Re: [PATCH v3 3/3] media: mtk-mdp: Fix mdp device tree

From: Matthias Brugger
Date: Mon May 15 2017 - 03:28:21 EST




On 15/05/17 04:31, Minghsiu Tsai wrote:
On Fri, 2017-05-12 at 17:05 +0200, Matthias Brugger wrote:

On 12/05/17 05:22, Minghsiu Tsai wrote:
From: Daniel Kurtz <djkurtz@xxxxxxxxxxxx>

If the mdp_* nodes are under an mdp sub-node, their corresponding
platform device does not automatically get its iommu assigned properly.

Fix this by moving the mdp component nodes up a level such that they are
siblings of mdp and all other SoC subsystems. This also simplifies the
device tree.

Although it fixes iommu assignment issue, it also break compatibility
with old device tree. So, the patch in driver is needed to iterate over
sibling mdp device nodes, not child ones, to keep driver work properly.


Couldn't we preserve backwards compatibility by doing something like this:
diff --git a/drivers/media/platform/mtk-mdp/mtk_mdp_core.c
b/drivers/media/platform/mtk-mdp/mtk_mdp_core.c
index 9e4eb7dcc424..277d8fe6eb76 100644
--- a/drivers/media/platform/mtk-mdp/mtk_mdp_core.c
+++ b/drivers/media/platform/mtk-mdp/mtk_mdp_core.c
@@ -103,7 +103,7 @@ static int mtk_mdp_probe(struct platform_device *pdev)
{
struct mtk_mdp_dev *mdp;
struct device *dev = &pdev->dev;
- struct device_node *node;
+ struct device_node *node, *parent;
int i, ret = 0;

mdp = devm_kzalloc(dev, sizeof(*mdp), GFP_KERNEL);
@@ -117,8 +117,14 @@ static int mtk_mdp_probe(struct platform_device *pdev)
mutex_init(&mdp->lock);
mutex_init(&mdp->vpulock);

+ /* Old dts had the components as child nodes */
+ if (of_get_next_child(dev->of_node, NULL))
+ parent = dev->of_node;
+ else
+ parent = dev->of_node->parent;
+
/* Iterate over sibling MDP function blocks */
- for_each_child_of_node(dev->of_node, node) {
+ for_each_child_of_node(parent, node) {
const struct of_device_id *of_id;
enum mtk_mdp_comp_type comp_type;
int comp_id;

Maybe even by putting a warning in the if branch to make sure, people
are aware of their out-of-date device tree blobs.

Regards,
Matthias


Hi Matthias,

It is a good idea to do compatible in such a way and put a warning the
device tree is out of date. People can find out cause soon if device
tree is old.

I modify the code as below:

+ /* Old dts had the components as child nodes */
+ if (of_get_next_child(dev->of_node, NULL)) {
+ parent = dev->of_node;
+ dev_warn(dev, "device tree is out of date\n");
+ } else {
+ parent = dev->of_node->parent;
+ }

Will you upload it in a separate patch?
If not, I can merge it in my patch series and upload v4.


Please integrate it into your patch series.

Mauro, are you ok with the dev_warn about the out-of-date device-tree?

Regards,
Matthias



Best Regards,

Ming Hsiu

Signed-off-by: Daniel Kurtz <djkurtz@xxxxxxxxxxxx>
Signed-off-by: Minghsiu Tsai <minghsiu.tsai@xxxxxxxxxxxx>

---
drivers/media/platform/mtk-mdp/mtk_mdp_core.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/media/platform/mtk-mdp/mtk_mdp_core.c b/drivers/media/platform/mtk-mdp/mtk_mdp_core.c
index 9e4eb7d..a5ad586 100644
--- a/drivers/media/platform/mtk-mdp/mtk_mdp_core.c
+++ b/drivers/media/platform/mtk-mdp/mtk_mdp_core.c
@@ -118,7 +118,7 @@ static int mtk_mdp_probe(struct platform_device *pdev)
mutex_init(&mdp->vpulock);
/* Iterate over sibling MDP function blocks */
- for_each_child_of_node(dev->of_node, node) {
+ for_each_child_of_node(dev->of_node->parent, node) {
const struct of_device_id *of_id;
enum mtk_mdp_comp_type comp_type;
int comp_id;