Re: [PATCH] [v2] Bluetooth: avoid memcmp() out of bounds warning

[Kees Cook]

On Mon, Oct 09, 2023 at 10:31:31PM +0200, Arnd Bergmann wrote:
> From: Arnd Bergmann <arnd@xxxxxxxx>
> 
> bacmp() is a wrapper around memcpy(), which contain compile-time
> checks for buffer overflow. Since the hci_conn_request_evt() also calls
> bt_dev_dbg() with an implicit NULL pointer check, the compiler is now
> aware of a case where 'hdev' is NULL and treats this as meaning that
> zero bytes are available:
> 
> In file included from net/bluetooth/hci_event.c:32:
> In function 'bacmp',
>     inlined from 'hci_conn_request_evt' at net/bluetooth/hci_event.c:3276:7:
> include/net/bluetooth/bluetooth.h:364:16: error: 'memcmp' specified bound 6 exceeds source size 0 [-Werror=stringop-overread]
>   364 |         return memcmp(ba1, ba2, sizeof(bdaddr_t));
>       |                ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> 
> Add another NULL pointer check before the bacmp() to ensure the compiler
> understands the code flow enough to not warn about it.  Since the patch
> that introduced the warning is marked for stable backports, this one
> should also go that way to avoid introducing build regressions.
> 
> Fixes: d70e44fef8621 ("Bluetooth: Reject connection with the device which has same BD_ADDR")
> Cc: Kees Cook <keescook@xxxxxxxxxxxx>
> Cc: "Lee, Chun-Yi" <jlee@xxxxxxxx>
> Cc: Luiz Augusto von Dentz <luiz.von.dentz@xxxxxxxxx>
> Cc: Marcel Holtmann <marcel@xxxxxxxxxxxx>
> Cc: stable@xxxxxxxxxxxxxxx
> Signed-off-by: Arnd Bergmann <arnd@xxxxxxxx>

A weird side-effect of the NULL check, but not unreasonable. :)

Reviewed-by: Kees Cook <keescook@xxxxxxxxxxxx>

-- 
Kees Cook