Re: git: 1ccf543b21ef - main - ifconfig: Add SR-IOV VF status output

From: Gleb Smirnoff <glebius_at_freebsd.org>
Date: Tue, 11 Aug 2026 14:59:54 UTC
On Mon, Aug 10, 2026 at 12:03:00PM -0700, Kevin Bowling wrote:
K> > On Sun, Aug 09, 2026 at 06:47:43AM +0000, Kevin Bowling wrote:
K> > K> The branch main has been updated by kbowling:
K> > K>
K> > K> URL: https://cgit.FreeBSD.org/src/commit/?id=1ccf543b21eff6e0828142e5c1d09519247143f4
K> > K>
K> > K> commit 1ccf543b21eff6e0828142e5c1d09519247143f4
K> > K> Author:     Eric Joyner <erj@FreeBSD.org>
K> > K> AuthorDate: 2019-03-18 18:30:00 +0000
K> > K> Commit:     Kevin Bowling <kbowling@FreeBSD.org>
K> > K> CommitDate: 2026-08-09 06:46:29 +0000
K> > K>
K> > K>     ifconfig: Add SR-IOV VF status output
K> > K>
K> > K>     - Adds SR-IOV VF status to the existing ifconfig "-v" output
K> > K>     - Adds ioctl command for reporting VF status info from drivers
K> > K>     - Adds support to iflib for drivers to handle this new ioctl
K> > K>     - Add support for ioctl in ixl(4)
K> > K>
K> > K>     Signed-off-by: Eric Joyner <erj@freebsd.org>
K> > K>
K> > K>     Relnotes:       yes
K> > K>     Differential Revision:  https://reviews.freebsd.org/D19647
K> >
K> > This is not a reviewed change.  The D19647 itself was not accepted 7 years ago.
K> > Just an URL to reviews.freebsd.org doesn't warrant that change is reviewed.
K> > Even if it was accepted, just 7 year lag itself requires a re-review.
K> 
K> I reviewed it.
K> 
K> > This particular change brings a lot of new ioctl(2) APIs.  It is the opposite
K> > of what other people were doing with ifconfig in the last years - reducing
K> > ioctl use and switching to netlink.
K> 
K> This was to give Eric proper credit and provenance, it is immediately
K> replaced by an nvlist.  A netlink interface is provided.  What
K> concretely is your suggestion?
K> 
K> > Please revert this and start a proper review process.

My suggestion was to revert the change and move towards reducing ioctl use in
favor of netlink.  Also, as Pouria rightfully noticed in response to other
change, that packing of parameters should be bare netlink without nvlist as
intermediate packing.

-- 
Gleb Smirnoff