svn commit: r334104 - in head/sys: netinet sys

[Jonathan T. Looney]

On Wed, May 23, 2018 at 7:13 PM, Matthew Macy <mmacy at freebsd.org> wrote:
>
> On Wed, May 23, 2018 at 11:52 AM, John Baldwin <jhb at freebsd.org> wrote:
> > On Wednesday, May 23, 2018 05:00:05 PM Matt Macy wrote:
> >> Author: mmacy
> >> Date: Wed May 23 17:00:05 2018
> >> New Revision: 334104
> >> URL: https://svnweb.freebsd.org/changeset/base/334104
> >>
> >> Log:
> >>   epoch: allow for conditionally asserting that the epoch context
fields
> >>   are unused by zeroing on INVARIANTS builds
> >
> > Is M_ZERO really so bad that you need to make it conditional?
>
> In this case not at all. It's only exercised by sysctl handlers. I'm
> mostly responding to an inquiry by jtl. However, gratuitous M_ZERO
> usage does have a cumulative adverse performance impact.

I appreciate you making this change. And, I do think it is worth avoiding
M_ZERO where it is unnecessary, for the reason you state.

> > I would probably have preferred something like 'M_ZERO_INVARIANTS'
> > instead perhaps (or M_ZERO_EPOCH) that only controls M_ZERO and is
> > still or'd with M_WAITOK or M_NOWAIT.
>
> Yes. I like that better too. Thanks.

Yes, that does seem better.

Thanks!

Jonathan