Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Fri, 12 May 2023 16:30:25 -0700
In-Reply-To: <ZF7KbsCXBQHnOv7g@google.com>
Mime-Version: 1.0
References: <20230511235917.639770-1-seanjc@google.com> <20230511235917.639770-5-seanjc@google.com>
 <ZF7KbsCXBQHnOv7g@google.com>
Message-ID: <ZF7MES4qEKd8T6OW@google.com>
Subject: Re: [PATCH 4/9] KVM: x86/mmu: Rename MMU_WARN_ON() to KVM_MMU_WARN_ON()
From:   Sean Christopherson <seanjc@google.com>
To:     David Matlack <dmatlack@google.com>
Cc:     Paolo Bonzini <pbonzini@redhat.com>, kvm@vger.kernel.org,
        linux-kernel@vger.kernel.org, Mingwei Zhang <mizhang@google.com>,
        Jim Mattson <jmattson@google.com>
Content-Type: text/plain; charset="us-ascii"
Precedence: bulk

On Fri, May 12, 2023, David Matlack wrote:
> On Thu, May 11, 2023 at 04:59:12PM -0700, Sean Christopherson wrote:
> > Rename MMU_WARN_ON() to make it super obvious that the assertions are
> > all about KVM's MMU, not the primary MMU.
> 
> I think adding KVM is a step in the right direction but I have 2
> remaining problems with KVM_MMU_WARN_ON():
> 
>  - Reminds me of VM_WARN_ON(), which toggles between WARN_ON() and
>    BUG_ON(), whereas KVM_MMU_WARN_ON() toggles between no-op and
>    WARN_ON().

No, VM_WARN_ON() bounces between WARN_ON() and nop, just like KVM_MMU_WARN_ON().
There's an extra bit of magic that adds a static assert that the code is valid
(which I can/should/will add), but the runtime behavior is a nop.

  #define VM_WARN_ON(cond) (void)WARN_ON(cond)
  #else
  #define VM_WARN_ON(cond) BUILD_BUG_ON_INVALID(cond)

/*
 * BUILD_BUG_ON_INVALID() permits the compiler to check the validity of the
 * expression but avoids the generation of any code, even if that expression
 * has side-effects.
 */
#define BUILD_BUG_ON_INVALID(e) ((void)(sizeof((__force long)(e))))

>  - It's not obvious from the name that it's a no-op most of the time.
> 
> Naming is hard so I might just make things worse by trying but...
> 
> How about KVM_MMU_PROVE(condition). That directly pairs it with the new
> CONFIG_KVM_PROVE_MMU(), makes it sufficiently different from
> VM_WARN_ON() and WARN_ON() that readers will not make assumptions about
> what's happening under the hood. Also "PROVE" sounds like a high bar
> which conveys this might not always be enabled.

It inverts the checks though.  Contexting switching between "WARN_ON" and "ASSERT"
is hard enough, I don't want to add a third flavor.

> That also will allow us to convert this to a WARN_ON_ONCE() (my
> suggestion on the other patch) without having to make the name any
> longer.