Re: [PATCH] hyperv: make HYPERV a menuconfig to ease disabling it all
From: Stephen Hemminger
Date: Sat Dec 16 2017 - 12:51:28 EST
On Thu, 14 Dec 2017 09:43:48 -0800
Randy Dunlap <rdunlap@xxxxxxxxxxxxx> wrote:
> On 12/13/2017 11:51 PM, Greg KH wrote:
> > On Wed, Dec 13, 2017 at 01:23:38PM -0800, Stephen Hemminger wrote:
> >> On Wed, 13 Dec 2017 09:54:19 +0100
> >> Vincent Legoll <vincent.legoll@xxxxxxxxx> wrote:
> >>
> >>> Hello,
> >>>
> >>> On Sun, Dec 10, 2017 at 6:50 AM, Stephen Hemminger
> >>> <stephen@xxxxxxxxxxxxxxxxxx> wrote:
> >>>> Will this break existing configs?
> >>>
> >>> I don't think so. Last time I did some similar changes, the kbuild
> >>> test robot found some warnings on some configurations, I hope
> >>> it will find problems (if any) for that series too (this one is not alone,
> >>> I've got a bunch of other similar patches in-flight)
> >>>
> >>> Thanks
> >>
> >> NAK
> >>
> >> Let me give a concrete example of how this will break users.
> >>
> >> 1. Assume user has a working .config file in their kernel build directory
> >> which builds a kernel that works on Hyper-V.
> >>
> >> 2. Add your patch (or assume it makes into a later version).
> >>
> >> 3. User then does
> >>
> >> $ make oldconfig
> >> scripts/kconfig/conf --oldconfig Kconfig
> >> *
> >> * Restart config...
> >> *
> >> *
> >> * Microsoft Hyper-V guest support
> >> *
> >> Microsoft Hyper-V guest support (HYPERV_MENU) [N/y] (NEW)
> >>
> >> If they hit return, the default value is not enabling HyperV and they
> >> will then go on to build a kernel that will not boot on your system.
> >>
> >> The default MUST be set to Yes.
>
> That should work.
>
> > Or you can just not take these types of odd and silly changes to the
> > Kconfig files, and leave it as-is. I have yet to see the good reason
> > why these are needed at all.
>
> Some of us would like to be able to disable many like drivers at one time
> instead of having to go down a list of say 20-30 drivers and disable them
> one at a time.
>
It makes sense to organize the config if you dont break old configs.
It would be more logical to group and treat all para-virtualized guest
support in same way. Hyper-V should be next to KVM and Xen.