Date: Fri, 22 Apr 2016 09:16:12 +0200
From: Ingo Molnar <mingo@kernel.org>
To: Kees Cook <keescook@chromium.org>
Cc: Borislav Petkov <bp@suse.de>, Baoquan He <bhe@redhat.com>,
        Yinghai Lu <yinghai@kernel.org>, Ingo Molnar <mingo@redhat.com>,
        "x86@kernel.org" <x86@kernel.org>,
        Andrew Morton <akpm@linux-foundation.org>,
        Andrey Ryabinin <aryabinin@virtuozzo.com>,
        Dmitry Vyukov <dvyukov@google.com>, "H.J. Lu" <hjl.tools@gmail.com>,
        Josh Poimboeuf <jpoimboe@redhat.com>,
        Andy Lutomirski <luto@kernel.org>, LKML <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH 2/5] x86, KASLR: Drop CONFIG_RANDOMIZE_BASE_MAX_OFFSET
Message-ID: <20160422071612.GA6819@gmail.com>
References: <1461185746-8017-1-git-send-email-keescook@chromium.org>
 <1461185746-8017-3-git-send-email-keescook@chromium.org>
 <20160421174423.GD29616@pd.tnic>
 <CAGXu5j+n3GdKV9xHHTa1siD=C_+63oUm5Rby-qXCvVAD1Og9Fw@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <CAGXu5j+n3GdKV9xHHTa1siD=C_+63oUm5Rby-qXCvVAD1Og9Fw@mail.gmail.com>
User-Agent: Mutt/1.5.23 (2014-03-12)
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1085
Lines: 30


* Kees Cook <keescook@chromium.org> wrote:

> >> +       Since the kernel is built using 2GB addressing,
> >
> > Does that try to refer to the 1G kernel and 1G fixmap pagetable
> > mappings? I.e., level2_kernel_pgt and level2_fixmap_pgt in
> > arch/x86/kernel/head_64.S?
> 
> The "2GB addressing" part is in reference to:
> 
>        -mcmodel=kernel
>            Generate code for the kernel code model.  The kernel runs in the
>            negative 2 GB of the address space.  This model has to be used for
>            Linux kernel code.

On x86-64 this is a special GCC compiler small memory model, it is called the 
'kernel code model', which is rather generic and no 'real name' ever stuck.

Due to RIP-relative addressing and the sign-extension of 48 bit virtual addresses, 
this allows nearly as compact kernel code and (static) kernel data definitions as 
a 32-bit kernel would allow.

The (positive) 0-4GB virtual memory range has similar advantages, but is of course 
frequently used by user-space code. Negative addresses are reserved for the kernel 
only.

Thanks,

	Ingo