Date:   Tue, 21 Feb 2023 09:21:53 -0800
From:   Krister Johansen <kjlx@templeofstupid.com>
To:     Thomas Gleixner <tglx@linutronix.de>
Cc:     Krister Johansen <kjlx@templeofstupid.com>,
        xen-devel@lists.xenproject.org, linux-kernel@vger.kernel.org,
        Juergen Gross <jgross@suse.com>,
        Jan Beulich <jbeulich@suse.com>,
        Boris Ostrovsky <boris.ostrovsky@oracle.com>,
        Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
        Dave Hansen <dave.hansen@linux.intel.com>, x86@kernel.org,
        "H. Peter Anvin" <hpa@zytor.com>,
        Marcelo Tosatti <mtosatti@redhat.com>,
        Anthony Liguori <aliguori@amazon.com>,
        David Reaver <me@davidreaver.com>,
        Brendan Gregg <brendan@intel.com>
Subject: Re: [PATCH linux-next 2/2] x86/xen/time: cleanup
 xen_tsc_safe_clocksource
Message-ID: <20230221172153.GA1958@templeofstupid.com>
References: <cover.1676610413.git.kjlx@templeofstupid.com>
 <f6bab47230b00cecb67f2c5d94716c8236609967.1676610413.git.kjlx@templeofstupid.com>
 <87h6vgov2p.ffs@tglx>
 <20230221041440.GA1934@templeofstupid.com>
 <20230221055117.GA1934@templeofstupid.com>
 <871qmjpegh.ffs@tglx>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <871qmjpegh.ffs@tglx>
Precedence: bulk

On Tue, Feb 21, 2023 at 10:14:54AM +0100, Thomas Gleixner wrote:
> On Mon, Feb 20 2023 at 21:51, Krister Johansen wrote:
> > On Mon, Feb 20, 2023 at 08:14:40PM -0800, Krister Johansen wrote:
> >> > static bool __init xen_tsc_safe_clocksource(void)
> >> > {
> >> > 	u32 eax, ebx. ecx, edx;
> >> >  
> >> > 	/* Leaf 4, sub-leaf 0 (0x40000x03) */
> >> > 	cpuid_count(xen_cpuid_base() + 3, 0, &eax, &ebx, &ecx, &edx);
> >> > 
> >> > 	return ebx == XEN_CPUID_TSC_MODE_NEVER_EMULATE;
> >> > }
> >> 
> >> I'm all for simplifying.  I'm happy to clean up that return to be more
> >> idiomatic.  I was under the impression, perhaps mistaken, though, that
> >> the X86_FEATURE_CONSTANT_TSC, X86_FEATURE_NONSTOP_TSC, and
> >> check_tsc_unstable() checks were actually serving a purpose: to ensure
> >> that we don't rely on the tsc in environments where it's being emulated
> >> and the OS would be better served by using a PV clock.  Specifically,
> >> kvmclock_init() makes a very similar set of checks that I also thought
> >> were load-bearing.
> >
> > Bah, what I meant to say was emulated, unstable, or otherwise unsuitable
> > for use as a clocksource.  IOW, even if TSC_MODE_NEVER_EMULATE is
> > set, it's possible that a user is attempting a migration from a cpu
> > that's not invariant, and we'd still want to check for that case and
> > fall back to a PV clocksource, correct?
> 
> Sure. But a life migration from a NEVER_EMULATE to a non-invariant host
> is a patently bad idea and has nothing to do with the __init function,
> which is gone at that point already.
> 
> What I wanted to say:
> 
> static bool __init xen_tsc_safe_clocksource(void)
> {
>         ......        
> 
> 	/* Leaf 4, sub-leaf 0 (0x40000x03) */
> 	cpuid_count(xen_cpuid_base() + 3, 0, &eax, &ebx, &ecx, &edx);
> 
> 	return ebx == XEN_CPUID_TSC_MODE_NEVER_EMULATE;
> }

Thanks, I'm happy to make it look like ^ that.  I should have thought to
do this myself. :/

I'll send out a v2 making this correction.

> I didn't have the full context and was just looking at the condition.
> Now I checked the full context and I think that except for the
> 
> 	if (check_tsc_unstable())
> 
> check everything else can go away unless you do not trust the hypervisor
> that it only sets the NEVER_EMULATE bit when CONSTANT and NONSTOP are
> set as well. But yeah, you might prefer to be paranoid. It's virt after
> all.

Unless there are objections, I think I'd prefer to err on the side of
paranoia here.  Sorry for the confusion.

-K