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> <20220921145422.437618-3-ardb@kernel.org>
 <Yz6xBROUBPyaUSoB@zn.tnic> <CAMj1kXGCWmay_=cncZpZwXoyLgzt7=2dVuXHaaQU=K6NEXrezQ@mail.gmail.com>
 <Yz64aMVo4W+D70Fz@zn.tnic> <CAMj1kXHi4TgGd=vicfxVOUnzDrhSfsiR=572L6cEwi47JO17jw@mail.gmail.com>
In-Reply-To: <CAMj1kXHi4TgGd=vicfxVOUnzDrhSfsiR=572L6cEwi47JO17jw@mail.gmail.com>
From:   Ard Biesheuvel <ardb@kernel.org>
Date:   Thu, 6 Oct 2022 14:27:54 +0200
Message-ID: <CAMj1kXEkX=dZ0uUAdXoe5JC_Zv2+ndTMYwDiixYmWT-ip8OOOg@mail.gmail.com>
Subject: Re: [PATCH v2 02/16] x86/compressed: efi-mixed: move 32-bit
 entrypoint code 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, 6 Oct 2022 at 13:19, Ard Biesheuvel <ardb@kernel.org> wrote:
>
> On Thu, 6 Oct 2022 at 13:13, Borislav Petkov <bp@alien8.de> wrote:
> >
> > On Thu, Oct 06, 2022 at 12:56:09PM +0200, Ard Biesheuvel wrote:
> > > efi32_pe_entry() preserves and restores the caller's value of %ebx,
> > > because from there, we might actually return control to the firmware.
> > > The value it keeps in %ebx itself is not live when it jumps to
> > > efi32_entry - it stores its value into image_offset, which is reloaded
> > > from memory at a later point.
> >
> > Hmm, might be prudent to have a comment there because it is using %ebx a
> > couple of insns before the JMP:
> >
> >         subl    %esi, %ebx
> >                       ^^^^
> >         movl    %ebx, rva(image_offset)(%ebp)   // save image_offset
> >
> > <--- I think you mean that after this, %ebx is not needed anymore?
> >
>
> Exactly.
>
> >         xorl    %esi, %esi
> >         jmp     efi32_entry
> >
> > 2:      popl    %edi                            // restore callee-save registers
> >         popl    %ebx
> >
> > and this restores its original value ofc.
> >
> > > efi32_stub_entry() is the 'EFI handover protocol' entry point, which
> > > cannot return to the firmware (and we discard the return address
> > > already) so %ebx can be clobbered.
> >
> > That info would be good to have in a comment above it.
> >
>
> Fair enough.

I'll add the below in the next revision

--- a/arch/x86/boot/compressed/head_64.S
+++ b/arch/x86/boot/compressed/head_64.S
@@ -307,6 +307,19 @@ SYM_FUNC_START(efi32_stub_entry)
 SYM_FUNC_END(efi32_stub_entry)

        .text
+/*
+ * This is the common EFI stub entry point for mixed mode.
+ *
+ * Arguments:  %ecx    image handle
+ *             %edx    EFI system table pointer
+ *             %esi    struct bootparams pointer (or NULL when not using
+ *                     the EFI handover protocol)
+ *
+ * Since this is the point of no return for ordinary execution, no registers
+ * are considered live except for the function parameters. [Note that the EFI
+ * stub may still exit and return to the firmware using the Exit() EFI boot
+ * service.]
+ */
 SYM_FUNC_START_LOCAL(efi32_entry)
        call    1f
 1:     pop     %ebx
@@ -837,7 +850,8 @@ SYM_FUNC_START(efi32_pe_entry)
        subl    %esi, %ebx
        movl    %ebx, rva(image_offset)(%ebp)   // save image_offset
        xorl    %esi, %esi
-       jmp     efi32_entry
+       jmp     efi32_entry                     // pass %ecx, %edx, %esi
+                                               // no other registers
remain live

 2:     popl    %edi                            // restore callee-save registers
        popl    %ebx