extract-{handshakes,keys}: rework for upstream kernel

Now that WireGuard has been upstreamed and the repos split, we have to
look elsewhere for these headers.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

man: document dynamic debug trick for Linux

This comes up occasionally, so it may be useful to mention its
possibility in the man page. At least the Arch Linux and Ubuntu kernels
support dynamic debugging, so this advise will at least help somebody.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: android: split uids into multiple commands

Different versions of netd have different limits on how many can be
passed at once.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Reported-by: Alexey <zaranecc@bk.ru>

version: bump

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

Makefile: sort inputs to linker so that build is reproducible

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

netlink: make sure to clear return value when trying again

Otherwise this runs in an infinite loop if at some point a dump was
interrupted.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

fuzz: add set and setconf fuzzers

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

Makefile: evaluate git version lazily

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

fuzz: add generic command argument fuzzer

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

ipc: simplify inflatable buffer and add fuzzer

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

Makefile: add standard 'all' target

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Reported-by: Bruno Wolff III <bruno@wolff.to>

Makefile: remove pwd from compile output

We previously included $(pwd) in the compile output pretty printer,
because it matched our parent out-of-tree module build. Since we're no
longer coupled to the module, we can return to a prettier scheme of just
using the object name.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Fixes: eb68ad07 ("Makefile: even prettier output")

version: bump

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

global: bump copyright

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: linux: quote ifname for nft

Otherwise nft(8) has strange ideas of what a string is.

Suggested-by: RistiCore <RistiCore@mail.ee>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

Makefile: rework automatic version.h mangling

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Reported-by: Joe Doss <joe@solidadmin.com>

fuzz: find bugs when parsing uapi input

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

fuzz: find bugs in the config syntax parser

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

man: add documentation about removing explicit listen-port

Signed-off-by: Devin Smith <thundza@gmail.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

dns-hatchet: adjust path for new repo layout

Reported-by: Joe Doss <joe@solidadmin.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

Makefile: port static analysis check

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

Makefile: DEBUG_TOOLS -> DEBUG and document

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

systemd: update documentation URL

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

version: bump

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

Makefile: add git versioning to dev builds

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

README: consolidate with INSTALL and rewrite

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg: include tools version

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg: add back source formerly shared with kernel module

We used to reach back into parent directories for this, but with the
repo split, we now require our own copy.

We use -idirafter in case system headers are installed for the
wireguard.h netlink definitions.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

gitignore: trim down to basics

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: linux: use already configured addresses instead of in-memory

The ADDRESSES array might not have addresses added during PreUp. But
moreover, nft(8) and iptables(8) don't like ip addresses in the form
somev6prefix::someipv4suffix, such as fd00::1.2.3.4, while ip(8) can
handle it. So by adding these first and then asking for them back, we
always get normalized addresses suitable for nft(8) and iptables(8).

Reported-by: Silvan Nagl <mail@53c70r.de>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg: adjust wg.8 syntax for consistency in COMMANDS section

Signed-off-by: Kai Haberzettl <khaberz@gmail.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: linux: try both iptables(8) and nft(8) on teardown

Daniel argues that technically a package manager could install nft(8)
after previously having started wg-quick(8) using iptables(8).

Suggested-by: Daniel Kahn Gillmor <dkg@fifthhorseman.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: linux: support older nft(8)

Older nft(8), such as that on Ubuntu, does not accept the - parameter to
the -f argument and doesn't accept symbolic priority names. So instead
use the canonical numeric priority forms and use <(echo) instead of -.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

global: fix up spelling

Signed-off-by: Josh Soref <jsoref@gmail.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: linux: add support for nft and prefer it

If nft(8) is installed, use it. These rules should be identical to the
iptables-restore(8) ones, with the advantage that cleanup is easy
because we use custom table names.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: linux: ignore save warnings for iptables-nft

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: linux: suppress more warnings on weird kernels

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: linux: some iptables don't like empty lines

Reported-by: Kenneth R. Crudup <kenny@panix.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: linux: iptables-* -w is not widely supported

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

ipc: make sure userspace communication frees wgdevice

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: linux: have remove_iptables return true

Reported-by: Thomas Sattler <sattler@med.uni-frankfurt.de>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: linux: ensure postdown hooks execute

Reported-by: Thomas Sattler <sattler@med.uni-frankfurt.de>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: linux: suppress error when finding unused table

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg: add syncconf command

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

reresolve-dns: remove invalid anchors on regex match

Reported-by: Conrad Meyer <cem@freebsd.org>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: linux: filter bogus injected packets and don't disable rpfilter

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: linux: only touch net.ipv4 for v4

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: android: check for null in binder cleanup functions

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: android: use Binder for setting DNS on Android 10

Signed-off-by: Nicolas Douma <nicolas@serveur.io>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg: windows: enforce named pipe ownership and use protected prefix

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: linux: don't fail down when using systemd-resolved

systemd-resolved has a compatibility interface for use with resolvconf
scripts when resolvectl is called from a symlink from resolvconf.
However, when tearing down the interface, cmd_down calls del_if and then
unset_dns. In the case of systemd-resolved, deleting the interface also
removes the systemd-resolved entry and causes resolvconf -d to fail when
resolvconf really is a symlink to resolvectl. This causes `wg-quick
down` and 'wg-quick@.service' to exit with failure.

Instead we use the resolvconf '-f' flag to ignore non-existent
interfaces, supported by both openresolv and sd-resolved resolvconf.

Signed-off-by: Ronan Pigott <rpigott@berkeley.edu>
[zx2c4: moved -f argument to end to remain compatible with Debian's resolvconf]
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: openbsd: fix alternate routing table syntax

route(8) has always used the `-T` option to specify the
routing table; there is no `rdomain` option.

Signed-off-by: Ankur Kothari <ankur@lipidity.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: android: refactor and add incoming allow rules

Suggested-by: Yağmur Oymak <yagmur.oymak@gmail.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: darwin: support being called from launchd

This causes wg-quick up to wait for the monitor to exit before it exits,
so that launchd can correctly wait on it.

Reported-by: Cameron Palmer <cameron@promon.no>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg: pass WG_ENDPOINT_RESOLUTION_RETRIES=infinity to systemd unit

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg: add wincompat layer to wg(8)

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg: allow setting WG_ENDPOINT_RESOLUTION_RETRIES

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: specify protocol to ip(8), because of inconsistencies

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: look up existing routes properly

This was never really correct, and then 5.1 broke it entirely.

Reported-by: piraty1@inbox.ru
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: make darwin and freebsd path search strict like linux

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: freebsd: workaround SIOCGIFSTATUS race in FreeBSD kernel

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg: avoid unneccessary next_peer assignments in sort_peers()

Signed-off-by: Luis Ressel <aranea@aixah.de>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: add 'strip' subcommand

`wg-quick strip` prints the config file to stdout after stripping it of
all wg-quick-specific options.

This enables tricks such as `wg addconf $DEV <(wg-quick strip $DEV)`.

Signed-off-by: Luis Ressel <aranea@aixah.de>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg: warn if an AllowedIP has a nonzero host part

Signed-off-by: Luis Ressel <aranea@aixah.de>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: freebsd: export TMPDIR when restoring and don't make empty

Otherwise mktemp doesn't see it, and if it's empty we wind up in /.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg: add support for Haiku

Signed-off-by: Alexander von Gluck IV <kallisti5@unixzen.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg: genkey: account for short reads of /dev/urandom

Apparently Haiku has a misbehaving /dev/urandom.

While we're at it, simplify the function signature to completely succeed
or completely fail and make sure the caller checks the result.

Reported-by: Alexander von Gluck IV <kallisti5@unixzen.com>
Nitpicked-by: Aaron Jones <aaronmdjones@gmail.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: freebsd: rebreak interface loopback, while fixing localhost

The commit 7c833642 ("wg-quick: freebsd: allow loopback to work") was
supposed to make things better, but actually it just started sending
legitimate localhost traffic over the WireGuard interface, which is
really quite bad.

This reverts commit 7c833642dfa342218602ab18e7091e86408d2982.

Reported-by: Matt Smith <matt.xtaz@gmail.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg: c_acc doesn't need to be initialized

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg: fight compiler slightly harder

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

noise: store clamped key instead of raw key

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

highlighter: when subtracting char, cast to unsigned

Windows.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

systemd: wg-quick should depend on nss-lookup.target

Since wg-quick(8) calls wg(8) which does hostname lookups, we should
probably only run this after we're allowed to look up hostnames.

Reported-by: Anton Castelli <anton.c42@gmail.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg: remove unused check phony declaration

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: freebsd: allow loopback to work

FreeBSD adds a route for point-to-point destination addresses. We don't
really want to specify any destination address, but unfortunately we
have to. Before we tried to cheat by giving our own address as the
destination, but this had the unfortunate effect of preventing
loopback from working on our local ip address. We work around this with
yet another kludge: we set the destination address to 127.0.0.1. Since
127.0.0.1 is already assigned to an interface, this has the same effect
of not specifying a destination address, and therefore we accomplish the
intended behavior.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

netlink: use __kernel_timespec for handshake time

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

contrib: introduce simple highlighter library

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

global: normalize -> clamp

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

keygen-html: bring back pure javascript implementation

This reverts commit 9d5baf7d1d14ca7eb0852b41566330259229d489.

Benoît Viguier has proofs that values will stay well within 2^53. We
also have an improved carry function that's much simpler.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

Kconfig: IPsec isn't IPSec

Reported-by: Raf Czlonka <rczlonka@gmail.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

global: update copyright

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg: curve25519: handle unaligned loads/stores safely

Reported-by: Chris Hewitt <chris@chrishewitt.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

makefile: use immediate expansion and use correct template patterns

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: bring interface up while setting MTU

This avoids another ip(8) invocation for little benefit.
Confirmed to work with iproute2 and busybox.

Signed-off-by: Aaron Jones <aaronmdjones@gmail.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

embeddable-wg-library: do not warn on unrecognized netlink attributes

This is a follow up of bcf8684c9ec90fe0d283a67d1654d05fb3eae019.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

global: various formatting tweeks

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: android: do not choke on empty allowed-ips

Reported-by: Samuel Holland <samuel@sholland.org>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

keygen-html: add missing glue macro

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg.8: AllowedIPs isn't actually required

An empty allowed IPs is totally valid, for folks wishing to move IP
addresses between multiple peers atomically.

Suggested-by: Comex <comexk@gmail.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg.8: specify that wg(8) shows runtime info too

Suggested-by: Comex <comexk@gmail.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick: wait for interface to disappear on freebsd

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg: don't fail if a netlink interface dump is inconsistent

Netlink returns NLM_F_DUMP_INTR if the set of all tunnels changed
during the dump. That's unfortunate, but is pretty common on busy
systems that are adding and removing tunnels all the time. Rather
than retrying, potentially indefinitely, we just work with the
partial results.

Reported-by: Robert Gerus <ar@is-a.cat>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg: compile on gnu99

We don't actually use any C11 features, so we can at least compile with
ancient gcc.

Reported-by: Aaron M. D. Jones <aaronmdjones@gmail.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg: use libc's endianness macro if no compiler macro

This lets us be compiled with ancient gcc.

Reported-by: Jeff Brandt <jeff@jeffcolo.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

global: rename struct wireguard_ to struct wg_

This required a bit of pruning of our christmas trees.

Suggested-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

netlink: do not stuff index into nla type

It's not used for anything, and LKML doesn't like the type being used as
an index value.

Suggested-by: Eugene Syromiatnikov <esyr@redhat.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

crypto: clean up remaining .h->.c

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

wg-quick.8: add policy routing example

Suggested-by: Toke Høiland-Jørgensen <toke@toke.dk>
Suggested-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

crypto: make constant naming scheme consistent

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

global: put SPDX identifier on its own line

The kernel has very specific rules correlating file type with comment
type, and also SPDX identifiers can't be merged with other comments.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>