On Wed, Feb 25, 2026 at 07:03:29PM +0100, Thorsten Blum wrote:
> Replace kcalloc() with kzalloc_objs() when allocating the endpoint
> address array to keep the size type-safe and match nearby allocations.
> Reformat ->busy_urbs allocation to a single line. No functional change.
>
> Signed-off-by: Thorsten Blum <[email protected]>
> ---
> drivers/most/most_usb.c | 6 ++----
> 1 file changed, 2 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/most/most_usb.c b/drivers/most/most_usb.c
> index d2c0875727a3..6437733afee0 100644
> --- a/drivers/most/most_usb.c
> +++ b/drivers/most/most_usb.c
> @@ -1009,13 +1009,11 @@ hdm_probe(struct usb_interface *interface, const
> struct usb_device_id *id)
> goto err_free_conf;
>
> mdev->iface.channel_vector = mdev->cap;
> - mdev->ep_address =
> - kcalloc(num_endpoints, sizeof(*mdev->ep_address), GFP_KERNEL);
> + mdev->ep_address = kzalloc_objs(*mdev->ep_address, num_endpoints);
This case was skipped because mdev->ep_address is "u8 *". This is a
pretty long way to go about allocating "num_endpoints"-many bytes.
Perhaps this should just be:
mdev->ep_address = kzalloc(num_endpoints, GFP_KERNEL);
Though maybe in keeping with all the other num_endpoints allocations,
your patch is fine as-is. I'd be nice to use __counted_by_ptr() here,
but the num_endpoints is in a sub-structure, which isn't supported yet.
Reviewed-by: Kees Cook <[email protected]>
--
Kees Cook
