Skip to contentSkip to conversation

leds: lp5523: fix out-of-bounds bug in lp5523_selftest()

PATCH Applied Maarten Zanders 2 participants 3 messages Oct 28, 2022 – Nov 2, 2022 LED

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 leds: lp5523: fix out-of-bounds bug in lp5523_selftest()

When not all LED channels of the led chip are configured, the sysfs selftest functionality gives erroneous results and tries to test all channels of the chip.
There is a potential for LED overcurrent conditions since the test current will be set to values from out-of-bound regions.

It is wrong to use pdata->led_config[i].led_current to skip absent channels as led_config[] only contains the configured LED channels.

Instead of iterating over all the physical channels of the device, loop over the available LED configurations and use led->chan_nr to access the correct i2c registers. Keep the zero-check for the LED current as existing users might depend on this to disable a channel.

---
 drivers/leds/leds-lp5523.c | 27 +++++++++++++++------------
 1 file changed, 15 insertions(+), 12 deletions(-)
drivers/leds/leds-lp5523.c +15 -12
index 369d40b0b65b..e08e3de1428d 100644
@@ -581,8 +581,8 @@ static ssize_t lp5523_selftest(struct device *dev,
 	struct lp55xx_led *led = i2c_get_clientdata(to_i2c_client(dev));
 	struct lp55xx_chip *chip = led->chip;
 	struct lp55xx_platform_data *pdata = chip->pdata;
-	int i, ret, pos = 0;
-	u8 status, adc, vdd;
+	int ret, pos = 0;
+	u8 status, adc, vdd, i;
 
 	mutex_lock(&chip->lock);
 
@@ -612,20 +612,21 @@ static ssize_t lp5523_selftest(struct device *dev,
 
 	vdd--;	/* There may be some fluctuation in measurement */
 
-	for (i = 0; i < LP5523_MAX_LEDS; i++) {
-		/* Skip non-existing channels */
+	for (i = 0; i < pdata->num_channels; i++) {
+		/* Skip disabled channels */
 		if (pdata->led_config[i].led_current == 0)
 			continue;
 
 		/* Set default current */
-		lp55xx_write(chip, LP5523_REG_LED_CURRENT_BASE + i,
+		lp55xx_write(chip, LP5523_REG_LED_CURRENT_BASE + led->chan_nr,
 			pdata->led_config[i].led_current);
 
-		lp55xx_write(chip, LP5523_REG_LED_PWM_BASE + i, 0xff);
+		lp55xx_write(chip, LP5523_REG_LED_PWM_BASE + led->chan_nr,
+			     0xff);
 		/* let current stabilize 2 - 4ms before measurements start */
 		usleep_range(2000, 4000);
 		lp55xx_write(chip, LP5523_REG_LED_TEST_CTRL,
-			     LP5523_EN_LEDTEST | i);
+			     LP5523_EN_LEDTEST | led->chan_nr);
 		/* ADC conversion time is 2.7 ms typically */
 		usleep_range(3000, 6000);
 		ret = lp55xx_read(chip, LP5523_REG_STATUS, &status);
@@ -633,20 +634,22 @@ static ssize_t lp5523_selftest(struct device *dev,
 			goto fail;
 
 		if (!(status & LP5523_LEDTEST_DONE))
-			usleep_range(3000, 6000);/* Was not ready. Wait. */
+			usleep_range(3000, 6000); /* Was not ready. Wait. */
 
 		ret = lp55xx_read(chip, LP5523_REG_LED_TEST_ADC, &adc);
 		if (ret < 0)
 			goto fail;
 
 		if (adc >= vdd || adc < LP5523_ADC_SHORTCIRC_LIM)
-			pos += sprintf(buf + pos, "LED %d FAIL\n", i);
+			pos += sprintf(buf + pos, "LED %d FAIL\n",
+				       led->chan_nr);
 
-		lp55xx_write(chip, LP5523_REG_LED_PWM_BASE + i, 0x00);
+		lp55xx_write(chip, LP5523_REG_LED_PWM_BASE + led->chan_nr,
+			     0x00);
 
 		/* Restore current */
-		lp55xx_write(chip, LP5523_REG_LED_CURRENT_BASE + i,
-			led->led_current);
+		lp55xx_write(chip, LP5523_REG_LED_CURRENT_BASE + led->chan_nr,
+			     led->led_current);
 		led++;
 	}
 	if (pos == 0)
Signature · 1 line
2.37.3
Applied
#1 lore raw

Hi!

Quoted 6 lines

When not all LED channels of the led chip are configured, the sysfs selftest functionality gives erroneous results and tries to test all channels of the chip.
There is a potential for LED overcurrent conditions since the test current will be set to values from out-of-bound regions.

Quoted 7 lines

It is wrong to use pdata->led_config[i].led_current to skip absent channels as led_config[] only contains the configured LED channels.

Instead of iterating over all the physical channels of the device, loop over the available LED configurations and use led->chan_nr to access the correct i2c registers. Keep the zero-check for the LED current as existing users might depend on this to disable a channel.

Thanks, applied, I'll push eventually.

I wonder, should we do these kind of actions on attribute _read_? I'd preffer some kind of write action to trigger this.

Best regards,
Pavel

Quoted 78 lines
---
 drivers/leds/leds-lp5523.c | 27 +++++++++++++++------------
 1 file changed, 15 insertions(+), 12 deletions(-)
drivers/leds/leds-lp5523.c +15 -12
index 369d40b0b65b..e08e3de1428d 100644
@@ -581,8 +581,8 @@ static ssize_t lp5523_selftest(struct device *dev,
 	struct lp55xx_led *led = i2c_get_clientdata(to_i2c_client(dev));
 	struct lp55xx_chip *chip = led->chip;
 	struct lp55xx_platform_data *pdata = chip->pdata;
-	int i, ret, pos = 0;
-	u8 status, adc, vdd;
+	int ret, pos = 0;
+	u8 status, adc, vdd, i;
 
 	mutex_lock(&chip->lock);
 
@@ -612,20 +612,21 @@ static ssize_t lp5523_selftest(struct device *dev,
 
 	vdd--;	/* There may be some fluctuation in measurement */
 
-	for (i = 0; i < LP5523_MAX_LEDS; i++) {
-		/* Skip non-existing channels */
+	for (i = 0; i < pdata->num_channels; i++) {
+		/* Skip disabled channels */
 		if (pdata->led_config[i].led_current == 0)
 			continue;
 
 		/* Set default current */
-		lp55xx_write(chip, LP5523_REG_LED_CURRENT_BASE + i,
+		lp55xx_write(chip, LP5523_REG_LED_CURRENT_BASE + led->chan_nr,
 			pdata->led_config[i].led_current);
 
-		lp55xx_write(chip, LP5523_REG_LED_PWM_BASE + i, 0xff);
+		lp55xx_write(chip, LP5523_REG_LED_PWM_BASE + led->chan_nr,
+			     0xff);
 		/* let current stabilize 2 - 4ms before measurements start */
 		usleep_range(2000, 4000);
 		lp55xx_write(chip, LP5523_REG_LED_TEST_CTRL,
-			     LP5523_EN_LEDTEST | i);
+			     LP5523_EN_LEDTEST | led->chan_nr);
 		/* ADC conversion time is 2.7 ms typically */
 		usleep_range(3000, 6000);
 		ret = lp55xx_read(chip, LP5523_REG_STATUS, &status);
@@ -633,20 +634,22 @@ static ssize_t lp5523_selftest(struct device *dev,
 			goto fail;
 
 		if (!(status & LP5523_LEDTEST_DONE))
-			usleep_range(3000, 6000);/* Was not ready. Wait. */
+			usleep_range(3000, 6000); /* Was not ready. Wait. */
 
 		ret = lp55xx_read(chip, LP5523_REG_LED_TEST_ADC, &adc);
 		if (ret < 0)
 			goto fail;
 
 		if (adc >= vdd || adc < LP5523_ADC_SHORTCIRC_LIM)
-			pos += sprintf(buf + pos, "LED %d FAIL\n", i);
+			pos += sprintf(buf + pos, "LED %d FAIL\n",
+				       led->chan_nr);
 
-		lp55xx_write(chip, LP5523_REG_LED_PWM_BASE + i, 0x00);
+		lp55xx_write(chip, LP5523_REG_LED_PWM_BASE + led->chan_nr,
+			     0x00);
 
 		/* Restore current */
-		lp55xx_write(chip, LP5523_REG_LED_CURRENT_BASE + i,
-			led->led_current);
+		lp55xx_write(chip, LP5523_REG_LED_CURRENT_BASE + led->chan_nr,
+			     led->led_current);
 		led++;
 	}
 	if (pos == 0)
Signature · 1 line
2.37.3
Signature · 1 line
People of Russia, stop Putin before his war on Ukraine escalates.
#2 lore raw

Hi Pavel,

Quoted 3 lines · Pavel Machek

I wonder, should we do these kind of actions on attribute _read_? I'd preffer some kind of write action to trigger this.

Yes I agree that doing these kind of things on a file read feels somehow strange. Strictly speaking, I guess you would implement an IOCTL to initiate the action and then perform a read to get the result. But that would needlessly complicate things on both ends? I'll keep it in mind and when I encounter similar interfaces in other drivers get back to this if needed.