Re: [PATCH v4 08/21] lib/crc: Switch ARM and arm64 to 'ksimd' scoped guard API

[Jonathan Cameron <jonathan.cameron@huawei.com>]

On Fri, 31 Oct 2025 15:05:19 +0100
Ard Biesheuvel <ardb@kernel.org> wrote:

> On Fri, 31 Oct 2025 at 14:52, Ard Biesheuvel <ardb@kernel.org> wrote:
> >
> > On Fri, 31 Oct 2025 at 14:49, Jonathan Cameron
> > <jonathan.cameron@huawei.com> wrote:  
> > >
> > > On Fri, 31 Oct 2025 11:39:07 +0100
> > > Ard Biesheuvel <ardb+git@google.com> wrote:
> > >  
> > > > From: Ard Biesheuvel <ardb@kernel.org>
> > > >
> > > > Before modifying the prototypes of kernel_neon_begin() and
> > > > kernel_neon_end() to accommodate kernel mode FP/SIMD state buffers
> > > > allocated on the stack, move arm64 to the new 'ksimd' scoped guard API,
> > > > which encapsulates the calls to those functions.
> > > >
> > > > For symmetry, do the same for 32-bit ARM too.
> > > >
> > > > Reviewed-by: Eric Biggers <ebiggers@kernel.org>
> > > > Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
> > > > ---
> > > >  lib/crc/arm/crc-t10dif.h   | 16 +++++-----------
> > > >  lib/crc/arm/crc32.h        | 11 ++++-------
> > > >  lib/crc/arm64/crc-t10dif.h | 16 +++++-----------
> > > >  lib/crc/arm64/crc32.h      | 16 ++++++----------
> > > >  4 files changed, 20 insertions(+), 39 deletions(-)
> > > >
> > > > diff --git a/lib/crc/arm/crc-t10dif.h b/lib/crc/arm/crc-t10dif.h
> > > > index 63441de5e3f1..aaeeab0defb5 100644
> > > > --- a/lib/crc/arm/crc-t10dif.h
> > > > +++ b/lib/crc/arm/crc-t10dif.h  
> > >  
> > > >  static __ro_after_init DEFINE_STATIC_KEY_FALSE(have_neon);
> > > > @@ -20,21 +19,16 @@ asmlinkage void crc_t10dif_pmull8(u16 init_crc, const u8 *buf, size_t len,
> > > >  static inline u16 crc_t10dif_arch(u16 crc, const u8 *data, size_t length)
> > > >  {
> > > >       if (length >= CRC_T10DIF_PMULL_CHUNK_SIZE) {
> > > > -             if (static_branch_likely(&have_pmull)) {
> > > > -                     if (likely(may_use_simd())) {
> > > > -                             kernel_neon_begin();
> > > > -                             crc = crc_t10dif_pmull64(crc, data, length);
> > > > -                             kernel_neon_end();
> > > > -                             return crc;
> > > > -                     }
> > > > +             if (static_branch_likely(&have_pmull) && likely(may_use_simd())) {
> > > > +                     scoped_ksimd()
> > > > +                             return crc_t10dif_pmull64(crc, data, length);  
> > >  
> > > >               } else if (length > CRC_T10DIF_PMULL_CHUNK_SIZE &&
> > > >                          static_branch_likely(&have_neon) &&
> > > >                          likely(may_use_simd())) {  
> > >
> > > I briefly thought this was a functional change but it's not because
> > > of may_use_simd() being something that isn't going to change between
> > > the two evaluations.
> > >
> > > Would it hurt at all to pull that up to be
> > >         if (length >= CRC_T10DIF_PMULL_CHUNK_SIZE && likely(may_use_simd())) {
> > >                 if (static_branch_likely(&have_pmull)) {
> > >                         scoped_ksimd()
> > >                                 return crc_t10dif_pmull64(crc, data, length);
> > >
> > >                 } else if (length > CRC_T10DIF_PMULL_CHUNK_SIZE &&
> > >                            static_branch_likely(&have_neon)) {
> > >                 ...
> > >
> > > ?
> > >  
> >
> > Yeah that would be a reasonable cleanup, I guess.  
> 
> Actually, looking more closely, that would result in may_use_simd()
> being evaluated even when the static keys are set to false, given that
> the compiler is unlikely to be able to figure out by itself that
> may_use_simd() has no side effects.
Yeah. That was why it was a question :)
Given everything is marked as likely I wasn't sure if we cared about when
the static keys aren't set.

Jonathan

>