Skip to contentSkip to conversation

perf/x86: Use sysfs_emit() in show() callback function

PATCH Superseded Deepak R Varma 2 participants 7 messages Feb 21, 2023 – Feb 26, 2023

Subsystems (2)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

v2 cover perf/x86: Use sysfs_emit() in show() callback function

The patch set proposes to replace use of sprintf family function usage for formatting data to be returned to the user space using show callback functions.

The modified files have the same maintainer list though they correspond to different drivers. Since its the same, I clubbed these into a patch set. If this is not correct, I can send in individual driver specific patches.

Changes in v2:
1. Update the patch log message to provide details on the problem associated with current implementation and how the proposal is a better solution. Feedback provided by Peter Zijlstra <[email protected]>

Deepak R Varma (3):
perf/x86/core: Use sysfs_emit() in show() callback function perf/x86/intel/pt: Use sysfs_emit() in show() callback function perf/x86/intel: Use sysfs_emit() in show() callback function

 arch/x86/events/core.c       | 8 +++-----
 arch/x86/events/intel/core.c | 6 +++---
 arch/x86/events/intel/pt.c   | 2 +-
 3 files changed, 7 insertions(+), 9 deletions(-)
Signature · 1 line
2.34.1
#1 lore raw

PATCH v2 1/3 perf/x86/core: Use sysfs_emit() in show() callback function

As per C99 standard, snprintf() returns the number of bytes that would be encoded in the destination buffer when it is sufficiently large. This return value may be different from what the caller is expecting and hence may lead to potential errors in the program.
Kernel release 2.6.2 introduced scnprintf() & vscnprintf() which precisely return the actual bytes encoded into the destination buffer.

For the sysfs attribute show() callback functions, which returns the number of bytes to the user space, a more recent recommendation is to use sysfs_emit() or sysfs_emit_at() instead of sprintf() family of functions. This is recorded in the Documentation/filesystems/sysfs.rst Kernel documentation file.

Issue identified using the coccinelle device_attr_show.cocci script.

---
Changes in v2:
- Revise patch log message to include details on the potential issues with current implementation and how the proposal is a better solution. Feedback provided by Peter Zijlstra <[email protected]>

 arch/x86/events/core.c | 8 +++-----
 1 file changed, 3 insertions(+), 5 deletions(-)
arch/x86/events/core.c +3 -5
index 85a63a41c471..27c03e6dcb5d 100644
@@ -1896,9 +1896,7 @@ ssize_t events_hybrid_sysfs_show(struct device *dev,
 		if (x86_pmu.hybrid_pmu[i].cpu_type & pmu->cpu_type) {
 			next_str = strchr(str, ';');
 			if (next_str)
-				return snprintf(page, next_str - str + 1, "%s", str);
-			else
-				return sprintf(page, "%s", str);
+				return sysfs_emit(page, "%s", str);
 		}
 		str = strchr(str, ';');
 		str++;
@@ -2544,7 +2542,7 @@ static ssize_t get_attr_rdpmc(struct device *cdev,
 			      struct device_attribute *attr,
 			      char *buf)
 {
-	return snprintf(buf, 40, "%d\n", x86_pmu.attr_rdpmc);
+	return sysfs_emit(buf, "%d\n", x86_pmu.attr_rdpmc);
 }
 
 static ssize_t set_attr_rdpmc(struct device *cdev,
@@ -2602,7 +2600,7 @@ static ssize_t max_precise_show(struct device *cdev,
 				  struct device_attribute *attr,
 				  char *buf)
 {
-	return snprintf(buf, PAGE_SIZE, "%d\n", x86_pmu_max_precise());
+	return sysfs_emit(buf, "%d\n", x86_pmu_max_precise());
 }
 
 static DEVICE_ATTR_RO(max_precise);
Signature · 1 line
2.34.1
#5 lore raw

PATCH v2 2/3 perf/x86/intel/pt: Use sysfs_emit() in show() callback function

In reply to Deepak R Varma

As per C99 standard, snprintf() returns the number of bytes that would be encoded in the destination buffer when it is sufficiently large. This return value may be different from what the caller is expecting and hence may lead to potential errors in the program.
Kernel release 2.6.2 introduced scnprintf() & vscnprintf() which precisely return the actual bytes encoded into the destination buffer.

For the sysfs attribute show() callback functions, which returns the number of bytes to the user space, a more recent recommendation is to use sysfs_emit() or sysfs_emit_at() instead of sprintf() family of functions. This is recorded in the Documentation/filesystems/sysfs.rst Kernel documentation file.

Issue identified using the coccinelle device_attr_show.cocci script.

---
Changes in v2:
- Revise patch log message to include details on the potential issues with current implementation and how the proposal is a better solution. Feedback provided by Peter Zijlstra <[email protected]>

 arch/x86/events/intel/pt.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/x86/events/intel/pt.c +1 -1
index 42a55794004a..d9e6d771b458 100644
@@ -96,7 +96,7 @@ static ssize_t pt_cap_show(struct device *cdev,
 		container_of(attr, struct dev_ext_attribute, attr);
 	enum pt_capabilities cap = (long)ea->var;
 
-	return snprintf(buf, PAGE_SIZE, "%x\n", intel_pt_validate_hw_cap(cap));
+	return sysfs_emit(buf, "%x\n", intel_pt_validate_hw_cap(cap));
 }
 
 static struct attribute_group pt_cap_group __ro_after_init = {
Signature · 1 line
2.34.1
#6 lore raw

PATCH v2 3/3 perf/x86/intel: Use sysfs_emit() in show() callback function

In reply to Deepak R Varma

As per C99 standard, snprintf() returns the number of bytes that would be encoded in the destination buffer when it is sufficiently large. This return value may be different from what the caller is expecting and hence may lead to potential errors in the program.
Kernel release 2.6.2 introduced scnprintf() & vscnprintf() which precisely return the actual bytes encoded into the destination buffer.

For the sysfs attribute show() callback functions, which returns the number of bytes to the user space, a more recent recommendation is to use sysfs_emit() or sysfs_emit_at() instead of sprintf() family of functions. This is recorded in the Documentation/filesystems/sysfs.rst Kernel documentation file.

Issue identified using the coccinelle device_attr_show.cocci script.

---
Changes in v2:
- Revise patch log message to include details on the potential issues with current implementation and how the proposal is a better solution. Feedback provided by Peter Zijlstra <[email protected]>

 arch/x86/events/intel/core.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
arch/x86/events/intel/core.c +3 -3
index bafdc2be479a..8fb1225123ef 100644
@@ -5273,7 +5273,7 @@ static ssize_t show_sysctl_tfa(struct device *cdev,
 			      struct device_attribute *attr,
 			      char *buf)
 {
-	return snprintf(buf, 40, "%d\n", allow_tsx_force_abort);
+	return sysfs_emit(buf, "%d\n", allow_tsx_force_abort);
 }
 
 static ssize_t set_sysctl_tfa(struct device *cdev,
@@ -5307,7 +5307,7 @@ static ssize_t branches_show(struct device *cdev,
 			     struct device_attribute *attr,
 			     char *buf)
 {
-	return snprintf(buf, PAGE_SIZE, "%d\n", x86_pmu.lbr_nr);
+	return sysfs_emit(buf, "%d\n", x86_pmu.lbr_nr);
 }
 
 static DEVICE_ATTR_RO(branches);
@@ -5323,7 +5323,7 @@ static ssize_t pmu_name_show(struct device *cdev,
 			     struct device_attribute *attr,
 			     char *buf)
 {
-	return snprintf(buf, PAGE_SIZE, "%s\n", pmu_name_str);
+	return sysfs_emit(buf, "%s\n", pmu_name_str);
 }
 
 static DEVICE_ATTR_RO(pmu_name);
Signature · 1 line
2.34.1
maintainer
#2 lore raw

Re: [PATCH v2 1/3] perf/x86/core: Use sysfs_emit() in show() callback function

In reply to Deepak R Varma
Quoted 5 lines · Deepak R Varma
arch/x86/events/core.c +0 -0
index 85a63a41c471..27c03e6dcb5d 100644

arch/x86/events/core.c ssize_t events_hybrid_sysfs_show(struct device *dev, · line 1902

Quoted 10 lines
arch/x86/events/core.c +1 -3
@@ -1896,9 +1896,7 @@ ssize_t events_hybrid_sysfs_show(struct device *dev,
 		if (x86_pmu.hybrid_pmu[i].cpu_type & pmu->cpu_type) {
 			next_str = strchr(str, ';');
 			if (next_str)
-				return snprintf(page, next_str - str + 1, "%s", str);
-			else
-				return sprintf(page, "%s", str);
+				return sysfs_emit(page, "%s", str);
 		}
 		str = strchr(str, ';');
 		str++;

How is this correct ?!?

#3 lore raw

Re: [PATCH v2 1/3] perf/x86/core: Use sysfs_emit() in show() callback function

Quoted 10 lines · Peter Zijlstra
Quoted 9 lines · Deepak R Varma
arch/x86/events/core.c +0 -1
index 85a63a41c471..27c03e6dcb5d 100644
@@ -1896,9 +1896,7 @@ ssize_t events_hybrid_sysfs_show(struct device *dev,
 		if (x86_pmu.hybrid_pmu[i].cpu_type & pmu->cpu_type) {
 			next_str = strchr(str, ';');
 			if (next_str)
-				return snprintf(page, next_str - str + 1, "%s", str);
Quoted 8 lines

arch/x86/events/core.c ssize_t events_hybrid_sysfs_show(struct device *dev, · line 1902

Quoted 6 lines
arch/x86/events/core.c +1 -2
-			else
-				return sprintf(page, "%s", str);
+				return sysfs_emit(page, "%s", str);
 		}
 		str = strchr(str, ';');
 		str++;

How is this correct ?!?

oops.. that is bad on my part. My apologies for the wrong code. I will correct it and send in v3.

Thank you Peter.

Regards,
./drv

#4 lore raw

Re: [PATCH v2 1/3] perf/x86/core: Use sysfs_emit() in show() callback function

Quoted 20 lines · Deepak R Varma
Quoted 10 lines · Peter Zijlstra
Quoted 9 lines · Deepak R Varma
arch/x86/events/core.c +0 -1
index 85a63a41c471..27c03e6dcb5d 100644
@@ -1896,9 +1896,7 @@ ssize_t events_hybrid_sysfs_show(struct device *dev,
 		if (x86_pmu.hybrid_pmu[i].cpu_type & pmu->cpu_type) {
 			next_str = strchr(str, ';');
 			if (next_str)
-				return snprintf(page, next_str - str + 1, "%s", str);
Quoted 8 lines

arch/x86/events/core.c ssize_t events_hybrid_sysfs_show(struct device *dev, · line 1902

Quoted 6 lines
arch/x86/events/core.c +1 -2
-			else
-				return sprintf(page, "%s", str);
+				return sysfs_emit(page, "%s", str);
 		}
 		str = strchr(str, ';');
 		str++;

How is this correct ?!?

Quoted 2 lines

oops.. that is bad on my part. My apologies for the wrong code. I will correct it and send in v3.

Hello Peter,
I reviewed the code more closely and concluded that the current implementation is better as is. I sent in a v3 with necessary correction for your review.

I do have another observation from this area that I will send in as a separate patch soon.

Thank you again.
./drv

Quoted 5 lines

Thank you Peter.

Regards,
./drv