From: Alexandr Nedvedicky <sashan@fastmail.net>
Subject: Re: PF Queue bandwidth now 64bit for >4Gbps queues
To: Andrew Lemin <andrew.lemin@gmail.com>, tech@openbsd.org, Theo Buehler <tb@openbsd.org>
Date: Fri, 20 Mar 2026 08:49:06 +0100

Hello,

On Thu, Mar 19, 2026 at 12:46:03PM +0000, Stuart Henderson wrote:
> On 2026/03/18 11:28, Stuart Henderson wrote:
> > On 2026/03/18 11:01, Stuart Henderson wrote:
> > > On 2026/03/18 00:57, Andrew Lemin wrote:
> > > > struct hfsc_sc grows from 12 to 24 bytes (with a 4-byte padding hole
> > > > between d and m2 due to alignment - should we reorder?).
> > > 
> > > yes.
> > > 
> > > from style(9):
> > > 
> > >    When declaring variables in structures, declare them sorted by use, then
> > >    by size (largest to smallest), then by alphabetical order.  The first
> > >    category normally doesn't apply, but there are exceptions.
> > > 
> > > (+ i'll see if I can figure out what's needed in ports-land for this).
> > > 
> > 
> > +Values up to 999G are supported, allowing configuration of PF queues on
> > +10G, 25G, 40G, and 100G interfaces.
> > 
> > I don't think the clause about interface speeds is needed
> 
> Actually I think the manual change should just be dropped, it's not
> needed. Diff below drops that and reorders the struct. This version is
> ok sthen and I would like this to coincide with library bumps that are
> about to take place, so that most packages will get updated anyway
> (i.e. commit this first or at the same time as the bumps).

    the diff also works for me. OK sashan

> 
> +Values up to 999G are supported.
> 
> There is a pre-existing issue with pfctl silently mishandling bandwidth
> specs above this. That should be fixed rather than rely on someone
> reading docs. I think not a blocker for committing the uint64 change.
> I'll send a separate mail about this.
> 

    ... and the fix is separate commit now.

thanks a lot
regards
sashan