Skip to contentSkip to conversation

drm/bridge: cdns-dsi: Replace deprecated UNIVERSAL_DEV_PM_OPS()

PATCH Superseded Vitor Soares 3 participants 5 messages Apr 7 – Apr 30

Subsystems (4)

v4 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

PATCH v4 1/1 drm/bridge: cdns-dsi: Replace deprecated UNIVERSAL_DEV_PM_OPS()

From: Vitor Soares <[email protected]>

The deprecated UNIVERSAL_DEV_PM_OPS() macro uses the provided callbacks for both runtime PM and system sleep. This causes the DSI clocks to be disabled twice: once during runtime suspend and again during system suspend, resulting in a WARN message from the clock framework when attempting to disable already-disabled clocks.

[   84.384540] clk:231:5 already disabled
[   84.388314] WARNING: CPU: 2 PID: 531 at /drivers/clk/clk.c:1181 clk_core_disable+0xa4/0xac
...
[   84.579183] Call trace:
[   84.581624]  clk_core_disable+0xa4/0xac
[   84.585457]  clk_disable+0x30/0x4c
[   84.588857]  cdns_dsi_suspend+0x20/0x58 [cdns_dsi]
[   84.593651]  pm_generic_suspend+0x2c/0x44
[   84.597661]  ti_sci_pd_suspend+0xbc/0x15c
[   84.601670]  dpm_run_callback+0x8c/0x14c
[   84.605588]  __device_suspend+0x1a0/0x56c
[   84.609594]  dpm_suspend+0x17c/0x21c
[   84.613165]  dpm_suspend_start+0xa0/0xa8
[   84.617083]  suspend_devices_and_enter+0x12c/0x634
[   84.621872]  pm_suspend+0x1fc/0x368

To address this issue, replace UNIVERSAL_DEV_PM_OPS() with SET_RUNTIME_PM_OPS(), enabling suspend/resume handling through the _enable()/_disable() hooks managed by the DRM framework for both runtime and system-wide PM.

Fixes e19233955d9e ("drm/bridge: Add Cadence DSI driver")

---
v3 -> v4
- Add Reviewed-by from Tomi Valkeinen
- Rebase on top of drm-misc-fixes
- Verified issue still present on current mainline

v2 -> v3
- Fix warning: 'cdns_dsi_suspend' defined but not used [-Wunused-function]
- Fix warning: 'cdns_dsi_resume' defined but not used [-Wunused-function]

v1 -> v2
- Rely only on SET_RUNTIME_PM_OPS() for the PM.

 drivers/gpu/drm/bridge/cadence/cdns-dsi-core.c | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)
drivers/gpu/drm/bridge/cadence/cdns-dsi-core.c +6 -5
index 0dd85e26248c..e07a9892df4e 100644
@@ -1230,7 +1230,7 @@ static const struct mipi_dsi_host_ops cdns_dsi_ops = {
 	.transfer = cdns_dsi_transfer,
 };
 
-static int __maybe_unused cdns_dsi_resume(struct device *dev)
+static int cdns_dsi_resume(struct device *dev)
 {
 	struct cdns_dsi *dsi = dev_get_drvdata(dev);
 
@@ -1241,7 +1241,7 @@ static int __maybe_unused cdns_dsi_resume(struct device *dev)
 	return 0;
 }
 
-static int __maybe_unused cdns_dsi_suspend(struct device *dev)
+static int cdns_dsi_suspend(struct device *dev)
 {
 	struct cdns_dsi *dsi = dev_get_drvdata(dev);
 
@@ -1251,8 +1251,9 @@ static int __maybe_unused cdns_dsi_suspend(struct device *dev)
 	return 0;
 }
 
-static UNIVERSAL_DEV_PM_OPS(cdns_dsi_pm_ops, cdns_dsi_suspend, cdns_dsi_resume,
-			    NULL);
+static const struct dev_pm_ops cdns_dsi_pm_ops = {
+	RUNTIME_PM_OPS(cdns_dsi_suspend, cdns_dsi_resume, NULL)
+};
 
 static int cdns_dsi_drm_probe(struct platform_device *pdev)
 {
@@ -1399,7 +1400,7 @@ static struct platform_driver cdns_dsi_platform_driver = {
 	.driver = {
 		.name   = "cdns-dsi",
 		.of_match_table = cdns_dsi_of_match,
-		.pm = &cdns_dsi_pm_ops,
+		.pm = pm_ptr(&cdns_dsi_pm_ops),
 	},
 };
 module_platform_driver(cdns_dsi_platform_driver);
Signature · 1 line
2.53.0
#1 lore raw

+Dmitry, Luca

Hello,

Quoted 29 lines · Vitor Soares

From: Vitor Soares <[email protected]>

The deprecated UNIVERSAL_DEV_PM_OPS() macro uses the provided callbacks for both runtime PM and system sleep. This causes the DSI clocks to be disabled twice: once during runtime suspend and again during system suspend, resulting in a WARN message from the clock framework when attempting to disable already-disabled clocks.

[   84.384540] clk:231:5 already disabled
[   84.388314] WARNING: CPU: 2 PID: 531 at /drivers/clk/clk.c:1181 clk_core_disable+0xa4/0xac
...
[   84.579183] Call trace:
[   84.581624]  clk_core_disable+0xa4/0xac
[   84.585457]  clk_disable+0x30/0x4c
[   84.588857]  cdns_dsi_suspend+0x20/0x58 [cdns_dsi]
[   84.593651]  pm_generic_suspend+0x2c/0x44
[   84.597661]  ti_sci_pd_suspend+0xbc/0x15c
[   84.601670]  dpm_run_callback+0x8c/0x14c
[   84.605588]  __device_suspend+0x1a0/0x56c
[   84.609594]  dpm_suspend+0x17c/0x21c
[   84.613165]  dpm_suspend_start+0xa0/0xa8
[   84.617083]  suspend_devices_and_enter+0x12c/0x634
[   84.621872]  pm_suspend+0x1fc/0x368

To address this issue, replace UNIVERSAL_DEV_PM_OPS() with SET_RUNTIME_PM_OPS(), enabling suspend/resume handling through the _enable()/_disable() hooks managed by the DRM framework for both runtime and system-wide PM.

Quoted 4 lines
Fixes e19233955d9e ("drm/bridge: Add Cadence DSI driver")

Luca, Dmitry: any chance you can help with this patch?

Thanks,
Francesco

reviewer
#2 lore raw
In reply to Vitor Soares
Quoted 24 lines · Vitor Soares

From: Vitor Soares <[email protected]>

The deprecated UNIVERSAL_DEV_PM_OPS() macro uses the provided callbacks for both runtime PM and system sleep. This causes the DSI clocks to be disabled twice: once during runtime suspend and again during system suspend, resulting in a WARN message from the clock framework when attempting to disable already-disabled clocks.

[   84.384540] clk:231:5 already disabled
[   84.388314] WARNING: CPU: 2 PID: 531 at /drivers/clk/clk.c:1181 clk_core_disable+0xa4/0xac
...
[   84.579183] Call trace:
[   84.581624]  clk_core_disable+0xa4/0xac
[   84.585457]  clk_disable+0x30/0x4c
[   84.588857]  cdns_dsi_suspend+0x20/0x58 [cdns_dsi]
[   84.593651]  pm_generic_suspend+0x2c/0x44
[   84.597661]  ti_sci_pd_suspend+0xbc/0x15c
[   84.601670]  dpm_run_callback+0x8c/0x14c
[   84.605588]  __device_suspend+0x1a0/0x56c
[   84.609594]  dpm_suspend+0x17c/0x21c
[   84.613165]  dpm_suspend_start+0xa0/0xa8
[   84.617083]  suspend_devices_and_enter+0x12c/0x634
[   84.621872]  pm_suspend+0x1fc/0x368
Quoted 2 lines

To address this issue, replace UNIVERSAL_DEV_PM_OPS() with SET_RUNTIME_PM_OPS(), enabling suspend/resume handling through the

This is not what the patch does, the patch uses RUNTIME_PM_OPS.

Quoted 46 lines

_enable()/_disable() hooks managed by the DRM framework for both runtime and system-wide PM.

Fixes e19233955d9e ("drm/bridge: Add Cadence DSI driver")

---
v3 -> v4
- Add Reviewed-by from Tomi Valkeinen
- Rebase on top of drm-misc-fixes
- Verified issue still present on current mainline

v2 -> v3
- Fix warning: 'cdns_dsi_suspend' defined but not used [-Wunused-function]
- Fix warning: 'cdns_dsi_resume' defined but not used [-Wunused-function]

v1 -> v2
- Rely only on SET_RUNTIME_PM_OPS() for the PM.

 drivers/gpu/drm/bridge/cadence/cdns-dsi-core.c | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)
drivers/gpu/drm/bridge/cadence/cdns-dsi-core.c +2 -2
index 0dd85e26248c..e07a9892df4e 100644
@@ -1230,7 +1230,7 @@ static const struct mipi_dsi_host_ops cdns_dsi_ops = {
 	.transfer = cdns_dsi_transfer,
 };
 
-static int __maybe_unused cdns_dsi_resume(struct device *dev)
+static int cdns_dsi_resume(struct device *dev)
 {
 	struct cdns_dsi *dsi = dev_get_drvdata(dev);
 
@@ -1241,7 +1241,7 @@ static int __maybe_unused cdns_dsi_resume(struct device *dev)
 	return 0;
 }
 
-static int __maybe_unused cdns_dsi_suspend(struct device *dev)
+static int cdns_dsi_suspend(struct device *dev)
 {
 	struct cdns_dsi *dsi = dev_get_drvdata(dev);
 

drivers/gpu/drm/bridge/cadence/cdns-dsi-core.c static int __maybe_unused cdns_dsi_suspend(struct device *dev) · line 1256

Quoted 8 lines
drivers/gpu/drm/bridge/cadence/cdns-dsi-core.c +3 -2
@@ -1251,8 +1251,9 @@ static int __maybe_unused cdns_dsi_suspend(struct device *dev)
 	return 0;
 }
 
-static UNIVERSAL_DEV_PM_OPS(cdns_dsi_pm_ops, cdns_dsi_suspend, cdns_dsi_resume,
-			    NULL);
+static const struct dev_pm_ops cdns_dsi_pm_ops = {
+	RUNTIME_PM_OPS(cdns_dsi_suspend, cdns_dsi_resume, NULL)
+};

Not an expert here, but the docs [0] suggest using DEFINE_RUNTIME_DEV_PM_OPS(). Is there a good reason to not do so?

[0] https://elixir.bootlin.com/linux/v7.0.1/source/include/linux/pm.h#L455-L456

Luca

--
Luca Ceresoli, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com

#3 lore raw

Hi Luca,

Thanks you for the feedback.

Quoted 26 lines · Luca Ceresoli
Quoted 25 lines · Vitor Soares

From: Vitor Soares <[email protected]>

The deprecated UNIVERSAL_DEV_PM_OPS() macro uses the provided callbacks for both runtime PM and system sleep. This causes the DSI clocks to be disabled twice: once during runtime suspend and again during system suspend, resulting in a WARN message from the clock framework when attempting to disable already-disabled clocks.

[   84.384540] clk:231:5 already disabled
[   84.388314] WARNING: CPU: 2 PID: 531 at /drivers/clk/clk.c:1181 clk_core_disable+0xa4/0xac
...
[   84.579183] Call trace:
[   84.581624]  clk_core_disable+0xa4/0xac
[   84.585457]  clk_disable+0x30/0x4c
[   84.588857]  cdns_dsi_suspend+0x20/0x58 [cdns_dsi]
[   84.593651]  pm_generic_suspend+0x2c/0x44
[   84.597661]  ti_sci_pd_suspend+0xbc/0x15c
[   84.601670]  dpm_run_callback+0x8c/0x14c
[   84.605588]  __device_suspend+0x1a0/0x56c
[   84.609594]  dpm_suspend+0x17c/0x21c
[   84.613165]  dpm_suspend_start+0xa0/0xa8
[   84.617083]  suspend_devices_and_enter+0x12c/0x634
[   84.621872]  pm_suspend+0x1fc/0x368

Quoted 5 lines
Quoted 2 lines

To address this issue, replace UNIVERSAL_DEV_PM_OPS() with SET_RUNTIME_PM_OPS(), enabling suspend/resume handling through the

This is not what the patch does, the patch uses RUNTIME_PM_OPS.

I missed changing it. I will send another version fixing the commit message to RUNTIME_PM_OPS().

Quoted 62 lines
Quoted 48 lines

_enable()/_disable() hooks managed by the DRM framework for both runtime and system-wide PM.

Cc [email protected] # 6.1.x
Fixes e19233955d9e ("drm/bridge: Add Cadence DSI driver")

---
v3 -> v4
 - Add Reviewed-by from Tomi Valkeinen
 - Rebase on top of drm-misc-fixes
 - Verified issue still present on current mainline

v2 -> v3
 - Fix warning: 'cdns_dsi_suspend' defined but not used [-Wunused-function]
 - Fix warning: 'cdns_dsi_resume' defined but not used [-Wunused-function]

v1 -> v2
 - Rely only on SET_RUNTIME_PM_OPS() for the PM.

 drivers/gpu/drm/bridge/cadence/cdns-dsi-core.c | 11 ++++++-----

 1 file changed, 6 insertions(+), 5 deletions(-)
drivers/gpu/drm/bridge/cadence/cdns-dsi-core.c +2 -2
index 0dd85e26248c..e07a9892df4e 100644
@@ -1230,7 +1230,7 @@ static const struct mipi_dsi_host_ops cdns_dsi_ops = {
         .transfer = cdns_dsi_transfer,
  };
 
-static int __maybe_unused cdns_dsi_resume(struct device *dev)
+static int cdns_dsi_resume(struct device *dev)
  {
         struct cdns_dsi *dsi = dev_get_drvdata(dev);
 
@@ -1241,7 +1241,7 @@ static int __maybe_unused cdns_dsi_resume(struct
 device *dev)
         return 0;
  }
 
-static int __maybe_unused cdns_dsi_suspend(struct device *dev)
+static int cdns_dsi_suspend(struct device *dev)
  {
         struct cdns_dsi *dsi = dev_get_drvdata(dev);

drivers/gpu/drm/bridge/cadence/cdns-dsi-core.c static int __maybe_unused cdns_dsi_suspend(struct · line 1258

Quoted 10 lines
drivers/gpu/drm/bridge/cadence/cdns-dsi-core.c +3 -2
@@ -1251,8 +1251,9 @@ static int __maybe_unused cdns_dsi_suspend(struct
 device *dev)
         return 0;
  }
 
-static UNIVERSAL_DEV_PM_OPS(cdns_dsi_pm_ops, cdns_dsi_suspend,
 cdns_dsi_resume,
-                           NULL);
+static const struct dev_pm_ops cdns_dsi_pm_ops = {
+       RUNTIME_PM_OPS(cdns_dsi_suspend, cdns_dsi_resume, NULL)
+};

Not an expert here, but the docs [0] suggest using DEFINE_RUNTIME_DEV_PM_OPS(). Is there a good reason to not do so?

Quoted 5 lines

[0]
https://elixir.bootlin.com/linux/v7.0.1/source/include/linux/pm.h#L455-L456

Luca

In an earlier discussion [0], we concluded that bridges/panels should only deal with runtime PM:

[0]
https://lore.kernel.org/all/[email protected]/

Best regards,
Vitor Soares

reviewer
#4 lore raw

Hello Vitor,

Quoted 10 lines · Vitor Soares
Quoted 10 lines

drivers/gpu/drm/bridge/cadence/cdns-dsi-core.c static int __maybe_unused cdns_dsi_suspend(struct device *dev) · line 1256

Quoted 6 lines

-static UNIVERSAL_DEV_PM_OPS(cdns_dsi_pm_ops, cdns_dsi_suspend, cdns_dsi_resume,
-                           NULL);

drivers/gpu/drm/bridge/cadence/cdns-dsi-core.c +3 -0
+static const struct dev_pm_ops cdns_dsi_pm_ops = {
+       RUNTIME_PM_OPS(cdns_dsi_suspend, cdns_dsi_resume, NULL)
+};

Not an expert here, but the docs [0] suggest using DEFINE_RUNTIME_DEV_PM_OPS(). Is there a good reason to not do so?

Quoted 8 lines
Quoted 5 lines

[0]
https://elixir.bootlin.com/linux/v7.0.1/source/include/linux/pm.h#L455-L456

Luca

In an earlier discussion [0], we concluded that bridges/panels should only deal with runtime PM:

Ah, good. Then please mention this in the commit message in v5. It can help others doing the right thing in future similar cases.

Luca

--
Luca Ceresoli, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com