Exempt 'ports' from the rule which asserts that nodes with single child
node having reg = 0, should not have the '#size-cells' and
'#address-cells' properties.

Ports of certain hardware do need to be described as only having a
single child node 'port@0', especially when hardware has multiple ports,
and the other ports 'port@x' are planned to be added subsequently. In
such cases, just using 'port', would be an inaccurate hardware
description.

For example, Texas Instruments' DSS (display-subsystem), which has 2 or
4 video ports depending on the SoC. Describing the first video port with
just 'port' under ports would be inaccurate and even slightly
misleading. Simply using port@0 (when other ports are not added)
produces the following warning, while making dtbs with W=1 flag set[0].

code-block ::

	Warning (graph_child_address): /bus@100000/dss@4a00000/ports:
	graph node has single child node 'port@0',
	#address-cells/#size-cells are not necessary

Signed-off-by: Aradhya Bhatia <a-bhatia1@ti.com>

[0]: https://lore.kernel.org/all/570903b6-8239-d44a-5fac-71700804cb5d@ti.com/
---
 scripts/dtc/checks.c | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

diff --git a/scripts/dtc/checks.c b/scripts/dtc/checks.c
index 9f31d2607182..705aa0fbcfa2 100644
--- a/scripts/dtc/checks.c
+++ b/scripts/dtc/checks.c
@@ -1797,9 +1797,18 @@ static void check_graph_child_address(struct check *c, struct dt_info *dti,
 		cnt++;
 	}
 
-	if (cnt == 1 && node->addr_cells != -1)
+	if (cnt == 1 && node->addr_cells != -1) {
+		/*
+		 * The graph node "ports" are exempt from this rule, because
+		 * certain hardware do need to be described as only having a
+		 * signle port with reg = 0.
+		 */
+		if (!strcmp(node->name, "ports"))
+			return;
+
 		FAIL(c, dti, node, "graph node has single child node '%s', #address-cells/#size-cells are not necessary",
 		     node->children->name);
+	}
 }
 WARNING(graph_child_address, check_graph_child_address, NULL, &graph_nodes);
 

base-commit: e3b18f7200f45d66f7141136c25554ac1e82009b
-- 
2.40.1

On Fri, Oct 13, 2023 at 02:24:24PM +0530, Aradhya Bhatia wrote:
> Exempt 'ports' from the rule which asserts that nodes with single child
> node having reg = 0, should not have the '#size-cells' and
> '#address-cells' properties.
> 
> Ports of certain hardware do need to be described as only having a
> single child node 'port@0', especially when hardware has multiple ports,
> and the other ports 'port@x' are planned to be added subsequently. In
> such cases, just using 'port', would be an inaccurate hardware
> description.
> 
> For example, Texas Instruments' DSS (display-subsystem), which has 2 or
> 4 video ports depending on the SoC. Describing the first video port with
> just 'port' under ports would be inaccurate and even slightly
> misleading. Simply using port@0 (when other ports are not added)
> produces the following warning, while making dtbs with W=1 flag set[0].

There's a reason this is behind W=1.

In general, if you only have a single 'port' it should be just 'port' 
which is equivalent to port 0. There's exceptions to that, so the 
warning is off by default.

> code-block ::
> 
> 	Warning (graph_child_address): /bus@100000/dss@4a00000/ports:
> 	graph node has single child node 'port@0',
> 	#address-cells/#size-cells are not necessary
> 
> Signed-off-by: Aradhya Bhatia <a-bhatia1@ti.com>
> 
> [0]: https://lore.kernel.org/all/570903b6-8239-d44a-5fac-71700804cb5d@ti.com/
> ---
>  scripts/dtc/checks.c | 11 ++++++++++-
>  1 file changed, 10 insertions(+), 1 deletion(-)

This is a copy of upstream dtc. We don't take patches for it (except in 
emergency). Look at the commit history.

Rob

On 26-Oct-23 01:01, Rob Herring wrote:
> On Fri, Oct 13, 2023 at 02:24:24PM +0530, Aradhya Bhatia wrote:
>> Exempt 'ports' from the rule which asserts that nodes with single child
>> node having reg = 0, should not have the '#size-cells' and
>> '#address-cells' properties.
>>
>> Ports of certain hardware do need to be described as only having a
>> single child node 'port@0', especially when hardware has multiple ports,
>> and the other ports 'port@x' are planned to be added subsequently. In
>> such cases, just using 'port', would be an inaccurate hardware
>> description.
>>
>> For example, Texas Instruments' DSS (display-subsystem), which has 2 or
>> 4 video ports depending on the SoC. Describing the first video port with
>> just 'port' under ports would be inaccurate and even slightly
>> misleading. Simply using port@0 (when other ports are not added)
>> produces the following warning, while making dtbs with W=1 flag set[0].
> 
> There's a reason this is behind W=1.
> 
> In general, if you only have a single 'port' it should be just 'port' 
> which is equivalent to port 0. There's exceptions to that, so the 
> warning is off by default.

Thank you for reviewing the patch, Rob!

I had a discussion offline, and I agree that the patch may not be needed
after all.

Moreover, upon looking at the tests provided in upstream dtc tree, I
also realized that the check is exclusively limited to port@0, and does
not include any random 'child@0'. This makes the patch make a lot less
sense too.

Regards
Aradhya

> 
>> code-block ::
>>
>> 	Warning (graph_child_address): /bus@100000/dss@4a00000/ports:
>> 	graph node has single child node 'port@0',
>> 	#address-cells/#size-cells are not necessary
>>
>> Signed-off-by: Aradhya Bhatia <a-bhatia1@ti.com>
>>
>> [0]: https://lore.kernel.org/all/570903b6-8239-d44a-5fac-71700804cb5d@ti.com/
>> ---
>>  scripts/dtc/checks.c | 11 ++++++++++-
>>  1 file changed, 10 insertions(+), 1 deletion(-)
> 
> This is a copy of upstream dtc. We don't take patches for it (except in 
> emergency). Look at the commit history.
> 
> Rob