Re: [patch 0/6] Cure kexec() vs. mwait_play_dead() troubles

From: Sean Christopherson <seanjc@google.com>
To: Ashok Raj <ashok_raj@linux.intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>,
	LKML <linux-kernel@vger.kernel.org>,
	x86@kernel.org, Ashok Raj <ashok.raj@linux.intel.com>,
	Dave Hansen <dave.hansen@linux.intel.com>,
	Tony Luck <tony.luck@intel.com>,
	Arjan van de Veen <arjan@linux.intel.com>,
	Peter Zijlstra <peterz@infradead.org>,
	Eric Biederman <ebiederm@xmission.com>,
	Ashok Raj <ashok.raj@intel.com>
Subject: Re: [patch 0/6] Cure kexec() vs. mwait_play_dead() troubles
Date: Wed, 7 Jun 2023 10:33:35 -0700	[thread overview]
Message-ID: <ZIC/b+AwvH2wIz/o@google.com> (raw)
In-Reply-To: <ZICuhZHCqSYvR4IO@araj-dh-work>

On Wed, Jun 07, 2023, Ashok Raj wrote:
> On Tue, Jun 06, 2023 at 12:41:43AM +0200, Thomas Gleixner wrote:
> > On Mon, Jun 05 2023 at 10:41, Sean Christopherson wrote:
> > > On Sat, Jun 03, 2023, Thomas Gleixner wrote:
> > >> This is only half safe because HLT can resume execution due to NMI, SMI and
> > >> MCE. Unfortunately there is no real safe mechanism to "park" a CPU reliably,
> > >
> > > On Intel.  On AMD, enabling EFER.SVME and doing CLGI will block everything except
> > > single-step #DB (lol) and RESET.  #MC handling is implementation-dependent and
> > > *might* cause shutdown, but at least there's a chance it will work.  And presumably
> > > modern CPUs do pend the #MC until GIF=1.
> > 
> > Abusing SVME for that is definitely in the realm of creative bonus
> > points, but not necessarily a general purpose solution.
> > 
> > >> So parking them via INIT is not completely solving the problem, but it
> > >> takes at least NMI and SMI out of the picture.
> > >
> > > Don't most SMM handlers rendezvous all CPUs?  I.e. won't blocking SMIs indefinitely
> > > potentially cause problems too?
> > 
> > Not that I'm aware of. If so then this would be a hideous firmware bug
> > as firmware must be aware of CPUs which hang around in INIT independent
> > of this.
> 
> SMM does do the rendezvous of all CPUs, but also has a way to detect the
> blocked ones, in WFS via some package scoped ubox register. So it knows to
> skip those. I can find this in internal sources, but they aren't available
> in the edk2 open reference code. They happen to be documented only in the
> BWG, which isn't available freely.

Ah, so putting CPUs into WFS shouldn't result in odd delays.  At least not on
bare metal.  Hmm, and AFAIK the primary use case for SMM in VMs is for secure
boot, so taking SMIs after booting and putting CPUs back into WFS should be ok-ish.

Finding a victim to test this in a QEMU VM w/ Secure Boot would be nice to have.