ASoC: rockchip: Reorder clock enable sequence
Subsystems (4)
v2 cover ASoC: rockchip: Reorder clock enable sequence
From: bui duc phuc <[email protected]>
Hi all,
This series reorders the runtime resume clock enable sequence in the Rockchip SPDIF and PDM drivers to enable the bus clock before the functional controller clock.
It also updates the SPDIF DT binding clock descriptions to match the actual clock usage in the driver.
Additionally, this v2 adds two new patches addressing issues reported by the Sashiko AI Review tool regarding regcache sync failure handling and runtime PM resume status validation.
Testing:
- Patch 1: Verified (dt_binding_check passed).
- Patches 2 to 5: Compile tested only. Please help test if you have the relevant Rockchip hardware.
Changes in v2:
- Include two new patches to handle runtime PM resume and regcache sync failures based on Sashiko AI Review.
- Update commit message based on Krzysztof's review
- Clarify in the commit message that the resume sequence becomes the reverse of the suspend sequence.
Best Regards,
Phuc
bui duc phuc (5):
ASoC: dt-bindings: rockchip-spdif: Correct SPDIF clock descriptions
ASoC: rockchip: spdif: Reorder clock enable sequence
ASoC: rockchip: rockchip_pdm: Reorder clock enable sequence
ASoC: rockchip: spdif: Restore regcache cache-only mode on sync
failure
ASoC: rockchip: rockchip_pdm: Handle runtime PM resume failures in
set_fmt
.../bindings/sound/rockchip-spdif.yaml | 2 +- sound/soc/rockchip/rockchip_pdm.c | 16 ++++++++++------ sound/soc/rockchip/rockchip_spdif.c | 11 ++++++----- 3 files changed, 17 insertions(+), 12 deletions(-)
Signature · 1 line
2.43.0
PATCH v2 1/5 ASoC: dt-bindings: rockchip-spdif: Correct SPDIF clock descriptions
From: bui duc phuc <[email protected]>
Update the binding descriptions to match the actual clock usage, where 'mclk' is the controller clock and 'hclk' is the bus clock.
---
Changes in v2:
- Update commit message based on Krzysztof's review
Documentation/devicetree/bindings/sound/rockchip-spdif.yaml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Documentation/devicetree/bindings/sound/rockchip-spdif.yaml +1 -1
@@ -45,8 +45,8 @@ properties: clocks: items: - - description: clock for SPDIF bus - description: clock for SPDIF controller + - description: clock for SPDIF bus clock-names: items:
Signature · 1 line
2.43.0
Re: [PATCH v2 1/5] ASoC: dt-bindings: rockchip-spdif: Correct SPDIF clock descriptions
Quoted 5 lines · On Tue, 02 Jun 2026 17:16:04 +0700, [email protected] wrote:
From: bui duc phuc <[email protected]>
Update the binding descriptions to match the actual clock usage, where 'mclk' is the controller clock and 'hclk' is the bus clock.
Quoted 9 lines
Signed-off-by bui duc phuc <[email protected]>---
Changes in v2:
- Update commit message based on Krzysztof's reviewDocumentation/devicetree/bindings/sound/rockchip-spdif.yaml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
PATCH v2 2/5 ASoC: rockchip: spdif: Reorder clock enable sequence
In reply to [email protected]From: bui duc phuc <[email protected]>
Enable the 'hclk' bus clock before the 'mclk' controller clock during runtime resume.
The bus clock provides the register access interface, so enable it before the controller clock. This also makes the resume sequence the reverse of the suspend sequence, which keeps the clock ordering consistent.
---
NOTE: This patch is compile-tested only. Please help test if you have
the relevant Rockchip hardware.
Changes in v2:
- Clarify in the commit message that the resume sequence becomes the reverse of the suspend sequence.
sound/soc/rockchip/rockchip_spdif.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-)
sound/soc/rockchip/rockchip_spdif.c +5 -5
@@ -76,16 +76,16 @@ static int rk_spdif_runtime_resume(struct device *dev)struct rk_spdif_dev *spdif =dev_get_drvdata (dev);int ret; - ret =clk_prepare_enable (spdif->mclk); + ret =clk_prepare_enable (spdif->hclk);if (ret) { -dev_err (spdif->dev,"mclk clock enable failed %d\n" , ret); +dev_err (spdif->dev,"hclk clock enable failed %d\n" , ret);return ret; } - ret =clk_prepare_enable (spdif->hclk); + ret =clk_prepare_enable (spdif->mclk);if (ret) { -clk_disable_unprepare (spdif->mclk); -dev_err (spdif->dev,"hclk clock enable failed %d\n" , ret); +clk_disable_unprepare (spdif->hclk); +dev_err (spdif->dev,"mclk clock enable failed %d\n" , ret);return ret; }
Signature · 1 line
2.43.0
[email protected]
AI reviewRe: [PATCH v2 2/5] ASoC: rockchip: spdif: Reorder clock enable sequence
Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
Pre-existing issues:
- [High] Reading channel status bytes via a u16 pointer causes unaligned memory accesses and endianness bugs.
- [High] The MCLK domain reset bit is asserted but never de-asserted, which poisons the regmap cache and causes unintended resets on resume. --
commit d98bc7bb278e4f820ed1d93795c2185544cad12b
Author: bui duc phuc <[email protected]>
ASoC: rockchip: spdif: Reorder clock enable sequence
This commit reorders the clock enable sequence during runtime resume for the rockchip spdif driver. It enables the hclk bus clock before the mclk controller clock, keeping the clock ordering consistent with the suspend sequence.
Quoted 5 lines
sound/soc/rockchip/rockchip_spdif.c +0
-0@@ -76,16 +76,16 @@ static int rk_spdif_runtime_resume(struct device *dev)
[ ... ]
Quoted 4 lines
sound/soc/rockchip/rockchip_spdif.c +1
-2- ret =clk_prepare_enable (spdif->hclk); + ret =clk_prepare_enable (spdif->mclk);if (ret) { -clk_disable_unprepare (spdif->mclk);
sound/soc/rockchip/rockchip_spdif.c static int rk_spdif_runtime_resume(struct device *dev)
Quoted 10 lines
sound/soc/rockchip/rockchip_spdif.c +2
-1-dev_err (spdif->dev,"hclk clock enable failed %d\n" , ret); +clk_disable_unprepare (spdif->hclk); +dev_err (spdif->dev,"mclk clock enable failed %d\n" , ret);return ret; }regcache_cache_only (spdif->regmap, false);regcache_mark_dirty (spdif->regmap); ret =regcache_sync (spdif->regmap);
[Severity: High]
This is a pre-existing issue, but does this regcache_sync() call cause unintended resets due to a poisoned regmap cache?
In rk_spdif_hw_params(), the driver sets the SPDIF_CFGR_CLR_EN bit to clear the MCLK domain logic:
sound/soc/rockchip/rockchip_spdif.c:rk_spdif_hw_params()
... regmap_update_bits(spdif->regmap, SPDIF_CFGR, SPDIF_CFGR_CLR_MASK, SPDIF_CFGR_CLR_EN); udelay(1); ret = regmap_update_bits(spdif->regmap, SPDIF_CFGR, ...
This bit is never explicitly de-asserted.
Because SPDIF_CFGR is a cached register, the cache permanently stores the value with the clear bit set.
During pm_runtime_resume(), does writing this cached value back to the hardware unintentionally re-trigger the MCLK domain reset and corrupt the hardware state?
Also, looking elsewhere in the same file, there is another pre-existing issue in rk_spdif_hw_params():
sound/soc/rockchip/rockchip_spdif.c:rk_spdif_hw_params()
... ret = snd_pcm_create_iec958_consumer_hw_params(params, cs, sizeof(cs)); if (ret < 0) return ret; fc = (u16 *)cs; for (i = 0; i < CS_BYTE / 2; i++) regmap_write(spdif->regmap, SPDIF_CHNSRn(i), CS_FRAME(fc[i])); ...
[Severity: High]
This isn't a bug introduced by this patch, but does casting the u8 array to a u16 pointer cause unaligned memory accesses and endianness bugs?
The 6-byte cs array is cast to a u16 pointer and directly dereferenced.
Can this trigger alignment faults or undefined behavior depending on the architecture and compiler optimizations?
Additionally, since it performs a host-endian read of the byte stream, does this cause the IEC958 channel status bits to be incorrectly swapped when running on big-endian architectures?
Could get_unaligned_le16() be used here instead?
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=2
PATCH v2 3/5 ASoC: rockchip: rockchip_pdm: Reorder clock enable sequence
In reply to [email protected]From: bui duc phuc <[email protected]>
Enable the 'hclk' bus clock before the 'clk' controller clock during runtime resume.
The bus clock provides the register access interface, so enable it before the controller clock. This also makes the resume sequence the reverse of the suspend sequence, which keeps the clock ordering consistent.
---
NOTE: This patch is compile-tested only. Please help test if you have
the relevant Rockchip hardware.
Changes in v2:
- Clarify in the commit message that the resume sequence becomes the reverse of the suspend sequence.
sound/soc/rockchip/rockchip_pdm.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-)
sound/soc/rockchip/rockchip_pdm.c +5 -5
@@ -422,16 +422,16 @@ static int rockchip_pdm_runtime_resume(struct device *dev)struct rk_pdm_dev *pdm =dev_get_drvdata (dev);int ret; - ret =clk_prepare_enable (pdm->clk); + ret =clk_prepare_enable (pdm->hclk);if (ret) { -dev_err (pdm->dev,"clock enable failed %d\n" , ret); +dev_err (pdm->dev,"hclock enable failed %d\n" , ret);return ret; } - ret =clk_prepare_enable (pdm->hclk); + ret =clk_prepare_enable (pdm->clk);if (ret) { -clk_disable_unprepare (pdm->clk); -dev_err (pdm->dev,"hclock enable failed %d\n" , ret); +clk_disable_unprepare (pdm->hclk); +dev_err (pdm->dev,"clock enable failed %d\n" , ret);return ret; }
Signature · 1 line
2.43.0
[email protected]
AI reviewRe: [PATCH v2 3/5] ASoC: rockchip: rockchip_pdm: Reorder clock enable sequence
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
Pre-existing issues:
- [Critical] If get_pdm_clk() falls back to the signoff clock rate, it leaves clk_out at 0, resulting in a division by zero panic in rockchip_pdm_hw_params(). --
commit d4ba65106b95a91e9e92f74421222069efac6183
Author: bui duc phuc <[email protected]>
ASoC: rockchip: rockchip_pdm: Reorder clock enable sequence
Enable the 'hclk' bus clock before the 'clk' controller clock during runtime resume. This ensures the bus clock is available for register access before the controller is clocked, keeping the resume sequence symmetrical with suspend.
sound/soc/rockchip/rockchip_pdm.c
Quoted 4 lines
sound/soc/rockchip/rockchip_pdm.c +0
-0
[Severity: Critical]
This is a pre-existing issue, but I noticed a potential division by zero in this file when the clock tree cannot synthesize an exact frequency match.
In get_pdm_clk(), if the code falls back to the signoff clock rate, it updates *clk_src but appears to leave *clk_out at 0:
sound/soc/rockchip/rockchip_pdm.c:get_pdm_clk() {
...
if (!clk) {
clk = clk_round_rate(pdm->clk, PDM_SIGNOFF_CLK_RATE);
*clk_src = clk;
}...
}
This 0 value is then passed into rational_best_approximation() as the given_numerator, which evaluates it and sets m = 0. Later in rockchip_pdm_hw_params(), the driver divides by m:
sound/soc/rockchip/rockchip_pdm.c:rockchip_pdm_hw_params() {
...
rational_best_approximation(clk_out, clk_src, GENMASK(16 - 1, 0), GENMASK(16 - 1, 0), &m, &n); ... clk_div = n / m;
...
}
Could an unprivileged userspace application requesting a standard audio format trigger this division by zero and panic the kernel? Should *clk_out be updated in the fallback path, or should rockchip_pdm_hw_params() check for m == 0 before dividing?
[ ... ]
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=3
PATCH v2 4/5 ASoC: rockchip: spdif: Restore regcache cache-only mode on sync failure
In reply to [email protected]From: bui duc phuc <[email protected]>
If regcache_sync() fails during runtime resume, the driver disables the clocks and returns an error. However, the regmap cache-only mode is left disabled.
Restore cache-only mode in the error path so subsequent register accesses continue to use the cache while the device is inactive.
---
NOTE: This patch is compile-tested only. Please help test if you have
the relevant Rockchip hardware.
sound/soc/rockchip/rockchip_spdif.c | 1 + 1 file changed, 1 insertion(+)
sound/soc/rockchip/rockchip_spdif.c +1 -0
@@ -94,6 +94,7 @@ static int rk_spdif_runtime_resume(struct device *dev) ret =regcache_sync (spdif->regmap);if (ret) { +regcache_cache_only (spdif->regmap, true);clk_disable_unprepare (spdif->mclk);clk_disable_unprepare (spdif->hclk); }
Signature · 1 line
2.43.0
[email protected]
AI reviewRe: [PATCH v2 4/5] ASoC: rockchip: spdif: Restore regcache cache-only mode on sync failure
Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
Pre-existing issues:
- [High] The error paths in i2s_runtime_resume, i2s_tdm_runtime_resume, and rockchip_sai_runtime_resume fail to restore regcache_cache_only(..., true) when regcache_sync() fails.
- [High] The regmap is initialized with cache_only = false but the device is placed in RPM_SUSPENDED state at probe completion without setting cache_only = true. --
commit 92568295e5eeb9d7ffb897ac4754fcc60c989c10
Author: bui duc phuc <[email protected]>
ASoC: rockchip: spdif: Restore regcache cache-only mode on sync failure
If regcache_sync() fails during runtime resume, the driver disables the clocks and returns an error. This restores cache-only mode in the error path so subsequent register accesses continue to use the cache while the device is inactive.
sound/soc/rockchip/rockchip_spdif.c static int rk_spdif_runtime_resume(struct device *dev) · line 100
Quoted 11 lines
sound/soc/rockchip/rockchip_spdif.c +1
-0@@ -94,6 +94,7 @@ static int rk_spdif_runtime_resume(struct device *dev) ret =regcache_sync (spdif->regmap);if (ret) { +regcache_cache_only (spdif->regmap, true);clk_disable_unprepare (spdif->mclk);clk_disable_unprepare (spdif->hclk); }
[Severity: High]
This isn't a bug introduced by this patch, but while this change fixes the issue in rockchip_spdif.c, the exact same error path exists in other Rockchip ASoC drivers.
For example, in sound/soc/rockchip/rockchip_i2s.c:i2s_runtime_resume():
ret = regcache_sync(i2s->regmap);
if (ret)
clk_disable_unprepare(i2s->mclk);Because the error path drops the device back to the suspended state without restoring cache-only mode, subsequent regmap accesses (like reading from debugfs /sys/kernel/debug/regmap/) will bypass the -EBUSY check in _regmap_read() and attempt to read directly from memory-mapped registers. Accessing registers in a powered-off domain causes the AHB bus to hang, resulting in an imprecise external abort and kernel panic.
Similar paths exist in rockchip_i2s_tdm.c:i2s_tdm_runtime_resume() and rockchip_sai.c:rockchip_sai_runtime_resume(). Should these drivers be updated as well to fully resolve the issue across the subsystem?
[Severity: High]
This is also a pre-existing issue, but does rk_spdif_probe() leave the device exposed to a similar abort before the first resume?
In sound/soc/rockchip/rockchip_spdif.c:rk_spdif_probe():
devm_pm_runtime_enable(&pdev->dev);
if (!pm_runtime_enabled(&pdev->dev)) {The regmap is initialized with cache_only = false by default. When runtime PM is enabled, the device is placed in the RPM_SUSPENDED state and the hardware power domain can be powered off.
Because config->pm_runtime is not set to true in the regmap configuration, if a regmap read occurs (e.g., via debugfs) before the device is first resumed via ALSA, wouldn't regmap attempt to read physical registers while the power domain is off? Should rk_spdif_probe() explicitly call regcache_cache_only(spdif->regmap, true) before completing?
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=4
PATCH v2 5/5 ASoC: rockchip: rockchip_pdm: Handle runtime PM resume failures in set_fmt
In reply to [email protected]From: bui duc phuc <[email protected]>
rockchip_pdm_set_fmt() calls pm_runtime_get_sync() before accessing hardware registers, but ignores its return value.
If the runtime resume fails, the function continues to perform register accesses while the device state is undefined.
Replace pm_runtime_get_sync() with pm_runtime_resume_and_get() and return early on failure to avoid unpowered register accesses.
---
NOTE: This patch is compile-tested only. Please help test if you have
the relevant Rockchip hardware.
sound/soc/rockchip/rockchip_pdm.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-)
sound/soc/rockchip/rockchip_pdm.c +5 -1
@@ -321,6 +321,7 @@ static int rockchip_pdm_set_fmt(struct snd_soc_dai *cpu_dai, {struct rk_pdm_dev *pdm =to_info (cpu_dai);unsigned int mask =0 , val =0 ; +int ret; mask =PDM_CKP_MSK ;switch (fmt &SND_SOC_DAIFMT_INV_MASK ) {
@@ -334,7 +335,10 @@ static int rockchip_pdm_set_fmt(struct snd_soc_dai *cpu_dai,return -EINVAL ; } -pm_runtime_get_sync (cpu_dai->dev); + ret =pm_runtime_resume_and_get (cpu_dai->dev); +if (ret) +return ret; +regmap_update_bits (pdm->regmap,PDM_CLK_CTRL , mask, val);pm_runtime_put (cpu_dai->dev);
Signature · 1 line
2.43.0
[email protected]
AI reviewRe: [PATCH v2 5/5] ASoC: rockchip: rockchip_pdm: Handle runtime PM resume failures in set_fmt
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
Pre-existing issues:
- [High] Unbalanced clock disable in rockchip_pdm_remove for pdm->clk causes an enable count underflow, leading to a kernel warning. --
commit ecd4c53ebd06077db2d8eab0ef6ff0fdc098c863
Author: bui duc phuc <[email protected]>
ASoC: rockchip: rockchip_pdm: Handle runtime PM resume failures in set_fmt
This patch replaces pm_runtime_get_sync() with pm_runtime_resume_and_get() in rockchip_pdm_set_fmt() and adds an early return on failure. This ensures register accesses are not performed when the device state is undefined.
Quoted 4 lines
sound/soc/rockchip/rockchip_pdm.c +0
-0
[ ... ]
Quoted 9 lines
static void
rockchip_pdm_remove(struct platform_device *pdev){ struct rk_pdm_dev *pdm = dev_get_drvdata(&pdev->dev); pm_runtime_disable(&pdev->dev); if (!pm_runtime_status_suspended(&pdev->dev)) rockchip_pdm_runtime_suspend(&pdev->dev); clk_disable_unprepare(pdm->clk);
[Severity: High]
This is a pre-existing issue, but does this unconditional disable cause an enable count underflow for pdm->clk?
In rockchip_pdm_probe(), pdm->clk is acquired but not enabled. It appears to only be enabled dynamically via rockchip_pdm_runtime_resume(). During rockchip_pdm_remove(), the driver correctly calls rockchip_pdm_runtime_suspend() if the device is active, which disables the clock and drops its reference count to 0.
However, calling clk_disable_unprepare() again immediately after would push the reference count below zero, triggering a Common Clock Framework warning in clk_core_disable(). Could this be repeatedly triggered via sysfs driver unbind to cause a denial of service if panic_on_warn is set?
Quoted 2 lines
clk_disable_unprepare(pdm->hclk); }
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=5
Mark Brown
maintainerQuoted 11 lines · On Tue, 02 Jun 2026 17:16:03 +0700, [email protected] wrote:
ASoC: rockchip: Reorder clock enable sequence
From: bui duc phuc <[email protected]>
Hi all,
This series reorders the runtime resume clock enable sequence in the Rockchip SPDIF and PDM drivers to enable the bus clock before the functional controller clock.
[...]
Applied to
https://git.kernel.org/pub/scm/linux/kernel/git/broonie/sound.git for-7.2
Thanks!
[1/5] ASoC: dt-bindings: rockchip-spdif: Correct SPDIF clock descriptions
https://git.kernel.org/broonie/sound/c/d057cbc218ac [2/5] ASoC: rockchip: spdif: Reorder clock enable sequence https://git.kernel.org/broonie/sound/c/74d3f01a90ca [3/5] ASoC: rockchip: rockchip_pdm: Reorder clock enable sequence https://git.kernel.org/broonie/sound/c/3168721d6ec3 [4/5] ASoC: rockchip: spdif: Restore regcache cache-only mode on sync failure https://git.kernel.org/broonie/sound/c/3546e9aa691a [5/5] ASoC: rockchip: rockchip_pdm: Handle runtime PM resume failures in set_fmt https://git.kernel.org/broonie/sound/c/ee7b5f7b3933
All being well this means that it will be integrated into the linux-next tree (usually sometime in the next 24 hours) and sent to Linus during the next merge window (or sooner if it is a bug fix), however if problems are discovered then the patch may be dropped or reverted.
You may get further e-mails resulting from automated or manual testing and review of the tree, please engage with people reporting problems and send followup patches addressing any issues that are reported if needed.
If any updates are required or you are submitting further changes they should be sent as incremental updates against current git, existing patches will not be replaced.
Please add any relevant lists and maintainers to the CCs when replying to this mail.
Thanks,
Mark