linux-kernel - Re: [PATCH 0/2] UEFI panic notification mechanism

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: Windows password security audit tool. GUI, reports in PDF.

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

Message-ID: <b11ff46a-2e20-01b6-cbd9-2038b7bf4bc9@igalia.com>
Date:   Tue, 28 Jun 2022 20:10:09 +0200
From:   "Guilherme G. Piccoli" <gpiccoli@...lia.com>
To:     Ard Biesheuvel <ardb@...nel.org>
Cc:     Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        kernel-dev@...lia.com, kernel@...ccoli.net,
        Anton Vorontsov <anton@...msg.org>,
        Colin Cross <ccross@...roid.com>,
        Kees Cook <keescook@...omium.org>,
        Matt Fleming <matt@...eblueprint.co.uk>,
        Matthew Garrett <matthew.garrett@...ula.com>,
        Tony Luck <tony.luck@...el.com>,
        linux-efi <linux-efi@...r.kernel.org>
Subject: Re: [PATCH 0/2] UEFI panic notification mechanism

On 28/06/2022 09:49, Ard Biesheuvel wrote:
> On Thu, 2 Jun 2022 at 19:40, Guilherme G. Piccoli <gpiccoli@...lia.com> wrote:
>>
>> Hi Ard, apologies for annoying!
>>
> 
> No worries, just very busy :-)
> 
>> Just a friendly ping asking if you have any opinions about these patches.
>>
> 
> Honestly, I'm not sure I see the value of this. You want to 'notify
> the UEFI firmware' but the firmware doesn't really care about these
> variables, only your bespoke tooling does. EFI pstore captures far
> more data, so if you just wipe /sys/fs/pstore after each clean boot,
> you already have all the data you need, no?
> 
> Also, I'm in the process of removing the public efivar_entry_xxx() API
> - please look at the efi/next tree for a peek. This is related to your
> point 3), i.e., the efivar layer is a disaster in terms of consistency
> between different observers of the EFI variable store. Switching to
> efivar_set_variable() [the new API] should fix that.

Hi Ard, thanks a lot for your review! Lemme split my points in some
bullets below:

a) What about patch 1, is it good as-is?

b) Cool about efivar_set_variable(), could fix that in V2.

c) Now, to the most relevant thing, the value of this. I might be
mistaken, but is there any known way to let UEFI know a panic happened?
For now, maybe only my UEFI fw might use that (and the usage is very
nice, showing a panic logo), but this opens a myriad of potential uses.
We can think in RAM preserving mechanism conditional to panic scenarios,
special resets for panic vs. cold boot, and even maybe a firmware vmcore
capturing.

If there is any other way to let UEFI know about a panic, let me know
and that will likely be more than enough for me. Otherwise, do you think
such small code would be a big burden to carry, considering the
cost/benefit for my use case?

Thanks in advance for your analysis.
Cheers,

Guilherme