On Wed, 25 Feb 2026 at 13:09, Bruce Richardson
<[email protected]> wrote:
>
> On Wed, Feb 25, 2026 at 12:29:21PM +0100, David Marchand wrote:
> > Some applications use port hotplug as their primary way for using DPDK
> > resources.
> > Having a systematic device probing is a problem when not all available
> > resources will be used by the application, as such applications won't set
> > an explicit allow list at startup.
> >
> > This is the case for OVS on systems with multiple mlx5 devices:
> > one device can be used by the kernel while the other(s) are used by DPDK.
> > In such a setup, the kernel used device may get reconfigured in
> > unexpected ways and trigger issues like the one described by Kevin
> > not so long ago in bugzilla 1873.
> >
> > Add an EAL option to disable device probing, allowing to select per bus.
> >
> > Note: the current implementation does not take combinations of the new
> > option with -a/-b.
> >
> > Signed-off-by: David Marchand <[email protected]>
> > ---
> Wondering about this problem and fix options. Rather than a new flag for
> no-probe, would an alternative flag to switch the default mode from
> block-listing to allow-listing work?
>
> Right now, by default, EAL probes all buses and tries to take all
> non-blocklisted devices. We could add a flag to switch that so that we
> allow-list by default, which I think should have the same effect as this if
> no -a flags are passed. We already have support for managing -a and -b flag
> interactions, which should make interacting with the new flag easier.

I like the idea. The devil is in the details though and I'll have to
look carefully at all bus probe() functions.
I'll try to implement as a RFC v2.


-- 
David Marchand

Reply via email to