Re: [PATCH v2 00/19] `zerocopy` support
Greg Kroah-Hartman <[email protected]>
| Newsgroups | gmane.linux.kernel.rust,gmane.linux.kbuild.devel |
|---|---|
| Message-ID | <2026061250-negligent-bagging-9ed8@gregkh> |
On Tue, Jun 09, 2026 at 03:08:20PM +0200, Miguel Ojeda wrote: > On Tue, Jun 9, 2026 at 2:43 PM Alice Ryhl <[email protected]> wrote: > > > > > It's not clear to me if changing the uapi headers like this is even > > > allowed to begin with. It's a somewhat non-trivial change. > > After our call I recalled the tool I was thinking about: > > scripts/check-uapi.sh > Documentation/dev-tools/checkuapi.rst > > It turns out it uses `abidiff` (abigail). > > >From a quick look, it seems to complain if I add explicit padding > where implicit padding was there before, i.e. as a "data member > insertion", even if the types etc. are the same. > > However, for this change, i.e. when inside the `union`, it doesn't > seem to complain. check-uapi is there just to determine if we actually change the user/kernel api, by adding explicit padding you didn't change the size or location of anything, so all is good. Ideally yes, we should have explicit padding everywhere, so we know to catch this, but for older structures sometimes we didn't :( thanks, greg k-h