Skip to contentSkip to conversation

iio: ssp_sensors: avoid uninitialized variable usage

PATCH In review Arnd Bergmann 3 participants 3 messages Mar 22, 2019 – Mar 24, 2019 IIO

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 iio: ssp_sensors: avoid uninitialized variable usage

clang points out that 'calculated_time' is only sometimes initialized here, which leads to incorrect data being passed into another function:

drivers/iio/common/ssp_sensors/ssp_iio.c:95:6: error: variable 'calculated_time' is used uninitialized whenever 'if' condition is false [-Werror,-Wsometimes-uninitialized]

        if (indio_dev->scan_timestamp) {
            ^~~~~~~~~~~~~~~~~~~~~~~~~
drivers/iio/common/ssp_sensors/ssp_iio.c:102:9: note: uninitialized use occurs here
                                                  calculated_time);
                                                  ^~~~~~~~~~~~~~~
drivers/iio/common/ssp_sensors/ssp_iio.c:95:2: note: remove the 'if' if its condition is always true
        if (indio_dev->scan_timestamp) {
        ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/iio/common/ssp_sensors/ssp_iio.c:84:25: note: initialize the variable 'calculated_time' to silence this warning
        int64_t calculated_time;
                               ^

The data is subsequently ignored by iio_push_to_buffers_with_timestamp(), but the warning still feels legitimate and to work around it, we can initialize the time in the other case.

---
 drivers/iio/common/ssp_sensors/ssp_iio.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/iio/common/ssp_sensors/ssp_iio.c +2 -0
index 645f2e3975db..81e8f4844c90 100644
@@ -96,6 +96,8 @@ int ssp_common_process_data(struct iio_dev *indio_dev, void *buf,
 		memcpy(&time, &((char *)buf)[len], SSP_TIME_SIZE);
 		calculated_time =
 			timestamp + (int64_t)le32_to_cpu(time) * 1000000;
+	} else {
+		calculated_time = 0;
 	}
 
 	return iio_push_to_buffers_with_timestamp(indio_dev, spd->buffer,
Signature · 1 line
2.20.0
Reviewed-by on v1 1/1
#1 lore raw
Quoted 42 lines · Arnd Bergmann

clang points out that 'calculated_time' is only sometimes initialized here, which leads to incorrect data being passed into another function:

drivers/iio/common/ssp_sensors/ssp_iio.c:95:6: error: variable 'calculated_time' is used uninitialized whenever 'if' condition is false [-Werror,-Wsometimes-uninitialized]

        if (indio_dev->scan_timestamp) {
            ^~~~~~~~~~~~~~~~~~~~~~~~~
drivers/iio/common/ssp_sensors/ssp_iio.c:102:9: note: uninitialized use occurs here
                                                  calculated_time);
                                                  ^~~~~~~~~~~~~~~
drivers/iio/common/ssp_sensors/ssp_iio.c:95:2: note: remove the 'if' if its condition is always true
        if (indio_dev->scan_timestamp) {
        ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/iio/common/ssp_sensors/ssp_iio.c:84:25: note: initialize the variable 'calculated_time' to silence this warning
        int64_t calculated_time;
                               ^

The data is subsequently ignored by iio_push_to_buffers_with_timestamp(), but the warning still feels legitimate and to work around it, we can initialize the time in the other case.

---
 drivers/iio/common/ssp_sensors/ssp_iio.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/iio/common/ssp_sensors/ssp_iio.c +2 -0
index 645f2e3975db..81e8f4844c90 100644
@@ -96,6 +96,8 @@ int ssp_common_process_data(struct iio_dev *indio_dev, void *buf,
 		memcpy(&time, &((char *)buf)[len], SSP_TIME_SIZE);
 		calculated_time =
 			timestamp + (int64_t)le32_to_cpu(time) * 1000000;
+	} else {
+		calculated_time = 0;
 	}
 
 	return iio_push_to_buffers_with_timestamp(indio_dev, spd->buffer,
Signature · 1 line
2.20.0

I sent a similar change, which is sitting in Jonathan's testing branch:

https://git.kernel.org/pub/scm/linux/kernel/git/jic23/iio.git/commit/?id=0643039b4fee4aa54a233ead15dc0b2286f059d7

You made a good point previously that initializing the variable at the beginning of a function may not always be the best choice. I don't have a personal preference for which patch stays around so:

Just in case.

Nathan

maintainer
#2 lore raw
Quoted 52 lines · On Fri, 22 Mar 2019 08:01:19 -0700 Nathan Chancellor…
Quoted 42 lines · Arnd Bergmann

clang points out that 'calculated_time' is only sometimes initialized here, which leads to incorrect data being passed into another function:

drivers/iio/common/ssp_sensors/ssp_iio.c:95:6: error: variable 'calculated_time' is used uninitialized whenever 'if' condition is false [-Werror,-Wsometimes-uninitialized]

        if (indio_dev->scan_timestamp) {
            ^~~~~~~~~~~~~~~~~~~~~~~~~
drivers/iio/common/ssp_sensors/ssp_iio.c:102:9: note: uninitialized use occurs here
                                                  calculated_time);
                                                  ^~~~~~~~~~~~~~~
drivers/iio/common/ssp_sensors/ssp_iio.c:95:2: note: remove the 'if' if its condition is always true
        if (indio_dev->scan_timestamp) {
        ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/iio/common/ssp_sensors/ssp_iio.c:84:25: note: initialize the variable 'calculated_time' to silence this warning
        int64_t calculated_time;
                               ^

The data is subsequently ignored by iio_push_to_buffers_with_timestamp(), but the warning still feels legitimate and to work around it, we can initialize the time in the other case.

---
 drivers/iio/common/ssp_sensors/ssp_iio.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/iio/common/ssp_sensors/ssp_iio.c +2 -0
index 645f2e3975db..81e8f4844c90 100644
@@ -96,6 +96,8 @@ int ssp_common_process_data(struct iio_dev *indio_dev, void *buf,
 		memcpy(&time, &((char *)buf)[len], SSP_TIME_SIZE);
 		calculated_time =
 			timestamp + (int64_t)le32_to_cpu(time) * 1000000;
+	} else {
+		calculated_time = 0;
 	}
 
 	return iio_push_to_buffers_with_timestamp(indio_dev, spd->buffer,
Signature · 1 line
2.20.0

I sent a similar change, which is sitting in Jonathan's testing branch:

https://git.kernel.org/pub/scm/linux/kernel/git/jic23/iio.git/commit/?id=0643039b4fee4aa54a233ead15dc0b2286f059d7

You made a good point previously that initializing the variable at the beginning of a function may not always be the best choice. I don't have a personal preference for which patch stays around so:

Quoted 5 lines

Just in case.

Nathan

I'll stick to the original, mostly to avoid unnecessary churn. Apologies for the delay in pull requests making these more generally visible. Always takes me a while to get one out after RC1. Would imagine these will go to Greg and hence linux-next sometime in the next week.

Jonathan