Re: [Freedreno] [PATCH v1 1/5] drm/msm/dp: remove pm_runtime_xxx() from dp_power.c
On 10/07/2023 20:25, Kuogee Hsieh wrote: On 7/9/2023 1:32 PM, Abhinav Kumar wrote: On 7/9/2023 11:00 AM, Dmitry Baryshkov wrote: On Sun, 9 Jul 2023 at 20:22, Abhinav Kumar wrote: On 7/7/2023 5:06 PM, Dmitry Baryshkov wrote: On 08/07/2023 02:52, Kuogee Hsieh wrote: Since both pm_runtime_resume() and pm_runtime_suspend() are not populated at dp_pm_ops. Those pm_runtime_get/put() functions within dp_power.c will not have any effects in addition to increase/decrease power counter. Lie. Even if the commit text is incorrect, review comments like this are not helping the patch nor the author and will just get ignored anyway. The review comment might be overreacting, excuse me. I was really impressed by the commit message, which contradicts the basic source code. pm_runtime_get() does a lot more than just increasing the power counter. It says within dp_power.c. Nonetheless, please let us know what is missing in the context of this patch like Bjorn did to make it an effective review and we can correct it. In its current form, the review comment is adding no value. I am new in pm. Any recommendation to revise this commit test? I'd say, squash patches 1 and 2 and then state in the commit message that you are changing pm_runtime code paths because you want to power up the device from the runtime callbacks rather than just waking up the device in the power up path. Generally it is much easier to justify changing from A to B rather than just dropping A and then adding B. Also pm_runtime_xxx() should be executed at top layer. Why? I guess he meant to centralize this around dp_display.c. Will elaborate while posting the next rev. Signed-off-by: Kuogee Hsieh --- drivers/gpu/drm/msm/dp/dp_power.c | 9 - 1 file changed, 9 deletions(-) diff --git a/drivers/gpu/drm/msm/dp/dp_power.c b/drivers/gpu/drm/msm/dp/dp_power.c index 5cb84ca..ed2f62a 100644 --- a/drivers/gpu/drm/msm/dp/dp_power.c +++ b/drivers/gpu/drm/msm/dp/dp_power.c @@ -152,8 +152,6 @@ int dp_power_client_init(struct dp_power *dp_power) power = container_of(dp_power, struct dp_power_private, dp_power); - pm_runtime_enable(power->dev); - return dp_power_clk_init(power); } @@ -162,8 +160,6 @@ void dp_power_client_deinit(struct dp_power *dp_power) struct dp_power_private *power; power = container_of(dp_power, struct dp_power_private, dp_power); - - pm_runtime_disable(power->dev); } int dp_power_init(struct dp_power *dp_power) @@ -173,11 +169,7 @@ int dp_power_init(struct dp_power *dp_power) power = container_of(dp_power, struct dp_power_private, dp_power); - pm_runtime_get_sync(power->dev); - rc = dp_power_clk_enable(dp_power, DP_CORE_PM, true); - if (rc) - pm_runtime_put_sync(power->dev); return rc; } @@ -189,7 +181,6 @@ int dp_power_deinit(struct dp_power *dp_power) power = container_of(dp_power, struct dp_power_private, dp_power); dp_power_clk_enable(dp_power, DP_CORE_PM, false); - pm_runtime_put_sync(power->dev); return 0; } -- With best wishes Dmitry
Re: [Freedreno] [PATCH v1 1/5] drm/msm/dp: remove pm_runtime_xxx() from dp_power.c
On 7/9/2023 1:32 PM, Abhinav Kumar wrote: On 7/9/2023 11:00 AM, Dmitry Baryshkov wrote: On Sun, 9 Jul 2023 at 20:22, Abhinav Kumar wrote: On 7/7/2023 5:06 PM, Dmitry Baryshkov wrote: On 08/07/2023 02:52, Kuogee Hsieh wrote: Since both pm_runtime_resume() and pm_runtime_suspend() are not populated at dp_pm_ops. Those pm_runtime_get/put() functions within dp_power.c will not have any effects in addition to increase/decrease power counter. Lie. Even if the commit text is incorrect, review comments like this are not helping the patch nor the author and will just get ignored anyway. The review comment might be overreacting, excuse me. I was really impressed by the commit message, which contradicts the basic source code. pm_runtime_get() does a lot more than just increasing the power counter. It says within dp_power.c. Nonetheless, please let us know what is missing in the context of this patch like Bjorn did to make it an effective review and we can correct it. In its current form, the review comment is adding no value. I am new in pm. Any recommendation to revise this commit test? Also pm_runtime_xxx() should be executed at top layer. Why? I guess he meant to centralize this around dp_display.c. Will elaborate while posting the next rev. Signed-off-by: Kuogee Hsieh --- drivers/gpu/drm/msm/dp/dp_power.c | 9 - 1 file changed, 9 deletions(-) diff --git a/drivers/gpu/drm/msm/dp/dp_power.c b/drivers/gpu/drm/msm/dp/dp_power.c index 5cb84ca..ed2f62a 100644 --- a/drivers/gpu/drm/msm/dp/dp_power.c +++ b/drivers/gpu/drm/msm/dp/dp_power.c @@ -152,8 +152,6 @@ int dp_power_client_init(struct dp_power *dp_power) power = container_of(dp_power, struct dp_power_private, dp_power); - pm_runtime_enable(power->dev); - return dp_power_clk_init(power); } @@ -162,8 +160,6 @@ void dp_power_client_deinit(struct dp_power *dp_power) struct dp_power_private *power; power = container_of(dp_power, struct dp_power_private, dp_power); - - pm_runtime_disable(power->dev); } int dp_power_init(struct dp_power *dp_power) @@ -173,11 +169,7 @@ int dp_power_init(struct dp_power *dp_power) power = container_of(dp_power, struct dp_power_private, dp_power); - pm_runtime_get_sync(power->dev); - rc = dp_power_clk_enable(dp_power, DP_CORE_PM, true); - if (rc) - pm_runtime_put_sync(power->dev); return rc; } @@ -189,7 +181,6 @@ int dp_power_deinit(struct dp_power *dp_power) power = container_of(dp_power, struct dp_power_private, dp_power); dp_power_clk_enable(dp_power, DP_CORE_PM, false); - pm_runtime_put_sync(power->dev); return 0; }
Re: [Freedreno] [PATCH v1 1/5] drm/msm/dp: remove pm_runtime_xxx() from dp_power.c
On 7/9/2023 11:00 AM, Dmitry Baryshkov wrote: On Sun, 9 Jul 2023 at 20:22, Abhinav Kumar wrote: On 7/7/2023 5:06 PM, Dmitry Baryshkov wrote: On 08/07/2023 02:52, Kuogee Hsieh wrote: Since both pm_runtime_resume() and pm_runtime_suspend() are not populated at dp_pm_ops. Those pm_runtime_get/put() functions within dp_power.c will not have any effects in addition to increase/decrease power counter. Lie. Even if the commit text is incorrect, review comments like this are not helping the patch nor the author and will just get ignored anyway. The review comment might be overreacting, excuse me. I was really impressed by the commit message, which contradicts the basic source code. pm_runtime_get() does a lot more than just increasing the power counter. It says within dp_power.c. Nonetheless, please let us know what is missing in the context of this patch like Bjorn did to make it an effective review and we can correct it. In its current form, the review comment is adding no value. Also pm_runtime_xxx() should be executed at top layer. Why? I guess he meant to centralize this around dp_display.c. Will elaborate while posting the next rev. Signed-off-by: Kuogee Hsieh --- drivers/gpu/drm/msm/dp/dp_power.c | 9 - 1 file changed, 9 deletions(-) diff --git a/drivers/gpu/drm/msm/dp/dp_power.c b/drivers/gpu/drm/msm/dp/dp_power.c index 5cb84ca..ed2f62a 100644 --- a/drivers/gpu/drm/msm/dp/dp_power.c +++ b/drivers/gpu/drm/msm/dp/dp_power.c @@ -152,8 +152,6 @@ int dp_power_client_init(struct dp_power *dp_power) power = container_of(dp_power, struct dp_power_private, dp_power); -pm_runtime_enable(power->dev); - return dp_power_clk_init(power); } @@ -162,8 +160,6 @@ void dp_power_client_deinit(struct dp_power *dp_power) struct dp_power_private *power; power = container_of(dp_power, struct dp_power_private, dp_power); - -pm_runtime_disable(power->dev); } int dp_power_init(struct dp_power *dp_power) @@ -173,11 +169,7 @@ int dp_power_init(struct dp_power *dp_power) power = container_of(dp_power, struct dp_power_private, dp_power); -pm_runtime_get_sync(power->dev); - rc = dp_power_clk_enable(dp_power, DP_CORE_PM, true); -if (rc) -pm_runtime_put_sync(power->dev); return rc; } @@ -189,7 +181,6 @@ int dp_power_deinit(struct dp_power *dp_power) power = container_of(dp_power, struct dp_power_private, dp_power); dp_power_clk_enable(dp_power, DP_CORE_PM, false); -pm_runtime_put_sync(power->dev); return 0; }
Re: [Freedreno] [PATCH v1 1/5] drm/msm/dp: remove pm_runtime_xxx() from dp_power.c
On Sun, 9 Jul 2023 at 20:22, Abhinav Kumar wrote: > > > > On 7/7/2023 5:06 PM, Dmitry Baryshkov wrote: > > On 08/07/2023 02:52, Kuogee Hsieh wrote: > >> Since both pm_runtime_resume() and pm_runtime_suspend() are not > >> populated at dp_pm_ops. Those pm_runtime_get/put() functions within > >> dp_power.c will not have any effects in addition to increase/decrease > >> power counter. > > > > Lie. > > > > Even if the commit text is incorrect, review comments like this are not > helping the patch nor the author and will just get ignored anyway. The review comment might be overreacting, excuse me. I was really impressed by the commit message, which contradicts the basic source code. pm_runtime_get() does a lot more than just increasing the power counter. > >> Also pm_runtime_xxx() should be executed at top > >> layer. > > > > Why? > > > > I guess he meant to centralize this around dp_display.c. Will elaborate > while posting the next rev. > > >> > >> Signed-off-by: Kuogee Hsieh > >> --- > >> drivers/gpu/drm/msm/dp/dp_power.c | 9 - > >> 1 file changed, 9 deletions(-) > >> > >> diff --git a/drivers/gpu/drm/msm/dp/dp_power.c > >> b/drivers/gpu/drm/msm/dp/dp_power.c > >> index 5cb84ca..ed2f62a 100644 > >> --- a/drivers/gpu/drm/msm/dp/dp_power.c > >> +++ b/drivers/gpu/drm/msm/dp/dp_power.c > >> @@ -152,8 +152,6 @@ int dp_power_client_init(struct dp_power *dp_power) > >> power = container_of(dp_power, struct dp_power_private, dp_power); > >> -pm_runtime_enable(power->dev); > >> - > >> return dp_power_clk_init(power); > >> } > >> @@ -162,8 +160,6 @@ void dp_power_client_deinit(struct dp_power > >> *dp_power) > >> struct dp_power_private *power; > >> power = container_of(dp_power, struct dp_power_private, dp_power); > >> - > >> -pm_runtime_disable(power->dev); > >> } > >> int dp_power_init(struct dp_power *dp_power) > >> @@ -173,11 +169,7 @@ int dp_power_init(struct dp_power *dp_power) > >> power = container_of(dp_power, struct dp_power_private, dp_power); > >> -pm_runtime_get_sync(power->dev); > >> - > >> rc = dp_power_clk_enable(dp_power, DP_CORE_PM, true); > >> -if (rc) > >> -pm_runtime_put_sync(power->dev); > >> return rc; > >> } > >> @@ -189,7 +181,6 @@ int dp_power_deinit(struct dp_power *dp_power) > >> power = container_of(dp_power, struct dp_power_private, dp_power); > >> dp_power_clk_enable(dp_power, DP_CORE_PM, false); > >> -pm_runtime_put_sync(power->dev); > >> return 0; > >> } > > -- With best wishes Dmitry
Re: [PATCH v1 1/5] drm/msm/dp: remove pm_runtime_xxx() from dp_power.c
On 7/8/2023 7:34 PM, Bjorn Andersson wrote: On Fri, Jul 07, 2023 at 04:52:19PM -0700, Kuogee Hsieh wrote: Since both pm_runtime_resume() and pm_runtime_suspend() are not populated at dp_pm_ops. Those pm_runtime_get/put() functions within dp_power.c will not have any effects in addition to increase/decrease power counter. Also pm_runtime_xxx() should be executed at top layer. Getting/putting the runtime PM state affects the vote for the GDSC. So I would suggest that you move this after patch 2, to not create a gap in the git history of lacking GDSC votes. Regards, Bjorn the mdss_dp node has rpmhpd SC7180_CX in its power domains. the parent device has the GDSC. So just so that I understand this right, the DP's vote on this will still count because the parent's power domain wont get collapsed till the child PM votes are removed too? If so, I see your point and yes it will make sense to move this later to avoid the gap. Signed-off-by: Kuogee Hsieh --- drivers/gpu/drm/msm/dp/dp_power.c | 9 - 1 file changed, 9 deletions(-) diff --git a/drivers/gpu/drm/msm/dp/dp_power.c b/drivers/gpu/drm/msm/dp/dp_power.c index 5cb84ca..ed2f62a 100644 --- a/drivers/gpu/drm/msm/dp/dp_power.c +++ b/drivers/gpu/drm/msm/dp/dp_power.c @@ -152,8 +152,6 @@ int dp_power_client_init(struct dp_power *dp_power) power = container_of(dp_power, struct dp_power_private, dp_power); - pm_runtime_enable(power->dev); - return dp_power_clk_init(power); } @@ -162,8 +160,6 @@ void dp_power_client_deinit(struct dp_power *dp_power) struct dp_power_private *power; power = container_of(dp_power, struct dp_power_private, dp_power); - - pm_runtime_disable(power->dev); } int dp_power_init(struct dp_power *dp_power) @@ -173,11 +169,7 @@ int dp_power_init(struct dp_power *dp_power) power = container_of(dp_power, struct dp_power_private, dp_power); - pm_runtime_get_sync(power->dev); - rc = dp_power_clk_enable(dp_power, DP_CORE_PM, true); - if (rc) - pm_runtime_put_sync(power->dev); return rc; } @@ -189,7 +181,6 @@ int dp_power_deinit(struct dp_power *dp_power) power = container_of(dp_power, struct dp_power_private, dp_power); dp_power_clk_enable(dp_power, DP_CORE_PM, false); - pm_runtime_put_sync(power->dev); return 0; } -- 2.7.4
Re: [Freedreno] [PATCH v1 1/5] drm/msm/dp: remove pm_runtime_xxx() from dp_power.c
On 7/7/2023 5:06 PM, Dmitry Baryshkov wrote: On 08/07/2023 02:52, Kuogee Hsieh wrote: Since both pm_runtime_resume() and pm_runtime_suspend() are not populated at dp_pm_ops. Those pm_runtime_get/put() functions within dp_power.c will not have any effects in addition to increase/decrease power counter. Lie. Even if the commit text is incorrect, review comments like this are not helping the patch nor the author and will just get ignored anyway. Also pm_runtime_xxx() should be executed at top layer. Why? I guess he meant to centralize this around dp_display.c. Will elaborate while posting the next rev. Signed-off-by: Kuogee Hsieh --- drivers/gpu/drm/msm/dp/dp_power.c | 9 - 1 file changed, 9 deletions(-) diff --git a/drivers/gpu/drm/msm/dp/dp_power.c b/drivers/gpu/drm/msm/dp/dp_power.c index 5cb84ca..ed2f62a 100644 --- a/drivers/gpu/drm/msm/dp/dp_power.c +++ b/drivers/gpu/drm/msm/dp/dp_power.c @@ -152,8 +152,6 @@ int dp_power_client_init(struct dp_power *dp_power) power = container_of(dp_power, struct dp_power_private, dp_power); - pm_runtime_enable(power->dev); - return dp_power_clk_init(power); } @@ -162,8 +160,6 @@ void dp_power_client_deinit(struct dp_power *dp_power) struct dp_power_private *power; power = container_of(dp_power, struct dp_power_private, dp_power); - - pm_runtime_disable(power->dev); } int dp_power_init(struct dp_power *dp_power) @@ -173,11 +169,7 @@ int dp_power_init(struct dp_power *dp_power) power = container_of(dp_power, struct dp_power_private, dp_power); - pm_runtime_get_sync(power->dev); - rc = dp_power_clk_enable(dp_power, DP_CORE_PM, true); - if (rc) - pm_runtime_put_sync(power->dev); return rc; } @@ -189,7 +181,6 @@ int dp_power_deinit(struct dp_power *dp_power) power = container_of(dp_power, struct dp_power_private, dp_power); dp_power_clk_enable(dp_power, DP_CORE_PM, false); - pm_runtime_put_sync(power->dev); return 0; }
Re: [PATCH v1 1/5] drm/msm/dp: remove pm_runtime_xxx() from dp_power.c
On Fri, Jul 07, 2023 at 04:52:19PM -0700, Kuogee Hsieh wrote: > Since both pm_runtime_resume() and pm_runtime_suspend() are not > populated at dp_pm_ops. Those pm_runtime_get/put() functions within > dp_power.c will not have any effects in addition to increase/decrease > power counter. Also pm_runtime_xxx() should be executed at top > layer. > Getting/putting the runtime PM state affects the vote for the GDSC. So I would suggest that you move this after patch 2, to not create a gap in the git history of lacking GDSC votes. Regards, Bjorn > Signed-off-by: Kuogee Hsieh > --- > drivers/gpu/drm/msm/dp/dp_power.c | 9 - > 1 file changed, 9 deletions(-) > > diff --git a/drivers/gpu/drm/msm/dp/dp_power.c > b/drivers/gpu/drm/msm/dp/dp_power.c > index 5cb84ca..ed2f62a 100644 > --- a/drivers/gpu/drm/msm/dp/dp_power.c > +++ b/drivers/gpu/drm/msm/dp/dp_power.c > @@ -152,8 +152,6 @@ int dp_power_client_init(struct dp_power *dp_power) > > power = container_of(dp_power, struct dp_power_private, dp_power); > > - pm_runtime_enable(power->dev); > - > return dp_power_clk_init(power); > } > > @@ -162,8 +160,6 @@ void dp_power_client_deinit(struct dp_power *dp_power) > struct dp_power_private *power; > > power = container_of(dp_power, struct dp_power_private, dp_power); > - > - pm_runtime_disable(power->dev); > } > > int dp_power_init(struct dp_power *dp_power) > @@ -173,11 +169,7 @@ int dp_power_init(struct dp_power *dp_power) > > power = container_of(dp_power, struct dp_power_private, dp_power); > > - pm_runtime_get_sync(power->dev); > - > rc = dp_power_clk_enable(dp_power, DP_CORE_PM, true); > - if (rc) > - pm_runtime_put_sync(power->dev); > > return rc; > } > @@ -189,7 +181,6 @@ int dp_power_deinit(struct dp_power *dp_power) > power = container_of(dp_power, struct dp_power_private, dp_power); > > dp_power_clk_enable(dp_power, DP_CORE_PM, false); > - pm_runtime_put_sync(power->dev); > return 0; > } > > -- > 2.7.4 >
Re: [PATCH v1 1/5] drm/msm/dp: remove pm_runtime_xxx() from dp_power.c
On 08/07/2023 02:52, Kuogee Hsieh wrote: Since both pm_runtime_resume() and pm_runtime_suspend() are not populated at dp_pm_ops. Those pm_runtime_get/put() functions within dp_power.c will not have any effects in addition to increase/decrease power counter. Lie. Also pm_runtime_xxx() should be executed at top layer. Why? Signed-off-by: Kuogee Hsieh --- drivers/gpu/drm/msm/dp/dp_power.c | 9 - 1 file changed, 9 deletions(-) diff --git a/drivers/gpu/drm/msm/dp/dp_power.c b/drivers/gpu/drm/msm/dp/dp_power.c index 5cb84ca..ed2f62a 100644 --- a/drivers/gpu/drm/msm/dp/dp_power.c +++ b/drivers/gpu/drm/msm/dp/dp_power.c @@ -152,8 +152,6 @@ int dp_power_client_init(struct dp_power *dp_power) power = container_of(dp_power, struct dp_power_private, dp_power); - pm_runtime_enable(power->dev); - return dp_power_clk_init(power); } @@ -162,8 +160,6 @@ void dp_power_client_deinit(struct dp_power *dp_power) struct dp_power_private *power; power = container_of(dp_power, struct dp_power_private, dp_power); - - pm_runtime_disable(power->dev); } int dp_power_init(struct dp_power *dp_power) @@ -173,11 +169,7 @@ int dp_power_init(struct dp_power *dp_power) power = container_of(dp_power, struct dp_power_private, dp_power); - pm_runtime_get_sync(power->dev); - rc = dp_power_clk_enable(dp_power, DP_CORE_PM, true); - if (rc) - pm_runtime_put_sync(power->dev); return rc; } @@ -189,7 +181,6 @@ int dp_power_deinit(struct dp_power *dp_power) power = container_of(dp_power, struct dp_power_private, dp_power); dp_power_clk_enable(dp_power, DP_CORE_PM, false); - pm_runtime_put_sync(power->dev); return 0; } -- With best wishes Dmitry
[PATCH v1 1/5] drm/msm/dp: remove pm_runtime_xxx() from dp_power.c
Since both pm_runtime_resume() and pm_runtime_suspend() are not populated at dp_pm_ops. Those pm_runtime_get/put() functions within dp_power.c will not have any effects in addition to increase/decrease power counter. Also pm_runtime_xxx() should be executed at top layer. Signed-off-by: Kuogee Hsieh --- drivers/gpu/drm/msm/dp/dp_power.c | 9 - 1 file changed, 9 deletions(-) diff --git a/drivers/gpu/drm/msm/dp/dp_power.c b/drivers/gpu/drm/msm/dp/dp_power.c index 5cb84ca..ed2f62a 100644 --- a/drivers/gpu/drm/msm/dp/dp_power.c +++ b/drivers/gpu/drm/msm/dp/dp_power.c @@ -152,8 +152,6 @@ int dp_power_client_init(struct dp_power *dp_power) power = container_of(dp_power, struct dp_power_private, dp_power); - pm_runtime_enable(power->dev); - return dp_power_clk_init(power); } @@ -162,8 +160,6 @@ void dp_power_client_deinit(struct dp_power *dp_power) struct dp_power_private *power; power = container_of(dp_power, struct dp_power_private, dp_power); - - pm_runtime_disable(power->dev); } int dp_power_init(struct dp_power *dp_power) @@ -173,11 +169,7 @@ int dp_power_init(struct dp_power *dp_power) power = container_of(dp_power, struct dp_power_private, dp_power); - pm_runtime_get_sync(power->dev); - rc = dp_power_clk_enable(dp_power, DP_CORE_PM, true); - if (rc) - pm_runtime_put_sync(power->dev); return rc; } @@ -189,7 +181,6 @@ int dp_power_deinit(struct dp_power *dp_power) power = container_of(dp_power, struct dp_power_private, dp_power); dp_power_clk_enable(dp_power, DP_CORE_PM, false); - pm_runtime_put_sync(power->dev); return 0; } -- 2.7.4
