Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:5 as permitted sender) client-ip=2620:137:e000::3:5;
MIME-Version: 1.0
References: <20231011192528.262425-1-nik.borisov@suse.com> <20231012101456.goamenepqlte65jv@box.shutemov.name>
 <8e36be57-58a6-404e-8828-6c777b8d2196@suse.com> <20231012112523.6koxkdrk72srx5og@box.shutemov.name>
In-Reply-To: <20231012112523.6koxkdrk72srx5og@box.shutemov.name>
From:   Ard Biesheuvel <ardb@kernel.org>
Date:   Thu, 12 Oct 2023 13:26:40 +0200
Message-ID: <CAMj1kXFrhLMvnsQagj7HpeieKhywyCgFpw7vRS1x-nwDT=mUow@mail.gmail.com>
Subject: Re: [PATCH] x86/efistub: Don't try to print after ExitBootService()
To:     kirill.shutemov@linux.intel.com
Cc:     Nikolay Borisov <nik.borisov@suse.com>, linux-efi@vger.kernel.org,
        linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Thu, 12 Oct 2023 at 13:25, <kirill.shutemov@linux.intel.com> wrote:
>
> On Thu, Oct 12, 2023 at 01:51:13PM +0300, Nikolay Borisov wrote:
> >
> >
> > On 12.10.23 =D0=B3. 13:14 =D1=87., kirill.shutemov@linux.intel.com wrot=
e:
> > > On Wed, Oct 11, 2023 at 10:25:28PM +0300, Nikolay Borisov wrote:
> > > > setup_e820() is executed after UEFI's ExitBootService has been call=
ed.
> > > > This causes the firmware to throw an exception because Console IO
> > > > protocol handler is supposed to work only during boot service
> > > > environment. As per UEFI 2.9, section 12.1:
> > > >
> > > >   "This protocol isused to handle input and output of text-based
> > > >   information intended for the system user during the operation of =
code
> > > >   in the boot services environment."
> > > >
> > > > Running a TDX guest with TDVF with unaccepted memory disabled resul=
ts in
> > > > the following output:
> > >
> > > Oh. My bad.
> > >
> > > But there's other codepath that does the same. If setup_e820() fails =
with
> > > EFI_BUFFER_TOO_SMALL, efi_stub_entry() would try to print "exit_boot(=
)
> > > failed\n".
> > >
> > > I wouldner if it is feasible to hook up earlyprintk console into
> > > efi_printk() machinery for after ExitBootService() case? Silent boot
> > > failure is not the best UX.
> > >
> >
> >
> > In my testing I was able to transpose setup_e820 and efi exit_boot_serv=
ice
> > by calling exit_boot_func before setup_e820 which ensures the various m=
emory
> > variables are populated. Is there any specific reason why ExitBootServi=
ces
> > is called before setting up the e820 table? AFAIU this is an arbitrary
> > choice?
>
> Because if you allocate memory with EFI service it can alter EFI memory
> map and we need the last version to convert it to e820.
>

Indeed, and note that the memory map may change due to asynchronous
events, which only get shut down when ExitBootServices() is called.
This is the reason for this complicated dance around
ExitBootServices() with the callback etc