2015-02-12 20:49:34

by Sylvain Rochet

[permalink] [raw]
Subject: [PATCHv2 0/2] drm: atmel-hlcdc: PM support

This series depends on Boris' "[PATCH v2] drm: atmel-hlcdc: Atomic
mode-setting conversion"
<1423236143-6494-1-git-send-email-boris.brezillon@free-electrons.com>
plus a few fixes which are going to be in v3 of Boris' patch.

This series adds basic PM support for Atmel HLCDC.

Changes since v1:
* (*crtc_funcs->disable)(crtc) replaced to crtc_funcs->disable(crtc)

Sylvain Rochet (2):
drm: atmel-hlcdc: Add PM suspend/resume support
drm: atmel-hlcdc: Add pinctrl PM select sleep,default state in CRTC
suspend/resume

drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_crtc.c | 3 ++
drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_dc.c | 60 ++++++++++++++++++++++++++
2 files changed, 63 insertions(+)

--
2.1.4


2015-02-12 20:49:35

by Sylvain Rochet

[permalink] [raw]
Subject: [PATCHv2 1/2] drm: atmel-hlcdc: Add PM suspend/resume support

On suspend: switch off CRTC if not already suspended with runtime PM

On resume: switch on CRTC if we were not already suspended from runtime
PM while suspending.

Signed-off-by: Sylvain Rochet <[email protected]>
---
drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_dc.c | 60 ++++++++++++++++++++++++++++
1 file changed, 60 insertions(+)

diff --git a/drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_dc.c b/drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_dc.c
index 22c3cca..e22883b 100644
--- a/drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_dc.c
+++ b/drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_dc.c
@@ -427,6 +427,36 @@ static void atmel_hlcdc_dc_lastclose(struct drm_device *dev)
drm_fbdev_cma_restore_mode(dc->fbdev);
}

+static int atmel_hlcdc_dc_suspend(struct drm_device *dev, pm_message_t state)
+{
+ struct drm_crtc *crtc;
+
+ drm_modeset_lock_all(dev);
+
+ list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
+ struct drm_crtc_helper_funcs *crtc_funcs = crtc->helper_private;
+ crtc_funcs->disable(crtc);
+ }
+
+ drm_modeset_unlock_all(dev);
+ return 0;
+}
+
+static int atmel_hlcdc_dc_resume(struct drm_device *dev)
+{
+ struct drm_crtc *crtc;
+
+ drm_modeset_lock_all(dev);
+
+ list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
+ struct drm_crtc_helper_funcs *crtc_funcs = crtc->helper_private;
+ crtc_funcs->enable(crtc);
+ }
+
+ drm_modeset_unlock_all(dev);
+ return 0;
+}
+
static int atmel_hlcdc_dc_irq_postinstall(struct drm_device *dev)
{
struct atmel_hlcdc_dc *dc = dev->dev_private;
@@ -488,6 +518,8 @@ static struct drm_driver atmel_hlcdc_dc_driver = {
.driver_features = DRIVER_HAVE_IRQ | DRIVER_GEM | DRIVER_MODESET,
.preclose = atmel_hlcdc_dc_preclose,
.lastclose = atmel_hlcdc_dc_lastclose,
+ .suspend = atmel_hlcdc_dc_suspend,
+ .resume = atmel_hlcdc_dc_resume,
.irq_handler = atmel_hlcdc_dc_irq_handler,
.irq_preinstall = atmel_hlcdc_dc_irq_uninstall,
.irq_postinstall = atmel_hlcdc_dc_irq_postinstall,
@@ -559,6 +591,33 @@ static int atmel_hlcdc_dc_drm_remove(struct platform_device *pdev)
return 0;
}

+#ifdef CONFIG_PM
+static int atmel_hlcdc_dc_drm_suspend(struct device *dev)
+{
+ struct drm_device *drm_dev = dev_get_drvdata(dev);
+ pm_message_t message;
+
+ if (pm_runtime_suspended(dev) || !drm_dev)
+ return 0;
+
+ message.event = PM_EVENT_SUSPEND;
+ return atmel_hlcdc_dc_suspend(drm_dev, message);
+}
+
+static int atmel_hlcdc_dc_drm_resume(struct device *dev)
+{
+ struct drm_device *drm_dev = dev_get_drvdata(dev);
+
+ if (pm_runtime_suspended(dev) || !drm_dev)
+ return 0;
+
+ return atmel_hlcdc_dc_resume(drm_dev);
+}
+#endif
+
+static SIMPLE_DEV_PM_OPS(atmel_hlcdc_dc_drm_pm_ops,
+ atmel_hlcdc_dc_drm_suspend, atmel_hlcdc_dc_drm_resume);
+
static const struct of_device_id atmel_hlcdc_dc_of_match[] = {
{ .compatible = "atmel,hlcdc-display-controller" },
{ },
@@ -569,6 +628,7 @@ static struct platform_driver atmel_hlcdc_dc_platform_driver = {
.remove = atmel_hlcdc_dc_drm_remove,
.driver = {
.name = "atmel-hlcdc-display-controller",
+ .pm = &atmel_hlcdc_dc_drm_pm_ops,
.of_match_table = atmel_hlcdc_dc_of_match,
},
};
--
2.1.4

2015-02-12 20:49:36

by Sylvain Rochet

[permalink] [raw]
Subject: [PATCHv2 2/2] drm: atmel-hlcdc: Add pinctrl PM select sleep,default state in CRTC suspend/resume

Some LCD panels have back-powering issue when un-powered, allows users
to use an alternate pinctrl "sleep" in order to clamp outputs to a
wanted state at suspend.

Signed-off-by: Sylvain Rochet <[email protected]>
Acked-by: Boris Brezillon <[email protected]>
---
drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_crtc.c | 3 +++
1 file changed, 3 insertions(+)

diff --git a/drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_crtc.c b/drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_crtc.c
index 79c9a19..decd9ea 100644
--- a/drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_crtc.c
+++ b/drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_crtc.c
@@ -21,6 +21,7 @@
#include <linux/clk.h>
#include <linux/pm.h>
#include <linux/pm_runtime.h>
+#include <linux/pinctrl/consumer.h>

#include <drm/drm_crtc.h>
#include <drm/drm_crtc_helper.h>
@@ -156,6 +157,7 @@ static void atmel_hlcdc_crtc_disable(struct drm_crtc *c)
cpu_relax();

clk_disable_unprepare(crtc->dc->hlcdc->sys_clk);
+ pinctrl_pm_select_sleep_state(dev->dev);

pm_runtime_allow(dev->dev);

@@ -178,6 +180,7 @@ static void atmel_hlcdc_crtc_enable(struct drm_crtc *c)

pm_runtime_forbid(dev->dev);

+ pinctrl_pm_select_default_state(dev->dev);
clk_prepare_enable(crtc->dc->hlcdc->sys_clk);

regmap_write(regmap, ATMEL_HLCDC_EN, ATMEL_HLCDC_PIXEL_CLK);
--
2.1.4

2015-02-17 09:47:40

by Andrzej Hajda

[permalink] [raw]
Subject: Re: [PATCHv2 1/2] drm: atmel-hlcdc: Add PM suspend/resume support

Hi Sylvain,

Few comments below.

On 02/12/2015 09:49 PM, Sylvain Rochet wrote:
> On suspend: switch off CRTC if not already suspended with runtime PM
>
> On resume: switch on CRTC if we were not already suspended from runtime
> PM while suspending.
>
> Signed-off-by: Sylvain Rochet <[email protected]>
> ---
> drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_dc.c | 60 ++++++++++++++++++++++++++++
> 1 file changed, 60 insertions(+)
>
> diff --git a/drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_dc.c b/drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_dc.c
> index 22c3cca..e22883b 100644
> --- a/drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_dc.c
> +++ b/drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_dc.c
> @@ -427,6 +427,36 @@ static void atmel_hlcdc_dc_lastclose(struct drm_device *dev)
> drm_fbdev_cma_restore_mode(dc->fbdev);
> }
>
> +static int atmel_hlcdc_dc_suspend(struct drm_device *dev, pm_message_t state)
> +{
> + struct drm_crtc *crtc;
> +
> + drm_modeset_lock_all(dev);
> +
> + list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
> + struct drm_crtc_helper_funcs *crtc_funcs = crtc->helper_private;
> + crtc_funcs->disable(crtc);
> + }
> +
> + drm_modeset_unlock_all(dev);
> + return 0;
> +}
> +
> +static int atmel_hlcdc_dc_resume(struct drm_device *dev)
> +{
> + struct drm_crtc *crtc;
> +
> + drm_modeset_lock_all(dev);
> +
> + list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
> + struct drm_crtc_helper_funcs *crtc_funcs = crtc->helper_private;
> + crtc_funcs->enable(crtc);

What about crtc's which were disabled before suspend, they will be
enabled here unconditionally.


> + }
> +
> + drm_modeset_unlock_all(dev);
> + return 0;
> +}
> +
> static int atmel_hlcdc_dc_irq_postinstall(struct drm_device *dev)
> {
> struct atmel_hlcdc_dc *dc = dev->dev_private;
> @@ -488,6 +518,8 @@ static struct drm_driver atmel_hlcdc_dc_driver = {
> .driver_features = DRIVER_HAVE_IRQ | DRIVER_GEM | DRIVER_MODESET,
> .preclose = atmel_hlcdc_dc_preclose,
> .lastclose = atmel_hlcdc_dc_lastclose,
> + .suspend = atmel_hlcdc_dc_suspend,
> + .resume = atmel_hlcdc_dc_resume,

These callbacks are obsolete and should not be used, additionally core
don't call them if drm_device have DRIVER_MODESET flag.

> .irq_handler = atmel_hlcdc_dc_irq_handler,
> .irq_preinstall = atmel_hlcdc_dc_irq_uninstall,
> .irq_postinstall = atmel_hlcdc_dc_irq_postinstall,
> @@ -559,6 +591,33 @@ static int atmel_hlcdc_dc_drm_remove(struct platform_device *pdev)
> return 0;
> }
>
> +#ifdef CONFIG_PM
> +static int atmel_hlcdc_dc_drm_suspend(struct device *dev)
> +{
> + struct drm_device *drm_dev = dev_get_drvdata(dev);
> + pm_message_t message;
> +
> + if (pm_runtime_suspended(dev) || !drm_dev)

Nitpick, !drm_dev is always false, is not?

> + return 0;
> +
> + message.event = PM_EVENT_SUSPEND;
> + return atmel_hlcdc_dc_suspend(drm_dev, message);

If you remove obsolete callbacks you can move suspend code here,
the same for resume.

> +}
> +
> +static int atmel_hlcdc_dc_drm_resume(struct device *dev)
> +{
> + struct drm_device *drm_dev = dev_get_drvdata(dev);
> +
> + if (pm_runtime_suspended(dev) || !drm_dev)
> + return 0;
> +
> + return atmel_hlcdc_dc_resume(drm_dev);

Ditto x2

Regards
Andrzej

> +}
> +#endif
> +
> +static SIMPLE_DEV_PM_OPS(atmel_hlcdc_dc_drm_pm_ops,
> + atmel_hlcdc_dc_drm_suspend, atmel_hlcdc_dc_drm_resume);
> +
> static const struct of_device_id atmel_hlcdc_dc_of_match[] = {
> { .compatible = "atmel,hlcdc-display-controller" },
> { },
> @@ -569,6 +628,7 @@ static struct platform_driver atmel_hlcdc_dc_platform_driver = {
> .remove = atmel_hlcdc_dc_drm_remove,
> .driver = {
> .name = "atmel-hlcdc-display-controller",
> + .pm = &atmel_hlcdc_dc_drm_pm_ops,
> .of_match_table = atmel_hlcdc_dc_of_match,
> },
> };
>

2015-02-22 18:03:33

by Sylvain Rochet

[permalink] [raw]
Subject: Re: [PATCHv2 1/2] drm: atmel-hlcdc: Add PM suspend/resume support

Hello Andrzej,

On Tue, Feb 17, 2015 at 10:47:23AM +0100, Andrzej Hajda wrote:
> On 02/12/2015 09:49 PM, Sylvain Rochet wrote:
> > +static int atmel_hlcdc_dc_resume(struct drm_device *dev)
> > +{
> > + struct drm_crtc *crtc;
> > +
> > + drm_modeset_lock_all(dev);
> > +
> > + list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
> > + struct drm_crtc_helper_funcs *crtc_funcs = crtc->helper_private;
> > + crtc_funcs->enable(crtc);
>
> What about crtc's which were disabled before suspend, they will be
> enabled here unconditionally.

Indeed, fixed in v3.


> > struct atmel_hlcdc_dc *dc = dev->dev_private;
> > @@ -488,6 +518,8 @@ static struct drm_driver atmel_hlcdc_dc_driver = {
> > .driver_features = DRIVER_HAVE_IRQ | DRIVER_GEM | DRIVER_MODESET,
> > .preclose = atmel_hlcdc_dc_preclose,
> > .lastclose = atmel_hlcdc_dc_lastclose,
> > + .suspend = atmel_hlcdc_dc_suspend,
> > + .resume = atmel_hlcdc_dc_resume,
>
> These callbacks are obsolete and should not be used, additionally core
> don't call them if drm_device have DRIVER_MODESET flag.

Oh!, that might explain the struggle I had understanding when they were
actually called. Removed in v3.


> > +#ifdef CONFIG_PM
> > +static int atmel_hlcdc_dc_drm_suspend(struct device *dev)
> > +{
> > + struct drm_device *drm_dev = dev_get_drvdata(dev);
> > + pm_message_t message;
> > +
> > + if (pm_runtime_suspended(dev) || !drm_dev)
>
> Nitpick, !drm_dev is always false, is not?

I guess so, platform_set_drvdata() is called during probe so it can't be
null here. Removed in v3.


> > + return 0;
> > +
> > + message.event = PM_EVENT_SUSPEND;
> > + return atmel_hlcdc_dc_suspend(drm_dev, message);
>
> If you remove obsolete callbacks you can move suspend code here,
> the same for resume.
>
> > +}
> > +
> > +static int atmel_hlcdc_dc_drm_resume(struct device *dev)
> > +{
> > + struct drm_device *drm_dev = dev_get_drvdata(dev);
> > +
> > + if (pm_runtime_suspended(dev) || !drm_dev)
> > + return 0;
> > +
> > + return atmel_hlcdc_dc_resume(drm_dev);
>
> Ditto x2

Indeed, merged in v3.


Thank you very much for this review ! :)

Sylvain