On Mon, Sep 06, 2021 at 12:43:50PM +0100, Mark Brown wrote:

If you're looking for coverage on this stuff it's also good to check
with clang as well, it's sufficiently different that it often triggers
extra stuff [...]

Yup, I tested across multiple GCC and Clang versions, which is
why the failures came as such a surprise. And specifically, these
overflow changes have been designed with Clang in mind (as well as GCC
obviously). (i.e. see the patch[0] in this series, and the last patch[1]
in the coming series.)

-Kees

[0] https://lore.kernel.org/lkml/20210822075122.864511-17-keescook@chromium.org/ (local)
[1] https://lore.kernel.org/lkml/20210818060533.3569517-64-keescook@chromium.org/ (local)

-- 
Kees Cook