Skip to contentSkip to conversation

iio: adc: qcom-spmi-adc5: Fix the channel name

PATCH Superseded Andy Shevchenko 2 participants 5 messages Jan 17, 2023 – Jan 18, 2023

Subsystems (2)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 iio: adc: qcom-spmi-adc5: Fix the channel name

The node name can contain an address part which is not used by the driver. Cut it out before assigning the channel name.

Fixes 4f47a236a23d ("iio: adc: qcom-spmi-adc5: convert to device properties")
---
 drivers/iio/adc/qcom-spmi-adc5.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)
drivers/iio/adc/qcom-spmi-adc5.c +8 -1
index e90c299c913a..c52bf6722a6a 100644
@@ -628,12 +628,19 @@ static int adc5_get_fw_channel_data(struct adc5_chip *adc,
 				    struct fwnode_handle *fwnode,
 				    const struct adc5_data *data)
 {
-	const char *name = fwnode_get_name(fwnode), *channel_name;
+	const char *name, *channel_name;
 	u32 chan, value, varr[2];
 	u32 sid = 0;
 	int ret;
 	struct device *dev = adc->dev;
 
+	name = devm_kasprintf(dev, GFP_KERNEL, "%pfwP", fwnode);
+	if (!name)
+		return -ENOMEM;
+
+	/* Cut the address part */
+	name[strchrnul(name, '@') - name] = '\0';
+
 	ret = fwnode_property_read_u32(fwnode, "reg", &chan);
 	if (ret) {
 		dev_err(dev, "invalid channel number %s\n", name);
Signature · 1 line
2.39.0
#1 lore raw
Quoted 2 lines · Andy Shevchenko

The node name can contain an address part which is not used by the driver. Cut it out before assigning the channel name.

This explanation doesn't cut it. It's not that the driver "doesn't use" the address part, it is that this string is propagated into the userspace label, sysfs /filenames/ *and breaking ABI*.

Quoted 12 lines
Fixes 4f47a236a23d ("iio: adc: qcom-spmi-adc5: convert to device properties")
---
 drivers/iio/adc/qcom-spmi-adc5.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)
drivers/iio/adc/qcom-spmi-adc5.c +0 -0
index e90c299c913a..c52bf6722a6a 100644

drivers/iio/adc/qcom-spmi-adc5.c static int adc5_get_fw_channel_data(struct adc5_chip *adc, · line 631

Quoted 5 lines
drivers/iio/adc/qcom-spmi-adc5.c +1 -1
@@ -628,12 +628,19 @@ static int adc5_get_fw_channel_data(struct adc5_chip *adc,
 				    struct fwnode_handle *fwnode,
 				    const struct adc5_data *data)
 {
-	const char *name = fwnode_get_name(fwnode), *channel_name;
+	const char *name, *channel_name;

I don't think this'll compile as name is still a pointer to const data, while you're assigning (a '\0' char) to it below.

- Marijn

Quoted 18 lines
drivers/iio/adc/qcom-spmi-adc5.c +7 -0
 	u32 chan, value, varr[2];
 	u32 sid = 0;
 	int ret;
 	struct device *dev = adc->dev;
 
+	name = devm_kasprintf(dev, GFP_KERNEL, "%pfwP", fwnode);
+	if (!name)
+		return -ENOMEM;
+
+	/* Cut the address part */
+	name[strchrnul(name, '@') - name] = '\0';
+
 	ret = fwnode_property_read_u32(fwnode, "reg", &chan);
 	if (ret) {
 		dev_err(dev, "invalid channel number %s\n", name);
Signature · 1 line
2.39.0
#2 lore raw
Quoted 5 lines · Marijn Suijten
Quoted 2 lines · Andy Shevchenko

The node name can contain an address part which is not used by the driver. Cut it out before assigning the channel name.

This explanation doesn't cut it. It's not that the driver "doesn't use"

Driver doesn't use it still. There is no contradiction, but I agree that below part is good to have in the commit message.

Quoted 2 lines

the address part, it is that this string is propagated into the userspace label, sysfs /filenames/ *and breaking ABI*.

So I will add it into v2 in case the fix works (see below).

...

Quoted 5 lines

drivers/iio/adc/qcom-spmi-adc5.c static int adc5_get_fw_channel_data(struct adc5_chip *adc, · line 631

Quoted 2 lines
drivers/iio/adc/qcom-spmi-adc5.c +1 -1
-	const char *name = fwnode_get_name(fwnode), *channel_name;
+	const char *name, *channel_name;

I don't think this'll compile as name is still a pointer to const data, while you're assigning (a '\0' char) to it below.

Right, it's always hard for me to compile things for ARM on x86 :-) Thanks for catching this up!

But does this fix the issue after compilation fix?

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#3 lore raw
Quoted 9 lines · Andy Shevchenko
Quoted 5 lines · Marijn Suijten
Quoted 2 lines · Andy Shevchenko

The node name can contain an address part which is not used by the driver. Cut it out before assigning the channel name.

This explanation doesn't cut it. It's not that the driver "doesn't use"

Driver doesn't use it still. There is no contradiction, but I agree that below part is good to have in the commit message.

You can leave that in if you want but that's not the issue that I reported/described. Having both describes the situation in full.

Quoted 7 lines
Quoted 2 lines

the address part, it is that this string is propagated into the userspace label, sysfs /filenames/ *and breaking ABI*.

So I will add it into v2 in case the fix works (see below).

...

Quoted 8 lines
Quoted 5 lines

drivers/iio/adc/qcom-spmi-adc5.c static int adc5_get_fw_channel_data(struct adc5_chip *adc, · line 631

Quoted 2 lines
drivers/iio/adc/qcom-spmi-adc5.c +1 -1
-	const char *name = fwnode_get_name(fwnode), *channel_name;
+	const char *name, *channel_name;

I don't think this'll compile as name is still a pointer to const data, while you're assigning (a '\0' char) to it below.

Right, it's always hard for me to compile things for ARM on x86 :-) Thanks for catching this up!

Thanks for sending this in regardless; as said before I rather break ABI and clean the driver up properly (no more extend_name...) than sending a fix like this :)

Quoted 1 line

But does this fix the issue after compilation fix?

It does, no more @xx in sysfs filenames nor label contents!

- Marijn

#4 lore raw
Quoted 3 lines · Marijn Suijten
Quoted 2 lines · Andy Shevchenko
Quoted 1 line · Marijn Suijten

On 2023-01-17 11:39:44, Andy Shevchenko wrote:

...

Quoted 6 lines
Quoted 6 lines
Quoted 5 lines

drivers/iio/adc/qcom-spmi-adc5.c static int adc5_get_fw_channel_data(struct adc5_chip *adc, · line 631

Quoted 2 lines
drivers/iio/adc/qcom-spmi-adc5.c +1 -1
-	const char *name = fwnode_get_name(fwnode), *channel_name;
+	const char *name, *channel_name;

I don't think this'll compile as name is still a pointer to const data, while you're assigning (a '\0' char) to it below.

Quoted 6 lines
Quoted 2 lines

Right, it's always hard for me to compile things for ARM on x86 :-) Thanks for catching this up!

Thanks for sending this in regardless; as said before I rather break ABI and clean the driver up properly (no more extend_name...) than sending a fix like this :)

Yes, but we need to backport something and ABI breakage is definitely not an option for that.

Quoted 3 lines
Quoted 1 line

But does this fix the issue after compilation fix?

It does, no more @xx in sysfs filenames nor label contents!

Okay, I'm about to send v2 and hopefully you can give your Tested-by.

Signature · 2 lines
With Best Regards,
Andy Shevchenko