linux-kernel - Re: [PATCH -v2] x86/boot/compressed: Register dummy NMI handler in EFI boot loader, to avoid kdump crashes

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]

Date:   Tue, 10 Jan 2023 20:32:07 +0800
From:   Zeng Heng <zengheng4@...wei.com>
To:     Borislav Petkov <bp@...en8.de>, Ingo Molnar <mingo@...nel.org>
CC:     <michael.roth@....com>, <hpa@...or.com>, <tglx@...utronix.de>,
        <sathyanarayanan.kuppuswamy@...ux.intel.com>,
        <kirill.shutemov@...ux.intel.com>, <jroedel@...e.de>,
        <keescook@...omium.org>, <mingo@...hat.com>,
        <dave.hansen@...ux.intel.com>, <brijesh.singh@....com>,
        <linux-kernel@...r.kernel.org>, <x86@...nel.org>,
        <liwei391@...wei.com>
Subject: Re: [PATCH -v2] x86/boot/compressed: Register dummy NMI handler in
 EFI boot loader, to avoid kdump crashes


On 2023/1/10 20:11, Borislav Petkov wrote:
> On Tue, Jan 10, 2023 at 01:01:06PM +0100, Ingo Molnar wrote:
>> From: Zeng Heng <zengheng4@...wei.com>
>> Date: Tue, 10 Jan 2023 18:27:45 +0800
>> Subject: [PATCH] x86/boot/compressed: Register dummy NMI handler in EFI boot loader, to avoid kdump crashes
>>
>> If kdump is enabled, when using mce_inject to inject errors, EFI
> Why does "EFI" matter here? Any boot loader would do...
>
>> boot loader would decompress & load second kernel for saving the
> s/&/and/
>
>> vmcore file.
>>
>> For normal errors that is fine.
> Useless sentence.
>
>> However, in the MCE case, the panic
>> CPU that firstly enters into mce_panic() is running within NMI
>> interrupt context,
> "#MC context" it is non-maskable but that's not "NMI interrupt context"

mce is registered on NMI handler by inject_init().

And here is the context of mce-inject:

#0  relocate_kernel () at arch/x86/kernel/relocate_kernel_64.S:55
#1  0xffffffff81a57fc2 in machine_kexec (image=0xffff888101ef8400)
     at arch/x86/kernel/machine_kexec_64.c:391
#2  0xffffffff811a9573 in __crash_kexec (regs=regs@...ry=0x0 
<fixed_percpu_data>)
     at kernel/kexec_core.c:1057
#3  0xffffffff81a5b4e4 in panic (fmt=fmt@...ry=0xffffffff823211c8 "Fatal 
machine check")
     at kernel/panic.c:393
#4  0xffffffff81aa65f5 in mce_panic (
     msg=msg@...ry=0xffffffff823211c8 "Fatal machine check",
     final=final@...ry=0xffff88813ac9eec0, exp=<optimized out>)
     at arch/x86/kernel/cpu/mce/core.c:380
#5  0xffffffff8103863b in mce_reign () at 
arch/x86/kernel/cpu/mce/core.c:1042
#6  0xffffffff81aa682f in mce_end (order=order@...ry=1)
     at arch/x86/kernel/cpu/mce/core.c:1175
#7  0xffffffff81aa6d57 in do_machine_check (regs=0xfffffe00000beef8)
     at arch/x86/kernel/cpu/mce/core.c:1567
#8  0xffffffffc0495167 in raise_exception (m=m@...ry=0xffff88813ad9ef40,
     pregs=<optimized out>) at arch/x86/kernel/cpu/mce/inject.c:152
#9  0xffffffffc0495e7f in mce_raise_notify (cmd=<optimized out>, 
regs=<optimized out>)
     at arch/x86/kernel/cpu/mce/inject.c:168
#10 0xffffffff810204b8 in nmi_handle ()
#11 0xffffffff81aa5e62 in default_do_nmi 
(regs=regs@...ry=0xfffffe00000beef8)
     at arch/x86/kernel/nmi.c:335
#12 0xffffffff81aa608d in exc_nmi (regs=0xfffffe00000beef8) at 
arch/x86/kernel/nmi.c:517
#13 0xffffffff81c014e8 in asm_exc_nmi () at arch/x86/entry/entry_64.S:1440


>
>> and the processor blocks delivery of subsequent
>> NMIs until the next execution of the IRET instruction.
>>
>> When the panic CPU takes long time in the panic processing route,
> I'm still unclear on the order of events here. It sounds like
>
> 1. MCE injected
> 2. panic
> 3. kdump gets loaded
>
> If that is the case, then I presume the flow is:
>
> mce_panic -> panic -> __crash_kexec()
>
> Yes?

Yes, exactly. The following procedure is like:

panic() -> relocate_kernel() -> identity_mapped() -> x86 purgatory image 
-> EFI loader -> secondary kernel

>
> If so, then we should make sure we have *exited* #MC context before calling
> panic() and not have to add hacks like this one of adding an empty NMI handler.
>
> But I'm only speculating as it is hard to make sense of all this text.
>
> Thx.
>