Received-SPF: pass (google.com: domain of linux-crypto+bounces-1411-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.48.161 as permitted sender) client-ip=147.75.48.161;
Date: Fri, 12 Jan 2024 20:48:49 +0100
From: Borislav Petkov <bp@alien8.de>
To: Michael Roth <michael.roth@amd.com>
Cc: x86@kernel.org, kvm@vger.kernel.org, linux-coco@lists.linux.dev,
	linux-mm@kvack.org, linux-crypto@vger.kernel.org,
	linux-kernel@vger.kernel.org, tglx@linutronix.de, mingo@redhat.com,
	jroedel@suse.de, thomas.lendacky@amd.com, hpa@zytor.com,
	ardb@kernel.org, pbonzini@redhat.com, seanjc@google.com,
	vkuznets@redhat.com, jmattson@google.com, luto@kernel.org,
	dave.hansen@linux.intel.com, slp@redhat.com, pgonda@google.com,
	peterz@infradead.org, srinivas.pandruvada@linux.intel.com,
	rientjes@google.com, tobin@ibm.com, vbabka@suse.cz,
	kirill@shutemov.name, ak@linux.intel.com, tony.luck@intel.com,
	sathyanarayanan.kuppuswamy@linux.intel.com, alpergun@google.com,
	jarkko@kernel.org, ashish.kalra@amd.com, nikunj.dadhania@amd.com,
	pankaj.gupta@amd.com, liam.merwick@oracle.com, zhi.a.wang@intel.com,
	Brijesh Singh <brijesh.singh@amd.com>
Subject: Re: [PATCH v1 11/26] x86/sev: Invalidate pages from the direct map
 when adding them to the RMP table
Message-ID: <20240112194849.GDZaGXoTeYkx3GYSsQ@fat_crate.local>
References: <20231230161954.569267-1-michael.roth@amd.com>
 <20231230161954.569267-12-michael.roth@amd.com>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
In-Reply-To: <20231230161954.569267-12-michael.roth@amd.com>

On Sat, Dec 30, 2023 at 10:19:39AM -0600, Michael Roth wrote:
>  static int rmpupdate(u64 pfn, struct rmp_state *state)
>  {
>  	unsigned long paddr = pfn << PAGE_SHIFT;
> -	int ret;
> +	int ret, level, npages;
>  
>  	if (!cpu_feature_enabled(X86_FEATURE_SEV_SNP))
>  		return -ENODEV;
>  
> +	level = RMP_TO_PG_LEVEL(state->pagesize);
> +	npages = page_level_size(level) / PAGE_SIZE;
> +
> +	/*
> +	 * If the kernel uses a 2MB directmap mapping to write to an address,
> +	 * and that 2MB range happens to contain a 4KB page that set to private
> +	 * in the RMP table, an RMP #PF will trigger and cause a host crash.
> +	 *
> +	 * Prevent this by removing pages from the directmap prior to setting
> +	 * them as private in the RMP table.
> +	 */
> +	if (state->assigned && invalidate_direct_map(pfn, npages))
> +		return -EFAULT;

Well, the comment says one thing but the code does not do quite what the
text says:

* where is the check that pfn is part of the kernel direct map?

* where is the check that this address is part of a 2M PMD in the direct
map so that the situation is even given that you can have a 4K private
PTE there?

What this does is simply invalidate @npages unconditionally.

Then, __set_pages_np() already takes a number of pages and a start
address. Why is this thing iterating instead of sending *all* npages in
one go?

Yes, you'd need two new helpers, something like:

set_pages_present(struct page *page, int numpages)
set_pages_non_present(struct page *page, int numpages)

which call the lowlevel counterparts.

For some reason I remember asking this a while ago...

-- 
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette