I2C IRQ Probe Improvements

PATCH In review Charles Keepax 2 participants 10 messages May 21, 2019 – May 22, 2019

Subsystems (4)

ACPI I2C subsystem host drivers I2C I2C ACPI support

lore

Share & tools

lore

Charles Keepax

author

2019-05-21 15:04 UTC #0 lore raw

cover I2C IRQ Probe Improvements

Apologies for the resend adding Andy and Jarkko.

This series attempts to align as much IRQ handling into the probe path as possible. Note that I don't have a great setup for testing these patches so they are mostly just build tested and need careful review and testing before any of them are merged.

The series brings the ACPI path inline with the way the device tree path handles the IRQ entirely at probe time. However, it still leaves any IRQ specified through the board_info as being handled at device time. In that case we need to cache something from the board_info until probe time, which leaves any alternative solution with something basically the same as the current handling although perhaps caching more stuff.

Thanks,
Charles

See previous discussions:
- https://lkml.org/lkml/2019/2/15/989
- https://www.spinics.net/lists/linux-i2c/msg39541.html

Charles Keepax (5):
i2c: acpi: Factor out getting the IRQ from ACPI
i2c: acpi: Use available IRQ helper functions
i2c: core: Move ACPI IRQ handling to probe time
i2c: core: Move ACPI gpio IRQ handling into i2c_acpi_get_irq
i2c: core: Tidy up handling of init_irq

 drivers/i2c/i2c-core-acpi.c | 50 ++++++++++++++++++++++++++++++---------------
 drivers/i2c/i2c-core-base.c | 11 +++++-----
 drivers/i2c/i2c-core.h      |  9 ++++++++
 3 files changed, 48 insertions(+), 22 deletions(-)

Signature · 1 line

2.11.0

Charles Keepax

author

2019-05-21 15:04 UTC #1 lore raw

PATCH 1/5 i2c: acpi: Factor out getting the IRQ from ACPI

In preparation for future refactoring factor out the fetch of the IRQ into its own helper function.

Signed-off-by Charles Keepax <[email protected]>

---
 drivers/i2c/i2c-core-acpi.c | 40 +++++++++++++++++++++++++---------------
 1 file changed, 25 insertions(+), 15 deletions(-)

drivers/i2c/i2c-core-acpi.c +25 ~~-15~~

index 2728006920888..c3ac3ea184317 100644
@@ -137,13 +137,35 @@ static int i2c_acpi_do_lookup(struct acpi_device *adev,
 	return 0;
 }
 
+static int i2c_acpi_get_irq(struct acpi_device *adev, int *irq)
+{
+	struct list_head resource_list;
+	struct resource_entry *entry;
+	int ret;
+
+	INIT_LIST_HEAD(&resource_list);
+
+	ret = acpi_dev_get_resources(adev, &resource_list, NULL, NULL);
+	if (ret < 0)
+		return -EINVAL;
+
+	resource_list_for_each_entry(entry, &resource_list) {
+		if (resource_type(entry->res) == IORESOURCE_IRQ) {
+			*irq = entry->res->start;
+			break;
+		}
+	}
+
+	acpi_dev_free_resource_list(&resource_list);
+
+	return 0;
+}
+
 static int i2c_acpi_get_info(struct acpi_device *adev,
 			     struct i2c_board_info *info,
 			     struct i2c_adapter *adapter,
 			     acpi_handle *adapter_handle)
 {
-	struct list_head resource_list;
-	struct resource_entry *entry;
 	struct i2c_acpi_lookup lookup;
 	int ret;

@@ -175,19 +197,7 @@ static int i2c_acpi_get_info(struct acpi_device *adev,
 		*adapter_handle = lookup.adapter_handle;
 
 	/* Then fill IRQ number if any */
-	INIT_LIST_HEAD(&resource_list);
-	ret = acpi_dev_get_resources(adev, &resource_list, NULL, NULL);
-	if (ret < 0)
-		return -EINVAL;
-
-	resource_list_for_each_entry(entry, &resource_list) {
-		if (resource_type(entry->res) == IORESOURCE_IRQ) {
-			info->irq = entry->res->start;
-			break;
-		}
-	}
-
-	acpi_dev_free_resource_list(&resource_list);
+	i2c_acpi_get_irq(adev, &info->irq);
 
 	acpi_set_modalias(adev, dev_name(&adev->dev), info->type,
 			  sizeof(info->type));

Signature · 1 line

2.11.0

Charles Keepax

author

2019-05-21 15:04 UTC #2 lore raw

PATCH 2/5 i2c: acpi: Use available IRQ helper functions

In reply to Charles Keepax

Use the available IRQ helper functions, most of the functions have additional helpful side affects like configuring the trigger type of the IRQ.

Signed-off-by Charles Keepax <[email protected]>

---
 drivers/i2c/i2c-core-acpi.c | 23 ++++++++++++++---------
 drivers/i2c/i2c-core-base.c |  4 ++--
 drivers/i2c/i2c-core.h      |  2 ++
 3 files changed, 18 insertions(+), 11 deletions(-)

drivers/i2c/i2c-core-acpi.c +14 -9

index c3ac3ea184317..764cd10420a74 100644
@@ -137,25 +137,30 @@ static int i2c_acpi_do_lookup(struct acpi_device *adev,
 	return 0;
 }
 
+static int i2c_acpi_add_resource(struct acpi_resource *ares, void *data)
+{
+	int *irq = data;
+	struct resource r;
+
+	if (*irq < 0 && acpi_dev_resource_interrupt(ares, 0, &r))
+		*irq = i2c_dev_irq_from_resources(&r, 1);
+
+	return 1; /* No need to add resource to the list */
+}
+
 static int i2c_acpi_get_irq(struct acpi_device *adev, int *irq)
 {
 	struct list_head resource_list;
-	struct resource_entry *entry;
 	int ret;
 
 	INIT_LIST_HEAD(&resource_list);
+	*irq = -ENOENT;
 
-	ret = acpi_dev_get_resources(adev, &resource_list, NULL, NULL);
+	ret = acpi_dev_get_resources(adev, &resource_list,
+				     i2c_acpi_add_resource, &irq);
 	if (ret < 0)
 		return -EINVAL;
 
-	resource_list_for_each_entry(entry, &resource_list) {
-		if (resource_type(entry->res) == IORESOURCE_IRQ) {
-			*irq = entry->res->start;
-			break;
-		}
-	}
-
 	acpi_dev_free_resource_list(&resource_list);
 
 	return 0;

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

index d389d4fb0623a..84bf11b25a120 100644
@@ -687,8 +687,8 @@ static void i2c_dev_set_name(struct i2c_adapter *adap,
 		     i2c_encode_flags_to_addr(client));
 }
 
-static int i2c_dev_irq_from_resources(const struct resource *resources,
-				      unsigned int num_resources)
+int i2c_dev_irq_from_resources(const struct resource *resources,
+			       unsigned int num_resources)
 {
 	struct irq_data *irqd;
 	int i;

drivers/i2c/i2c-core.h +2 -0

index c88cfef813431..8f3a08dc73a25 100644
@@ -28,6 +28,8 @@ extern struct list_head	__i2c_board_list;
 extern int		__i2c_first_dynamic_bus_num;
 
 int i2c_check_7bit_addr_validity_strict(unsigned short addr);
+int i2c_dev_irq_from_resources(const struct resource *resources,
+			       unsigned int num_resources);
 
 /*
  * We only allow atomic transfers for very late communication, e.g. to send

Signature · 1 line

2.11.0

Andy Shevchenko

2019-05-21 17:16 UTC #3 lore raw

Re: [PATCH 2/5] i2c: acpi: Use available IRQ helper functions

Quoted 3 lines · Charles Keepax

Use the available IRQ helper functions, most of the functions have additional helpful side affects like configuring the trigger type of the IRQ.

More logical to me is to do this as a first patch in the series. Besides that, I would really split it into two:
- exporting helper;
- using it.

Signature · 2 lines

With Best Regards,
Andy Shevchenko

Charles Keepax

author

2019-05-21 15:05 UTC #4 lore raw

PATCH 3/5 i2c: core: Move ACPI IRQ handling to probe time

In reply to Charles Keepax

Bring the ACPI path in sync with the device tree path and handle all the IRQ fetching at probe time. This leaves the only IRQ handling at device registration time being that which is passed directly through the board info as either a resource or an actual IRQ number.

Signed-off-by Charles Keepax <[email protected]>

---
 drivers/i2c/i2c-core-acpi.c | 6 ++----
 drivers/i2c/i2c-core-base.c | 6 +++++-
 drivers/i2c/i2c-core.h      | 7 +++++++
 3 files changed, 14 insertions(+), 5 deletions(-)

drivers/i2c/i2c-core-acpi.c +2 -4

index 764cd10420a74..e332760bf9ebc 100644
@@ -148,8 +148,9 @@ static int i2c_acpi_add_resource(struct acpi_resource *ares, void *data)
 	return 1; /* No need to add resource to the list */
 }
 
-static int i2c_acpi_get_irq(struct acpi_device *adev, int *irq)
+int i2c_acpi_get_irq(struct i2c_client *client, int *irq)
 {
+	struct acpi_device *adev = ACPI_COMPANION(&client->adapter->dev);
 	struct list_head resource_list;
 	int ret;

@@ -201,9 +202,6 @@ static int i2c_acpi_get_info(struct acpi_device *adev,
 	if (adapter_handle)
 		*adapter_handle = lookup.adapter_handle;
 
-	/* Then fill IRQ number if any */
-	i2c_acpi_get_irq(adev, &info->irq);
-
 	acpi_set_modalias(adev, dev_name(&adev->dev), info->type,
 			  sizeof(info->type));

drivers/i2c/i2c-core-base.c +5 -1

index 84bf11b25a120..c1afa17a76bfc 100644
@@ -335,7 +335,11 @@ static int i2c_device_probe(struct device *dev)
 			if (irq == -EINVAL || irq == -ENODATA)
 				irq = of_irq_get(dev->of_node, 0);
 		} else if (ACPI_COMPANION(dev)) {
-			irq = acpi_dev_gpio_irq_get(ACPI_COMPANION(dev), 0);
+			i2c_acpi_get_irq(client, &irq);
+
+			if (irq == -ENOENT)
+				irq = acpi_dev_gpio_irq_get(ACPI_COMPANION(dev),
+							    0);
 		}
 		if (irq == -EPROBE_DEFER)
 			return irq;

drivers/i2c/i2c-core.h +7 -0

index 8f3a08dc73a25..6bec145ab7d74 100644
@@ -72,6 +72,8 @@ const struct acpi_device_id *
 i2c_acpi_match_device(const struct acpi_device_id *matches,
 		      struct i2c_client *client);
 void i2c_acpi_register_devices(struct i2c_adapter *adap);
+
+int i2c_acpi_get_irq(struct i2c_client *client, int *irq);
 #else /* CONFIG_ACPI */
 static inline void i2c_acpi_register_devices(struct i2c_adapter *adap) { }
 static inline const struct acpi_device_id *

@@ -80,6 +82,11 @@ i2c_acpi_match_device(const struct acpi_device_id *matches,
 {
 	return NULL;
 }
+
+static inline int i2c_acpi_get_irq(struct i2c_client *client, int *irq)
+{
+	return 0;
+}
 #endif /* CONFIG_ACPI */
 extern struct notifier_block i2c_acpi_notifier;

Signature · 1 line

2.11.0

Andy Shevchenko

2019-05-21 17:18 UTC #5 lore raw

Re: [PATCH 3/5] i2c: core: Move ACPI IRQ handling to probe time

drivers/i2c/i2c-core-acpi.c static int i2c_acpi_add_resource(struct acpi_resource *ares, void *data) · line 151

Quoted 7 lines · Charles Keepax

Bring the ACPI path in sync with the device tree path and handle all the IRQ fetching at probe time. This leaves the only IRQ handling at device registration time being that which is passed directly through the board info as either a resource or an actual IRQ number.
drivers/i2c/i2c-core-acpi.c +1 -1
-static int i2c_acpi_get_irq(struct acpi_device *adev, int *irq)
+int i2c_acpi_get_irq(struct i2c_client *client, int *irq)

Instead of ping-pong style of the programming, I would rather do this in the same patch, where you introduce a helper (patch 2 I suppose, after my comments against current patch 2 in the series).

Also possible to split exporting itself to a separate patch.

Signature · 2 lines

With Best Regards,
Andy Shevchenko

Charles Keepax

author

2019-05-21 15:05 UTC #6 lore raw

PATCH 4/5 i2c: core: Move ACPI gpio IRQ handling into i2c_acpi_get_irq

In reply to Charles Keepax

It makes sense to contain all the ACPI IRQ handling in a single helper function.

Signed-off-by Charles Keepax <[email protected]>

---

Note that this one is somewhat interesting, it seems the search through the resource list is done against the companion device of the adapter but the GPIO search is done against the companion device of the client. It feels to me like these really should be done on the same device, and certainly this is what SPI does (both against the equivalent of the adapter). Perhaps someone with more ACPI knowledge than myself could comment?

Thanks,
Charles

 drivers/i2c/i2c-core-acpi.c | 3 +++
 drivers/i2c/i2c-core-base.c | 4 ----
 2 files changed, 3 insertions(+), 4 deletions(-)

drivers/i2c/i2c-core-acpi.c +3 -0

index e332760bf9ebc..0c882d956e9a4 100644
@@ -164,6 +164,9 @@ int i2c_acpi_get_irq(struct i2c_client *client, int *irq)
 
 	acpi_dev_free_resource_list(&resource_list);
 
+	if (*irq < 0)
+		*irq = acpi_dev_gpio_irq_get(ACPI_COMPANION(&client->dev), 0);
+
 	return 0;
 }

drivers/i2c/i2c-core-base.c +0 -4

index c1afa17a76bfc..f958b50c78c04 100644
@@ -336,10 +336,6 @@ static int i2c_device_probe(struct device *dev)
 				irq = of_irq_get(dev->of_node, 0);
 		} else if (ACPI_COMPANION(dev)) {
 			i2c_acpi_get_irq(client, &irq);
-
-			if (irq == -ENOENT)
-				irq = acpi_dev_gpio_irq_get(ACPI_COMPANION(dev),
-							    0);
 		}
 		if (irq == -EPROBE_DEFER)
 			return irq;

Signature · 1 line

2.11.0

Andy Shevchenko

2019-05-21 17:27 UTC #7 lore raw

Re: [PATCH 4/5] i2c: core: Move ACPI gpio IRQ handling into i2c_acpi_get_irq

Quoted 10 lines · Charles Keepax

It makes sense to contain all the ACPI IRQ handling in a single helper function.
Note that this one is somewhat interesting, it seems the search through the resource list is done against the companion device of the adapter but the GPIO search is done against the companion device of the client. It feels to me like these really should be done on the same device, and certainly this is what SPI does (both against the equivalent of the adapter). Perhaps someone with more ACPI knowledge than myself could comment?

It would be interesting to see the path how you come to this conclusion.

drivers/i2c/i2c-core-acpi.c int i2c_acpi_get_irq(struct i2c_client *client, int *irq) · line 168

Quoted 4 lines

drivers/i2c/i2c-core-acpi.c +2 -0

 	acpi_dev_free_resource_list(&resource_list);
 
+	if (*irq < 0)
+		*irq = acpi_dev_gpio_irq_get(ACPI_COMPANION(&client->dev), 0);

I think adev here is what we may use here.

You may put assert here and see if it happens when you test your series.

Signature · 2 lines

With Best Regards,
Andy Shevchenko

Charles Keepax

author

2019-05-22 08:19 UTC #8 lore raw

Re: [PATCH 4/5] i2c: core: Move ACPI gpio IRQ handling into i2c_acpi_get_irq

Quoted 12 lines · Andy Shevchenko

Quoted 10 lines · Charles Keepax
It makes sense to contain all the ACPI IRQ handling in a single helper function.
Note that this one is somewhat interesting, it seems the search through the resource list is done against the companion device of the adapter but the GPIO search is done against the companion device of the client. It feels to me like these really should be done on the same device, and certainly this is what SPI does (both against the equivalent of the adapter). Perhaps someone with more ACPI knowledge than myself could comment?

Quoted 2 lines

It would be interesting to see the path how you come to this conclusion.

Apologies but I am not sure which conclusion you are referencing. Assuming it is them being called with different acpi_device's. It is perhaps me misunderstanding things but it looks like i2c_acpi_get_info implies the adev should correspond to the adapter. Where as acpi_dev_gpio_irq_get is called with the result of ACPI_COMPANION(dev) where dev is client->dev.

Quoted 7 lines

drivers/i2c/i2c-core-acpi.c int i2c_acpi_get_irq(struct i2c_client *client, int *irq) · line 168
Quoted 4 lines
drivers/i2c/i2c-core-acpi.c +2 -0
 	acpi_dev_free_resource_list(&resource_list);
 
+	if (*irq < 0)
+		*irq = acpi_dev_gpio_irq_get(ACPI_COMPANION(&client->dev), 0);
I think adev here is what we may use here.

Indeed that is what I would have expected as well, I will update the code to do so and hopefully any issues will come out in testing.

Quoted 2 lines

You may put assert here and see if it happens when you test your series.

Alas I don't have a good way to test this series, they come out of some additional work Wolfram wanted based on some issues caused by a device tree fix I made a while back.

Thanks,
Charles

Charles Keepax

author

2019-05-21 15:05 UTC #9 lore raw

PATCH 5/5 i2c: core: Tidy up handling of init_irq

In reply to Charles Keepax

Only set init_irq during i2c_device_new and only handle client->irq on the probe/remove paths.

Signed-off-by Charles Keepax <[email protected]>

---
 drivers/i2c/i2c-core-base.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

drivers/i2c/i2c-core-base.c +3 -2

index f958b50c78c04..c0a52802d23e7 100644
@@ -322,6 +322,8 @@ static int i2c_device_probe(struct device *dev)
 
 	driver = to_i2c_driver(dev->driver);
 
+	client->irq = client->init_irq;
+
 	if (!client->irq && !driver->disable_i2c_core_irq_mapping) {
 		int irq = -ENOENT;

@@ -432,7 +434,7 @@ static int i2c_device_remove(struct device *dev)
 	dev_pm_clear_wake_irq(&client->dev);
 	device_init_wakeup(&client->dev, false);
 
-	client->irq = client->init_irq;
+	client->irq = 0;
 	if (client->flags & I2C_CLIENT_HOST_NOTIFY)
 		pm_runtime_put(&client->adapter->dev);

@@ -749,7 +751,6 @@ i2c_new_client_device(struct i2c_adapter *adap, struct i2c_board_info const *inf
 	if (!client->init_irq)
 		client->init_irq = i2c_dev_irq_from_resources(info->resources,
 							 info->num_resources);
-	client->irq = client->init_irq;
 
 	strlcpy(client->name, info->type, sizeof(client->name));

Signature · 1 line

2.11.0