qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH] microvm: turn off io reservations for pcie root ports


From: Michael S. Tsirkin
Subject: Re: [PATCH] microvm: turn off io reservations for pcie root ports
Date: Wed, 29 Jun 2022 10:04:19 -0400

On Wed, Jun 29, 2022 at 02:20:50PM +0200, Gerd Hoffmann wrote:
> On Wed, Jun 29, 2022 at 03:16:17AM -0400, Michael S. Tsirkin wrote:
> > On Wed, Jun 29, 2022 at 09:10:23AM +0200, Gerd Hoffmann wrote:
> > > On Mon, Jun 27, 2022 at 06:37:50PM -0400, Michael S. Tsirkin wrote:
> > > > On Thu, Jun 09, 2022 at 09:28:38AM +0200, Gerd Hoffmann wrote:
> > > > > On Wed, Jun 08, 2022 at 12:06:17PM -0400, Michael S. Tsirkin wrote:
> > > > > > On Fri, Jun 03, 2022 at 10:59:20AM +0200, Gerd Hoffmann wrote:
> > > > > > > The pcie host bridge has no io window on microvm,
> > > > > > > so io reservations will not work.
> > > > > > > 
> > > > > > > Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
> > > > > > 
> > > > > > I don't much like overriding user like this. We end up users
> > > > > > setting it to silly values and then if we do want to
> > > > > > support this things just break. Thoughts?
> > > > > 
> > > > > Well, it just looked like the simplest way to tell the firmware that
> > > > > io reservations are pointless.  Do you have a better idea?
> > > > > 
> > > > > take care,
> > > > >   Gerd
> > > > 
> > > > Fail if user supplies values we can't support.
> > > 
> > > Well, it is the *default* value which doesn't work on microvm.
> > > 
> > > take care,
> > >   Gerd
> > 
> > Changing defaults is ok of course. Let's just not override the user.
> 
> Ok, so I could use a compat property instead and change the default
> for microvm that way.  That would allow the user set any value it
> wants.

And if you like check the value and fail init if not 0.

> I still don't see the point though.  There is only a single value which
> makes sense on microvm.  Which is zero.  The only effect the user could
> archive is make the firmware throwing errors ...
> 
> take care,
>   Gerd

My concern is simple: if right now we override the value then
some users might set it to != 0 by mistake. Then if down the road
we want to intepret != 0 in some way, these setups
will start failing. I don't claim to see a use-case why we'd want to
but it's hard to predict the future with certainty.

-- 
MST




reply via email to

[Prev in Thread] Current Thread [Next in Thread]