Skip to contentSkip to conversation

ACPI: bus: Use OF match data for PRP0001 matched devices

PATCH Superseded Kartik Rajput 4 participants 8 messages Jan 7 – Jan 8

Subsystems (2)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 ACPI: bus: Use OF match data for PRP0001 matched devices

When a device is matched via PRP0001, the driver's OF (DT) match table must be used to obtain the device match data. If a driver provides both an acpi_match_table and an of_match_table, the current acpi_device_get_match_data() path consults the driver's acpi_match_table and returns NULL (no ACPI ID matches).

Explicitly detect PRP0001 and fetch match data from the driver's of_match_table via acpi_of_device_get_match_data().

Fixes 886ca88be6b3 ("ACPI / bus: Respect PRP0001 when retrieving device match data")

---
Changes in v2:
* Fix build errors.

---
 drivers/acpi/bus.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/acpi/bus.c +2 -1
index 5e110badac7b..6658c4339656 100644
@@ -1031,8 +1031,9 @@ const void *acpi_device_get_match_data(const struct device *dev)
 {
 	const struct acpi_device_id *acpi_ids = dev->driver->acpi_match_table;
 	const struct acpi_device_id *match;
+	struct acpi_device *adev = ACPI_COMPANION(dev);
 
-	if (!acpi_ids)
+	if (!strcmp(ACPI_DT_NAMESPACE_HID, acpi_device_hid(adev)))
 		return acpi_of_device_get_match_data(dev);
 
 	match = acpi_match_device(acpi_ids, dev);
Signature · 1 line
2.43.0
#1 lore raw

Hi Kartik,

Quoted 9 lines · Kartik Rajput

When a device is matched via PRP0001, the driver's OF (DT) match table must be used to obtain the device match data. If a driver provides both an acpi_match_table and an of_match_table, the current acpi_device_get_match_data() path consults the driver's acpi_match_table and returns NULL (no ACPI ID matches).

Explicitly detect PRP0001 and fetch match data from the driver's of_match_table via acpi_of_device_get_match_data().

Quoted 6 lines
Fixes 886ca88be6b3 ("ACPI / bus: Respect PRP0001 when retrieving device match data")

---
Changes in v2:
* Fix build errors.

Thanks for the update.

Quoted 9 lines
---
 drivers/acpi/bus.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/acpi/bus.c +0 -0
index 5e110badac7b..6658c4339656 100644

drivers/acpi/bus.c const void *acpi_device_get_match_data(const struct device *dev) · line 1036

Quoted 7 lines
drivers/acpi/bus.c +2 -1
@@ -1031,8 +1031,9 @@ const void *acpi_device_get_match_data(const struct device *dev)
 {
 	const struct acpi_device_id *acpi_ids = dev->driver->acpi_match_table;
 	const struct acpi_device_id *match;
+	struct acpi_device *adev = ACPI_COMPANION(dev);
 
-	if (!acpi_ids)
+	if (!strcmp(ACPI_DT_NAMESPACE_HID, acpi_device_hid(adev)))

I'd swap the arguments to have the static one on the right, i.e.

if (!strcmp(acpi_device_hid(adev), ACPI_DT_NAMESPACE_HID))

The patch looks good apart from that IMO.

Quoted 3 lines
drivers/acpi/bus.c +0 -0
 		return acpi_of_device_get_match_data(dev);
 
 	match = acpi_match_device(acpi_ids, dev);
Signature · 3 lines
Kind regards,

Sakari Ailus
#2 lore raw
Quoted 47 lines · Sakari Ailus

External email: Use caution opening links or attachments

Hi Kartik,

Quoted 9 lines · Kartik Rajput

When a device is matched via PRP0001, the driver's OF (DT) match table must be used to obtain the device match data. If a driver provides both an acpi_match_table and an of_match_table, the current acpi_device_get_match_data() path consults the driver's acpi_match_table and returns NULL (no ACPI ID matches).

Explicitly detect PRP0001 and fetch match data from the driver's of_match_table via acpi_of_device_get_match_data().

Quoted 6 lines
Fixes 886ca88be6b3 ("ACPI / bus: Respect PRP0001 when retrieving device match data")

---
Changes in v2:
* Fix build errors.

Thanks for the update.

Quoted 9 lines
---
  drivers/acpi/bus.c | 3 ++-
  1 file changed, 2 insertions(+), 1 deletion(-)
drivers/acpi/bus.c +0 -0
index 5e110badac7b..6658c4339656 100644

drivers/acpi/bus.c const void *acpi_device_get_match_data(const struct device *dev) · line 1036

Quoted 7 lines
drivers/acpi/bus.c +2 -1
@@ -1031,8 +1031,9 @@ const void *acpi_device_get_match_data(const struct device *dev)
  {
       const struct acpi_device_id *acpi_ids = dev->driver->acpi_match_table;
       const struct acpi_device_id *match;
+     struct acpi_device *adev = ACPI_COMPANION(dev);
 
-     if (!acpi_ids)
+     if (!strcmp(ACPI_DT_NAMESPACE_HID, acpi_device_hid(adev)))

I'd swap the arguments to have the static one on the right, i.e.

if (!strcmp(acpi_device_hid(adev), ACPI_DT_NAMESPACE_HID))

The patch looks good apart from that IMO.

Quoted 8 lines

drivers/acpi/bus.c const void *acpi_device_get_match_data(const struct device *dev) · line 1039

Quoted 3 lines
drivers/acpi/bus.c +0 -0
               return acpi_of_device_get_match_data(dev);
 
       match = acpi_match_device(acpi_ids, dev);

--
Kind regards,

Sakari Ailus

Hi Sakari,

Thanks for reviewing the patch!

I will make this change in the next version of this patch.

Thanks,
Kartik

#3 lore raw
In reply to Kartik Rajput
Quoted 25 lines · Kartik Rajput

When a device is matched via PRP0001, the driver's OF (DT) match table must be used to obtain the device match data. If a driver provides both an acpi_match_table and an of_match_table, the current acpi_device_get_match_data() path consults the driver's acpi_match_table and returns NULL (no ACPI ID matches).

Explicitly detect PRP0001 and fetch match data from the driver's of_match_table via acpi_of_device_get_match_data().

Fixes 886ca88be6b3 ("ACPI / bus: Respect PRP0001 when retrieving device match data")

---
Changes in v2:
* Fix build errors.

---
 drivers/acpi/bus.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/acpi/bus.c +0 -0
index 5e110badac7b..6658c4339656 100644

drivers/acpi/bus.c const void *acpi_device_get_match_data(const struct device *dev) · line 1035

Quoted 6 lines
drivers/acpi/bus.c +1 -1
@@ -1031,8 +1031,9 @@ const void *acpi_device_get_match_data(const struct device *dev)
 {
        const struct acpi_device_id *acpi_ids = dev->driver->acpi_match_table;
        const struct acpi_device_id *match;
+       struct acpi_device *adev = ACPI_COMPANION(dev);
 
-       if (!acpi_ids)

You still need to check acpi_ids against NULL or (better) check adev against NULL.

Quoted 5 lines
drivers/acpi/bus.c +1 -1
+       if (!strcmp(ACPI_DT_NAMESPACE_HID, acpi_device_hid(adev)))
                return acpi_of_device_get_match_data(dev);
 
        match = acpi_match_device(acpi_ids, dev);
--
#4 lore raw
Quoted 34 lines · Rafael J. Wysocki

External email: Use caution opening links or attachments

Quoted 30 lines · Kartik Rajput

When a device is matched via PRP0001, the driver's OF (DT) match table must be used to obtain the device match data. If a driver provides both an acpi_match_table and an of_match_table, the current acpi_device_get_match_data() path consults the driver's acpi_match_table and returns NULL (no ACPI ID matches).

Explicitly detect PRP0001 and fetch match data from the driver's of_match_table via acpi_of_device_get_match_data().

Fixes 886ca88be6b3 ("ACPI / bus: Respect PRP0001 when retrieving device match data")

---
Changes in v2:
* Fix build errors.

---
  drivers/acpi/bus.c | 3 ++-
  1 file changed, 2 insertions(+), 1 deletion(-)
drivers/acpi/bus.c +1 -0
index 5e110badac7b..6658c4339656 100644
@@ -1031,8 +1031,9 @@ const void *acpi_device_get_match_data(const struct device *dev)
  {
         const struct acpi_device_id *acpi_ids = dev->driver->acpi_match_table;
         const struct acpi_device_id *match;
+       struct acpi_device *adev = ACPI_COMPANION(dev);
 
Quoted 5 lines

drivers/acpi/bus.c const void *acpi_device_get_match_data(const struct device *dev) · line 1035

Quoted 1 line
drivers/acpi/bus.c +0 -1
-       if (!acpi_ids)

You still need to check acpi_ids against NULL or (better) check adev against NULL.

Hi Rafael,

Thanks for reviewing the patch.

Yes, it looks like adev should be NULL-checked. acpi_of_device_get_match_data() also uses the adev reference and passes it to acpi_of_match_device(), which checks it against NULL.

I will make this change in the next revision.

Thanks,
Kartik

#5 lore raw
In reply to Kartik Rajput
Quoted 5 lines · Kartik Rajput

When a device is matched via PRP0001, the driver's OF (DT) match table must be used to obtain the device match data. If a driver provides both an acpi_match_table and an of_match_table, the current acpi_device_get_match_data() path consults the driver's acpi_match_table and returns NULL (no ACPI ID matches).

Since we have both tables, why the actual ACPI HID of the device in question (actually which one?) can't be used?

Quoted 2 lines

Explicitly detect PRP0001 and fetch match data from the driver's of_match_table via acpi_of_device_get_match_data().

In principle we can go this way, but can you tell a bit more of a story? Why the device in question can't use existed or a newly allocated ACPI HID for that?

...

drivers/acpi/bus.c const void *acpi_device_get_match_data(const struct device *dev) · line 1034

Quoted 5 lines
drivers/acpi/bus.c +1 -0
@@ -1031,8 +1031,9 @@ const void *acpi_device_get_match_data(const struct device *dev)
 {
 	const struct acpi_device_id *acpi_ids = dev->driver->acpi_match_table;
 	const struct acpi_device_id *match;
+	struct acpi_device *adev = ACPI_COMPANION(dev);

Please, keep it in reversed xmas tree order.

Quoted 7 lines
drivers/acpi/bus.c +1 -1
-	if (!acpi_ids)
+	if (!strcmp(ACPI_DT_NAMESPACE_HID, acpi_device_hid(adev)))
 
 
 		return acpi_of_device_get_match_data(dev);
 

match = acpi_match_device(acpi_ids, dev);

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#6 lore raw

Hi Andy,

Thanks for reviewing the patch!

Quoted 13 lines · Andy Shevchenko

External email: Use caution opening links or attachments

Quoted 5 lines · Kartik Rajput

When a device is matched via PRP0001, the driver's OF (DT) match table must be used to obtain the device match data. If a driver provides both an acpi_match_table and an of_match_table, the current acpi_device_get_match_data() path consults the driver's acpi_match_table and returns NULL (no ACPI ID matches).

Since we have both tables, why the actual ACPI HID of the device in question (actually which one?) can't be used?

Quoted 7 lines
Quoted 2 lines

Explicitly detect PRP0001 and fetch match data from the driver's of_match_table via acpi_of_device_get_match_data().

In principle we can go this way, but can you tell a bit more of a story? Why the device in question can't use existed or a newly allocated ACPI HID for that?

While testing PRP0001-based matching with the Tegra fuse driver on an SoC that does not yet have an allocated ACPI HID, device_get_match_data() returned NULL because the driver also provides an acpi_match_table.

Commit 886ca88be6b3 ("ACPI / bus: Respect PRP0001 when retrieving device match data") was intended to address this by honoring PRP0001 when retrieving match data. However, when a driver provides an ACPI match table, acpi_device_get_match_data() currently consults only that table, resulting in NULL match data despite a successful PRP0001 match.

Quoted 10 lines

...

drivers/acpi/bus.c const void *acpi_device_get_match_data(const struct device *dev) · line 1034

Quoted 5 lines
drivers/acpi/bus.c +1 -0
@@ -1031,8 +1031,9 @@ const void *acpi_device_get_match_data(const struct device *dev)
  {
       const struct acpi_device_id *acpi_ids = dev->driver->acpi_match_table;
       const struct acpi_device_id *match;
+     struct acpi_device *adev = ACPI_COMPANION(dev);

Please, keep it in reversed xmas tree order.

Ack. I will update this in the next revision.

Thanks,
Kartik

#7 lore raw
Quoted 15 lines · Kartik Rajput
Quoted 14 lines · Andy Shevchenko
Quoted 5 lines · Kartik Rajput

When a device is matched via PRP0001, the driver's OF (DT) match table must be used to obtain the device match data. If a driver provides both an acpi_match_table and an of_match_table, the current acpi_device_get_match_data() path consults the driver's acpi_match_table and returns NULL (no ACPI ID matches).

Since we have both tables, why the actual ACPI HID of the device in question (actually which one?) can't be used?

Quoted 2 lines

Explicitly detect PRP0001 and fetch match data from the driver's of_match_table via acpi_of_device_get_match_data().

Quoted 6 lines
Quoted 3 lines

In principle we can go this way, but can you tell a bit more of a story? Why the device in question can't use existed or a newly allocated ACPI HID for that?

While testing PRP0001-based matching with the Tegra fuse driver on an SoC that does not yet have an allocated ACPI HID,

Which means in the production you do not need this patch. Allocate ID and go with it.

Quoted 8 lines

device_get_match_data() returned NULL because
the driver also provides an acpi_match_table.

Commit 886ca88be6b3 ("ACPI / bus: Respect PRP0001 when retrieving device match data") was intended to address this by honoring PRP0001 when retrieving match data. However, when a driver provides an ACPI match table, acpi_device_get_match_data() currently consults only that table, resulting in NULL match data despite a successful PRP0001 match.

Which is expected behaviour. So there are two cases I can see that might make this patch valid:

1) the preproduction development when the driver has both tables and formal ACPI HID is not allocated yet (this what has to be the main "why" point in the commit message);

2) a driver that got ACPI HID from somebody else and breaks the PRP0001 setups for others (no evidences so far, but I admit there is a potential to have a such).

That said, having the Fixes tag is unrequired.

Signature · 2 lines
With Best Regards,
Andy Shevchenko