[PATCH v6 1/3] arm64: mte: make the per-task SCTLR_EL1 field usable elsewhere

[Peter Collingbourne]

On Tue, Jan 26, 2021 at 4:50 AM Will Deacon <will at kernel.org> wrote:
>
> On Tue, Dec 29, 2020 at 10:59:13PM -0800, Peter Collingbourne wrote:
> > In an upcoming change we are going to introduce per-task SCTLR_EL1
> > bits for PAC. Move the existing per-task SCTLR_EL1 field out of the
> > MTE-specific code so that we will be able to use it from both the
> > PAC and MTE code paths and make the task switching code more efficient.
> >
> > Signed-off-by: Peter Collingbourne <pcc at google.com>
> > Link: https://linux-review.googlesource.com/id/Ic65fac78a7926168fa68f9e8da591c9e04ff7278
> > ---
> >  arch/arm64/Kconfig                 |  4 +++
> >  arch/arm64/include/asm/mte.h       |  4 ---
> >  arch/arm64/include/asm/processor.h | 10 ++++++-
> >  arch/arm64/kernel/mte.c            | 47 ++++++------------------------
> >  arch/arm64/kernel/process.c        | 34 +++++++++++++++++----
> >  5 files changed, 50 insertions(+), 49 deletions(-)
> >
> > diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
> > index 05e17351e4f3..82e38d1ca012 100644
> > --- a/arch/arm64/Kconfig
> > +++ b/arch/arm64/Kconfig
> > @@ -348,6 +348,9 @@ config KASAN_SHADOW_OFFSET
> >       default 0xeffffff800000000 if ARM64_VA_BITS_36 && KASAN_SW_TAGS
> >       default 0xffffffffffffffff
> >
> > +config ARM64_NEED_SCTLR_USER
> > +     bool
> > +
> >  source "arch/arm64/Kconfig.platforms"
> >
> >  menu "Kernel Features"
> > @@ -1653,6 +1656,7 @@ config ARM64_MTE
> >       # Required for tag checking in the uaccess routines
> >       depends on ARM64_PAN
> >       select ARCH_USES_HIGH_VMA_FLAGS
> > +     select ARM64_NEED_SCTLR_USER
>
> I'm not sure I'd bother with this; is it the end of the world if we do this
> unconditionally?

I think I'd be fine with doing it unconditionally. If both PAC and MTE
are disabled then I believe that the only additional code that we
would end up executing is the comparison between prev and next
sctlr_user on task switch in process.c (which would always yield
false) and since task switch is much less common than entry/exit it
seems acceptable to me.

Peter