Skip to contentSkip to conversation

Add bridged amplifiers to cs42l43

PATCH In review Charles Keepax 2 participants 23 messages Apr 11, 2024 – Apr 15, 2024

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

cover Add bridged amplifiers to cs42l43

On some cs42l43 systems a couple of cs35l56 amplifiers are attached to the cs42l43's SPI and I2S. On Windows the cs42l43 is controlled by a SDCA class driver and these two amplifiers are controlled by firmware running on the cs42l43. However, under Linux the decision was made to interact with the cs42l43 directly, affording the user greater control over the audio system. However, this has resulted in an issue where these two bridged cs35l56 amplifiers are not populated in ACPI and must be added manually. There is at least an SDCA extension unit DT entry we can key off.

The process of adding this is handled using a software node, firstly the ability to add native chip selects to software nodes must be added. Secondly, an additional flag for naming the SPI devices is added this allows the machine driver to key to the correct amplifier. Then finally, the cs42l43 SPI driver adds the two amplifiers directly onto its SPI bus.

An additional series will follow soon to add the audio machine driver parts (in the sof-sdw driver), however that is fairly orthogonal to this part of the process, getting the actual amplifiers registered.

Thanks,
Charles

Series changes since v4:
- Remove extraneous fwnode_handle_puts in driver/spi/spi-cs42l43.c
- Make Kconfig for swnode undef gpios not user visible
- Add some missing headers
- Add patch to update handling in spi_dev_set_name
- Remove stray blank line
- Use ACPI_HANDLE_FWNODE

Series changes since v3:
- Add Kconfig to make swnode conditionally built
- Add define for swnode name
- Add custom init and exit calls to register swnode
- Use export namespaces
- Always name swnode SPI devices after the node name
- Correct some header includes
- Use HZ_PER_MHZ
- Use some swnode helper macros
- Use acpi_get_local_address
- Correct some handle puts
- Add some dev_err_probes

Series changes since v2:
- Add missing fwnode_handle_puts in driver/spi/spi-cs423l43.c

Series changes since v1:
- Add missing statics in driver/spi/spi-cs42l43.c

Charles Keepax (2):
gpio: swnode: Add ability to specify native chip selects for SPI
spi: Add a mechanism to use the fwnode name for the SPI device

Maciej Strozek (1):
spi: cs42l43: Add bridged cs35l56 amplifiers

Charles Keepax (3):
gpio: swnode: Add ability to specify native chip selects for SPI
spi: Switch to using is_acpi_device_node() in spi_dev_set_name()
spi: Update swnode based SPI devices to use the fwnode name

Maciej Strozek (1):
spi: cs42l43: Add bridged cs35l56 amplifiers

 drivers/gpio/Kconfig          |   3 +
 drivers/gpio/gpiolib-swnode.c |  40 ++++++++++
 drivers/spi/Kconfig           |   1 +
 drivers/spi/spi-cs42l43.c     | 135 +++++++++++++++++++++++++++++++++-
 drivers/spi/spi.c             |  13 +++-
 include/linux/gpio/consumer.h |   4 +
 6 files changed, 191 insertions(+), 5 deletions(-)
Signature · 1 line
2.39.2
#1 lore raw

PATCH v5 1/4 gpio: swnode: Add ability to specify native chip selects for SPI

SPI devices can specify a cs-gpios property to enumerate their chip selects. Under device tree, a zero entry in this property can be used to specify that a particular chip select is using the SPI controllers native chip select, for example:

        cs-gpios = <&gpio1 0 0>, <0>;

Here the second chip select is native. However, when using swnodes there is currently no way to specify a native chip select. The proposal here is to register a swnode_gpio_undefined software node, that can be specified to allow the indication of a native chip select. For example:

static const struct software_node_ref_args device_cs_refs[] = {
	{
		.node  = &device_gpiochip_swnode,
		.nargs = 2,
		.args  = { 0, GPIO_ACTIVE_LOW },
	},
	{
		.node  = &swnode_gpio_undefined,
		.nargs = 0,
	},
};

Register the swnode as the gpiolib is initialised and check in swnode_get_gpio_device() if the returned node matches swnode_gpio_undefined and return -ENOENT, which matches the behaviour of the device tree system when it encounters a 0 phandle.

---

Changes since v4:
- Make GPIO_SWNODE_UNDEFINED not user visible
- Minor fixes to commit message
- Add some missing headers

Thanks,
Charles

 drivers/gpio/Kconfig          |  3 +++
 drivers/gpio/gpiolib-swnode.c | 40 +++++++++++++++++++++++++++++++++++
 include/linux/gpio/consumer.h |  4 ++++
 3 files changed, 47 insertions(+)
drivers/gpio/Kconfig +3 -0
index b50d0b470849..c44a6b57aefa 100644
@@ -103,6 +103,9 @@ config GPIO_REGMAP
 	select REGMAP
 	tristate
 
+config GPIO_SWNODE_UNDEFINED
+	bool
+
 # put drivers in the right section, in alphabetical order
 
 # This symbol is selected by both I2C and SPI expanders
drivers/gpio/gpiolib-swnode.c +40 -0
index fa52bdb1a29a..add5f8962e8d 100644
@@ -6,6 +6,8 @@
  */
 #include <linux/err.h>
 #include <linux/errno.h>
+#include <linux/export.h>
+#include <linux/init.h>
 #include <linux/kernel.h>
 #include <linux/printk.h>
 #include <linux/property.h>
@@ -17,6 +19,8 @@
 #include "gpiolib.h"
 #include "gpiolib-swnode.h"
 
+#define GPIOLIB_SWNODE_UNDEFINED_NAME "swnode-gpio-undefined"
+
 static void swnode_format_propname(const char *con_id, char *propname,
 				   size_t max_size)
 {
@@ -40,6 +44,13 @@ static struct gpio_device *swnode_get_gpio_device(struct fwnode_handle *fwnode)
 	if (!gdev_node || !gdev_node->name)
 		return ERR_PTR(-EINVAL);
 
+	/*
+	 * Check for special node that identifies undefined GPIOs, this is
+	 * primarily used as a key for internal chip selects in SPI bindings.
+	 */
+	if (!strcmp(gdev_node->name, GPIOLIB_SWNODE_UNDEFINED_NAME))
+		return ERR_PTR(-ENOENT);
+
 	gdev = gpio_device_find_by_label(gdev_node->name);
 	return gdev ?: ERR_PTR(-EPROBE_DEFER);
 }
@@ -121,3 +132,32 @@ int swnode_gpio_count(const struct fwnode_handle *fwnode, const char *con_id)
 
 	return count ?: -ENOENT;
 }
+
+#if IS_ENABLED(CONFIG_GPIO_SWNODE_UNDEFINED)
+/*
+ * A special node that identifies undefined GPIOs, this is primarily used as
+ * a key for internal chip selects in SPI bindings.
+ */
+const struct software_node swnode_gpio_undefined = {
+	.name = GPIOLIB_SWNODE_UNDEFINED_NAME,
+};
+EXPORT_SYMBOL_NS_GPL(swnode_gpio_undefined, GPIO_SWNODE);
+
+static int __init swnode_gpio_init(void)
+{
+	int ret;
+
+	ret = software_node_register(&swnode_gpio_undefined);
+	if (ret < 0)
+		pr_err("gpiolib: failed to register swnode: %d\n", ret);
+
+	return ret;
+}
+subsys_initcall(swnode_gpio_init);
+
+static void __exit swnode_gpio_cleanup(void)
+{
+	software_node_unregister(&swnode_gpio_undefined);
+}
+__exitcall(swnode_gpio_cleanup);
+#endif
include/linux/gpio/consumer.h +4 -0
index db2dfbae8edb..e685fac43398 100644
@@ -12,6 +12,8 @@ struct fwnode_handle;
 struct gpio_array;
 struct gpio_desc;
 
+struct software_node;
+
 /**
  * struct gpio_descs - Struct containing an array of descriptors that can be
  *                     obtained using gpiod_get_array()
@@ -54,6 +56,8 @@ enum gpiod_flags {
 	GPIOD_OUT_HIGH_OPEN_DRAIN = GPIOD_OUT_HIGH | GPIOD_FLAGS_BIT_OPEN_DRAIN,
 };
 
+extern const struct software_node swnode_gpio_undefined;
+
 #ifdef CONFIG_GPIOLIB
 
 /* Return the number of GPIOs associated with a device / function */
Signature · 1 line
2.39.2
#2 lore raw

Re: [PATCH v5 1/4] gpio: swnode: Add ability to specify native chip selects for SPI

Quoted 9 lines · Charles Keepax

SPI devices can specify a cs-gpios property to enumerate their chip selects. Under device tree, a zero entry in this property can be used to specify that a particular chip select is using the SPI controllers native chip select, for example:

        cs-gpios = <&gpio1 0 0>, <0>;

Here the second chip select is native. However, when using swnodes

Here, the

Quoted 21 lines

there is currently no way to specify a native chip select. The proposal here is to register a swnode_gpio_undefined software node, that can be specified to allow the indication of a native chip select. For example:

static const struct software_node_ref_args device_cs_refs[] = {
        {
                .node  = &device_gpiochip_swnode,
                .nargs = 2,
                .args  = { 0, GPIO_ACTIVE_LOW },
        },
        {
                .node  = &swnode_gpio_undefined,
                .nargs = 0,
        },
};

Register the swnode as the gpiolib is initialised and check in swnode_get_gpio_device() if the returned node matches swnode_gpio_undefined and return -ENOENT, which matches the behaviour of the device tree system when it encounters a 0 phandle.

...

drivers/gpio/Kconfig config GPIO_REGMAP · line 107

Quoted 2 lines
drivers/gpio/Kconfig +2 -0
+config GPIO_SWNODE_UNDEFINED
+       bool

But why did you remove the help? Please, put it back.

...

drivers/gpio/gpiolib-swnode.c static struct gpio_device *swnode_get_gpio_device(struct fwnode_handle *fwnode) · line 48

Quoted 2 lines
drivers/gpio/gpiolib-swnode.c +2 -0
+       /*
+        * Check for special node that identifies undefined GPIOs, this is

for a special

drivers/gpio/gpiolib-swnode.c static struct gpio_device *swnode_get_gpio_device(struct fwnode_handle *fwnode) · line 52

Quoted 4 lines
drivers/gpio/gpiolib-swnode.c +4 -0
+        * primarily used as a key for internal chip selects in SPI bindings.
+        */
+       if (!strcmp(gdev_node->name, GPIOLIB_SWNODE_UNDEFINED_NAME))
+               return ERR_PTR(-ENOENT);

This is a dead code when the respective config option is not selected. Or actually a potential flaw if somebody else names their swnode like this.

...

drivers/gpio/gpiolib-swnode.c int swnode_gpio_count(const struct fwnode_handle *fwnode, const char *con_id) · line 152

Quoted 3 lines
drivers/gpio/gpiolib-swnode.c +3 -0
+       ret = software_node_register(&swnode_gpio_undefined);
+       if (ret < 0)
+               pr_err("gpiolib: failed to register swnode: %d\n", ret);

Instead of this prefix, define pr_fmt()

Quoted 1 line

+ return ret;

...

include/linux/gpio/consumer.h

Quoted 2 lines
include/linux/gpio/consumer.h +0 -0

Why? I already said that the best place is gpio/property.h as it's exactly for swnode related code.

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

Re: [PATCH v5 1/4] gpio: swnode: Add ability to specify native chip selects for SPI

Quoted 6 lines · Andy Shevchenko
Quoted 1 line · Charles Keepax

Here the second chip select is native. However, when using swnodes

Here, the

Can do.

Quoted 5 lines

drivers/gpio/Kconfig config GPIO_REGMAP · line 107

Quoted 2 lines
drivers/gpio/Kconfig +2 -0
+config GPIO_SWNODE_UNDEFINED
+       bool

But why did you remove the help? Please, put it back.

Sorry didn't realise you still wanted it will pop it back.

Quoted 6 lines

...

drivers/gpio/gpiolib-swnode.c static struct gpio_device *swnode_get_gpio_device(struct fwnode_handle *fwnode) · line 48

Quoted 2 lines
drivers/gpio/gpiolib-swnode.c +2 -0
+       /*
+        * Check for special node that identifies undefined GPIOs, this is

for a special

Can do.

Quoted 10 lines

drivers/gpio/gpiolib-swnode.c static struct gpio_device *swnode_get_gpio_device(struct fwnode_handle *fwnode) · line 52

Quoted 5 lines
drivers/gpio/gpiolib-swnode.c +4 -0
+        * primarily used as a key for internal chip selects in SPI bindings.
+        */
+       if (!strcmp(gdev_node->name, GPIOLIB_SWNODE_UNDEFINED_NAME))
+               return ERR_PTR(-ENOENT);

This is a dead code when the respective config option is not selected. Or actually a potential flaw if somebody else names their swnode like this.

Can add a check for the config.

Quoted 8 lines

...

drivers/gpio/gpiolib-swnode.c int swnode_gpio_count(const struct fwnode_handle *fwnode, const char *con_id) · line 152

Quoted 3 lines
drivers/gpio/gpiolib-swnode.c +3 -0
+       ret = software_node_register(&swnode_gpio_undefined);
+       if (ret < 0)
+               pr_err("gpiolib: failed to register swnode: %d\n", ret);

Instead of this prefix, define pr_fmt()

Little iffy on this, there are other prints in gpiolib that do it this way as well, I guess I could add a patch to convert everything but its starting to get a little out of the thrust of what I am doing here.

Quoted 10 lines
Quoted 1 line

+ return ret;

...

include/linux/gpio/consumer.h

Quoted 2 lines
include/linux/gpio/consumer.h +0 -0

Why? I already said that the best place is gpio/property.h as it's exactly for swnode related code.

Oh I see that's what you meant by those comments, as per my email I wasn't exactly sure. Will move them I don't mind where they go.

Thanks,
Charles

#4 lore raw

Re: [PATCH v5 1/4] gpio: swnode: Add ability to specify native chip selects for SPI

Quoted 3 lines · Charles Keepax
Quoted 2 lines · Andy Shevchenko

On Thu, Apr 11, 2024 at 12:06 PM Charles Keepax
<[email protected]> wrote:

...

Quoted 6 lines
Quoted 4 lines

drivers/gpio/Kconfig config GPIO_REGMAP · line 107

Quoted 2 lines
drivers/gpio/Kconfig +2 -0
+config GPIO_SWNODE_UNDEFINED
+       bool

But why did you remove the help? Please, put it back.

Sorry didn't realise you still wanted it will pop it back.

No, I don't want it to be user-selectable.

This is defined by a non-empty summary near to the type of the option (bool, tristate). The help text is orthogonal to this.

...

Quoted 8 lines
Quoted 6 lines

drivers/gpio/gpiolib-swnode.c static struct gpio_device *swnode_get_gpio_device(struct fwnode_handle *fwnode) · line 52

Quoted 2 lines
drivers/gpio/gpiolib-swnode.c +2 -0
+       if (!strcmp(gdev_node->name, GPIOLIB_SWNODE_UNDEFINED_NAME))
+               return ERR_PTR(-ENOENT);

This is a dead code when the respective config option is not selected. Or actually a potential flaw if somebody else names their swnode like this.

Can add a check for the config.

Maybe something like

    if (IS_ENABLED(...) &&
        !strcmp(...))

...

Quoted 10 lines
Quoted 5 lines

drivers/gpio/gpiolib-swnode.c int swnode_gpio_count(const struct fwnode_handle *fwnode, const char *con_id) · line 152

Quoted 3 lines
drivers/gpio/gpiolib-swnode.c +3 -0
+       ret = software_node_register(&swnode_gpio_undefined);
+       if (ret < 0)
+               pr_err("gpiolib: failed to register swnode: %d\n", ret);

Instead of this prefix, define pr_fmt()

Little iffy on this, there are other prints in gpiolib that do it this way as well, I guess I could add a patch to convert everything but its starting to get a little out of the thrust of what I am doing here.

That's why I'm talking only about this (gpiolib-swnode) module where you can have it as

"gpiolib: swnode: " fmt

or alike

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

Re: [PATCH v5 1/4] gpio: swnode: Add ability to specify native chip selects for SPI

Quoted 8 lines · Andy Shevchenko
Quoted 3 lines · Charles Keepax
Quoted 2 lines · Andy Shevchenko

On Thu, Apr 11, 2024 at 12:06 PM Charles Keepax
<[email protected]> wrote:

...

Quoted 9 lines
Quoted 6 lines
Quoted 4 lines

drivers/gpio/Kconfig config GPIO_REGMAP · line 107

Quoted 2 lines
drivers/gpio/Kconfig +2 -0
+config GPIO_SWNODE_UNDEFINED
+       bool

But why did you remove the help? Please, put it back.

Sorry didn't realise you still wanted it will pop it back.

No, I don't want it to be user-selectable.

Yeah I get that just didn't realise you also wanted the help, they are technically orthogonal but most non-user selectable things don't define a help.

Quoted 5 lines

Maybe something like

    if (IS_ENABLED(...) &&
        !strcmp(...))

Aye that is what I just added.

Quoted 11 lines
Quoted 10 lines
Quoted 5 lines

drivers/gpio/gpiolib-swnode.c int swnode_gpio_count(const struct fwnode_handle *fwnode, const char *con_id) · line 152

Quoted 3 lines
drivers/gpio/gpiolib-swnode.c +3 -0
+       ret = software_node_register(&swnode_gpio_undefined);
+       if (ret < 0)
+               pr_err("gpiolib: failed to register swnode: %d\n", ret);

Instead of this prefix, define pr_fmt()

Little iffy on this, there are other prints in gpiolib that do it this way as well, I guess I could add a patch to convert everything but its starting to get a little out of the thrust of what I am doing here.

Quoted 7 lines

That's why I'm talking only about this (gpiolib-swnode) module where you can have it as

"gpiolib: swnode: " fmt

or alike

Alright will add this too.

Thanks,
Charles

#6 lore raw

PATCH v5 2/4 spi: Switch to using is_acpi_device_node() in spi_dev_set_name()

In reply to Charles Keepax

Use the more modern is_acpi_device_node() rather than checking ACPI_COMPANION().

---

New since v4 of the series.

Thanks,
Charles

 drivers/spi/spi.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)
drivers/spi/spi.c +5 -3
index a2f01116ba09..05b33901eaa9 100644
@@ -12,6 +12,7 @@
 #include <linux/dmaengine.h>
 #include <linux/dma-mapping.h>
 #include <linux/export.h>
+#include <linux/fwnode.h>
 #include <linux/gpio/consumer.h>
 #include <linux/highmem.h>
 #include <linux/idr.h>
@@ -597,10 +598,11 @@ EXPORT_SYMBOL_GPL(spi_alloc_device);
 
 static void spi_dev_set_name(struct spi_device *spi)
 {
-	struct acpi_device *adev = ACPI_COMPANION(&spi->dev);
+	struct device *dev = &spi->dev;
+	struct fwnode_handle *fwnode = dev_fwnode(dev);
 
-	if (adev) {
-		dev_set_name(&spi->dev, "spi-%s", acpi_dev_name(adev));
+	if (is_acpi_device_node(fwnode)) {
+		dev_set_name(dev, "spi-%s", acpi_dev_name(to_acpi_device_node(fwnode)));
 		return;
 	}
 
Signature · 1 line
2.39.2
#7 lore raw

Re: [PATCH v5 2/4] spi: Switch to using is_acpi_device_node() in spi_dev_set_name()

Quoted 3 lines · Charles Keepax

Use the more modern is_acpi_device_node() rather than checking ACPI_COMPANION().

I don't think it's valuable on its own. There is no clear motivation why to do that, I suggested it exactly in the conjunction of not introducing two ways of fwnode type check. That said, you probably want to elaborate the motivation in the commit message if you want to keep it separate.

...

Quoted 1 line

+#include <linux/fwnode.h>

This header is not supposed to be included by the end users. property.h is.

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

Re: [PATCH v5 2/4] spi: Switch to using is_acpi_device_node() in spi_dev_set_name()

Quoted 12 lines · Andy Shevchenko
Quoted 3 lines · Charles Keepax

Use the more modern is_acpi_device_node() rather than checking ACPI_COMPANION().

I don't think it's valuable on its own. There is no clear motivation why to do that, I suggested it exactly in the conjunction of not introducing two ways of fwnode type check. That said, you probably want to elaborate the motivation in the commit message if you want to keep it separate.

I am really tempted to just drop this, its not necessary for my changes and changes something that is unrelated to them. At the least it belongs in a separate patch.

Quoted 6 lines

...

Quoted 1 line

+#include <linux/fwnode.h>

This header is not supposed to be included by the end users. property.h is.

Fair enough will update, although I really feel these headers could use some annotation if they are not supposed to be directly included. Either include everything you use or just include a top level header makes sense but this weird mixture we seem to use is very confusing and I don't have a big enough brain to remember every header.

Thanks,
Charles

#9 lore raw

Re: [PATCH v5 2/4] spi: Switch to using is_acpi_device_node() in spi_dev_set_name()

Quoted 13 lines · Charles Keepax
Quoted 11 lines · Andy Shevchenko
Quoted 3 lines · Charles Keepax

Use the more modern is_acpi_device_node() rather than checking ACPI_COMPANION().

I don't think it's valuable on its own. There is no clear motivation why to do that, I suggested it exactly in the conjunction of not introducing two ways of fwnode type check. That said, you probably want to elaborate the motivation in the commit message if you want to keep it separate.

Quoted 3 lines

I am really tempted to just drop this, its not necessary for my changes and changes something that is unrelated to them. At the least it belongs in a separate patch.

Okay, just elaborate in the commit message that this is done to make new comer not to invent its own way for fwnode type check,

...

Quoted 10 lines
Quoted 3 lines
Quoted 1 line

+#include <linux/fwnode.h>

This header is not supposed to be included by the end users. property.h is.

Fair enough will update, although I really feel these headers could use some annotation if they are not supposed to be directly included. Either include everything you use or just include a top level header makes sense but this weird mixture we seem to use is very confusing and I don't have a big enough brain to remember every header.

Rough hint is to look into the contents. But I agree, that is a complete mess.

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

PATCH v5 3/4 spi: Update swnode based SPI devices to use the fwnode name

In reply to Charles Keepax

Update the name for software node based SPI devices to use the fwnode name as the device name. This is helpful since swnode devices are usually added within the kernel, and the kernel often then requires a predictable name such that it can refer back to the device.

---

No changes since v4.

Thanks,
Charles

 drivers/spi/spi.c | 5 +++++
 1 file changed, 5 insertions(+)
drivers/spi/spi.c +5 -0
index 05b33901eaa9..bcc9d3ab7cd9 100644
@@ -606,6 +606,11 @@ static void spi_dev_set_name(struct spi_device *spi)
 		return;
 	}
 
+	if (is_software_node(fwnode)) {
+		dev_set_name(dev, "spi-%s", fwnode_get_name(fwnode));
+		return;
+	}
+
 	dev_set_name(&spi->dev, "%s.%u", dev_name(&spi->controller->dev),
 		     spi_get_chipselect(spi, 0));
 }
Signature · 1 line
2.39.2
#11 lore raw

Re: [PATCH v5 3/4] spi: Update swnode based SPI devices to use the fwnode name

Quoted 5 lines · Charles Keepax

Update the name for software node based SPI devices to use the fwnode name as the device name. This is helpful since swnode devices are usually added within the kernel, and the kernel often then requires a predictable name such that it can refer back to the device.

...

drivers/spi/spi.c static void spi_dev_set_name(struct spi_device *spi) · line 610

Quoted 2 lines
drivers/spi/spi.c +2 -0
+       if (is_software_node(fwnode)) {
+               dev_set_name(dev, "spi-%s", fwnode_get_name(fwnode));

Wouldn't %pfwP / %pfw work?

Thinking more about this, maybe even the ACPI case also can be combined? See for the details

87526603c892 ("irqdomain: Get rid of special treatment for ACPI in __irq_domain_add()")
9ed78b05f998 ("irqdomain: Allow software nodes for IRQ domain creation")

Quoted 2 lines
drivers/spi/spi.c +2 -0
+               return;
+       }
Signature · 2 lines
With Best Regards,
Andy Shevchenko
#12 lore raw

Re: [PATCH v5 3/4] spi: Update swnode based SPI devices to use the fwnode name

Quoted 8 lines · Andy Shevchenko
Quoted 5 lines · Charles Keepax

Update the name for software node based SPI devices to use the fwnode name as the device name. This is helpful since swnode devices are usually added within the kernel, and the kernel often then requires a predictable name such that it can refer back to the device.

Quoted 7 lines

...

drivers/spi/spi.c static void spi_dev_set_name(struct spi_device *spi) · line 610

Quoted 2 lines
drivers/spi/spi.c +2 -0
+       if (is_software_node(fwnode)) {
+               dev_set_name(dev, "spi-%s", fwnode_get_name(fwnode));

Wouldn't %pfwP / %pfw work?

Oh I was totally unaware of those will have a look, thanks.

Quoted 7 lines

Thinking more about this, maybe even the ACPI case also can be combined? See for the details

87526603c892 ("irqdomain: Get rid of special treatment for ACPI in __irq_domain_add()")
9ed78b05f998 ("irqdomain: Allow software nodes for IRQ domain creation")

Maybe, I am not super confident that acpi_dev_name will always return the same thing as fwnode_get_name and I don't really want to risk renaming everything on the ACPI side.

Thanks,
Charles

#13 lore raw

Re: [PATCH v5 3/4] spi: Update swnode based SPI devices to use the fwnode name

Quoted 3 lines · Charles Keepax
Quoted 2 lines · Andy Shevchenko

On Thu, Apr 11, 2024 at 12:06 PM Charles Keepax
<[email protected]> wrote:

...

Quoted 10 lines
Quoted 6 lines

Thinking more about this, maybe even the ACPI case also can be combined? See for the details

87526603c892 ("irqdomain: Get rid of special treatment for ACPI in __irq_domain_add()")
9ed78b05f998 ("irqdomain: Allow software nodes for IRQ domain creation")

Maybe, I am not super confident that acpi_dev_name will always return the same thing as fwnode_get_name and I don't really want to risk renaming everything on the ACPI side.

Scratch this. They are completely different.

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

PATCH v5 4/4 spi: cs42l43: Add bridged cs35l56 amplifiers

In reply to Charles Keepax

From: Maciej Strozek <[email protected]>

On some cs42l43 systems a couple of cs35l56 amplifiers are attached to the cs42l43's SPI and I2S. On Windows the cs42l43 is controlled by a SDCA class driver and these two amplifiers are controlled by firmware running on the cs42l43. However, under Linux the decision was made to interact with the cs42l43 directly, affording the user greater control over the audio system. However, this has resulted in an issue where these two bridged cs35l56 amplifiers are not populated in ACPI and must be added manually.

Check for the presence of the "01fa-cirrus-sidecar-instances" property in the SDCA extension unit's ACPI properties to confirm the presence of these two amplifiers and if they exist add them manually onto the SPI bus.

---

Changes since v4:
- Drop some redundant fwnode_handle_puts
- Remove stray blank line
- Use ACPI_HANDLE_FWNODE

Thanks,
Charles

 drivers/spi/Kconfig       |   1 +
 drivers/spi/spi-cs42l43.c | 135 +++++++++++++++++++++++++++++++++++++-
 2 files changed, 134 insertions(+), 2 deletions(-)
drivers/spi/Kconfig +1 -0
index 554664efda86..17325e0b7bd5 100644
@@ -284,6 +284,7 @@ config SPI_COLDFIRE_QSPI
 config SPI_CS42L43
 	tristate "Cirrus Logic CS42L43 SPI controller"
 	depends on MFD_CS42L43 && PINCTRL_CS42L43
+	select GPIO_SWNODE_UNDEFINED
 	help
 	  This enables support for the SPI controller inside the Cirrus Logic
 	  CS42L43 audio codec.
drivers/spi/spi-cs42l43.c +133 -2

192 lines not on this page · raw message

Signature · 1 line
2.39.2
#15 lore raw

Re: [PATCH v5 4/4] spi: cs42l43: Add bridged cs35l56 amplifiers

Quoted 15 lines · Charles Keepax

From: Maciej Strozek <[email protected]>

On some cs42l43 systems a couple of cs35l56 amplifiers are attached to the cs42l43's SPI and I2S. On Windows the cs42l43 is controlled by a SDCA class driver and these two amplifiers are controlled by firmware running on the cs42l43. However, under Linux the decision was made to interact with the cs42l43 directly, affording the user greater control over the audio system. However, this has resulted in an issue where these two bridged cs35l56 amplifiers are not populated in ACPI and must be added manually.

Check for the presence of the "01fa-cirrus-sidecar-instances" property in the SDCA extension unit's ACPI properties to confirm the presence of these two amplifiers and if they exist add them manually onto the SPI bus.

...

drivers/spi/spi-cs42l43.c static const unsigned int cs42l43_clock_divs[] = { · line 53

Quoted 7 lines
drivers/spi/spi-cs42l43.c +7 -0
+static const struct software_node ampl = {
+       .name                   = "cs35l56-left",
+};
+
+static const struct software_node ampr = {
+       .name                   = "cs35l56-right",
+};

Still I fail to see how these are used. They provide just a static swnode with name and no properties. How is that different from the no-fwnode case? Can you test without these being defined?

...

drivers/spi/spi-cs42l43.c static const unsigned int cs42l43_clock_divs[] = { · line 73

Quoted 3 lines
drivers/spi/spi-cs42l43.c +3 -0
+static const struct software_node cs42l43_gpiochip_swnode = {
+       .name                   = "cs42l43-pinctrl",
+};

On the contrary I understand this one (however, using that generic name prevents more than one or two drivers from being instantiated).

...

Quoted 1 line

+ SOFTWARE_NODE_REFERENCE(&swnode_gpio_undefined),

gpio/property.h for this one.

...

drivers/spi/spi-cs42l43.c static size_t cs42l43_spi_max_length(struct spi_device *spi) · line 261

Quoted 13 lines
drivers/spi/spi-cs42l43.c +12 -0
+static bool cs42l43_has_sidecar(struct fwnode_handle *fwnode)
+{
+       static const u32 func_smart_amp = 0x1;
+       struct fwnode_handle *child_fwnode, *ext_fwnode;
+       unsigned int val;
+       u32 function;
+       int ret;
+
+       fwnode_for_each_child_node(fwnode, child_fwnode) {
+               acpi_handle handle = ACPI_HANDLE_FWNODE(child_fwnode);
 
+               if (!handle)
+                       continue;

This is _almost_ redundant check. handle == NULL is for the global root object which quite unlikely will have the _ADR method. The specification is clear about this: "The _ADR object is valid only within an Augmented Device Descriptor." That said, the check makes sense against the (very) ill-formed DSDT.

Quoted 22 lines
drivers/spi/spi-cs42l43.c +22 -0
+               ret = acpi_get_local_address(handle, &function);
+               if (ret || function != func_smart_amp)
+                       continue;
+
+               ext_fwnode = fwnode_get_named_child_node(child_fwnode,
+                               "mipi-sdca-function-expansion-subproperties");
+               if (!ext_fwnode)
+                       continue;
+
+               ret = fwnode_property_read_u32(ext_fwnode,
+                                              "01fa-cirrus-sidecar-instances",
+                                              &val);
+
+               fwnode_handle_put(ext_fwnode);
+               fwnode_handle_put(child_fwnode);
+
+               if (!ret)
+                       return !!val;
+       }
+
+       return false;
+}

...

drivers/spi/spi-cs42l43.c static int cs42l43_spi_probe(struct platform_device *pdev) · line 362

Quoted 13 lines
drivers/spi/spi-cs42l43.c +13 -0
+       if (has_sidecar) {
+               ret = software_node_register(&cs42l43_gpiochip_swnode);
+               if (ret) {
+                       return dev_err_probe(priv->dev, ret,
+                                            "Failed to register gpio swnode\n");
+               }
+
+               ret = device_create_managed_software_node(&priv->ctlr->dev,
+                                                         cs42l43_cs_props, NULL);
+               if (ret) {
+                       dev_err_probe(priv->dev, ret, "Failed to add swnode\n");
+                       goto err;
+               }

Wouldn't it miss the parent fwnode? I mean that you might probably need to call...

drivers/spi/spi-cs42l43.c static int cs42l43_spi_probe(struct platform_device *pdev) · line 364

Quoted 2 lines
drivers/spi/spi-cs42l43.c +2 -0
+       } else {
+               device_set_node(&priv->ctlr->dev, fwnode);

...this one always. Have you checked it? How does sysfs look like before and after this change on the device in question?

Quoted 1 line

+ }

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

Re: [PATCH v5 4/4] spi: cs42l43: Add bridged cs35l56 amplifiers

Quoted 2 lines · Andy Shevchenko

On Thu, Apr 11, 2024 at 12:06 PM Charles Keepax
<[email protected]> wrote:

...

Quoted 28 lines

drivers/spi/spi-cs42l43.c static size_t cs42l43_spi_max_length(struct spi_device *spi) · line 261

Quoted 5 lines
drivers/spi/spi-cs42l43.c +4 -0
+       fwnode_for_each_child_node(fwnode, child_fwnode) {
+               acpi_handle handle = ACPI_HANDLE_FWNODE(child_fwnode);
 
+               if (!handle)
+                       continue;

This is _almost_ redundant check. handle == NULL is for the global root object which quite unlikely will have the _ADR method. The specification is clear about this: "The _ADR object is valid only within an Augmented Device Descriptor." That said, the check makes sense against the (very) ill-formed DSDT.

Quoted 16 lines
drivers/spi/spi-cs42l43.c +15 -0
+               ret = acpi_get_local_address(handle, &function);
+               if (ret || function != func_smart_amp)
+                       continue;
+
+               ext_fwnode = fwnode_get_named_child_node(child_fwnode,
+                               "mipi-sdca-function-expansion-subproperties");
+               if (!ext_fwnode)
+                       continue;
+
+               ret = fwnode_property_read_u32(ext_fwnode,
+                                              "01fa-cirrus-sidecar-instances",
+                                              &val);
+
+               fwnode_handle_put(ext_fwnode);
 
+               fwnode_handle_put(child_fwnode);

And still this leftover...

Quoted 2 lines
Quoted 2 lines
drivers/spi/spi-cs42l43.c +2 -0
+               if (!ret)
+                       return !!val;
Signature · 2 lines
With Best Regards,
Andy Shevchenko
#17 lore raw

Re: [PATCH v5 4/4] spi: cs42l43: Add bridged cs35l56 amplifiers

Quoted 4 lines · Andy Shevchenko
Quoted 2 lines · Andy Shevchenko

On Thu, Apr 11, 2024 at 12:06 PM Charles Keepax
<[email protected]> wrote:

...

Quoted 6 lines
Quoted 1 line
Quoted 1 line

+ fwnode_handle_put(child_fwnode);

And still this leftover...

Quoted 2 lines
Quoted 2 lines
drivers/spi/spi-cs42l43.c +2 -0
+               if (!ret)
+                       return !!val;

To be precise it has to be refactored like

if (ret)

    continue;

put()
return ...;

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

Re: [PATCH v5 4/4] spi: cs42l43: Add bridged cs35l56 amplifiers

In reply to Andy Shevchenko
Quoted 20 lines · Andy Shevchenko
Quoted 15 lines · Charles Keepax

From: Maciej Strozek <[email protected]>

On some cs42l43 systems a couple of cs35l56 amplifiers are attached to the cs42l43's SPI and I2S. On Windows the cs42l43 is controlled by a SDCA class driver and these two amplifiers are controlled by firmware running on the cs42l43. However, under Linux the decision was made to interact with the cs42l43 directly, affording the user greater control over the audio system. However, this has resulted in an issue where these two bridged cs35l56 amplifiers are not populated in ACPI and must be added manually.

Check for the presence of the "01fa-cirrus-sidecar-instances" property in the SDCA extension unit's ACPI properties to confirm the presence of these two amplifiers and if they exist add them manually onto the SPI bus.

...

Quoted 12 lines

drivers/spi/spi-cs42l43.c static const unsigned int cs42l43_clock_divs[] = { · line 53

Quoted 7 lines
drivers/spi/spi-cs42l43.c +7 -0
+static const struct software_node ampl = {
+       .name                   = "cs35l56-left",
+};
+
+static const struct software_node ampr = {
+       .name                   = "cs35l56-right",
+};

Still I fail to see how these are used. They provide just a static swnode with name and no properties. How is that different from the no-fwnode case? Can you test without these being defined?

The code in the last patch will pick up the name and use it to name the amps that are registered. This means when those amps are referred to by the audio machine driver code we will know what they are called. Admittedly that audio machine driver change isn't in this series as it is a bit of a work in progress and not technically necessary for just registering the amps as this series does.

Quoted 9 lines

...

drivers/spi/spi-cs42l43.c static const unsigned int cs42l43_clock_divs[] = { · line 73

Quoted 3 lines
drivers/spi/spi-cs42l43.c +3 -0
+static const struct software_node cs42l43_gpiochip_swnode = {
+       .name                   = "cs42l43-pinctrl",
+};

On the contrary I understand this one (however, using that generic name prevents more than one or two drivers from being instantiated).

Yeah that might need to change in the future, however there is no obvious use-cases for using multiple cs42l43's in a single system so at the moment we are not doing the work to support that case. There are plenty other things that would need fixed to support that as well.

Quoted 5 lines

...

Quoted 1 line

+ SOFTWARE_NODE_REFERENCE(&swnode_gpio_undefined),

gpio/property.h for this one.

Sorry, still not quite following this one, are you just reminding me to include the header when I move the swnode_gpio_undefined definition or are you asking for something more?

Quoted 11 lines
Quoted 10 lines
drivers/spi/spi-cs42l43.c +10 -0
+static bool cs42l43_has_sidecar(struct fwnode_handle *fwnode)
+{
+       static const u32 func_smart_amp = 0x1;
+       struct fwnode_handle *child_fwnode, *ext_fwnode;
+       unsigned int val;
+       u32 function;
+       int ret;
+
+       fwnode_for_each_child_node(fwnode, child_fwnode) {
+               acpi_handle handle = ACPI_HANDLE_FWNODE(child_fwnode);
Quoted 9 lines

Review

Quoted 2 lines
(quoted diff) +2 -0
+               if (!handle)
+                       continue;

This is _almost_ redundant check. handle == NULL is for the global root object which quite unlikely will have the _ADR method. The specification is clear about this: "The _ADR object is valid only within an Augmented Device Descriptor." That said, the check makes sense against the (very) ill-formed DSDT.

I am willing to be guided here, but given it would result in a null pointer dereference I am inclined to keep the check personally.

Quoted 14 lines

drivers/spi/spi-cs42l43.c static int cs42l43_spi_probe(struct platform_device *pdev) · line 362

Quoted 13 lines
drivers/spi/spi-cs42l43.c +13 -0
+       if (has_sidecar) {
+               ret = software_node_register(&cs42l43_gpiochip_swnode);
+               if (ret) {
+                       return dev_err_probe(priv->dev, ret,
+                                            "Failed to register gpio swnode\n");
+               }
+
+               ret = device_create_managed_software_node(&priv->ctlr->dev,
+                                                         cs42l43_cs_props, NULL);
+               if (ret) {
+                       dev_err_probe(priv->dev, ret, "Failed to add swnode\n");
+                       goto err;
+               }
Quoted 8 lines

Wouldn't it miss the parent fwnode? I mean that you might probably need to call...

drivers/spi/spi-cs42l43.c static int cs42l43_spi_probe(struct platform_device *pdev) · line 364

Quoted 2 lines
drivers/spi/spi-cs42l43.c +2 -0
+       } else {
+               device_set_node(&priv->ctlr->dev, fwnode);

...this one always. Have you checked it? How does sysfs look like before and after this change on the device in question?

I will check this.

Thanks,
Charles

#19 lore raw

Re: [PATCH v5 4/4] spi: cs42l43: Add bridged cs35l56 amplifiers

Quoted 3 lines · Charles Keepax
Quoted 2 lines · Andy Shevchenko

On Thu, Apr 11, 2024 at 12:06 PM Charles Keepax
<[email protected]> wrote:

...

Quoted 12 lines
Quoted 11 lines

drivers/spi/spi-cs42l43.c static const unsigned int cs42l43_clock_divs[] = { · line 53

Quoted 7 lines
drivers/spi/spi-cs42l43.c +7 -0
+static const struct software_node ampl = {
+       .name                   = "cs35l56-left",
+};
+
+static const struct software_node ampr = {
+       .name                   = "cs35l56-right",
+};

Still I fail to see how these are used. They provide just a static swnode with name and no properties. How is that different from the no-fwnode case? Can you test without these being defined?

Quoted 7 lines

The code in the last patch will pick up the name and use it to name the amps that are registered. This means when those amps are referred to by the audio machine driver code we will know what they are called. Admittedly that audio machine driver change isn't in this series as it is a bit of a work in progress and not technically necessary for just registering the amps as this series does.

Thank you for the patience, now I realise how it's connected. But wouldn't the simple spi-<SPI ID>-<bus number>.<chip select> work? Thinking loudly: Probably not as bus number is dynamic nowadays,

...

Quoted 7 lines
Quoted 6 lines

drivers/spi/spi-cs42l43.c static const unsigned int cs42l43_clock_divs[] = { · line 73

Quoted 3 lines
drivers/spi/spi-cs42l43.c +3 -0
+static const struct software_node cs42l43_gpiochip_swnode = {
+       .name                   = "cs42l43-pinctrl",
+};

On the contrary I understand this one (however, using that generic name prevents more than one or two drivers from being instantiated).

Quoted 5 lines

Yeah that might need to change in the future, however there is no obvious use-cases for using multiple cs42l43's in a single system so at the moment we are not doing the work to support that case. There are plenty other things that would need fixed to support that as well.

I see, just be aware that names for "root" swnodes must be globally unique. Otherwise they will clash over sysfs folder namings.

...

Quoted 7 lines
Quoted 3 lines
Quoted 1 line

+ SOFTWARE_NODE_REFERENCE(&swnode_gpio_undefined),

gpio/property.h for this one.

Sorry, still not quite following this one, are you just reminding me to include the header when I move the swnode_gpio_undefined definition or are you asking for something more?

Yes, when you have moved the newly added exported variable there, include itt in addition to gpio/consumer.h.

...

Quoted 11 lines
Quoted 10 lines

drivers/spi/spi-cs42l43.c static size_t cs42l43_spi_max_length(struct spi_device *spi) · line 261

Quoted 4 lines
drivers/spi/spi-cs42l43.c +3 -0
+               acpi_handle handle = ACPI_HANDLE_FWNODE(child_fwnode);
 
+               if (!handle)
+                       continue;

This is _almost_ redundant check. handle == NULL is for the global root object which quite unlikely will have the _ADR method. The specification is clear about this: "The _ADR object is valid only within an Augmented Device Descriptor." That said, the check makes sense against the (very) ill-formed DSDT.

Quoted 3 lines

I am willing to be guided here, but given it would result in a null pointer dereference I am inclined to keep the check personally.

There is no NULL pointer dereference. That's the point. And I explained how ACPICA treats this.

...

Quoted 17 lines
Quoted 17 lines

drivers/spi/spi-cs42l43.c static int cs42l43_spi_probe(struct platform_device *pdev) · line 362

Quoted 13 lines
drivers/spi/spi-cs42l43.c +13 -0
+       if (has_sidecar) {
+               ret = software_node_register(&cs42l43_gpiochip_swnode);
+               if (ret) {
+                       return dev_err_probe(priv->dev, ret,
+                                            "Failed to register gpio swnode\n");
+               }
+
+               ret = device_create_managed_software_node(&priv->ctlr->dev,
+                                                         cs42l43_cs_props, NULL);
+               if (ret) {
+                       dev_err_probe(priv->dev, ret, "Failed to add swnode\n");
+                       goto err;
+               }

Wouldn't it miss the parent fwnode? I mean that you might probably need to call...

Quoted 7 lines
Quoted 5 lines

drivers/spi/spi-cs42l43.c static int cs42l43_spi_probe(struct platform_device *pdev) · line 364

Quoted 2 lines
drivers/spi/spi-cs42l43.c +2 -0
+       } else {
+               device_set_node(&priv->ctlr->dev, fwnode);

...this one always. Have you checked it? How does sysfs look like before and after this change on the device in question?

I will check this.

Basically in the expected case there should be two symlinks: to physical node and to swnode.

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

Re: [PATCH v5 4/4] spi: cs42l43: Add bridged cs35l56 amplifiers

Quoted 19 lines · Andy Shevchenko
Quoted 17 lines · Charles Keepax
Quoted 16 lines · Andy Shevchenko
Quoted 13 lines · Charles Keepax
drivers/spi/spi-cs42l43.c +13 -0
+       if (has_sidecar) {
+               ret = software_node_register(&cs42l43_gpiochip_swnode);
+               if (ret) {
+                       return dev_err_probe(priv->dev, ret,
+                                            "Failed to register gpio swnode\n");
+               }
+
+               ret = device_create_managed_software_node(&priv->ctlr->dev,
+                                                         cs42l43_cs_props, NULL);
+               if (ret) {
+                       dev_err_probe(priv->dev, ret, "Failed to add swnode\n");
+                       goto err;
+               }
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

Wouldn't it miss the parent fwnode? I mean that you might probably need to call...

Ok I am pretty sure this is all fine, I don't think we can pass a parent into device_create_managed_software_node since it requires a parent software node, but in this case there isn't one. This is the root node here, since the "parent" would be ACPI stuff here.

Quoted 8 lines
Quoted 7 lines
Quoted 5 lines

drivers/spi/spi-cs42l43.c static int cs42l43_spi_probe(struct platform_device *pdev) · line 364

Quoted 2 lines
drivers/spi/spi-cs42l43.c +2 -0
+       } else {
+               device_set_node(&priv->ctlr->dev, fwnode);

...this one always. Have you checked it? How does sysfs look like before and after this change on the device in question?

I will check this.

We can't always call device_set_node. Firstly, we would need to set it to the software node, however that is never returned from device_create_managed_software_node. Secondly, the set_secondary_node called in device_create_managed_software_node will set the primary node anyway since there isn't a valid primary node on the device. Finally, we don't want the primary node set to the ACPI node anyway since we want to override those settings here with our bridged amp settings.

Quoted 3 lines

Basically in the expected case there should be two symlinks: to physical node and to swnode.

I think the sysfs all looks reasonable to me, I can see the SPI devices in /sys/bus/spi/devices, under those devices I can see a symlink to the software node.

Thanks,
Charles

#21 lore raw

Re: [PATCH v5 4/4] spi: cs42l43: Add bridged cs35l56 amplifiers

Quoted 6 lines · Charles Keepax
Quoted 5 lines · Andy Shevchenko
Quoted 3 lines · Charles Keepax
Quoted 2 lines · Andy Shevchenko

On Thu, Apr 11, 2024 at 12:06 PM Charles Keepax
<[email protected]> wrote:

...

Quoted 13 lines
Quoted 13 lines
Quoted 13 lines
Quoted 13 lines
Quoted 12 lines
drivers/spi/spi-cs42l43.c +12 -0
+               ret = software_node_register(&cs42l43_gpiochip_swnode);
+               if (ret) {
+                       return dev_err_probe(priv->dev, ret,
+                                            "Failed to register gpio swnode\n");
+               }
+
+               ret = device_create_managed_software_node(&priv->ctlr->dev,
+                                                         cs42l43_cs_props, NULL);
+               if (ret) {
+                       dev_err_probe(priv->dev, ret, "Failed to add swnode\n");
+                       goto err;
+               }
Quoted 4 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

Wouldn't it miss the parent fwnode? I mean that you might probably need to call...

Ok I am pretty sure this is all fine,

But have you checked this?

Quoted 4 lines

I don't think we can pass a
parent into device_create_managed_software_node since it requires a parent software node, but in this case there isn't one. This is the root node here, since the "parent" would be ACPI stuff here.

No, this is done implicitly by so called primary and secondary fwnode. If you have no fwnode is added to the device (via let's say device_set_node() call), it most likely has no "primary" fwnode which is usually points to the "physical" one (from ACPI or DT), while secondary one will be pointing to swnode:

Ex.

# ls -ld /sys/devices/pci0000:00/0000:00:11.0/dwc3.0.auto/*_node
lrwxrwxrwx    1 root     root             0 Jan  1 00:01
/sys/devices/pci0000:00/0000:00:11.0/dwc3.0.auto/firmwar
e_node -> ../../../LNXSYSTM:00/LNXSYBUS:00/PNP0A08:00/device:08
lrwxrwxrwx    1 root     root             0 Jan  1 00:01
/sys/devices/pci0000:00/0000:00:11.0/dwc3.0.auto/softwar
e_node -> ../../../../kernel/software_nodes/node0
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 5 lines

drivers/spi/spi-cs42l43.c static int cs42l43_spi_probe(struct platform_device *pdev) · line 364

Quoted 2 lines
drivers/spi/spi-cs42l43.c +2 -0
+       } else {
+               device_set_node(&priv->ctlr->dev, fwnode);

...this one always. Have you checked it? How does sysfs look like before and after this change on the device in question?

Quoted 7 lines
Quoted 1 line
Quoted 1 line

I will check this.

We can't always call device_set_node. Firstly, we would need to set it to the software node, however that is never returned from device_create_managed_software_node. Secondly, the set_secondary_node called in device_create_managed_software_node will set the primary node anyway since there isn't a valid primary node on the device.

That was basically my question above. If the device has a primary fwnode (or one shared with a parent) it would be nice to propagate it. OTOH it might have a side effect of using properties from that in the code.

Quoted 10 lines

Finally, we don't want the primary node set to the ACPI node anyway since we want to override those settings here with our bridged amp settings.

Quoted 2 lines

Basically in the expected case there should be two symlinks: to physical node and to swnode.

I think the sysfs all looks reasonable to me, I can see the SPI devices in /sys/bus/spi/devices, under those devices I can see a symlink to the software node.

OK.

--
With Best Regards,
Andy Shevchenko

#22 lore raw
In reply to Charles Keepax
Quoted 18 lines · Charles Keepax

On some cs42l43 systems a couple of cs35l56 amplifiers are attached to the cs42l43's SPI and I2S. On Windows the cs42l43 is controlled by a SDCA class driver and these two amplifiers are controlled by firmware running on the cs42l43. However, under Linux the decision was made to interact with the cs42l43 directly, affording the user greater control over the audio system. However, this has resulted in an issue where these two bridged cs35l56 amplifiers are not populated in ACPI and must be added manually. There is at least an SDCA extension unit DT entry we can key off.

The process of adding this is handled using a software node, firstly the ability to add native chip selects to software nodes must be added. Secondly, an additional flag for naming the SPI devices is added this allows the machine driver to key to the correct amplifier. Then finally, the cs42l43 SPI driver adds the two amplifiers directly onto its SPI bus.

Quoted 3 lines

An additional series will follow soon to add the audio machine driver parts (in the sof-sdw driver), however that is fairly orthogonal to this part of the process, getting the actual amplifiers registered.

Quick note: I dunno how you prepared your series, but the cover letter missed the versioning. git format-patch -v<N> ... does it correctly, where <X> is the version number. You may also see how I do with a script [1].

[1]: https://github.com/andy-shev/home-bin-tools/blob/master/ge2maintainer.sh

Signature · 2 lines
With Best Regards,
Andy Shevchenko