linux-kernel - Re: [PATCH v4 20/38] KVM: x86/pmu: Check if mediated vPMU can intercept rdpmc

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: <aJKm3zawEkrHT6Ms@google.com>
Date: Tue, 5 Aug 2025 17:50:39 -0700
From: Sean Christopherson <seanjc@...gle.com>
To: Dapeng Mi <dapeng1.mi@...ux.intel.com>
Cc: Sandipan Das <sandipan.das@....com>, Mingwei Zhang <mizhang@...gle.com>, 
	linux-kernel@...r.kernel.org, kvm@...r.kernel.org
Subject: Re: [PATCH v4 20/38] KVM: x86/pmu: Check if mediated vPMU can
 intercept rdpmc

Trimmed the Cc substantially, I'm guessing most folks don't need a play-by-play
update of misguided optimism. 

On Mon, Aug 04, 2025, Sean Christopherson wrote:
> On Fri, Aug 01, 2025, Sean Christopherson wrote:
> > On Wed, Jul 30, 2025, Dapeng Mi wrote:
> > > 
> > > On 7/30/2025 8:38 AM, Sean Christopherson wrote:
> > > > On Tue, Jul 29, 2025, Dapeng Mi wrote:
> > > >> BTW, Sean, may I know your plan about the mediated vPMU v5 patch set? Thanks.
> > > > I'll get it out this week (hopefully tomorrow).
> > > 
> > > Thumbs up! Thanks.
> > 
> > I lied, I'm not going to get it out until Monday.  Figuring out how to deal with
> > instruction emulation in the fastpath VM-Exit handlers took me longer than I was
> > hoping/expecting.
> > 
> > It's fully tested, and I have all but one changelog written, but I'm out of time
> > for today (I made a stupid goof (inverted a !) that cost me an ~hour today, *sigh*).
> > 
> > Unless I get hit by a meteor, I'll get it out Monday.
> 
> *sigh*
> 
> Wrong again (fortunately, I didn't get hit by a meteor).  Long story short, I
> revisited (yet again) how to deal with enabling the mediated PMU.  I had been
> doing almost all of my testing with a hack to force-enable a mediated PMU, and
> when it came time to rip that out, I just couldn't convince myself that requiring
> userspace to enable KVM_CAP_PMU_CAPABILITY was the best behavior (I especially
> hated that PMU support would silently disappear).
> 
> So, bad news is, v5 isn't happening today.  Good news is that I think I figured
> out a not-awful solution for enabling the mediated PMU.  I'll post details (and
> hopefully v5) tomorrow.

Too hopeful yet again.  I realized at the last minute that I neglected to add
support for disabling PMU MSR interception for L2 after rebasing on the MSR
interception cleanups.  Adding the code back was easy, but I need to test it.

I'm buried in meetings tomorrow, but there's a small chance I'll be able to squeak
out v5.