[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Tue, 5 May 2020 19:42:43 +0200
From: Torsten Duwe <duwe@....de>
To: Mark Rutland <mark.rutland@....com>
Cc: Arnd Bergmann <arnd@...db.de>,
Catalin Marinas <catalin.marinas@....com>,
Will Deacon <will@...nel.org>,
Ard Biesheuvel <ardb@...nel.org>,
Amit Daniel Kachhap <amit.kachhap@....com>,
Torsten Duwe <duwe@...e.de>,
Ard Biesheuvel <ard.biesheuvel@...aro.org>,
AKASHI Takahiro <takahiro.akashi@...aro.org>,
Josh Poimboeuf <jpoimboe@...hat.com>,
Julien Thierry <jthierry@...hat.com>,
Andrew Morton <akpm@...ux-foundation.org>,
Marc Zyngier <maz@...nel.org>,
Kees Cook <keescook@...omium.org>,
Alexandre Ghiti <alex@...ti.fr>,
Kristina Martsenko <kristina.martsenko@....com>,
Ionela Voinescu <ionela.voinescu@....com>,
Steve Capper <steve.capper@....com>,
linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
clang-built-linux@...glegroups.com
Subject: Re: [PATCH] arm64: disable patchable function entry on big-endian
clang builds
Hi Arnd, Mark and others,
this may not be worth arguing but I'm currently fighting excessive
workarounds in another area and so this triggers me, so I have to make
a remark ;-)
On Tue, 5 May 2020 15:25:56 +0100
Mark Rutland <mark.rutland@....com> wrote:
> On Tue, May 05, 2020 at 04:12:36PM +0200, Arnd Bergmann wrote:
> > Clang only supports the patchable_function_entry attribute on
> > little-endian arm64 builds, but not on big-endian:
> >
> > include/linux/kasan-checks.h:16:8: error: unknown attribute
> > 'patchable_function_entry' ignored [-Werror,-Wunknown-attributes]
> >
> > Disable that configuration with another dependency. Unfortunately
> > the existing check is not enough, as $(cc-option) at this point does
> > not pass the -mbig-endian flag.
>
> Well that's unfortunate. :(
>
> Do we know if this is deliberate and/or likely to change in future?
> This practically rules out a BE distro kernel with things like PAC,
> which isn't ideal.
But still better than cumulating workarounds. If clang's flags aren't
orthogonal then that's a bug in clang. If I get a vote here I'm against
it.
> > Fixes: 3b23e4991fb6 ("arm64: implement ftrace with regs")
> > Signed-off-by: Arnd Bergmann <arnd@...db.de>
>
> This looks fine for now, and we can add a version check in future, so:
^^^^^^^^^^^^^^^^^^^
see what I mean? And in the end another line of code you'll never again
get rid of.
I suggest to get a quote from clang folks first about their schedule and
regarded importance of patchable-function-entries on BE, and leave it as
is: broken on certain clang configurations. It's not the kernel's fault.
> Acked-by: Mark Rutland <mark.rutland@....com>
>
> Mark.
>
> > ---
> > arch/arm64/Kconfig | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
> > index 4b256fa6db7a..a33d6402b934 100644
> > --- a/arch/arm64/Kconfig
> > +++ b/arch/arm64/Kconfig
> > @@ -151,7 +151,7 @@ config ARM64
> > select HAVE_DMA_CONTIGUOUS
> > select HAVE_DYNAMIC_FTRACE
> > select HAVE_DYNAMIC_FTRACE_WITH_REGS \
> > - if $(cc-option,-fies on y=2)
> > + if $(cc-option,-fpatchable-function-entry=2) &&
> > !(CC_IS_CLANG && CPU_BIG_ENDIAN) select
> > HAVE_EFFICIENT_UNALIGNED_ACCESS select HAVE_FAST_GUP
> > select HAVE_FTRACE_MCOUNT_RECORD
> > --
> > 2.26.0
> >
Powered by blists - more mailing lists