[PATCH] drm/meson: enable runtime PM
Neil Armstrong
narmstrong at baylibre.com
Tue Oct 1 08:41:08 UTC 2019
On 30/09/2019 17:34, Kevin Hilman wrote:
> Neil Armstrong <narmstrong at baylibre.com> writes:
>
>> Hi Kevin,
>>
>> On 25/09/2019 21:31, Kevin Hilman wrote:
>>> From: Kevin Hilman <khilman at baylibre.com>
>>>
>>> On some SoCs, the VPU is in a power-domain and needs runtime PM
>>> enabled and used in order to keep the power domain on.
>>>
>>> Signed-off-by: Kevin Hilman <khilman at baylibre.com>
>>> ---
>>> drivers/gpu/drm/meson/meson_drv.c | 4 ++++
>>> 1 file changed, 4 insertions(+)
>>>
>>> diff --git a/drivers/gpu/drm/meson/meson_drv.c b/drivers/gpu/drm/meson/meson_drv.c
>>> index 2310c96fff46..256b6a0e9c6b 100644
>>> --- a/drivers/gpu/drm/meson/meson_drv.c
>>> +++ b/drivers/gpu/drm/meson/meson_drv.c
>>> @@ -14,6 +14,7 @@
>>> #include <linux/platform_device.h>
>>> #include <linux/component.h>
>>> #include <linux/of_graph.h>
>>> +#include <linux/pm_runtime.h>
>>>
>>> #include <drm/drmP.h>
>>> #include <drm/drm_atomic.h>
>>> @@ -274,6 +275,7 @@ static int meson_drv_bind_master(struct device *dev, bool has_components)
>>>
>>> /* Hardware Initialization */
>>>
>>> + pm_runtime_get_sync(dev);
>>> meson_vpu_init(priv);
>>> meson_venc_init(priv);
>>> meson_vpp_init(priv);
>>> @@ -416,6 +418,7 @@ static int meson_drv_probe(struct platform_device *pdev)
>>> struct device_node *ep, *remote;
>>> int count = 0;
>>>
>>> + pm_runtime_enable(&pdev->dev);
>>> for_each_endpoint_of_node(np, ep) {
>>> remote = of_graph_get_remote_port_parent(ep);
>>> if (!remote || !of_device_is_available(remote)) {
>>> @@ -440,6 +443,7 @@ static int meson_drv_probe(struct platform_device *pdev)
>>> }
>>>
>>> /* If no output endpoints were available, simply bail out */
>>> + pm_runtime_disable(&pdev->dev);
>>> return 0;
>>> };
>>>
>>>
>>
>> I'll rather implement true runtime PM instead,
>
> While this is a minimum implementation, can you explain what you mean by
> "true" runtime PM?
Implementing runtime_sync()/_put when the display is used/unused.
ARM did a good job in their malidp driver.
Neil
>
> Kevin
>
More information about the dri-devel
mailing list