Hi Michał,
>
> Hi,
>
> On 05/29, Brian Gix wrote:
> > This will have one major side effect in that it will no longer support
> > kernels older than v4.9 (I am personally using v5.0.17). I am told
> > that (some?) Freescale platforms also will have problems regardless of
> > kernel version.
>
> Alright, so I agree that it would be up to the vendor (i.e. me ;) to provide a
> patch for older kernels.
>
> I think the most sensible option would be to patch ELL to use userspace fallback,
> instead of patching bluez. So if you could just make sure that an appropriate
> unit test exists within ELL, I have no objections.
The unit test for AEAD in .../ell/unit/test-cipher.c already makes this report on encryption failures:
if (!success) {
printf("* Some kernel versions before v4.9 have a known AEAD\n"
"* bug. If the system running this test is using a\n"
"* v4.8 or earlier kernel, a failure here is likely\n"
"* due to that kernel bug.\n");
}
So I will likely be applying the patch as written by this weekend to convert all of mesh crypto to ELL.
Any patching (to ELL or BlueZ) will again become the responsibility of the vendor offering support to these older platforms. To emphasize this, I will probably also add a note to .../bluez/mesh/README to emphasize both the problem with these older platforms, usage of the ELL unit test to identify, and the required work-around.
Best Regards,
Brian
