[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAK8P3a0DX6Eh0OwUh975RL2Pq_ux_AQeBPSakPFxZf+t62EJ=w@mail.gmail.com>
Date: Tue, 4 Dec 2018 22:33:13 +0100
From: Arnd Bergmann <arnd@...db.de>
To: Firoz Khan <firoz.khan@...aro.org>
Cc: linuxppc-dev <linuxppc-dev@...ts.ozlabs.org>,
Benjamin Herrenschmidt <benh@...nel.crashing.org>,
Paul Mackerras <paulus@...ba.org>,
Michael Ellerman <mpe@...erman.id.au>,
Geert Uytterhoeven <geert@...ux-m68k.org>, linuxram@...ibm.com,
Mathieu Desnoyers <mathieu.desnoyers@...icios.com>,
leitao@...ian.org, Boqun Feng <boqun.feng@...il.com>,
gregkh <gregkh@...uxfoundation.org>,
Philippe Ombredanne <pombredanne@...b.com>,
Thomas Gleixner <tglx@...utronix.de>,
Kate Stewart <kstewart@...uxfoundation.org>,
y2038 Mailman List <y2038@...ts.linaro.org>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
linux-arch <linux-arch@...r.kernel.org>,
Deepa Dinamani <deepa.kernel@...il.com>,
Marcin Juszkiewicz <marcin.juszkiewicz@...aro.org>
Subject: Re: [PATCH v3 4/4] powerpc: generate uapi header and system call
table files
On Tue, Dec 4, 2018 at 10:17 PM Arnd Bergmann <arnd@...db.de> wrote:
>
> > diff --git a/arch/powerpc/kernel/systbl.S b/arch/powerpc/kernel/systbl.S
> > index 9ff1913..d756348 100644
> > --- a/arch/powerpc/kernel/systbl.S
> > +++ b/arch/powerpc/kernel/systbl.S
>
> > #ifdef CONFIG_PPC64
> > @@ -47,5 +25,17 @@
> > .globl sys_call_table
> > sys_call_table:
> >
> > +#if defined CONFIG_PPC64 && defined(CONFIG_COMPAT)
> > +#define __SYSCALL(nr, entry, nargs) .8byte entry
> > #define compat_sys_sigsuspend sys_sigsuspend
> > -#include <asm/systbl.h>
> > +#include <asm/syscall_table_c32.h>
> > +#undef __SYSCALL
> > +#elif defined CONFIG_PPC64 && !defined(CONFIG_COMPAT)
> > +#define __SYSCALL(nr, entry, nargs) .8byte entry
> > +#include <asm/syscall_table_64.h>
> > +#undef __SYSCALL
>
> This part looks wrong: where the old sys_call_table contained
> both the native and compat calls side by side, the new table
> contains only the compat versions (note that CONFIG_COMPAT is
> always set on PPC64). I think you copy-pasted this incorrectly from
> one of the other architectures that uses a different layout for the
> compat calls.
>
> The way I think this could work best is if you add a preparation
> patch that splits out the 64-bit table into two separate ones
> (and make sure that works). After that, you can add your
> existing patch, rebased to use the two generated tables
> instead of the old-style tables. I'll follow up with an untested
> patch for illustration.
One more thing I noticed while writing up that patch: you also need
to use the DOTSYM() macro in __SYSCALL() to generate the correct
symbol name for the old ELF ABI (when PPC64_ELF_ABI_v2 is
not set).
Arnd
Powered by blists - more mailing lists