Re: [Xen-devel] [PATCH] x86/viridian: Reword HV_X64_MSR_CRASH_CTL print message

[Wei Liu <wl@xxxxxxx>]

On Tue, 17 Sep 2019 at 17:31, Andrew Cooper <andrew.cooper3@xxxxxxxxxx> wrote:
>
> On 16/09/2019 14:56, Paul Durrant wrote:
> >> -----Original Message-----
> >> From: Wei Liu <wl@xxxxxxx>
> >> Sent: 16 September 2019 14:29
> >> To: Andrew Cooper <Andrew.Cooper3@xxxxxxxxxx>
> >> Cc: Paul Durrant <Paul.Durrant@xxxxxxxxxx>; Xen-devel 
> >> <xen-devel@xxxxxxxxxxxxxxxxxxxx>; Jan Beulich
> >> <JBeulich@xxxxxxxx>; Wei Liu <wl@xxxxxxx>; Roger Pau Monne 
> >> <roger.pau@xxxxxxxxxx>
> >> Subject: Re: [PATCH] x86/viridian: Reword HV_X64_MSR_CRASH_CTL print 
> >> message
> >>
> >> On Mon, 16 Sep 2019 at 14:13, Andrew Cooper <andrew.cooper3@xxxxxxxxxx> 
> >> wrote:
> >> [...]
> >>>>>>> Replace the VIRIDIAN prefix with 'reported' to reduce the confusion to
> >>>>>>> non-xen-developers trying to interpret the message.
> >>>>>> This is a message that is peculiar to Windows VMs, so how about 
> >>>>>> "Windows VM CRASH"?
> >>>>> I presume you mean particular, but no - it isn't windows which is the
> >>>>> exclusive user of this interface.  Linux has a driver to use it when
> >>>>> running under HyperV.
> >>>> Hmm, that's a bit odd. I thought the crash codes are Windows specific. 
> >>>> Perhaps they can be
> >> distinguished in some way.
> >>> Linux sets its own guest OS identifier, which is some function of
> >>> HV_LINUX_VENDOR_ID.
> >>>
> >>>>  All the same, the log line needs to lead people to some way of decoding 
> >>>> the magic numbers I
> >> think. How about:
> >>>> "VIRIDIAN REPORTED CRASH"
> >>>>
> >>>> ?
> >>> That is still just as confusing to read.
> >>>
> >>> There is no way to decode the numbers without knowing what OS is
> >>> running, and simply saying "Viridian" doesn't help with that.
> >>>
> >> Would it make sense to call dump_guest_os_id here as well? Seeing that
> >> it is only printed when it was first set.
> > Yes, that's not a bad idea.
>
> This is as maybe, but still doesn't help with ambiguity because you
> can't expect people to recognise guest-id's by their hex value.  It also
> doesn't help with the confusion of having the word viridian in the
> printed string.
>

I look closer today. They can be interpreted at least for Linux. It is
useful information to have. If we want to print crash control values
at all, OS information shouldn't be omitted.

I don't really have an opinion on whether the viridian prefix is useful or not.

Wei.

> ~Andrew

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel