Skip to contentSkip to conversation

drivers/perf: hisi: Fixed some issues with hisi pmu

PATCH Applied Junhao He 4 participants 11 messages Apr 25, 2024 – Apr 28, 2024

Subsystems (3)
lore
Share & tools
lore

Junhao He

author
#0 lore raw

cover drivers/perf: hisi: Fixed some issues with hisi pmu

This patchset includes 3 bug fixes changes to hisi PMU:
- Fix out-of-bound access when valid event group in hns pmu
- Fixes the memory leak in hns pmu
- Fix out-of-bound access when valid event group in pcie pmu

Hao Chen (1):
drivers/perf: hisi: hns3: Actually use devm_add_action_or_reset()

Junhao He (2):
drivers/perf: hisi_pcie: Fix out-of-bound access when valid event
group
drivers/perf: hisi: hns3: Fix out-of-bound access when valid event
group

 drivers/perf/hisilicon/hisi_pcie_pmu.c | 14 +++++++++++++-
 drivers/perf/hisilicon/hns3_pmu.c      | 16 ++++++++++++++--
 2 files changed, 27 insertions(+), 3 deletions(-)
Signature · 1 line
2.33.0
author
#1 lore raw

PATCH 1/3 drivers/perf: hisi_pcie: Fix out-of-bound access when valid event group

The perf tool allows users to create event groups through following cmd [1], but the driver does not check whether the array index is out of bounds when writing data to the event_group array. If the number of events in an event_group is greater than HISI_PCIE_MAX_COUNTERS, the memory write overflow of event_group array occurs.

Add array index check to fix the possible array out of bounds violation, and return directly when write new events are written to array bounds.

There are 9 different events in an event_group.
[1] perf stat -e '{pmu/event1/, ... ,pmu/event9/}'

Fixes 8404b0fbc7fb ("drivers/perf: hisi: Add driver for HiSilicon PCIe PMU")
---
 drivers/perf/hisilicon/hisi_pcie_pmu.c | 14 +++++++++++++-
 1 file changed, 13 insertions(+), 1 deletion(-)
drivers/perf/hisilicon/hisi_pcie_pmu.c +13 -1
index 5d1f0e9fdb08..dba399125658 100644
@@ -350,15 +350,27 @@ static bool hisi_pcie_pmu_validate_event_group(struct perf_event *event)
 			return false;
 
 		for (num = 0; num < counters; num++) {
+			/*
+			 * If we find a related event, then it's a valid group
+			 * since we don't need to allocate a new counter for it.
+			 */
 			if (hisi_pcie_pmu_cmp_event(event_group[num], sibling))
 				break;
 		}
 
+		/*
+		 * Otherwise it's a new event but if there's no available counter,
+		 * fail the check since we cannot schedule all the events in
+		 * the group simultaneously.
+		 */
+		if (num == HISI_PCIE_MAX_COUNTERS)
+			return false;
+
 		if (num == counters)
 			event_group[counters++] = sibling;
 	}
 
-	return counters <= HISI_PCIE_MAX_COUNTERS;
+	return true;
 }
 
 static int hisi_pcie_pmu_event_init(struct perf_event *event)
Signature · 1 line
2.33.0
maintainer Reviewed-by on v1 1/3
#2 lore raw

Re: [PATCH 1/3] drivers/perf: hisi_pcie: Fix out-of-bound access when valid event group

Quoted 51 lines · Junhao He

The perf tool allows users to create event groups through following cmd [1], but the driver does not check whether the array index is out of bounds when writing data to the event_group array. If the number of events in an event_group is greater than HISI_PCIE_MAX_COUNTERS, the memory write overflow of event_group array occurs.

Add array index check to fix the possible array out of bounds violation, and return directly when write new events are written to array bounds.

There are 9 different events in an event_group.
[1] perf stat -e '{pmu/event1/, ... ,pmu/event9/}'

Fixes 8404b0fbc7fb ("drivers/perf: hisi: Add driver for HiSilicon PCIe PMU")
---
  drivers/perf/hisilicon/hisi_pcie_pmu.c | 14 +++++++++++++-
  1 file changed, 13 insertions(+), 1 deletion(-)
drivers/perf/hisilicon/hisi_pcie_pmu.c +13 -1
index 5d1f0e9fdb08..dba399125658 100644
@@ -350,15 +350,27 @@ static bool hisi_pcie_pmu_validate_event_group(struct perf_event *event)
  			return false;
 
  		for (num = 0; num < counters; num++) {
+			/*
+			 * If we find a related event, then it's a valid group
+			 * since we don't need to allocate a new counter for it.
+			 */
  			if (hisi_pcie_pmu_cmp_event(event_group[num], sibling))
  				break;
  		}
 
+		/*
+		 * Otherwise it's a new event but if there's no available counter,
+		 * fail the check since we cannot schedule all the events in
+		 * the group simultaneously.
+		 */
+		if (num == HISI_PCIE_MAX_COUNTERS)
+			return false;
+
  		if (num == counters)
  			event_group[counters++] = sibling;
  	}
 
-	return counters <= HISI_PCIE_MAX_COUNTERS;
+	return true;
  }
 
  static int hisi_pcie_pmu_event_init(struct perf_event *event)
Acked-by on v1 1/3
#3 lore raw

Re: [PATCH 1/3] drivers/perf: hisi_pcie: Fix out-of-bound access when valid event group

In reply to Junhao He
Quoted 6 lines · On Thu, 25 Apr 2024 20:46:25 +0800 Junhao He <[email protected]> wrote:

The perf tool allows users to create event groups through following cmd [1], but the driver does not check whether the array index is out of bounds when writing data to the event_group array. If the number of events in an event_group is greater than HISI_PCIE_MAX_COUNTERS, the memory write overflow of event_group array occurs.

Quoted 8 lines

Add array index check to fix the possible array out of bounds violation, and return directly when write new events are written to array bounds.

There are 9 different events in an event_group.
[1] perf stat -e '{pmu/event1/, ... ,pmu/event9/}'

Fixes 8404b0fbc7fb ("drivers/perf: hisi: Add driver for HiSilicon PCIe PMU")

Thanks,

Quoted 37 lines
---
 drivers/perf/hisilicon/hisi_pcie_pmu.c | 14 +++++++++++++-
 1 file changed, 13 insertions(+), 1 deletion(-)
drivers/perf/hisilicon/hisi_pcie_pmu.c +13 -1
index 5d1f0e9fdb08..dba399125658 100644
@@ -350,15 +350,27 @@ static bool hisi_pcie_pmu_validate_event_group(struct perf_event *event)
 			return false;
 
 		for (num = 0; num < counters; num++) {
+			/*
+			 * If we find a related event, then it's a valid group
+			 * since we don't need to allocate a new counter for it.
+			 */
 			if (hisi_pcie_pmu_cmp_event(event_group[num], sibling))
 				break;
 		}
 
+		/*
+		 * Otherwise it's a new event but if there's no available counter,
+		 * fail the check since we cannot schedule all the events in
+		 * the group simultaneously.
+		 */
+		if (num == HISI_PCIE_MAX_COUNTERS)
+			return false;
+
 		if (num == counters)
 			event_group[counters++] = sibling;
 	}
 
-	return counters <= HISI_PCIE_MAX_COUNTERS;
+	return true;
 }
 
 static int hisi_pcie_pmu_event_init(struct perf_event *event)
author
#4 lore raw

PATCH 2/3 drivers/perf: hisi: hns3: Fix out-of-bound access when valid event group

In reply to Junhao He

The perf tool allows users to create event groups through following cmd [1], but the driver does not check whether the array index is out of bounds when writing data to the event_group array. If the number of events in an event_group is greater than HNS3_PMU_MAX_HW_EVENTS, the memory write overflow of event_group array occurs.

Add array index check to fix the possible array out of bounds violation, and return directly when write new events are written to array bounds.

There are 9 different events in an event_group.
[1] perf stat -e '{pmu/event1/, ... ,pmu/event9/}

Fixes 66637ab137b4 ("drivers/perf: hisi: add driver for HNS3 PMU")
---
 drivers/perf/hisilicon/hns3_pmu.c | 14 +++++++++++++-
 1 file changed, 13 insertions(+), 1 deletion(-)
drivers/perf/hisilicon/hns3_pmu.c +13 -1
index 16869bf5bf4c..cbdd53b0a034 100644
@@ -1085,15 +1085,27 @@ static bool hns3_pmu_validate_event_group(struct perf_event *event)
 			return false;
 
 		for (num = 0; num < counters; num++) {
+			/*
+			 * If we find a related event, then it's a valid group
+			 * since we don't need to allocate a new counter for it.
+			 */
 			if (hns3_pmu_cmp_event(event_group[num], sibling))
 				break;
 		}
 
+		/*
+		 * Otherwise it's a new event but if there's no available counter,
+		 * fail the check since we cannot schedule all the events in
+		 * the group simultaneously.
+		 */
+		if (num == HNS3_PMU_MAX_HW_EVENTS)
+			return false;
+
 		if (num == counters)
 			event_group[counters++] = sibling;
 	}
 
-	return counters <= HNS3_PMU_MAX_HW_EVENTS;
+	return true;
 }
 
 static u32 hns3_pmu_get_filter_condition(struct perf_event *event)
Signature · 1 line
2.33.0
maintainer Reviewed-by on v1 2/3
#5 lore raw

Re: [PATCH 2/3] drivers/perf: hisi: hns3: Fix out-of-bound access when valid event group

Quoted 52 lines · Junhao He

The perf tool allows users to create event groups through following cmd [1], but the driver does not check whether the array index is out of bounds when writing data to the event_group array. If the number of events in an event_group is greater than HNS3_PMU_MAX_HW_EVENTS, the memory write overflow of event_group array occurs.

Add array index check to fix the possible array out of bounds violation, and return directly when write new events are written to array bounds.

There are 9 different events in an event_group.
[1] perf stat -e '{pmu/event1/, ... ,pmu/event9/}

Fixes 66637ab137b4 ("drivers/perf: hisi: add driver for HNS3 PMU")
---
  drivers/perf/hisilicon/hns3_pmu.c | 14 +++++++++++++-
  1 file changed, 13 insertions(+), 1 deletion(-)
drivers/perf/hisilicon/hns3_pmu.c +13 -1
index 16869bf5bf4c..cbdd53b0a034 100644
@@ -1085,15 +1085,27 @@ static bool hns3_pmu_validate_event_group(struct perf_event *event)
  			return false;
 
  		for (num = 0; num < counters; num++) {
+			/*
+			 * If we find a related event, then it's a valid group
+			 * since we don't need to allocate a new counter for it.
+			 */
  			if (hns3_pmu_cmp_event(event_group[num], sibling))
  				break;
  		}
 
+		/*
+		 * Otherwise it's a new event but if there's no available counter,
+		 * fail the check since we cannot schedule all the events in
+		 * the group simultaneously.
+		 */
+		if (num == HNS3_PMU_MAX_HW_EVENTS)
+			return false;
+
  		if (num == counters)
  			event_group[counters++] = sibling;
  	}
 
-	return counters <= HNS3_PMU_MAX_HW_EVENTS;
+	return true;
  }
 
  static u32 hns3_pmu_get_filter_condition(struct perf_event *event)
Acked-by on v1 2/3
#6 lore raw

Re: [PATCH 2/3] drivers/perf: hisi: hns3: Fix out-of-bound access when valid event group

In reply to Junhao He
Quoted 9 lines · On Thu, 25 Apr 2024 20:46:26 +0800 Junhao He <[email protected]> wrote:

The perf tool allows users to create event groups through following cmd [1], but the driver does not check whether the array index is out of bounds when writing data to the event_group array. If the number of events in an event_group is greater than HNS3_PMU_MAX_HW_EVENTS, the memory write overflow of event_group array occurs.

Add array index check to fix the possible array out of bounds violation, and return directly when write new events are written to array bounds.

Quoted 6 lines

There are 9 different events in an event_group.
[1] perf stat -e '{pmu/event1/, ... ,pmu/event9/}

Fixes 66637ab137b4 ("drivers/perf: hisi: add driver for HNS3 PMU")
Quoted 37 lines
---
 drivers/perf/hisilicon/hns3_pmu.c | 14 +++++++++++++-
 1 file changed, 13 insertions(+), 1 deletion(-)
drivers/perf/hisilicon/hns3_pmu.c +13 -1
index 16869bf5bf4c..cbdd53b0a034 100644
@@ -1085,15 +1085,27 @@ static bool hns3_pmu_validate_event_group(struct perf_event *event)
 			return false;
 
 		for (num = 0; num < counters; num++) {
+			/*
+			 * If we find a related event, then it's a valid group
+			 * since we don't need to allocate a new counter for it.
+			 */
 			if (hns3_pmu_cmp_event(event_group[num], sibling))
 				break;
 		}
 
+		/*
+		 * Otherwise it's a new event but if there's no available counter,
+		 * fail the check since we cannot schedule all the events in
+		 * the group simultaneously.
+		 */
+		if (num == HNS3_PMU_MAX_HW_EVENTS)
+			return false;
+
 		if (num == counters)
 			event_group[counters++] = sibling;
 	}
 
-	return counters <= HNS3_PMU_MAX_HW_EVENTS;
+	return true;
 }
 
 static u32 hns3_pmu_get_filter_condition(struct perf_event *event)
author
#7 lore raw

PATCH 3/3 drivers/perf: hisi: hns3: Actually use devm_add_action_or_reset()

In reply to Junhao He

From: Hao Chen <[email protected]>

pci_alloc_irq_vectors() allocates an irq vector. When devm_add_action() fails, the irq vector is not freed, which leads to a memory leak.

Replace the devm_add_action with devm_add_action_or_reset to ensure the irq vector can be destroyed when it fails.

Fixes 66637ab137b4 ("drivers/perf: hisi: add driver for HNS3 PMU")
---
 drivers/perf/hisilicon/hns3_pmu.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/perf/hisilicon/hns3_pmu.c +1 -1
index cbdd53b0a034..60062eaa342a 100644
@@ -1527,7 +1527,7 @@ static int hns3_pmu_irq_register(struct pci_dev *pdev,
 		return ret;
 	}
 
-	ret = devm_add_action(&pdev->dev, hns3_pmu_free_irq, pdev);
+	ret = devm_add_action_or_reset(&pdev->dev, hns3_pmu_free_irq, pdev);
 	if (ret) {
 		pci_err(pdev, "failed to add free irq action, ret = %d.\n", ret);
 		return ret;
Signature · 1 line
2.33.0
maintainer Reviewed-by on v1 3/3
#8 lore raw

Re: [PATCH 3/3] drivers/perf: hisi: hns3: Actually use devm_add_action_or_reset()

Quoted 28 lines · Junhao He

From: Hao Chen <[email protected]>

pci_alloc_irq_vectors() allocates an irq vector. When devm_add_action() fails, the irq vector is not freed, which leads to a memory leak.

Replace the devm_add_action with devm_add_action_or_reset to ensure the irq vector can be destroyed when it fails.

Fixes 66637ab137b4 ("drivers/perf: hisi: add driver for HNS3 PMU")
---
  drivers/perf/hisilicon/hns3_pmu.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
drivers/perf/hisilicon/hns3_pmu.c +1 -1
index cbdd53b0a034..60062eaa342a 100644
@@ -1527,7 +1527,7 @@ static int hns3_pmu_irq_register(struct pci_dev *pdev,
  		return ret;
  	}
 
-	ret = devm_add_action(&pdev->dev, hns3_pmu_free_irq, pdev);
+	ret = devm_add_action_or_reset(&pdev->dev, hns3_pmu_free_irq, pdev);
  	if (ret) {
  		pci_err(pdev, "failed to add free irq action, ret = %d.\n", ret);
  		return ret;
Acked-by on v1 3/3
#9 lore raw

Re: [PATCH 3/3] drivers/perf: hisi: hns3: Actually use devm_add_action_or_reset()

In reply to Junhao He
Quoted 11 lines · On Thu, 25 Apr 2024 20:46:27 +0800 Junhao He <[email protected]> wrote:

From: Hao Chen <[email protected]>

pci_alloc_irq_vectors() allocates an irq vector. When devm_add_action() fails, the irq vector is not freed, which leads to a memory leak.

Replace the devm_add_action with devm_add_action_or_reset to ensure the irq vector can be destroyed when it fails.

Fixes 66637ab137b4 ("drivers/perf: hisi: add driver for HNS3 PMU")

Thanks,

Quoted 17 lines
---
 drivers/perf/hisilicon/hns3_pmu.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/perf/hisilicon/hns3_pmu.c +1 -1
index cbdd53b0a034..60062eaa342a 100644
@@ -1527,7 +1527,7 @@ static int hns3_pmu_irq_register(struct pci_dev *pdev,
 		return ret;
 	}
 
-	ret = devm_add_action(&pdev->dev, hns3_pmu_free_irq, pdev);
+	ret = devm_add_action_or_reset(&pdev->dev, hns3_pmu_free_irq, pdev);
 	if (ret) {
 		pci_err(pdev, "failed to add free irq action, ret = %d.\n", ret);
 		return ret;
maintainer Applied
#10 lore raw
In reply to Junhao He
Quoted 9 lines · Junhao He

This patchset includes 3 bug fixes changes to hisi PMU:
- Fix out-of-bound access when valid event group in hns pmu
- Fixes the memory leak in hns pmu
- Fix out-of-bound access when valid event group in pcie pmu

Hao Chen (1):
drivers/perf: hisi: hns3: Actually use devm_add_action_or_reset()

[...]

Applied to will (for-next/perf), thanks!

[1/3] drivers/perf: hisi_pcie: Fix out-of-bound access when valid event group

      https://git.kernel.org/will/c/77fce82678ea
[2/3] drivers/perf: hisi: hns3: Fix out-of-bound access when valid event group
      https://git.kernel.org/will/c/81bdd60a3d1d
[3/3] drivers/perf: hisi: hns3: Actually use devm_add_action_or_reset()
      https://git.kernel.org/will/c/582c1aeee0a9

Cheers,

Signature · 5 lines
Will

https://fixes.arm64.dev
https://next.arm64.dev
https://will.arm64.dev