Re: [Qemu-devel] [PATCH v2 1/4] hostmem-memfd: disable for systems wihtout sealing support

[Eduardo Habkost]

On Wed, Jan 16, 2019 at 03:48:57PM +0000, Daniel P. Berrangé wrote:
> On Wed, Jan 16, 2019 at 06:46:39PM +0300, Ilya Maximets wrote:
> > 
> > 
> > On 16.01.2019 18:30, Eduardo Habkost wrote:
> > > On Wed, Dec 12, 2018 at 07:49:36AM +0100, Gerd Hoffmann wrote:
> > >> On Tue, Dec 11, 2018 at 02:09:11PM +0300, Ilya Maximets wrote:
> > >>> On 11.12.2018 13:53, Daniel P. Berrangé wrote:
> > >>>>>
> > >>>>> Let's restrict memfd backend to systems with sealing support.
> > >>>>
> > >>>> I don't think we need todo that - sealing is optional in the QEMU code,
> > >>>> we simply have it set to the wrong default when sealing is not 
> > >>>> available.
> > >>>
> > >>> That was literally what I've fixed in v1:
> > >>>     
> > >>> https://lists.nongnu.org/archive/html/qemu-devel/2018-11/msg05483.html
> > >>>
> > >>> but 2 people suggested me to disable memfd entirely for this case.
> > >>> Do you think I need to get patch from v1 back ?
> > >>>
> > >>> Gerd, Marc-André, what do you think?
> > >>
> > >> I still think it makes sense to require sealing support.  Sealing is
> > >> very useful, and there are only a few kernel versions with memfd but
> > >> without sealing.  So finding such kernels in the wild will become more
> > >> rare over time.  I wouldn't worry too much about them.
> > > 
> > > -object memory-backend-memfd,id=mem,size=2M,seal=off still
> > > works on those systems, doesn't it?  What's the rationale for
> > > breaking a working configuration without following the
> > > deprecation policy?
> > > 
> > 
> > See the commit message.
> > '.seal' property is not registered if sealing is not supported.
> > So, there is no way to disable sealing on the system that does not support 
> > it.
> 
> As I pointed out a few lines up, this is simply because QEMU has a bug
> setting seal=true as the built-in default value even when it isn't
> supported. 

Changing to seal=false by default may make it work on some hosts,
but I don't see the point of increasing our support burden just
for a few kernel versions.  I agree with Gerd, I think it's
simpler to keep it unsupported.

-- 
Eduardo