Received-SPF: pass (google.com: domain of linux-kernel+bounces-67892-linux.lists.archive=gmail.com@vger.kernel.org designates 139.178.88.99 as permitted sender) client-ip=139.178.88.99;
Date: Thu, 15 Feb 2024 16:26:02 -0800
In-Reply-To: <Zc6d6fwakreoVtN5@linux.dev>
Precedence: bulk
Mime-Version: 1.0
References: <20240215010004.1456078-1-seanjc@google.com> <20240215010004.1456078-3-seanjc@google.com>
 <Zc3JcNVhghB0Chlz@linux.dev> <Zc5c7Af-N71_RYq0@google.com>
 <Zc5wTHuphbg3peZ9@linux.dev> <Zc6DPEWcHh-TKCSD@google.com> <Zc6d6fwakreoVtN5@linux.dev>
Message-ID: <Zc6rmksmgZ31fd-K@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 Thu, Feb 15, 2024 at 01:33:48PM -0800, Sean Christopherson wrote:
> 
> [...]
> 
> > +/* TODO: Expand this madness to also support u8, u16, and u32 operands. */
> > +#define vcpu_arch_put_guest(mem, val, rand) 						\
> > +do {											\
> > +	if (!is_forced_emulation_enabled || !(rand & 1)) {				\
> > +		*mem = val;								\
> > +	} else if (rand & 2) {								\
> > +		__asm__ __volatile__(KVM_FEP "movq %1, %0"				\
> > +				     : "+m" (*mem)					\
> > +				     : "r" (val) : "memory");				\
> > +	} else {									\
> > +		uint64_t __old = READ_ONCE(*mem);					\
> > +											\
> > +		__asm__ __volatile__(KVM_FEP LOCK_PREFIX "cmpxchgq %[new], %[ptr]"	\
> > +				     : [ptr] "+m" (*mem), [old] "+a" (__old)		\
> > +				     : [new]"r" (val) : "memory", "cc");		\
> > +	}										\
> > +} while (0)
> > +
> 
> Last bit of bikeshedding then I'll go... Can you just use a C function
> and #define it so you can still do ifdeffery to slam in a default
> implementation?

Yes, but the macro shenanigans aren't to create a default, they're to set the
stage for expanding to other sizes without having to do:

  vcpu_arch_put_guest{8,16,32,64}()

or if we like bytes instead of bits:

  vcpu_arch_put_guest{1,2,4,8}()

I'm not completely against that approach; it's not _that_ much copy+paste
boilerplate, but it's enough that I think that macros would be a clear win,
especially if we want to expand what instructions are used.

<me fiddles around>

Actually, I take that back, I am against that approach :-)

I was expecting to have to do some switch() explosion to get the CMPXCHG stuff
working, but I'm pretty sure the mess that is the kernel's unsafe_try_cmpxchg_user()
and __put_user_size() is is almost entirely due to needing to support 32-bit kernels,
or maybe some need to strictly control the asm constraints.

For selftests, AFAICT the below Just Works on gcc and clang for legal sizes.  And
as a bonus, we can sanity check that the pointer and value are of the same size.
Which we definitely should do, otherwise the compiler has a nasty habit of using
the size of the value of the right hand side for the asm blobs, e.g. this

	vcpu_arch_put_guest((u8 *)addr, (u32)val, rand);

generates 32-bit accesses.  Oof.

#define vcpu_arch_put_guest(mem, val, rand) 					\
do {										\
	kvm_static_assert(sizeof(*mem) == sizeof(val));				\
	if (!is_forced_emulation_enabled || !(rand & 1)) {			\
		*mem = val;							\
	} else if (rand & 2) {							\
		__asm__ __volatile__(KVM_FEP "mov %1, %0"			\
				     : "+m" (*mem)				\
				     : "r" (val) : "memory");			\
	} else {								\
		uint64_t __old = READ_ONCE(*mem);				\
										\
		__asm__ __volatile__(LOCK_PREFIX "cmpxchg %[new], %[ptr]"	\
				     : [ptr] "+m" (*mem), [old] "+a" (__old)	\
				     : [new]"r" (val) : "memory", "cc");	\
	}									\
} while (0)