From mboxrd@z Thu Jan 1 00:00:00 1970 From: Zhang Yanfei Subject: Re: [PATCH 0/3] x86: clear vmcss on all cpus when doing kdump if necessary Date: Thu, 18 Oct 2012 09:12:02 +0800 Message-ID: <507F5762.3060208@cn.fujitsu.com> References: <5077BB7B.4030203@cn.fujitsu.com> <507C2F18.8080904@redhat.com> <507E17C0.9070305@cn.fujitsu.com> <507E8575.4040908@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: x86@kernel.org, kexec@lists.infradead.org, linux-doc@vger.kernel.org, mtosatti@redhat.com, linux-kernel@vger.kernel.org, "kvm@vger.kernel.org" , "Hao, Xudong" , "Nakajima, Jun" To: Avi Kivity Return-path: In-Reply-To: <507E8575.4040908@redhat.com> Sender: linux-doc-owner@vger.kernel.org List-Id: kvm.vger.kernel.org =E4=BA=8E 2012=E5=B9=B410=E6=9C=8817=E6=97=A5 18:16, Avi Kivity =E5=86=99= =E9=81=93: > On 10/17/2012 04:28 AM, Zhang Yanfei wrote: >> =E4=BA=8E 2012=E5=B9=B410=E6=9C=8815=E6=97=A5 23:43, Avi Kivity =E5=86= =99=E9=81=93: >>> On 10/12/2012 08:40 AM, Zhang Yanfei wrote: >>>> Currently, kdump just makes all the logical processors leave VMX o= peration by >>>> executing VMXOFF instruction, so any VMCSs active on the logical p= rocessors may >>>> be corrupted. But, sometimes, we need the VMCSs to debug guest ima= ges contained >>>> in the host vmcore. To prevent the corruption, we should VMCLEAR t= he VMCSs before >>>> executing the VMXOFF instruction. >>> >>> How have you verified that VMXOFF doesn't flush cached VMCSs alread= y? >>> >> >> I tried some tests, for example, I made copies for every vmcs, and i= n the kdump >> path, I backed up all the loaded vmcs into the copies before vmxoff. >> After generating the vmcore, I retrieve the vmcss and their copies, = and compare them, >> no differences. >> >> Another test is using VMCLEAR to clear all the loaded vmcs before VM= XOFF, >> and compare the vmcss and their copies, there are indeed differences= between the >> vmcs and its copy. >> >> I know the tests may be not so convincing, for example, I used memcp= y to back up >> the vmcss and it is an ordinary memory operation. But to ensure the = non-corruption >> of the vmcss in the vmcore, I think we should VMCLEAR the vmcss befo= re VMXOFF just >> as the Intel spec says. >=20 > Sorry, I was unclear -- I was referring to the spec, I wasn't sure > whether VMXOFF is defined to flush VMCSes or whether it just invalida= tes > on-chip caches so that it won't flush them out in the future, corrupt= ing > memory. We don't want to depend on actual behaviour as it may change > with future version. >=20 > Copying some Intel folk, maybe they can clarify it. >=20 Yes, the Intel spec says "may be" about the VMCS-corruption thing. From chapter 24.10.1 in Intel=C2=AE 64 and IA-32 Architectures Software Deve= loper=E2=80=99s Manual Volume 3C:System Programming Guide, Part 3, there is the descrip= tion: "If a logical processor leaves VMX operation, any VMCSs active on that = logical processor may be corrupted (see below). To prevent such corruption of a= VMCS that may be used either after a return to VMX operation or on another logica= l processor, software should VMCLEAR that VMCS before executing the VMXOFF instructi= on or removing power from the processor (e.g., as part of a transition to the= S3 and S4 power states)." Our purpose is to make sure the VMCSs in the vmcore are updated and non= -corrupted. So according to the description above, no matter whether VMXOFF is defined= to flush VMCSs or whether it just invalidates on-chip caches, we'd better VMCLEA= R the VMCSs before executing the VMXOFF. Thanks Zhang Yanfei