Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Wed, 26 Jul 2023 01:23:37 +0700
From:   Ammar Faizi <ammarfaizi2@gnuweeb.org>
To:     Zhangjin Wu <falcon@tinylab.org>
Cc:     Arnd Bergmann <arnd@arndb.de>, Sven Schnelle <svens@linux.ibm.com>,
        Thomas =?iso-8859-1?Q?Wei=DFschuh?= <thomas@t-8ch.de>,
        Willy Tarreau <w@1wt.eu>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        Linux Kselftest Mailing List 
        <linux-kselftest@vger.kernel.org>
Subject: Re: [PATCH v1 1/8] tools/nolibc: add support for powerpc
Message-ID: <ZMATKempQBPGCY2v@biznet-home.integral.gnuweeb.org>
References: <ZL/gAORYFijgjfvH@biznet-home.integral.gnuweeb.org>
 <20230725170426.58050-1-falcon@tinylab.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20230725170426.58050-1-falcon@tinylab.org>
Precedence: bulk

On Wed, Jul 26, 2023 at 01:04:26AM +0800, Zhangjin Wu wrote:
> My old 'reply' is not rigorous, since the syscall6() uses stack to pass
> the 6th argument, so, our new syscall.h didn't support it currently,
> the syscalls I have tested about "=r" instead of "=a" were only syscall1-5().

Yeah, it won't fit with the new design.

i386 runs out of GPRs very quickly. Given that, it had a hard time
implementing syscall6() properly in nolibc. The calling convention
itself actually doesn't require stack for executing 'int $0x80'.

The reason of why it uses stack is because the %ebp register cannot be
listed in the clobber list nor in the constraint if -fomit-frame-pointer
is not activated. Thus, we have to carefully preserve the value on the
stack before using %ebp as the 6-th argument to the syscall. It's a hack
to make it work on i386.

> Ok, so, with the new syscalls.h proposed, we'd better keep i386
> syscall6() as-is.
> 
> For the left syscall1-5(), is there any risk when use '=r' instead of 'r'?

Using "=r" instead of "r" doesn't make sense.

Did you mean "=r" instead of "=a"?

If that's what you mean:

So far I don't see the risk of using "=r" instead of "=a" as long as the
variable is properly marked as 'register' + asm("eax").

-- 
Ammar Faizi