linux-kernel - Re: [PATCH 2/3] x86/cpu: fix unbootable VMs by inlining memcmp in hypervisor_cpuid

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <824e522d-c9e2-4e24-8ce6-aca6573db06a@suse.com>
Date: Fri, 2 Aug 2024 16:25:56 +0300
From: Nikolay Borisov <nik.borisov@...e.com>
To: Alexey Dobriyan <adobriyan@...il.com>,
 Thomas Gleixner <tglx@...utronix.de>, Ingo Molnar <mingo@...hat.com>,
 Borislav Petkov <bp@...en8.de>, Dave Hansen <dave.hansen@...ux.intel.com>
Cc: xen-devel@...ts.xenproject.org, linux-kernel@...r.kernel.org,
 x86@...nel.org, "H. Peter Anvin" <hpa@...or.com>,
 Juergen Gross <jgross@...e.com>, Boris Ostrovsky <boris.ostrovsky@...cle.com>
Subject: Re: [PATCH 2/3] x86/cpu: fix unbootable VMs by inlining memcmp in
 hypervisor_cpuid_base

On 2.08.24 г. 11:50 ч., Alexey Dobriyan wrote:
> If this memcmp() is not inlined then PVH early boot code can call
> into KASAN-instrumented memcmp() which results in unbootable VMs:
> 
> 	pvh_start_xen
> 	xen_prepare_pvh
> 	xen_cpuid_base
> 	hypervisor_cpuid_base
> 	memcmp
> 
> Ubuntu's gcc version 11.4.0 (Ubuntu 11.4.0-1ubuntu1~22.04) inlines
> memcmp with patch and the bug is partially fixed.
> 
> Leave FIXME just in case someone cares enough to compare 3 pairs of
> integers like 3 pairs of integers.
> 
> Signed-off-by: Alexey Dobriyan <adobriyan@...il.com>
> ---
> 
>   arch/x86/include/asm/cpuid.h | 15 ++++++++++++++-
>   1 file changed, 14 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/x86/include/asm/cpuid.h b/arch/x86/include/asm/cpuid.h
> index 6b122a31da06..3eca7824430e 100644
> --- a/arch/x86/include/asm/cpuid.h
> +++ b/arch/x86/include/asm/cpuid.h
> @@ -196,7 +196,20 @@ static inline uint32_t hypervisor_cpuid_base(const char *sig, uint32_t leaves)
>   	for_each_possible_hypervisor_cpuid_base(base) {
>   		cpuid(base, &eax, &signature[0], &signature[1], &signature[2]);
>   
> -		if (!memcmp(sig, signature, 12) &&
> +		/*
> +		 * FIXME rewrite cpuid comparators to accept uint32_t[3].
> +		 *
> +		 * This memcmp()
> +		 * a) is called from PVH early boot code
> +		 *    before instrumentation is set up,
> +		 * b) may be compiled to "call memcmp" (not inlined),
> +		 * c) memcmp() itself may be instrumented.
> +		 *
> +		 * Any combination of 2 is fine, but all 3 aren't.
> +		 *
> +		 * Force inline this function call.
> +		 */
> +		if (!__builtin_memcmp(sig, signature, 12) &&

Instead of putting this giant FIXME, why not simply do the comparison as 
ints, i.e ((uint32_t)&sig[0]) == signature1 && ((uitn32_t)&sig[4]) == 
signature2 && ((uint32_t)&sig[8] == signature_3  and be done with it?

>   		    (leaves == 0 || ((eax - base) >= leaves)))
>   			return base;
>   	}
>