Received-SPF: pass (google.com: domain of linux-kernel+bounces-178228-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.80.249 as permitted sender) client-ip=147.75.80.249;
Precedence: bulk
MIME-Version: 1.0
References: <ZkG4cMe1TFI5w7sc@gmail.com>
In-Reply-To: <ZkG4cMe1TFI5w7sc@gmail.com>
From: Linus Torvalds <torvalds@linux-foundation.org>
Date: Mon, 13 May 2024 18:02:57 -0700
Message-ID: <CAHk-=wgf=jwo1HZjQN7UeFw7iYPtQ_i0ri7JNOci+7Rn5-pDcg@mail.gmail.com>
Subject: Re: [GIT PULL] x86/boot changes for v6.10
To: Ingo Molnar <mingo@kernel.org>
Cc: linux-kernel@vger.kernel.org, "the arch/x86 maintainers" <x86@kernel.org>, 
	Ard Biesheuvel <ardb@kernel.org>, "H. Peter Anvin" <hpa@zytor.com>
Content-Type: text/plain; charset="UTF-8"

On Sun, 12 May 2024 at 23:51, Ingo Molnar <mingo@kernel.org> wrote:
>
>  - Re-introduce a bootloader quirk wrt. CR4 handling

I've pulled this, but shouldn't the compressed boot also just stop
setting the G flag that it didn't understand?

For example, arch/x86/kernel/head64.c seems to do this:

        pmd_entry = __PAGE_KERNEL_LARGE_EXEC & ~_PAGE_GLOBAL;

but arch/x86/boot/compressed/ident_map_64.c does the somewhat suspect

        mapping_info.page_flag = __PAGE_KERNEL_LARGE_EXEC | sme_me_mask;

without masking off _PAGE_GLOBAL.

The hibernation code does

        pgprot_t pmd_text_prot = __pgprot(__PAGE_KERNEL_LARGE_EXEC);
        pgprot_val(pmd_text_prot) &= __default_kernel_pte_mask;

and again there are several situations where __default_kernel_pte_mask
does not have _PAGE_GLOBAL.

So again, the boot/compressed code seems a bit at an odds with other
code paths. The cr4 games seem to work around the fact that this code
is just buggy.

Hmm?

          Linus