These can only be accessed with CAP_SYS_ADMIN so it's not a critical
security issue.  The problem is that "page" is controlled by the user in
the ioctl().  The test to see if the bit is set in state->info->page_mask
is not sufficient because "page" can be very high and shift wrap around
to a bit which is set.

Signed-off-by: Dan Carpenter <dan.carpen...@oracle.com>

diff --git a/drivers/media/i2c/adv7604.c b/drivers/media/i2c/adv7604.c
index 660bacb8f7d9..8c633b8f30e7 100644
--- a/drivers/media/i2c/adv7604.c
+++ b/drivers/media/i2c/adv7604.c
@@ -618,7 +618,7 @@ static int adv76xx_read_reg(struct v4l2_subdev *sd, 
unsigned int reg)
        unsigned int val;
        int err;
 
-       if (!(BIT(page) & state->info->page_mask))
+       if (page >= ADV76XX_PAGE_MAX || !(BIT(page) & state->info->page_mask))
                return -EINVAL;
 
        reg &= 0xff;
@@ -633,7 +633,7 @@ static int adv76xx_write_reg(struct v4l2_subdev *sd, 
unsigned int reg, u8 val)
        struct adv76xx_state *state = to_state(sd);
        unsigned int page = reg >> 8;
 
-       if (!(BIT(page) & state->info->page_mask))
+       if (page >= ADV76XX_PAGE_MAX || !(BIT(page) & state->info->page_mask))
                return -EINVAL;
 
        reg &= 0xff;

Reply via email to