On Sat, Dec 7, 2019 at 11:48 AM Kees Cook wrote:
>
> Please pull this mostly mechanical treewide conversion to the single and
> more accurately named sizeof_member() macro for the end of v5.5-rc1.

So this one I'm _still_ not convinced about. It makes yet another name
for something we've had before, which just annoys me. And maybe it's
the 13-year old in me, but "sizeof_member()" just makes me go "that's
puerile".

I _can_ see why we'd want to standardize on one of the tree versions
we have, but I can't really see the problem with the existing #define
that we have, and that is used (admittedly not all that much):
sizeof_field().

This weekend was the last-minute pull request by Google's Kees Cook to introduce the new sizeof_member() macro that had been previously rejected from Linux 5.4. Well, it was again rejected by Linus Torvalds prior to tagging the Linux 5.5-rc1 kernel.

The sizeof_member() macro has been aimed to unify 2~3 other macros within the kernel tree currently and using the size-of-field moniker, but Cook argued that for measuring the size of a member of a C struct, the new macro is more appropriate and converted usage of the old macros to this new single macro.

While there are some pretty great features for Linux 5.5, one that didn't make it quite in time was the long-awaited introduction of WireGuard as the in-kernel secure VPN tunnel. While it was a bummer it didn't make 5.5, all indications are at this point is that it will be in Linux 5.6.

With Linux 5.5 the crypto subsystem adopted some elements of WireGuard's "Zinc" crypto code and that in turn opened the door for merging WireGuard now that the cryptography side was sorted out. But WireGuard was too late for introduction in net-next even with a last minute attempt trying to get it into 5.5, but instead it's aiming early for merging to net-next to ensure it's timely introduction with Linux 5.6.