Skip to contentSkip to conversation

i2c: Fix modalias for ACPI enumerated I2C devices

PATCH In review Jarkko Nikula 11 participants 23 messages Oct 10, 2013 – Oct 16, 2013 8 messages not in the archive

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 i2c: Fix modalias for ACPI enumerated I2C devices

There is a minor fault about ACPI enumerated I2C devices with their modalias attribute. Now modalias is set by device instance not by hardware ID. For example "i2c:INTABCD:00", "i2c:INTABCD:01" etc.

This means each device instance gets different modalias which does match with generated modules.alias. Currently this is not problem as matching can happen also with "acpi:INTABCD" modalias.

Fix this by using ACPI hardware ID.

Cc Mika Westerberg <[email protected]>

---
Generated on top of v3.12-rc4-29-g0e7a3ed.

---
 drivers/i2c/i2c-core.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/i2c/i2c-core.c +1 -1
index 29d3f04..6dd0c53 100644
@@ -1111,7 +1111,7 @@ static acpi_status acpi_i2c_add_device(acpi_handle handle, u32 level,
 	if (ret < 0 || !info.addr)
 		return AE_OK;
 
-	strlcpy(info.type, dev_name(&adev->dev), sizeof(info.type));
+	strlcpy(info.type, acpi_device_hid(adev), sizeof(info.type));
 	if (!i2c_new_device(adapter, &info)) {
 		dev_err(&adapter->dev,
 			"failed to add I2C device %s from ACPI\n",
Signature · 1 line
1.8.4.rc3

A message that is not in the archive (sent off-list or not yet received): 1381414669-26115-1-git-send-email-jarkko.nikula-VuQAYsv1563Yd54FQh9/[email protected]

#2 lore raw

+Rafael

Quoted 3 lines · Jarkko Nikula

There is a minor fault about ACPI enumerated I2C devices with their modalias attribute. Now modalias is set by device instance not by hardware ID. For example "i2c:INTABCD:00", "i2c:INTABCD:01" etc.

I think that this is intentional. We don't want that the i2c modalias matches with the ACPI device (like with the i2c:INTABCD). Instead use ACPI IDs that are added to the driver to match with the ACPI device.

Quoted 29 lines

This means each device instance gets different modalias which does match with generated modules.alias. Currently this is not problem as matching can happen also with "acpi:INTABCD" modalias.

Fix this by using ACPI hardware ID.

Cc Mika Westerberg <mika.westerberg-VuQAYsv1563Yd54FQh9/[email protected]>

---
Generated on top of v3.12-rc4-29-g0e7a3ed.

---
 drivers/i2c/i2c-core.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/i2c/i2c-core.c +1 -1
index 29d3f04..6dd0c53 100644
@@ -1111,7 +1111,7 @@ static acpi_status acpi_i2c_add_device(acpi_handle handle, u32 level,
 	if (ret < 0 || !info.addr)
 		return AE_OK;
 
-	strlcpy(info.type, dev_name(&adev->dev), sizeof(info.type));
+	strlcpy(info.type, acpi_device_hid(adev), sizeof(info.type));
 	if (!i2c_new_device(adapter, &info)) {
 		dev_err(&adapter->dev,
 			"failed to add I2C device %s from ACPI\n",
Signature · 1 line
1.8.4.rc3

A message that is not in the archive (sent off-list or not yet received): [email protected]

#4 lore raw
Quoted 10 lines · Mika Westerberg

+Rafael

Quoted 3 lines · Jarkko Nikula

There is a minor fault about ACPI enumerated I2C devices with their modalias attribute. Now modalias is set by device instance not by hardware ID. For example "i2c:INTABCD:00", "i2c:INTABCD:01" etc.

I think that this is intentional. We don't want that the i2c modalias matches with the ACPI device (like with the i2c:INTABCD). Instead use ACPI IDs that are added to the driver to match with the ACPI device.

Well, I'm not really sure this was intentional, but I wonder how other bus types work in that respect?

Quoted 28 lines
Quoted 27 lines

This means each device instance gets different modalias which does match with generated modules.alias. Currently this is not problem as matching can happen also with "acpi:INTABCD" modalias.

Fix this by using ACPI hardware ID.

Cc Mika Westerberg <mika.westerberg-VuQAYsv1563Yd54FQh9/[email protected]>

---
Generated on top of v3.12-rc4-29-g0e7a3ed.

---
 drivers/i2c/i2c-core.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/i2c/i2c-core.c +1 -1
index 29d3f04..6dd0c53 100644
@@ -1111,7 +1111,7 @@ static acpi_status acpi_i2c_add_device(acpi_handle handle, u32 level,
 	if (ret < 0 || !info.addr)
 		return AE_OK;
 
-	strlcpy(info.type, dev_name(&adev->dev), sizeof(info.type));
+	strlcpy(info.type, acpi_device_hid(adev), sizeof(info.type));
 	if (!i2c_new_device(adapter, &info)) {
 		dev_err(&adapter->dev,
 			"failed to add I2C device %s from ACPI\n",
Signature · 2 lines
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.
#5 lore raw
Quoted 6 lines · Rafael J. Wysocki
Quoted 3 lines

I think that this is intentional. We don't want that the i2c modalias matches with the ACPI device (like with the i2c:INTABCD). Instead use ACPI IDs that are added to the driver to match with the ACPI device.

Well, I'm not really sure this was intentional, but I wonder how other bus types work in that respect?

We have the same for platform bus, if that's what you are asking.

It probably doesn't hurt to have this patch applied but it might cause inadvertent match if for some reason there is an I2C client driver that happens to have INTABCD I2C id in its list.

A message that is not in the archive (sent off-list or not yet received): [email protected]

#7 lore raw
Quoted 5 lines · Mika Westerberg
Quoted 4 lines · Rafael J. Wysocki
Quoted 3 lines

I think that this is intentional. We don't want that the i2c modalias matches with the ACPI device (like with the i2c:INTABCD). Instead use ACPI IDs that are added to the driver to match with the ACPI device.

Quoted 8 lines
Quoted 2 lines

Well, I'm not really sure this was intentional, but I wonder how other bus types work in that respect?

We have the same for platform bus, if that's what you are asking.

It probably doesn't hurt to have this patch applied but it might cause inadvertent match if for some reason there is an I2C client driver that happens to have INTABCD I2C id in its list.

Well, if they have that id in their lists, they are supposed to be able to handle this device, aren't they? What other reason may be there for them to put that id into their lists?

Signature · 2 lines
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.

A message that is not in the archive (sent off-list or not yet received): [email protected]

#9 lore raw
Quoted 11 lines · Rafael J. Wysocki
Quoted 10 lines · Mika Westerberg
Quoted 6 lines · Rafael J. Wysocki
Quoted 3 lines

I think that this is intentional. We don't want that the i2c modalias matches with the ACPI device (like with the i2c:INTABCD). Instead use ACPI IDs that are added to the driver to match with the ACPI device.

Well, I'm not really sure this was intentional, but I wonder how other bus types work in that respect?

We have the same for platform bus, if that's what you are asking.

Quoted 7 lines
Quoted 3 lines

It probably doesn't hurt to have this patch applied but it might cause inadvertent match if for some reason there is an I2C client driver that happens to have INTABCD I2C id in its list.

Well, if they have that id in their lists, they are supposed to be able to handle this device, aren't they? What other reason may be there for them to put that id into their lists?

If we have two ACPI enumerated devices, they have following modalias:

i2c-device0: i2c:INTABCD:00
acpi:INTABCD

i2c-device1: i2c:INTABCD:01
acpi:INTABCD

Likelihood that some random I2C driver has INTABCD:00 or INTABCD:01 ids in their list is minimal. However, when you turn it to this:

i2c-device0: i2c:INTABCD
acpi:INTABCD

i2c-device1: i2c:INTABCD
acpi:INTABCD

It might be possible that we get a match that isn't supposed to happen. Well, OK it is pretty remote but anyway :-)

#10 lore raw
Quoted 12 lines · Mika Westerberg

If we have two ACPI enumerated devices, they have following modalias:

   i2c-device0:	i2c:INTABCD:00
		acpi:INTABCD

   i2c-device1:	i2c:INTABCD:01
   		acpi:INTABCD

Likelihood that some random I2C driver has INTABCD:00 or INTABCD:01 ids in their list is minimal. However, when you turn it to this:

Quoted 8 lines
   i2c-device0:	i2c:INTABCD
		acpi:INTABCD

   i2c-device1:	i2c:INTABCD
   		acpi:INTABCD

It might be possible that we get a match that isn't supposed to happen. Well, OK it is pretty remote but anyway :-)

Well, name conflicts could occur of course but still I don't think we should generate illegal or wrong modaliases. I'm not an udev expert but I suppose trying to load nonexisting drivers (i2c_INTABCD:xy) could slow booting a little and perhaps pollute needlessly error log compared to if it can see that driver is already loaded or tries to load the same driver again.

I don't think name conflicts can pose too big risk as they are trivial to fix in sources and can be queued to stable too.

Signature · 1 line
Jarkko

A message that is not in the archive (sent off-list or not yet received): 525B9321.5000906-VuQAYsv1563Yd54FQh9/[email protected]

#12 lore raw
Quoted 28 lines · Jarkko Nikula
Quoted 12 lines · Mika Westerberg

If we have two ACPI enumerated devices, they have following modalias:

i2c-device0: i2c:INTABCD:00
acpi:INTABCD

i2c-device1: i2c:INTABCD:01
acpi:INTABCD

Likelihood that some random I2C driver has INTABCD:00 or INTABCD:01 ids in their list is minimal. However, when you turn it to this:

Quoted 8 lines

i2c-device0: i2c:INTABCD
acpi:INTABCD

i2c-device1: i2c:INTABCD
acpi:INTABCD

It might be possible that we get a match that isn't supposed to happen. Well, OK it is pretty remote but anyway :-)

Well, name conflicts could occur of course but still I don't think we should generate illegal or wrong modaliases. I'm not an udev expert but I suppose trying to load nonexisting drivers (i2c_INTABCD:xy) could slow booting a little and perhaps pollute needlessly error log compared to if it can see that driver is already loaded or tries to load the same driver again.

Quoted 2 lines

I don't think name conflicts can pose too big risk as they are trivial to fix in sources and can be queued to stable too.

OK, you got me convinced :-)

No further objections from me.

#13 lore raw
In reply to Mika Westerberg
Quoted 8 lines · Mika Westerberg
Quoted 5 lines · Rafael J. Wysocki
Quoted 3 lines

I think that this is intentional. We don't want that the i2c modalias matches with the ACPI device (like with the i2c:INTABCD). Instead use ACPI IDs that are added to the driver to match with the ACPI device.

Well, I'm not really sure this was intentional, but I wonder how other bus types work in that respect?

We have the same for platform bus, if that's what you are asking.

Do we? I don't recall seeing per device modaliases on other platforms on their platform buses.

And actually I don't see that happening in drivers/base/platform.c: platform_uevent() either where just pdev->name is used but not pdev->id (which is used with pdev->name for dev_set_name()).

This makes me thinking that perhaps "pdevinfo.name = dev_name(&adev->dev);" in drivers/acpi/acpi_platform.c: acpi_create_platform_device() should be fixed too as now modalias for ACPI registered platform devices differ from platform devices that are registered in other subsystems (e.g. regulatory, pcspkr, alarmtimer, etc devices)?

I can send a patch for that.

Signature · 1 line
Jarkko
#14 lore raw
Quoted 11 lines · Jarkko Nikula
Quoted 8 lines · Mika Westerberg
Quoted 5 lines · Rafael J. Wysocki
Quoted 3 lines

I think that this is intentional. We don't want that the i2c modalias matches with the ACPI device (like with the i2c:INTABCD). Instead use ACPI IDs that are added to the driver to match with the ACPI device.

Well, I'm not really sure this was intentional, but I wonder how other bus types work in that respect?

We have the same for platform bus, if that's what you are asking.

Do we? I don't recall seeing per device modaliases on other platforms on their platform buses.

I mean for platform devices enumerated from ACPI.

Quoted 10 lines

And actually I don't see that happening in drivers/base/platform.c:
platform_uevent() either where just pdev->name is used but not pdev->id (which is used with pdev->name for dev_set_name()).

This makes me thinking that perhaps "pdevinfo.name = dev_name(&adev->dev);" in drivers/acpi/acpi_platform.c:
acpi_create_platform_device() should be fixed too as now modalias for ACPI registered platform devices differ from platform devices that are registered in other subsystems (e.g. regulatory, pcspkr, alarmtimer, etc devices)?

Well, if you think that it doesn't hit us back later if we get a match that isn't supposed to happen.

#15 lore raw
In reply to a message not in the archive

Hi,

Quoted 8 lines · Jarkko Nikula

There is a minor fault about ACPI enumerated I2C devices with their modalias attribute. Now modalias is set by device instance not by hardware ID. For example "i2c:INTABCD:00", "i2c:INTABCD:01" etc.

This means each device instance gets different modalias which does match with generated modules.alias. Currently this is not problem as matching can happen also with "acpi:INTABCD" modalias.

IMO, this is not the proper fix for the modalias problem because ACPI enumerated I2C device may have compatible ids.
Instead, we should export all the compatible ids as the modules alias of the ACPI enumerated I2C device.

can you please take a look at the patch I sent out earlier?
https://patchwork.kernel.org/patch/3034991/
https://patchwork.kernel.org/patch/3035041/
https://patchwork.kernel.org/patch/3035021/

thanks,
rui

Quoted 25 lines

Fix this by using ACPI hardware ID.

Cc Mika Westerberg <mika.westerberg-VuQAYsv1563Yd54FQh9/[email protected]>

---
Generated on top of v3.12-rc4-29-g0e7a3ed.

---
 drivers/i2c/i2c-core.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/i2c/i2c-core.c +1 -1
index 29d3f04..6dd0c53 100644
@@ -1111,7 +1111,7 @@ static acpi_status acpi_i2c_add_device(acpi_handle handle, u32 level,
 	if (ret < 0 || !info.addr)
 		return AE_OK;
 
-	strlcpy(info.type, dev_name(&adev->dev), sizeof(info.type));
+	strlcpy(info.type, acpi_device_hid(adev), sizeof(info.type));
 	if (!i2c_new_device(adapter, &info)) {
 		dev_err(&adapter->dev,
 			"failed to add I2C device %s from ACPI\n",
#16 lore raw
Quoted 16 lines · Zhang Rui

Hi,

Quoted 8 lines · Jarkko Nikula

There is a minor fault about ACPI enumerated I2C devices with their modalias attribute. Now modalias is set by device instance not by hardware ID. For example "i2c:INTABCD:00", "i2c:INTABCD:01" etc.

This means each device instance gets different modalias which does match with generated modules.alias. Currently this is not problem as matching can happen also with "acpi:INTABCD" modalias.

IMO, this is not the proper fix for the modalias problem because ACPI enumerated I2C device may have compatible ids.
Instead, we should export all the compatible ids as the modules alias of the ACPI enumerated I2C device.

Quoted 4 lines

can you please take a look at the patch I sent out earlier?
https://patchwork.kernel.org/patch/3034991/
https://patchwork.kernel.org/patch/3035041/
https://patchwork.kernel.org/patch/3035021/

I see. This makes sense as it avoids that same device has two different modaliases from both acpi and other subsystem.

How about modalias nodes in sysfs, should they also reflect what is matching uvent?

Signature · 1 line
Jarkko
#17 lore raw
Quoted 21 lines · Jarkko Nikula
Quoted 16 lines · Zhang Rui

Hi,

Quoted 8 lines · Jarkko Nikula

There is a minor fault about ACPI enumerated I2C devices with their modalias attribute. Now modalias is set by device instance not by hardware ID. For example "i2c:INTABCD:00", "i2c:INTABCD:01" etc.

This means each device instance gets different modalias which does match with generated modules.alias. Currently this is not problem as matching can happen also with "acpi:INTABCD" modalias.

IMO, this is not the proper fix for the modalias problem because ACPI enumerated I2C device may have compatible ids.
Instead, we should export all the compatible ids as the modules alias of the ACPI enumerated I2C device.

Quoted 4 lines

can you please take a look at the patch I sent out earlier?
https://patchwork.kernel.org/patch/3034991/
https://patchwork.kernel.org/patch/3035041/
https://patchwork.kernel.org/patch/3035021/

Quoted 6 lines

I see. This makes sense as it avoids that same device has two different modaliases from both acpi and other subsystem.

How about modalias nodes in sysfs, should they also reflect what is matching uvent?

good catch, will fix "modalias" as well in next version.

thanks,
rui

#18 lore raw
Quoted 22 lines · Zhang Rui
Quoted 21 lines · Jarkko Nikula
Quoted 20 lines · Zhang Rui

Hi,

Quoted 8 lines · Jarkko Nikula

There is a minor fault about ACPI enumerated I2C devices with their modalias attribute. Now modalias is set by device instance not by hardware ID. For example "i2c:INTABCD:00", "i2c:INTABCD:01" etc.

This means each device instance gets different modalias which does match with generated modules.alias. Currently this is not problem as matching can happen also with "acpi:INTABCD" modalias.

IMO, this is not the proper fix for the modalias problem because ACPI enumerated I2C device may have compatible ids.
Instead, we should export all the compatible ids as the modules alias of the ACPI enumerated I2C device.

can you please take a look at the patch I sent out earlier?
https://patchwork.kernel.org/patch/3034991/
https://patchwork.kernel.org/patch/3035041/
https://patchwork.kernel.org/patch/3035021/

Quoted 7 lines
Quoted 6 lines

I see. This makes sense as it avoids that same device has two different modaliases from both acpi and other subsystem.

How about modalias nodes in sysfs, should they also reflect what is matching uvent?

good catch, will fix "modalias" as well in next version.

Hi,

I have a question about the device "uevent" and "modalias" sysfs attributes.
what is the relationship between these two?
Am I right to say that, if there is the "MODALIAS" field in uevent file, this field must be consistent with the content in "modalias" attribute?

I checked the code in drivers/base/platform.c,
static ssize_t modalias_show(struct device *dev, struct device_attribute *a,
char *buf)

{
        struct platform_device  *pdev = to_platform_device(dev);
        int len = snprintf(buf, PAGE_SIZE, "platform:%s\n", pdev->name);

        return (len >= PAGE_SIZE) ? (PAGE_SIZE - 1) : len;
}

static int platform_uevent(struct device *dev, struct kobj_uevent_env *env)

{
        struct platform_device  *pdev = to_platform_device(dev);
        int rc;

        /* Some devices have extra OF data and an OF-style MODALIAS */
        rc = of_device_uevent_modalias(dev, env);
        if (rc != -ENODEV)
                return rc;

        add_uevent_var(env, "MODALIAS=%s%s", PLATFORM_MODULE_PREFIX,
                        pdev->name);
        return 0;
}

This means that the OF-style MODALIAS is not shown in "modalias" sysfs attribute.
is this a bug?

thanks,
rui

#19 lore raw
Quoted 66 lines · Zhang Rui
Quoted 22 lines · Zhang Rui
Quoted 21 lines · Jarkko Nikula
Quoted 20 lines · Zhang Rui

Hi,

Quoted 8 lines · Jarkko Nikula

There is a minor fault about ACPI enumerated I2C devices with their modalias attribute. Now modalias is set by device instance not by hardware ID. For example "i2c:INTABCD:00", "i2c:INTABCD:01" etc.

This means each device instance gets different modalias which does match with generated modules.alias. Currently this is not problem as matching can happen also with "acpi:INTABCD" modalias.

IMO, this is not the proper fix for the modalias problem because ACPI enumerated I2C device may have compatible ids.
Instead, we should export all the compatible ids as the modules alias of the ACPI enumerated I2C device.

can you please take a look at the patch I sent out earlier?
https://patchwork.kernel.org/patch/3034991/
https://patchwork.kernel.org/patch/3035041/
https://patchwork.kernel.org/patch/3035021/

Quoted 7 lines
Quoted 6 lines

I see. This makes sense as it avoids that same device has two different modaliases from both acpi and other subsystem.

How about modalias nodes in sysfs, should they also reflect what is matching uvent?

good catch, will fix "modalias" as well in next version.

Hi,

I have a question about the device "uevent" and "modalias" sysfs attributes.
what is the relationship between these two?
Am I right to say that, if there is the "MODALIAS" field in uevent file, this field must be consistent with the content in "modalias" attribute?

I checked the code in drivers/base/platform.c,
static ssize_t modalias_show(struct device *dev, struct device_attribute *a,
char *buf)

{
        struct platform_device  *pdev = to_platform_device(dev);
        int len = snprintf(buf, PAGE_SIZE, "platform:%s\n", pdev->name);

        return (len >= PAGE_SIZE) ? (PAGE_SIZE - 1) : len;
}

static int platform_uevent(struct device *dev, struct kobj_uevent_env *env)

{
        struct platform_device  *pdev = to_platform_device(dev);
        int rc;

        /* Some devices have extra OF data and an OF-style MODALIAS */
        rc = of_device_uevent_modalias(dev, env);
        if (rc != -ENODEV)
                return rc;

        add_uevent_var(env, "MODALIAS=%s%s", PLATFORM_MODULE_PREFIX,
                        pdev->name);
        return 0;
}
Quoted 3 lines

This means that the OF-style MODALIAS is not shown in "modalias" sysfs attribute.
is this a bug?

I would consider that as a bug, but I'm not sure what the recommended practice is. Greg?

Signature · 2 lines
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.

A message that is not in the archive (sent off-list or not yet received): [email protected]

#21 lore raw
Quoted 32 lines · Rafael J. Wysocki
Quoted 31 lines · Zhang Rui
Quoted 29 lines · Zhang Rui
Quoted 21 lines · Jarkko Nikula
Quoted 16 lines · Zhang Rui

Hi,

Quoted 8 lines · Jarkko Nikula

There is a minor fault about ACPI enumerated I2C devices with their modalias attribute. Now modalias is set by device instance not by hardware ID. For example "i2c:INTABCD:00", "i2c:INTABCD:01" etc.

This means each device instance gets different modalias which does match with generated modules.alias. Currently this is not problem as matching can happen also with "acpi:INTABCD" modalias.

IMO, this is not the proper fix for the modalias problem because ACPI enumerated I2C device may have compatible ids.
Instead, we should export all the compatible ids as the modules alias of the ACPI enumerated I2C device.

Quoted 4 lines

can you please take a look at the patch I sent out earlier?
https://patchwork.kernel.org/patch/3034991/
https://patchwork.kernel.org/patch/3035041/
https://patchwork.kernel.org/patch/3035021/

Quoted 6 lines

I see. This makes sense as it avoids that same device has two different modaliases from both acpi and other subsystem.

How about modalias nodes in sysfs, should they also reflect what is matching uvent?

good catch, will fix "modalias" as well in next version.

Quoted 7 lines
Quoted 7 lines

Hi,

I have a question about the device "uevent" and "modalias" sysfs attributes.
what is the relationship between these two?
Am I right to say that, if there is the "MODALIAS" field in uevent file, this field must be consistent with the content in "modalias" attribute?

Well, if it isn't, it's pretty pointless, right?

Quoted 27 lines
Quoted 27 lines

I checked the code in drivers/base/platform.c,
static ssize_t modalias_show(struct device *dev, struct device_attribute *a,
char *buf)

{
        struct platform_device  *pdev = to_platform_device(dev);
        int len = snprintf(buf, PAGE_SIZE, "platform:%s\n", pdev->name);

        return (len >= PAGE_SIZE) ? (PAGE_SIZE - 1) : len;
}

static int platform_uevent(struct device *dev, struct kobj_uevent_env *env)

{
        struct platform_device  *pdev = to_platform_device(dev);
        int rc;

        /* Some devices have extra OF data and an OF-style MODALIAS */
        rc = of_device_uevent_modalias(dev, env);
        if (rc != -ENODEV)
                return rc;

        add_uevent_var(env, "MODALIAS=%s%s", PLATFORM_MODULE_PREFIX,
                        pdev->name);
        return 0;
}
Quoted 6 lines
Quoted 3 lines

This means that the OF-style MODALIAS is not shown in "modalias" sysfs attribute.
is this a bug?

I would consider that as a bug, but I'm not sure what the recommended practice is. Greg?

I have no idea how the OF stuff is working, and honestly, I really have no wish to ever know anything about it. Especially when it comes to platform devices/drivers, something that I personally hate and wish would be deleted.

So go ask the OF maintainers/developers, this is their domain :)

sorry,

greg k-h

#22 lore raw
Quoted 77 lines · Greg Kroah-Hartman
Quoted 32 lines · Rafael J. Wysocki
Quoted 31 lines · Zhang Rui
Quoted 29 lines · Zhang Rui
Quoted 21 lines · Jarkko Nikula
Quoted 16 lines · Zhang Rui

Hi,

Quoted 8 lines · Jarkko Nikula

There is a minor fault about ACPI enumerated I2C devices with their modalias attribute. Now modalias is set by device instance not by hardware ID. For example "i2c:INTABCD:00", "i2c:INTABCD:01" etc.

This means each device instance gets different modalias which does match with generated modules.alias. Currently this is not problem as matching can happen also with "acpi:INTABCD" modalias.

IMO, this is not the proper fix for the modalias problem because ACPI enumerated I2C device may have compatible ids.
Instead, we should export all the compatible ids as the modules alias of the ACPI enumerated I2C device.

Quoted 4 lines

can you please take a look at the patch I sent out earlier?
https://patchwork.kernel.org/patch/3034991/
https://patchwork.kernel.org/patch/3035041/
https://patchwork.kernel.org/patch/3035021/

Quoted 6 lines

I see. This makes sense as it avoids that same device has two different modaliases from both acpi and other subsystem.

How about modalias nodes in sysfs, should they also reflect what is matching uvent?

good catch, will fix "modalias" as well in next version.

Quoted 7 lines
Quoted 7 lines

Hi,

I have a question about the device "uevent" and "modalias" sysfs attributes.
what is the relationship between these two?
Am I right to say that, if there is the "MODALIAS" field in uevent file, this field must be consistent with the content in "modalias" attribute?

Well, if it isn't, it's pretty pointless, right?

Quoted 27 lines
Quoted 27 lines

I checked the code in drivers/base/platform.c,
static ssize_t modalias_show(struct device *dev, struct device_attribute *a,
char *buf)

{
        struct platform_device  *pdev = to_platform_device(dev);
        int len = snprintf(buf, PAGE_SIZE, "platform:%s\n", pdev->name);

        return (len >= PAGE_SIZE) ? (PAGE_SIZE - 1) : len;
}

static int platform_uevent(struct device *dev, struct kobj_uevent_env *env)

{
        struct platform_device  *pdev = to_platform_device(dev);
        int rc;

        /* Some devices have extra OF data and an OF-style MODALIAS */
        rc = of_device_uevent_modalias(dev, env);
        if (rc != -ENODEV)
                return rc;

        add_uevent_var(env, "MODALIAS=%s%s", PLATFORM_MODULE_PREFIX,
                        pdev->name);
        return 0;
}
Quoted 6 lines
Quoted 3 lines

This means that the OF-style MODALIAS is not shown in "modalias" sysfs attribute.
is this a bug?

I would consider that as a bug, but I'm not sure what the recommended practice is. Greg?

Quoted 6 lines

I have no idea how the OF stuff is working, and honestly, I really have no wish to ever know anything about it. Especially when it comes to platform devices/drivers, something that I personally hate and wish would be deleted.

So go ask the OF maintainers/developers, this is their domain :)

Well, OK. Whom in particular?

Rafael

#23 lore raw
Quoted 78 lines · Rafael J. Wysocki
Quoted 77 lines · Greg Kroah-Hartman
Quoted 32 lines · Rafael J. Wysocki
Quoted 31 lines · Zhang Rui
Quoted 29 lines · Zhang Rui
Quoted 21 lines · Jarkko Nikula
Quoted 16 lines · Zhang Rui

Hi,

Quoted 8 lines · Jarkko Nikula

There is a minor fault about ACPI enumerated I2C devices with their modalias attribute. Now modalias is set by device instance not by hardware ID. For example "i2c:INTABCD:00", "i2c:INTABCD:01" etc.

This means each device instance gets different modalias which does match with generated modules.alias. Currently this is not problem as matching can happen also with "acpi:INTABCD" modalias.

IMO, this is not the proper fix for the modalias problem because ACPI enumerated I2C device may have compatible ids.
Instead, we should export all the compatible ids as the modules alias of the ACPI enumerated I2C device.

Quoted 4 lines

can you please take a look at the patch I sent out earlier?
https://patchwork.kernel.org/patch/3034991/
https://patchwork.kernel.org/patch/3035041/
https://patchwork.kernel.org/patch/3035021/

Quoted 6 lines

I see. This makes sense as it avoids that same device has two different modaliases from both acpi and other subsystem.

How about modalias nodes in sysfs, should they also reflect what is matching uvent?

good catch, will fix "modalias" as well in next version.

Quoted 7 lines
Quoted 7 lines

Hi,

I have a question about the device "uevent" and "modalias" sysfs attributes.
what is the relationship between these two?
Am I right to say that, if there is the "MODALIAS" field in uevent file, this field must be consistent with the content in "modalias" attribute?

Well, if it isn't, it's pretty pointless, right?

Quoted 33 lines
Quoted 27 lines

I checked the code in drivers/base/platform.c,
static ssize_t modalias_show(struct device *dev, struct device_attribute *a,
char *buf)

{
        struct platform_device  *pdev = to_platform_device(dev);
        int len = snprintf(buf, PAGE_SIZE, "platform:%s\n", pdev->name);

        return (len >= PAGE_SIZE) ? (PAGE_SIZE - 1) : len;
}

static int platform_uevent(struct device *dev, struct kobj_uevent_env *env)

{
        struct platform_device  *pdev = to_platform_device(dev);
        int rc;

        /* Some devices have extra OF data and an OF-style MODALIAS */
        rc = of_device_uevent_modalias(dev, env);
        if (rc != -ENODEV)
                return rc;

        add_uevent_var(env, "MODALIAS=%s%s", PLATFORM_MODULE_PREFIX,
                        pdev->name);
        return 0;
}
Quoted 3 lines

This means that the OF-style MODALIAS is not shown in "modalias" sysfs attribute.
is this a bug?

I would consider that as a bug, but I'm not sure what the recommended practice is. Greg?

Quoted 8 lines
Quoted 6 lines

I have no idea how the OF stuff is working, and honestly, I really have no wish to ever know anything about it. Especially when it comes to platform devices/drivers, something that I personally hate and wish would be deleted.

So go ask the OF maintainers/developers, this is their domain :)

Well, OK. Whom in particular?

The "OPEN FIRMWARE AND FLATTENED DEVICE TREE" entries in MAINTAINERS?

#24 lore raw
Quoted 84 lines · Greg Kroah-Hartman
Quoted 83 lines · Rafael J. Wysocki
Quoted 82 lines · Greg Kroah-Hartman
Quoted 32 lines · Rafael J. Wysocki
Quoted 31 lines · Zhang Rui
Quoted 29 lines · Zhang Rui
Quoted 21 lines · Jarkko Nikula
Quoted 16 lines · Zhang Rui

Hi,

Quoted 8 lines · Jarkko Nikula

There is a minor fault about ACPI enumerated I2C devices with their modalias attribute. Now modalias is set by device instance not by hardware ID. For example "i2c:INTABCD:00", "i2c:INTABCD:01" etc.

This means each device instance gets different modalias which does match with generated modules.alias. Currently this is not problem as matching can happen also with "acpi:INTABCD" modalias.

IMO, this is not the proper fix for the modalias problem because ACPI enumerated I2C device may have compatible ids.
Instead, we should export all the compatible ids as the modules alias of the ACPI enumerated I2C device.

Quoted 4 lines

can you please take a look at the patch I sent out earlier?
https://patchwork.kernel.org/patch/3034991/
https://patchwork.kernel.org/patch/3035041/
https://patchwork.kernel.org/patch/3035021/

Quoted 6 lines

I see. This makes sense as it avoids that same device has two different modaliases from both acpi and other subsystem.

How about modalias nodes in sysfs, should they also reflect what is matching uvent?

good catch, will fix "modalias" as well in next version.

Quoted 7 lines
Quoted 7 lines

Hi,

I have a question about the device "uevent" and "modalias" sysfs attributes.
what is the relationship between these two?
Am I right to say that, if there is the "MODALIAS" field in uevent file, this field must be consistent with the content in "modalias" attribute?

Well, if it isn't, it's pretty pointless, right?

Quoted 27 lines
Quoted 27 lines

I checked the code in drivers/base/platform.c,
static ssize_t modalias_show(struct device *dev, struct device_attribute *a,
char *buf)

{
        struct platform_device  *pdev = to_platform_device(dev);
        int len = snprintf(buf, PAGE_SIZE, "platform:%s\n", pdev->name);

        return (len >= PAGE_SIZE) ? (PAGE_SIZE - 1) : len;
}

static int platform_uevent(struct device *dev, struct kobj_uevent_env *env)

{
        struct platform_device  *pdev = to_platform_device(dev);
        int rc;

        /* Some devices have extra OF data and an OF-style MODALIAS */
        rc = of_device_uevent_modalias(dev, env);
        if (rc != -ENODEV)
                return rc;

        add_uevent_var(env, "MODALIAS=%s%s", PLATFORM_MODULE_PREFIX,
                        pdev->name);
        return 0;
}
Quoted 6 lines
Quoted 3 lines

This means that the OF-style MODALIAS is not shown in "modalias" sysfs attribute.
is this a bug?

I would consider that as a bug, but I'm not sure what the recommended practice is. Greg?

I have no idea how the OF stuff is working, and honestly, I really have no wish to ever know anything about it. Especially when it comes to platform devices/drivers, something that I personally hate and wish would be deleted.

Quoted 5 lines
Quoted 3 lines
Quoted 1 line

So go ask the OF maintainers/developers, this is their domain :)

Well, OK. Whom in particular?

The "OPEN FIRMWARE AND FLATTENED DEVICE TREE" entries in MAINTAINERS?

Hmm, that looks like Grant and Rob Herring.

Grant, Rob, do we want the OF-style MODALIAS to be shown in "modalias" sysfs attributes of platform devices?

Rafael

#25 lore raw
Quoted 77 lines · Rafael J. Wysocki
Quoted 76 lines · Greg Kroah-Hartman
Quoted 75 lines · Rafael J. Wysocki
Quoted 74 lines · Greg Kroah-Hartman
Quoted 32 lines · Rafael J. Wysocki
Quoted 31 lines · Zhang Rui
Quoted 29 lines · Zhang Rui
Quoted 21 lines · Jarkko Nikula
Quoted 16 lines · Zhang Rui

Hi,

Quoted 8 lines · Jarkko Nikula

There is a minor fault about ACPI enumerated I2C devices with their modalias attribute. Now modalias is set by device instance not by hardware ID. For example "i2c:INTABCD:00", "i2c:INTABCD:01" etc.

This means each device instance gets different modalias which does match with generated modules.alias. Currently this is not problem as matching can happen also with "acpi:INTABCD" modalias.

IMO, this is not the proper fix for the modalias problem because ACPI enumerated I2C device may have compatible ids.
Instead, we should export all the compatible ids as the modules alias of the ACPI enumerated I2C device.

Quoted 4 lines

can you please take a look at the patch I sent out earlier?
https://patchwork.kernel.org/patch/3034991/
https://patchwork.kernel.org/patch/3035041/
https://patchwork.kernel.org/patch/3035021/

Quoted 6 lines

I see. This makes sense as it avoids that same device has two different modaliases from both acpi and other subsystem.

How about modalias nodes in sysfs, should they also reflect what is matching uvent?

good catch, will fix "modalias" as well in next version.

Quoted 7 lines
Quoted 7 lines

Hi,

I have a question about the device "uevent" and "modalias" sysfs attributes.
what is the relationship between these two?
Am I right to say that, if there is the "MODALIAS" field in uevent file, this field must be consistent with the content in "modalias" attribute?

Well, if it isn't, it's pretty pointless, right?

Quoted 31 lines
Quoted 30 lines

I checked the code in drivers/base/platform.c,
static ssize_t modalias_show(struct device *dev, struct device_attribute *a,
char *buf)

{
        struct platform_device  *pdev = to_platform_device(dev);
        int len = snprintf(buf, PAGE_SIZE, "platform:%s\n", pdev->name);

        return (len >= PAGE_SIZE) ? (PAGE_SIZE - 1) : len;
}

static int platform_uevent(struct device *dev, struct kobj_uevent_env *env)

{
        struct platform_device  *pdev = to_platform_device(dev);
        int rc;

        /* Some devices have extra OF data and an OF-style MODALIAS */
        rc = of_device_uevent_modalias(dev, env);
        if (rc != -ENODEV)
                return rc;

        add_uevent_var(env, "MODALIAS=%s%s", PLATFORM_MODULE_PREFIX,
                        pdev->name);
        return 0;
}

This means that the OF-style MODALIAS is not shown in "modalias" sysfs attribute.
is this a bug?

Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 2 lines

I would consider that as a bug, but I'm not sure what the recommended practice is. Greg?

I have no idea how the OF stuff is working, and honestly, I really have no wish to ever know anything about it. Especially when it comes to platform devices/drivers, something that I personally hate and wish would be deleted.

<digress>Greg, I've heard you say that a lot, but regardless of what platform devices/drivers were originally designed for, it is pretty much exactly what we need for non-discoverable memory mapped busses. I've yet to heard a viable alternative proposed. I've heard the proposal of creating new bus types and new driver binding to that bus type for each variant of a non-discoverable memory mapped bus, but I think it is a non-starter. There are too many combinations. What /might/ work to replace the platform_bus_type would be to have a mechanism for drivers to transparently bind to multiple bus types, but then I suspect that it will end up looking an awful lot like the existing platform_bus_type.

Probably worth discussing over beer next week</digress>

Quoted 11 lines
Quoted 6 lines
Quoted 4 lines
Quoted 2 lines

So go ask the OF maintainers/developers, this is their domain :)

Well, OK. Whom in particular?

The "OPEN FIRMWARE AND FLATTENED DEVICE TREE" entries in MAINTAINERS?

Hmm, that looks like Grant and Rob Herring.

Grant, Rob, do we want the OF-style MODALIAS to be shown in "modalias" sysfs attributes of platform devices?

I've not actually had to deal with modutils and modalias on any of the platforms I've worked with so I've only looked at it superficially. Yes, that does look like a bug, but Ben would probably have a much better idea [cc'd].

g.

A message that is not in the archive (sent off-list or not yet received): CACxGe6v3JHLKHqBXux=1mgm227S2dLafaqohKvzqsw1uo6tHyA-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org

#27 lore raw
Quoted 98 lines · Grant Likely
Quoted 77 lines · Rafael J. Wysocki
Quoted 76 lines · Greg Kroah-Hartman
Quoted 75 lines · Rafael J. Wysocki
Quoted 74 lines · Greg Kroah-Hartman
Quoted 32 lines · Rafael J. Wysocki
Quoted 31 lines · Zhang Rui
Quoted 29 lines · Zhang Rui
Quoted 21 lines · Jarkko Nikula
Quoted 16 lines · Zhang Rui

Hi,

Quoted 8 lines · Jarkko Nikula

There is a minor fault about ACPI enumerated I2C devices with their modalias attribute. Now modalias is set by device instance not by hardware ID. For example "i2c:INTABCD:00", "i2c:INTABCD:01" etc.

This means each device instance gets different modalias which does match with generated modules.alias. Currently this is not problem as matching can happen also with "acpi:INTABCD" modalias.

IMO, this is not the proper fix for the modalias problem because ACPI enumerated I2C device may have compatible ids.
Instead, we should export all the compatible ids as the modules alias of the ACPI enumerated I2C device.

Quoted 4 lines

can you please take a look at the patch I sent out earlier?
https://patchwork.kernel.org/patch/3034991/
https://patchwork.kernel.org/patch/3035041/
https://patchwork.kernel.org/patch/3035021/

Quoted 6 lines

I see. This makes sense as it avoids that same device has two different modaliases from both acpi and other subsystem.

How about modalias nodes in sysfs, should they also reflect what is matching uvent?

good catch, will fix "modalias" as well in next version.

Quoted 7 lines
Quoted 7 lines

Hi,

I have a question about the device "uevent" and "modalias" sysfs attributes.
what is the relationship between these two?
Am I right to say that, if there is the "MODALIAS" field in uevent file, this field must be consistent with the content in "modalias" attribute?

Well, if it isn't, it's pretty pointless, right?

Quoted 31 lines
Quoted 30 lines

I checked the code in drivers/base/platform.c,
static ssize_t modalias_show(struct device *dev, struct device_attribute *a,
char *buf)

{
        struct platform_device  *pdev = to_platform_device(dev);
        int len = snprintf(buf, PAGE_SIZE, "platform:%s\n", pdev->name);

        return (len >= PAGE_SIZE) ? (PAGE_SIZE - 1) : len;
}

static int platform_uevent(struct device *dev, struct kobj_uevent_env *env)

{
        struct platform_device  *pdev = to_platform_device(dev);
        int rc;

        /* Some devices have extra OF data and an OF-style MODALIAS */
        rc = of_device_uevent_modalias(dev, env);
        if (rc != -ENODEV)
                return rc;

        add_uevent_var(env, "MODALIAS=%s%s", PLATFORM_MODULE_PREFIX,
                        pdev->name);
        return 0;
}

This means that the OF-style MODALIAS is not shown in "modalias" sysfs attribute.
is this a bug?

Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 2 lines

I would consider that as a bug, but I'm not sure what the recommended practice is. Greg?

I have no idea how the OF stuff is working, and honestly, I really have no wish to ever know anything about it. Especially when it comes to platform devices/drivers, something that I personally hate and wish would be deleted.

<digress>Greg, I've heard you say that a lot, but regardless of what platform devices/drivers were originally designed for, it is pretty much exactly what we need for non-discoverable memory mapped busses. I've yet to heard a viable alternative proposed. I've heard the proposal of creating new bus types and new driver binding to that bus type for each variant of a non-discoverable memory mapped bus, but I think it is a non-starter. There are too many combinations. What /might/ work to replace the platform_bus_type would be to have a mechanism for drivers to transparently bind to multiple bus types, but then I suspect that it will end up looking an awful lot like the existing platform_bus_type.

Quoted 1 line

Probably worth discussing over beer next week</digress>

I think we have a whole session about this at the ARM summit next week, and if someone wants to bring beer into it, that will make me happy!

#28 lore raw
In reply to Grant Likely
Quoted 23 lines · Grant Likely
Quoted 22 lines · Rafael J. Wysocki
Quoted 21 lines · Greg Kroah-Hartman
Quoted 20 lines · Rafael J. Wysocki
Quoted 19 lines · Greg Kroah-Hartman
Quoted 6 lines · Rafael J. Wysocki
Quoted 5 lines · Zhang Rui

I have a question about the device "uevent" and "modalias" sysfs attributes.
what is the relationship between these two?
Am I right to say that, if there is the "MODALIAS" field in uevent file, this field must be consistent with the content in "modalias" attribute?

Well, if it isn't, it's pretty pointless, right?

Quoted 11 lines
Quoted 11 lines

static int platform_uevent(struct device *dev, struct kobj_uevent_env *env)

{
         struct platform_device  *pdev = to_platform_device(dev);
         int rc;

         /* Some devices have extra OF data and an OF-style MODALIAS */
         rc = of_device_uevent_modalias(dev, env);
         if (rc != -ENODEV)
                 return rc;
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
         add_uevent_var(env, "MODALIAS=%s%s", PLATFORM_MODULE_PREFIX,
                         pdev->name);
         return 0;
}

This means that the OF-style MODALIAS is not shown in "modalias" sysfs attribute.
is this a bug?

Here is an example from one DT based system:

cat /sys/bus/platform/devices/48070000.i2c/uevent
DRIVER=omap_i2c
OF_NAME=i2c
OF_FULLNAME=/ocp/i2c@48070000
OF_COMPATIBLE_0=ti,omap4-i2c
OF_COMPATIBLE_N=1
MODALIAS=of:Ni2cT<NULL>Cti,omap4-i2c

cat /sys/bus/platform/devices/48070000.i2c/modalias platform:48070000.i2c

And a device on that I2C bus:

cat /sys/bus/platform/devices/rtc.11/uevent
DRIVER=twl_rtc
OF_NAME=rtc
OF_FULLNAME=/ocp/i2c@48070000/twl@48/rtc
OF_COMPATIBLE_0=ti,twl4030-rtc
OF_COMPATIBLE_N=1
MODALIAS=of:NrtcT<NULL>Cti,twl4030-rtc

cat /sys/bus/platform/devices/rtc.11/modalias
platform:rtc.11

Unfortunately I cannot debug above example further at the moment is there failing or needless modprobe calls. Maybe device tree experts know better?

Signature · 1 line
Jarkko

A message that is not in the archive (sent off-list or not yet received): [email protected]

#30 lore raw
Quoted 77 lines · Greg Kroah-Hartman
Quoted 32 lines · Rafael J. Wysocki
Quoted 31 lines · Zhang Rui
Quoted 29 lines · Zhang Rui
Quoted 21 lines · Jarkko Nikula
Quoted 16 lines · Zhang Rui

Hi,

Quoted 8 lines · Jarkko Nikula

There is a minor fault about ACPI enumerated I2C devices with their modalias attribute. Now modalias is set by device instance not by hardware ID. For example "i2c:INTABCD:00", "i2c:INTABCD:01" etc.

This means each device instance gets different modalias which does match with generated modules.alias. Currently this is not problem as matching can happen also with "acpi:INTABCD" modalias.

IMO, this is not the proper fix for the modalias problem because ACPI enumerated I2C device may have compatible ids.
Instead, we should export all the compatible ids as the modules alias of the ACPI enumerated I2C device.

Quoted 4 lines

can you please take a look at the patch I sent out earlier?
https://patchwork.kernel.org/patch/3034991/
https://patchwork.kernel.org/patch/3035041/
https://patchwork.kernel.org/patch/3035021/

Quoted 6 lines

I see. This makes sense as it avoids that same device has two different modaliases from both acpi and other subsystem.

How about modalias nodes in sysfs, should they also reflect what is matching uvent?

good catch, will fix "modalias" as well in next version.

Quoted 7 lines
Quoted 7 lines

Hi,

I have a question about the device "uevent" and "modalias" sysfs attributes.
what is the relationship between these two?
Am I right to say that, if there is the "MODALIAS" field in uevent file, this field must be consistent with the content in "modalias" attribute?

Well, if it isn't, it's pretty pointless, right?

Quoted 27 lines
Quoted 27 lines

I checked the code in drivers/base/platform.c,
static ssize_t modalias_show(struct device *dev, struct device_attribute *a,
char *buf)

{
        struct platform_device  *pdev = to_platform_device(dev);
        int len = snprintf(buf, PAGE_SIZE, "platform:%s\n", pdev->name);

        return (len >= PAGE_SIZE) ? (PAGE_SIZE - 1) : len;
}

static int platform_uevent(struct device *dev, struct kobj_uevent_env *env)

{
        struct platform_device  *pdev = to_platform_device(dev);
        int rc;

        /* Some devices have extra OF data and an OF-style MODALIAS */
        rc = of_device_uevent_modalias(dev, env);
        if (rc != -ENODEV)
                return rc;

        add_uevent_var(env, "MODALIAS=%s%s", PLATFORM_MODULE_PREFIX,
                        pdev->name);
        return 0;
}
Quoted 6 lines
Quoted 3 lines

This means that the OF-style MODALIAS is not shown in "modalias" sysfs attribute.
is this a bug?

I would consider that as a bug, but I'm not sure what the recommended practice is. Greg?

Quoted 6 lines

I have no idea how the OF stuff is working, and honestly, I really have no wish to ever know anything about it. Especially when it comes to platform devices/drivers, something that I personally hate and wish would be deleted.

So go ask the OF maintainers/developers, this is their domain :)

So Rui, let's consider that as a bug and please prepare a patch to fix it.

Thanks!

Signature · 2 lines
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.