linux-kernel - Re: [PATCH 2/3] x86: add guest_late_init hook to hypervisor

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: <7ee9d814-9845-8145-d30a-9570f7cb43d8@suse.com>
Date:   Wed, 8 Nov 2017 10:49:28 +0100
From:   Juergen Gross <jgross@...e.com>
To:     Ingo Molnar <mingo@...nel.org>
Cc:     linux-kernel@...r.kernel.org, x86@...nel.org, kvm@...r.kernel.org,
        xen-devel@...ts.xenproject.org, tglx@...utronix.de,
        mingo@...hat.com, hpa@...or.com, boris.ostrovsky@...cle.com,
        pbonzini@...hat.com, rkrcmar@...hat.com, rjw@...ysocki.net,
        len.brown@...el.com, pavel@....cz
Subject: Re: [PATCH 2/3] x86: add guest_late_init hook to hypervisor_x86
 structure

On 08/11/17 10:40, Ingo Molnar wrote:
> 
> * Juergen Gross <jgross@...e.com> wrote:
> 
>>> Plus add a default empty function (which hypervisors can override). This avoids 
>>> all the hidden conditions and wrappery.
>>
>> Hmm, x86_hyper is just a pointer being NULL on bare metal. So we would
>> have to add a pre-filled struct with dummy functions and in case a
>> hypervisor is detected we'd need to copy all non-NULL pointers of the
>> hypervisor specific struct to the pre-filled one.
> 
> Ok, I missed that detail - but yeah, since this is mostly about boot code,
> where readability is king, I still think it would be an overall improvement.
> 
> This is the pattern we are trying to use with x86_platform ops for example, and 
> doing:
> 
>   git grep -w x86_platform arch/x86
> 
> gives a pretty clear idea about how it's used - while if it was all within 
> wrappers it would be a lot more difficult to discover all the callsites.
> 
> Admittedly it's not done totally consistently:
> 
>  arch/x86/kernel/apic/probe_32.c:        if (x86_platform.apic_post_init)
>  arch/x86/kernel/apic/probe_64.c:        if (x86_platform.apic_post_init)
>  arch/x86/kernel/ebda.c: if (!x86_platform.legacy.reserve_bios_regions)
>  arch/x86/kernel/platform-quirks.c:      if (x86_platform.set_legacy_features)
>  arch/x86/platform/intel-mid/device_libs/platform_mrfld_rtc.c:   if (!x86_platform.legacy.rtc)
> 
> ... but the _idea_ behind it is consistent ;-)
> 
>> In case there are no objections I can add a patch to modify the current
>> way x86_hyper is used to the pre-filled variant.
> 
> Yeah, sounds good to me!

Okay. With you mentioning x86_platform: should I make x86_hyper a member
of x86_platform (e.g. x86_platform.hyper.) ?

I think this would fit quite nice.


Juergen