Re: [PATCH 2/3] drm/msm/dsi: move next bridge acquisition to dsi_bind
On 4/5/2024 11:15 AM, Dmitry Baryshkov wrote:
On Fri, 5 Apr 2024 at 20:35, Abhinav Kumar wrote:
On 3/9/2024 7:09 AM, Dmitry Baryshkov wrote:
Currently the MSM DSI driver looks for the next bridge during
msm_dsi_modeset_init(). If the bridge is not registered at that point,
this might result in -EPROBE_DEFER, which can be problematic that late
during the device probe process. Move next bridge acquisition to the
dsi_bind state so that probe deferral is returned as early as possible.
But msm_dsi_modeset)init() is also called during msm_drm_bind() only.
What issue are you suggesting will be fixed by moving this from
msm_drm_bind() to dsi_bind()?
The goal is to return as early as possible as not not cause
probe-deferral loops. See commit fbc35b45f9f6 ("Add documentation on
meaning of -EPROBE_DEFER"). It discusses returning from probe() but
the same logic applies to bind().
Understood. I was trying to make sure the purpose of the patch is that
"deferral in component_bind() is better than component_master_bind()"
But yes, overall that is better since the unbounding path will be more
in the master case.
Signed-off-by: Dmitry Baryshkov
---
drivers/gpu/drm/msm/dsi/dsi.c | 16
drivers/gpu/drm/msm/dsi/dsi.h | 2 ++
drivers/gpu/drm/msm/dsi/dsi_manager.c | 8 +---
3 files changed, 19 insertions(+), 7 deletions(-)
Reviewed-by: Abhinav Kumar
Re: [PATCH 2/3] drm/msm/dsi: move next bridge acquisition to dsi_bind
On Fri, 5 Apr 2024 at 20:35, Abhinav Kumar wrote:
>
>
>
> On 3/9/2024 7:09 AM, Dmitry Baryshkov wrote:
> > Currently the MSM DSI driver looks for the next bridge during
> > msm_dsi_modeset_init(). If the bridge is not registered at that point,
> > this might result in -EPROBE_DEFER, which can be problematic that late
> > during the device probe process. Move next bridge acquisition to the
> > dsi_bind state so that probe deferral is returned as early as possible.
> >
>
> But msm_dsi_modeset)init() is also called during msm_drm_bind() only.
>
> What issue are you suggesting will be fixed by moving this from
> msm_drm_bind() to dsi_bind()?
The goal is to return as early as possible as not not cause
probe-deferral loops. See commit fbc35b45f9f6 ("Add documentation on
meaning of -EPROBE_DEFER"). It discusses returning from probe() but
the same logic applies to bind().
> > Signed-off-by: Dmitry Baryshkov
> > ---
> > drivers/gpu/drm/msm/dsi/dsi.c | 16
> > drivers/gpu/drm/msm/dsi/dsi.h | 2 ++
> > drivers/gpu/drm/msm/dsi/dsi_manager.c | 8 +---
> > 3 files changed, 19 insertions(+), 7 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/msm/dsi/dsi.c b/drivers/gpu/drm/msm/dsi/dsi.c
> > index 37c4c07005fe..38f10f7a10d3 100644
> > --- a/drivers/gpu/drm/msm/dsi/dsi.c
> > +++ b/drivers/gpu/drm/msm/dsi/dsi.c
> > @@ -120,6 +120,22 @@ static int dsi_bind(struct device *dev, struct device
> > *master, void *data)
> > struct msm_drm_private *priv = dev_get_drvdata(master);
> > struct msm_dsi *msm_dsi = dev_get_drvdata(dev);
> >
> > + /*
> > + * Next bridge doesn't exist for the secondary DSI host in a bonded
> > + * pair.
> > + */
> > + if (!msm_dsi_is_bonded_dsi(msm_dsi) ||
> > + msm_dsi_is_master_dsi(msm_dsi)) {
> > + struct drm_bridge *ext_bridge;
> > +
> > + ext_bridge = devm_drm_of_get_bridge(_dsi->pdev->dev,
> > +
> > msm_dsi->pdev->dev.of_node, 1, 0);
> > + if (IS_ERR(ext_bridge))
> > + return PTR_ERR(ext_bridge);
> > +
> > + msm_dsi->next_bridge = ext_bridge;
> > + }
> > +
> > priv->dsi[msm_dsi->id] = msm_dsi;
> >
> > return 0;
> > diff --git a/drivers/gpu/drm/msm/dsi/dsi.h b/drivers/gpu/drm/msm/dsi/dsi.h
> > index 2ad9a842c678..0adef65be1de 100644
> > --- a/drivers/gpu/drm/msm/dsi/dsi.h
> > +++ b/drivers/gpu/drm/msm/dsi/dsi.h
> > @@ -38,6 +38,8 @@ struct msm_dsi {
> > struct mipi_dsi_host *host;
> > struct msm_dsi_phy *phy;
> >
> > + struct drm_bridge *next_bridge;
> > +
> > struct device *phy_dev;
> > bool phy_enabled;
> >
> > diff --git a/drivers/gpu/drm/msm/dsi/dsi_manager.c
> > b/drivers/gpu/drm/msm/dsi/dsi_manager.c
> > index a7c7f85b73e4..b7c52b14c790 100644
> > --- a/drivers/gpu/drm/msm/dsi/dsi_manager.c
> > +++ b/drivers/gpu/drm/msm/dsi/dsi_manager.c
> > @@ -464,18 +464,12 @@ int msm_dsi_manager_ext_bridge_init(u8 id, struct
> > drm_bridge *int_bridge)
> > struct msm_dsi *msm_dsi = dsi_mgr_get_dsi(id);
> > struct drm_device *dev = msm_dsi->dev;
> > struct drm_encoder *encoder;
> > - struct drm_bridge *ext_bridge;
> > struct drm_connector *connector;
> > int ret;
> >
> > - ext_bridge = devm_drm_of_get_bridge(_dsi->pdev->dev,
> > - msm_dsi->pdev->dev.of_node, 1, 0);
> > - if (IS_ERR(ext_bridge))
> > - return PTR_ERR(ext_bridge);
> > -
> > encoder = int_bridge->encoder;
> >
> > - ret = drm_bridge_attach(encoder, ext_bridge, int_bridge,
> > + ret = drm_bridge_attach(encoder, msm_dsi->next_bridge, int_bridge,
> > DRM_BRIDGE_ATTACH_NO_CONNECTOR);
> > if (ret)
> > return ret;
> >
--
With best wishes
Dmitry
Re: [PATCH 2/3] drm/msm/dsi: move next bridge acquisition to dsi_bind
On 3/9/2024 7:09 AM, Dmitry Baryshkov wrote:
Currently the MSM DSI driver looks for the next bridge during
msm_dsi_modeset_init(). If the bridge is not registered at that point,
this might result in -EPROBE_DEFER, which can be problematic that late
during the device probe process. Move next bridge acquisition to the
dsi_bind state so that probe deferral is returned as early as possible.
But msm_dsi_modeset)init() is also called during msm_drm_bind() only.
What issue are you suggesting will be fixed by moving this from
msm_drm_bind() to dsi_bind()?
Signed-off-by: Dmitry Baryshkov
---
drivers/gpu/drm/msm/dsi/dsi.c | 16
drivers/gpu/drm/msm/dsi/dsi.h | 2 ++
drivers/gpu/drm/msm/dsi/dsi_manager.c | 8 +---
3 files changed, 19 insertions(+), 7 deletions(-)
diff --git a/drivers/gpu/drm/msm/dsi/dsi.c b/drivers/gpu/drm/msm/dsi/dsi.c
index 37c4c07005fe..38f10f7a10d3 100644
--- a/drivers/gpu/drm/msm/dsi/dsi.c
+++ b/drivers/gpu/drm/msm/dsi/dsi.c
@@ -120,6 +120,22 @@ static int dsi_bind(struct device *dev, struct device
*master, void *data)
struct msm_drm_private *priv = dev_get_drvdata(master);
struct msm_dsi *msm_dsi = dev_get_drvdata(dev);
+ /*
+* Next bridge doesn't exist for the secondary DSI host in a bonded
+* pair.
+*/
+ if (!msm_dsi_is_bonded_dsi(msm_dsi) ||
+ msm_dsi_is_master_dsi(msm_dsi)) {
+ struct drm_bridge *ext_bridge;
+
+ ext_bridge = devm_drm_of_get_bridge(_dsi->pdev->dev,
+ msm_dsi->pdev->dev.of_node,
1, 0);
+ if (IS_ERR(ext_bridge))
+ return PTR_ERR(ext_bridge);
+
+ msm_dsi->next_bridge = ext_bridge;
+ }
+
priv->dsi[msm_dsi->id] = msm_dsi;
return 0;
diff --git a/drivers/gpu/drm/msm/dsi/dsi.h b/drivers/gpu/drm/msm/dsi/dsi.h
index 2ad9a842c678..0adef65be1de 100644
--- a/drivers/gpu/drm/msm/dsi/dsi.h
+++ b/drivers/gpu/drm/msm/dsi/dsi.h
@@ -38,6 +38,8 @@ struct msm_dsi {
struct mipi_dsi_host *host;
struct msm_dsi_phy *phy;
+ struct drm_bridge *next_bridge;
+
struct device *phy_dev;
bool phy_enabled;
diff --git a/drivers/gpu/drm/msm/dsi/dsi_manager.c b/drivers/gpu/drm/msm/dsi/dsi_manager.c
index a7c7f85b73e4..b7c52b14c790 100644
--- a/drivers/gpu/drm/msm/dsi/dsi_manager.c
+++ b/drivers/gpu/drm/msm/dsi/dsi_manager.c
@@ -464,18 +464,12 @@ int msm_dsi_manager_ext_bridge_init(u8 id, struct
drm_bridge *int_bridge)
struct msm_dsi *msm_dsi = dsi_mgr_get_dsi(id);
struct drm_device *dev = msm_dsi->dev;
struct drm_encoder *encoder;
- struct drm_bridge *ext_bridge;
struct drm_connector *connector;
int ret;
- ext_bridge = devm_drm_of_get_bridge(_dsi->pdev->dev,
- msm_dsi->pdev->dev.of_node, 1, 0);
- if (IS_ERR(ext_bridge))
- return PTR_ERR(ext_bridge);
-
encoder = int_bridge->encoder;
- ret = drm_bridge_attach(encoder, ext_bridge, int_bridge,
+ ret = drm_bridge_attach(encoder, msm_dsi->next_bridge, int_bridge,
DRM_BRIDGE_ATTACH_NO_CONNECTOR);
if (ret)
return ret;
