RE: [PATCH v4 2/5] x86/hyper-v: Add synthetic debugger definitions

From: Michael Kelley <mikelley@microsoft.com>
To: Jon Doron <arilou@gmail.com>
Cc: vkuznets <vkuznets@redhat.com>, Wei Liu <wei.liu@kernel.org>,
	"kvm@vger.kernel.org" <kvm@vger.kernel.org>,
	"linux-hyperv@vger.kernel.org" <linux-hyperv@vger.kernel.org>
Subject: RE: [PATCH v4 2/5] x86/hyper-v: Add synthetic debugger definitions
Date: Thu, 12 Mar 2020 19:32:46 +0000	[thread overview]
Message-ID: <MW2PR2101MB1052572531F218BF7DAFBA2AD7FD0@MW2PR2101MB1052.namprd21.prod.outlook.com> (raw)
In-Reply-To: <20200312191216.GA2950@jondnuc>

From: Jon Doron <arilou@gmail.com>  Sent: Thursday, March 12, 2020 12:12 PM
> 
> On 12/03/2020, Michael Kelley wrote:
> >From: Vitaly Kuznetsov <vkuznets@redhat.com>  Sent: Thursday, March 12, 2020 6:51 AM
> >>
> >> Michael Kelley <mikelley@microsoft.com> writes:
> >>
> >> > I'm flexible, and trying to not be a pain-in-the-neck. :-)  What would
> >> > the KVM guys think about putting the definitions in a KVM specific
> >> > #include file, and clearly marking them as deprecated, mostly
> >> > undocumented, and used only to support debugging old Windows
> >> > versions?
> >>
> >> I *think* we should do the following: defines which *are* present in
> >> TLFS doc (e.g. HV_FEATURE_DEBUG_MSRS_AVAILABLE,
> >> HV_STATUS_OPERATION_DENIED, ...) go to asm/hyperv-tlfs.h, the rest
> >> (syndbg) stuff goes to kvm-specific include (I'd suggest we just use
> >> hyperv.h we already have).
> >>
> >> What do you think?
> >>
> >
> >I could live with this proposal, since they *are* in the TLFS v6.0 as it
> >exists today. However, v6.0 seems inconsistent in what parts of this
> >debugging functionality it exposes, probably just because someone
> >hasn't thought comprehensively about the topic across the whole
> >document.   I'll make sure that it gets looked at in the next revision
> >(which should be a lot sooner that the 2+ years it took to get the v6.0
> >revision done).   But I won't be surprised if the remaining vestiges are
> >removed at that time, in which case we would want to move the
> >definitions from hyperv-tlfs.h to KVM's hyper.h.
> >
> >Michael
> 
> Hi guys, just a quick note I went over the old HyperV TLFS and it seems
> like all the Syndbg MSRs are documented (under Appendix F: Hypervisor
> Synthetic MSRs, from v5.0b).
> 
> It seems like the undocumented stuff is HV_X64_MSR_SYNDBG_OPTIONS which
> seems kinda odd because that's how you enable the hypercalls debugging
> interface which is documented.
> 
> And the syndbg CPUID leafs are not documented as well.
> 
> So would you like me to put all the MSRs in the tlfs omitting the
> HV_X64_MSR_SYNDBG_OPTIONS.
> 
> So in hyperv.h we will have HV_X64_MSR_SYNDBG_OPTIONS and the CPUID
> leafs.
> 

Could you make the decision based on the new v6.0 of the Hyper-V TLFS?
See https://docs.microsoft.com/en-us/virtualization/hyper-v-on-windows/reference/tlfs
to get a copy.  I think some of the synthetic debugger stuff has been dropped
from the v6.0 version compared with the earlier v5.0 versions, and I'd like the
updates to hyperv-tlfs.h to reflect that newest version.

Michael