Re: [PATCH] iio: imu: st_lsm6dsx: Fix reading array out of bounds

[Stephan Gerhold <stephan@gerhold.net>]

On Thu, Apr 09, 2020 at 01:50:24PM +0200, Alexandre Bard wrote:
> Le 09.04.20 à 13:01, Stephan Gerhold a écrit :
> > On Thu, Apr 09, 2020 at 10:58:18AM +0200, Alexandre Bard wrote:
> >> Former code was iterating through all possible IDs whereas only a few
> >> per settings array are really available. Leading to several out of
> >> bounds readings.
> >>
> >> Line is now longer than 80 characters. But since it is a classic for
> >> loop I think it is better to keep it like this than splitting it.
> >>
> >> Signed-off-by: Alexandre Bard <alexandre.bard@netmodule.com>
> >> ---
> >>  drivers/iio/imu/st_lsm6dsx/st_lsm6dsx_core.c | 2 +-
> >>  1 file changed, 1 insertion(+), 1 deletion(-)
> >>
> >> diff --git a/drivers/iio/imu/st_lsm6dsx/st_lsm6dsx_core.c b/drivers/iio/imu/st_lsm6dsx/st_lsm6dsx_core.c
> >> index 84d219ae6aee..be8882ff30eb 100644
> >> --- a/drivers/iio/imu/st_lsm6dsx/st_lsm6dsx_core.c
> >> +++ b/drivers/iio/imu/st_lsm6dsx/st_lsm6dsx_core.c
> >> @@ -1350,7 +1350,7 @@ static int st_lsm6dsx_check_whoami(struct st_lsm6dsx_hw *hw, int id,
> >>  	int err, i, j, data;
> >>  
> >>  	for (i = 0; i < ARRAY_SIZE(st_lsm6dsx_sensor_settings); i++) {
> >> -		for (j = 0; j < ST_LSM6DSX_MAX_ID; j++) {
> >> +		for (j = 0; j < ARRAY_SIZE(st_lsm6dsx_sensor_settings[i].id); j++) {
> > id in st_lsm6dsx_settings is declared as:
> >
> > 	struct {
> > 		enum st_lsm6dsx_hw_id hw_id;
> > 		const char *name;
> > 	} id[ST_LSM6DSX_MAX_ID];
> >
> > so it's always ST_LSM6DSX_MAX_ID long
> > (additional entries are just zero-initialized).
> >
> > Isn't ARRAY_SIZE(st_lsm6dsx_sensor_settings[i].id) == ST_LSM6DSX_MAX_ID
> > in this case?
> Yes, you are right, I missed that. But there is still a problem :
> parsing 0-initialized fields can lead to a false positive when looking for the
> value ST_LSM6DS3_ID which is the first element of an enum. So either the enum
> must be patched to start at 1 or the length of valid ids in a settings must be
> retrieved somehow.
> 
> Or is there another way ? Or am I wrong ?

ST_LSM6DS3_ID was indeed broken, which is why I added a .name != NULL
check in commit fb4fbc8904e7 ("iio: imu: st_lsm6dsx: Fix selection of ST_LSM6DS3_ID").

.name is only set for properly initialized IDs, so this ensures that we
do not match any zero-initialized entries. :)

> >
> >>  			if (st_lsm6dsx_sensor_settings[i].id[j].name &&
> >>  			    id == st_lsm6dsx_sensor_settings[i].id[j].hw_id)
> >>  				break;
> >> -- 
> >> 2.20.1
> >>