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;
MIME-Version: 1.0
References: <20220921145422.437618-1-ardb@kernel.org> <20220921145422.437618-5-ardb@kernel.org>
 <Y3ZZ35vJBh4fQu9M@zn.tnic>
In-Reply-To: <Y3ZZ35vJBh4fQu9M@zn.tnic>
From:   Ard Biesheuvel <ardb@kernel.org>
Date:   Thu, 17 Nov 2022 17:06:37 +0100
Message-ID: <CAMj1kXEf=fCwGLakttfAkyGYj4PFrahcAj0GiMCUFVnOyaZtRw@mail.gmail.com>
Subject: Re: [PATCH v2 04/16] x86/compressed: efi-mixed: move efi32_pe_entry
 into .text section
To:     Borislav Petkov <bp@alien8.de>
Cc:     linux-efi@vger.kernel.org, linux-kernel@vger.kernel.org,
        Thomas Gleixner <tglx@linutronix.de>,
        Ingo Molnar <mingo@redhat.com>,
        Dave Hansen <dave.hansen@linux.intel.com>,
        Michael Roth <michael.roth@amd.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Thu, 17 Nov 2022 at 16:57, Borislav Petkov <bp@alien8.de> wrote:
>
> On Wed, Sep 21, 2022 at 04:54:10PM +0200, Ard Biesheuvel wrote:
> >       /*
> >        * We need to set the image_offset variable here since startup_32() will
> >        * use it before we get to the 64-bit efi_pe_entry() in C code.
> >        */
> > -     subl    %esi, %ebx
> > -     movl    %ebx, rva(image_offset)(%ebp)   // save image_offset
> > +     subl    %esi, %ebp                      // calculate image_offset
> > +     movl    %ebp, (image_offset - 1b)(%ebx) // save image_offset
>
> All looks ok, just one question: what was the reason for that
> image_offset thing?
>
> I see:
>
> 1887c9b653f9 ("efi/x86: Decompress at start of PE image load address")
>
> It says that if the kernel is loaded as a PE executable using
> LoadImage() we don't know where that image will be loaded each time so
> we're saving that offset for later when relocating (or not) the kernel?
>
> All part of those improvements:
>
> https://lore.kernel.org/all/20200301230537.2247550-1-nivedita@alum.mit.edu/
>
> Am I close?
>

Yes.

The x86 boot protocol does not require that the setup data block comes
right before the image, it just receives the address in %esi

When doing PE boot, this is guaranteed, and so we can reuse the memory
before the image.

> I.e., that image_offset is purely a kernel thing and not something EFI
> LoadImage's inner workings mandate...? It doesn't seem so from where I'm
> standing but lemme doublecheck still.
>

No this has nothing do with the EFI in particular, only with how the
x86 boot image is constructed and wrapped into a PE/COFF executable.