Skip to contentSkip to conversation

platform: set of_node in platform_device_register_full()

PATCH In review Mans Rullgard 3 participants 15 messages Feb 16, 2019 – Feb 20, 2019 Driver core

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 platform: set of_node in platform_device_register_full()

If the provided fwnode is an OF node, set dev.of_node as well.

---
 drivers/base/platform.c | 1 +
 1 file changed, 1 insertion(+)
drivers/base/platform.c +1 -0
index dff82a3c2caa..853a1d0e5845 100644
@@ -512,6 +512,7 @@ struct platform_device *platform_device_register_full(
 
 	pdev->dev.parent = pdevinfo->parent;
 	pdev->dev.fwnode = pdevinfo->fwnode;
+	pdev->dev.of_node = of_node_get(to_of_node(pdev->dev.fwnode));
 
 	if (pdevinfo->dma_mask) {
 		/*
Signature · 1 line
2.20.1
#1 lore raw
Quoted 13 lines · Mans Rullgard

If the provided fwnode is an OF node, set dev.of_node as well.

---
 drivers/base/platform.c | 1 +
 1 file changed, 1 insertion(+)
drivers/base/platform.c +0 -0
index dff82a3c2caa..853a1d0e5845 100644

drivers/base/platform.c struct platform_device *platform_device_register_full( · line 515

Quoted 4 lines
drivers/base/platform.c +1 -0
@@ -512,6 +512,7 @@ struct platform_device *platform_device_register_full(
 
        pdev->dev.parent = pdevinfo->parent;
        pdev->dev.fwnode = pdevinfo->fwnode;
+       pdev->dev.of_node = of_node_get(to_of_node(pdev->dev.fwnode));

of_node_get() generally does a kobject_get() on the node's kobject, so when is that reference dropped? Or if it doesn't need to be dropped at all, why is this the case?

Quoted 6 lines
drivers/base/platform.c +0 -1
        if (pdevinfo->dma_mask) {
                /*
--

2.20.1

#2 lore raw
Quoted 15 lines · Rafael J. Wysocki
Quoted 14 lines · Mans Rullgard

If the provided fwnode is an OF node, set dev.of_node as well.

---
 drivers/base/platform.c | 1 +
 1 file changed, 1 insertion(+)
drivers/base/platform.c +0 -0
index dff82a3c2caa..853a1d0e5845 100644
@@ -512,6 +512,7 @@ struct platform_device *platform_device_register_full(
 
Quoted 7 lines

drivers/base/platform.c struct platform_device *platform_device_register_full( · line 515

Quoted 3 lines
drivers/base/platform.c +1 -0
        pdev->dev.parent = pdevinfo->parent;
        pdev->dev.fwnode = pdevinfo->fwnode;
+       pdev->dev.of_node = of_node_get(to_of_node(pdev->dev.fwnode));

of_node_get() generally does a kobject_get() on the node's kobject, so when is that reference dropped? Or if it doesn't need to be dropped at all, why is this the case?

platform_device_release() calls of_device_node_put().

Signature · 1 line
Måns Rullgård
#3 lore raw
Quoted 26 lines · Måns Rullgård
Quoted 15 lines · Rafael J. Wysocki
Quoted 14 lines · Mans Rullgard

If the provided fwnode is an OF node, set dev.of_node as well.

---
 drivers/base/platform.c | 1 +
 1 file changed, 1 insertion(+)
drivers/base/platform.c +0 -0
index dff82a3c2caa..853a1d0e5845 100644
@@ -512,6 +512,7 @@ struct platform_device *platform_device_register_full(
 
Quoted 7 lines

drivers/base/platform.c struct platform_device *platform_device_register_full( · line 515

Quoted 3 lines
drivers/base/platform.c +1 -0
        pdev->dev.parent = pdevinfo->parent;
        pdev->dev.fwnode = pdevinfo->fwnode;
+       pdev->dev.of_node = of_node_get(to_of_node(pdev->dev.fwnode));

of_node_get() generally does a kobject_get() on the node's kobject, so when is that reference dropped? Or if it doesn't need to be dropped at all, why is this the case?

Quoted 1 line

platform_device_release() calls of_device_node_put().

Yes, it does, but this is the reference that's already acquired for devices added while parsing DT, isn't it?

Your change adds an extra reference AFAICS.

Also, why is this patch needed?

#4 lore raw
Quoted 27 lines · Rafael J. Wysocki
Quoted 26 lines · Måns Rullgård
Quoted 22 lines · Rafael J. Wysocki
Quoted 13 lines · Mans Rullgard

If the provided fwnode is an OF node, set dev.of_node as well.

---
 drivers/base/platform.c | 1 +
 1 file changed, 1 insertion(+)
drivers/base/platform.c +0 -0
index dff82a3c2caa..853a1d0e5845 100644

drivers/base/platform.c struct platform_device *platform_device_register_full( · line 515

Quoted 4 lines
drivers/base/platform.c +1 -0
@@ -512,6 +512,7 @@ struct platform_device *platform_device_register_full(
 
        pdev->dev.parent = pdevinfo->parent;
        pdev->dev.fwnode = pdevinfo->fwnode;
+       pdev->dev.of_node = of_node_get(to_of_node(pdev->dev.fwnode));

of_node_get() generally does a kobject_get() on the node's kobject, so when is that reference dropped? Or if it doesn't need to be dropped at all, why is this the case?

Quoted 8 lines
Quoted 1 line

platform_device_release() calls of_device_node_put().

Yes, it does, but this is the reference that's already acquired for devices added while parsing DT, isn't it?

Your change adds an extra reference AFAICS.

Also, why is this patch needed?

Some drivers are just shims that create extra "glue" devices with the DT device as parent and have the real driver bind to these. In other cases, the same real driver matches the DT node directly. When a glue device is used, it needs to get a reference to the original DT node in order for the main driver to access properties and child nodes.

Right now, my problem is that the suxi-musb driver creates such a glue device for the musb core driver to bind to without setting of_node. This means devices attached to this USB interface don't get associated with DT nodes, if present, the way they do with EHCI.

The sunxi-musb driver uses platform_device_register_full(), so this seemed like the easiest way to let it set of_node of the new device. Since this creates a second reference to the same node, of_node_get() is required.

If you'd prefer solving this in some other way, please tell me how.

Signature · 1 line
Måns Rullgård
#5 lore raw
Quoted 45 lines · Måns Rullgård
Quoted 27 lines · Rafael J. Wysocki
Quoted 26 lines · Måns Rullgård
Quoted 22 lines · Rafael J. Wysocki
Quoted 13 lines · Mans Rullgard

If the provided fwnode is an OF node, set dev.of_node as well.

---
 drivers/base/platform.c | 1 +
 1 file changed, 1 insertion(+)
drivers/base/platform.c +0 -0
index dff82a3c2caa..853a1d0e5845 100644

drivers/base/platform.c struct platform_device *platform_device_register_full( · line 515

Quoted 4 lines
drivers/base/platform.c +1 -0
@@ -512,6 +512,7 @@ struct platform_device *platform_device_register_full(
 
        pdev->dev.parent = pdevinfo->parent;
        pdev->dev.fwnode = pdevinfo->fwnode;
+       pdev->dev.of_node = of_node_get(to_of_node(pdev->dev.fwnode));

of_node_get() generally does a kobject_get() on the node's kobject, so when is that reference dropped? Or if it doesn't need to be dropped at all, why is this the case?

Quoted 8 lines
Quoted 1 line

platform_device_release() calls of_device_node_put().

Yes, it does, but this is the reference that's already acquired for devices added while parsing DT, isn't it?

Your change adds an extra reference AFAICS.

Also, why is this patch needed?

Some drivers are just shims that create extra "glue" devices with the DT device as parent and have the real driver bind to these. In other cases, the same real driver matches the DT node directly. When a glue device is used, it needs to get a reference to the original DT node in order for the main driver to access properties and child nodes.

Quoted 4 lines

Right now, my problem is that the suxi-musb driver creates such a glue device for the musb core driver to bind to without setting of_node. This means devices attached to this USB interface don't get associated with DT nodes, if present, the way they do with EHCI.

You really should describe problems that you want to address in patch changelogs. This helps a lot to understand the motivation for the changes.

Quoted 4 lines

The sunxi-musb driver uses platform_device_register_full(), so this seemed like the easiest way to let it set of_node of the new device. Since this creates a second reference to the same node, of_node_get() is required.

But what about devices that already have of_node set at this point?

Maybe check if of_node is NULL before trying to set it?

#6 lore raw
Quoted 46 lines · Rafael J. Wysocki
Quoted 45 lines · Måns Rullgård
Quoted 27 lines · Rafael J. Wysocki
Quoted 26 lines · Måns Rullgård
Quoted 22 lines · Rafael J. Wysocki
Quoted 13 lines · Mans Rullgard

If the provided fwnode is an OF node, set dev.of_node as well.

---
 drivers/base/platform.c | 1 +
 1 file changed, 1 insertion(+)
drivers/base/platform.c +0 -0
index dff82a3c2caa..853a1d0e5845 100644

drivers/base/platform.c struct platform_device *platform_device_register_full( · line 515

Quoted 4 lines
drivers/base/platform.c +1 -0
@@ -512,6 +512,7 @@ struct platform_device *platform_device_register_full(
 
        pdev->dev.parent = pdevinfo->parent;
        pdev->dev.fwnode = pdevinfo->fwnode;
+       pdev->dev.of_node = of_node_get(to_of_node(pdev->dev.fwnode));

of_node_get() generally does a kobject_get() on the node's kobject, so when is that reference dropped? Or if it doesn't need to be dropped at all, why is this the case?

Quoted 8 lines
Quoted 1 line

platform_device_release() calls of_device_node_put().

Yes, it does, but this is the reference that's already acquired for devices added while parsing DT, isn't it?

Your change adds an extra reference AFAICS.

Also, why is this patch needed?

Some drivers are just shims that create extra "glue" devices with the DT device as parent and have the real driver bind to these. In other cases, the same real driver matches the DT node directly. When a glue device is used, it needs to get a reference to the original DT node in order for the main driver to access properties and child nodes.

Quoted 8 lines
Quoted 4 lines

Right now, my problem is that the suxi-musb driver creates such a glue device for the musb core driver to bind to without setting of_node. This means devices attached to this USB interface don't get associated with DT nodes, if present, the way they do with EHCI.

You really should describe problems that you want to address in patch changelogs. This helps a lot to understand the motivation for the changes.

Do you want me to send a new patch with the above explanation in the commit message?

Quoted 8 lines
Quoted 4 lines

The sunxi-musb driver uses platform_device_register_full(), so this seemed like the easiest way to let it set of_node of the new device. Since this creates a second reference to the same node, of_node_get() is required.

But what about devices that already have of_node set at this point?

Maybe check if of_node is NULL before trying to set it?

It's a brand new device allocated a few lines above. of_node has to be null here.

Signature · 1 line
Måns Rullgård
#7 lore raw
Quoted 54 lines · Måns Rullgård
Quoted 51 lines · Rafael J. Wysocki
Quoted 49 lines · Måns Rullgård
Quoted 27 lines · Rafael J. Wysocki
Quoted 26 lines · Måns Rullgård
Quoted 22 lines · Rafael J. Wysocki
Quoted 13 lines · Mans Rullgard

If the provided fwnode is an OF node, set dev.of_node as well.

---
 drivers/base/platform.c | 1 +
 1 file changed, 1 insertion(+)
drivers/base/platform.c +0 -0
index dff82a3c2caa..853a1d0e5845 100644

drivers/base/platform.c struct platform_device *platform_device_register_full( · line 515

Quoted 4 lines
drivers/base/platform.c +1 -0
@@ -512,6 +512,7 @@ struct platform_device *platform_device_register_full(
 
        pdev->dev.parent = pdevinfo->parent;
        pdev->dev.fwnode = pdevinfo->fwnode;
+       pdev->dev.of_node = of_node_get(to_of_node(pdev->dev.fwnode));

of_node_get() generally does a kobject_get() on the node's kobject, so when is that reference dropped? Or if it doesn't need to be dropped at all, why is this the case?

Quoted 8 lines
Quoted 1 line

platform_device_release() calls of_device_node_put().

Yes, it does, but this is the reference that's already acquired for devices added while parsing DT, isn't it?

Your change adds an extra reference AFAICS.

Also, why is this patch needed?

Some drivers are just shims that create extra "glue" devices with the DT device as parent and have the real driver bind to these. In other cases, the same real driver matches the DT node directly. When a glue device is used, it needs to get a reference to the original DT node in order for the main driver to access properties and child nodes.

Right now, my problem is that the suxi-musb driver creates such a glue device for the musb core driver to bind to without setting of_node. This means devices attached to this USB interface don't get associated with DT nodes, if present, the way they do with EHCI.

Quoted 6 lines
Quoted 3 lines

You really should describe problems that you want to address in patch changelogs. This helps a lot to understand the motivation for the changes.

Do you want me to send a new patch with the above explanation in the commit message?

Yes, please.

Quoted 11 lines
Quoted 8 lines
Quoted 4 lines

The sunxi-musb driver uses platform_device_register_full(), so this seemed like the easiest way to let it set of_node of the new device. Since this creates a second reference to the same node, of_node_get() is required.

But what about devices that already have of_node set at this point?

Maybe check if of_node is NULL before trying to set it?

It's a brand new device allocated a few lines above. of_node has to be null here.

Right, sorry for the confusion.

#8 lore raw

PATCH 1/1 platform: set of_node in platform_device_register_full()

In reply to Mans Rullgard

If the provided fwnode is an OF node, set dev.of_node as well.

Some drivers are just shims that create extra "glue" devices with the DT device as parent and have the real driver bind to these. In these cases, the glue device needs to get a reference to the original DT node in order for the main driver to access properties and child nodes.

For example, the sunxi-musb driver creates such a glue device using platform_device_register_full(). Consequently, devices attached to this USB interface don't get associated with DT nodes, if present, the way they do with EHCI.

This change will allow sunxi-musb and similar driver to easily propagate the DT node to child devices as required.

---
 drivers/base/platform.c | 1 +
 1 file changed, 1 insertion(+)
drivers/base/platform.c +1 -0
index dff82a3c2caa..853a1d0e5845 100644
@@ -512,6 +512,7 @@ struct platform_device *platform_device_register_full(
 
 	pdev->dev.parent = pdevinfo->parent;
 	pdev->dev.fwnode = pdevinfo->fwnode;
+	pdev->dev.of_node = of_node_get(to_of_node(pdev->dev.fwnode));
 
 	if (pdevinfo->dma_mask) {
 		/*
Signature · 1 line
2.20.1
maintainer
#9 lore raw
Quoted 7 lines · Mans Rullgard

If the provided fwnode is an OF node, set dev.of_node as well.

Some drivers are just shims that create extra "glue" devices with the DT device as parent and have the real driver bind to these. In these cases, the glue device needs to get a reference to the original DT node in order for the main driver to access properties and child nodes.

Quoted 7 lines

For example, the sunxi-musb driver creates such a glue device using platform_device_register_full(). Consequently, devices attached to this USB interface don't get associated with DT nodes, if present, the way they do with EHCI.

This change will allow sunxi-musb and similar driver to easily propagate the DT node to child devices as required.

Just a drive-by comment, didn't look to closely at this patch, but this all sounds familiar.

Note that if both platform devices are bound to drivers you may end up with some resources like pinctrl which are handled automatically by driver core at probe time to be requested twice (and failing the second time).

Take a look at 4e75e1d7dac9 ("driver core: add helper to reuse a device-tree node"), which provides a means to avoid this, and 49484abd93ab ("USB: musb: dsps: propagate device-tree node").

Quoted 10 lines
---
 drivers/base/platform.c | 1 +
 1 file changed, 1 insertion(+)
drivers/base/platform.c +0 -0
index dff82a3c2caa..853a1d0e5845 100644

drivers/base/platform.c struct platform_device *platform_device_register_full( · line 518

Quoted 7 lines
drivers/base/platform.c +1 -0
@@ -512,6 +512,7 @@ struct platform_device *platform_device_register_full(
 
 	pdev->dev.parent = pdevinfo->parent;
 	pdev->dev.fwnode = pdevinfo->fwnode;
+	pdev->dev.of_node = of_node_get(to_of_node(pdev->dev.fwnode));
 
 	if (pdevinfo->dma_mask) {
 		/*

Johan

#10 lore raw
Quoted 19 lines · Johan Hovold
Quoted 7 lines · Mans Rullgard

If the provided fwnode is an OF node, set dev.of_node as well.

Some drivers are just shims that create extra "glue" devices with the DT device as parent and have the real driver bind to these. In these cases, the glue device needs to get a reference to the original DT node in order for the main driver to access properties and child nodes.

Quoted 7 lines

For example, the sunxi-musb driver creates such a glue device using platform_device_register_full(). Consequently, devices attached to this USB interface don't get associated with DT nodes, if present, the way they do with EHCI.

This change will allow sunxi-musb and similar driver to easily propagate the DT node to child devices as required.

Just a drive-by comment, didn't look to closely at this patch, but this all sounds familiar.

Quoted 8 lines

Note that if both platform devices are bound to drivers you may end up with some resources like pinctrl which are handled automatically by driver core at probe time to be requested twice (and failing the second time).

Take a look at 4e75e1d7dac9 ("driver core: add helper to reuse a device-tree node"), which provides a means to avoid this, and 49484abd93ab ("USB: musb: dsps: propagate device-tree node").

Thanks, and ugh. So we should be setting the of_node_reused flag when this is the case. It's easy for the musb-dsps driver since it doesn't use platform_device_register_full() and can do this before the device_add() call. How can we convey that this flag needs to be set?

Quoted 19 lines
Quoted 10 lines
---
 drivers/base/platform.c | 1 +
 1 file changed, 1 insertion(+)
drivers/base/platform.c +0 -0
index dff82a3c2caa..853a1d0e5845 100644

drivers/base/platform.c struct platform_device *platform_device_register_full( · line 518

Quoted 7 lines
drivers/base/platform.c +1 -0
@@ -512,6 +512,7 @@ struct platform_device *platform_device_register_full(
 
 	pdev->dev.parent = pdevinfo->parent;
 	pdev->dev.fwnode = pdevinfo->fwnode;
+	pdev->dev.of_node = of_node_get(to_of_node(pdev->dev.fwnode));
 
 	if (pdevinfo->dma_mask) {
 		/*

Johan

Signature · 1 line
Måns Rullgård
#11 lore raw
Quoted 27 lines · Måns Rullgård
Quoted 24 lines · Johan Hovold
Quoted 7 lines · Mans Rullgard

If the provided fwnode is an OF node, set dev.of_node as well.

Some drivers are just shims that create extra "glue" devices with the DT device as parent and have the real driver bind to these. In these cases, the glue device needs to get a reference to the original DT node in order for the main driver to access properties and child nodes.

Quoted 7 lines

For example, the sunxi-musb driver creates such a glue device using platform_device_register_full(). Consequently, devices attached to this USB interface don't get associated with DT nodes, if present, the way they do with EHCI.

This change will allow sunxi-musb and similar driver to easily propagate the DT node to child devices as required.

Just a drive-by comment, didn't look to closely at this patch, but this all sounds familiar.

Note that if both platform devices are bound to drivers you may end up with some resources like pinctrl which are handled automatically by driver core at probe time to be requested twice (and failing the second time).

Quoted 8 lines
Quoted 3 lines

Take a look at 4e75e1d7dac9 ("driver core: add helper to reuse a device-tree node"), which provides a means to avoid this, and 49484abd93ab ("USB: musb: dsps: propagate device-tree node").

Thanks, and ugh. So we should be setting the of_node_reused flag when this is the case. It's easy for the musb-dsps driver since it doesn't use platform_device_register_full() and can do this before the device_add() call. How can we convey that this flag needs to be set?

Through pdevinfo I guess?

#12 lore raw
Quoted 32 lines · Rafael J. Wysocki
Quoted 31 lines · Måns Rullgård
Quoted 27 lines · Johan Hovold
Quoted 7 lines · Mans Rullgard

If the provided fwnode is an OF node, set dev.of_node as well.

Some drivers are just shims that create extra "glue" devices with the DT device as parent and have the real driver bind to these. In these cases, the glue device needs to get a reference to the original DT node in order for the main driver to access properties and child nodes.

Quoted 7 lines

For example, the sunxi-musb driver creates such a glue device using platform_device_register_full(). Consequently, devices attached to this USB interface don't get associated with DT nodes, if present, the way they do with EHCI.

This change will allow sunxi-musb and similar driver to easily propagate the DT node to child devices as required.

Just a drive-by comment, didn't look to closely at this patch, but this all sounds familiar.

Note that if both platform devices are bound to drivers you may end up with some resources like pinctrl which are handled automatically by driver core at probe time to be requested twice (and failing the second time).

Take a look at 4e75e1d7dac9 ("driver core: add helper to reuse a device-tree node"), which provides a means to avoid this, and 49484abd93ab ("USB: musb: dsps: propagate device-tree node").

Quoted 6 lines
Quoted 4 lines

Thanks, and ugh. So we should be setting the of_node_reused flag when this is the case. It's easy for the musb-dsps driver since it doesn't use platform_device_register_full() and can do this before the device_add() call. How can we convey that this flag needs to be set?

Through pdevinfo I guess?

Not without adding another field to it. The most direct is of course to simply add an of_node_reused flag there too and copy it over. Would that be OK, or is there a better way?

Signature · 1 line
Måns Rullgård
#13 lore raw
Quoted 40 lines · Måns Rullgård
Quoted 37 lines · Rafael J. Wysocki
Quoted 35 lines · Måns Rullgård
Quoted 19 lines · Johan Hovold
Quoted 7 lines · Mans Rullgard

If the provided fwnode is an OF node, set dev.of_node as well.

Some drivers are just shims that create extra "glue" devices with the DT device as parent and have the real driver bind to these. In these cases, the glue device needs to get a reference to the original DT node in order for the main driver to access properties and child nodes.

Quoted 7 lines

For example, the sunxi-musb driver creates such a glue device using platform_device_register_full(). Consequently, devices attached to this USB interface don't get associated with DT nodes, if present, the way they do with EHCI.

This change will allow sunxi-musb and similar driver to easily propagate the DT node to child devices as required.

Just a drive-by comment, didn't look to closely at this patch, but this all sounds familiar.

Quoted 8 lines

Note that if both platform devices are bound to drivers you may end up with some resources like pinctrl which are handled automatically by driver core at probe time to be requested twice (and failing the second time).

Take a look at 4e75e1d7dac9 ("driver core: add helper to reuse a device-tree node"), which provides a means to avoid this, and 49484abd93ab ("USB: musb: dsps: propagate device-tree node").

Thanks, and ugh. So we should be setting the of_node_reused flag when this is the case. It's easy for the musb-dsps driver since it doesn't use platform_device_register_full() and can do this before the device_add() call. How can we convey that this flag needs to be set?

Quoted 5 lines
Quoted 1 line

Through pdevinfo I guess?

Not without adding another field to it. The most direct is of course to simply add an of_node_reused flag there too and copy it over. Would that be OK, or is there a better way?

That's what I meant. :-)

#14 lore raw
In reply to Mans Rullgard
Quoted 26 lines · Mans Rullgard

If the provided fwnode is an OF node, set dev.of_node as well.

Some drivers are just shims that create extra "glue" devices with the DT device as parent and have the real driver bind to these. In these cases, the glue device needs to get a reference to the original DT node in order for the main driver to access properties and child nodes.

For example, the sunxi-musb driver creates such a glue device using platform_device_register_full(). Consequently, devices attached to this USB interface don't get associated with DT nodes, if present, the way they do with EHCI.

This change will allow sunxi-musb and similar driver to easily propagate the DT node to child devices as required.

---
 drivers/base/platform.c | 1 +
 1 file changed, 1 insertion(+)
drivers/base/platform.c +0 -0
index dff82a3c2caa..853a1d0e5845 100644

drivers/base/platform.c struct platform_device *platform_device_register_full( · line 518

Quoted 7 lines
drivers/base/platform.c +1 -0
@@ -512,6 +512,7 @@ struct platform_device *platform_device_register_full(
 
 	pdev->dev.parent = pdevinfo->parent;
 	pdev->dev.fwnode = pdevinfo->fwnode;
+	pdev->dev.of_node = of_node_get(to_of_node(pdev->dev.fwnode));
 
 	if (pdevinfo->dma_mask) {
 		/*

Sorry, I forgot to add a v2 to this. Anyway, the only change is the commit message.

Signature · 1 line
Måns Rullgård