Address the remaining manpage comments.

Address emaste@ manpage comments.

• Use STAILQ mbuf macros in nl_io_queue
• Update man pages to reflect the comments

I’m going to commit this tomorrow, Friday 30, unless there are any objections

Update headers to reflect the comments.

In D36002#835201, @imp wrote:

From a licensing point of view, thinks look generally OK (modulo the one different language for the man pages I noticed). My approval is conditional on you fixing or explaining why you used these.
Most of these headers appear to have little creative content
Copying from the RFC is the only possible issue, since each RFC has its own permission statement. Please verify text from there can be copied verbatim.
Otherwise it appears that most of these items are covered by 'only one way to do it' plus 'gotta have these names and values for compatibility' exceptions.

Their "Full Copyright Statement" states that it can be used freely provided their copyright statement has to be included: "provided that the above copyright notice and this paragraph are

included on all such copies and derivative works". I've updated `netlink/netlink.h` to include this header. Does it look right to you?

Address header files feedback.

• Add basic support for NETLINK_GET_STRICT_CHK
• Fix getsockopt() for NETLINK_CAP_ACK / NETLINK_EXT_ACK
• Move interface creation logic to the iface_drivers.c
• Fix manpage lint errors

In D36651#833505, @kp wrote:

That's an interesting idea, but I'm not clear on what the long-term plan is.
This will work (and might indeed be an improvement) for cloned interfaces, but we'd still have to maintain the previous way of handling things for "real" (e.g. igb0) interfaces as they're moved between vnets.

I worry that we'll end up doing things one way for cloned interfaces, and another for other interfaces, which gives us twice the opportunity for bugs, and half as much testing.

Thank you for the feedback!
Indeed, I should have included the proposal for the physical interfaces.
The primary goal of this change in general is to replace the current mechanism. I'm currently thinking of the following:

• Slightly re-purpose if_reassign() to perform the same function as vmove_f handler (potentially even use if_reassign() directly even for cloned interfaces)
• Write the handlers for the "relevant" (e.g. modern, likely to be used in routers/cloud setups) interfaces such as vnet/iflib/mellanox/chelsio
• Add a deprecation warning with a wiki link, specifying how to report it, to the current cloning method
• Wait for a couple of months and remove the current cloning method

Abort in favour of D36746

• Add custom SO_RCVBUF handler
• Sync to the recent HEAD
• Add manpage for netlink(4) and a manpage draft for rtnetlink(4).

Simplify patch.

Update patch

Update patch

add ifc_copyout()

.

Clean up headers a bit.

Update to reflect netlink KPI changes.

Fix panic when adding vlan without id.

• Add (basic) link deletion/creation
• Simplify parsing
• Update headers
• Sync to recent HEAD

• Fix (eventual) panic on module reload by deregistering neigh eventhandler

• Retire event bridge logic in favor of direct rtsock<>netlink calls
• Add neighbor notifications
• Update to latest -HEAD

• Provide notifications for generic netlink families load/unload

Update to recent -HEAD.

• Add generic netlink support
• Simplify error checking when writing attributes
• Add extended ACK support (text error messages & error offset)

Address most of glebius@ comments.

• Add support for multipath operations w/o user nexthops.
• Generalise parsing a bit

• Add netlink -> rtsock route event translation
• Sync to recent HEAD