Re: [Xen-devel] [PATCH 3/4] x86/linker: add a reloc section to ELF binary

From: "Jan Beulich" <JBeulich@suse.com>
To: "Roger Pau Monne" <roger.pau@citrix.com>
Cc: Andrew Cooper <andrew.cooper3@citrix.com>,
	Daniel Kiper <daniel.kiper@oracle.com>, WeiLiu <wl@xen.org>,
	xen-devel <xen-devel@lists.xenproject.org>
Subject: Re: [Xen-devel] [PATCH 3/4] x86/linker: add a reloc section to ELF binary
Date: Tue, 25 Jun 2019 08:18:35 -0600	[thread overview]
Message-ID: <5D122D3B020000780023AFBA@prv1-mh.provo.novell.com> (raw)
In-Reply-To: <20190625124813.hnvxyvbvjfzdbfib@MacBook-Air-de-Roger.local>

>>> On 25.06.19 at 14:48, <roger.pau@citrix.com> wrote:
> On Tue, Jun 25, 2019 at 01:08:49PM +0200, Roger Pau Monné wrote:
>> Sorry for not being clear. By remove I mean `git rm
>> xen/arch/x86/efi/relocs-dummy.S` and fix the build, like the diff
>> appended below.
> 
> The chunk below will not work because relocs-dummy is also needed
> by the EFI build. I'm however lost at why this is required, and the
> commit message that introduced the file (bf6501a62e) doesn't add any
> reasoning.
> 
> Is maybe .reloc mandatory for PE format?

Yes, almost. You _can_ have one without .reloc, but then you're tied
to it loading at the linked address. That's fine with an ordinary boot
loader, but it's not an option when this is to get loaded just like a
normal binary, without knowing at which address it'll be placed.
Remember that the EFI boot environment runs in (pseudo)physical
mode, i.e. there's a 1:1 mapping between linear and physical
addresses. Therefore there's no way to predict a memory range
that's always going to be available (and that hence xen.efi could be
linked to load at).

Jan

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel