Subject: Re: [RFC PATCH] x86: Make sure verify_cpu has a good stack
To: Borislav Petkov <bp@alien8.de>
References: <1456917606-4979-1-git-send-email-bp@alien8.de>
 <56D70CE2.1050500@nextfour.com> <20160302161532.GK16954@pd.tnic>
 <56D716F7.3080407@nextfour.com> <20160302165516.GC8362@pd.tnic>
CC: "H. Peter Anvin" <hpa@zytor.com>, X86 ML <x86@kernel.org>,
        LKML <linux-kernel@vger.kernel.org>,
        Tom Lendacky <thomas.lendacky@amd.com>
From: =?UTF-8?Q?Mika_Penttil=c3=a4?= <mika.penttila@nextfour.com>
Message-ID: <56D7266B.9030009@nextfour.com>
Date: Wed, 2 Mar 2016 19:44:11 +0200
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101
 Thunderbird/38.5.1
MIME-Version: 1.0
In-Reply-To: <20160302165516.GC8362@pd.tnic>
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: 8bit
SpamDiagnosticOutput: 1:23
SpamDiagnosticMetadata: NSPM
X-MS-Exchange-CrossTenant-OriginalArrivalTime: 02 Mar 2016 17:44:15.9764 (UTC)
X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted
X-MS-Exchange-Transport-CrossTenantHeadersStamped: VI1PR07MB0910
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org
Content-Length: 1523
Lines: 57


On 02.03.2016 18:55, Borislav Petkov wrote:
> On Wed, Mar 02, 2016 at 06:38:15PM +0200, Mika Penttilä wrote:
>> I actually looked at it a while too...
>>
>> The
>>   movq stack_start - __START_KERNEL_map, %rsp
>>
>> turns into (objdump disassembly)
>>
>>   mov    0x0,%rsp
>>
>> with relocation
>> 0000000000000004 R_X86_64_32S      stack_start+0x0000000080000000
>>
>> Now stack_start is at ffffffff81ef3380, so the relocation gives 1ef3380 which would be correct, so why the
>> second subq ?
>>
>> You may explain :)
> Here it is :-)
>
> $ readelf -a vmlinux | grep stack_start
>  70526: ffffffff81cbabf8     0 NOTYPE  GLOBAL DEFAULT   14 stack_start
>
> 0xffffffff81cbabf8 - __START_KERNEL_map =
> 0xffffffff81cbabf8 - 0xffffffff80000000 =
> 0x1cbabf8
>
> (gdb) x/x 0x1cbabf8
> 0x1cbabf8:      0xffffffff81c03ff8
>
> (You don't need gdb for that - you can hexdump or objdump vmlinux).
>
> Now stack_start is:
>
>         GLOBAL(stack_start)
>         .quad  init_thread_union+THREAD_SIZE-8
>
> which is
>
> $ readelf -a vmlinux | grep init_thread_union
>  82491: ffffffff81c00000 16384 OBJECT  GLOBAL DEFAULT   14 init_thread_union
>
> so init_thread_union+THREAD_SIZE-8 = 0xffffffff81c00000 + 4*4096-8 = 0xffffffff81c03ff8
>
> So you have to subtract __START_KERNEL_map again because it has there a
> virtual address and we haven't enabled paging yet:
>
> 0xffffffff81c03ff8 - 0xffffffff80000000 = 0x1c03ff8.
>
> Makes sense?
>
Ah missed completely that stack_start is effectively a pointer to stack..

Thanks,
Mika