Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20200221050934.719152-1-brgerst@gmail.com> <20200221060756.GA3368@light.dominikbrodowski.net>
In-Reply-To: <20200221060756.GA3368@light.dominikbrodowski.net>
From:   Brian Gerst <brgerst@gmail.com>
Date:   Fri, 21 Feb 2020 08:15:19 -0500
Message-ID: <CAMzpN2iQuaNdTdL6G1rGbUFo+r16iRFo1zbiD_VMrrjtGf0acw@mail.gmail.com>
Subject: Re: [PATCH 0/5] Enable pt_regs based syscalls for x86-32 native
To:     Dominik Brodowski <linux@dominikbrodowski.net>
Cc:     Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        "the arch/x86 maintainers" <x86@kernel.org>,
        Thomas Gleixner <tglx@linutronix.de>,
        Ingo Molnar <mingo@kernel.org>, Borislav Petkov <bp@alien8.de>,
        "H . Peter Anvin" <hpa@zytor.com>,
        Andy Lutomirski <luto@kernel.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Fri, Feb 21, 2020 at 2:07 AM Dominik Brodowski
<linux@dominikbrodowski.net> wrote:
>
> Brian,
>
> On Fri, Feb 21, 2020 at 12:09:29AM -0500, Brian Gerst wrote:
> > This patch series cleans up the x86 syscall wrapper code and converts
> > the 32-bit native kernel over to pt_regs based syscalls.
>
> thanks for your patchset. Could you explain a bit more what the rationale
> is. Due to asmlinkage, it doesn't leak "random user-provided register
> content down the call chain" (as was the case for x86-64). But it may be
> cleaner, and you mention in patch 5/5 that the new way is "a bit more
> efficient" -- do you have numbers?

The main rationale for this patch set is to make the 32-bit native
kernel consistent with the 64-bit kernel.  It's also slightly more
efficient because the old code pushed all 6 arguments onto the stack
whereas the new code only reads the args the syscall needs, with the
pt_regs pointer passed in through a register.  By efficient I mean
that it uses fewer instructions and stack accesses, not that it will
actually have a significant difference on a benchmark.

--
Brian Gerst