Received-SPF: pass (google.com: domain of linux-kernel+bounces-67538-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:4601:e00::3 as permitted sender) client-ip=2604:1380:4601:e00::3;
Date: Thu, 15 Feb 2024 10:50:20 -0800
In-Reply-To: <Zc3JcNVhghB0Chlz@linux.dev>
Precedence: bulk
Mime-Version: 1.0
References: <20240215010004.1456078-1-seanjc@google.com> <20240215010004.1456078-3-seanjc@google.com>
 <Zc3JcNVhghB0Chlz@linux.dev>
Message-ID: <Zc5c7Af-N71_RYq0@google.com>
Subject: Re: [PATCH 2/2] KVM: selftests: Test forced instruction emulation in
 dirty log test (x86 only)
From: Sean Christopherson <seanjc@google.com>
To: Oliver Upton <oliver.upton@linux.dev>
Cc: Paolo Bonzini <pbonzini@redhat.com>, kvm@vger.kernel.org, linux-kernel@vger.kernel.org, 
	David Matlack <dmatlack@google.com>, Pasha Tatashin <tatashin@google.com>, 
	Michael Krebs <mkrebs@google.com>
Content-Type: text/plain; charset="us-ascii"

On Thu, Feb 15, 2024, Oliver Upton wrote:
> On Wed, Feb 14, 2024 at 05:00:04PM -0800, Sean Christopherson wrote:
> > Add forced emulation of MOV and LOCK CMPXCHG instructions in the dirty log
> > test's guest code to verify that KVM's emulator marks pages dirty as
> > expected (and obviously to verify the emulator works at all).  In the long
> > term, the guest code would ideally hammer more of KVM's emulator, but for
> > starters, cover the two major paths: writes and atomics.
> > 
> > To minimize #ifdeffery, wrap only the related code that is x86 specific,
> > unnecessariliy synchronizing an extra boolean to the guest is far from the
> > end of the world.
> 
> Meh, I wouldn't say the end result in guest_write_memory() is that
> pretty. Just ifdef the whole function and provide a generic implementation
> for the other architectures.
> 
> > Signed-off-by: Sean Christopherson <seanjc@google.com>
> > ---
> >  tools/testing/selftests/kvm/dirty_log_test.c | 36 ++++++++++++++++++--
> >  1 file changed, 33 insertions(+), 3 deletions(-)
> > 
> > diff --git a/tools/testing/selftests/kvm/dirty_log_test.c b/tools/testing/selftests/kvm/dirty_log_test.c
> > index eaad5b20854c..ff1d1c7f05d8 100644
> > --- a/tools/testing/selftests/kvm/dirty_log_test.c
> > +++ b/tools/testing/selftests/kvm/dirty_log_test.c
> > @@ -92,6 +92,29 @@ static uint64_t guest_test_phys_mem;
> >   */
> >  static uint64_t guest_test_virt_mem = DEFAULT_GUEST_TEST_MEM;
> >  
> > +static bool is_forced_emulation_enabled;
> > +
> > +static void guest_write_memory(uint64_t *mem, uint64_t val, uint64_t rand)
> > +{
> > +#ifdef __x86_64__
> > +	if (is_forced_emulation_enabled && (rand & 1)) {
> > +		if (rand & 2) {
> 
> Can't you invert the logic and drop a level of indentation?
> 
> 	if (!(is_forced_emulation_enabled && (rand & 1))) {
> 		*mem = val;
> 	} else if (rand & 2) {
> 		movq
> 	} else {
> 		cmpxchg8b
> 	}

Yeah, the funky flow I concocted was done purely to have the "no emulation" path
fall through to the common "*mem = val".  I don't have a strong preference, I
mentally flipped a coin on doing that versus what you suggested, and apparently
chose poorly :-)