Message ID | 20250424-drm-bridge-convert-to-alloc-api-v2-34-8f91a404d86b@bootlin.com |
---|---|
State | Superseded |
Headers | show |
Series | drm: convert all bridges to devm_drm_bridge_alloc() | expand |
On Thu, Apr 24, 2025 at 10:05:49PM +0200, Luca Ceresoli wrote: > This is the new API for allocating DRM bridges. > > The devm lifetime management of this driver is peculiar. The underlying > device for the panel_bridge is the panel, and the devm lifetime is tied the > panel device (panel->dev). However the panel_bridge allocation is not > performed by the panel driver, but rather by a separate entity (typically > the previous bridge in the encoder chain). > > Thus when that separate entoty is destroyed, the panel_bridge is not > removed automatically by devm, so it is rather done explicitly by calling > drm_panel_bridge_remove(). This is the function that does devm_kfree() the > panel_bridge in current code, so update it as well to put the bridge > reference instead. > > Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com> This looks fine, but we need a TODO entry to clean this up later on, and a comment on devm_drm_put_bridge that this is inherently unsafe and must not be used. Maxime
Hi Maxime, On Mon, 28 Apr 2025 13:39:23 +0200 Maxime Ripard <mripard@kernel.org> wrote: > On Thu, Apr 24, 2025 at 10:05:49PM +0200, Luca Ceresoli wrote: > > This is the new API for allocating DRM bridges. > > > > The devm lifetime management of this driver is peculiar. The underlying > > device for the panel_bridge is the panel, and the devm lifetime is tied the > > panel device (panel->dev). However the panel_bridge allocation is not > > performed by the panel driver, but rather by a separate entity (typically > > the previous bridge in the encoder chain). > > > > Thus when that separate entoty is destroyed, the panel_bridge is not > > removed automatically by devm, so it is rather done explicitly by calling > > drm_panel_bridge_remove(). This is the function that does devm_kfree() the > > panel_bridge in current code, so update it as well to put the bridge > > reference instead. > > > > Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com> > > This looks fine, but we need a TODO entry to clean this up later on, and > a comment on devm_drm_put_bridge that this is inherently unsafe and > must not be used. Ah, I see, OK. Quick draft: /** * devm_drm_put_bridge - Release a bridge reference obtained via devm * @dev: device that got the bridge via devm * @bridge: pointer to a struct drm_bridge obtained via devm * * Same as drm_bridge_put() for bridge pointers obtained via devm functions * such as devm_drm_bridge_alloc(). + * + * This function is a temporary workaround and MUST NOT be used. Manual + * handling of bridge lifetime is inherently unsafe. */ and: - devm_kfree(panel_bridge->panel->dev, bridge); + /* TODO remove this after reworking panel_bridge lifetime */ + devm_drm_put_bridge(panel_bridge->panel->dev, bridge); } Does it look good enough? Luca
On Mon, Apr 28, 2025 at 05:25:16PM +0200, Luca Ceresoli wrote: > Hi Maxime, > > On Mon, 28 Apr 2025 13:39:23 +0200 > Maxime Ripard <mripard@kernel.org> wrote: > > > On Thu, Apr 24, 2025 at 10:05:49PM +0200, Luca Ceresoli wrote: > > > This is the new API for allocating DRM bridges. > > > > > > The devm lifetime management of this driver is peculiar. The underlying > > > device for the panel_bridge is the panel, and the devm lifetime is tied the > > > panel device (panel->dev). However the panel_bridge allocation is not > > > performed by the panel driver, but rather by a separate entity (typically > > > the previous bridge in the encoder chain). > > > > > > Thus when that separate entoty is destroyed, the panel_bridge is not > > > removed automatically by devm, so it is rather done explicitly by calling > > > drm_panel_bridge_remove(). This is the function that does devm_kfree() the > > > panel_bridge in current code, so update it as well to put the bridge > > > reference instead. > > > > > > Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com> > > > > This looks fine, but we need a TODO entry to clean this up later on, and > > a comment on devm_drm_put_bridge that this is inherently unsafe and > > must not be used. > > Ah, I see, OK. > > Quick draft: > > /** > * devm_drm_put_bridge - Release a bridge reference obtained via devm > * @dev: device that got the bridge via devm > * @bridge: pointer to a struct drm_bridge obtained via devm > * > * Same as drm_bridge_put() for bridge pointers obtained via devm functions > * such as devm_drm_bridge_alloc(). > + * > + * This function is a temporary workaround and MUST NOT be used. Manual > + * handling of bridge lifetime is inherently unsafe. > */ That part looks good to me > and: > > - devm_kfree(panel_bridge->panel->dev, bridge); > + /* TODO remove this after reworking panel_bridge lifetime */ > + devm_drm_put_bridge(panel_bridge->panel->dev, bridge); > } > > Does it look good enough? That too, but I was talking about an entry in https://www.kernel.org/doc/html/latest/gpu/todo.html Maxime
On Mon, 5 May 2025 08:23:26 +0200 Maxime Ripard <mripard@kernel.org> wrote: > On Mon, Apr 28, 2025 at 05:25:16PM +0200, Luca Ceresoli wrote: > > Hi Maxime, > > > > On Mon, 28 Apr 2025 13:39:23 +0200 > > Maxime Ripard <mripard@kernel.org> wrote: > > > > > On Thu, Apr 24, 2025 at 10:05:49PM +0200, Luca Ceresoli wrote: > > > > This is the new API for allocating DRM bridges. > > > > > > > > The devm lifetime management of this driver is peculiar. The underlying > > > > device for the panel_bridge is the panel, and the devm lifetime is tied the > > > > panel device (panel->dev). However the panel_bridge allocation is not > > > > performed by the panel driver, but rather by a separate entity (typically > > > > the previous bridge in the encoder chain). > > > > > > > > Thus when that separate entoty is destroyed, the panel_bridge is not > > > > removed automatically by devm, so it is rather done explicitly by calling > > > > drm_panel_bridge_remove(). This is the function that does devm_kfree() the > > > > panel_bridge in current code, so update it as well to put the bridge > > > > reference instead. > > > > > > > > Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com> > > > > > > This looks fine, but we need a TODO entry to clean this up later on, and > > > a comment on devm_drm_put_bridge that this is inherently unsafe and > > > must not be used. > > > > Ah, I see, OK. > > > > Quick draft: > > > > /** > > * devm_drm_put_bridge - Release a bridge reference obtained via devm > > * @dev: device that got the bridge via devm > > * @bridge: pointer to a struct drm_bridge obtained via devm > > * > > * Same as drm_bridge_put() for bridge pointers obtained via devm functions > > * such as devm_drm_bridge_alloc(). > > + * > > + * This function is a temporary workaround and MUST NOT be used. Manual > > + * handling of bridge lifetime is inherently unsafe. > > */ > > That part looks good to me > > > and: > > > > - devm_kfree(panel_bridge->panel->dev, bridge); > > + /* TODO remove this after reworking panel_bridge lifetime */ > > + devm_drm_put_bridge(panel_bridge->panel->dev, bridge); > > } > > > > Does it look good enough? > > That too, but I was talking about an entry in > https://www.kernel.org/doc/html/latest/gpu/todo.html Ah, sure! I queued this for v3, if you have better suggestions don't hesitate to let me know: -----8<----- Remove devm_drm_put_bridge() ---------------------------- Due to how the panel bridge handles the drm_bridge object lifetime, special care must be taken to dispose of the drm_bridge object when the panel_bridge is removed. This is currently managed using devm_drm_put_bridge(), but that is an unsafe, temporary workaround. To fix that, the DRM panel lifetime needs to be reworked. After the rework is done, remove devm_drm_put_bridge() and the TODO in drm_panel_bridge_remove(). Contact: Maxime Ripard <mripard@kernel.org>, Luca Ceresoli <luca.ceresoli@bootlin.com> Level: Intermediate -----8<----- Luca
diff --git a/drivers/gpu/drm/bridge/panel.c b/drivers/gpu/drm/bridge/panel.c index 79b009ab9396048eac57ad47631a902e949d77c6..ddd1e91970d09b93aa64f50cd9155939a12a2c6f 100644 --- a/drivers/gpu/drm/bridge/panel.c +++ b/drivers/gpu/drm/bridge/panel.c @@ -287,15 +287,14 @@ struct drm_bridge *drm_panel_bridge_add_typed(struct drm_panel *panel, if (!panel) return ERR_PTR(-EINVAL); - panel_bridge = devm_kzalloc(panel->dev, sizeof(*panel_bridge), - GFP_KERNEL); - if (!panel_bridge) - return ERR_PTR(-ENOMEM); + panel_bridge = devm_drm_bridge_alloc(panel->dev, struct panel_bridge, bridge, + &panel_bridge_bridge_funcs); + if (IS_ERR(panel_bridge)) + return (void *)panel_bridge; panel_bridge->connector_type = connector_type; panel_bridge->panel = panel; - panel_bridge->bridge.funcs = &panel_bridge_bridge_funcs; panel_bridge->bridge.of_node = panel->dev->of_node; panel_bridge->bridge.ops = DRM_BRIDGE_OP_MODES; panel_bridge->bridge.type = connector_type; @@ -327,7 +326,7 @@ void drm_panel_bridge_remove(struct drm_bridge *bridge) panel_bridge = drm_bridge_to_panel_bridge(bridge); drm_bridge_remove(bridge); - devm_kfree(panel_bridge->panel->dev, bridge); + devm_drm_put_bridge(panel_bridge->panel->dev, bridge); } EXPORT_SYMBOL(drm_panel_bridge_remove);
This is the new API for allocating DRM bridges. The devm lifetime management of this driver is peculiar. The underlying device for the panel_bridge is the panel, and the devm lifetime is tied the panel device (panel->dev). However the panel_bridge allocation is not performed by the panel driver, but rather by a separate entity (typically the previous bridge in the encoder chain). Thus when that separate entoty is destroyed, the panel_bridge is not removed automatically by devm, so it is rather done explicitly by calling drm_panel_bridge_remove(). This is the function that does devm_kfree() the panel_bridge in current code, so update it as well to put the bridge reference instead. Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com> --- drivers/gpu/drm/bridge/panel.c | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-)