arm: omap: omap2plus_defconfig: remove unwanted ethernet drivers

Many of AM335x and AM437x hook backlight to
one of these two devices. By enabling their
drivers we make sure pwm-backlight can do
its thing.

i2c devices match against struct i2c_device_id even for CONFIG_OF case, so adding a struct of_device_id doesn't change anything. As a result, currently, edt-ft5x06 will not autoprobe if built as a module.

To fix the issue and still maintain backwards compatibility with all DTS files currently in tree, we're just moving all ids from of_device_id to i2c_device_id while also adding the following specific ids which should be used from now on:

	{ "edt-ft5206", 0, }
	{ "edt-ft5306", 0, }
	{ "edt-ft5406", 0, }

Hi Felipe,

Quoted 4 lines · Felipe Balbi

i2c devices match against struct i2c_device_id even for CONFIG_OF case, so adding a struct of_device_id doesn't change anything. As a result, currently, edt-ft5x06 will not autoprobe if built as a module.

Why doe snot it autoprobe? We properly declare MODULE_DEVICE_TABLE for OF, is it because we are missing some data in device uevent?

Quoted 9 lines

To fix the issue and still maintain backwards compatibility with all DTS files currently in tree, we're just moving all ids from of_device_id to i2c_device_id while also adding the following specific ids which should be used from now on:

	{ "edt-ft5206", 0, }
	{ "edt-ft5306", 0, }
	{ "edt-ft5406", 0, }

Is this a tee-wide change? Link to the discussion?

Quoted 10 lines · Dmitry Torokhov

Hi Felipe,

Quoted 4 lines · Felipe Balbi

i2c devices match against struct i2c_device_id even for CONFIG_OF case, so adding a struct of_device_id doesn't change anything. As a result, currently, edt-ft5x06 will not autoprobe if built as a module.

Why doe snot it autoprobe? We properly declare MODULE_DEVICE_TABLE for OF, is it because we are missing some data in device uevent?

because of of_i2c_register_devices(). Maybe Wolfram can give a better explanation here, but it just doesn't match through of_device_id.

Apply this:

drivers/i2c/i2c-core.c +8 -2

index 2f90ac6..f0dc16e 100644
@@ -449,8 +449,10 @@ static const struct i2c_device_id *i2c_match_id(const struct i2c_device_id *id,
 						const struct i2c_client *client)
 {
 	while (id->name[0]) {
-		if (strcmp(client->name, id->name) == 0)
+		if (strcmp(client->name, id->name) == 0) {
+			dev_info(&client->dev, "i2c_device_id match\n");
 			return id;
+		}
 		id++;
 	}
 	return NULL;

@@ -465,8 +467,10 @@ static int i2c_device_match(struct device *dev, struct device_driver *drv)
 		return 0;
 
 	/* Attempt an OF style match */
-	if (of_driver_match_device(dev, drv))
+	if (of_driver_match_device(dev, drv)) {
+		dev_info(dev, "of driver match\n");
 		return 1;
+	}
 
 	/* Then ACPI style match */
 	if (acpi_driver_match_device(dev, drv))

@@ -1081,6 +1085,7 @@ i2c_new_device(struct i2c_adapter *adap, struct i2c_board_info const *info)
 	client->dev.bus = &i2c_bus_type;
 	client->dev.type = &i2c_client_type;
 	client->dev.of_node = info->of_node;
+	dev_info(&adap->dev, "%s: of_node %p\n", __func__, info->of_node);
 	ACPI_COMPANION_SET(&client->dev, info->acpi_node.companion);
 
 	i2c_dev_set_name(adap, client);

@@ -1411,6 +1416,7 @@ static void of_i2c_register_devices(struct i2c_adapter *adap)
 
 		info.irq = irq_of_parse_and_map(node, 0);
 		info.of_node = of_node_get(node);
+		dev_info(&adap->dev, "%s: of_node %p\n", __func__, info.of_node);
 		info.archdata = &dev_ad;
 
 		if (of_get_property(node, "wakeup-source", NULL))

then boot the board and you get http://hastebin.com/oqemezajez

Interesting, it's matching against of but only when I modprobe.

Let me debug this one a little more.

Because with OF we can pass more specific compatible flags (such as edt-ft5306) instead of generic edt-ft5x06, when i2c-core's of_i2c_register_devices() tries to request_module(), it'll request it with a non-existent specific module alias.

In order to have this driver autoprobing again, we just need to add missing MODULE_ALIAS() entries to edt-ft5x06 driver.

Thanks to Dmitry for noticing that it actually should autoprobe even with of_device_id.

Cc Dmitry Torokhov <[email protected]>

Cc <[email protected]>

Cc <[email protected]> # v3.15+

Fixes dac90dc2 (Input: edt-ft5x06 - add DT support)

Signed-off-by Felipe Balbi <[email protected]>

---

Because with OF we can pass more specific compatible flags (such as edt-ft5306) instead of generic edt-ft5x06, when i2c-core's of_i2c_register_devices() tries to request_module(), it'll request it with a non-existent specific module alias.

In order to have this driver autoprobing again, we just need to add missing MODULE_ALIAS() entries to edt-ft5x06 driver.

Thanks to Dmitry for noticing that it actually should autoprobe even with of_device_id.

Cc Dmitry Torokhov <[email protected]>

Cc <[email protected]>

Cc <[email protected]> # v3.15+

Fixes dac90dc2 (Input: edt-ft5x06 - add DT support)

Signed-off-by Felipe Balbi <[email protected]>

---
 drivers/input/touchscreen/edt-ft5x06.c | 3 +++
 1 file changed, 3 insertions(+)

drivers/input/touchscreen/edt-ft5x06.c +3 -0

index ee3434f..bcbf688 100644
@@ -1147,6 +1147,9 @@ static struct i2c_driver edt_ft5x06_ts_driver = {
 
 module_i2c_driver(edt_ft5x06_ts_driver);
 
+MODULE_ALIAS("i2c:edt-ft5206");
+MODULE_ALIAS("i2c:edt-ft5306");
+MODULE_ALIAS("i2c:edt-ft5406");
 MODULE_AUTHOR("Simon Budig <[email protected]>");
 MODULE_DESCRIPTION("EDT FT5x06 I2C Touchscreen Driver");
 MODULE_LICENSE("GPL");

Signature · 1 line

2.1.0.GIT

Quoted 40 lines · Felipe Balbi

Because with OF we can pass more specific compatible flags (such as edt-ft5306) instead of generic edt-ft5x06, when i2c-core's of_i2c_register_devices() tries to request_module(), it'll request it with a non-existent specific module alias.

In order to have this driver autoprobing again, we just need to add missing MODULE_ALIAS() entries to edt-ft5x06 driver.

Thanks to Dmitry for noticing that it actually should autoprobe even with of_device_id.

Cc Dmitry Torokhov <[email protected]>

Cc <[email protected]>

Cc <[email protected]> # v3.15+

Fixes dac90dc2 (Input: edt-ft5x06 - add DT support)

Signed-off-by Felipe Balbi <[email protected]>

---
 drivers/input/touchscreen/edt-ft5x06.c | 3 +++
 1 file changed, 3 insertions(+)

drivers/input/touchscreen/edt-ft5x06.c +3 -0

index ee3434f..bcbf688 100644
@@ -1147,6 +1147,9 @@ static struct i2c_driver edt_ft5x06_ts_driver = {
 
 module_i2c_driver(edt_ft5x06_ts_driver);
 
+MODULE_ALIAS("i2c:edt-ft5206");
+MODULE_ALIAS("i2c:edt-ft5306");
+MODULE_ALIAS("i2c:edt-ft5406");
 MODULE_AUTHOR("Simon Budig <[email protected]>");
 MODULE_DESCRIPTION("EDT FT5x06 I2C Touchscreen Driver");
 MODULE_LICENSE("GPL");

Signature · 1 line

2.1.0.GIT

Here is what I see if I run modinfo on the module:

dtor@dtor-ws:~/kernel/work$ modinfo
drivers/input/touchscreen/edt-ft5x06.ko
filename:
/home/dtor/kernel/work/drivers/input/touchscreen/edt-ft5x06.ko
license:        GPL
description:    EDT FT5x06 I2C Touchscreen Driver
author:         Simon Budig <[email protected]>
alias:          i2c:edt-ft5x06
alias:          of:N*T*Cedt,edt-ft5406*
alias:          of:N*T*Cedt,edt-ft5306*
alias:          of:N*T*Cedt,edt-ft5206*
depends:        i2c-core
intree:         Y
vermagic:       3.17.0+ SMP preempt mod_unload 

As you can see we already have what I consider proper modaliases for the driver. Why don't they work? Is it because modprobe doe snot know how to handle OF-style modaliases or device's uevents are missing OF data in them?

Quoted 59 lines · Dmitry Torokhov

Quoted 40 lines · Felipe Balbi

Because with OF we can pass more specific compatible flags (such as edt-ft5306) instead of generic edt-ft5x06, when i2c-core's of_i2c_register_devices() tries to request_module(), it'll request it with a non-existent specific module alias.

In order to have this driver autoprobing again, we just need to add missing MODULE_ALIAS() entries to edt-ft5x06 driver.

Thanks to Dmitry for noticing that it actually should autoprobe even with of_device_id.

Cc Dmitry Torokhov <[email protected]>

Cc <[email protected]>

Cc <[email protected]> # v3.15+

Fixes dac90dc2 (Input: edt-ft5x06 - add DT support)

Signed-off-by Felipe Balbi <[email protected]>

---
 drivers/input/touchscreen/edt-ft5x06.c | 3 +++
 1 file changed, 3 insertions(+)

drivers/input/touchscreen/edt-ft5x06.c +3 -0

index ee3434f..bcbf688 100644
@@ -1147,6 +1147,9 @@ static struct i2c_driver edt_ft5x06_ts_driver = {
 
 module_i2c_driver(edt_ft5x06_ts_driver);
 
+MODULE_ALIAS("i2c:edt-ft5206");
+MODULE_ALIAS("i2c:edt-ft5306");
+MODULE_ALIAS("i2c:edt-ft5406");
 MODULE_AUTHOR("Simon Budig <[email protected]>");
 MODULE_DESCRIPTION("EDT FT5x06 I2C Touchscreen Driver");
 MODULE_LICENSE("GPL");

Signature · 1 line

2.1.0.GIT

Here is what I see if I run modinfo on the module:

dtor@dtor-ws:~/kernel/work$ modinfo
drivers/input/touchscreen/edt-ft5x06.ko
filename:
/home/dtor/kernel/work/drivers/input/touchscreen/edt-ft5x06.ko
license:        GPL
description:    EDT FT5x06 I2C Touchscreen Driver
author:         Simon Budig <[email protected]>
alias:          i2c:edt-ft5x06
alias:          of:N*T*Cedt,edt-ft5406*
alias:          of:N*T*Cedt,edt-ft5306*
alias:          of:N*T*Cedt,edt-ft5206*
depends:        i2c-core
intree:         Y
vermagic:       3.17.0+ SMP preempt mod_unload 

Quoted 4 lines

As you can see we already have what I consider proper modaliases for the driver. Why don't they work? Is it because modprobe doe snot know how to handle OF-style modaliases or device's uevents are missing OF data in them?

no, it's because i2c call request_module for i2c:edt-ft5306. i2c core never uses any of the of aliases.

Quoted 60 lines · Felipe Balbi

Quoted 59 lines · Dmitry Torokhov

Quoted 40 lines · Felipe Balbi

Because with OF we can pass more specific compatible flags (such as edt-ft5306) instead of generic edt-ft5x06, when i2c-core's of_i2c_register_devices() tries to request_module(), it'll request it with a non-existent specific module alias.

In order to have this driver autoprobing again, we just need to add missing MODULE_ALIAS() entries to edt-ft5x06 driver.

Thanks to Dmitry for noticing that it actually should autoprobe even with of_device_id.

Cc Dmitry Torokhov <[email protected]>

Cc <[email protected]>

Cc <[email protected]> # v3.15+

Fixes dac90dc2 (Input: edt-ft5x06 - add DT support)

Signed-off-by Felipe Balbi <[email protected]>

---
 drivers/input/touchscreen/edt-ft5x06.c | 3 +++
 1 file changed, 3 insertions(+)

drivers/input/touchscreen/edt-ft5x06.c +3 -0

index ee3434f..bcbf688 100644
@@ -1147,6 +1147,9 @@ static struct i2c_driver edt_ft5x06_ts_driver = {
 
 module_i2c_driver(edt_ft5x06_ts_driver);
 
+MODULE_ALIAS("i2c:edt-ft5206");
+MODULE_ALIAS("i2c:edt-ft5306");
+MODULE_ALIAS("i2c:edt-ft5406");
 MODULE_AUTHOR("Simon Budig <[email protected]>");
 MODULE_DESCRIPTION("EDT FT5x06 I2C Touchscreen Driver");
 MODULE_LICENSE("GPL");

Signature · 1 line

2.1.0.GIT

Here is what I see if I run modinfo on the module:

dtor@dtor-ws:~/kernel/work$ modinfo
drivers/input/touchscreen/edt-ft5x06.ko
filename:
/home/dtor/kernel/work/drivers/input/touchscreen/edt-ft5x06.ko
license:        GPL
description:    EDT FT5x06 I2C Touchscreen Driver
author:         Simon Budig <[email protected]>
alias:          i2c:edt-ft5x06
alias:          of:N*T*Cedt,edt-ft5406*
alias:          of:N*T*Cedt,edt-ft5306*
alias:          of:N*T*Cedt,edt-ft5206*
depends:        i2c-core
intree:         Y
vermagic:       3.17.0+ SMP preempt mod_unload 

Quoted 7 lines

Quoted 4 lines

As you can see we already have what I consider proper modaliases for the driver. Why don't they work? Is it because modprobe doe snot know how to handle OF-style modaliases or device's uevents are missing OF data in them?

no, it's because i2c call request_module for i2c:edt-ft5306. i2c core never uses any of the of aliases.

1. I think both i2c and spi are cheating in this regard: they should rely on normal driver request paths through udev.

2. I2C should emit proper modalias for devices coming by the way of devicetree. It already does special handling for ACPI, it needs to do the same for OF. I think SPI does the right thing there.

Quoted 69 lines · Dmitry Torokhov

Quoted 67 lines · Felipe Balbi

Quoted 59 lines · Dmitry Torokhov

Quoted 40 lines · Felipe Balbi

Because with OF we can pass more specific compatible flags (such as edt-ft5306) instead of generic edt-ft5x06, when i2c-core's of_i2c_register_devices() tries to request_module(), it'll request it with a non-existent specific module alias.

In order to have this driver autoprobing again, we just need to add missing MODULE_ALIAS() entries to edt-ft5x06 driver.

Thanks to Dmitry for noticing that it actually should autoprobe even with of_device_id.

Cc Dmitry Torokhov <[email protected]>

Cc <[email protected]>

Cc <[email protected]> # v3.15+

Fixes dac90dc2 (Input: edt-ft5x06 - add DT support)

Signed-off-by Felipe Balbi <[email protected]>

---
 drivers/input/touchscreen/edt-ft5x06.c | 3 +++
 1 file changed, 3 insertions(+)

drivers/input/touchscreen/edt-ft5x06.c +3 -0

index ee3434f..bcbf688 100644
@@ -1147,6 +1147,9 @@ static struct i2c_driver edt_ft5x06_ts_driver = {
 
 module_i2c_driver(edt_ft5x06_ts_driver);
 
+MODULE_ALIAS("i2c:edt-ft5206");
+MODULE_ALIAS("i2c:edt-ft5306");
+MODULE_ALIAS("i2c:edt-ft5406");
 MODULE_AUTHOR("Simon Budig <[email protected]>");
 MODULE_DESCRIPTION("EDT FT5x06 I2C Touchscreen Driver");
 MODULE_LICENSE("GPL");

Signature · 1 line

2.1.0.GIT

Here is what I see if I run modinfo on the module:

dtor@dtor-ws:~/kernel/work$ modinfo
drivers/input/touchscreen/edt-ft5x06.ko
filename:
/home/dtor/kernel/work/drivers/input/touchscreen/edt-ft5x06.ko
license:        GPL
description:    EDT FT5x06 I2C Touchscreen Driver
author:         Simon Budig <[email protected]>
alias:          i2c:edt-ft5x06
alias:          of:N*T*Cedt,edt-ft5406*
alias:          of:N*T*Cedt,edt-ft5306*
alias:          of:N*T*Cedt,edt-ft5206*
depends:        i2c-core
intree:         Y
vermagic:       3.17.0+ SMP preempt mod_unload 

Quoted 4 lines

As you can see we already have what I consider proper modaliases for the driver. Why don't they work? Is it because modprobe doe snot know how to handle OF-style modaliases or device's uevents are missing OF data in them?

no, it's because i2c call request_module for i2c:edt-ft5306. i2c core never uses any of the of aliases.

Quoted 6 lines

1. I think both i2c and spi are cheating in this regard: they should rely on normal driver request paths through udev.

2. I2C should emit proper modalias for devices coming by the way of devicetree. It already does special handling for ACPI, it needs to do the same for OF. I think SPI does the right thing there.

I'll let Wolfram comment here for I2C.

Can you comment on this ? As of today i2c-based drivers will probe with DT-based boot if DT's compatible is set to the same id as in struct i2c_device_id. No i2c driver with of_device_id can autoprobe using OF ids.