The encoder drivers use drm_of_find_possible_crtcs to get upstream crtcs from the device tree using of_graph. For the results to be correct, encoders must be probed/bound after _all_ crtcs have been created. The existing code uses a depth first recursive traversal of the of_graph, which means the encoders downstream of the TCON get add right after the first TCON. The second TCON or CRTC will never be properly associated with encoders connected to it.

Other platforms, such as Rockchip, deal with this by probing all CRTCs first, then all subsequent components. This is easy to do since the CRTCs correspond to just one device node, and are the first nodes in the pipeline.

However with Allwinner SoCs, the function of the CRTC is split between the display backend (DE 1.0) or mixer (DE 2.0), which does scan-out and compositing, and the TCON, which generates the display timing signals. Further complicating the process, there may be a Dynamic Range Controller between the backend and the TCON. Also, the backend is preceded by the frontend, with a Display Enhancement Unit possibly in between.

In a dual display pipeline setup, both frontends can feed either backend, and both backends can feed either TCON. We want all components of the same type to be added before the next type in the pipeline. Fortunately, the pipelines are perfectly symmetric, i.e. components of the same type are at the same depth when counted from the frontend. The only exception is the third pipeline in the A80 SoC, which we do not support anyway.

Hence we can use a breadth first search traversal order to add components. We do not need to check for duplicates. The component matching system handles this for us.

Signed-off-by: Chen-Yu Tsai wens@csie.org --- drivers/gpu/drm/sun4i/sun4i_drv.c | 81 ++++++++++++++++++++++++++++++++++----- 1 file changed, 72 insertions(+), 9 deletions(-)

diff --git a/drivers/gpu/drm/sun4i/sun4i_drv.c b/drivers/gpu/drm/sun4i/sun4i_drv.c index ace59651892f..2ff4233fe2da 100644 --- a/drivers/gpu/drm/sun4i/sun4i_drv.c +++ b/drivers/gpu/drm/sun4i/sun4i_drv.c @@ -200,11 +200,39 @@ static int compare_of(struct device *dev, void *data) return dev->of_node == data; }

+/* + * The encoder drivers use drm_of_find_possible_crtcs to get upstream + * crtcs from the device tree using of_graph. For the results to be + * correct, encoders must be probed/bound after _all_ crtcs have been + * created. The existing code uses a depth first recursive traversal + * of the of_graph, which means the encoders downstream of the TCON + * get add right after the first TCON. The second TCON or CRTC will + * never be properly associated with encoders connected to it. + * + * Also, in a dual display pipeline setup, both frontends can feed + * either backend, and both backends can feed either TCON, we want + * all components of the same type to be added before the next type + * in the pipeline. Fortunately, the pipelines are perfectly symmetric, + * i.e. components of the same type are at the same depth when counted + * from the frontend. The only exception is the third pipeline in + * the A80 SoC, which we do not support anyway. + * + * Hence we can use a breadth first search traversal order to add + * components. We do not need to check for duplicates. The component + * matching system handles this for us. + */ +struct endpoint_list { + struct device_node *node; + struct list_head list; +}; + static int sun4i_drv_add_endpoints(struct device *dev, + struct list_head *endpoints, struct component_match **match, struct device_node *node) { struct device_node *port, *ep, *remote; + struct endpoint_list *endpoint; int count = 0;

/* @@ -264,10 +292,15 @@ static int sun4i_drv_add_endpoints(struct device *dev, } }

- /* Walk down our tree */ - count += sun4i_drv_add_endpoints(dev, match, remote); + /* Add downstream nodes to the queue */ + endpoint = kzalloc(sizeof(*endpoint), GFP_KERNEL); + if (!endpoint) { + of_node_put(remote); + return -ENOMEM; + }

- of_node_put(remote); + endpoint->node = remote; + list_add_tail(&endpoint->list, endpoints); }

return count; @@ -277,7 +310,9 @@ static int sun4i_drv_probe(struct platform_device *pdev) { struct component_match *match = NULL; struct device_node *np = pdev->dev.of_node; - int i, count = 0; + struct endpoint_list *endpoint, *endpoint_temp; + int i, ret, count = 0; + LIST_HEAD(endpoints);

for (i = 0;; i++) { struct device_node *pipeline = of_parse_phandle(np, @@ -286,12 +321,31 @@ static int sun4i_drv_probe(struct platform_device *pdev) if (!pipeline) break;

- count += sun4i_drv_add_endpoints(&pdev->dev, &match, - pipeline); - of_node_put(pipeline); + endpoint = kzalloc(sizeof(*endpoint), GFP_KERNEL); + if (!endpoint) { + ret = -ENOMEM; + goto err_free_endpoints; + } + + endpoint->node = pipeline; + list_add_tail(&endpoint->list, &endpoints); + } + + list_for_each_entry_safe(endpoint, endpoint_temp, &endpoints, list) { + /* process this endpoint */ + ret = sun4i_drv_add_endpoints(&pdev->dev, &endpoints, &match, + endpoint->node); + + /* sun4i_drv_add_endpoints can fail to allocate memory */ + if (ret < 0) + goto err_free_endpoints; + + count += ret;

- DRM_DEBUG_DRIVER("Queued %d outputs on pipeline %d\n", - count, i); + /* delete and cleanup the current entry */ + list_del(&endpoint->list); + of_node_put(endpoint->node); + kfree(endpoint); }

if (count) @@ -300,6 +354,15 @@ static int sun4i_drv_probe(struct platform_device *pdev) match); else return 0; + +err_free_endpoints: + list_for_each_entry_safe(endpoint, endpoint_temp, &endpoints, list) { + list_del(&endpoint->list); + of_node_put(endpoint->node); + kfree(endpoint); + } + + return ret; }

static int sun4i_drv_remove(struct platform_device *pdev)

The device tree binding for sun4i-drm says:

For all connections between components up to the TCONs in the display pipeline, when there are multiple components of the same type at the same depth, the local endpoint ID must be the same as the remote component's index. For example, if the remote endpoint is Frontend 1, then the local endpoint ID must be 1.

We should be able to get the TCON's ID directly from any of the remote endpoints from its input port. With the ID, we can then go through the list of registered engines and find a matching one by ID.

However the A31 device tree is incorrect. We assumed that there were no cross pipeline connections between the backends and TCONs. As a result, in all the endpoints between the backends and TCONs of the second display pipeline, the endpoint IDs were incorrectly set to 0, when in fact they should've been set to 1.

To maintain compatibility with this incorrect device tree, we first check if the TCON's input port has 1 or many endpoints. If there are more than 1, then it is likely a fixed version, and we can proceed with the new method. If there is only 1 endpoint, then it is possibly an incorrect version, or it could be the SoC only has one pipeline. In either case we fall back to using the old method of traversing the input connections to find a matching engine, and then get its ID.

Signed-off-by: Chen-Yu Tsai wens@csie.org --- drivers/gpu/drm/sun4i/sun4i_tcon.c | 121 ++++++++++++++++++++++++++++++++++++- 1 file changed, 118 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/sun4i/sun4i_tcon.c b/drivers/gpu/drm/sun4i/sun4i_tcon.c index 065654dbfb2c..33c20d2f9fb1 100644 --- a/drivers/gpu/drm/sun4i/sun4i_tcon.c +++ b/drivers/gpu/drm/sun4i/sun4i_tcon.c @@ -463,8 +463,9 @@ static int sun4i_tcon_init_regmap(struct device *dev, * function in fact searches the corresponding engine, and the ID is * requested via the get_id function of the engine. */ -static struct sunxi_engine *sun4i_tcon_find_engine(struct sun4i_drv *drv, - struct device_node *node) +static struct sunxi_engine * +sun4i_tcon_find_engine_traverse(struct sun4i_drv *drv, + struct device_node *node) { struct device_node *port, *ep, *remote; struct sunxi_engine *engine; @@ -502,7 +503,7 @@ static struct sunxi_engine *sun4i_tcon_find_engine(struct sun4i_drv *drv, }

/* keep looking through upstream ports */ - engine = sun4i_tcon_find_engine(drv, remote); + engine = sun4i_tcon_find_engine_traverse(drv, remote); if (!IS_ERR(engine)) { of_node_put(remote); of_node_put(port); @@ -513,6 +514,120 @@ static struct sunxi_engine *sun4i_tcon_find_engine(struct sun4i_drv *drv, return ERR_PTR(-EINVAL); }

+/* + * The device tree binding says that the remote endpoint ID of any + * connection between components, up to and including the TCON, of + * the display pipeline should be equal to the actual ID of the local + * component. Thus we can look at any one of the input connections of + * the TCONs, and use that connection's remote endpoint ID as our own. + * + * Since the user of this function already finds the input port, + * the port is passed in directly without further checks. + */ +static int sun4i_tcon_of_get_id_from_port(struct device_node *port) +{ + struct device_node *ep; + int ret = -EINVAL; + + /* try finding an upstream endpoint */ + for_each_available_child_of_node(port, ep) { + struct device_node *remote; + u32 reg; + + remote = of_graph_get_remote_endpoint(ep); + if (!remote) + continue; + + ret = of_property_read_u32(remote, "reg", &reg); + if (ret) + continue; + + ret = reg; + } + + return ret; +} + +/* + * Once we know the TCON's id, we can look through the list of + * engines to find a matching one. We assume all engines have + * been probed and added to the list. + */ +static struct sunxi_engine *sun4i_tcon_get_engine_by_id(struct sun4i_drv *drv, + int id) +{ + struct sunxi_engine *engine; + + list_for_each_entry(engine, &drv->engine_list, list) + if (engine->id == id) + return engine; + + return ERR_PTR(-EINVAL); +} + +/* + * On SoCs with the old display pipeline design (Display Engine 1.0), + * we assumed the TCON was always tied to just one backend. However + * this proved not to be the case. On the A31, the TCON can select + * either backend as its source. On the A20 (and likely on the A10), + * the backend can choose which TCON to output to. + * + * The device tree binding says that the remote endpoint ID of any + * connection between components, up to and including the TCON, of + * the display pipeline should be equal to the actual ID of the local + * component. Thus we should be able to look at any one of the input + * connections of the TCONs, and use that connection's remote endpoint + * ID as our own. + * + * However the connections between the backend and TCON were assumed + * to be always singular, and their endpoit IDs were all incorrectly + * set to 0. This means for these old device trees, we cannot just look + * up the remote endpoint ID of a TCON input endpoint. TCON1 would be + * incorrectly identified as TCON0. + * + * This function first checks if the TCON node has 2 input endpoints. + * If so, then the device tree is a corrected version, and it will use + * sun4i_tcon_of_get_id() and sun4i_tcon_get_engine_by_id() from above + * to fetch the ID and engine directly. If not, then it is likely an + * old device trees, where the endpoint IDs were incorrect, but did not + * have endpoint connections between the backend and TCON across + * different display pipelines. It will fall back to the old method of + * traversing the of_graph to try and find a matching engine by device + * node. + * + * In the case of single display pipeline device trees, either method + * works. + */ +static struct sunxi_engine *sun4i_tcon_find_engine(struct sun4i_drv *drv, + struct device_node *node) +{ + struct device_node *port; + struct sunxi_engine *engine; + + port = of_graph_get_port_by_id(node, 0); + if (!port) + return ERR_PTR(-EINVAL); + + /* + * Is this a corrected device tree with cross pipeline + * connections between the backend and TCON? + */ + if (of_get_child_count(port) > 1) { + /* Get our ID directly from an upstream endpoint */ + int id = sun4i_tcon_of_get_id_from_port(port); + + /* Get our engine by matching our ID */ + engine = sun4i_tcon_get_engine_by_id(drv, id); + + of_node_put(port); + return engine; + } + + /* Fallback to old method by traversing input endpoints */ + of_node_put(port); + return sun4i_tcon_find_engine_traverse(drv, node); +} + static int sun4i_tcon_bind(struct device *dev, struct device *master, void *data) {

Now that sun4i_tcon_find_engine_traverse() usage is restricted to the single input case, we can remove the for_each_available_child_of_node loop.

While at it, consolidate all the of_node_put calls into a common exit path.

Signed-off-by: Chen-Yu Tsai wens@csie.org --- drivers/gpu/drm/sun4i/sun4i_tcon.c | 51 +++++++++++++++++++------------------- 1 file changed, 25 insertions(+), 26 deletions(-)

diff --git a/drivers/gpu/drm/sun4i/sun4i_tcon.c b/drivers/gpu/drm/sun4i/sun4i_tcon.c index 33c20d2f9fb1..e9ab03f0bf6e 100644 --- a/drivers/gpu/drm/sun4i/sun4i_tcon.c +++ b/drivers/gpu/drm/sun4i/sun4i_tcon.c @@ -468,7 +468,7 @@ sun4i_tcon_find_engine_traverse(struct sun4i_drv *drv, struct device_node *node) { struct device_node *port, *ep, *remote; - struct sunxi_engine *engine; + struct sunxi_engine *engine = ERR_PTR(-EINVAL);

port = of_graph_get_port_by_id(node, 0); if (!port) @@ -483,35 +483,34 @@ sun4i_tcon_find_engine_traverse(struct sun4i_drv *drv, * * Bail out if there are multiple input connections. */ - if (of_get_available_child_count(port) != 1) { - of_node_put(port); - return ERR_PTR(-EINVAL); - } + if (of_get_available_child_count(port) != 1) + goto out_put_port;

- for_each_available_child_of_node(port, ep) { - remote = of_graph_get_remote_port_parent(ep); - if (!remote) - continue; + /* Get the first connection without specifying an ID */ + ep = of_get_next_available_child(port, NULL); + if (!ep) + goto out_put_port;

- /* does this node match any registered engines? */ - list_for_each_entry(engine, &drv->engine_list, list) { - if (remote == engine->node) { - of_node_put(remote); - of_node_put(port); - return engine; - } - } + remote = of_graph_get_remote_port_parent(ep); + if (!remote) + goto out_put_ep;

- /* keep looking through upstream ports */ - engine = sun4i_tcon_find_engine_traverse(drv, remote); - if (!IS_ERR(engine)) { - of_node_put(remote); - of_node_put(port); - return engine; - } - } + /* does this node match any registered engines? */ + list_for_each_entry(engine, &drv->engine_list, list) + if (remote == engine->node) + goto out_put_remote;

- return ERR_PTR(-EINVAL); + /* keep looking through upstream ports */ + engine = sun4i_tcon_find_engine_traverse(drv, remote); + +out_put_remote: + of_node_put(remote); +out_put_ep: + of_node_put(ep); +out_put_port: + of_node_put(port); + + return engine; }

/*

If we want to have vblank on both pipelines at the same time, we need to call drm_vblank_init with num_crtcs = 2.

Instead, since the crtc init calls correctly set mode_config.num_crtc, we can move the drm_vblank_init call to after the crtc init code is called, which is the component bind part. Then we can just pass mode_config.num_crtc in.

Signed-off-by: Chen-Yu Tsai wens@csie.org --- drivers/gpu/drm/sun4i/sun4i_drv.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/drivers/gpu/drm/sun4i/sun4i_drv.c b/drivers/gpu/drm/sun4i/sun4i_drv.c index 2ff4233fe2da..a2012638d5f7 100644 --- a/drivers/gpu/drm/sun4i/sun4i_drv.c +++ b/drivers/gpu/drm/sun4i/sun4i_drv.c @@ -106,11 +106,6 @@ static int sun4i_drv_bind(struct device *dev) goto free_drm; }

- /* drm_vblank_init calls kcalloc, which can fail */ - ret = drm_vblank_init(drm, 1); - if (ret) - goto free_mem_region; - drm_mode_config_init(drm);

ret = component_bind_all(drm->dev, drm); @@ -119,6 +114,11 @@ static int sun4i_drv_bind(struct device *dev) goto cleanup_mode_config; }

+ /* drm_vblank_init calls kcalloc, which can fail */ + ret = drm_vblank_init(drm, drm->mode_config.num_crtc); + if (ret) + goto free_mem_region; + drm->irq_enabled = true;

/* Remove early framebuffers (ie. simplefb) */