NAT/Network: Add missing Id/@file/copyright headers.

scm automatic cleanups.

NAT/Net: Cosmetics - prefer ip_addr_* APIs to memcpy/memset.

NAT/Net: Cosmetic follow up to r94469 - use ipoff variable in debug printf, don't fetch it from the header again.

NAT/Net: OS X converts total length field of the inner IP header inside ICMP error message to host byte order too (but not the flags/fragment offset field, unlike outer header).

NAT/Net: When using IP_HDRINCL we need to rewrite mapped ping destination ourselves.

NAT/Net: pxping_pcb_rtstrfmt: for mapped addresses print both original and mapped destianations.

NAT/Net: OS X reports IP datagrams on raw sockets with IP header fragment offset field in host byte order too.

NAT/Net: Solaris doesn't compute IPv4 header checksum for ICMP echo replies from host's own interfaces. Compute it from scratch instead of doing incremental update if checksum is zero.

NAT/Net: init pxping::hdrincl appropriately as we know IP_HDRINCL is off initially, so -1 (erroneously intended for "not inited") doesn't make sense here and is wrong since it would be later interpreted as if IP_HDRINCL was set.

NAT/Net: convert perror/warn/warnx to DPRINTFs.

NAT/Net: Start untangling errno vs. winsock mess. Don't refer errno directly, while here convert some perror() calls to DPRINTFs.

NAT/Net: Get rid of inet_ntop(3). Add formatter for %R[ping_pcb] and use it.

NAT/Net: #define LOG_GROUP LOG_GROUP_NAT_SERVICE

G/c temporary local redefinitions of DPRINTF*.

Get rid of plain printfs: convert to DPRINTFs or g/c.

pxping_recv6: translate errors from proxy_sendto() into ICMPv6 error messages.

pxping_recv6: check if pbuf_header() succeeded.

pxping_pcb_deregister: clear ping_pcb::next to satisfy assertion (now fixed) in pxping_pcb_delete() happy.

pxping_pcb_delete: fix assertions (that were not compiled).

We still own pbuf after ip_output_if/ip6_output_if so call pbuf_free always, not just on errors.

pxping_recv4: comment update for previous after testing - OS X really wants IP header checksum set to checksum of all data, in host order.

pxping_recv4: according to ping sources at www.opensource.apple.com and contrary to icmp(4) man page, when sending with IP_HDRINCL the IP header checksum must be set, and set to the host-order checksum of the whole IP datagram including payload.

pxping_pmgr_icmp6_error: g/c unused variable.

Use either IP_HDRINCL or OS-specific socket option to deal with IPv4 DF flag.

pxping_recv4, pxping_recv6: free incoming pbuf after reporting time exceeded.

pxping_recv4: disable setsockopt-based code to set DF - it's too fragile w.r.t. different versions of system headers that don't have corresponding options defined (older Linux, older Solaris).

Disable DF code on Linux for now to unbreak the build. Apparently, IP_PMTUDISC_PROBE is not defined everywhere.

pxping_recv4: propagate DF flag if OS provides a socket option for that.

Try to fix in6_pktinfo kludge again. Test for __GLIBC_PREREQ in a separate conditional directive.

Try to fix in6_pktinfo kludge. Apparently it was hidden under __USE_GNU in glibc 2.8, so check glibc version too. This should hopefully fix build on RHEL5.

Do the song and dance to use WSARecvMsg on Windows.

G/c unused variable.

Init sin6_flowinfo.

Set sin6_port through correct union member.

Complete ICMPv6.

Refactor a bit.

Add a few comments.

Delay TTL check until after loopback remapping check. Fix loopback remapping logic for ICMP errors.

Add a comment that RFC 3542 IPV6_RECVTCLASS has no RFC 2292 counterpart (this only affects Windows, I think).

On OS X request RFC 3542 API (neither old RFC 2292, nor new RFC 3542 is provided by default).

Oops. Revert previous. <netinet/icmp6.h> that it needs (and that I forgotten to add in previous) conflicts with lwIP headers. I've been waiting for this to happen.

Use ICMP filters if present.

Snapshot work in progress on IPv6 ping now that it seems to work for the basic case. Raw sockets API for IPv6 is somewhat unwieldy, so I'll do more refactoring when dust settles a bit.

pxping_timeout_del: mark pxping argument as unused.

Use rand(3) on Windows. It's not good, but we don't need cryptographically strong random number for ping id.

G/c unused variables. Pass iplen around as u16_t. This makes Windows happier.

Handle mapped loopbacks. Propagate TTL, TOS.

Snapshot work in progress on ping proxy. Not hooked into the build yet.