[PATCH v3 0/5] platform/x86: intel_scu_ipc: Avoid working around IO and cleanups

[PATCH v3 0/5] platform/x86: intel_scu_ipc: Avoid working around IO and cleanups

[Andy Shevchenko]

The first patch avoids using a workaround for IO (which seems unneeded).
The rest is a batch of cleanups. Has been tested on Intel Merrifield
(thanks, Ferry).

v3:
- split patches 1 & 2 to two patches each (Ilpo)
- preserve a comment for reading buffer (Ilpo)
- inherited tags for new patches as the result is the same code wise

v2:
- dropped stray change (Mika)
- elaborated conditional changes in the commit message (Mika)
- gathered tags (Mika, Ferry)

Andy Shevchenko (5):
  platform/x86: intel_scu_ipc: Unify the flow in pwr_reg_rdwr()
  platform/x86: intel_scu_ipc: Replace workaround by 32-bit IO
  platform/x86: intel_scu_ipc: Simplify code with cleanup helpers
  platform/x86: intel_scu_ipc: Convert to check for errors first
  platform/x86: intel_scu_ipc: Save a copy of the entire struct
    intel_scu_ipc_data

 drivers/platform/x86/intel_scu_ipc.c | 142 ++++++++++++---------------
 1 file changed, 61 insertions(+), 81 deletions(-)

-- 
2.43.0.rc1.1336.g36b5255a03ac

[PATCH v3 1/5] platform/x86: intel_scu_ipc: Unify the flow in pwr_reg_rdwr()

[Andy Shevchenko]

Unify the flow in pwr_reg_rdwr() with intel_scu_ipc_dev_command_with_size().
While read buffer is not currently being used by the callers of
pwr_reg_rdwr(), this might change in the future. For the existing code
it's a harmless change because the read data for the write operation
is simply discarded.

Tested-by: Ferry Toth <fntoth@gmail.com>
Acked-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
---
 drivers/platform/x86/intel_scu_ipc.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/platform/x86/intel_scu_ipc.c b/drivers/platform/x86/intel_scu_ipc.c
index 5b16d29c93d7..8d11d18701a4 100644
--- a/drivers/platform/x86/intel_scu_ipc.c
+++ b/drivers/platform/x86/intel_scu_ipc.c
@@ -325,7 +325,7 @@ static int pwr_reg_rdwr(struct intel_scu_ipc_dev *scu, u16 *addr, u8 *data,
 	}
 
 	err = intel_scu_ipc_check_status(scu);
-	if (!err && id == IPC_CMD_PCNTRL_R) { /* Read rbuf */
+	if (!err) { /* Read rbuf */
 		/* Workaround: values are read as 0 without memcpy_fromio */
 		memcpy_fromio(cbuf, scu->ipc_base + 0x90, 16);
 		for (nc = 0; nc < count; nc++)
-- 
2.43.0.rc1.1336.g36b5255a03ac

[PATCH v3 2/5] platform/x86: intel_scu_ipc: Replace workaround by 32-bit IO

[Andy Shevchenko]

The theory is that the so called workaround in pwr_reg_rdwr() is
the actual reader of the data in 32-bit chunks. For some reason
the 8-bit IO won't fail after that. Replace the workaround by using
32-bit IO explicitly and then memcpy() as much data as was requested
by the user. The same approach is already in use in
intel_scu_ipc_dev_command_with_size().

Tested-by: Ferry Toth <fntoth@gmail.com>
Acked-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
---
 drivers/platform/x86/intel_scu_ipc.c | 13 +++----------
 1 file changed, 3 insertions(+), 10 deletions(-)

diff --git a/drivers/platform/x86/intel_scu_ipc.c b/drivers/platform/x86/intel_scu_ipc.c
index 8d11d18701a4..290b38627542 100644
--- a/drivers/platform/x86/intel_scu_ipc.c
+++ b/drivers/platform/x86/intel_scu_ipc.c
@@ -217,12 +217,6 @@ static inline u8 ipc_read_status(struct intel_scu_ipc_dev *scu)
 	return __raw_readl(scu->ipc_base + IPC_STATUS);
 }
 
-/* Read ipc byte data */
-static inline u8 ipc_data_readb(struct intel_scu_ipc_dev *scu, u32 offset)
-{
-	return readb(scu->ipc_base + IPC_READ_BUFFER + offset);
-}
-
 /* Read ipc u32 data */
 static inline u32 ipc_data_readl(struct intel_scu_ipc_dev *scu, u32 offset)
 {
@@ -326,10 +320,9 @@ static int pwr_reg_rdwr(struct intel_scu_ipc_dev *scu, u16 *addr, u8 *data,
 
 	err = intel_scu_ipc_check_status(scu);
 	if (!err) { /* Read rbuf */
-		/* Workaround: values are read as 0 without memcpy_fromio */
-		memcpy_fromio(cbuf, scu->ipc_base + 0x90, 16);
-		for (nc = 0; nc < count; nc++)
-			data[nc] = ipc_data_readb(scu, nc);
+		for (nc = 0, offset = 0; nc < 4; nc++, offset += 4)
+			wbuf[nc] = ipc_data_readl(scu, offset);
+		memcpy(data, wbuf, count);
 	}
 	mutex_unlock(&ipclock);
 	return err;
-- 
2.43.0.rc1.1336.g36b5255a03ac

[PATCH v3 3/5] platform/x86: intel_scu_ipc: Simplify code with cleanup helpers

[Andy Shevchenko]

Use macros defined in linux/cleanup.h to automate resource lifetime
control in the driver. The byproduct of this change is that the
negative conditionals are swapped by positive ones that are being
attached to the respective calls, hence the code uses the regular
pattern, i.e. checking for the error first.

Tested-by: Ferry Toth <fntoth@gmail.com>
Acked-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
---
 drivers/platform/x86/intel_scu_ipc.c | 65 +++++++++++-----------------
 1 file changed, 25 insertions(+), 40 deletions(-)

diff --git a/drivers/platform/x86/intel_scu_ipc.c b/drivers/platform/x86/intel_scu_ipc.c
index 290b38627542..9829d4145c58 100644
--- a/drivers/platform/x86/intel_scu_ipc.c
+++ b/drivers/platform/x86/intel_scu_ipc.c
@@ -13,6 +13,7 @@
  * along with other APIs.
  */
 
+#include <linux/cleanup.h>
 #include <linux/delay.h>
 #include <linux/device.h>
 #include <linux/errno.h>
@@ -99,23 +100,21 @@ static struct class intel_scu_ipc_class = {
  */
 struct intel_scu_ipc_dev *intel_scu_ipc_dev_get(void)
 {
-	struct intel_scu_ipc_dev *scu = NULL;
+	guard(mutex)(&ipclock);
 
-	mutex_lock(&ipclock);
 	if (ipcdev) {
 		get_device(&ipcdev->dev);
 		/*
 		 * Prevent the IPC provider from being unloaded while it
 		 * is being used.
 		 */
-		if (!try_module_get(ipcdev->owner))
-			put_device(&ipcdev->dev);
-		else
-			scu = ipcdev;
+		if (try_module_get(ipcdev->owner))
+			return ipcdev;
+
+		put_device(&ipcdev->dev);
 	}
 
-	mutex_unlock(&ipclock);
-	return scu;
+	return NULL;
 }
 EXPORT_SYMBOL_GPL(intel_scu_ipc_dev_get);
 
@@ -289,12 +288,11 @@ static int pwr_reg_rdwr(struct intel_scu_ipc_dev *scu, u16 *addr, u8 *data,
 
 	memset(cbuf, 0, sizeof(cbuf));
 
-	mutex_lock(&ipclock);
+	guard(mutex)(&ipclock);
+
 	scu = intel_scu_ipc_get(scu);
-	if (IS_ERR(scu)) {
-		mutex_unlock(&ipclock);
+	if (IS_ERR(scu))
 		return PTR_ERR(scu);
-	}
 
 	for (nc = 0; nc < count; nc++, offset += 2) {
 		cbuf[offset] = addr[nc];
@@ -324,7 +322,6 @@ static int pwr_reg_rdwr(struct intel_scu_ipc_dev *scu, u16 *addr, u8 *data,
 			wbuf[nc] = ipc_data_readl(scu, offset);
 		memcpy(data, wbuf, count);
 	}
-	mutex_unlock(&ipclock);
 	return err;
 }
 
@@ -446,17 +443,15 @@ int intel_scu_ipc_dev_simple_command(struct intel_scu_ipc_dev *scu, int cmd,
 	u32 cmdval;
 	int err;
 
-	mutex_lock(&ipclock);
+	guard(mutex)(&ipclock);
+
 	scu = intel_scu_ipc_get(scu);
-	if (IS_ERR(scu)) {
-		mutex_unlock(&ipclock);
+	if (IS_ERR(scu))
 		return PTR_ERR(scu);
-	}
 
 	cmdval = sub << 12 | cmd;
 	ipc_command(scu, cmdval);
 	err = intel_scu_ipc_check_status(scu);
-	mutex_unlock(&ipclock);
 	if (err)
 		dev_err(&scu->dev, "IPC command %#x failed with %d\n", cmdval, err);
 	return err;
@@ -491,12 +486,11 @@ int intel_scu_ipc_dev_command_with_size(struct intel_scu_ipc_dev *scu, int cmd,
 	if (inbuflen > 4 || outbuflen > 4)
 		return -EINVAL;
 
-	mutex_lock(&ipclock);
+	guard(mutex)(&ipclock);
+
 	scu = intel_scu_ipc_get(scu);
-	if (IS_ERR(scu)) {
-		mutex_unlock(&ipclock);
+	if (IS_ERR(scu))
 		return PTR_ERR(scu);
-	}
 
 	memcpy(inbuf, in, inlen);
 	for (i = 0; i < inbuflen; i++)
@@ -515,7 +509,6 @@ int intel_scu_ipc_dev_command_with_size(struct intel_scu_ipc_dev *scu, int cmd,
 		memcpy(out, outbuf, outlen);
 	}
 
-	mutex_unlock(&ipclock);
 	if (err)
 		dev_err(&scu->dev, "IPC command %#x failed with %d\n", cmdval, err);
 	return err;
@@ -572,18 +565,15 @@ __intel_scu_ipc_register(struct device *parent,
 	struct intel_scu_ipc_dev *scu;
 	void __iomem *ipc_base;
 
-	mutex_lock(&ipclock);
+	guard(mutex)(&ipclock);
+
 	/* We support only one IPC */
-	if (ipcdev) {
-		err = -EBUSY;
-		goto err_unlock;
-	}
+	if (ipcdev)
+		return ERR_PTR(-EBUSY);
 
 	scu = kzalloc(sizeof(*scu), GFP_KERNEL);
-	if (!scu) {
-		err = -ENOMEM;
-		goto err_unlock;
-	}
+	if (!scu)
+		return ERR_PTR(-ENOMEM);
 
 	scu->owner = owner;
 	scu->dev.parent = parent;
@@ -621,13 +611,11 @@ __intel_scu_ipc_register(struct device *parent,
 	err = device_register(&scu->dev);
 	if (err) {
 		put_device(&scu->dev);
-		goto err_unlock;
+		return ERR_PTR(err);
 	}
 
 	/* Assign device at last */
 	ipcdev = scu;
-	mutex_unlock(&ipclock);
-
 	return scu;
 
 err_unmap:
@@ -636,9 +624,6 @@ __intel_scu_ipc_register(struct device *parent,
 	release_mem_region(scu_data->mem.start, resource_size(&scu_data->mem));
 err_free:
 	kfree(scu);
-err_unlock:
-	mutex_unlock(&ipclock);
-
 	return ERR_PTR(err);
 }
 EXPORT_SYMBOL_GPL(__intel_scu_ipc_register);
@@ -652,12 +637,12 @@ EXPORT_SYMBOL_GPL(__intel_scu_ipc_register);
  */
 void intel_scu_ipc_unregister(struct intel_scu_ipc_dev *scu)
 {
-	mutex_lock(&ipclock);
+	guard(mutex)(&ipclock);
+
 	if (!WARN_ON(!ipcdev)) {
 		ipcdev = NULL;
 		device_unregister(&scu->dev);
 	}
-	mutex_unlock(&ipclock);
 }
 EXPORT_SYMBOL_GPL(intel_scu_ipc_unregister);
 
-- 
2.43.0.rc1.1336.g36b5255a03ac

[PATCH v3 4/5] platform/x86: intel_scu_ipc: Convert to check for errors first

[Andy Shevchenko]

Convert to use usual pattern, i.e. check for errors first. This makes
code less indented and easier to read after all.

Tested-by: Ferry Toth <fntoth@gmail.com>
Acked-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
---
 drivers/platform/x86/intel_scu_ipc.c | 37 ++++++++++++++--------------
 1 file changed, 19 insertions(+), 18 deletions(-)

diff --git a/drivers/platform/x86/intel_scu_ipc.c b/drivers/platform/x86/intel_scu_ipc.c
index 9829d4145c58..2d25a0a86143 100644
--- a/drivers/platform/x86/intel_scu_ipc.c
+++ b/drivers/platform/x86/intel_scu_ipc.c
@@ -317,12 +317,15 @@ static int pwr_reg_rdwr(struct intel_scu_ipc_dev *scu, u16 *addr, u8 *data,
 	}
 
 	err = intel_scu_ipc_check_status(scu);
-	if (!err) { /* Read rbuf */
-		for (nc = 0, offset = 0; nc < 4; nc++, offset += 4)
-			wbuf[nc] = ipc_data_readl(scu, offset);
-		memcpy(data, wbuf, count);
-	}
-	return err;
+	if (err)
+		return err;
+
+	/* Read rbuf */
+	for (nc = 0, offset = 0; nc < 4; nc++, offset += 4)
+		wbuf[nc] = ipc_data_readl(scu, offset);
+	memcpy(data, wbuf, count);
+
+	return 0;
 }
 
 /**
@@ -480,7 +483,7 @@ int intel_scu_ipc_dev_command_with_size(struct intel_scu_ipc_dev *scu, int cmd,
 {
 	size_t outbuflen = DIV_ROUND_UP(outlen, sizeof(u32));
 	size_t inbuflen = DIV_ROUND_UP(inlen, sizeof(u32));
-	u32 cmdval, inbuf[4] = {};
+	u32 cmdval, inbuf[4] = {}, outbuf[4] = {};
 	int i, err;
 
 	if (inbuflen > 4 || outbuflen > 4)
@@ -499,19 +502,17 @@ int intel_scu_ipc_dev_command_with_size(struct intel_scu_ipc_dev *scu, int cmd,
 	cmdval = (size << 16) | (sub << 12) | cmd;
 	ipc_command(scu, cmdval);
 	err = intel_scu_ipc_check_status(scu);
-
-	if (!err) {
-		u32 outbuf[4] = {};
-
-		for (i = 0; i < outbuflen; i++)
-			outbuf[i] = ipc_data_readl(scu, 4 * i);
-
-		memcpy(out, outbuf, outlen);
+	if (err) {
+		dev_err(&scu->dev, "IPC command %#x failed with %d\n", cmdval, err);
+		return err;
 	}
 
-	if (err)
-		dev_err(&scu->dev, "IPC command %#x failed with %d\n", cmdval, err);
-	return err;
+	for (i = 0; i < outbuflen; i++)
+		outbuf[i] = ipc_data_readl(scu, 4 * i);
+
+	memcpy(out, outbuf, outlen);
+
+	return 0;
 }
 EXPORT_SYMBOL(intel_scu_ipc_dev_command_with_size);
 
-- 
2.43.0.rc1.1336.g36b5255a03ac

[PATCH v3 5/5] platform/x86: intel_scu_ipc: Save a copy of the entire struct intel_scu_ipc_data

[Andy Shevchenko]

Save a copy of the entire struct intel_scu_ipc_data for easier
maintenance in case of expanding (adding new members become simpler).

Acked-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Tested-by: Ferry Toth <fntoth@gmail.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
---
 drivers/platform/x86/intel_scu_ipc.c | 33 ++++++++++++++--------------
 1 file changed, 17 insertions(+), 16 deletions(-)

diff --git a/drivers/platform/x86/intel_scu_ipc.c b/drivers/platform/x86/intel_scu_ipc.c
index 2d25a0a86143..3acf6149a9ec 100644
--- a/drivers/platform/x86/intel_scu_ipc.c
+++ b/drivers/platform/x86/intel_scu_ipc.c
@@ -57,11 +57,11 @@
 
 struct intel_scu_ipc_dev {
 	struct device dev;
-	struct resource mem;
 	struct module *owner;
-	int irq;
 	void __iomem *ipc_base;
 	struct completion cmd_complete;
+
+	struct intel_scu_ipc_data data;
 };
 
 #define IPC_STATUS		0x04
@@ -255,7 +255,7 @@ static inline int ipc_wait_for_interrupt(struct intel_scu_ipc_dev *scu)
 
 static int intel_scu_ipc_check_status(struct intel_scu_ipc_dev *scu)
 {
-	return scu->irq > 0 ? ipc_wait_for_interrupt(scu) : busy_loop(scu);
+	return scu->data.irq > 0 ? ipc_wait_for_interrupt(scu) : busy_loop(scu);
 }
 
 static struct intel_scu_ipc_dev *intel_scu_ipc_get(struct intel_scu_ipc_dev *scu)
@@ -536,13 +536,13 @@ static irqreturn_t ioc(int irq, void *dev_id)
 
 static void intel_scu_ipc_release(struct device *dev)
 {
-	struct intel_scu_ipc_dev *scu;
+	struct intel_scu_ipc_dev *scu = container_of(dev, struct intel_scu_ipc_dev, dev);
+	struct intel_scu_ipc_data *data = &scu->data;
 
-	scu = container_of(dev, struct intel_scu_ipc_dev, dev);
-	if (scu->irq > 0)
-		free_irq(scu->irq, scu);
+	if (data->irq > 0)
+		free_irq(data->irq, scu);
 	iounmap(scu->ipc_base);
-	release_mem_region(scu->mem.start, resource_size(&scu->mem));
+	release_mem_region(data->mem.start, resource_size(&data->mem));
 	kfree(scu);
 }
 
@@ -563,6 +563,7 @@ __intel_scu_ipc_register(struct device *parent,
 			 struct module *owner)
 {
 	int err;
+	struct intel_scu_ipc_data *data;
 	struct intel_scu_ipc_dev *scu;
 	void __iomem *ipc_base;
 
@@ -581,25 +582,25 @@ __intel_scu_ipc_register(struct device *parent,
 	scu->dev.class = &intel_scu_ipc_class;
 	scu->dev.release = intel_scu_ipc_release;
 
-	if (!request_mem_region(scu_data->mem.start, resource_size(&scu_data->mem),
-				"intel_scu_ipc")) {
+	memcpy(&scu->data, scu_data, sizeof(scu->data));
+	data = &scu->data;
+
+	if (!request_mem_region(data->mem.start, resource_size(&data->mem), "intel_scu_ipc")) {
 		err = -EBUSY;
 		goto err_free;
 	}
 
-	ipc_base = ioremap(scu_data->mem.start, resource_size(&scu_data->mem));
+	ipc_base = ioremap(data->mem.start, resource_size(&data->mem));
 	if (!ipc_base) {
 		err = -ENOMEM;
 		goto err_release;
 	}
 
 	scu->ipc_base = ipc_base;
-	scu->mem = scu_data->mem;
-	scu->irq = scu_data->irq;
 	init_completion(&scu->cmd_complete);
 
-	if (scu->irq > 0) {
-		err = request_irq(scu->irq, ioc, 0, "intel_scu_ipc", scu);
+	if (data->irq > 0) {
+		err = request_irq(data->irq, ioc, 0, "intel_scu_ipc", scu);
 		if (err)
 			goto err_unmap;
 	}
@@ -622,7 +623,7 @@ __intel_scu_ipc_register(struct device *parent,
 err_unmap:
 	iounmap(ipc_base);
 err_release:
-	release_mem_region(scu_data->mem.start, resource_size(&scu_data->mem));
+	release_mem_region(data->mem.start, resource_size(&data->mem));
 err_free:
 	kfree(scu);
 	return ERR_PTR(err);
-- 
2.43.0.rc1.1336.g36b5255a03ac

Re: [PATCH v3 0/5] platform/x86: intel_scu_ipc: Avoid working around IO and cleanups

[Ilpo Järvinen]

On Mon, 21 Oct 2024 16:34:28 +0300, Andy Shevchenko wrote:

> The first patch avoids using a workaround for IO (which seems unneeded).
> The rest is a batch of cleanups. Has been tested on Intel Merrifield
> (thanks, Ferry).
> 
> v3:
> - split patches 1 & 2 to two patches each (Ilpo)
> - preserve a comment for reading buffer (Ilpo)
> - inherited tags for new patches as the result is the same code wise
> 
> [...]


Thank you for your contribution, it has been applied to my local
review-ilpo branch. Note it will show up in the public
platform-drivers-x86/review-ilpo branch only once I've pushed my
local branch there, which might take a while.

The list of commits applied:
[1/5] platform/x86: intel_scu_ipc: Unify the flow in pwr_reg_rdwr()
      commit: 6bd35252102e2b79acb54c74527c81c09f690759
[2/5] platform/x86: intel_scu_ipc: Replace workaround by 32-bit IO
      commit: 8b142950a5a718d55b09f08900306c75ceed2cf9
[3/5] platform/x86: intel_scu_ipc: Simplify code with cleanup helpers
      commit: ce44b96261a6af318a9e28a16f0d9749e89179a6
[4/5] platform/x86: intel_scu_ipc: Convert to check for errors first
      commit: acf1b04c6890aef9bb356195221e75ce3862f84e
[5/5] platform/x86: intel_scu_ipc: Save a copy of the entire struct intel_scu_ipc_data
      commit: f02fcc7c12a61cfcb65937a570915672fa33ce73

--
 i.