Re: git: 238271f4a66b - main - stand: Add a snarky note about the upstream ZFS situation

On 18 Apr 2023, at 23:49, Warner Losh <imp@bsdimp.com> wrote:
> 
> 
> 
> On Tue, Apr 18, 2023, 3:34 PM Jessica Clarke <jrtc27@freebsd.org> wrote:
>> On 18 Apr 2023, at 22:31, Warner Losh <imp@FreeBSD.org> wrote:
>> > 
>> > The branch main has been updated by imp:
>> > 
>> > URL: https://cgit.FreeBSD.org/src/commit/?id=238271f4a66bd06b8b9a232a82f3ee0882e4cbb9
>> > 
>> > commit 238271f4a66bd06b8b9a232a82f3ee0882e4cbb9
>> > Author:     Warner Losh <imp@FreeBSD.org>
>> > AuthorDate: 2023-04-18 21:29:45 +0000
>> > Commit:     Warner Losh <imp@FreeBSD.org>
>> > CommitDate: 2023-04-18 21:31:17 +0000
>> > 
>> >    stand: Add a snarky note about the upstream ZFS situation
>> > 
>> >    The latest import of openzfs broke the hacks that we used to omit the
>> >    special registers being used on arm64. Add snarky note documenting this
>> >    situation since it's a mess now since the hack was only partially
>> >    undone, leaving behind a mess.
>> > 
>> >    Sponsored by:           Netflix
>> > ---
>> > stand/libsa/zfs/Makefile.inc | 4 ++++
>> > 1 file changed, 4 insertions(+)
>> > 
>> > diff --git a/stand/libsa/zfs/Makefile.inc b/stand/libsa/zfs/Makefile.inc
>> > index f4cecdbc3085..7660f4ab7baf 100644
>> > --- a/stand/libsa/zfs/Makefile.inc
>> > +++ b/stand/libsa/zfs/Makefile.inc
>> > @@ -19,6 +19,7 @@ ZSTD_SRC+=  zstd_common.c
>> > ZSTD_SRC+=    zstd_ddict.c zstd_decompress.c zstd_decompress_block.c
>> > ZSTD_SRC+=    zstd_double_fast.c zstd_fast.c zstd_lazy.c zstd_ldm.c
>> > 
>> > +# This is completely bogus: We should be able to omit this code completely.
>> > .if ${MACHINE_ARCH} == "aarch64"
>> > ZFS_SRC_AS =  b3_aarch64_sse2.S b3_aarch64_sse41.S
>> > .endif
>> > @@ -90,10 +91,13 @@ CFLAGS.skein_block.c+=    -DSKEIN_LOOP=111
>> > 
>> > # To find blake3_impl.c in OpenZFS tree for our somehat ugly blake3_impl_hack.c
>> > # that's needed until the necessary tweaks can be upstreamed.
>> > +# XXX the last import gutted all this since upstream changes broke this hack.
>> > CFLAGS.blake3_impl_hack.c+= -I${OZFS}/module/icp/algs/blake3 -I${OZFS}/module/icp/include
>> > 
>> > CWARNFLAGS.zfs.c+= ${NO_WDANGLING_POINTER}
>> > 
>> > +# Needing to remove the -mgeneral-regs-only is a red flag that this is not quite
>> > +# right. But it's needed at the moment due to the muddled upstream.
>> 
>> This one isn’t bogus? The file is deliberately using NEON so needs
>> access to floating-point registers, which LLVM (mostly) enforces for
>> the assembler, unlike GNU as.
> 
> No. It's bogus because we should not be using it at all. The generic implementation is fast enough for the boot loader and we have a blanket policy against using extra register sets. The change wasn't discussed before hand and what's really needed are upstream changes to be able to omit it entirely.

Oh I see, I misunderstood the point the comment was making.

Thanks,
Jess

> > b3_aarch64_sse2.o: b3_aarch64_sse2.S
> >       ${CC} -c ${CFLAGS:N-mgeneral-regs-only} ${WERROR} ${.IMPSRC} \
> >           -o ${.TARGET}