just and fyi, I've broken this down to 5 commits and only pushed the first so far

From my testing this seems good. From there I will break this down a bit so the commits are smaller and start adding stuff next week

In D46689#1082245, @damjan.jov_gmail.com wrote:

This update:

• Makes EIM opt-in, off by default, although I still think this decision needs broader discussion, or even a vote.

Is there a reason this moves in the ordering of nat rules? We tend to aim for roughly alphabetical ordering if there isn't a reason for something different?

Generally looks good, there are some small nits, but I'm happy for this to land once you've made the changes to the man page and the comment at the top.

Sorry one more thing, can you upload future diffs with full context. It makes reviewing on phabricator much easier.

Two small things, I will start testing now

Tweak some wording to imply reality

Sorry it is a bit late

In D16851#1066818, @bdcom.dragon_gmail.com wrote:

I think we just need to check first fragments in frag6_input() and check whole header chain everytime:

if (fragoff == 0)
{
    if (!ip6_valid_hdrchain(&m, off, nxt, frgpartlen))
    {
         icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_MISSING_HDR, 0);
         return (IPPROTO_DONE);
     }
}

This will conform with RFC 8200, page 21, last paragraph.

Thanks Ed, I think that is a useful clarification.

For anyone looking into this the ACPI reference table might help:

Thanks for the comments both, I'm sorry I rolled my confusion about the use of cdevpriv and clean up with this change.

• handle devfs_set_cdevpriv correctly
• actively initialize more of priv

Looks like an improvement to me, the big comment helps.

So far this generally looks ok to me and the tests seem to work.

Thanks for checking this over!

• Fix white space issues
• Fix line lengths
• test: use documentation addresses
• test: use tcpdump for verifying traffic

Use imx35-gpio rather than the soc specific compat string

In D46532#1060750, @manu wrote:

Any idea why it have a different compatible string ?
Is this because only of features that we don't use ?

Address most of Kp's comments
This adds a pfctl test case

Add a sysctl to control size of the udp mapping allocations

fix bad rebase

• Rebase change
• add pf rule syntax for endpoint-independent

I really like this change. The latest vm snapshots are triggering the upgrade message, but there are no google results for how to upgrade. Can you point me to the correct way to upgrade and I'll draft a change for the handbook with the upgrade message included.

I have a temp fix for this directly in ice and will work with erj@ to have the driver support the vlan ifconfig flags correctly

Thanks both for the comments, I didn't think this would be the correct approach, but it seemed easier to create a review for how I would like this to work.

Please fix the man page typo, but otherwise looks good to me

• access ext_arg1 via the accessor

updated

• swap order of arguments for destructor
• replace an m_freem with m_free

• Correct copyright statement
• Move then for if statements onto the same line

In D44371#1013215, @vmaffione wrote:

Any additional tests with pkt-gen in TX or RX mode over iflib interfaces and virtio-net ones?
E.g. two VMs/machines connected back to back with one transmitting and the other receiving, to stress test it?
Setting the sys.dev.netmap.admode sysctl to 2 will force emulated netmap even if the interface has native support.

• Add a macro for getting netmap adapter from ext_arg1

• - use SET_MBUF_DESTRUCTOR to set ext_arg1

• - Don't remove the pool lock until we have removed all mbufs

• - if the reference has gone after taking the lock don't free the mbuf

Address review comments