drm/tegra: gr2d/gr3d: Move pm_runtime_enable to gr*d_probe
cover drm/tegra: gr2d/gr3d: Move pm_runtime_enable to gr*d_probe
The gr*d_remove() has pm_runtime_disable, this indicates it should be paired with pm_runtime_enable in the probe instead of being inside gr*d_runtime_resume().
Ion Agorria (1):
drm/tegra: gr2d/gr3d: Move pm_runtime_enable to gr*d_probe
drivers/gpu/drm/tegra/gr2d.c | 8 ++++---- drivers/gpu/drm/tegra/gr3d.c | 8 ++++---- 2 files changed, 8 insertions(+), 8 deletions(-)
Signature · 1 line
2.51.0
PATCH 1/1 drm/tegra: gr2d/gr3d: Move pm_runtime_enable to gr*d_probe
From: Ion Agorria <[email protected]>
The gr*d_remove() has pm_runtime_disable, this indicates it should be paired with pm_runtime_enable in the probe instead of being inside gr*d_runtime_resume().
--- drivers/gpu/drm/tegra/gr2d.c | 8 ++++---- drivers/gpu/drm/tegra/gr3d.c | 8 ++++---- 2 files changed, 8 insertions(+), 8 deletions(-)
drivers/gpu/drm/tegra/gr2d.c +4 -4
@@ -286,6 +286,10 @@ static int gr2d_probe(struct platform_device *pdev)for (i =0 ; i <ARRAY_SIZE (gr2d_addr_regs); i++)set_bit (gr2d_addr_regs[i], gr2d->addr_regs); +pm_runtime_enable (dev); +pm_runtime_use_autosuspend (dev); +pm_runtime_set_autosuspend_delay (dev,500 ); +return 0 ; }
@@ -367,10 +371,6 @@ static int __maybe_unused gr2d_runtime_resume(struct device *dev) goto disable_clk; } -pm_runtime_enable (dev); -pm_runtime_use_autosuspend (dev); -pm_runtime_set_autosuspend_delay (dev,500 ); -return 0 ; disable_clk:
drivers/gpu/drm/tegra/gr3d.c +4 -4
@@ -517,6 +517,10 @@ static int gr3d_probe(struct platform_device *pdev)for (i =0 ; i <ARRAY_SIZE (gr3d_addr_regs); i++)set_bit (gr3d_addr_regs[i], gr3d->addr_regs); +pm_runtime_enable (&pdev->dev); +pm_runtime_use_autosuspend (&pdev->dev); +pm_runtime_set_autosuspend_delay (&pdev->dev,500 ); +return 0 ; }
@@ -578,10 +582,6 @@ static int __maybe_unused gr3d_runtime_resume(struct device *dev) goto disable_clk; } -pm_runtime_enable (dev); -pm_runtime_use_autosuspend (dev); -pm_runtime_set_autosuspend_delay (dev,500 ); -return 0 ; disable_clk:
Signature · 1 line
2.51.0
Quoted 69 lines · Svyatoslav Ryhel
From: Ion Agorria <[email protected]>
The gr*d_remove() has
pm_runtime_disable, this indicates it should be paired withpm_runtime_enablein the probe instead of being inside gr*d_runtime_resume().Signed-off-by Ion Agorria <[email protected]>Signed-off-by Svyatoslav Ryhel <[email protected]>--- drivers/gpu/drm/tegra/gr2d.c | 8 ++++---- drivers/gpu/drm/tegra/gr3d.c | 8 ++++---- 2 files changed, 8 insertions(+), 8 deletions(-)drivers/gpu/drm/tegra/gr2d.c +4
-4@@ -286,6 +286,10 @@ static int gr2d_probe(struct platform_device *pdev)for (i =0 ; i <ARRAY_SIZE (gr2d_addr_regs); i++)set_bit (gr2d_addr_regs[i], gr2d->addr_regs); +pm_runtime_enable (dev); +pm_runtime_use_autosuspend (dev); +pm_runtime_set_autosuspend_delay (dev,500 ); +return 0 ; }@@ -367,10 +371,6 @@ static int __maybe_unused gr2d_runtime_resume(struct device *dev) goto disable_clk; } -pm_runtime_enable (dev); -pm_runtime_use_autosuspend (dev); -pm_runtime_set_autosuspend_delay (dev,500 ); -return 0 ; disable_clk:drivers/gpu/drm/tegra/gr3d.c +4
-4@@ -517,6 +517,10 @@ static int gr3d_probe(struct platform_device *pdev)for (i =0 ; i <ARRAY_SIZE (gr3d_addr_regs); i++)set_bit (gr3d_addr_regs[i], gr3d->addr_regs); +pm_runtime_enable (&pdev->dev); +pm_runtime_use_autosuspend (&pdev->dev); +pm_runtime_set_autosuspend_delay (&pdev->dev,500 ); +return 0 ; }@@ -578,10 +582,6 @@ static int __maybe_unused gr3d_runtime_resume(struct device *dev) goto disable_clk; } -pm_runtime_enable (dev); -pm_runtime_use_autosuspend (dev); -pm_runtime_set_autosuspend_delay (dev,500 ); -return 0 ; disable_clk:Signature · 1 line
2.51.0
Oof, looks like I had managed to really bungle this with my earlier patch. Thanks for fixing it!
FWIW, I've been working on adding some nightly testing for Host1x/ TegraDRM, so hopefully we'll be able to catch such things easier in the future.
вт, 28 квіт. 2026 р. о 04:53 Mikko Perttunen <[email protected]> пише:
Quoted 66 lines · Svyatoslav Ryhel
Quoted 64 lines
From: Ion Agorria <[email protected]>
The gr*d_remove() has
pm_runtime_disable, this indicates it should be paired withpm_runtime_enablein the probe instead of being inside gr*d_runtime_resume().Signed-off-by Ion Agorria <[email protected]>Signed-off-by Svyatoslav Ryhel <[email protected]>--- drivers/gpu/drm/tegra/gr2d.c | 8 ++++---- drivers/gpu/drm/tegra/gr3d.c | 8 ++++---- 2 files changed, 8 insertions(+), 8 deletions(-)drivers/gpu/drm/tegra/gr2d.c +4
-4@@ -286,6 +286,10 @@ static int gr2d_probe(struct platform_device *pdev)for (i =0 ; i <ARRAY_SIZE (gr2d_addr_regs); i++)set_bit (gr2d_addr_regs[i], gr2d->addr_regs); +pm_runtime_enable (dev); +pm_runtime_use_autosuspend (dev); +pm_runtime_set_autosuspend_delay (dev,500 ); +return 0 ; }@@ -367,10 +371,6 @@ static int __maybe_unused gr2d_runtime_resume(struct device *dev) goto disable_clk; } -pm_runtime_enable (dev); -pm_runtime_use_autosuspend (dev); -pm_runtime_set_autosuspend_delay (dev,500 ); -return 0 ; disable_clk:drivers/gpu/drm/tegra/gr3d.c +4
-4@@ -517,6 +517,10 @@ static int gr3d_probe(struct platform_device *pdev)for (i =0 ; i <ARRAY_SIZE (gr3d_addr_regs); i++)set_bit (gr3d_addr_regs[i], gr3d->addr_regs); +pm_runtime_enable (&pdev->dev); +pm_runtime_use_autosuspend (&pdev->dev); +pm_runtime_set_autosuspend_delay (&pdev->dev,500 ); +return 0 ; }@@ -578,10 +582,6 @@ static int __maybe_unused gr3d_runtime_resume(struct device *dev) goto disable_clk; } -pm_runtime_enable (dev); -pm_runtime_use_autosuspend (dev); -pm_runtime_set_autosuspend_delay (dev,500 ); -return 0 ;
Quoted 9 lines
Quoted 5 lines
drivers/gpu/drm/tegra/gr3d.c +0
-1disable_clk: --2.51.0
Oof, looks like I had managed to really bungle this with my earlier patch. Thanks for fixing it!
Hello Mikko!
Thank you for taking time and looking into this patch. Don't be so harsh to yourself, PM is easy to mess and hard to set properly. This patch does fix gr*d access but it does not resolve the issue itself. PM should be set in the init/exit rather then probe/remove. I have v2 which fixes this and one more minor issue and I will send them later on.
So for now this patch should not be picked.
Best regards,
Svyatoslav R.
Quoted 9 lines
FWIW, I've been working on adding some nightly testing for Host1x/ TegraDRM, so hopefully we'll be able to catch such things easier in the future.
Fixes 62fa0a985e2c ("drm/tegra: Enable runtime PM during probe")Acked-by Mikko Perttunen <[email protected]>
Mikko Perttunen
maintainerQuoted 77 lines · Svyatoslav Ryhel
вт, 28 квіт. 2026 р. о 04:53 Mikko Perttunen <[email protected]> пише:
Quoted 66 lines · Svyatoslav Ryhel
Quoted 64 lines
From: Ion Agorria <[email protected]>
The gr*d_remove() has
pm_runtime_disable, this indicates it should be paired withpm_runtime_enablein the probe instead of being inside gr*d_runtime_resume().Signed-off-by Ion Agorria <[email protected]>Signed-off-by Svyatoslav Ryhel <[email protected]>--- drivers/gpu/drm/tegra/gr2d.c | 8 ++++---- drivers/gpu/drm/tegra/gr3d.c | 8 ++++---- 2 files changed, 8 insertions(+), 8 deletions(-)drivers/gpu/drm/tegra/gr2d.c +4
-4@@ -286,6 +286,10 @@ static int gr2d_probe(struct platform_device *pdev)for (i =0 ; i <ARRAY_SIZE (gr2d_addr_regs); i++)set_bit (gr2d_addr_regs[i], gr2d->addr_regs); +pm_runtime_enable (dev); +pm_runtime_use_autosuspend (dev); +pm_runtime_set_autosuspend_delay (dev,500 ); +return 0 ; }@@ -367,10 +371,6 @@ static int __maybe_unused gr2d_runtime_resume(struct device *dev) goto disable_clk; } -pm_runtime_enable (dev); -pm_runtime_use_autosuspend (dev); -pm_runtime_set_autosuspend_delay (dev,500 ); -return 0 ; disable_clk:drivers/gpu/drm/tegra/gr3d.c +4
-4@@ -517,6 +517,10 @@ static int gr3d_probe(struct platform_device *pdev)for (i =0 ; i <ARRAY_SIZE (gr3d_addr_regs); i++)set_bit (gr3d_addr_regs[i], gr3d->addr_regs); +pm_runtime_enable (&pdev->dev); +pm_runtime_use_autosuspend (&pdev->dev); +pm_runtime_set_autosuspend_delay (&pdev->dev,500 ); +return 0 ; }@@ -578,10 +582,6 @@ static int __maybe_unused gr3d_runtime_resume(struct device *dev) goto disable_clk; } -pm_runtime_enable (dev); -pm_runtime_use_autosuspend (dev); -pm_runtime_set_autosuspend_delay (dev,500 ); -return 0 ;Quoted 9 lines
Quoted 5 lines
drivers/gpu/drm/tegra/gr3d.c +0
-1disable_clk: --2.51.0
Oof, looks like I had managed to really bungle this with my earlier patch. Thanks for fixing it!
Quoted 8 lines
Hello Mikko!
Thank you for taking time and looking into this patch. Don't be so harsh to yourself, PM is easy to mess and hard to set properly. This patch does fix gr*d access but it does not resolve the issue itself. PM should be set in the init/exit rather then probe/remove. I have v2 which fixes this and one more minor issue and I will send them later on.
Thanks! Why do you think it's necessary to enable runtime PM in init? If you look at the commit I referenced below (in 'Fixes'), we've had some issues in the past with doing pm_runtime_enable outside of probe, where the engine's power domain would be left enabled even when it is idle.
gr2d/gr3d I suppose wouldn't be in practice affected by that issue though given they aren't in their own power domains.
Cheers
Mikko
Quoted 15 lines
So for now this patch should not be picked.
Best regards,
Svyatoslav R.Quoted 9 lines
FWIW, I've been working on adding some nightly testing for Host1x/ TegraDRM, so hopefully we'll be able to catch such things easier in the future.
Fixes 62fa0a985e2c ("drm/tegra: Enable runtime PM during probe")Acked-by Mikko Perttunen <[email protected]>
чт, 30 квіт. 2026 р. о 09:23 Mikko Perttunen <[email protected]> пише:
Quoted 88 lines · Svyatoslav Ryhel
Quoted 77 lines
вт, 28 квіт. 2026 р. о 04:53 Mikko Perttunen <[email protected]> пише:
Quoted 66 lines · Svyatoslav Ryhel
Quoted 64 lines
From: Ion Agorria <[email protected]>
The gr*d_remove() has
pm_runtime_disable, this indicates it should be paired withpm_runtime_enablein the probe instead of being inside gr*d_runtime_resume().Signed-off-by Ion Agorria <[email protected]>Signed-off-by Svyatoslav Ryhel <[email protected]>--- drivers/gpu/drm/tegra/gr2d.c | 8 ++++---- drivers/gpu/drm/tegra/gr3d.c | 8 ++++---- 2 files changed, 8 insertions(+), 8 deletions(-)drivers/gpu/drm/tegra/gr2d.c +4
-4@@ -286,6 +286,10 @@ static int gr2d_probe(struct platform_device *pdev)for (i =0 ; i <ARRAY_SIZE (gr2d_addr_regs); i++)set_bit (gr2d_addr_regs[i], gr2d->addr_regs); +pm_runtime_enable (dev); +pm_runtime_use_autosuspend (dev); +pm_runtime_set_autosuspend_delay (dev,500 ); +return 0 ; }@@ -367,10 +371,6 @@ static int __maybe_unused gr2d_runtime_resume(struct device *dev) goto disable_clk; } -pm_runtime_enable (dev); -pm_runtime_use_autosuspend (dev); -pm_runtime_set_autosuspend_delay (dev,500 ); -return 0 ; disable_clk:drivers/gpu/drm/tegra/gr3d.c +4
-4@@ -517,6 +517,10 @@ static int gr3d_probe(struct platform_device *pdev)for (i =0 ; i <ARRAY_SIZE (gr3d_addr_regs); i++)set_bit (gr3d_addr_regs[i], gr3d->addr_regs); +pm_runtime_enable (&pdev->dev); +pm_runtime_use_autosuspend (&pdev->dev); +pm_runtime_set_autosuspend_delay (&pdev->dev,500 ); +return 0 ; }@@ -578,10 +582,6 @@ static int __maybe_unused gr3d_runtime_resume(struct device *dev) goto disable_clk; } -pm_runtime_enable (dev); -pm_runtime_use_autosuspend (dev); -pm_runtime_set_autosuspend_delay (dev,500 ); -return 0 ;Quoted 9 lines
Quoted 5 lines
drivers/gpu/drm/tegra/gr3d.c +0
-1disable_clk: --2.51.0
Oof, looks like I had managed to really bungle this with my earlier patch. Thanks for fixing it!
Quoted 8 lines
Hello Mikko!
Thank you for taking time and looking into this patch. Don't be so harsh to yourself, PM is easy to mess and hard to set properly. This patch does fix gr*d access but it does not resolve the issue itself. PM should be set in the init/exit rather then probe/remove. I have v2 which fixes this and one more minor issue and I will send them later on.
Quoted 8 lines
Thanks! Why do you think it's necessary to enable runtime PM in init? If you look at the commit I referenced below (in 'Fixes'), we've had some issues in the past with doing
pm_runtime_enableoutside of probe, where the engine's power domain would be left enabled even when it is idle.gr2d/gr3d I suppose wouldn't be in practice affected by that issue though given they aren't in their own power domains.
WDYM, gr2d and gr3d have their own power domains.
If the master device is unbound and rebound, gr2d_init() will run again, but pm configuration is only located in this probe function which will not run again, while pm disable are both in exit and remove. This results in pm issue we are observing.
Solution would be either do everyting in probe/remove, or init/exit. Probe/remove will lead to domain being turned on even if engines are idle. Init/exit seems to me more suitable and we have tasted this configuration for a quite while in the grate DRM version.
Best regards,
Svyatoslav R.
Quoted 23 lines
Cheers
MikkoQuoted 15 lines
So for now this patch should not be picked.
Best regards,
Svyatoslav R.Quoted 9 lines
FWIW, I've been working on adding some nightly testing for Host1x/ TegraDRM, so hopefully we'll be able to catch such things easier in the future.
Fixes 62fa0a985e2c ("drm/tegra: Enable runtime PM during probe")Acked-by Mikko Perttunen <[email protected]>
Mikko Perttunen
maintainerQuoted 95 lines · Svyatoslav Ryhel
чт, 30 квіт. 2026 р. о 09:23 Mikko Perttunen <[email protected]> пише:
Quoted 94 lines · Svyatoslav Ryhel
Quoted 77 lines
вт, 28 квіт. 2026 р. о 04:53 Mikko Perttunen <[email protected]> пише:
Quoted 66 lines · Svyatoslav Ryhel
Quoted 64 lines
From: Ion Agorria <[email protected]>
The gr*d_remove() has
pm_runtime_disable, this indicates it should be paired withpm_runtime_enablein the probe instead of being inside gr*d_runtime_resume().Signed-off-by Ion Agorria <[email protected]>Signed-off-by Svyatoslav Ryhel <[email protected]>--- drivers/gpu/drm/tegra/gr2d.c | 8 ++++---- drivers/gpu/drm/tegra/gr3d.c | 8 ++++---- 2 files changed, 8 insertions(+), 8 deletions(-)drivers/gpu/drm/tegra/gr2d.c +4
-4@@ -286,6 +286,10 @@ static int gr2d_probe(struct platform_device *pdev)for (i =0 ; i <ARRAY_SIZE (gr2d_addr_regs); i++)set_bit (gr2d_addr_regs[i], gr2d->addr_regs); +pm_runtime_enable (dev); +pm_runtime_use_autosuspend (dev); +pm_runtime_set_autosuspend_delay (dev,500 ); +return 0 ; }@@ -367,10 +371,6 @@ static int __maybe_unused gr2d_runtime_resume(struct device *dev) goto disable_clk; } -pm_runtime_enable (dev); -pm_runtime_use_autosuspend (dev); -pm_runtime_set_autosuspend_delay (dev,500 ); -return 0 ; disable_clk:drivers/gpu/drm/tegra/gr3d.c +4
-4@@ -517,6 +517,10 @@ static int gr3d_probe(struct platform_device *pdev)for (i =0 ; i <ARRAY_SIZE (gr3d_addr_regs); i++)set_bit (gr3d_addr_regs[i], gr3d->addr_regs); +pm_runtime_enable (&pdev->dev); +pm_runtime_use_autosuspend (&pdev->dev); +pm_runtime_set_autosuspend_delay (&pdev->dev,500 ); +return 0 ; }@@ -578,10 +582,6 @@ static int __maybe_unused gr3d_runtime_resume(struct device *dev) goto disable_clk; } -pm_runtime_enable (dev); -pm_runtime_use_autosuspend (dev); -pm_runtime_set_autosuspend_delay (dev,500 ); -return 0 ;Quoted 9 lines
Quoted 5 lines
drivers/gpu/drm/tegra/gr3d.c +0
-1disable_clk: --2.51.0
Oof, looks like I had managed to really bungle this with my earlier patch. Thanks for fixing it!
Quoted 8 lines
Hello Mikko!
Thank you for taking time and looking into this patch. Don't be so harsh to yourself, PM is easy to mess and hard to set properly. This patch does fix gr*d access but it does not resolve the issue itself. PM should be set in the init/exit rather then probe/remove. I have v2 which fixes this and one more minor issue and I will send them later on.
Thanks! Why do you think it's necessary to enable runtime PM in init? If you look at the commit I referenced below (in 'Fixes'), we've had some issues in the past with doing
pm_runtime_enableoutside of probe, where the engine's power domain would be left enabled even when it is idle.
Quoted 4 lines
Quoted 2 lines
gr2d/gr3d I suppose wouldn't be in practice affected by that issue though given they aren't in their own power domains.
WDYM, gr2d and gr3d have their own power domains.
Ah, my bad, didn't look far enough in the device tree.
Quoted 10 lines
If the master device is unbound and rebound,
gr2d_init()will run again, but pm configuration is only located in this probe function which will not run again, while pm disable are both in exit and remove. This results in pm issue we are observing.Solution would be either do everyting in probe/remove, or init/exit. Probe/remove will lead to domain being turned on even if engines are idle. Init/exit seems to me more suitable and we have tasted this configuration for a quite while in the grate DRM version.
Yes, we should only do it in one place. Where are you observing the domain being turned on when idle?
Genpd turns on the domain before probe is called, and it should turn it off again after probe if probe had enabled runtime PM and the usage count is zero. That's what we have currently for the newer engines like VIC and it's working.
We used to have it in init/exit, but then we were seeing the domain sometimes being left on at boot, which makes some sense since as probe didn't enable runtime PM, genpd would leave the domain powered after probe. The domain would then also never get powered off when init enabled runtime PM. I assume the reason this only happened sometimes was that it was dependent on if the init function or the genpd post-probe code ran first.
Based on that we should have the runtime PM handling in probe/remove, to ensure genpd turns off the domain after probe when there are no users.
Mikko
Quoted 27 lines
Best regards,
Svyatoslav R.Quoted 23 lines
Cheers
MikkoQuoted 15 lines
So for now this patch should not be picked.
Best regards,
Svyatoslav R.Quoted 9 lines
FWIW, I've been working on adding some nightly testing for Host1x/ TegraDRM, so hopefully we'll be able to catch such things easier in the future.
Fixes 62fa0a985e2c ("drm/tegra: Enable runtime PM during probe")Acked-by Mikko Perttunen <[email protected]>
чт, 30 квіт. 2026 р. о 11:38 Mikko Perttunen <[email protected]> пише:
Quoted 131 lines · Svyatoslav Ryhel
Quoted 95 lines
чт, 30 квіт. 2026 р. о 09:23 Mikko Perttunen <[email protected]> пише:
Quoted 94 lines · Svyatoslav Ryhel
Quoted 77 lines
вт, 28 квіт. 2026 р. о 04:53 Mikko Perttunen <[email protected]> пише:
Quoted 66 lines · Svyatoslav Ryhel
Quoted 64 lines
From: Ion Agorria <[email protected]>
The gr*d_remove() has
pm_runtime_disable, this indicates it should be paired withpm_runtime_enablein the probe instead of being inside gr*d_runtime_resume().Signed-off-by Ion Agorria <[email protected]>Signed-off-by Svyatoslav Ryhel <[email protected]>--- drivers/gpu/drm/tegra/gr2d.c | 8 ++++---- drivers/gpu/drm/tegra/gr3d.c | 8 ++++---- 2 files changed, 8 insertions(+), 8 deletions(-)drivers/gpu/drm/tegra/gr2d.c +4
-4@@ -286,6 +286,10 @@ static int gr2d_probe(struct platform_device *pdev)for (i =0 ; i <ARRAY_SIZE (gr2d_addr_regs); i++)set_bit (gr2d_addr_regs[i], gr2d->addr_regs); +pm_runtime_enable (dev); +pm_runtime_use_autosuspend (dev); +pm_runtime_set_autosuspend_delay (dev,500 ); +return 0 ; }@@ -367,10 +371,6 @@ static int __maybe_unused gr2d_runtime_resume(struct device *dev) goto disable_clk; } -pm_runtime_enable (dev); -pm_runtime_use_autosuspend (dev); -pm_runtime_set_autosuspend_delay (dev,500 ); -return 0 ; disable_clk:drivers/gpu/drm/tegra/gr3d.c +4
-4@@ -517,6 +517,10 @@ static int gr3d_probe(struct platform_device *pdev)for (i =0 ; i <ARRAY_SIZE (gr3d_addr_regs); i++)set_bit (gr3d_addr_regs[i], gr3d->addr_regs); +pm_runtime_enable (&pdev->dev); +pm_runtime_use_autosuspend (&pdev->dev); +pm_runtime_set_autosuspend_delay (&pdev->dev,500 ); +return 0 ; }@@ -578,10 +582,6 @@ static int __maybe_unused gr3d_runtime_resume(struct device *dev) goto disable_clk; } -pm_runtime_enable (dev); -pm_runtime_use_autosuspend (dev); -pm_runtime_set_autosuspend_delay (dev,500 ); -return 0 ;Quoted 9 lines
Quoted 5 lines
drivers/gpu/drm/tegra/gr3d.c +0
-1disable_clk: --2.51.0
Oof, looks like I had managed to really bungle this with my earlier patch. Thanks for fixing it!
Quoted 8 lines
Hello Mikko!
Thank you for taking time and looking into this patch. Don't be so harsh to yourself, PM is easy to mess and hard to set properly. This patch does fix gr*d access but it does not resolve the issue itself. PM should be set in the init/exit rather then probe/remove. I have v2 which fixes this and one more minor issue and I will send them later on.
Thanks! Why do you think it's necessary to enable runtime PM in init? If you look at the commit I referenced below (in 'Fixes'), we've had some issues in the past with doing
pm_runtime_enableoutside of probe, where the engine's power domain would be left enabled even when it is idle.Quoted 4 lines
Quoted 2 lines
gr2d/gr3d I suppose wouldn't be in practice affected by that issue though given they aren't in their own power domains.
WDYM, gr2d and gr3d have their own power domains.
Ah, my bad, didn't look far enough in the device tree.
Quoted 10 lines
If the master device is unbound and rebound,
gr2d_init()will run again, but pm configuration is only located in this probe function which will not run again, while pm disable are both in exit and remove. This results in pm issue we are observing.Solution would be either do everyting in probe/remove, or init/exit. Probe/remove will lead to domain being turned on even if engines are idle. Init/exit seems to me more suitable and we have tasted this configuration for a quite while in the grate DRM version.
Yes, we should only do it in one place. Where are you observing the domain being turned on when idle?
Genpd turns on the domain before probe is called, and it should turn it off again after probe if probe had enabled runtime PM and the usage count is zero. That's what we have currently for the newer engines like VIC and it's working.
We used to have it in init/exit, but then we were seeing the domain sometimes being left on at boot, which makes some sense since as probe didn't enable runtime PM, genpd would leave the domain powered after probe. The domain would then also never get powered off when init enabled runtime PM. I assume the reason this only happened sometimes was that it was dependent on if the init function or the genpd post-probe code ran first.
Quoted 6 lines
Based on that we should have the runtime PM handling in probe/remove, to ensure genpd turns off the domain after probe when there are no users.
Mikko
Thank you for an explanation, I will adjust v2 accordingly (configure PM in probe/remove).
Quoted 31 lines
Quoted 27 lines
Best regards,
Svyatoslav R.Quoted 23 lines
Cheers
MikkoQuoted 15 lines
So for now this patch should not be picked.
Best regards,
Svyatoslav R.Quoted 9 lines
FWIW, I've been working on adding some nightly testing for Host1x/ TegraDRM, so hopefully we'll be able to catch such things easier in the future.
Fixes 62fa0a985e2c ("drm/tegra: Enable runtime PM during probe")Acked-by Mikko Perttunen <[email protected]>