf_core_remove,
> +	.driver = {
> +		.name = "stm32-mdf",
> +		.of_match_table = stm32_mdf_of_match,
> +		.pm = &stm32_mdf_core_pm_ops,
> +	},
> +};

[Severity: Low]
Building with CONFIG_PM_SLEEP=n or CONFIG_PM=n (for example under
COMPILE_TEST) will warn about unused static PM callbacks because
SET_SYSTEM_SLEEP_PM_OPS() and SET_RUNTIME_PM_OPS() compile out the
references.  Using SYSTEM_SLEEP_PM_OPS(), RUNTIME_PM_OPS(), and
pm_ptr(&stm32_mdf_core_pm_ops) allows the compiler to drop them cleanly
when PM is disabled.

> diff --git a/drivers/iio/adc/stm32-mdf.h b/drivers/iio/adc/stm32-mdf.h
> new file mode 100644
> index 0000000000000..b334769f8b175
> --- /dev/null
> +++ b/drivers/iio/adc/stm32-mdf.h
[ ... ]
> +/* MDF_OR: Option Register */
> +#define MDF_OR_OPTION_MASK OPTION(31, 0)

[Severity: Low]
Should OPTION(31, 0) be GENMASK(31, 0)?  OPTION() is not defined anywhere,
so any use of MDF_OR_OPTION_MASK will fail to build.

[ ... ]
> +struct stm32_mdf {
> +	struct list_head filter_list;
> +	struct clk *kclk;
> +	unsigned long fproc;
> +	unsigned int nbf;
> +	unsigned int nb_interleave;
> +};

[Severity: Low]
stm32-mdf.h has no #ifndef/#define/#endif include guards, so including it
more than once in a translation unit will cause redefinition errors for
struct stm32_mdf and stm32_mdf_regmap_init().

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261118-upstream_mdf_v2-v2-0-7939d3c3720a@foss.st.com?part=2

  reply	other threads:[~2026-11-18 15:45 UTC|newest]

Thread overview: 22+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-18 15:17 [PATCH v2 00/11] iio: adc: stm32: add mdf support for stm32mp2 Olivier Moysan
2026-11-18 15:17 ` [PATCH v2 01/11] dt-bindings: iio: adc: add stm32mp2 mdf filter Olivier Moysan
2026-11-18 15:17 ` [PATCH v2 02/11] iio: adc: stm32-mdf: add core registration Olivier Moysan
2026-11-18 15:45   ` sashiko-bot [this message]
2026-11-18 15:17 ` [PATCH v2 03/11] iio: adc: stm32-mdf: add cck clocks provider Olivier Moysan
2026-11-18 15:50   ` sashiko-bot
2026-11-18 15:17 ` [PATCH v2 04/11] iio: adc: stm32-mdf: add serial interfaces support Olivier Moysan
2026-11-18 15:42   ` sashiko-bot
2026-11-18 15:17 ` [PATCH v2 05/11] iio: adc: stm32-mdf: add digital filter device support Olivier Moysan
2026-11-18 15:40   ` sashiko-bot
2026-11-18 15:17 ` [PATCH v2 06/11] iio: adc: stm32-mdf: configure cic filters and sample rates Olivier Moysan
2026-11-18 15:34   ` sashiko-bot
2026-11-18 15:17 ` [PATCH v2 07/11] iio: adc: stm32-mdf: add direct iio conversion support Olivier Moysan
2026-11-18 15:51   ` sashiko-bot
2026-11-18 15:17 ` [PATCH v2 08/11] iio: adc: stm32-mdf: add dma buffered acquisition support Olivier Moysan
2026-11-18 15:55   ` sashiko-bot
2026-11-18 15:17 ` [PATCH v2 09/11] iio: adc: stm32-mdf: add external timer trigger support Olivier Moysan
2026-11-18 15:37   ` sashiko-bot
2026-11-18 15:17 ` [PATCH v2 10/11] iio: adc: stm32-mdf: add dmic audio filter support Olivier Moysan
2026-11-18 16:06   ` sashiko-bot
2026-11-18 15:17 ` [PATCH v2 11/11] ASoC: stm32: add mdf dai support Olivier Moysan
2026-11-18 16:08   ` sashiko-bot

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-221370@kernel.org \
    --to=sashiko-bot@kernel.org \
    --cc=conor+dt@kernel.org \
    --cc=devicetree@vger.kernel.org \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=linux-sound@vger.kernel.org \
    --cc=media-ci@linuxtv.org \
    --cc=olivier.moysan@foss.st.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