Skip to contentSkip to conversation

mfd: syscon: Added support for using platform driver resources

PATCH In review Alexander Shiyan 3 participants 7 messages Feb 4, 2013 – Feb 18, 2013

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 mfd: syscon: Added support for using platform driver resources

This patch adds support usage platform driver resources, i.e. possibility works without oftree support. Additionally patch removes CONFIG_OF dependency and adds helper for accessing regmap by searching device by its name.

---
 drivers/mfd/Kconfig        |  1 -
 drivers/mfd/syscon.c       | 62 +++++++++++++++++++++++++++++++++++-----------
 include/linux/mfd/syscon.h |  1 +
 3 files changed, 48 insertions(+), 16 deletions(-)
drivers/mfd/Kconfig +0 -1
index 47ad4e2..389f5e2 100644
@@ -1065,7 +1065,6 @@ config MFD_STA2X11
 
 config MFD_SYSCON
 	bool "System Controller Register R/W Based on Regmap"
-	depends on OF
 	select REGMAP_MMIO
 	help
 	  Select this option to enable accessing system control registers
drivers/mfd/syscon.c +47 -15

94 lines not on this page · raw message

include/linux/mfd/syscon.h +1 -0
index 6aeb6b8..e14f7fd 100644
@@ -15,6 +15,7 @@
 #ifndef __LINUX_MFD_SYSCON_H__
 #define __LINUX_MFD_SYSCON_H__
 
+extern struct regmap *syscon_regmap_lookup_by_name(const char *name);
 extern struct regmap *syscon_node_to_regmap(struct device_node *np);
 extern struct regmap *syscon_regmap_lookup_by_compatible(const char *s);
 extern struct regmap *syscon_regmap_lookup_by_phandle(
Signature · 1 line
1.7.12.4
#1 lore raw

Hi Alexander,

Thanks for the patch adding non-dt support. :-)

Quoted 43 lines · Alexander Shiyan

This patch adds support usage platform driver resources, i.e. possibility works without oftree support. Additionally patch removes CONFIG_OF dependency and adds helper for accessing regmap by searching device by its name.

---
 drivers/mfd/Kconfig        |  1 -
 drivers/mfd/syscon.c       | 62 +++++++++++++++++++++++++++++++++++-----------
 include/linux/mfd/syscon.h |  1 +
 3 files changed, 48 insertions(+), 16 deletions(-)
drivers/mfd/Kconfig +0 -1
index 47ad4e2..389f5e2 100644
@@ -1065,7 +1065,6 @@ config MFD_STA2X11
 
 config MFD_SYSCON
 	bool "System Controller Register R/W Based on Regmap"
-	depends on OF
 	select REGMAP_MMIO
 	help
 	  Select this option to enable accessing system control registers
drivers/mfd/syscon.c +10 -1
index 3f10591..27d0a08 100644
@@ -29,7 +29,27 @@ struct syscon {
 	struct regmap *regmap;
 };
 
-static int syscon_match(struct device *dev, void *data)
+static int syscon_match_name(struct device *dev, void *data)
+{
+	return !strcmp(dev_name(dev), (const char *)data);
+}
+
+struct regmap *syscon_regmap_lookup_by_name(const char *name)
+{
+	struct syscon *syscon;
+	struct device *dev;
+

drivers/mfd/syscon.c struct syscon { · line 51

Quoted 10 lines
drivers/mfd/syscon.c +10 -0
+	dev = driver_find_device(&syscon_driver.driver, NULL, (void *)name,
+				 syscon_match_name);
+	if (!dev)
+		return ERR_PTR(-EPROBE_DEFER);
+
+	syscon = dev_get_drvdata(dev);
+
+	return syscon->regmap;
+}
+

How about syscon_dev_to_regmap(struct device *dev) as the exist dt version syscon_node_to_regmap since it's not affected by the name change of devices?

Quoted 47 lines
drivers/mfd/syscon.c +14 -12
+static int syscon_match_of(struct device *dev, void *data)
 {
 	struct syscon *syscon = dev_get_drvdata(dev);
 	struct device_node *dn = data;
@@ -43,7 +63,7 @@ struct regmap *syscon_node_to_regmap(struct device_node *np)
 	struct device *dev;
 
 	dev = driver_find_device(&syscon_driver.driver, NULL, np,
-				 syscon_match);
+				 syscon_match_of);
 	if (!dev)
 		return ERR_PTR(-EPROBE_DEFER);
 
@@ -102,26 +122,38 @@ static int syscon_probe(struct platform_device *pdev)
 	struct device *dev = &pdev->dev;
 	struct device_node *np = dev->of_node;
 	struct syscon *syscon;
-	struct resource res;
+	struct resource *res, res_of;
 	int ret;
 
-	if (!np)
-		return -ENOENT;
-
 	syscon = devm_kzalloc(dev, sizeof(struct syscon),
 			    GFP_KERNEL);
 	if (!syscon)
 		return -ENOMEM;
 
-	syscon->base = of_iomap(np, 0);
-	if (!syscon->base)
-		return -EADDRNOTAVAIL;
-
-	ret = of_address_to_resource(np, 0, &res);
-	if (ret)
-		return ret;
+	if (IS_ENABLED(CONFIG_OF) && np) {
+		syscon->base = of_iomap(np, 0);
+		if (!syscon->base)
+			return -EADDRNOTAVAIL;
+
+		res = &res_of;
+		ret = of_address_to_resource(np, 0, res);
+		if (ret)
+			return ret;
+	} else {
+		res = platform_get_resource(pdev, IORESOURCE_MEM, 0);

drivers/mfd/syscon.c static int syscon_probe(struct platform_device *pdev) · line 153

Quoted 10 lines
drivers/mfd/syscon.c +10 -0
+		if (!res)
+			return -ENXIO;
+
+		if (!request_mem_region(res->start, resource_size(res),
+					dev_name(&pdev->dev)))
+			return -EBUSY;
+
+		syscon->base = ioremap(res->start, resource_size(res));
+		if (!syscon->base)
+			return -EADDRNOTAVAIL;

devm_request_and_ioremap?

Regards
Dong Aisheng

Quoted 37 lines
drivers/mfd/syscon.c +3 -2
+	}
 
-	syscon_regmap_config.max_register = res.end - res.start - 3;
+	syscon_regmap_config.max_register = res->end - res->start - 3;
 	syscon->regmap = devm_regmap_init_mmio(dev, syscon->base,
 					&syscon_regmap_config);
 	if (IS_ERR(syscon->regmap)) {
@@ -133,7 +165,7 @@ static int syscon_probe(struct platform_device *pdev)
 	platform_set_drvdata(pdev, syscon);
 
 	dev_info(dev, "syscon regmap start 0x%x end 0x%x registered\n",
-		res.start, res.end);
+		res->start, res->end);
 
 	return 0;
 }
include/linux/mfd/syscon.h +1 -0
index 6aeb6b8..e14f7fd 100644
@@ -15,6 +15,7 @@
 #ifndef __LINUX_MFD_SYSCON_H__
 #define __LINUX_MFD_SYSCON_H__
 
+extern struct regmap *syscon_regmap_lookup_by_name(const char *name);
 extern struct regmap *syscon_node_to_regmap(struct device_node *np);
 extern struct regmap *syscon_regmap_lookup_by_compatible(const char *s);
 extern struct regmap *syscon_regmap_lookup_by_phandle(
Signature · 7 lines
1.7.12.4

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
#2 lore raw

Hello.
...

Quoted 7 lines

Thanks for the patch adding non-dt support. :-)

Quoted 4 lines · Alexander Shiyan

This patch adds support usage platform driver resources, i.e. possibility works without oftree support. Additionally patch removes CONFIG_OF dependency and adds helper for accessing regmap by searching device by its name.

...

Quoted 15 lines
Quoted 15 lines
drivers/mfd/syscon.c +15 -0
+static int syscon_match_name(struct device *dev, void *data)
+{
+	return !strcmp(dev_name(dev), (const char *)data);
+}
+
+struct regmap *syscon_regmap_lookup_by_name(const char *name)
+{
+	struct syscon *syscon;
+	struct device *dev;
+
+	dev = driver_find_device(&syscon_driver.driver, NULL, (void *)name,
+				 syscon_match_name);
+	if (!dev)
+		return ERR_PTR(-EPROBE_DEFER);
+
Quoted 8 lines

drivers/mfd/syscon.c struct syscon { · line 51

Quoted 5 lines
drivers/mfd/syscon.c +5 -0
+	syscon = dev_get_drvdata(dev);
+
+	return syscon->regmap;
+}
+

How about syscon_dev_to_regmap(struct device *dev) as the exist dt version syscon_node_to_regmap since it's not affected by the name change of devices?

I am not completely understand what you mean. In my version which doing search regmap by name, we can call this function with desired device name, then use regmap:
struct regmap *r = syscon_regmap_lookup_by_name("syscon.1");

You suggest use "struct device" as parameter? I do not know what we should use as parameter to the function in this case, since we can get "struct device" only when register this device, i.e. in board support code, not from anywhere, for example from another driver.
Fixme please.

...

Quoted 15 lines
Quoted 15 lines
drivers/mfd/syscon.c +15 -0
+	if (IS_ENABLED(CONFIG_OF) && np) {
+		syscon->base = of_iomap(np, 0);
+		if (!syscon->base)
+			return -EADDRNOTAVAIL;
+
+		res = &res_of;
+		ret = of_address_to_resource(np, 0, res);
+		if (ret)
+			return ret;
+	} else {
+		res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+		if (!res)
+			return -ENXIO;
+
+		if (!request_mem_region(res->start, resource_size(res),
Quoted 8 lines

drivers/mfd/syscon.c static int syscon_probe(struct platform_device *pdev) · line 153

Quoted 6 lines
drivers/mfd/syscon.c +6 -0
+					dev_name(&pdev->dev)))
+			return -EBUSY;
+
+		syscon->base = ioremap(res->start, resource_size(res));
+		if (!syscon->base)
+			return -EADDRNOTAVAIL;

devm_request_and_ioremap?

We call of_iomap for DT-version, for removal procedure - iounmap. Will iounmap work properly with devm_-version, I'm not sure.
May be better to completely remove ".remove" (and module_exit) feature for driver? It is loaded at startup always once compiled, and should always be in the system.

Thanks!
---
���{.n�+�������+%�����ݶ��w��{.n�+����{��G�����{ay�ʇڙ�,j��f���h��������z_�(�階�ݢj"���m������G������?����&���~��iO��z��v�^�m���� ������?�I�

#3 lore raw

Hi Alexander,

Quoted 34 lines · Alexander Shiyan

Hello.
...

Quoted 7 lines

Thanks for the patch adding non-dt support. :-)

Quoted 4 lines · Alexander Shiyan

This patch adds support usage platform driver resources, i.e. possibility works without oftree support. Additionally patch removes CONFIG_OF dependency and adds helper for accessing regmap by searching device by its name.

...

Quoted 15 lines
Quoted 15 lines
drivers/mfd/syscon.c +15 -0
+static int syscon_match_name(struct device *dev, void *data)
+{
+   return !strcmp(dev_name(dev), (const char *)data);
+}
+
+struct regmap *syscon_regmap_lookup_by_name(const char *name)
+{
+   struct syscon *syscon;
+   struct device *dev;
+
+   dev = driver_find_device(&syscon_driver.driver, NULL, (void *)name,
+                            syscon_match_name);
+   if (!dev)
+           return ERR_PTR(-EPROBE_DEFER);
+
Quoted 8 lines

drivers/mfd/syscon.c struct syscon { · line 51

Quoted 5 lines
drivers/mfd/syscon.c +5 -0
+   syscon = dev_get_drvdata(dev);
+
+   return syscon->regmap;
+}
+

How about syscon_dev_to_regmap(struct device *dev) as the exist dt version syscon_node_to_regmap since it's not affected by the name change of devices?

Quoted 5 lines

I am not completely understand what you mean. In my version which doing search regmap by name, we can call this function with desired device name, then use regmap:
struct regmap *r = syscon_regmap_lookup_by_name("syscon.1");

My concern is that this API may be used by other client drivers, if we hardcode the device name in those drivers, once the device name is changed, all those drivers need change too.
For dt case, we use device node to find regmap, so does not care about the name.

Quoted 1 line

You suggest use "struct device" as parameter?

A device pointer.

Quoted 3 lines

I do not know what we should
use as parameter to the function in this case, since we can get "struct device" only when register this device,

If we have a platform_device, then we have its device pointer, right?

Quoted 4 lines

i.e. in board support code, not from anywhere,
for example from another driver.
Fixme please.

My understanding is that in board support code, we can have the platform device instance
of that syscon compatible device, then we can use it as the platform data parameter for those devices driver who wants to use it. Then in client driver, it can call:
syscon_dev_to_regmap(syscon_compatible_dev)
to find the regmap.
Just like dt working way, the device node usually uses a phandle pointing to the syscon compatible node which it wants to use.

Quoted 25 lines

...

Quoted 15 lines
Quoted 15 lines
drivers/mfd/syscon.c +15 -0
+   if (IS_ENABLED(CONFIG_OF) && np) {
+           syscon->base = of_iomap(np, 0);
+           if (!syscon->base)
+                   return -EADDRNOTAVAIL;
+
+           res = &res_of;
+           ret = of_address_to_resource(np, 0, res);
+           if (ret)
+                   return ret;
+   } else {
+           res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+           if (!res)
+                   return -ENXIO;
+
+           if (!request_mem_region(res->start, resource_size(res),
Quoted 8 lines

drivers/mfd/syscon.c static int syscon_probe(struct platform_device *pdev) · line 153

Quoted 6 lines
drivers/mfd/syscon.c +6 -0
+                                   dev_name(&pdev->dev)))
+                   return -EBUSY;
+
+           syscon->base = ioremap(res->start, resource_size(res));
+           if (!syscon->base)
+                   return -EADDRNOTAVAIL;

devm_request_and_ioremap?

Quoted 2 lines

We call of_iomap for DT-version, for removal procedure - iounmap. Will iounmap work properly with devm_-version, I'm not sure.

You're right, i'm afraid not.
If that, i wonder the ioremap resource may be freed twice for driver removal. So, i'm okay with your current way.
But one issue is that then we need take care of the resource free by ourselves for probe failed cases. e.g. ioremap and request_mem_region.
You may need add them.

Quoted 3 lines

May be better to completely remove ".remove" (and module_exit) feature for driver? It is loaded at startup always once compiled, and should always be in the system.

It supports to be a module, so it may be better to keep ".remove".

Regards
Dong Aisheng

#4 lore raw
In reply to Alexander Shiyan

Hello.

Quoted 59 lines
Quoted 33 lines

...

Quoted 7 lines

Thanks for the patch adding non-dt support. :-)

Quoted 4 lines · Alexander Shiyan

This patch adds support usage platform driver resources, i.e. possibility works without oftree support. Additionally patch removes CONFIG_OF dependency and adds helper for accessing regmap by searching device by its name.

...

Quoted 15 lines
Quoted 15 lines
drivers/mfd/syscon.c +15 -0
+static int syscon_match_name(struct device *dev, void *data)
+{
+   return !strcmp(dev_name(dev), (const char *)data);
+}
+
+struct regmap *syscon_regmap_lookup_by_name(const char *name)
+{
+   struct syscon *syscon;
+   struct device *dev;
+
+   dev = driver_find_device(&syscon_driver.driver, NULL, (void *)name,
+                            syscon_match_name);
+   if (!dev)
+           return ERR_PTR(-EPROBE_DEFER);
+
Quoted 8 lines

drivers/mfd/syscon.c struct syscon { · line 51

Quoted 5 lines
drivers/mfd/syscon.c +5 -0
+   syscon = dev_get_drvdata(dev);
+
+   return syscon->regmap;
+}
+

How about syscon_dev_to_regmap(struct device *dev) as the exist dt version syscon_node_to_regmap since it's not affected by the name change of devices?

Quoted 5 lines

I am not completely understand what you mean. In my version which doing search regmap by name, we can call this function with desired device name, then use regmap:
struct regmap *r = syscon_regmap_lookup_by_name("syscon.1");

My concern is that this API may be used by other client drivers, if we hardcode the device name in those drivers, once the device name is changed, all those drivers need change too.
For dt case, we use device node to find regmap, so does not care about the name.

Quoted 1 line

You suggest use "struct device" as parameter?

A device pointer.

Quoted 3 lines

I do not know what we should
use as parameter to the function in this case, since we can get "struct device" only when register this device,

If we have a platform_device, then we have its device pointer, right?

Quoted 4 lines

i.e. in board support code, not from anywhere,
for example from another driver.
Fixme please.

Quoted 9 lines

My understanding is that in board support code, we can have the platform device instance
of that syscon compatible device, then we can use it as the platform data parameter for those devices driver who wants to use it. Then in client driver, it can call:
syscon_dev_to_regmap(syscon_compatible_dev)
to find the regmap.
Just like dt working way, the device node usually uses a phandle pointing to the syscon compatible node which it wants to use.

This is not as easy as it seems.
All devices that will use "syscon" driver, in this case should have a platform_data record. I think that it can create problems with using these drivers as modules. Alternatively, we can create additional (virtual) "compatible" text property in syscon private data structure and use it to find the proper device. What is your opinion?

Quoted 28 lines
Quoted 25 lines

...

Quoted 15 lines
Quoted 15 lines
drivers/mfd/syscon.c +15 -0
+   if (IS_ENABLED(CONFIG_OF) && np) {
+           syscon->base = of_iomap(np, 0);
+           if (!syscon->base)
+                   return -EADDRNOTAVAIL;
+
+           res = &res_of;
+           ret = of_address_to_resource(np, 0, res);
+           if (ret)
+                   return ret;
+   } else {
+           res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+           if (!res)
+                   return -ENXIO;
+
+           if (!request_mem_region(res->start, resource_size(res),
Quoted 8 lines

drivers/mfd/syscon.c static int syscon_probe(struct platform_device *pdev) · line 153

Quoted 6 lines
drivers/mfd/syscon.c +6 -0
+                                   dev_name(&pdev->dev)))
+                   return -EBUSY;
+
+           syscon->base = ioremap(res->start, resource_size(res));
+           if (!syscon->base)
+                   return -EADDRNOTAVAIL;

devm_request_and_ioremap?

Quoted 2 lines

We call of_iomap for DT-version, for removal procedure - iounmap. Will iounmap work properly with devm_-version, I'm not sure.

Quoted 6 lines

You're right, i'm afraid not.
If that, i wonder the ioremap resource may be freed twice for driver removal. So, i'm okay with your current way.
But one issue is that then we need take care of the resource free by ourselves for probe failed cases. e.g. ioremap and request_mem_region.
You may need add them.

Yes, thanks. I forget about it.

Quoted 3 lines
Quoted 2 lines

May be better to completely remove ".remove" (and module_exit) feature for driver? It is loaded at startup always once compiled, and should always be in the system.

It supports to be a module, so it may be better to keep ".remove".

As far I can see, no. Kernel symbol defined as bool, so driver cannot be compiled as module:
config MFD_SYSCON
bool "System Controller Register R/W Based on Regmap"

Thanks!

---
���{.n�+�������+%�����ݶ��w��{.n�+����{��G�����{ay�ʇڙ�,j��f���h��������z_�(�階�ݢj"���m������G������?����&���~��iO��z��v�^�m���� ������?�I�

#5 lore raw

Hi Alexander,

Quoted 71 lines · Alexander Shiyan

Hello.

Quoted 59 lines
Quoted 33 lines

...

Quoted 7 lines

Thanks for the patch adding non-dt support. :-)

Quoted 4 lines · Alexander Shiyan

This patch adds support usage platform driver resources, i.e. possibility works without oftree support. Additionally patch removes CONFIG_OF dependency and adds helper for accessing regmap by searching device by its name.

...

Quoted 15 lines
Quoted 15 lines
drivers/mfd/syscon.c +15 -0
+static int syscon_match_name(struct device *dev, void *data)
+{
+   return !strcmp(dev_name(dev), (const char *)data);
+}
+
+struct regmap *syscon_regmap_lookup_by_name(const char *name)
+{
+   struct syscon *syscon;
+   struct device *dev;
+
+   dev = driver_find_device(&syscon_driver.driver, NULL, (void *)name,
+                            syscon_match_name);
+   if (!dev)
+           return ERR_PTR(-EPROBE_DEFER);
+
Quoted 8 lines

drivers/mfd/syscon.c struct syscon { · line 51

Quoted 5 lines
drivers/mfd/syscon.c +5 -0
+   syscon = dev_get_drvdata(dev);
+
+   return syscon->regmap;
+}
+

How about syscon_dev_to_regmap(struct device *dev) as the exist dt version syscon_node_to_regmap since it's not affected by the name change of devices?

Quoted 5 lines

I am not completely understand what you mean. In my version which doing search regmap by name, we can call this function with desired device name, then use regmap:
struct regmap *r = syscon_regmap_lookup_by_name("syscon.1");

My concern is that this API may be used by other client drivers, if we hardcode the device name in those drivers, once the device name is changed, all those drivers need change too.
For dt case, we use device node to find regmap, so does not care about the name.

Quoted 1 line

You suggest use "struct device" as parameter?

A device pointer.

Quoted 3 lines

I do not know what we should
use as parameter to the function in this case, since we can get "struct device" only when register this device,

If we have a platform_device, then we have its device pointer, right?

Quoted 4 lines

i.e. in board support code, not from anywhere,
for example from another driver.
Fixme please.

Quoted 9 lines

My understanding is that in board support code, we can have the platform device instance
of that syscon compatible device, then we can use it as the platform data parameter for those devices driver who wants to use it. Then in client driver, it can call:
syscon_dev_to_regmap(syscon_compatible_dev)
to find the regmap.
Just like dt working way, the device node usually uses a phandle pointing to the syscon compatible node which it wants to use.

Quoted 3 lines

This is not as easy as it seems.
All devices that will use "syscon" driver, in this case should have a platform_data record.

Yes, that's the same way as the dt version does.

Quoted 1 line

I think that it can create problems with using these drivers as modules.

What problems do you think?

Quoted 3 lines

Alternatively, we can create additional (virtual) "compatible" text property in syscon private data structure and use it to find the proper device. What is your opinion?

Hmm, i can't see why we need that.
IMO, for non-dt, we can just use platform_device_id to match devices.

Quoted 37 lines
Quoted 28 lines
Quoted 25 lines

...

Quoted 15 lines
Quoted 15 lines
drivers/mfd/syscon.c +15 -0
+   if (IS_ENABLED(CONFIG_OF) && np) {
+           syscon->base = of_iomap(np, 0);
+           if (!syscon->base)
+                   return -EADDRNOTAVAIL;
+
+           res = &res_of;
+           ret = of_address_to_resource(np, 0, res);
+           if (ret)
+                   return ret;
+   } else {
+           res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+           if (!res)
+                   return -ENXIO;
+
+           if (!request_mem_region(res->start, resource_size(res),
Quoted 8 lines

drivers/mfd/syscon.c static int syscon_probe(struct platform_device *pdev) · line 153

Quoted 6 lines
drivers/mfd/syscon.c +6 -0
+                                   dev_name(&pdev->dev)))
+                   return -EBUSY;
+
+           syscon->base = ioremap(res->start, resource_size(res));
+           if (!syscon->base)
+                   return -EADDRNOTAVAIL;

devm_request_and_ioremap?

Quoted 2 lines

We call of_iomap for DT-version, for removal procedure - iounmap. Will iounmap work properly with devm_-version, I'm not sure.

Quoted 6 lines

You're right, i'm afraid not.
If that, i wonder the ioremap resource may be freed twice for driver removal. So, i'm okay with your current way.
But one issue is that then we need take care of the resource free by ourselves for probe failed cases. e.g. ioremap and request_mem_region.
You may need add them.

Yes, thanks. I forget about it.

Quoted 9 lines
Quoted 3 lines
Quoted 2 lines

May be better to completely remove ".remove" (and module_exit) feature for driver? It is loaded at startup always once compiled, and should always be in the system.

It supports to be a module, so it may be better to keep ".remove".

As far I can see, no. Kernel symbol defined as bool, so driver cannot be compiled as module:
config MFD_SYSCON
bool "System Controller Register R/W Based on Regmap"

I see.
Thanks

Regards
Dong Aisheng

Quoted 3 lines

Thanks!

---

#6 lore raw
In reply to Alexander Shiyan

Hello.

...

Quoted 71 lines
Quoted 67 lines
Quoted 57 lines
Quoted 32 lines
Quoted 7 lines

Thanks for the patch adding non-dt support. :-)

Quoted 4 lines · Alexander Shiyan

This patch adds support usage platform driver resources, i.e. possibility works without oftree support. Additionally patch removes CONFIG_OF dependency and adds helper for accessing regmap by searching device by its name.

...

Quoted 15 lines
Quoted 15 lines
drivers/mfd/syscon.c +15 -0
+static int syscon_match_name(struct device *dev, void *data)
+{
+   return !strcmp(dev_name(dev), (const char *)data);
+}
+
+struct regmap *syscon_regmap_lookup_by_name(const char *name)
+{
+   struct syscon *syscon;
+   struct device *dev;
+
+   dev = driver_find_device(&syscon_driver.driver, NULL, (void *)name,
+                            syscon_match_name);
+   if (!dev)
+           return ERR_PTR(-EPROBE_DEFER);
+
Quoted 8 lines

drivers/mfd/syscon.c struct syscon { · line 51

Quoted 5 lines
drivers/mfd/syscon.c +5 -0
+   syscon = dev_get_drvdata(dev);
+
+   return syscon->regmap;
+}
+

How about syscon_dev_to_regmap(struct device *dev) as the exist dt version syscon_node_to_regmap since it's not affected by the name change of devices?

Quoted 5 lines

I am not completely understand what you mean. In my version which doing search regmap by name, we can call this function with desired device name, then use regmap:
struct regmap *r = syscon_regmap_lookup_by_name("syscon.1");

My concern is that this API may be used by other client drivers, if we hardcode the device name in those drivers, once the device name is changed, all those drivers need change too.
For dt case, we use device node to find regmap, so does not care about the name.

Quoted 1 line

You suggest use "struct device" as parameter?

A device pointer.

Quoted 3 lines

I do not know what we should
use as parameter to the function in this case, since we can get "struct device" only when register this device,

If we have a platform_device, then we have its device pointer, right?

Quoted 4 lines

i.e. in board support code, not from anywhere,
for example from another driver.
Fixme please.

Quoted 9 lines

My understanding is that in board support code, we can have the platform device instance
of that syscon compatible device, then we can use it as the platform data parameter for those devices driver who wants to use it. Then in client driver, it can call:
syscon_dev_to_regmap(syscon_compatible_dev)
to find the regmap.
Just like dt working way, the device node usually uses a phandle pointing to the syscon compatible node which it wants to use.

Quoted 3 lines

This is not as easy as it seems.
All devices that will use "syscon" driver, in this case should have a platform_data record.

Quoted 5 lines

Yes, that's the same way as the dt version does.

Quoted 1 line

I think that it can create problems with using these drivers as modules.

What problems do you think?

I've changed my opinion ;)

Quoted 6 lines
Quoted 3 lines

Alternatively, we can create additional (virtual) "compatible" text property in syscon private data structure and use it to find the proper device. What is your opinion?

Hmm, i can't see why we need that.
IMO, for non-dt, we can just use platform_device_id to match devices.

Unfortunately platform_device_id.name field have a length limitation. But, of course, this is a other theme for discussion. So, I'll send new version of patch. Thanks!

---
���{.n�+�������+%�����ݶ��w��{.n�+����{��G�����{ay�ʇڙ�,j��f���h��������z_�(�階�ݢj"���m������G������?����&���~��iO��z��v�^�m���� ������?�I�