linux-kernel - Re: [PATCH v4 2/6] x86/sev-es: Disable IRQs while GHCB is active

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

Date:   Fri, 11 Jun 2021 16:20:36 +0200
From:   Joerg Roedel <joro@...tes.org>
To:     Borislav Petkov <bp@...en8.de>
Cc:     x86@...nel.org, Joerg Roedel <jroedel@...e.de>, hpa@...or.com,
        Andy Lutomirski <luto@...nel.org>,
        Dave Hansen <dave.hansen@...ux.intel.com>,
        Peter Zijlstra <peterz@...radead.org>,
        Jiri Slaby <jslaby@...e.cz>,
        Dan Williams <dan.j.williams@...el.com>,
        Tom Lendacky <thomas.lendacky@....com>,
        Juergen Gross <jgross@...e.com>,
        Kees Cook <keescook@...omium.org>,
        David Rientjes <rientjes@...gle.com>,
        Cfir Cohen <cfir@...gle.com>,
        Erdem Aktas <erdemaktas@...gle.com>,
        Masami Hiramatsu <mhiramat@...nel.org>,
        Mike Stunes <mstunes@...are.com>,
        Sean Christopherson <seanjc@...gle.com>,
        Martin Radev <martin.b.radev@...il.com>,
        Arvind Sankar <nivedita@...m.mit.edu>,
        linux-coco@...ts.linux.dev, linux-kernel@...r.kernel.org,
        kvm@...r.kernel.org, virtualization@...ts.linux-foundation.org
Subject: Re: [PATCH v4 2/6] x86/sev-es: Disable IRQs while GHCB is active

On Fri, Jun 11, 2021 at 04:05:15PM +0200, Borislav Petkov wrote:
> On Thu, Jun 10, 2021 at 11:11:37AM +0200, Joerg Roedel wrote:
> Why not simply "sandwich" them:
> 
> 	local_irq_save()
> 	sev_es_get_ghcb()
> 
> 	...blablabla
> 
> 	sev_es_put_ghcb()
> 	local_irq_restore();
> 
> in every call site?

I am not a fan of this, because its easily forgotten to add
local_irq_save()/local_irq_restore() calls around those. Yes, we can add
irqs_disabled() assertions to the functions, but we can as well just
disable/enable IRQs in them. Only the previous value of EFLAGS.IF needs
to be carried from one function to the other.

> Hmm, so why aren't you accessing/setting data->ghcb_active and
> data->backup_ghcb_active safely using cmpxchg() if this path can be
> interrupted by an NMI?

Using cmpxchg is not necessary here. It is all per-cpu data, so local to
the current cpu. If an NMI happens anywhere in sev_es_get_ghcb() it can
still use the GHCB, because the interrupted #VC handler will not start
writing to it before sev_es_get_ghcb() returned.

Problems only come up when one path starts writing to the GHCB, but that
happens long after it is marked active.

Regards,

	Joerg