Received-SPF: pass (google.com: domain of linux-kernel+bounces-59389-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.199.223 as permitted sender) client-ip=147.75.199.223;
Precedence: bulk
MIME-Version: 1.0
References: <20240129180502.4069817-21-ardb+git@google.com>
 <20240129180502.4069817-27-ardb+git@google.com> <20240207132922.GSZcOFspSGaVluJo92@fat_crate.local>
In-Reply-To: <20240207132922.GSZcOFspSGaVluJo92@fat_crate.local>
From: Ard Biesheuvel <ardb@kernel.org>
Date: Fri, 9 Feb 2024 13:55:02 +0000
Message-ID: <CAMj1kXF+mHCYs08q58QFGuzZ4nzGd2sDr1gp2ydkOHHQ2LK5tQ@mail.gmail.com>
Subject: Re: [PATCH v3 06/19] x86/startup_64: Drop global variables keeping
 track of LA57 state
To: Borislav Petkov <bp@alien8.de>
Cc: Ard Biesheuvel <ardb+git@google.com>, linux-kernel@vger.kernel.org, 
	Kevin Loughlin <kevinloughlin@google.com>, Tom Lendacky <thomas.lendacky@amd.com>, 
	Dionna Glaze <dionnaglaze@google.com>, Thomas Gleixner <tglx@linutronix.de>, 
	Ingo Molnar <mingo@redhat.com>, Dave Hansen <dave.hansen@linux.intel.com>, 
	Andy Lutomirski <luto@kernel.org>, Arnd Bergmann <arnd@arndb.de>, Nathan Chancellor <nathan@kernel.org>, 
	Nick Desaulniers <ndesaulniers@google.com>, Justin Stitt <justinstitt@google.com>, 
	Kees Cook <keescook@chromium.org>, Brian Gerst <brgerst@gmail.com>, linux-arch@vger.kernel.org, 
	llvm@lists.linux.dev
Content-Type: text/plain; charset="UTF-8"

On Wed, 7 Feb 2024 at 13:29, Borislav Petkov <bp@alien8.de> wrote:
>
> On Mon, Jan 29, 2024 at 07:05:09PM +0100, Ard Biesheuvel wrote:
> >  static inline bool pgtable_l5_enabled(void)
> >  {
> >       return __pgtable_l5_enabled;
> >  }
> >  #else
> > -#define pgtable_l5_enabled() cpu_feature_enabled(X86_FEATURE_LA57)
> > +#define pgtable_l5_enabled() !!(native_read_cr4() & X86_CR4_LA57)
> >  #endif /* USE_EARLY_PGTABLE_L5 */
>
> Can we drop this ifdeffery and simply have __pgtable_l5_enabled always
> present and contain the correct value?
>

I was trying to get rid of global variable assignments and accesses
from the 1:1 mapping, but since we cannot get rid of those entirely,
we might just keep __pgtable_l5_enabled but use RIP_REL_REF() in the
accessors, and move the assignment to the asm startup code.

> So that we don't have an expensive CR4 read hidden in
> pgtable_l5_enabled()?
>

Yeah, I didn't realize it was expensive. Alternatively, we might do
something like

static __always_inline bool pgtable_l5_enabled(void)
{
   unsigned long r;
   bool ret;

   asm(ALTERNATIVE_TERNARY(
       "movq %%cr4, %[reg] \n\t btl %[la57], %k[reg]" CC_SET(c),
       %P[feat], "stc", "clc")
       : [reg] "=r" (r), CC_OUT(c) (ret)
       : [feat] "i" (X86_FEATURE_LA57),
         [la57] "i" (X86_CR4_LA57_BIT)
       : "cc");
   return ret;
}

but we'd still have two versions in that case.

> For the sake of simplicity, pgtable_l5_enabled() can be defined outside
> of CONFIG_X86_5LEVEL and since both vendors support 5level now, might as
> well start dropping the CONFIG ifdeffery slowly...
>
> Other than that - a nice cleanup!
>

Thanks.