Skip to contentSkip to conversation

ptp_ocp: set of small cleanups

PATCH In review Andy Shevchenko 5 participants 18 messages Jun 8, 2022 – Jun 10, 2022

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

cover ptp_ocp: set of small cleanups

The set of (independent) cleanups against ptp_ocp driver. Each patch has its own description, no need to repeat it here.

Andy Shevchenko (5):
ptp_ocp: use dev_err_probe()
ptp_ocp: use bits.h macros for all masks
ptp_ocp: drop duplicate NULL check in ptp_ocp_detach()
ptp_ocp: do not call pci_set_drvdata(pdev, NULL)
ptp_ocp: replace kzalloc(x*y) by kcalloc(y, x)

 drivers/ptp/ptp_ocp.c | 36 +++++++++++++++---------------------
 1 file changed, 15 insertions(+), 21 deletions(-)
Signature · 1 line
2.35.1
#1 lore raw

PATCH 1/5 ptp_ocp: use dev_err_probe()

Simplify the error path in ->probe() and unify message format a bit by using dev_err_probe().

---
 drivers/ptp/ptp_ocp.c | 13 +++++--------
 1 file changed, 5 insertions(+), 8 deletions(-)
drivers/ptp/ptp_ocp.c +5 -8
index 4519ef42b458..17930762fde9 100644
@@ -3722,14 +3722,12 @@ ptp_ocp_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 	int err;
 
 	devlink = devlink_alloc(&ptp_ocp_devlink_ops, sizeof(*bp), &pdev->dev);
-	if (!devlink) {
-		dev_err(&pdev->dev, "devlink_alloc failed\n");
-		return -ENOMEM;
-	}
+	if (!devlink)
+		return dev_err_probe(&pdev->dev, -ENOMEM, "devlink_alloc failed\n");
 
 	err = pci_enable_device(pdev);
 	if (err) {
-		dev_err(&pdev->dev, "pci_enable_device\n");
+		dev_err_probe(&pdev->dev, err, "pci_enable_device\n");
 		goto out_free;
 	}
 
@@ -3745,7 +3743,7 @@ ptp_ocp_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 	 */
 	err = pci_alloc_irq_vectors(pdev, 1, 17, PCI_IRQ_MSI | PCI_IRQ_MSIX);
 	if (err < 0) {
-		dev_err(&pdev->dev, "alloc_irq_vectors err: %d\n", err);
+		dev_err_probe(&pdev->dev, err, "alloc_irq_vectors\n");
 		goto out;
 	}
 	bp->n_irqs = err;
@@ -3757,8 +3755,7 @@ ptp_ocp_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 
 	bp->ptp = ptp_clock_register(&bp->ptp_info, &pdev->dev);
 	if (IS_ERR(bp->ptp)) {
-		err = PTR_ERR(bp->ptp);
-		dev_err(&pdev->dev, "ptp_clock_register: %d\n", err);
+		err = dev_err_probe(&pdev->dev, PTR_ERR(bp->ptp), "ptp_clock_register\n");
 		bp->ptp = NULL;
 		goto out;
 	}
Signature · 1 line
2.35.1
Acked-by on v1 1/5
#2 lore raw

Re: [PATCH net-next v1 1/5] ptp_ocp: use dev_err_probe()

Quoted 4 lines · Andy Shevchenko

Simplify the error path in ->probe() and unify message format a bit by using dev_err_probe().

LGTM

Quoted 45 lines
---
  drivers/ptp/ptp_ocp.c | 13 +++++--------
  1 file changed, 5 insertions(+), 8 deletions(-)
drivers/ptp/ptp_ocp.c +5 -8
index 4519ef42b458..17930762fde9 100644
@@ -3722,14 +3722,12 @@ ptp_ocp_probe(struct pci_dev *pdev, const struct pci_device_id *id)
  	int err;
 
  	devlink = devlink_alloc(&ptp_ocp_devlink_ops, sizeof(*bp), &pdev->dev);
-	if (!devlink) {
-		dev_err(&pdev->dev, "devlink_alloc failed\n");
-		return -ENOMEM;
-	}
+	if (!devlink)
+		return dev_err_probe(&pdev->dev, -ENOMEM, "devlink_alloc failed\n");
 
  	err = pci_enable_device(pdev);
  	if (err) {
-		dev_err(&pdev->dev, "pci_enable_device\n");
+		dev_err_probe(&pdev->dev, err, "pci_enable_device\n");
  		goto out_free;
  	}
 
@@ -3745,7 +3743,7 @@ ptp_ocp_probe(struct pci_dev *pdev, const struct pci_device_id *id)
  	 */
  	err = pci_alloc_irq_vectors(pdev, 1, 17, PCI_IRQ_MSI | PCI_IRQ_MSIX);
  	if (err < 0) {
-		dev_err(&pdev->dev, "alloc_irq_vectors err: %d\n", err);
+		dev_err_probe(&pdev->dev, err, "alloc_irq_vectors\n");
  		goto out;
  	}
  	bp->n_irqs = err;
@@ -3757,8 +3755,7 @@ ptp_ocp_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 
  	bp->ptp = ptp_clock_register(&bp->ptp_info, &pdev->dev);
  	if (IS_ERR(bp->ptp)) {
-		err = PTR_ERR(bp->ptp);
-		dev_err(&pdev->dev, "ptp_clock_register: %d\n", err);
+		err = dev_err_probe(&pdev->dev, PTR_ERR(bp->ptp), "ptp_clock_register\n");
  		bp->ptp = NULL;
  		goto out;
  	}
#3 lore raw

Re: [PATCH net-next v1 1/5] ptp_ocp: use dev_err_probe()

In reply to Andy Shevchenko
Quoted 2 lines · Andy Shevchenko

Simplify the error path in ->probe() and unify message format a bit by using dev_err_probe().

Line length.

Signature · 1 line
Jonathan
maintainer
#4 lore raw

Re: [PATCH net-next v1 1/5] ptp_ocp: use dev_err_probe()

In reply to Andy Shevchenko
Quoted 2 lines · On Wed, 8 Jun 2022 15:03:54 +0300 Andy Shevchenko wrote:

Simplify the error path in ->probe() and unify message format a bit by using dev_err_probe().

Let's not do this. I get that using dev_err_probe() even without possibility of -EPROBE_DEFER is acceptable, but converting existing drivers is too much IMO. Acceptable < best practice.

#5 lore raw

Re: [PATCH net-next v1 1/5] ptp_ocp: use dev_err_probe()

Quoted 7 lines · Jakub Kicinski
Quoted 2 lines · On Wed, 8 Jun 2022 15:03:54 +0300 Andy Shevchenko wrote:

Simplify the error path in ->probe() and unify message format a bit by using dev_err_probe().

Let's not do this. I get that using dev_err_probe() even without possibility of -EPROBE_DEFER is acceptable, but converting existing drivers is too much IMO. Acceptable < best practice.

Noted.

I have just checked that if you drop this patch the rest will be still applicable. If you have no objections, can you apply patches 2-5 then?

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

Re: [PATCH net-next v1 1/5] ptp_ocp: use dev_err_probe()

Quoted 2 lines · On Fri, 10 Jun 2022 14:09:24 +0300 Andy Shevchenko wrote:

I have just checked that if you drop this patch the rest will be still applicable. If you have no objections, can you apply patches 2-5 then?

It's tradition in netdev to ask people to repost. But looks completely safe for me to drop patch 1, so applied 2-5. Don't tell anyone I did this.

#7 lore raw

Re: [PATCH net-next v1 1/5] ptp_ocp: use dev_err_probe()

Quoted 6 lines · Jakub Kicinski
Quoted 2 lines · On Fri, 10 Jun 2022 14:09:24 +0300 Andy Shevchenko wrote:

I have just checked that if you drop this patch the rest will be still applicable. If you have no objections, can you apply patches 2-5 then?

It's tradition in netdev to ask people to repost. But looks completely safe for me to drop patch 1, so applied 2-5. Don't tell anyone I did this.

I see what you did there. :)

Signature · 1 line
Jonathan
#8 lore raw

Re: [PATCH net-next v1 1/5] ptp_ocp: use dev_err_probe()

In reply to Jakub Kicinski
Quoted 7 lines · Jakub Kicinski
Quoted 2 lines · On Fri, 10 Jun 2022 14:09:24 +0300 Andy Shevchenko wrote:

I have just checked that if you drop this patch the rest will be still applicable. If you have no objections, can you apply patches 2-5 then?

It's tradition in netdev to ask people to repost. But looks completely safe for me to drop patch 1, so applied 2-5.

Thanks!

Quoted 1 line

Don't tell anyone I did this.

Tschhh!

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

PATCH 2/5 ptp_ocp: use bits.h macros for all masks

In reply to Andy Shevchenko

Currently we are using BIT(), but GENMASK(). Make use of the latter one as well (far less error-prone, far more concise).

---
 drivers/ptp/ptp_ocp.c | 13 +++++++------
 1 file changed, 7 insertions(+), 6 deletions(-)
drivers/ptp/ptp_ocp.c +7 -6
index 17930762fde9..926add7be9a2 100644
@@ -1,6 +1,7 @@
 // SPDX-License-Identifier: GPL-2.0-only
 /* Copyright (c) 2020 Facebook */
 
+#include <linux/bits.h>
 #include <linux/err.h>
 #include <linux/kernel.h>
 #include <linux/module.h>
@@ -88,10 +89,10 @@ struct tod_reg {
 #define TOD_CTRL_DISABLE_FMT_A	BIT(17)
 #define TOD_CTRL_DISABLE_FMT_B	BIT(16)
 #define TOD_CTRL_ENABLE		BIT(0)
-#define TOD_CTRL_GNSS_MASK	((1U << 4) - 1)
+#define TOD_CTRL_GNSS_MASK	GENMASK(3, 0)
 #define TOD_CTRL_GNSS_SHIFT	24
 
-#define TOD_STATUS_UTC_MASK		0xff
+#define TOD_STATUS_UTC_MASK		GENMASK(7, 0)
 #define TOD_STATUS_UTC_VALID		BIT(8)
 #define TOD_STATUS_LEAP_ANNOUNCE	BIT(12)
 #define TOD_STATUS_LEAP_VALID		BIT(16)
@@ -205,7 +206,7 @@ struct frequency_reg {
 #define FREQ_STATUS_VALID	BIT(31)
 #define FREQ_STATUS_ERROR	BIT(30)
 #define FREQ_STATUS_OVERRUN	BIT(29)
-#define FREQ_STATUS_MASK	(BIT(24) - 1)
+#define FREQ_STATUS_MASK	GENMASK(23, 0)
 
 struct ptp_ocp_flash_info {
 	const char *name;
@@ -674,9 +675,9 @@ static const struct ocp_selector ptp_ocp_clock[] = {
 	{ }
 };
 
+#define SMA_DISABLE		BIT(16)
 #define SMA_ENABLE		BIT(15)
-#define SMA_SELECT_MASK		((1U << 15) - 1)
-#define SMA_DISABLE		0x10000
+#define SMA_SELECT_MASK		GENMASK(14, 0)
 
 static const struct ocp_selector ptp_ocp_sma_in[] = {
 	{ .name = "10Mhz",	.value = 0x0000 },
@@ -3440,7 +3441,7 @@ ptp_ocp_tod_status_show(struct seq_file *s, void *data)
 
 	val = ioread32(&bp->tod->utc_status);
 	seq_printf(s, "UTC status register: 0x%08X\n", val);
-	seq_printf(s, "UTC offset: %d  valid:%d\n",
+	seq_printf(s, "UTC offset: %ld  valid:%d\n",
 		val & TOD_STATUS_UTC_MASK, val & TOD_STATUS_UTC_VALID ? 1 : 0);
 	seq_printf(s, "Leap second info valid:%d, Leap second announce %d\n",
 		val & TOD_STATUS_LEAP_VALID ? 1 : 0,
Signature · 1 line
2.35.1
Acked-by on v1 2/5
#10 lore raw

Re: [PATCH v1 net-next 2/5] ptp_ocp: use bits.h macros for all masks

Quoted 4 lines · Andy Shevchenko

Currently we are using BIT(), but GENMASK(). Make use of the latter one as well (far less error-prone, far more concise).

LGTM

Quoted 59 lines
---
  drivers/ptp/ptp_ocp.c | 13 +++++++------
  1 file changed, 7 insertions(+), 6 deletions(-)
drivers/ptp/ptp_ocp.c +7 -6
index 17930762fde9..926add7be9a2 100644
@@ -1,6 +1,7 @@
  // SPDX-License-Identifier: GPL-2.0-only
  /* Copyright (c) 2020 Facebook */
 
+#include <linux/bits.h>
  #include <linux/err.h>
  #include <linux/kernel.h>
  #include <linux/module.h>
@@ -88,10 +89,10 @@ struct tod_reg {
  #define TOD_CTRL_DISABLE_FMT_A	BIT(17)
  #define TOD_CTRL_DISABLE_FMT_B	BIT(16)
  #define TOD_CTRL_ENABLE		BIT(0)
-#define TOD_CTRL_GNSS_MASK	((1U << 4) - 1)
+#define TOD_CTRL_GNSS_MASK	GENMASK(3, 0)
  #define TOD_CTRL_GNSS_SHIFT	24
 
-#define TOD_STATUS_UTC_MASK		0xff
+#define TOD_STATUS_UTC_MASK		GENMASK(7, 0)
  #define TOD_STATUS_UTC_VALID		BIT(8)
  #define TOD_STATUS_LEAP_ANNOUNCE	BIT(12)
  #define TOD_STATUS_LEAP_VALID		BIT(16)
@@ -205,7 +206,7 @@ struct frequency_reg {
  #define FREQ_STATUS_VALID	BIT(31)
  #define FREQ_STATUS_ERROR	BIT(30)
  #define FREQ_STATUS_OVERRUN	BIT(29)
-#define FREQ_STATUS_MASK	(BIT(24) - 1)
+#define FREQ_STATUS_MASK	GENMASK(23, 0)
 
  struct ptp_ocp_flash_info {
  	const char *name;
@@ -674,9 +675,9 @@ static const struct ocp_selector ptp_ocp_clock[] = {
  	{ }
  };
 
+#define SMA_DISABLE		BIT(16)
  #define SMA_ENABLE		BIT(15)
-#define SMA_SELECT_MASK		((1U << 15) - 1)
-#define SMA_DISABLE		0x10000
+#define SMA_SELECT_MASK		GENMASK(14, 0)
 
  static const struct ocp_selector ptp_ocp_sma_in[] = {
  	{ .name = "10Mhz",	.value = 0x0000 },
@@ -3440,7 +3441,7 @@ ptp_ocp_tod_status_show(struct seq_file *s, void *data)
 
  	val = ioread32(&bp->tod->utc_status);
  	seq_printf(s, "UTC status register: 0x%08X\n", val);
-	seq_printf(s, "UTC offset: %d  valid:%d\n",
+	seq_printf(s, "UTC offset: %ld  valid:%d\n",
  		val & TOD_STATUS_UTC_MASK, val & TOD_STATUS_UTC_VALID ? 1 : 0);
  	seq_printf(s, "Leap second info valid:%d, Leap second announce %d\n",
  		val & TOD_STATUS_LEAP_VALID ? 1 : 0,
#11 lore raw

PATCH 3/5 ptp_ocp: drop duplicate NULL check in ptp_ocp_detach()

In reply to Andy Shevchenko

Since platform_device_unregister() is NULL-aware, we don't need to duplicate this check. Remove it and fold the rest of the code.

---
 drivers/ptp/ptp_ocp.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)
drivers/ptp/ptp_ocp.c +2 -4
index 926add7be9a2..4e237f806085 100644
@@ -3701,10 +3701,8 @@ ptp_ocp_detach(struct ptp_ocp *bp)
 		serial8250_unregister_port(bp->mac_port);
 	if (bp->nmea_port != -1)
 		serial8250_unregister_port(bp->nmea_port);
-	if (bp->spi_flash)
-		platform_device_unregister(bp->spi_flash);
-	if (bp->i2c_ctrl)
-		platform_device_unregister(bp->i2c_ctrl);
+	platform_device_unregister(bp->spi_flash);
+	platform_device_unregister(bp->i2c_ctrl);
 	if (bp->i2c_clk)
 		clk_hw_unregister_fixed_rate(bp->i2c_clk);
 	if (bp->n_irqs)
Signature · 1 line
2.35.1
Acked-by on v1 3/5
#12 lore raw

Re: [PATCH v1 net-next 3/5] ptp_ocp: drop duplicate NULL check in ptp_ocp_detach()

Quoted 4 lines · Andy Shevchenko

Since platform_device_unregister() is NULL-aware, we don't need to duplicate this check. Remove it and fold the rest of the code.

Quoted 21 lines
---
  drivers/ptp/ptp_ocp.c | 6 ++----
  1 file changed, 2 insertions(+), 4 deletions(-)
drivers/ptp/ptp_ocp.c +2 -4
index 926add7be9a2..4e237f806085 100644
@@ -3701,10 +3701,8 @@ ptp_ocp_detach(struct ptp_ocp *bp)
  		serial8250_unregister_port(bp->mac_port);
  	if (bp->nmea_port != -1)
  		serial8250_unregister_port(bp->nmea_port);
-	if (bp->spi_flash)
-		platform_device_unregister(bp->spi_flash);
-	if (bp->i2c_ctrl)
-		platform_device_unregister(bp->i2c_ctrl);
+	platform_device_unregister(bp->spi_flash);
+	platform_device_unregister(bp->i2c_ctrl);
  	if (bp->i2c_clk)
  		clk_hw_unregister_fixed_rate(bp->i2c_clk);
  	if (bp->n_irqs)
#13 lore raw

PATCH 4/5 ptp_ocp: do not call pci_set_drvdata(pdev, NULL)

In reply to Andy Shevchenko

Cleaning up driver data is actually already handled by driver core, so there is no need to do it manually.

---
 drivers/ptp/ptp_ocp.c | 2 --
 1 file changed, 2 deletions(-)
drivers/ptp/ptp_ocp.c +0 -2
index 4e237f806085..857e35c68a04 100644
@@ -3769,7 +3769,6 @@ ptp_ocp_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 
 out:
 	ptp_ocp_detach(bp);
-	pci_set_drvdata(pdev, NULL);
 out_disable:
 	pci_disable_device(pdev);
 out_free:
@@ -3785,7 +3784,6 @@ ptp_ocp_remove(struct pci_dev *pdev)
 
 	devlink_unregister(devlink);
 	ptp_ocp_detach(bp);
-	pci_set_drvdata(pdev, NULL);
 	pci_disable_device(pdev);
 
 	devlink_free(devlink);
Signature · 1 line
2.35.1
Acked-by on v1 4/5
#14 lore raw

Re: [PATCH v1 net-next 4/5] ptp_ocp: do not call pci_set_drvdata(pdev, NULL)

Quoted 2 lines · Andy Shevchenko

Cleaning up driver data is actually already handled by driver core, so there is no need to do it manually.

I found a couple of places with exactly the same code in error path. For example Marvell's OcteonX drivers in crypto and net subsystems. Should we fix them too?

Quoted 1 line

Overall looks good.

Quoted 24 lines
---
  drivers/ptp/ptp_ocp.c | 2 --
  1 file changed, 2 deletions(-)
drivers/ptp/ptp_ocp.c +0 -2
index 4e237f806085..857e35c68a04 100644
@@ -3769,7 +3769,6 @@ ptp_ocp_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 
  out:
  	ptp_ocp_detach(bp);
-	pci_set_drvdata(pdev, NULL);
  out_disable:
  	pci_disable_device(pdev);
  out_free:
@@ -3785,7 +3784,6 @@ ptp_ocp_remove(struct pci_dev *pdev)
 
  	devlink_unregister(devlink);
  	ptp_ocp_detach(bp);
-	pci_set_drvdata(pdev, NULL);
  	pci_disable_device(pdev);
 
  	devlink_free(devlink);
#15 lore raw

Re: [PATCH v1 net-next 4/5] ptp_ocp: do not call pci_set_drvdata(pdev, NULL)

Quoted 8 lines · Vadim Fedorenko
Quoted 2 lines · Andy Shevchenko

Cleaning up driver data is actually already handled by driver core, so there is no need to do it manually.

I found a couple of places with exactly the same code in error path. For example Marvell's OcteonX drivers in crypto and net subsystems. Should we fix them too?

I believe there are even more, but feel free to fix them, they are not my POI.

...

Quoted 3 lines

Overall looks good.

Thanks!

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

PATCH 5/5 ptp_ocp: replace kzalloc(x*y) by kcalloc(y, x)

In reply to Andy Shevchenko

While here it may be no difference, the kcalloc() has some checks against overflow and it's logically correct to call it for an array.

---
 drivers/ptp/ptp_ocp.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/ptp/ptp_ocp.c +1 -1
index 857e35c68a04..83da36e69361 100644
@@ -2155,7 +2155,7 @@ ptp_ocp_fb_set_pins(struct ptp_ocp *bp)
 	struct ptp_pin_desc *config;
 	int i;
 
-	config = kzalloc(sizeof(*config) * 4, GFP_KERNEL);
+	config = kcalloc(4, sizeof(*config), GFP_KERNEL);
 	if (!config)
 		return -ENOMEM;
 
Signature · 1 line
2.35.1
Acked-by on v1 5/5
#17 lore raw

Re: [PATCH v1 net-next 5/5] ptp_ocp: replace kzalloc(x*y) by kcalloc(y, x)

Quoted 4 lines · Andy Shevchenko

While here it may be no difference, the kcalloc() has some checks against overflow and it's logically correct to call it for an array.

Quoted 17 lines
---
  drivers/ptp/ptp_ocp.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
drivers/ptp/ptp_ocp.c +1 -1
index 857e35c68a04..83da36e69361 100644
@@ -2155,7 +2155,7 @@ ptp_ocp_fb_set_pins(struct ptp_ocp *bp)
  	struct ptp_pin_desc *config;
  	int i;
 
-	config = kzalloc(sizeof(*config) * 4, GFP_KERNEL);
+	config = kcalloc(4, sizeof(*config), GFP_KERNEL);
  	if (!config)
  		return -ENOMEM;