[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20200604165703.GG3976@hirez.programming.kicks-ass.net>
Date: Thu, 4 Jun 2020 18:57:03 +0200
From: Peter Zijlstra <peterz@...radead.org>
To: David Miller <davem@...emloft.net>
Cc: mingo@...nel.org, will@...nel.org, tglx@...utronix.de,
x86@...nel.org, linux-kernel@...r.kernel.org,
a.darwish@...utronix.de, rostedt@...dmis.org,
bigeasy@...utronix.de, sparclinux@...r.kernel.org,
mpe@...erman.id.au, linuxppc-dev@...ts.ozlabs.org,
heiko.carstens@...ibm.com, linux-s390@...r.kernel.org
Subject: Re: [RFC][PATCH v3 1/5] sparc64: Fix asm/percpu.h build error
On Fri, May 29, 2020 at 04:29:17PM -0700, David Miller wrote:
> From: Peter Zijlstra <peterz@...radead.org>
> Date: Fri, 29 May 2020 23:35:51 +0200
>
> > ../arch/sparc/include/asm/percpu_64.h:7:24: warning: call-clobbered register used for global register variable
> > register unsigned long __local_per_cpu_offset asm("g5");
>
> The "-ffixed-g5" option on the command line tells gcc that we are
> using 'g5' as a fixed register, so some part of your build isn't using
> the:
>
> KBUILD_CFLAGS += -ffixed-g4 -ffixed-g5 -fcall-used-g7 -Wno-sign-compare
>
> from arch/sparc/Makefile for some reason.
Thanks, that was the clue I needed.
I think I see, what happens is that these headers end up in the VDSO
build, and that doesn't have these CFLAGS, because userspace.
Let me see what to do about that.
Powered by blists - more mailing lists