Skip to contentSkip to conversation

platform/x86: intel_pmc_core: Fix uninitialized pmc/map in pmc_core_send_ltr_ignore

PATCH In review Purva Yeshi 2 participants 3 messages Apr 17, 2025 – Apr 18, 2025

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 platform/x86: intel_pmc_core: Fix uninitialized pmc/map in pmc_core_send_ltr_ignore

Fix Smatch-detected issue:

drivers/platform/x86/intel/pmc/core.c:501 pmc_core_send_ltr_ignore()
error: uninitialized symbol 'pmc'.

drivers/platform/x86/intel/pmc/core.c:501 pmc_core_send_ltr_ignore()
error: uninitialized symbol 'map'.

drivers/platform/x86/intel/pmc/core.c:501 pmc_core_send_ltr_ignore()
error: we previously assumed 'pmc' could be null (see line 479)

Prevents uninitialized symbol warnings detected by smatch.

Ensures map is not accessed if pmc is NULL, preventing dereferencing of uninitialized pointers

Add defensive check for pmc and map to catch any unexpected edge cases and ensure all required pointers are valid.

---
 drivers/platform/x86/intel/pmc/core.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)
drivers/platform/x86/intel/pmc/core.c +6 -3
index 7a1d11f2914f..e674b940e29e 100644
@@ -462,8 +462,8 @@ DEFINE_SHOW_ATTRIBUTE(pmc_core_pll);
 
 int pmc_core_send_ltr_ignore(struct pmc_dev *pmcdev, u32 value, int ignore)
 {
-	struct pmc *pmc;
-	const struct pmc_reg_map *map;
+	struct pmc *pmc = NULL;
+	const struct pmc_reg_map *map = NULL;
 	u32 reg;
 	unsigned int pmc_index;
 	int ltr_index;
@@ -480,6 +480,9 @@ int pmc_core_send_ltr_ignore(struct pmc_dev *pmcdev, u32 value, int ignore)
 			continue;
 
 		map = pmc->map;
+		if (!map)
+			continue;
+
 		if (ltr_index <= map->ltr_ignore_max)
 			break;
 
@@ -491,7 +494,7 @@ int pmc_core_send_ltr_ignore(struct pmc_dev *pmcdev, u32 value, int ignore)
 		ltr_index = ltr_index - (map->ltr_ignore_max + 2) - 1;
 	}
 
-	if (pmc_index >= ARRAY_SIZE(pmcdev->pmcs) || ltr_index < 0)
+	if (pmc_index >= ARRAY_SIZE(pmcdev->pmcs) || ltr_index < 0 || !pmc || !map)
 		return -EINVAL;
 
 	pr_debug("ltr_ignore for pmc%d: ltr_index:%d\n", pmc_index, ltr_index);
Signature · 1 line
2.34.1
maintainer
#1 lore raw
Quoted 41 lines · Purva Yeshi

Fix Smatch-detected issue:

drivers/platform/x86/intel/pmc/core.c:501 pmc_core_send_ltr_ignore()
error: uninitialized symbol 'pmc'.

drivers/platform/x86/intel/pmc/core.c:501 pmc_core_send_ltr_ignore()
error: uninitialized symbol 'map'.

drivers/platform/x86/intel/pmc/core.c:501 pmc_core_send_ltr_ignore()
error: we previously assumed 'pmc' could be null (see line 479)

Prevents uninitialized symbol warnings detected by smatch.

Ensures map is not accessed if pmc is NULL, preventing dereferencing of uninitialized pointers

Add defensive check for pmc and map to catch any unexpected edge cases and ensure all required pointers are valid.

---
 drivers/platform/x86/intel/pmc/core.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)
drivers/platform/x86/intel/pmc/core.c +2 -2
index 7a1d11f2914f..e674b940e29e 100644
@@ -462,8 +462,8 @@ DEFINE_SHOW_ATTRIBUTE(pmc_core_pll);
 
 int pmc_core_send_ltr_ignore(struct pmc_dev *pmcdev, u32 value, int ignore)
 {
-	struct pmc *pmc;
-	const struct pmc_reg_map *map;
+	struct pmc *pmc = NULL;
+	const struct pmc_reg_map *map = NULL;
 	u32 reg;
 	unsigned int pmc_index;
 	int ltr_index;

drivers/platform/x86/intel/pmc/core.c int pmc_core_send_ltr_ignore(struct pmc_dev *pmcdev, u32 value, int ignore) · line 484

Quoted 5 lines
drivers/platform/x86/intel/pmc/core.c +2 -0
@@ -480,6 +480,9 @@ int pmc_core_send_ltr_ignore(struct pmc_dev *pmcdev, u32 value, int ignore)
 			continue;
 
 		map = pmc->map;
+		if (!map)
+			continue;

How can this happen?? If pmc is created, it should have a valid ->map AFAICT. Did you even read that code at all???

drivers/platform/x86/intel/pmc/core.c int pmc_core_send_ltr_ignore(struct pmc_dev *pmcdev, u32 value, int ignore) · line 497

Quoted 10 lines
drivers/platform/x86/intel/pmc/core.c +2 -1
+
 		if (ltr_index <= map->ltr_ignore_max)
 			break;
 
@@ -491,7 +494,7 @@ int pmc_core_send_ltr_ignore(struct pmc_dev *pmcdev, u32 value, int ignore)
 		ltr_index = ltr_index - (map->ltr_ignore_max + 2) - 1;
 	}
 
-	if (pmc_index >= ARRAY_SIZE(pmcdev->pmcs) || ltr_index < 0)
+	if (pmc_index >= ARRAY_SIZE(pmcdev->pmcs) || ltr_index < 0 || !pmc || !map)

What are the situations pmc_index >= ARRAY_SIZE(pmcdev->pmcs) check didn't catch where these new checks do something useful??

Lots of noise but little real substance in this patch?

Signature · 1 line
 i.
#2 lore raw
Quoted 44 lines · Ilpo Järvinen
Quoted 41 lines · Purva Yeshi

Fix Smatch-detected issue:

drivers/platform/x86/intel/pmc/core.c:501 pmc_core_send_ltr_ignore()
error: uninitialized symbol 'pmc'.

drivers/platform/x86/intel/pmc/core.c:501 pmc_core_send_ltr_ignore()
error: uninitialized symbol 'map'.

drivers/platform/x86/intel/pmc/core.c:501 pmc_core_send_ltr_ignore()
error: we previously assumed 'pmc' could be null (see line 479)

Prevents uninitialized symbol warnings detected by smatch.

Ensures map is not accessed if pmc is NULL, preventing dereferencing of uninitialized pointers

Add defensive check for pmc and map to catch any unexpected edge cases and ensure all required pointers are valid.

---
  drivers/platform/x86/intel/pmc/core.c | 9 ++++++---
  1 file changed, 6 insertions(+), 3 deletions(-)
drivers/platform/x86/intel/pmc/core.c +2 -2
index 7a1d11f2914f..e674b940e29e 100644
@@ -462,8 +462,8 @@ DEFINE_SHOW_ATTRIBUTE(pmc_core_pll);
 
  int pmc_core_send_ltr_ignore(struct pmc_dev *pmcdev, u32 value, int ignore)
  {
-	struct pmc *pmc;
-	const struct pmc_reg_map *map;
+	struct pmc *pmc = NULL;
+	const struct pmc_reg_map *map = NULL;
  	u32 reg;
  	unsigned int pmc_index;
  	int ltr_index;
@@ -480,6 +480,9 @@ int pmc_core_send_ltr_ignore(struct pmc_dev *pmcdev, u32 value, int ignore)
Quoted 8 lines

drivers/platform/x86/intel/pmc/core.c int pmc_core_send_ltr_ignore(struct pmc_dev *pmcdev, u32 value, int ignore) · line 484

Quoted 5 lines
drivers/platform/x86/intel/pmc/core.c +2 -0
  			continue;
 
  		map = pmc->map;
+		if (!map)
+			continue;

How can this happen?? If pmc is created, it should have a valid ->map AFAICT. Did you even read that code at all???

Hi,

Thanks for the feedback.

Yes, I did read through the code and I understand your point.

The motivation behind the patch was a Smatch warning about possible uninitialized use of map and pmc, even though they are logically guarded. I now see that these checks may not be necessary given the existing control flow.

Quoted 9 lines
Quoted 9 lines
drivers/platform/x86/intel/pmc/core.c +1 -0
+
  		if (ltr_index <= map->ltr_ignore_max)
  			break;
 
@@ -491,7 +494,7 @@ int pmc_core_send_ltr_ignore(struct pmc_dev *pmcdev, u32 value, int ignore)
  		ltr_index = ltr_index - (map->ltr_ignore_max + 2) - 1;
  	}
 
Quoted 7 lines

drivers/platform/x86/intel/pmc/core.c int pmc_core_send_ltr_ignore(struct pmc_dev *pmcdev, u32 value, int ignore) · line 497

Quoted 2 lines
drivers/platform/x86/intel/pmc/core.c +1 -1
-	if (pmc_index >= ARRAY_SIZE(pmcdev->pmcs) || ltr_index < 0)
+	if (pmc_index >= ARRAY_SIZE(pmcdev->pmcs) || ltr_index < 0 || !pmc || !map)

What are the situations pmc_index >= ARRAY_SIZE(pmcdev->pmcs) check didn't catch where these new checks do something useful??

Lots of noise but little real substance in this patch?

You're right, if pmc is non-NULL, then map should also be valid, and the bounds check on pmc_index already prevents out-of-bounds access. Adding further checks might just add noise.

I'll drop the patch unless there's a cleaner way to restructure the logic to make Smatch silence without redundant checks.

Thanks again for the clarification!

Best regards,
Purva

Quoted 1 line