On 2/17/26 15:01, Konstantin Belousov wrote:
On Tue, Feb 17, 2026 at 02:42:20PM -0500, John Baldwin wrote:
On 2/16/26 21:44, Konstantin Belousov wrote:
On Mon, Feb 16, 2026 at 05:01:58PM -0800, Enji Cooper (yaneurabeya) wrote:
On Feb 9, 2026, at 9:57 AM, Konstantin Belousov <[email protected]> wrote:
The branch main has been updated by kib:
URL:
https://cgit.FreeBSD.org/src/commit/?id=97ce8f7d2b8eff42460f60e8c49d9849b58b6e81
commit 97ce8f7d2b8eff42460f60e8c49d9849b58b6e81
Author: Konstantin Belousov <[email protected]>
AuthorDate: 2026-02-07 00:36:44 +0000
Commit: Konstantin Belousov <[email protected]>
CommitDate: 2026-02-09 17:57:00 +0000
amd64: define and use STACKALIGN and REDZONE_SZ
Reviewed by: brooks, emaste, jhb
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
Differential revision: https://reviews.freebsd.org/D55151
Hi kib@!
It looks like this broke the build with gcc12 on main:
https://ci.freebsd.org/job/FreeBSD-main-amd64-gcc14_build/2893/console .
https://reviews.freebsd.org/D55319
might be the fix.
I have the same thing at https://reviews.freebsd.org/D55161
Per my comments there, I do think it might break 32-bit kernels, but we probably
don't care about that. (Namely, if rounddown2() is used on a type larger than
uintptr_t such as trunc_page on a vm_paddr_t on i386, this change truncates the
value).
If we really care, we could use _Generic or the like to only use the uintptr_t
casts for pointer types.
I must admit that I initially wrote the cast to uintmax_t, and only then
looked at the nearby code (__builtin_align_up() and others). Since
uintptr_t was good enough for them, I used it in the patch as well.
I'm happy to just merge it as after my last batch of merges it is now the last
patch needed to build aarch64 kernels with GCC. Userspace for aarch64 for GCC
needs one more fix (though I think we want to fix it in the compiler rather than
in src, GCC isn't passing -lgcc when linking shared libraries for some reason).
--
John Baldwin
