linux-kernel - Re: Regression from efi: call get_event

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CAKv+Gu_EN2djoJQQ_QAxUQenD5=kKtyS8TFKg7cZaz+6xsbDRg@mail.gmail.com>
Date:   Tue, 13 Mar 2018 10:30:47 +0000
From:   Ard Biesheuvel <ard.biesheuvel@...aro.org>
To:     Thiebaud Weksteen <tweek@...gle.com>
Cc:     Hans de Goede <hdegoede@...hat.com>,
        Jeremy Cline <jeremy@...ine.org>,
        Javier Martinez Canillas <javierm@...hat.com>,
        Jarkko Sakkinen <jarkko.sakkinen@...ux.intel.com>,
        linux-efi@...r.kernel.org, linux-integrity@...r.kernel.org,
        tpmdd-devel@...ts.sourceforge.net,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: Regression from efi: call get_event_log before ExitBootServices

On 13 March 2018 at 10:23, Thiebaud Weksteen <tweek@...gle.com> wrote:
> On Tue, Mar 13, 2018 at 8:59 AM Ard Biesheuvel <ard.biesheuvel@...aro.org>
> wrote:
>
>> On 13 March 2018 at 07:47, Hans de Goede <hdegoede@...hat.com> wrote:
...
>> > Could the problem perhaps be that the new code for the TPM event-log is
>> > missing some handling to deal with running on a 32 bit firmware? I know
> the
>> > rest of the kernel has special code to deal with this.
>> >
>
> Yes, that was my guess as well.
>
>> That is a very good point, and I missed that this is a 64-bit kernel
>> running on 32-bit UEFI.
>
>> The TPM code does use efi_call_proto() directly, and now I am thinking
>> it is perhaps the allocate_pages() call that simply only initializes
>> the low 32-bits of log_tbl.
>
> That make sense. Would you know what happens to the arguments of the
> function in this case? (I'm thinking &log_location ?)

log_location and log_last_entry are always 64-bit quantities, and
efi_bool_t is always u8, so that shouldn't matter.

> Would it be safer to skip the code completely on EFI_MIXED systems?
>

Obviously, but I would like to avoid that if possible. Let's see first
if we've pinpointed it now.