On Thu, Jul 12, 2018 at 10:30 PM, Kees Cook <keescook@xxxxxxxxxxxx> wrote:
> On Thu, Jul 12, 2018 at 1:23 PM, Kees Cook <keescook@xxxxxxxxxxxx> wrote:
>> On Thu, Jul 12, 2018 at 8:11 AM, Arnd Bergmann <arnd@xxxxxxxx> wrote:
>>> On Wed, Jul 11, 2018 at 10:36 PM, Kees Cook <keescook@xxxxxxxxxxxx> wrote:
>>>> Two uses of SKCIPHER_REQUEST_ON_STACK() will trigger FRAME_WARN warnings
>>>> (when less than 2048) once the VLA is no longer hidden from the check:
>>>>
>>>> net/rxrpc/rxkad.c:398:1: warning: the frame size of 1152 bytes is larger than 1024 bytes [-Wframe-larger-than=]
>>>> net/rxrpc/rxkad.c:242:1: warning: the frame size of 1152 bytes is larger than 1024 bytes [-Wframe-larger-than=]
>>>>
>>>> This bumps the affected objects by 20% to silence the warnings while
>>>> still providing coverage is anything grows even more.
>>>>
>>>> Signed-off-by: Kees Cook <keescook@xxxxxxxxxxxx>
>>>
>>> (adding David Howells to cc)
>>>
>>> I don't think these are in a fast path, it should be possible to just use
>>> skcipher_alloc_req() instead of SKCIPHER_REQUEST_ON_STACK() here.
>>> From what I can tell, neither of the two are called in atomic context, so
>>> you should be able to use a GFP_KERNEL allocation.
>>
>> Sure, I can do that instead.
>
> Actually, I think this can actually be adjusted to just re-use the
> stack allocation, since rxkad_verify_packet() finishes one before
> doing another in rxkad_verify_packet_1():
That looks very nice, yes. The same thing is needed in
rxkad_secure_packet(), right?
Arnd
