Skip to contentSkip to conversation

regulator: rn5t618: Fix out of bounds array access

PATCH RESEND In review Axel Lin 2 participants 5 messages Apr 15, 2017 – Apr 16, 2017

Subsystems (2)
lore
Share & tools
lore

Axel Lin

author
#0 lore raw

PATCH RESEND 1/1 regulator: rn5t618: Fix out of bounds array access

The commit "regulator: rn5t618: Add RN5T567 PMIC support" added RN5T618_DCDC4 to the enum, then RN5T618_REG_NUM is also changed. So for rn5t618, there is out of bounds array access when checking regulators[i].name in the for loop.

The number of regulators is different for rn5t567 and rn5t618, so we had better remove RN5T618_REG_NUM and get the correct num_regulators during probe instead.

Fixes ed6d362d8dbc ("regulator: rn5t618: Add RN5T567 PMIC support")

---
RESEND: Correct subject line (remove double Fix)

 drivers/regulator/rn5t618-regulator.c | 8 ++++----
 include/linux/mfd/rn5t618.h           | 1 -
 2 files changed, 4 insertions(+), 5 deletions(-)
drivers/regulator/rn5t618-regulator.c +4 -4
index 8d2819e..0c09143 100644
@@ -85,14 +85,17 @@ static int rn5t618_regulator_probe(struct platform_device *pdev)
 	struct regulator_config config = { };
 	struct regulator_dev *rdev;
 	struct regulator_desc *regulators;
+	int num_regulators;
 	int i;
 
 	switch (rn5t618->variant) {
 	case RN5T567:
 		regulators = rn5t567_regulators;
+		num_regulators = ARRAY_SIZE(rn5t567_regulators);
 		break;
 	case RN5T618:
 		regulators = rn5t618_regulators;
+		num_regulators = ARRAY_SIZE(rn5t618_regulators);
 		break;
 	default:
 		return -EINVAL;
@@ -101,10 +104,7 @@ static int rn5t618_regulator_probe(struct platform_device *pdev)
 	config.dev = pdev->dev.parent;
 	config.regmap = rn5t618->regmap;
 
-	for (i = 0; i < RN5T618_REG_NUM; i++) {
-		if (!regulators[i].name)
-			continue;
-
+	for (i = 0; i < num_regulators; i++) {
 		rdev = devm_regulator_register(&pdev->dev,
 					       &regulators[i],
 					       &config);
include/linux/mfd/rn5t618.h +0 -1
index e5a6cde..d7b3155 100644
@@ -233,7 +233,6 @@ enum {
 	RN5T618_LDO5,
 	RN5T618_LDORTC1,
 	RN5T618_LDORTC2,
-	RN5T618_REG_NUM,
 };
 
 enum {
Signature · 1 line
2.9.3
#1 lore raw
Quoted 4 lines · Axel Lin

The commit "regulator: rn5t618: Add RN5T567 PMIC support" added RN5T618_DCDC4 to the enum, then RN5T618_REG_NUM is also changed. So for rn5t618, there is out of bounds array access when checking regulators[i].name in the for loop.

I use designated initializers ([RN5T618_##rid] = {..), which guarantee that the non initialized elements are zero. The highest element LDORTC2 is defined, hence the length of the array should be RN5T618_REG_NUM.

See also
https://gcc.gnu.org/onlinedocs/gcc/Designated-Inits.html

--
Stefan

Quoted 63 lines

The number of regulators is different for rn5t567 and rn5t618, so we had better remove RN5T618_REG_NUM and get the correct num_regulators during probe instead.

Fixes ed6d362d8dbc ("regulator: rn5t618: Add RN5T567 PMIC support")

---
RESEND: Correct subject line (remove double Fix)

 drivers/regulator/rn5t618-regulator.c | 8 ++++----
 include/linux/mfd/rn5t618.h           | 1 -
 2 files changed, 4 insertions(+), 5 deletions(-)
drivers/regulator/rn5t618-regulator.c +3 -0
index 8d2819e..0c09143 100644
@@ -85,14 +85,17 @@ static int rn5t618_regulator_probe(struct
 platform_device *pdev)
 	struct regulator_config config = { };
 	struct regulator_dev *rdev;
 	struct regulator_desc *regulators;
+	int num_regulators;
 	int i;
 
 	switch (rn5t618->variant) {
 	case RN5T567:
 		regulators = rn5t567_regulators;
+		num_regulators = ARRAY_SIZE(rn5t567_regulators);
 		break;
 	case RN5T618:
 		regulators = rn5t618_regulators;
+		num_regulators = ARRAY_SIZE(rn5t618_regulators);
 		break;
 	default:
drivers/regulator/rn5t618-regulator.c +1 -4
 		return -EINVAL;
@@ -101,10 +104,7 @@ static int rn5t618_regulator_probe(struct
 platform_device *pdev)
 	config.dev = pdev->dev.parent;
 	config.regmap = rn5t618->regmap;
 
-	for (i = 0; i < RN5T618_REG_NUM; i++) {
-		if (!regulators[i].name)
-			continue;
-
+	for (i = 0; i < num_regulators; i++) {
 		rdev = devm_regulator_register(&pdev->dev,
 					       &regulators[i],
drivers/regulator/rn5t618-regulator.c +0 -0
 					       &config);
include/linux/mfd/rn5t618.h +0 -1
index e5a6cde..d7b3155 100644
@@ -233,7 +233,6 @@ enum {
 	RN5T618_LDO5,
 	RN5T618_LDORTC1,
 	RN5T618_LDORTC2,
-	RN5T618_REG_NUM,
 };
 
 enum {
author
#2 lore raw

2017-04-16 0:53 GMT+08:00 Stefan Agner <[email protected]>:

Quoted 9 lines · Axel Lin
Quoted 4 lines

The commit "regulator: rn5t618: Add RN5T567 PMIC support" added RN5T618_DCDC4 to the enum, then RN5T618_REG_NUM is also changed. So for rn5t618, there is out of bounds array access when checking regulators[i].name in the for loop.

I use designated initializers ([RN5T618_##rid] = {..), which guarantee that the non initialized elements are zero. The highest element LDORTC2 is defined, hence the length of the array should be RN5T618_REG_NUM.

ok, I missed that. Then current code is fine.
Though the meaing of RN5T618_REG_NUM seems misleading to me as different variant has differnt number of regulators.

Thanks for the review,
Axel

author
#3 lore raw

2017-04-16 9:12 GMT+08:00 Axel Lin <[email protected]>:

Quoted 7 lines

2017-04-16 0:53 GMT+08:00 Stefan Agner <[email protected]>:

Quoted 6 lines · Axel Lin
Quoted 4 lines

The commit "regulator: rn5t618: Add RN5T567 PMIC support" added RN5T618_DCDC4 to the enum, then RN5T618_REG_NUM is also changed. So for rn5t618, there is out of bounds array access when checking regulators[i].name in the for loop.

Quoted 5 lines
Quoted 3 lines

I use designated initializers ([RN5T618_##rid] = {..), which guarantee that the non initialized elements are zero. The highest element LDORTC2 is defined, hence the length of the array should be RN5T618_REG_NUM.

ok, I missed that. Then current code is fine.

I just realize my patch is wrong due to the use of designated initializers.

Regards,
Axel

#4 lore raw
In reply to Axel Lin
Quoted 7 lines · Axel Lin

2017-04-16 0:53 GMT+08:00 Stefan Agner <[email protected]>:

Quoted 6 lines · Axel Lin
Quoted 4 lines

The commit "regulator: rn5t618: Add RN5T567 PMIC support" added RN5T618_DCDC4 to the enum, then RN5T618_REG_NUM is also changed. So for rn5t618, there is out of bounds array access when checking regulators[i].name in the for loop.

Quoted 7 lines
Quoted 3 lines

I use designated initializers ([RN5T618_##rid] = {..), which guarantee that the non initialized elements are zero. The highest element LDORTC2 is defined, hence the length of the array should be RN5T618_REG_NUM.

ok, I missed that. Then current code is fine.
Though the meaing of RN5T618_REG_NUM seems misleading to me as different variant has differnt number of regulators.

Yeah I admit the code is somewhat unobvious as it is now. But it allowed me to add RN5T567 support without changing the existing array and the preprocessor macro.

--
Stefan