Re: [PATCH] arch/x86/kernel/cpu/microcode/intel: don't store initrd's start

[Nicolai Stange <nicstange@gmail.com>]

Borislav Petkov <bp@alien8.de> writes:

> On Sun, Jul 24, 2016 at 05:05:49PM +0200, Nicolai Stange wrote:
>> On x86_64 with CONFIG_RANDOMIZE_MEMORY and CONFIG_MICROCODE_INTEL,
>> I get the following splat upon booting on an Intel i7-4800MQ:
>
> Before we dive into this, let's establish which tree you're exactly
> testing...

I tested on linux-next-20160722 (I wrote this below the '---' marker).

>
>> 
>>   Call Trace:
>>    [<ffffffffb6054cea>] ? find_microcode_patch+0x4a/0xa0
>>    [<ffffffffb6055487>] load_microcode.isra.1.constprop.12+0x37/0xa0
>>    [...]
>>    [<ffffffffb60557cd>] load_ucode_intel_ap+0x5d/0x80
>>    [<ffffffffb6054924>] load_ucode_ap+0x94/0xa0
>>    [<ffffffffb60481a8>] cpu_init+0x58/0x3e0
>>    [<ffffffffb60709bc>] ? set_pte_vaddr+0x5c/0x90
>>    [<ffffffffb6fac06c>] trap_init+0x2b6/0x328
>>    [<ffffffffb6fa0dba>] start_kernel+0x224/0x47f
>>    [<ffffffffb6fa0120>] ? early_idt_handler_array+0x120/0x120
>>    [<ffffffffb6fa02cf>] x86_64_start_reservations+0x29/0x2b
>>    [<ffffffffb6fa041e>] x86_64_start_kernel+0x14d/0x170
>>   [...]
>>   RIP  [<ffffffffb6055af5>] has_newer_microcode+0x5/0x20
>>    [...]
>>   ---[ end trace b163fd3960fd46fb ]---
>>   Kernel panic - not syncing: Attempted to kill the idle task!
>>   ---[ end Kernel panic - not syncing: Attempted to kill the idle task!
>> 
>> It can be bisected to commit 21ef9a5c3164 ("Merge branch 'x86/microcode'").
>
> This commit 21ef9a5c3164 does not exist - the current merge commit
> corresponding to this is
>
> commit 2a2a745e2d248498a21ba3876d2481e90c7fe0a5
> Merge: caaaa222620c eb06158ee145
> Author: Ingo Molnar <mingo@kernel.org>
> Date:   Sun Jul 24 08:26:41 2016 +0200
>
>     Merge branch 'x86/microcode'
>
>> Both of its parents, i.e.
>>   commit f5846c92b0a5 ("Merge branch 'x86/headers'")
>
> This commit does not exist either.
>
>> and
>>   commit eb06158ee145 ("x86/microcode: Remove unused symbol exports")
>
> This one does.
>
>> work fine by themselves.
>> "Cross-bisecting" between v4.7-rc6..f5846c92b0a5 and v4.7-rc6..eb06158ee145
>> reveals the conflicting commits:
>>   commit 021182e52fe0 ("x86/mm: Enable KASLR for physical mapping memory
>>                         regions")
>> and
>>   commit 6c5456474e7f ("x86/microcode: Fix loading precedence").
>
> Those two too.
>
> So it seems you're either testing an older tip/master or some linux-next.

See above.

> So let's clarify what exactly you're testing before we do anything else.

If you want me to try any particular tree, please just tell me which one.

Thanks,

Nicolai