diff --git a/net/wireguard/Kbuild b/net/wireguard/Kbuild
new file mode 100644
index 000000000000..f1b4ad76fdf6
--- /dev/null
+++ b/net/wireguard/Kbuild
@@ -0,0 +1,16 @@
+# SPDX-License-Identifier: GPL-2.0
+#
+# Copyright (C) 2015-2019 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved.
+
+ccflags-y := -O3 -fvisibility=hidden
+ccflags-$(CONFIG_WIREGUARD_DEBUG) += -DDEBUG -g
+ccflags-y += -D'pr_fmt(fmt)=KBUILD_MODNAME ": " fmt'
+ccflags-y += -Wframe-larger-than=2048
+ccflags-$(if $(WIREGUARD_VERSION),y,) += -D'WIREGUARD_VERSION="$(WIREGUARD_VERSION)"'
+
+wireguard-y := main.o noise.o device.o peer.o timers.o queueing.o send.o receive.o socket.o peerlookup.o allowedips.o ratelimiter.o cookie.o netlink.o
+
+include $(src)/crypto/Kbuild.include
+include $(src)/compat/Kbuild.include
+
+obj-$(if $(KBUILD_EXTMOD),m,$(CONFIG_WIREGUARD)) := wireguard.o
diff --git a/net/wireguard/Makefile b/net/wireguard/Makefile
new file mode 100644
index 000000000000..c20bfd39ae2d
--- /dev/null
+++ b/net/wireguard/Makefile
@@ -0,0 +1,58 @@
+# SPDX-License-Identifier: GPL-2.0
+#
+# Copyright (C) 2015-2019 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved.
+
+KERNELRELEASE ?= $(shell uname -r)
+KERNELDIR ?= /lib/modules/$(KERNELRELEASE)/build
+PREFIX ?= /usr
+DESTDIR ?=
+SRCDIR ?= $(PREFIX)/src
+DKMSDIR ?= $(SRCDIR)/wireguard
+DEPMOD ?= depmod
+
+PWD := $(shell pwd)
+
+all: module
+debug: module-debug
+
+ifneq ($(V),1)
+MAKEFLAGS += --no-print-directory
+endif
+
+WIREGUARD_VERSION = $(patsubst v%,%,$(shell GIT_CEILING_DIRECTORIES="$(PWD)/../.." git describe --dirty 2>/dev/null))
+
+module:
+	@$(MAKE) -C $(KERNELDIR) M=$(PWD) WIREGUARD_VERSION="$(WIREGUARD_VERSION)" modules
+
+module-debug:
+	@$(MAKE) -C $(KERNELDIR) M=$(PWD) V=1 CONFIG_WIREGUARD_DEBUG=y WIREGUARD_VERSION="$(WIREGUARD_VERSION)" modules
+
+clean:
+	@$(MAKE) -C $(KERNELDIR) M=$(PWD) clean
+
+module-install:
+	@$(MAKE) -C $(KERNELDIR) M=$(PWD) WIREGUARD_VERSION="$(WIREGUARD_VERSION)" modules_install
+	$(DEPMOD) -a $(KERNELRELEASE)
+
+install: module-install
+
+rwildcard=$(foreach d,$(wildcard $1*),$(call rwildcard,$d/,$2) $(filter $(subst *,%,$2),$d))
+DKMS_SOURCES := version.h Makefile Kbuild Kconfig dkms.conf $(filter-out version.h wireguard.mod.c tests/%,$(call rwildcard,,*.c *.h *.S *.pl *.include))
+dkms-install: $(DKMS_SOURCES)
+	@$(foreach f,$(DKMS_SOURCES),install -v -m0644 -D $(f) $(DESTDIR)$(DKMSDIR)/$(f);)
+
+style:
+	$(KERNELDIR)/scripts/checkpatch.pl -f --max-line-length=4000 --codespell --color=always $(filter-out wireguard.mod.c,$(wildcard *.c)) $(wildcard *.h) $(wildcard selftest/*.c)
+
+check: clean
+	scan-build --html-title=wireguard-linux-compat -maxloop 100 --view --keep-going $(MAKE) module CONFIG_WIREGUARD_DEBUG=y C=2 CF="-D__CHECK_ENDIAN__"
+
+coccicheck: clean
+	@$(MAKE) -C $(KERNELDIR) M=$(PWD) CONFIG_WIREGUARD_DEBUG=y coccicheck MODE=report
+
+cloc:
+	@cloc --skip-uniqueness --by-file --extract-with="$$(readlink -f ../kernel-tree-scripts/filter-compat-defines.sh) >FILE< > \$$(basename >FILE<)" $(filter-out wireguard.mod.c,$(wildcard *.c)) $(wildcard *.h)
+
+-include tests/debug.mk
+
+.PHONY: all module module-debug module-install install dkms-install clean cloc check style
diff --git a/net/wireguard/compat/compat.h b/net/wireguard/compat/compat.h
new file mode 100644
index 000000000000..0d109be3b8ff
--- /dev/null
+++ b/net/wireguard/compat/compat.h
@@ -0,0 +1,1070 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (C) 2015-2019 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved.
+ */
+
+#ifndef _WG_COMPAT_H
+#define _WG_COMPAT_H
+
+#include <linux/kconfig.h>
+#include <linux/version.h>
+#include <linux/types.h>
+#include <generated/utsrelease.h>
+
+#ifdef RHEL_MAJOR
+#if RHEL_MAJOR == 7
+#define ISRHEL7
+#elif RHEL_MAJOR == 8
+#define ISRHEL8
+#ifdef RHEL_MINOR
+#if RHEL_MINOR == 2
+#define ISRHEL82
+#endif
+#endif
+#endif
+#endif
+#ifdef UTS_UBUNTU_RELEASE_ABI
+#if LINUX_VERSION_CODE == KERNEL_VERSION(3, 13, 11)
+#define ISUBUNTU1404
+#elif LINUX_VERSION_CODE >= KERNEL_VERSION(4, 4, 0) && LINUX_VERSION_CODE < KERNEL_VERSION(4, 5, 0)
+#define ISUBUNTU1604
+#endif
+#endif
+#ifdef CONFIG_SUSE_KERNEL
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 5, 0) && LINUX_VERSION_CODE >= KERNEL_VERSION(4, 4, 0)
+#define ISOPENSUSE42
+#endif
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 13, 0) && LINUX_VERSION_CODE >= KERNEL_VERSION(4, 12, 0)
+#define ISOPENSUSE15
+#endif
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 10, 0)
+#error "WireGuard requires Linux >= 3.10"
+#endif
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 6, 0)
+#error "WireGuard has been merged into Linux >= 5.6 and therefore this compatibility module is no longer required."
+#endif
+
+#if defined(ISRHEL7)
+#include <linux/skbuff.h>
+#define headers_end headers_start
+#elif LINUX_VERSION_CODE < KERNEL_VERSION(3, 18, 0)
+#define headers_start data
+#define headers_end data
+#endif
+
+#include <linux/cache.h>
+#include <linux/init.h>
+#ifndef __ro_after_init
+#define __ro_after_init __read_mostly
+#endif
+
+#include <linux/compiler.h>
+#ifndef READ_ONCE
+#define READ_ONCE ACCESS_ONCE
+#endif
+#ifndef WRITE_ONCE
+#ifdef ACCESS_ONCE_RW
+#define WRITE_ONCE(p, v) (ACCESS_ONCE_RW(p) = (v))
+#else
+#define WRITE_ONCE(p, v) (ACCESS_ONCE(p) = (v))
+#endif
+#endif
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 17, 0)
+#include "udp_tunnel/udp_tunnel_partial_compat.h"
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 3, 0) && !defined(DEBUG) && defined(net_dbg_ratelimited)
+#undef net_dbg_ratelimited
+#define net_dbg_ratelimited(fmt, ...) do { if (0) no_printk(KERN_DEBUG pr_fmt(fmt), ##__VA_ARGS__); } while (0)
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 3, 0)
+#include <linux/rcupdate.h>
+#ifndef RCU_LOCKDEP_WARN
+#define RCU_LOCKDEP_WARN(cond, message) rcu_lockdep_assert(!(cond), message)
+#endif
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 3, 0) && !defined(ISRHEL7)
+#define ipv6_dst_lookup(a, b, c, d) ipv6_dst_lookup(b, c, d)
+#endif
+
+#if (LINUX_VERSION_CODE == KERNEL_VERSION(4, 4, 0) || \
+    (LINUX_VERSION_CODE < KERNEL_VERSION(4, 3, 5) && LINUX_VERSION_CODE >= KERNEL_VERSION(4, 2, 0)) || \
+    (LINUX_VERSION_CODE < KERNEL_VERSION(4, 1, 17) && LINUX_VERSION_CODE > KERNEL_VERSION(3, 19, 0)) || \
+    (LINUX_VERSION_CODE < KERNEL_VERSION(3, 18, 27) && LINUX_VERSION_CODE >= KERNEL_VERSION(3, 17, 0)) || \
+    (LINUX_VERSION_CODE < KERNEL_VERSION(3, 16, 8) && LINUX_VERSION_CODE >= KERNEL_VERSION(3, 15, 0)) || \
+    (LINUX_VERSION_CODE < KERNEL_VERSION(3, 14, 40) && LINUX_VERSION_CODE >= KERNEL_VERSION(3, 13, 0)) || \
+    (LINUX_VERSION_CODE < KERNEL_VERSION(3, 12, 54))) && !defined(ISUBUNTU1404) && !defined(ISRHEL7)
+#include <linux/if.h>
+#include <net/ip_tunnels.h>
+#define IP6_ECN_set_ce(a, b) IP6_ECN_set_ce(b)
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 12, 0) && IS_ENABLED(CONFIG_IPV6) && !defined(ISRHEL7)
+#include <net/ipv6.h>
+struct ipv6_stub_type {
+	void *udpv6_encap_enable;
+	int (*ipv6_dst_lookup)(struct sock *sk, struct dst_entry **dst, struct flowi6 *fl6);
+};
+static const struct ipv6_stub_type ipv6_stub_impl = {
+	.udpv6_encap_enable = (void *)1,
+	.ipv6_dst_lookup = ip6_dst_lookup
+};
+static const struct ipv6_stub_type *ipv6_stub = &ipv6_stub_impl;
+#endif
+
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 8, 0) && IS_ENABLED(CONFIG_IPV6) && !defined(ISOPENSUSE42) && !defined(ISRHEL7)
+#include <net/addrconf.h>
+static inline bool ipv6_mod_enabled(void)
+{
+	return ipv6_stub != NULL && ipv6_stub->udpv6_encap_enable != NULL;
+}
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 11, 0) && !defined(ISRHEL7)
+#include <linux/skbuff.h>
+static inline void skb_reset_tc(struct sk_buff *skb)
+{
+#ifdef CONFIG_NET_CLS_ACT
+	skb->tc_verd = 0;
+#endif
+}
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 11, 0)
+#include <linux/random.h>
+#include <linux/siphash.h>
+static inline u32 __compat_get_random_u32(void)
+{
+	static siphash_key_t key;
+	static u32 counter = 0;
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 4, 0)
+	static bool has_seeded = false;
+	if (unlikely(!has_seeded)) {
+		get_random_bytes(&key, sizeof(key));
+		has_seeded = true;
+	}
+#else
+	get_random_once(&key, sizeof(key));
+#endif
+	return siphash_2u32(counter++, get_random_int(), &key);
+}
+#define get_random_u32 __compat_get_random_u32
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 18, 0) && !defined(ISRHEL7)
+static inline void netif_keep_dst(struct net_device *dev)
+{
+	dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
+}
+#define COMPAT_CANNOT_USE_CSUM_LEVEL
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 14, 0) && !defined(ISRHEL7)
+#include <linux/netdevice.h>
+#ifndef netdev_alloc_pcpu_stats
+#define pcpu_sw_netstats pcpu_tstats
+#endif
+#ifndef netdev_alloc_pcpu_stats
+#define netdev_alloc_pcpu_stats alloc_percpu
+#endif
+#elif LINUX_VERSION_CODE < KERNEL_VERSION(3, 15, 0) && !defined(ISRHEL7)
+#include <linux/netdevice.h>
+#ifndef netdev_alloc_pcpu_stats
+#define netdev_alloc_pcpu_stats(type)					\
+({									\
+	typeof(type) __percpu *pcpu_stats = alloc_percpu(type);		\
+	if (pcpu_stats)	{						\
+		int __cpu;						\
+		for_each_possible_cpu(__cpu) {				\
+			typeof(type) *stat;				\
+			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
+			u64_stats_init(&stat->syncp);			\
+		}							\
+	}								\
+	pcpu_stats;							\
+})
+#endif
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 14, 0) && !defined(ISRHEL7)
+#include "checksum/checksum_partial_compat.h"
+static inline void *__compat_pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len)
+{
+	if (tail != skb) {
+		skb->data_len += len;
+		skb->len += len;
+	}
+	return skb_put(tail, len);
+}
+#define pskb_put __compat_pskb_put
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 11, 0) && !defined(ISRHEL7)
+#include <net/xfrm.h>
+static inline void skb_scrub_packet(struct sk_buff *skb, bool xnet)
+{
+#ifdef CONFIG_CAVIUM_OCTEON_IPFWD_OFFLOAD
+	memset(&skb->cvm_info, 0, sizeof(skb->cvm_info));
+	skb->cvm_reserved = 0;
+#endif
+	skb->tstamp.tv64 = 0;
+	skb->pkt_type = PACKET_HOST;
+	skb->skb_iif = 0;
+	skb_dst_drop(skb);
+	secpath_reset(skb);
+	nf_reset(skb);
+	nf_reset_trace(skb);
+	if (!xnet)
+		return;
+	skb_orphan(skb);
+	skb->mark = 0;
+}
+#endif
+
+#if (LINUX_VERSION_CODE < KERNEL_VERSION(3, 12, 0) || defined(ISUBUNTU1404)) && !defined(ISRHEL7)
+#include <linux/random.h>
+static inline u32 __compat_prandom_u32_max(u32 ep_ro)
+{
+	return (u32)(((u64)prandom_u32() * ep_ro) >> 32);
+}
+#define prandom_u32_max __compat_prandom_u32_max
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 10, 75) && !defined(ISRHEL7)
+#ifndef U8_MAX
+#define U8_MAX ((u8)~0U)
+#endif
+#ifndef S8_MAX
+#define S8_MAX ((s8)(U8_MAX >> 1))
+#endif
+#ifndef S8_MIN
+#define S8_MIN ((s8)(-S8_MAX - 1))
+#endif
+#ifndef U16_MAX
+#define U16_MAX ((u16)~0U)
+#endif
+#ifndef S16_MAX
+#define S16_MAX ((s16)(U16_MAX >> 1))
+#endif
+#ifndef S16_MIN
+#define S16_MIN ((s16)(-S16_MAX - 1))
+#endif
+#ifndef U32_MAX
+#define U32_MAX ((u32)~0U)
+#endif
+#ifndef S32_MAX
+#define S32_MAX ((s32)(U32_MAX >> 1))
+#endif
+#ifndef S32_MIN
+#define S32_MIN ((s32)(-S32_MAX - 1))
+#endif
+#ifndef U64_MAX
+#define U64_MAX ((u64)~0ULL)
+#endif
+#ifndef S64_MAX
+#define S64_MAX ((s64)(U64_MAX >> 1))
+#endif
+#ifndef S64_MIN
+#define S64_MIN ((s64)(-S64_MAX - 1))
+#endif
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 10, 60) && !defined(ISRHEL7)
+/* Making this static may very well invalidate its usefulness,
+ * but so it goes with compat code. */
+static inline void memzero_explicit(void *s, size_t count)
+{
+	memset(s, 0, count);
+	barrier();
+}
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 12, 0) && !defined(ISRHEL7)
+static const struct in6_addr __compat_in6addr_any = IN6ADDR_ANY_INIT;
+#define in6addr_any __compat_in6addr_any
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 13, 0) && LINUX_VERSION_CODE >= KERNEL_VERSION(4, 2, 0) && !defined(ISOPENSUSE15)
+#include <linux/completion.h>
+#include <linux/random.h>
+#include <linux/errno.h>
+struct rng_initializer {
+	struct completion done;
+	struct random_ready_callback cb;
+};
+static inline void rng_initialized_callback(struct random_ready_callback *cb)
+{
+	complete(&container_of(cb, struct rng_initializer, cb)->done);
+}
+static inline int wait_for_random_bytes(void)
+{
+	static bool rng_is_initialized = false;
+	int ret;
+	if (unlikely(!rng_is_initialized)) {
+		struct rng_initializer rng = {
+			.done = COMPLETION_INITIALIZER(rng.done),
+			.cb = { .owner = THIS_MODULE, .func = rng_initialized_callback }
+		};
+		ret = add_random_ready_callback(&rng.cb);
+		if (!ret) {
+			ret = wait_for_completion_interruptible(&rng.done);
+			if (ret) {
+				del_random_ready_callback(&rng.cb);
+				return ret;
+			}
+		} else if (ret != -EALREADY)
+			return ret;
+		rng_is_initialized = true;
+	}
+	return 0;
+}
+#elif LINUX_VERSION_CODE < KERNEL_VERSION(4, 2, 0)
+/* This is a disaster. Without this API, we really have no way of
+ * knowing if it's initialized. We just return that it has and hope
+ * for the best... */
+static inline int wait_for_random_bytes(void)
+{
+	return 0;
+}
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 19, 0) && LINUX_VERSION_CODE >= KERNEL_VERSION(4, 2, 0) && !defined(ISRHEL8)
+#include <linux/random.h>
+#include <linux/slab.h>
+struct rng_is_initialized_callback {
+	struct random_ready_callback cb;
+	atomic_t *rng_state;
+};
+static inline void rng_is_initialized_callback(struct random_ready_callback *cb)
+{
+	struct rng_is_initialized_callback *rdy = container_of(cb, struct rng_is_initialized_callback, cb);
+	atomic_set(rdy->rng_state, 2);
+	kfree(rdy);
+}
+static inline bool rng_is_initialized(void)
+{
+	static atomic_t rng_state = ATOMIC_INIT(0);
+
+	if (atomic_read(&rng_state) == 2)
+		return true;
+
+	if (atomic_cmpxchg(&rng_state, 0, 1) == 0) {
+		int ret;
+		struct rng_is_initialized_callback *rdy = kmalloc(sizeof(*rdy), GFP_ATOMIC);
+		if (!rdy) {
+			atomic_set(&rng_state, 0);
+			return false;
+		}
+		rdy->cb.owner = THIS_MODULE;
+		rdy->cb.func = rng_is_initialized_callback;
+		rdy->rng_state = &rng_state;
+		ret = add_random_ready_callback(&rdy->cb);
+		if (ret)
+			kfree(rdy);
+		if (ret == -EALREADY) {
+			atomic_set(&rng_state, 2);
+			return true;
+		} else if (ret)
+			atomic_set(&rng_state, 0);
+		return false;
+	}
+	return false;
+}
+#elif LINUX_VERSION_CODE < KERNEL_VERSION(4, 2, 0)
+/* This is a disaster. Without this API, we really have no way of
+ * knowing if it's initialized. We just return that it has and hope
+ * for the best... */
+static inline bool rng_is_initialized(void)
+{
+	return true;
+}
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 13, 0) && !defined(ISOPENSUSE15)
+static inline int get_random_bytes_wait(void *buf, int nbytes)
+{
+	int ret = wait_for_random_bytes();
+	if (unlikely(ret))
+		return ret;
+	get_random_bytes(buf, nbytes);
+	return 0;
+}
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 11, 0) && !defined(ISRHEL7)
+#define system_power_efficient_wq system_unbound_wq
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(5, 3, 0)
+#include <linux/ktime.h>
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 17, 0)
+#include <linux/hrtimer.h>
+#ifndef ktime_get_real_ts64
+#define timespec64 timespec
+#define ktime_get_real_ts64 ktime_get_real_ts
+#endif
+#else
+#include <linux/timekeeping.h>
+#endif
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 11, 0)
+static inline u64 __compat_jiffies64_to_nsecs(u64 j)
+{
+#if !(NSEC_PER_SEC % HZ)
+	return (NSEC_PER_SEC / HZ) * j;
+#else
+	return div_u64(j * HZ_TO_USEC_NUM, HZ_TO_USEC_DEN) * 1000;
+#endif
+}
+#define jiffies64_to_nsecs __compat_jiffies64_to_nsecs
+#endif
+static inline u64 ktime_get_coarse_boottime_ns(void)
+{
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 17, 0)
+	return ktime_to_ns(ktime_get_boottime());
+#elif (LINUX_VERSION_CODE < KERNEL_VERSION(5, 1, 12) && LINUX_VERSION_CODE >= KERNEL_VERSION(4, 20, 0)) || LINUX_VERSION_CODE < KERNEL_VERSION(4, 19, 53)
+	return ktime_to_ns(ktime_mono_to_any(ns_to_ktime(jiffies64_to_nsecs(get_jiffies_64())), TK_OFFS_BOOT));
+#else
+	return ktime_to_ns(ktime_get_coarse_boottime());
+#endif
+}
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 14, 0)
+#include <linux/inetdevice.h>
+static inline __be32 __compat_confirm_addr_indev(struct in_device *in_dev, __be32 dst,  __be32 local, int scope)
+{
+	int same = 0;
+	__be32 addr = 0;
+	for_ifa(in_dev) {
+		if (!addr && (local == ifa->ifa_local || !local) && ifa->ifa_scope <= scope) {
+			addr = ifa->ifa_local;
+			if (same)
+				break;
+		}
+		if (!same) {
+			same = (!local || inet_ifa_match(local, ifa)) && (!dst || inet_ifa_match(dst, ifa));
+			if (same && addr) {
+				if (local || !dst)
+					break;
+				if (inet_ifa_match(addr, ifa))
+					break;
+				if (ifa->ifa_scope <= scope) {
+					addr = ifa->ifa_local;
+					break;
+				}
+				same = 0;
+			}
+		}
+	} endfor_ifa(in_dev);
+	return same ? addr : 0;
+}
+static inline __be32 __compat_inet_confirm_addr(struct net *net, struct in_device *in_dev, __be32 dst, __be32 local, int scope)
+{
+	__be32 addr = 0;
+	struct net_device *dev;
+	if (in_dev)
+		return __compat_confirm_addr_indev(in_dev, dst, local, scope);
+	rcu_read_lock();
+	for_each_netdev_rcu(net, dev) {
+		in_dev = __in_dev_get_rcu(dev);
+		if (in_dev) {
+			addr = __compat_confirm_addr_indev(in_dev, dst, local, scope);
+			if (addr)
+				break;
+		}
+	}
+	rcu_read_unlock();
+	return addr;
+}
+#define inet_confirm_addr __compat_inet_confirm_addr
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 12, 0)
+#include <linux/vmalloc.h>
+#include <linux/mm.h>
+#include <linux/slab.h>
+static inline void *__compat_kvmalloc(size_t size, gfp_t flags)
+{
+	gfp_t kmalloc_flags = flags;
+	void *ret;
+	if (size > PAGE_SIZE) {
+		kmalloc_flags |= __GFP_NOWARN;
+		if (!(kmalloc_flags & __GFP_REPEAT) || (size <= PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER))
+			kmalloc_flags |= __GFP_NORETRY;
+	}
+	ret = kmalloc(size, kmalloc_flags);
+	if (ret || size <= PAGE_SIZE)
+		return ret;
+	return __vmalloc(size, flags, PAGE_KERNEL);
+}
+static inline void *__compat_kvzalloc(size_t size, gfp_t flags)
+{
+	return __compat_kvmalloc(size, flags | __GFP_ZERO);
+}
+#define kvmalloc __compat_kvmalloc
+#define kvzalloc __compat_kvzalloc
+#endif
+
+#if ((LINUX_VERSION_CODE < KERNEL_VERSION(3, 15, 0) && LINUX_VERSION_CODE >= KERNEL_VERSION(3, 13, 0)) || LINUX_VERSION_CODE < KERNEL_VERSION(3, 12, 41)) && !defined(ISUBUNTU1404)
+#include <linux/vmalloc.h>
+#include <linux/mm.h>
+static inline void __compat_kvfree(const void *addr)
+{
+	if (is_vmalloc_addr(addr))
+		vfree(addr);
+	else
+		kfree(addr);
+}
+#define kvfree __compat_kvfree
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 11, 9)
+#include <linux/netdevice.h>
+#define priv_destructor destructor
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 13, 0) && !defined(ISOPENSUSE15)
+#define wg_newlink(a,b,c,d,e) wg_newlink(a,b,c,d)
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 12, 0)
+#include <net/netlink.h>
+#include <net/genetlink.h>
+#define nlmsg_parse(a, b, c, d, e, f) nlmsg_parse(a, b, c, d, e)
+#define nla_parse_nested(a, b, c, d, e) nla_parse_nested(a, b, c, d)
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 10, 0) && !defined(ISRHEL7)
+static inline struct nlattr **genl_family_attrbuf(const struct genl_family *family)
+{
+	return family->attrbuf;
+}
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 12, 0)
+#define PTR_ERR_OR_ZERO(p) PTR_RET(p)
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 7, 0)
+#include <net/netlink.h>
+#define nla_put_u64_64bit(a, b, c, d) nla_put_u64(a, b, c)
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 6, 0)
+#include <net/genetlink.h>
+#ifndef GENL_UNS_ADMIN_PERM
+#define GENL_UNS_ADMIN_PERM GENL_ADMIN_PERM
+#endif
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 10, 0) && !defined(ISRHEL7)
+#include <net/genetlink.h>
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 13, 0)
+#define genl_register_family(a) genl_register_family_with_ops(a, genl_ops, ARRAY_SIZE(genl_ops))
+#define COMPAT_CANNOT_USE_CONST_GENL_OPS
+#else
+#define genl_register_family(a) genl_register_family_with_ops(a, genl_ops)
+#endif
+#define COMPAT_CANNOT_USE_GENL_NOPS
+#endif
+
+#if (LINUX_VERSION_CODE < KERNEL_VERSION(4, 14, 2) && LINUX_VERSION_CODE >= KERNEL_VERSION(4, 14, 0)) || (LINUX_VERSION_CODE < KERNEL_VERSION(4, 13, 16) && LINUX_VERSION_CODE >= KERNEL_VERSION(4, 10, 0)) || (LINUX_VERSION_CODE < KERNEL_VERSION(4, 9, 65) && LINUX_VERSION_CODE >= KERNEL_VERSION(4, 5, 0)) || (LINUX_VERSION_CODE < KERNEL_VERSION(4, 4, 101) && LINUX_VERSION_CODE >= KERNEL_VERSION(3, 19, 0)) || LINUX_VERSION_CODE < KERNEL_VERSION(3, 18, 84)
+#define __COMPAT_NETLINK_DUMP_BLOCK { \
+	int ret; \
+	skb->end -= nlmsg_total_size(sizeof(int)); \
+	ret = wg_get_device_dump_real(skb, cb); \
+	skb->end += nlmsg_total_size(sizeof(int)); \
+	return ret; \
+}
+#define __COMPAT_NETLINK_DUMP_OVERRIDE
+#else
+#define __COMPAT_NETLINK_DUMP_BLOCK return wg_get_device_dump_real(skb, cb);
+#endif
+#if (LINUX_VERSION_CODE < KERNEL_VERSION(4, 15, 8) && LINUX_VERSION_CODE >= KERNEL_VERSION(4, 15, 0)) || (LINUX_VERSION_CODE < KERNEL_VERSION(4, 14, 25) && LINUX_VERSION_CODE >= KERNEL_VERSION(4, 10, 0)) || LINUX_VERSION_CODE < KERNEL_VERSION(4, 9, 87)
+#define wg_get_device_dump(a, b) wg_get_device_dump_real(a, b); \
+static int wg_get_device_dump(a, b) { \
+	struct wg_device *wg = (struct wg_device *)cb->args[0]; \
+	if (!wg) { \
+		int ret = wg_get_device_start(cb); \
+		if (ret) \
+			return ret; \
+	} \
+	__COMPAT_NETLINK_DUMP_BLOCK \
+} \
+static int wg_get_device_dump_real(a, b)
+#define COMPAT_CANNOT_USE_NETLINK_START
+#elif defined(__COMPAT_NETLINK_DUMP_OVERRIDE)
+#define wg_get_device_dump(a, b) wg_get_device_dump_real(a, b); \
+static int wg_get_device_dump(a, b) { \
+	__COMPAT_NETLINK_DUMP_BLOCK \
+} \
+static int wg_get_device_dump_real(a, b)
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 17, 0)
+#define COMPAT_CANNOT_USE_IN6_DEV_GET
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 3, 0)
+#define COMPAT_CANNOT_USE_IFF_NO_QUEUE
+#endif
+
+#if defined(CONFIG_X86_64) && LINUX_VERSION_CODE < KERNEL_VERSION(4, 4, 0)
+#include <asm/user.h>
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 2, 0)
+#include <asm/xsave.h>
+#include <asm/xcr.h>
+static inline int cpu_has_xfeatures(u64 xfeatures_needed, const char **feature_name)
+{
+	return boot_cpu_has(X86_FEATURE_XSAVE) && xgetbv(XCR_XFEATURE_ENABLED_MASK) & xfeatures_needed;
+}
+#endif
+#ifndef XFEATURE_MASK_YMM
+#define XFEATURE_MASK_YMM XSTATE_YMM
+#endif
+#ifndef XFEATURE_MASK_SSE
+#define XFEATURE_MASK_SSE XSTATE_SSE
+#endif
+#ifndef XSTATE_AVX512
+#define XSTATE_AVX512 (XSTATE_OPMASK | XSTATE_ZMM_Hi256 | XSTATE_Hi16_ZMM)
+#endif
+#ifndef XFEATURE_MASK_AVX512
+#define XFEATURE_MASK_AVX512 XSTATE_AVX512
+#endif
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 6, 0) && defined(CONFIG_X86_64)
+/* This is incredibly dumb and reckless, but as it turns out, there's
+ * not really hardware Linux runs properly on that supports F but not BW
+ * and VL, so in practice this isn't so bad. Plus, this is compat layer,
+ * so the bar remains fairly low.
+ */
+#include <asm/cpufeature.h>
+#ifndef X86_FEATURE_AVX512BW
+#define X86_FEATURE_AVX512BW X86_FEATURE_AVX512F
+#endif
+#ifndef X86_FEATURE_AVX512VL
+#define X86_FEATURE_AVX512VL X86_FEATURE_AVX512F
+#endif
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 11, 0)
+struct __compat_dummy_container { char dev; };
+#define netdev_notifier_info net_device *)data); __attribute((unused)) char __compat_dummy_variable = ((struct __compat_dummy_container
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 14, 0)
+#define timer_setup(a, b, c) setup_timer(a, ((void (*)(unsigned long))b), ((unsigned long)a))
+#define from_timer(var, callback_timer, timer_fieldname) container_of(callback_timer, typeof(*var), timer_fieldname)
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 14, 3)
+#define COMPAT_CANNOT_USE_AVX512
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 15, 0)
+#include <net/genetlink.h>
+#define genl_dump_check_consistent(a, b) genl_dump_check_consistent(a, b, &genl_family)
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 13, 0) && !defined(ISRHEL7) && !defined(ISOPENSUSE15)
+static inline void *skb_put_data(struct sk_buff *skb, const void *data, unsigned int len)
+{
+	void *tmp = skb_put(skb, len);
+	memcpy(tmp, data, len);
+	return tmp;
+}
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 19, 0) && !defined(ISRHEL7)
+#define napi_complete_done(n, work_done) napi_complete(n)
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 5, 0)
+#include <linux/netdevice.h>
+/* NAPI_STATE_SCHED gets set by netif_napi_add anyway, so this is safe.
+ * Also, kernels without NAPI_STATE_NO_BUSY_POLL don't have a call to
+ * napi_hash_add inside of netif_napi_add.
+ */
+#define NAPI_STATE_NO_BUSY_POLL NAPI_STATE_SCHED
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 14, 0)
+#include <linux/atomic.h>
+#ifndef atomic_read_acquire
+#define atomic_read_acquire(v) ({ int __compat_p1 = atomic_read(v); smp_rmb(); __compat_p1; })
+#endif
+#ifndef atomic_set_release
+#define atomic_set_release(v, i) ({ smp_wmb(); atomic_set(v, i); })
+#endif
+#elif LINUX_VERSION_CODE < KERNEL_VERSION(4, 3, 0)
+#include <linux/atomic.h>
+#ifndef atomic_read_acquire
+#define atomic_read_acquire(v) smp_load_acquire(&(v)->counter)
+#endif
+#ifndef atomic_set_release
+#define atomic_set_release(v, i) smp_store_release(&(v)->counter, (i))
+#endif
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 17, 0)
+static inline void le32_to_cpu_array(u32 *buf, unsigned int words)
+{
+	while (words--) {
+		__le32_to_cpus(buf);
+		buf++;
+	}
+}
+static inline void cpu_to_le32_array(u32 *buf, unsigned int words)
+{
+	while (words--) {
+		__cpu_to_le32s(buf);
+		buf++;
+	}
+}
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 14, 0)
+#include <crypto/algapi.h>
+static inline void crypto_xor_cpy(u8 *dst, const u8 *src1, const u8 *src2,
+				  unsigned int size)
+{
+	if (IS_ENABLED(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) &&
+	    __builtin_constant_p(size) &&
+	    (size % sizeof(unsigned long)) == 0) {
+		unsigned long *d = (unsigned long *)dst;
+		unsigned long *s1 = (unsigned long *)src1;
+		unsigned long *s2 = (unsigned long *)src2;
+
+		while (size > 0) {
+			*d++ = *s1++ ^ *s2++;
+			size -= sizeof(unsigned long);
+		}
+	} else {
+		if (unlikely(dst != src1))
+			memmove(dst, src1, size);
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 11, 0)
+		crypto_xor(dst, src2, size);
+#else
+		__crypto_xor(dst, src2, size);
+#endif
+	}
+}
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 17, 0)
+#define read_cpuid_part() read_cpuid_part_number()
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 17, 0) && !defined(ISRHEL7)
+#define hlist_add_behind(a, b) hlist_add_after(b, a)
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(5, 0, 0)
+#define totalram_pages() totalram_pages
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 18, 0)
+struct __kernel_timespec {
+	int64_t tv_sec, tv_nsec;
+};
+#elif LINUX_VERSION_CODE < KERNEL_VERSION(5, 1, 0)
+#include <linux/time64.h>
+#ifdef __kernel_timespec
+#undef __kernel_timespec
+struct __kernel_timespec {
+	int64_t tv_sec, tv_nsec;
+};
+#endif
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 12, 0)
+#include <linux/kernel.h>
+#ifndef ALIGN_DOWN
+#define ALIGN_DOWN(x, a) __ALIGN_KERNEL((x) - ((a) - 1), (a))
+#endif
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(5, 1, 0) && !defined(ISRHEL82)
+#include <linux/skbuff.h>
+#define skb_probe_transport_header(a) skb_probe_transport_header(a, 0)
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 16, 0) && !defined(ISRHEL7)
+#define ignore_df local_df
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(5, 1, 0) && !defined(ISRHEL82)
+/* Note that all intentional uses of the non-_bh variety need to explicitly
+ * undef these, conditionalized on COMPAT_CANNOT_DEPRECIATE_BH_RCU.
+ */
+#include <linux/rcupdate.h>
+static __always_inline void old_synchronize_rcu(void)
+{
+	synchronize_rcu();
+}
+static __always_inline void old_call_rcu(void *a, void *b)
+{
+	call_rcu(a, b);
+}
+static __always_inline void old_rcu_barrier(void)
+{
+	rcu_barrier();
+}
+#ifdef synchronize_rcu
+#undef synchronize_rcu
+#endif
+#ifdef call_rcu
+#undef call_rcu
+#endif
+#ifdef rcu_barrier
+#undef rcu_barrier
+#endif
+#define synchronize_rcu synchronize_rcu_bh
+#define call_rcu call_rcu_bh
+#define rcu_barrier rcu_barrier_bh
+#define COMPAT_CANNOT_DEPRECIATE_BH_RCU
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 19, 10) && !defined(ISRHEL8)
+static inline void skb_mark_not_on_list(struct sk_buff *skb)
+{
+	skb->next = NULL;
+}
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 20, 0) && !defined(ISRHEL8)
+#define NLA_EXACT_LEN NLA_UNSPEC
+#endif
+#if LINUX_VERSION_CODE < KERNEL_VERSION(5, 2, 0) && !defined(ISRHEL82)
+#define NLA_MIN_LEN NLA_UNSPEC
+#define COMPAT_CANNOT_INDIVIDUAL_NETLINK_OPS_POLICY
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(5, 2, 0) && defined(__aarch64__)
+#define cpu_have_named_feature(name) (elf_hwcap & (HWCAP_ ## name))
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 1, 0)
+#include <linux/stddef.h>
+#ifndef offsetofend
+#define offsetofend(TYPE, MEMBER) (offsetof(TYPE, MEMBER) + sizeof(((TYPE *)0)->MEMBER))
+#endif
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(5, 5, 0)
+#define genl_dumpit_info(cb) ({ \
+	struct { struct nlattr **attrs; } *a = (void *)((u8 *)cb->args + offsetofend(struct dump_ctx, next_allowedip)); \
+	BUILD_BUG_ON(sizeof(cb->args) < offsetofend(struct dump_ctx, next_allowedip) + sizeof(*a)); \
+	a->attrs = genl_family_attrbuf(&genl_family); \
+	if (nlmsg_parse(cb->nlh, GENL_HDRLEN + genl_family.hdrsize, a->attrs, genl_family.maxattr, device_policy, NULL) < 0) \
+		memset(a->attrs, 0, (genl_family.maxattr + 1) * sizeof(struct nlattr *)); \
+	a; \
+})
+#endif
+
+#if (LINUX_VERSION_CODE < KERNEL_VERSION(5, 4, 5) && LINUX_VERSION_CODE >= KERNEL_VERSION(5, 4, 0)) || LINUX_VERSION_CODE < KERNEL_VERSION(5, 3, 18)
+#define ipv6_dst_lookup_flow(a, b, c, d) ipv6_dst_lookup(a, b, &dst, c) + (void *)0 ?: dst
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(5, 6, 0)
+#include <linux/skbuff.h>
+#ifndef skb_list_walk_safe
+#define skb_list_walk_safe(first, skb, next)                                   \
+	for ((skb) = (first), (next) = (skb) ? (skb)->next : NULL; (skb);      \
+	     (skb) = (next), (next) = (skb) ? (skb)->next : NULL)
+#endif
+#endif
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 5, 0)
+#define blake2s_init zinc_blake2s_init
+#define blake2s_init_key zinc_blake2s_init_key
+#define blake2s_update zinc_blake2s_update
+#define blake2s_final zinc_blake2s_final
+#define blake2s_hmac zinc_blake2s_hmac
+#define chacha20 zinc_chacha20
+#define hchacha20 zinc_hchacha20
+#define chacha20poly1305_encrypt zinc_chacha20poly1305_encrypt
+#define chacha20poly1305_encrypt_sg_inplace zinc_chacha20poly1305_encrypt_sg_inplace
+#define chacha20poly1305_decrypt zinc_chacha20poly1305_decrypt
+#define chacha20poly1305_decrypt_sg_inplace zinc_chacha20poly1305_decrypt_sg_inplace
+#define xchacha20poly1305_encrypt zinc_xchacha20poly1305_encrypt
+#define xchacha20poly1305_decrypt zinc_xchacha20poly1305_decrypt
+#define curve25519 zinc_curve25519
+#define curve25519_generate_secret zinc_curve25519_generate_secret
+#define curve25519_generate_public zinc_curve25519_generate_public
+#define poly1305_init zinc_poly1305_init
+#define poly1305_update zinc_poly1305_update
+#define poly1305_final zinc_poly1305_final
+#define blake2s_compress_ssse3 zinc_blake2s_compress_ssse3
+#define blake2s_compress_avx512 zinc_blake2s_compress_avx512
+#define poly1305_init_arm zinc_poly1305_init_arm
+#define poly1305_blocks_arm zinc_poly1305_blocks_arm
+#define poly1305_emit_arm zinc_poly1305_emit_arm
+#define poly1305_blocks_neon zinc_poly1305_blocks_neon
+#define poly1305_emit_neon zinc_poly1305_emit_neon
+#define poly1305_init_mips zinc_poly1305_init_mips
+#define poly1305_blocks_mips zinc_poly1305_blocks_mips
+#define poly1305_emit_mips zinc_poly1305_emit_mips
+#define poly1305_init_x86_64 zinc_poly1305_init_x86_64
+#define poly1305_blocks_x86_64 zinc_poly1305_blocks_x86_64
+#define poly1305_emit_x86_64 zinc_poly1305_emit_x86_64
+#define poly1305_emit_avx zinc_poly1305_emit_avx
+#define poly1305_blocks_avx zinc_poly1305_blocks_avx
+#define poly1305_blocks_avx2 zinc_poly1305_blocks_avx2
+#define poly1305_blocks_avx512 zinc_poly1305_blocks_avx512
+#define curve25519_neon zinc_curve25519_neon
+#define hchacha20_ssse3 zinc_hchacha20_ssse3
+#define chacha20_ssse3 zinc_chacha20_ssse3
+#define chacha20_avx2 zinc_chacha20_avx2
+#define chacha20_avx512 zinc_chacha20_avx512
+#define chacha20_avx512vl zinc_chacha20_avx512vl
+#define chacha20_mips zinc_chacha20_mips
+#define chacha20_arm zinc_chacha20_arm
+#define hchacha20_arm zinc_hchacha20_arm
+#define chacha20_neon zinc_chacha20_neon
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 19, 0)
+#include <linux/skbuff.h>
+static inline int skb_ensure_writable(struct sk_buff *skb, int write_len)
+{
+	if (!pskb_may_pull(skb, write_len))
+		return -ENOMEM;
+
+	if (!skb_cloned(skb) || skb_clone_writable(skb, write_len))
+		return 0;
+
+	return pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
+}
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(5, 6, 0)
+#if IS_ENABLED(CONFIG_NF_NAT)
+#include <linux/ip.h>
+#include <linux/icmpv6.h>
+#include <net/ipv6.h>
+#include <net/icmp.h>
+#include <net/netfilter/nf_conntrack.h>
+#if LINUX_VERSION_CODE < KERNEL_VERSION(5, 1, 0)
+#include <net/netfilter/nf_nat_core.h>
+#endif
+static inline void icmp_ndo_send(struct sk_buff *skb_in, int type, int code, __be32 info)
+{
+	struct sk_buff *cloned_skb = NULL;
+	enum ip_conntrack_info ctinfo;
+	struct nf_conn *ct;
+	__be32 orig_ip;
+
+	ct = nf_ct_get(skb_in, &ctinfo);
+	if (!ct || !(ct->status & IPS_SRC_NAT)) {
+		icmp_send(skb_in, type, code, info);
+		return;
+	}
+
+	if (skb_shared(skb_in))
+		skb_in = cloned_skb = skb_clone(skb_in, GFP_ATOMIC);
+
+	if (unlikely(!skb_in || skb_network_header(skb_in) < skb_in->head ||
+	    (skb_network_header(skb_in) + sizeof(struct iphdr)) >
+	    skb_tail_pointer(skb_in) || skb_ensure_writable(skb_in,
+	    skb_network_offset(skb_in) + sizeof(struct iphdr))))
+		goto out;
+
+	orig_ip = ip_hdr(skb_in)->saddr;
+	ip_hdr(skb_in)->saddr = ct->tuplehash[0].tuple.src.u3.ip;
+	icmp_send(skb_in, type, code, info);
+	ip_hdr(skb_in)->saddr = orig_ip;
+out:
+	consume_skb(cloned_skb);
+}
+static inline void icmpv6_ndo_send(struct sk_buff *skb_in, u8 type, u8 code, __u32 info)
+{
+	struct sk_buff *cloned_skb = NULL;
+	enum ip_conntrack_info ctinfo;
+	struct in6_addr orig_ip;
+	struct nf_conn *ct;
+
+	ct = nf_ct_get(skb_in, &ctinfo);
+	if (!ct || !(ct->status & IPS_SRC_NAT)) {
+		icmpv6_send(skb_in, type, code, info);
+		return;
+	}
+
+	if (skb_shared(skb_in))
+		skb_in = cloned_skb = skb_clone(skb_in, GFP_ATOMIC);
+
+	if (unlikely(!skb_in || skb_network_header(skb_in) < skb_in->head ||
+	    (skb_network_header(skb_in) + sizeof(struct ipv6hdr)) >
+	    skb_tail_pointer(skb_in) || skb_ensure_writable(skb_in,
+	    skb_network_offset(skb_in) + sizeof(struct ipv6hdr))))
+		goto out;
+
+	orig_ip = ipv6_hdr(skb_in)->saddr;
+	ipv6_hdr(skb_in)->saddr = ct->tuplehash[0].tuple.src.u3.in6;
+	icmpv6_send(skb_in, type, code, info);
+	ipv6_hdr(skb_in)->saddr = orig_ip;
+out:
+	consume_skb(cloned_skb);
+}
+#else
+#define icmp_ndo_send icmp_send
+#define icmpv6_ndo_send icmpv6_send
+#endif
+#endif
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 10, 0)
+#define COMPAT_CANNOT_USE_MAX_MTU
+#endif
+
+#if defined(ISUBUNTU1604)
+#include <linux/siphash.h>
+#ifndef _WG_LINUX_SIPHASH_H
+#define hsiphash_2u32 siphash_2u32
+#define hsiphash_3u32 siphash_3u32
+#define hsiphash_key_t siphash_key_t
+#endif
+#endif
+
+#ifdef CONFIG_VE
+#include <linux/netdev_features.h>
+#ifdef NETIF_F_VIRTUAL
+#undef NETIF_F_LLTX
+#define NETIF_F_LLTX (__NETIF_F(LLTX) | __NETIF_F(VIRTUAL))
+#endif
+#endif
+
+/* https://github.com/ClangBuiltLinux/linux/issues/7 */
+#if defined( __clang__) && (!defined(CONFIG_CLANG_VERSION) || CONFIG_CLANG_VERSION < 80000)
+#include <linux/bug.h>
+#undef BUILD_BUG_ON
+#define BUILD_BUG_ON(x)
+#endif
+
+/* PaX compatibility */
+#ifdef CONSTIFY_PLUGIN
+#include <linux/cache.h>
+#undef __read_mostly
+#define __read_mostly
+#endif
+#if (defined(RAP_PLUGIN) || defined(CONFIG_CFI_CLANG)) && LINUX_VERSION_CODE < KERNEL_VERSION(4, 15, 0)
+#include <linux/timer.h>
+#define wg_expired_retransmit_handshake(a) wg_expired_retransmit_handshake(unsigned long timer)
+#define wg_expired_send_keepalive(a) wg_expired_send_keepalive(unsigned long timer)
+#define wg_expired_new_handshake(a) wg_expired_new_handshake(unsigned long timer)
+#define wg_expired_zero_key_material(a) wg_expired_zero_key_material(unsigned long timer)
+#define wg_expired_send_persistent_keepalive(a) wg_expired_send_persistent_keepalive(unsigned long timer)
+#undef timer_setup
+#define timer_setup(a, b, c) setup_timer(a, ((void (*)(unsigned long))b), ((unsigned long)a))
+#undef from_timer
+#define from_timer(var, callback_timer, timer_fieldname) container_of((struct timer_list *)callback_timer, typeof(*var), timer_fieldname)
+#endif
+
+#endif /* _WG_COMPAT_H */
diff --git a/net/wireguard/crypto/zinc/curve25519/curve25519-hacl64.c b/net/wireguard/crypto/zinc/curve25519/curve25519-hacl64.c
new file mode 100644
index 000000000000..d6dcd0ce1892
--- /dev/null
+++ b/net/wireguard/crypto/zinc/curve25519/curve25519-hacl64.c
@@ -0,0 +1,779 @@
+// SPDX-License-Identifier: GPL-2.0 OR MIT
+/*
+ * Copyright (C) 2016-2017 INRIA and Microsoft Corporation.
+ * Copyright (C) 2018-2019 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved.
+ *
+ * This is a machine-generated formally verified implementation of Curve25519
+ * ECDH from: <https://github.com/mitls/hacl-star>. Though originally machine
+ * generated, it has been tweaked to be suitable for use in the kernel. It is
+ * optimized for 64-bit machines that can efficiently work with 128-bit
+ * integer types.
+ */
+
+typedef __uint128_t u128;
+
+static __always_inline u64 u64_eq_mask(u64 a, u64 b)
+{
+	u64 x = a ^ b;
+	u64 minus_x = ~x + (u64)1U;
+	u64 x_or_minus_x = x | minus_x;
+	u64 xnx = x_or_minus_x >> (u32)63U;
+	u64 c = xnx - (u64)1U;
+	return c;
+}
+
+static __always_inline u64 u64_gte_mask(u64 a, u64 b)
+{
+	u64 x = a;
+	u64 y = b;
+	u64 x_xor_y = x ^ y;
+	u64 x_sub_y = x - y;
+	u64 x_sub_y_xor_y = x_sub_y ^ y;
+	u64 q = x_xor_y | x_sub_y_xor_y;
+	u64 x_xor_q = x ^ q;
+	u64 x_xor_q_ = x_xor_q >> (u32)63U;
+	u64 c = x_xor_q_ - (u64)1U;
+	return c;
+}
+
+static __always_inline void modulo_carry_top(u64 *b)
+{
+	u64 b4 = b[4];
+	u64 b0 = b[0];
+	u64 b4_ = b4 & 0x7ffffffffffffLLU;
+	u64 b0_ = b0 + 19 * (b4 >> 51);
+	b[4] = b4_;
+	b[0] = b0_;
+}
+
+static __always_inline void fproduct_copy_from_wide_(u64 *output, u128 *input)
+{
+	{
+		u128 xi = input[0];
+		output[0] = ((u64)(xi));
+	}
+	{
+		u128 xi = input[1];
+		output[1] = ((u64)(xi));
+	}
+	{
+		u128 xi = input[2];
+		output[2] = ((u64)(xi));
+	}
+	{
+		u128 xi = input[3];
+		output[3] = ((u64)(xi));
+	}
+	{
+		u128 xi = input[4];
+		output[4] = ((u64)(xi));
+	}
+}
+
+static __always_inline void
+fproduct_sum_scalar_multiplication_(u128 *output, u64 *input, u64 s)
+{
+	output[0] += (u128)input[0] * s;
+	output[1] += (u128)input[1] * s;
+	output[2] += (u128)input[2] * s;
+	output[3] += (u128)input[3] * s;
+	output[4] += (u128)input[4] * s;
+}
+
+static __always_inline void fproduct_carry_wide_(u128 *tmp)
+{
+	{
+		u32 ctr = 0;
+		u128 tctr = tmp[ctr];
+		u128 tctrp1 = tmp[ctr + 1];
+		u64 r0 = ((u64)(tctr)) & 0x7ffffffffffffLLU;
+		u128 c = ((tctr) >> (51));
+		tmp[ctr] = ((u128)(r0));
+		tmp[ctr + 1] = ((tctrp1) + (c));
+	}
+	{
+		u32 ctr = 1;
+		u128 tctr = tmp[ctr];
+		u128 tctrp1 = tmp[ctr + 1];
+		u64 r0 = ((u64)(tctr)) & 0x7ffffffffffffLLU;
+		u128 c = ((tctr) >> (51));
+		tmp[ctr] = ((u128)(r0));
+		tmp[ctr + 1] = ((tctrp1) + (c));
+	}
+
+	{
+		u32 ctr = 2;
+		u128 tctr = tmp[ctr];
+		u128 tctrp1 = tmp[ctr + 1];
+		u64 r0 = ((u64)(tctr)) & 0x7ffffffffffffLLU;
+		u128 c = ((tctr) >> (51));
+		tmp[ctr] = ((u128)(r0));
+		tmp[ctr + 1] = ((tctrp1) + (c));
+	}
+	{
+		u32 ctr = 3;
+		u128 tctr = tmp[ctr];
+		u128 tctrp1 = tmp[ctr + 1];
+		u64 r0 = ((u64)(tctr)) & 0x7ffffffffffffLLU;
+		u128 c = ((tctr) >> (51));
+		tmp[ctr] = ((u128)(r0));
+		tmp[ctr + 1] = ((tctrp1) + (c));
+	}
+}
+
+static __always_inline void fmul_shift_reduce(u64 *output)
+{
+	u64 tmp = output[4];
+	u64 b0;
+	{
+		u32 ctr = 5 - 0 - 1;
+		u64 z = output[ctr - 1];
+		output[ctr] = z;
+	}
+	{
+		u32 ctr = 5 - 1 - 1;
+		u64 z = output[ctr - 1];
+		output[ctr] = z;
+	}
+	{
+		u32 ctr = 5 - 2 - 1;
+		u64 z = output[ctr - 1];
+		output[ctr] = z;
+	}
+	{
+		u32 ctr = 5 - 3 - 1;
+		u64 z = output[ctr - 1];
+		output[ctr] = z;
+	}
+	output[0] = tmp;
+	b0 = output[0];
+	output[0] = 19 * b0;
+}
+
+static __always_inline void fmul_mul_shift_reduce_(u128 *output, u64 *input,
+						   u64 *input21)
+{
+	u32 i;
+	u64 input2i;
+	{
+		u64 input2i = input21[0];
+		fproduct_sum_scalar_multiplication_(output, input, input2i);
+		fmul_shift_reduce(input);
+	}
+	{
+		u64 input2i = input21[1];
+		fproduct_sum_scalar_multiplication_(output, input, input2i);
+		fmul_shift_reduce(input);
+	}
+	{
+		u64 input2i = input21[2];
+		fproduct_sum_scalar_multiplication_(output, input, input2i);
+		fmul_shift_reduce(input);
+	}
+	{
+		u64 input2i = input21[3];
+		fproduct_sum_scalar_multiplication_(output, input, input2i);
+		fmul_shift_reduce(input);
+	}
+	i = 4;
+	input2i = input21[i];
+	fproduct_sum_scalar_multiplication_(output, input, input2i);
+}
+
+static __always_inline void fmul_fmul(u64 *output, u64 *input, u64 *input21)
+{
+	u64 tmp[5] = { input[0], input[1], input[2], input[3], input[4] };
+	{
+		u128 b4;
+		u128 b0;
+		u128 b4_;
+		u128 b0_;
+		u64 i0;
+		u64 i1;
+		u64 i0_;
+		u64 i1_;
+		u128 t[5] = { 0 };
+		fmul_mul_shift_reduce_(t, tmp, input21);
+		fproduct_carry_wide_(t);
+		b4 = t[4];
+		b0 = t[0];
+		b4_ = ((b4) & (((u128)(0x7ffffffffffffLLU))));
+		b0_ = ((b0) + (((u128)(19) * (((u64)(((b4) >> (51))))))));
+		t[4] = b4_;
+		t[0] = b0_;
+		fproduct_copy_from_wide_(output, t);
+		i0 = output[0];
+		i1 = output[1];
+		i0_ = i0 & 0x7ffffffffffffLLU;
+		i1_ = i1 + (i0 >> 51);
+		output[0] = i0_;
+		output[1] = i1_;
+	}
+}
+
+static __always_inline void fsquare_fsquare__(u128 *tmp, u64 *output)
+{
+	u64 r0 = output[0];
+	u64 r1 = output[1];
+	u64 r2 = output[2];
+	u64 r3 = output[3];
+	u64 r4 = output[4];
+	u64 d0 = r0 * 2;
+	u64 d1 = r1 * 2;
+	u64 d2 = r2 * 2 * 19;
+	u64 d419 = r4 * 19;
+	u64 d4 = d419 * 2;
+	u128 s0 = ((((((u128)(r0) * (r0))) + (((u128)(d4) * (r1))))) +
+		   (((u128)(d2) * (r3))));
+	u128 s1 = ((((((u128)(d0) * (r1))) + (((u128)(d4) * (r2))))) +
+		   (((u128)(r3 * 19) * (r3))));
+	u128 s2 = ((((((u128)(d0) * (r2))) + (((u128)(r1) * (r1))))) +
+		   (((u128)(d4) * (r3))));
+	u128 s3 = ((((((u128)(d0) * (r3))) + (((u128)(d1) * (r2))))) +
+		   (((u128)(r4) * (d419))));
+	u128 s4 = ((((((u128)(d0) * (r4))) + (((u128)(d1) * (r3))))) +
+		   (((u128)(r2) * (r2))));
+	tmp[0] = s0;
+	tmp[1] = s1;
+	tmp[2] = s2;
+	tmp[3] = s3;
+	tmp[4] = s4;
+}
+
+static __always_inline void fsquare_fsquare_(u128 *tmp, u64 *output)
+{
+	u128 b4;
+	u128 b0;
+	u128 b4_;
+	u128 b0_;
+	u64 i0;
+	u64 i1;
+	u64 i0_;
+	u64 i1_;
+	fsquare_fsquare__(tmp, output);
+	fproduct_carry_wide_(tmp);
+	b4 = tmp[4];
+	b0 = tmp[0];
+	b4_ = ((b4) & (((u128)(0x7ffffffffffffLLU))));
+	b0_ = ((b0) + (((u128)(19) * (((u64)(((b4) >> (51))))))));
+	tmp[4] = b4_;
+	tmp[0] = b0_;
+	fproduct_copy_from_wide_(output, tmp);
+	i0 = output[0];
+	i1 = output[1];
+	i0_ = i0 & 0x7ffffffffffffLLU;
+	i1_ = i1 + (i0 >> 51);
+	output[0] = i0_;
+	output[1] = i1_;
+}
+
+static __always_inline void fsquare_fsquare_times_(u64 *output, u128 *tmp,
+						   u32 count1)
+{
+	u32 i;
+	fsquare_fsquare_(tmp, output);
+	for (i = 1; i < count1; ++i)
+		fsquare_fsquare_(tmp, output);
+}
+
+static __always_inline void fsquare_fsquare_times(u64 *output, u64 *input,
+						  u32 count1)
+{
+	u128 t[5];
+	memcpy(output, input, 5 * sizeof(*input));
+	fsquare_fsquare_times_(output, t, count1);
+}
+
+static __always_inline void fsquare_fsquare_times_inplace(u64 *output,
+							  u32 count1)
+{
+	u128 t[5];
+	fsquare_fsquare_times_(output, t, count1);
+}
+
+static __always_inline void crecip_crecip(u64 *out, u64 *z)
+{
+	u64 buf[20] = { 0 };
+	u64 *a0 = buf;
+	u64 *t00 = buf + 5;
+	u64 *b0 = buf + 10;
+	u64 *t01;
+	u64 *b1;
+	u64 *c0;
+	u64 *a;
+	u64 *t0;
+	u64 *b;
+	u64 *c;
+	fsquare_fsquare_times(a0, z, 1);
+	fsquare_fsquare_times(t00, a0, 2);
+	fmul_fmul(b0, t00, z);
+	fmul_fmul(a0, b0, a0);
+	fsquare_fsquare_times(t00, a0, 1);
+	fmul_fmul(b0, t00, b0);
+	fsquare_fsquare_times(t00, b0, 5);
+	t01 = buf + 5;
+	b1 = buf + 10;
+	c0 = buf + 15;
+	fmul_fmul(b1, t01, b1);
+	fsquare_fsquare_times(t01, b1, 10);
+	fmul_fmul(c0, t01, b1);
+	fsquare_fsquare_times(t01, c0, 20);
+	fmul_fmul(t01, t01, c0);
+	fsquare_fsquare_times_inplace(t01, 10);
+	fmul_fmul(b1, t01, b1);
+	fsquare_fsquare_times(t01, b1, 50);
+	a = buf;
+	t0 = buf + 5;
+	b = buf + 10;
+	c = buf + 15;
+	fmul_fmul(c, t0, b);
+	fsquare_fsquare_times(t0, c, 100);
+	fmul_fmul(t0, t0, c);
+	fsquare_fsquare_times_inplace(t0, 50);
+	fmul_fmul(t0, t0, b);
+	fsquare_fsquare_times_inplace(t0, 5);
+	fmul_fmul(out, t0, a);
+}
+
+static __always_inline void fsum(u64 *a, u64 *b)
+{
+	a[0] += b[0];
+	a[1] += b[1];
+	a[2] += b[2];
+	a[3] += b[3];
+	a[4] += b[4];
+}
+
+static __always_inline void fdifference(u64 *a, u64 *b)
+{
+	u64 tmp[5] = { 0 };
+	u64 b0;
+	u64 b1;
+	u64 b2;
+	u64 b3;
+	u64 b4;
+	memcpy(tmp, b, 5 * sizeof(*b));
+	b0 = tmp[0];
+	b1 = tmp[1];
+	b2 = tmp[2];
+	b3 = tmp[3];
+	b4 = tmp[4];
+	tmp[0] = b0 + 0x3fffffffffff68LLU;
+	tmp[1] = b1 + 0x3ffffffffffff8LLU;
+	tmp[2] = b2 + 0x3ffffffffffff8LLU;
+	tmp[3] = b3 + 0x3ffffffffffff8LLU;
+	tmp[4] = b4 + 0x3ffffffffffff8LLU;
+	{
+		u64 xi = a[0];
+		u64 yi = tmp[0];
+		a[0] = yi - xi;
+	}
+	{
+		u64 xi = a[1];
+		u64 yi = tmp[1];
+		a[1] = yi - xi;
+	}
+	{
+		u64 xi = a[2];
+		u64 yi = tmp[2];
+		a[2] = yi - xi;
+	}
+	{
+		u64 xi = a[3];
+		u64 yi = tmp[3];
+		a[3] = yi - xi;
+	}
+	{
+		u64 xi = a[4];
+		u64 yi = tmp[4];
+		a[4] = yi - xi;
+	}
+}
+
+static __always_inline void fscalar(u64 *output, u64 *b, u64 s)
+{
+	u128 tmp[5];
+	u128 b4;
+	u128 b0;
+	u128 b4_;
+	u128 b0_;
+	{
+		u64 xi = b[0];
+		tmp[0] = ((u128)(xi) * (s));
+	}
+	{
+		u64 xi = b[1];
+		tmp[1] = ((u128)(xi) * (s));
+	}
+	{
+		u64 xi = b[2];
+		tmp[2] = ((u128)(xi) * (s));
+	}
+	{
+		u64 xi = b[3];
+		tmp[3] = ((u128)(xi) * (s));
+	}
+	{
+		u64 xi = b[4];
+		tmp[4] = ((u128)(xi) * (s));
+	}
+	fproduct_carry_wide_(tmp);
+	b4 = tmp[4];
+	b0 = tmp[0];
+	b4_ = ((b4) & (((u128)(0x7ffffffffffffLLU))));
+	b0_ = ((b0) + (((u128)(19) * (((u64)(((b4) >> (51))))))));
+	tmp[4] = b4_;
+	tmp[0] = b0_;
+	fproduct_copy_from_wide_(output, tmp);
+}
+
+static __always_inline void crecip(u64 *output, u64 *input)
+{
+	crecip_crecip(output, input);
+}
+
+static __always_inline void point_swap_conditional_step(u64 *a, u64 *b,
+							u64 swap1, u32 ctr)
+{
+	u32 i = ctr - 1;
+	u64 ai = a[i];
+	u64 bi = b[i];
+	u64 x = swap1 & (ai ^ bi);
+	u64 ai1 = ai ^ x;
+	u64 bi1 = bi ^ x;
+	a[i] = ai1;
+	b[i] = bi1;
+}
+
+static __always_inline void point_swap_conditional5(u64 *a, u64 *b, u64 swap1)
+{
+	point_swap_conditional_step(a, b, swap1, 5);
+	point_swap_conditional_step(a, b, swap1, 4);
+	point_swap_conditional_step(a, b, swap1, 3);
+	point_swap_conditional_step(a, b, swap1, 2);
+	point_swap_conditional_step(a, b, swap1, 1);
+}
+
+static __always_inline void point_swap_conditional(u64 *a, u64 *b, u64 iswap)
+{
+	u64 swap1 = 0 - iswap;
+	point_swap_conditional5(a, b, swap1);
+	point_swap_conditional5(a + 5, b + 5, swap1);
+}
+
+static __always_inline void point_copy(u64 *output, u64 *input)
+{
+	memcpy(output, input, 5 * sizeof(*input));
+	memcpy(output + 5, input + 5, 5 * sizeof(*input));
+}
+
+static __always_inline void addanddouble_fmonty(u64 *pp, u64 *ppq, u64 *p,
+						u64 *pq, u64 *qmqp)
+{
+	u64 *qx = qmqp;
+	u64 *x2 = pp;
+	u64 *z2 = pp + 5;
+	u64 *x3 = ppq;
+	u64 *z3 = ppq + 5;
+	u64 *x = p;
+	u64 *z = p + 5;
+	u64 *xprime = pq;
+	u64 *zprime = pq + 5;
+	u64 buf[40] = { 0 };
+	u64 *origx = buf;
+	u64 *origxprime0 = buf + 5;
+	u64 *xxprime0;
+	u64 *zzprime0;
+	u64 *origxprime;
+	xxprime0 = buf + 25;
+	zzprime0 = buf + 30;
+	memcpy(origx, x, 5 * sizeof(*x));
+	fsum(x, z);
+	fdifference(z, origx);
+	memcpy(origxprime0, xprime, 5 * sizeof(*xprime));
+	fsum(xprime, zprime);
+	fdifference(zprime, origxprime0);
+	fmul_fmul(xxprime0, xprime, z);
+	fmul_fmul(zzprime0, x, zprime);
+	origxprime = buf + 5;
+	{
+		u64 *xx0;
+		u64 *zz0;
+		u64 *xxprime;
+		u64 *zzprime;
+		u64 *zzzprime;
+		xx0 = buf + 15;
+		zz0 = buf + 20;
+		xxprime = buf + 25;
+		zzprime = buf + 30;
+		zzzprime = buf + 35;
+		memcpy(origxprime, xxprime, 5 * sizeof(*xxprime));
+		fsum(xxprime, zzprime);
+		fdifference(zzprime, origxprime);
+		fsquare_fsquare_times(x3, xxprime, 1);
+		fsquare_fsquare_times(zzzprime, zzprime, 1);
+		fmul_fmul(z3, zzzprime, qx);
+		fsquare_fsquare_times(xx0, x, 1);
+		fsquare_fsquare_times(zz0, z, 1);
+		{
+			u64 *zzz;
+			u64 *xx;
+			u64 *zz;
+			u64 scalar;
+			zzz = buf + 10;
+			xx = buf + 15;
+			zz = buf + 20;
+			fmul_fmul(x2, xx, zz);
+			fdifference(zz, xx);
+			scalar = 121665;
+			fscalar(zzz, zz, scalar);
+			fsum(zzz, xx);
+			fmul_fmul(z2, zzz, zz);
+		}
+	}
+}
+
+static __always_inline void
+ladder_smallloop_cmult_small_loop_step(u64 *nq, u64 *nqpq, u64 *nq2, u64 *nqpq2,
+				       u64 *q, u8 byt)
+{
+	u64 bit0 = (u64)(byt >> 7);
+	u64 bit;
+	point_swap_conditional(nq, nqpq, bit0);
+	addanddouble_fmonty(nq2, nqpq2, nq, nqpq, q);
+	bit = (u64)(byt >> 7);
+	point_swap_conditional(nq2, nqpq2, bit);
+}
+
+static __always_inline void
+ladder_smallloop_cmult_small_loop_double_step(u64 *nq, u64 *nqpq, u64 *nq2,
+					      u64 *nqpq2, u64 *q, u8 byt)
+{
+	u8 byt1;
+	ladder_smallloop_cmult_small_loop_step(nq, nqpq, nq2, nqpq2, q, byt);
+	byt1 = byt << 1;
+	ladder_smallloop_cmult_small_loop_step(nq2, nqpq2, nq, nqpq, q, byt1);
+}
+
+static __always_inline void
+ladder_smallloop_cmult_small_loop(u64 *nq, u64 *nqpq, u64 *nq2, u64 *nqpq2,
+				  u64 *q, u8 byt, u32 i)
+{
+	while (i--) {
+		ladder_smallloop_cmult_small_loop_double_step(nq, nqpq, nq2,
+							      nqpq2, q, byt);
+		byt <<= 2;
+	}
+}
+
+static __always_inline void ladder_bigloop_cmult_big_loop(u8 *n1, u64 *nq,
+							  u64 *nqpq, u64 *nq2,
+							  u64 *nqpq2, u64 *q,
+							  u32 i)
+{
+	while (i--) {
+		u8 byte = n1[i];
+		ladder_smallloop_cmult_small_loop(nq, nqpq, nq2, nqpq2, q,
+						  byte, 4);
+	}
+}
+
+static void ladder_cmult(u64 *result, u8 *n1, u64 *q)
+{
+	u64 point_buf[40] = { 0 };
+	u64 *nq = point_buf;
+	u64 *nqpq = point_buf + 10;
+	u64 *nq2 = point_buf + 20;
+	u64 *nqpq2 = point_buf + 30;
+	point_copy(nqpq, q);
+	nq[0] = 1;
+	ladder_bigloop_cmult_big_loop(n1, nq, nqpq, nq2, nqpq2, q, 32);
+	point_copy(result, nq);
+}
+
+static __always_inline void format_fexpand(u64 *output, const u8 *input)
+{
+	const u8 *x00 = input + 6;
+	const u8 *x01 = input + 12;
+	const u8 *x02 = input + 19;
+	const u8 *x0 = input + 24;
+	u64 i0, i1, i2, i3, i4, output0, output1, output2, output3, output4;
+	i0 = get_unaligned_le64(input);
+	i1 = get_unaligned_le64(x00);
+	i2 = get_unaligned_le64(x01);
+	i3 = get_unaligned_le64(x02);
+	i4 = get_unaligned_le64(x0);
+	output0 = i0 & 0x7ffffffffffffLLU;
+	output1 = i1 >> 3 & 0x7ffffffffffffLLU;
+	output2 = i2 >> 6 & 0x7ffffffffffffLLU;
+	output3 = i3 >> 1 & 0x7ffffffffffffLLU;
+	output4 = i4 >> 12 & 0x7ffffffffffffLLU;
+	output[0] = output0;
+	output[1] = output1;
+	output[2] = output2;
+	output[3] = output3;
+	output[4] = output4;
+}
+
+static __always_inline void format_fcontract_first_carry_pass(u64 *input)
+{
+	u64 t0 = input[0];
+	u64 t1 = input[1];
+	u64 t2 = input[2];
+	u64 t3 = input[3];
+	u64 t4 = input[4];
+	u64 t1_ = t1 + (t0 >> 51);
+	u64 t0_ = t0 & 0x7ffffffffffffLLU;
+	u64 t2_ = t2 + (t1_ >> 51);
+	u64 t1__ = t1_ & 0x7ffffffffffffLLU;
+	u64 t3_ = t3 + (t2_ >> 51);
+	u64 t2__ = t2_ & 0x7ffffffffffffLLU;
+	u64 t4_ = t4 + (t3_ >> 51);
+	u64 t3__ = t3_ & 0x7ffffffffffffLLU;
+	input[0] = t0_;
+	input[1] = t1__;
+	input[2] = t2__;
+	input[3] = t3__;
+	input[4] = t4_;
+}
+
+static __always_inline void format_fcontract_first_carry_full(u64 *input)
+{
+	format_fcontract_first_carry_pass(input);
+	modulo_carry_top(input);
+}
+
+static __always_inline void format_fcontract_second_carry_pass(u64 *input)
+{
+	u64 t0 = input[0];
+	u64 t1 = input[1];
+	u64 t2 = input[2];
+	u64 t3 = input[3];
+	u64 t4 = input[4];
+	u64 t1_ = t1 + (t0 >> 51);
+	u64 t0_ = t0 & 0x7ffffffffffffLLU;
+	u64 t2_ = t2 + (t1_ >> 51);
+	u64 t1__ = t1_ & 0x7ffffffffffffLLU;
+	u64 t3_ = t3 + (t2_ >> 51);
+	u64 t2__ = t2_ & 0x7ffffffffffffLLU;
+	u64 t4_ = t4 + (t3_ >> 51);
+	u64 t3__ = t3_ & 0x7ffffffffffffLLU;
+	input[0] = t0_;
+	input[1] = t1__;
+	input[2] = t2__;
+	input[3] = t3__;
+	input[4] = t4_;
+}
+
+static __always_inline void format_fcontract_second_carry_full(u64 *input)
+{
+	u64 i0;
+	u64 i1;
+	u64 i0_;
+	u64 i1_;
+	format_fcontract_second_carry_pass(input);
+	modulo_carry_top(input);
+	i0 = input[0];
+	i1 = input[1];
+	i0_ = i0 & 0x7ffffffffffffLLU;
+	i1_ = i1 + (i0 >> 51);
+	input[0] = i0_;
+	input[1] = i1_;
+}
+
+static __always_inline void format_fcontract_trim(u64 *input)
+{
+	u64 a0 = input[0];
+	u64 a1 = input[1];
+	u64 a2 = input[2];
+	u64 a3 = input[3];
+	u64 a4 = input[4];
+	u64 mask0 = u64_gte_mask(a0, 0x7ffffffffffedLLU);
+	u64 mask1 = u64_eq_mask(a1, 0x7ffffffffffffLLU);
+	u64 mask2 = u64_eq_mask(a2, 0x7ffffffffffffLLU);
+	u64 mask3 = u64_eq_mask(a3, 0x7ffffffffffffLLU);
+	u64 mask4 = u64_eq_mask(a4, 0x7ffffffffffffLLU);
+	u64 mask = (((mask0 & mask1) & mask2) & mask3) & mask4;
+	u64 a0_ = a0 - (0x7ffffffffffedLLU & mask);
+	u64 a1_ = a1 - (0x7ffffffffffffLLU & mask);
+	u64 a2_ = a2 - (0x7ffffffffffffLLU & mask);
+	u64 a3_ = a3 - (0x7ffffffffffffLLU & mask);
+	u64 a4_ = a4 - (0x7ffffffffffffLLU & mask);
+	input[0] = a0_;
+	input[1] = a1_;
+	input[2] = a2_;
+	input[3] = a3_;
+	input[4] = a4_;
+}
+
+static __always_inline void format_fcontract_store(u8 *output, u64 *input)
+{
+	u64 t0 = input[0];
+	u64 t1 = input[1];
+	u64 t2 = input[2];
+	u64 t3 = input[3];
+	u64 t4 = input[4];
+	u64 o0 = t1 << 51 | t0;
+	u64 o1 = t2 << 38 | t1 >> 13;
+	u64 o2 = t3 << 25 | t2 >> 26;
+	u64 o3 = t4 << 12 | t3 >> 39;
+	u8 *b0 = output;
+	u8 *b1 = output + 8;
+	u8 *b2 = output + 16;
+	u8 *b3 = output + 24;
+	put_unaligned_le64(o0, b0);
+	put_unaligned_le64(o1, b1);
+	put_unaligned_le64(o2, b2);
+	put_unaligned_le64(o3, b3);
+}
+
+static __always_inline void format_fcontract(u8 *output, u64 *input)
+{
+	format_fcontract_first_carry_full(input);
+	format_fcontract_second_carry_full(input);
+	format_fcontract_trim(input);
+	format_fcontract_store(output, input);
+}
+
+static __always_inline void format_scalar_of_point(u8 *scalar, u64 *point)
+{
+	u64 *x = point;
+	u64 *z = point + 5;
+	u64 buf[10] __aligned(32) = { 0 };
+	u64 *zmone = buf;
+	u64 *sc = buf + 5;
+	crecip(zmone, z);
+	fmul_fmul(sc, x, zmone);
+	format_fcontract(scalar, sc);
+}
+
+static void curve25519_generic(u8 mypublic[CURVE25519_KEY_SIZE],
+			       const u8 secret[CURVE25519_KEY_SIZE],
+			       const u8 basepoint[CURVE25519_KEY_SIZE])
+{
+	u64 buf0[10] __aligned(32) = { 0 };
+	u64 *x0 = buf0;
+	u64 *z = buf0 + 5;
+	u64 *q;
+	format_fexpand(x0, basepoint);
+	z[0] = 1;
+	q = buf0;
+	{
+		u8 e[32] __aligned(32) = { 0 };
+		u8 *scalar;
+		memcpy(e, secret, 32);
+		curve25519_clamp_secret(e);
+		scalar = e;
+		{
+			u64 buf[15] = { 0 };
+			u64 *nq = buf;
+			u64 *x = nq;
+			x[0] = 1;
+			ladder_cmult(nq, scalar, q);
+			format_scalar_of_point(mypublic, nq);
+			memzero_explicit(buf, sizeof(buf));
+		}
+		memzero_explicit(e, sizeof(e));
+	}
+	memzero_explicit(buf0, sizeof(buf0));
+}
diff --git a/net/wireguard/crypto/zinc/curve25519/curve25519-x86_64-glue.c b/net/wireguard/crypto/zinc/curve25519/curve25519-x86_64-glue.c
new file mode 100644
index 000000000000..e08cc2ba74f3
--- /dev/null
+++ b/net/wireguard/crypto/zinc/curve25519/curve25519-x86_64-glue.c
@@ -0,0 +1,44 @@
+// SPDX-License-Identifier: GPL-2.0 OR MIT
+/*
+ * Copyright (C) 2015-2019 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved.
+ */
+
+#include <asm/cpufeature.h>
+#include <asm/processor.h>
+
+#include "curve25519-x86_64.c"
+
+static bool curve25519_use_bmi2_adx __ro_after_init;
+static bool *const curve25519_nobs[] __initconst = {
+	&curve25519_use_bmi2_adx };
+
+static void __init curve25519_fpu_init(void)
+{
+	curve25519_use_bmi2_adx = IS_ENABLED(CONFIG_AS_BMI2) &&
+				  IS_ENABLED(CONFIG_AS_ADX) &&
+				  boot_cpu_has(X86_FEATURE_BMI2) &&
+				  boot_cpu_has(X86_FEATURE_ADX);
+}
+
+static inline bool curve25519_arch(u8 mypublic[CURVE25519_KEY_SIZE],
+				   const u8 secret[CURVE25519_KEY_SIZE],
+				   const u8 basepoint[CURVE25519_KEY_SIZE])
+{
+	if (IS_ENABLED(CONFIG_AS_ADX) && IS_ENABLED(CONFIG_AS_BMI2) &&
+	    curve25519_use_bmi2_adx) {
+		curve25519_ever64(mypublic, secret, basepoint);
+		return true;
+	}
+	return false;
+}
+
+static inline bool curve25519_base_arch(u8 pub[CURVE25519_KEY_SIZE],
+					const u8 secret[CURVE25519_KEY_SIZE])
+{
+	if (IS_ENABLED(CONFIG_AS_ADX) && IS_ENABLED(CONFIG_AS_BMI2) &&
+	    curve25519_use_bmi2_adx) {
+		curve25519_ever64_base(pub, secret);
+		return true;
+	}
+	return false;
+}
diff --git a/net/wireguard/crypto/zinc/curve25519/curve25519-x86_64.c b/net/wireguard/crypto/zinc/curve25519/curve25519-x86_64.c
new file mode 100644
index 000000000000..74570cbe3124
--- /dev/null
+++ b/net/wireguard/crypto/zinc/curve25519/curve25519-x86_64.c
@@ -0,0 +1,1369 @@
+// SPDX-License-Identifier: GPL-2.0 OR MIT
+/*
+ * Copyright (C) 2020 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved.
+ * Copyright (c) 2016-2020 INRIA, CMU and Microsoft Corporation
+ */
+
+static __always_inline u64 eq_mask(u64 a, u64 b)
+{
+	u64 x = a ^ b;
+	u64 minus_x = ~x + (u64)1U;
+	u64 x_or_minus_x = x | minus_x;
+	u64 xnx = x_or_minus_x >> (u32)63U;
+	return xnx - (u64)1U;
+}
+
+static __always_inline u64 gte_mask(u64 a, u64 b)
+{
+	u64 x = a;
+	u64 y = b;
+	u64 x_xor_y = x ^ y;
+	u64 x_sub_y = x - y;
+	u64 x_sub_y_xor_y = x_sub_y ^ y;
+	u64 q = x_xor_y | x_sub_y_xor_y;
+	u64 x_xor_q = x ^ q;
+	u64 x_xor_q_ = x_xor_q >> (u32)63U;
+	return x_xor_q_ - (u64)1U;
+}
+
+/* Computes the addition of four-element f1 with value in f2
+ * and returns the carry (if any) */
+static inline u64 add_scalar(u64 *out, const u64 *f1, u64 f2)
+{
+	u64 carry_r;
+
+	asm volatile(
+		/* Clear registers to propagate the carry bit */
+		"  xor %%r8, %%r8;"
+		"  xor %%r9, %%r9;"
+		"  xor %%r10, %%r10;"
+		"  xor %%r11, %%r11;"
+		"  xor %1, %1;"
+
+		/* Begin addition chain */
+		"  addq 0(%3), %0;"
+		"  movq %0, 0(%2);"
+		"  adcxq 8(%3), %%r8;"
+		"  movq %%r8, 8(%2);"
+		"  adcxq 16(%3), %%r9;"
+		"  movq %%r9, 16(%2);"
+		"  adcxq 24(%3), %%r10;"
+		"  movq %%r10, 24(%2);"
+
+		/* Return the carry bit in a register */
+		"  adcx %%r11, %1;"
+	: "+&r" (f2), "=&r" (carry_r)
+	: "r" (out), "r" (f1)
+	: "%r8", "%r9", "%r10", "%r11", "memory", "cc"
+	);
+
+	return carry_r;
+}
+
+/* Computes the field addition of two field elements */
+static inline void fadd(u64 *out, const u64 *f1, const u64 *f2)
+{
+	asm volatile(
+		/* Compute the raw addition of f1 + f2 */
+		"  movq 0(%0), %%r8;"
+		"  addq 0(%2), %%r8;"
+		"  movq 8(%0), %%r9;"
+		"  adcxq 8(%2), %%r9;"
+		"  movq 16(%0), %%r10;"
+		"  adcxq 16(%2), %%r10;"
+		"  movq 24(%0), %%r11;"
+		"  adcxq 24(%2), %%r11;"
+
+		/* Wrap the result back into the field */
+
+		/* Step 1: Compute carry*38 */
+		"  mov $0, %%rax;"
+		"  mov $38, %0;"
+		"  cmovc %0, %%rax;"
+
+		/* Step 2: Add carry*38 to the original sum */
+		"  xor %%rcx, %%rcx;"
+		"  add %%rax, %%r8;"
+		"  adcx %%rcx, %%r9;"
+		"  movq %%r9, 8(%1);"
+		"  adcx %%rcx, %%r10;"
+		"  movq %%r10, 16(%1);"
+		"  adcx %%rcx, %%r11;"
+		"  movq %%r11, 24(%1);"
+
+		/* Step 3: Fold the carry bit back in; guaranteed not to carry at this point */
+		"  mov $0, %%rax;"
+		"  cmovc %0, %%rax;"
+		"  add %%rax, %%r8;"
+		"  movq %%r8, 0(%1);"
+	: "+&r" (f2)
+	: "r" (out), "r" (f1)
+	: "%rax", "%rcx", "%r8", "%r9", "%r10", "%r11", "memory", "cc"
+	);
+}
+
+/* Computes the field substraction of two field elements */
+static inline void fsub(u64 *out, const u64 *f1, const u64 *f2)
+{
+	asm volatile(
+		/* Compute the raw substraction of f1-f2 */
+		"  movq 0(%1), %%r8;"
+		"  subq 0(%2), %%r8;"
+		"  movq 8(%1), %%r9;"
+		"  sbbq 8(%2), %%r9;"
+		"  movq 16(%1), %%r10;"
+		"  sbbq 16(%2), %%r10;"
+		"  movq 24(%1), %%r11;"
+		"  sbbq 24(%2), %%r11;"
+
+		/* Wrap the result back into the field */
+
+		/* Step 1: Compute carry*38 */
+		"  mov $0, %%rax;"
+		"  mov $38, %%rcx;"
+		"  cmovc %%rcx, %%rax;"
+
+		/* Step 2: Substract carry*38 from the original difference */
+		"  sub %%rax, %%r8;"
+		"  sbb $0, %%r9;"
+		"  sbb $0, %%r10;"
+		"  sbb $0, %%r11;"
+
+		/* Step 3: Fold the carry bit back in; guaranteed not to carry at this point */
+		"  mov $0, %%rax;"
+		"  cmovc %%rcx, %%rax;"
+		"  sub %%rax, %%r8;"
+
+		/* Store the result */
+		"  movq %%r8, 0(%0);"
+		"  movq %%r9, 8(%0);"
+		"  movq %%r10, 16(%0);"
+		"  movq %%r11, 24(%0);"
+	:
+	: "r" (out), "r" (f1), "r" (f2)
+	: "%rax", "%rcx", "%r8", "%r9", "%r10", "%r11", "memory", "cc"
+	);
+}
+
+/* Computes a field multiplication: out <- f1 * f2
+ * Uses the 8-element buffer tmp for intermediate results */
+static inline void fmul(u64 *out, const u64 *f1, const u64 *f2, u64 *tmp)
+{
+	asm volatile(
+		/* Compute the raw multiplication: tmp <- src1 * src2 */
+
+		/* Compute src1[0] * src2 */
+		"  movq 0(%1), %%rdx;"
+		"  mulxq 0(%3), %%r8, %%r9;"       "  xor %%r10, %%r10;"     "  movq %%r8, 0(%0);"
+		"  mulxq 8(%3), %%r10, %%r11;"     "  adox %%r9, %%r10;"     "  movq %%r10, 8(%0);"
+		"  mulxq 16(%3), %%r12, %%r13;"    "  adox %%r11, %%r12;"
+		"  mulxq 24(%3), %%r14, %%rdx;"    "  adox %%r13, %%r14;"    "  mov $0, %%rax;"
+		                                   "  adox %%rdx, %%rax;"
+		/* Compute src1[1] * src2 */
+		"  movq 8(%1), %%rdx;"
+		"  mulxq 0(%3), %%r8, %%r9;"       "  xor %%r10, %%r10;"     "  adcxq 8(%0), %%r8;"    "  movq %%r8, 8(%0);"
+		"  mulxq 8(%3), %%r10, %%r11;"     "  adox %%r9, %%r10;"     "  adcx %%r12, %%r10;"    "  movq %%r10, 16(%0);"
+		"  mulxq 16(%3), %%r12, %%r13;"    "  adox %%r11, %%r12;"    "  adcx %%r14, %%r12;"    "  mov $0, %%r8;"
+		"  mulxq 24(%3), %%r14, %%rdx;"    "  adox %%r13, %%r14;"    "  adcx %%rax, %%r14;"    "  mov $0, %%rax;"
+		                                   "  adox %%rdx, %%rax;"    "  adcx %%r8, %%rax;"
+		/* Compute src1[2] * src2 */
+		"  movq 16(%1), %%rdx;"
+		"  mulxq 0(%3), %%r8, %%r9;"       "  xor %%r10, %%r10;"    "  adcxq 16(%0), %%r8;"    "  movq %%r8, 16(%0);"
+		"  mulxq 8(%3), %%r10, %%r11;"     "  adox %%r9, %%r10;"     "  adcx %%r12, %%r10;"    "  movq %%r10, 24(%0);"
+		"  mulxq 16(%3), %%r12, %%r13;"    "  adox %%r11, %%r12;"    "  adcx %%r14, %%r12;"    "  mov $0, %%r8;"
+		"  mulxq 24(%3), %%r14, %%rdx;"    "  adox %%r13, %%r14;"    "  adcx %%rax, %%r14;"    "  mov $0, %%rax;"
+		                                   "  adox %%rdx, %%rax;"    "  adcx %%r8, %%rax;"
+		/* Compute src1[3] * src2 */
+		"  movq 24(%1), %%rdx;"
+		"  mulxq 0(%3), %%r8, %%r9;"       "  xor %%r10, %%r10;"    "  adcxq 24(%0), %%r8;"    "  movq %%r8, 24(%0);"
+		"  mulxq 8(%3), %%r10, %%r11;"     "  adox %%r9, %%r10;"     "  adcx %%r12, %%r10;"    "  movq %%r10, 32(%0);"
+		"  mulxq 16(%3), %%r12, %%r13;"    "  adox %%r11, %%r12;"    "  adcx %%r14, %%r12;"    "  movq %%r12, 40(%0);"    "  mov $0, %%r8;"
+		"  mulxq 24(%3), %%r14, %%rdx;"    "  adox %%r13, %%r14;"    "  adcx %%rax, %%r14;"    "  movq %%r14, 48(%0);"    "  mov $0, %%rax;"
+		                                   "  adox %%rdx, %%rax;"    "  adcx %%r8, %%rax;"     "  movq %%rax, 56(%0);"
+		/* Line up pointers */
+		"  mov %0, %1;"
+		"  mov %2, %0;"
+
+		/* Wrap the result back into the field */
+
+		/* Step 1: Compute dst + carry == tmp_hi * 38 + tmp_lo */
+		"  mov $38, %%rdx;"
+		"  mulxq 32(%1), %%r8, %%r13;"
+		"  xor %3, %3;"
+		"  adoxq 0(%1), %%r8;"
+		"  mulxq 40(%1), %%r9, %%r12;"
+		"  adcx %%r13, %%r9;"
+		"  adoxq 8(%1), %%r9;"
+		"  mulxq 48(%1), %%r10, %%r13;"
+		"  adcx %%r12, %%r10;"
+		"  adoxq 16(%1), %%r10;"
+		"  mulxq 56(%1), %%r11, %%rax;"
+		"  adcx %%r13, %%r11;"
+		"  adoxq 24(%1), %%r11;"
+		"  adcx %3, %%rax;"
+		"  adox %3, %%rax;"
+		"  imul %%rdx, %%rax;"
+
+		/* Step 2: Fold the carry back into dst */
+		"  add %%rax, %%r8;"
+		"  adcx %3, %%r9;"
+		"  movq %%r9, 8(%0);"
+		"  adcx %3, %%r10;"
+		"  movq %%r10, 16(%0);"
+		"  adcx %3, %%r11;"
+		"  movq %%r11, 24(%0);"
+
+		/* Step 3: Fold the carry bit back in; guaranteed not to carry at this point */
+		"  mov $0, %%rax;"
+		"  cmovc %%rdx, %%rax;"
+		"  add %%rax, %%r8;"
+		"  movq %%r8, 0(%0);"
+	: "+&r" (tmp), "+&r" (f1), "+&r" (out), "+&r" (f2)
+	:
+	: "%rax", "%rdx", "%r8", "%r9", "%r10", "%r11", "%r12", "%r13", "%r14", "memory", "cc"
+	);
+}
+
+/* Computes two field multiplications:
+ * out[0] <- f1[0] * f2[0]
+ * out[1] <- f1[1] * f2[1]
+ * Uses the 16-element buffer tmp for intermediate results. */
+static inline void fmul2(u64 *out, const u64 *f1, const u64 *f2, u64 *tmp)
+{
+	asm volatile(
+		/* Compute the raw multiplication tmp[0] <- f1[0] * f2[0] */
+
+		/* Compute src1[0] * src2 */
+		"  movq 0(%1), %%rdx;"
+		"  mulxq 0(%3), %%r8, %%r9;"       "  xor %%r10, %%r10;"     "  movq %%r8, 0(%0);"
+		"  mulxq 8(%3), %%r10, %%r11;"     "  adox %%r9, %%r10;"     "  movq %%r10, 8(%0);"
+		"  mulxq 16(%3), %%r12, %%r13;"    "  adox %%r11, %%r12;"
+		"  mulxq 24(%3), %%r14, %%rdx;"    "  adox %%r13, %%r14;"    "  mov $0, %%rax;"
+		                                   "  adox %%rdx, %%rax;"
+		/* Compute src1[1] * src2 */
+		"  movq 8(%1), %%rdx;"
+		"  mulxq 0(%3), %%r8, %%r9;"       "  xor %%r10, %%r10;"     "  adcxq 8(%0), %%r8;"    "  movq %%r8, 8(%0);"
+		"  mulxq 8(%3), %%r10, %%r11;"     "  adox %%r9, %%r10;"     "  adcx %%r12, %%r10;"    "  movq %%r10, 16(%0);"
+		"  mulxq 16(%3), %%r12, %%r13;"    "  adox %%r11, %%r12;"    "  adcx %%r14, %%r12;"    "  mov $0, %%r8;"
+		"  mulxq 24(%3), %%r14, %%rdx;"    "  adox %%r13, %%r14;"    "  adcx %%rax, %%r14;"    "  mov $0, %%rax;"
+		                                   "  adox %%rdx, %%rax;"    "  adcx %%r8, %%rax;"
+		/* Compute src1[2] * src2 */
+		"  movq 16(%1), %%rdx;"
+		"  mulxq 0(%3), %%r8, %%r9;"       "  xor %%r10, %%r10;"    "  adcxq 16(%0), %%r8;"    "  movq %%r8, 16(%0);"
+		"  mulxq 8(%3), %%r10, %%r11;"     "  adox %%r9, %%r10;"     "  adcx %%r12, %%r10;"    "  movq %%r10, 24(%0);"
+		"  mulxq 16(%3), %%r12, %%r13;"    "  adox %%r11, %%r12;"    "  adcx %%r14, %%r12;"    "  mov $0, %%r8;"
+		"  mulxq 24(%3), %%r14, %%rdx;"    "  adox %%r13, %%r14;"    "  adcx %%rax, %%r14;"    "  mov $0, %%rax;"
+		                                   "  adox %%rdx, %%rax;"    "  adcx %%r8, %%rax;"
+		/* Compute src1[3] * src2 */
+		"  movq 24(%1), %%rdx;"
+		"  mulxq 0(%3), %%r8, %%r9;"       "  xor %%r10, %%r10;"    "  adcxq 24(%0), %%r8;"    "  movq %%r8, 24(%0);"
+		"  mulxq 8(%3), %%r10, %%r11;"     "  adox %%r9, %%r10;"     "  adcx %%r12, %%r10;"    "  movq %%r10, 32(%0);"
+		"  mulxq 16(%3), %%r12, %%r13;"    "  adox %%r11, %%r12;"    "  adcx %%r14, %%r12;"    "  movq %%r12, 40(%0);"    "  mov $0, %%r8;"
+		"  mulxq 24(%3), %%r14, %%rdx;"    "  adox %%r13, %%r14;"    "  adcx %%rax, %%r14;"    "  movq %%r14, 48(%0);"    "  mov $0, %%rax;"
+		                                   "  adox %%rdx, %%rax;"    "  adcx %%r8, %%rax;"     "  movq %%rax, 56(%0);"
+
+		/* Compute the raw multiplication tmp[1] <- f1[1] * f2[1] */
+
+		/* Compute src1[0] * src2 */
+		"  movq 32(%1), %%rdx;"
+		"  mulxq 32(%3), %%r8, %%r9;"       "  xor %%r10, %%r10;"     "  movq %%r8, 64(%0);"
+		"  mulxq 40(%3), %%r10, %%r11;"     "  adox %%r9, %%r10;"     "  movq %%r10, 72(%0);"
+		"  mulxq 48(%3), %%r12, %%r13;"    "  adox %%r11, %%r12;"
+		"  mulxq 56(%3), %%r14, %%rdx;"    "  adox %%r13, %%r14;"    "  mov $0, %%rax;"
+		                                   "  adox %%rdx, %%rax;"
+		/* Compute src1[1] * src2 */
+		"  movq 40(%1), %%rdx;"
+		"  mulxq 32(%3), %%r8, %%r9;"       "  xor %%r10, %%r10;"     "  adcxq 72(%0), %%r8;"    "  movq %%r8, 72(%0);"
+		"  mulxq 40(%3), %%r10, %%r11;"     "  adox %%r9, %%r10;"     "  adcx %%r12, %%r10;"    "  movq %%r10, 80(%0);"
+		"  mulxq 48(%3), %%r12, %%r13;"    "  adox %%r11, %%r12;"    "  adcx %%r14, %%r12;"    "  mov $0, %%r8;"
+		"  mulxq 56(%3), %%r14, %%rdx;"    "  adox %%r13, %%r14;"    "  adcx %%rax, %%r14;"    "  mov $0, %%rax;"
+		                                   "  adox %%rdx, %%rax;"    "  adcx %%r8, %%rax;"
+		/* Compute src1[2] * src2 */
+		"  movq 48(%1), %%rdx;"
+		"  mulxq 32(%3), %%r8, %%r9;"       "  xor %%r10, %%r10;"    "  adcxq 80(%0), %%r8;"    "  movq %%r8, 80(%0);"
+		"  mulxq 40(%3), %%r10, %%r11;"     "  adox %%r9, %%r10;"     "  adcx %%r12, %%r10;"    "  movq %%r10, 88(%0);"
+		"  mulxq 48(%3), %%r12, %%r13;"    "  adox %%r11, %%r12;"    "  adcx %%r14, %%r12;"    "  mov $0, %%r8;"
+		"  mulxq 56(%3), %%r14, %%rdx;"    "  adox %%r13, %%r14;"    "  adcx %%rax, %%r14;"    "  mov $0, %%rax;"
+		                                   "  adox %%rdx, %%rax;"    "  adcx %%r8, %%rax;"
+		/* Compute src1[3] * src2 */
+		"  movq 56(%1), %%rdx;"
+		"  mulxq 32(%3), %%r8, %%r9;"       "  xor %%r10, %%r10;"    "  adcxq 88(%0), %%r8;"    "  movq %%r8, 88(%0);"
+		"  mulxq 40(%3), %%r10, %%r11;"     "  adox %%r9, %%r10;"     "  adcx %%r12, %%r10;"    "  movq %%r10, 96(%0);"
+		"  mulxq 48(%3), %%r12, %%r13;"    "  adox %%r11, %%r12;"    "  adcx %%r14, %%r12;"    "  movq %%r12, 104(%0);"    "  mov $0, %%r8;"
+		"  mulxq 56(%3), %%r14, %%rdx;"    "  adox %%r13, %%r14;"    "  adcx %%rax, %%r14;"    "  movq %%r14, 112(%0);"    "  mov $0, %%rax;"
+		                                   "  adox %%rdx, %%rax;"    "  adcx %%r8, %%rax;"     "  movq %%rax, 120(%0);"
+		/* Line up pointers */
+		"  mov %0, %1;"
+		"  mov %2, %0;"
+
+		/* Wrap the results back into the field */
+
+		/* Step 1: Compute dst + carry == tmp_hi * 38 + tmp_lo */
+		"  mov $38, %%rdx;"
+		"  mulxq 32(%1), %%r8, %%r13;"
+		"  xor %3, %3;"
+		"  adoxq 0(%1), %%r8;"
+		"  mulxq 40(%1), %%r9, %%r12;"
+		"  adcx %%r13, %%r9;"
+		"  adoxq 8(%1), %%r9;"
+		"  mulxq 48(%1), %%r10, %%r13;"
+		"  adcx %%r12, %%r10;"
+		"  adoxq 16(%1), %%r10;"
+		"  mulxq 56(%1), %%r11, %%rax;"
+		"  adcx %%r13, %%r11;"
+		"  adoxq 24(%1), %%r11;"
+		"  adcx %3, %%rax;"
+		"  adox %3, %%rax;"
+		"  imul %%rdx, %%rax;"
+
+		/* Step 2: Fold the carry back into dst */
+		"  add %%rax, %%r8;"
+		"  adcx %3, %%r9;"
+		"  movq %%r9, 8(%0);"
+		"  adcx %3, %%r10;"
+		"  movq %%r10, 16(%0);"
+		"  adcx %3, %%r11;"
+		"  movq %%r11, 24(%0);"
+
+		/* Step 3: Fold the carry bit back in; guaranteed not to carry at this point */
+		"  mov $0, %%rax;"
+		"  cmovc %%rdx, %%rax;"
+		"  add %%rax, %%r8;"
+		"  movq %%r8, 0(%0);"
+
+		/* Step 1: Compute dst + carry == tmp_hi * 38 + tmp_lo */
+		"  mov $38, %%rdx;"
+		"  mulxq 96(%1), %%r8, %%r13;"
+		"  xor %3, %3;"
+		"  adoxq 64(%1), %%r8;"
+		"  mulxq 104(%1), %%r9, %%r12;"
+		"  adcx %%r13, %%r9;"
+		"  adoxq 72(%1), %%r9;"
+		"  mulxq 112(%1), %%r10, %%r13;"
+		"  adcx %%r12, %%r10;"
+		"  adoxq 80(%1), %%r10;"
+		"  mulxq 120(%1), %%r11, %%rax;"
+		"  adcx %%r13, %%r11;"
+		"  adoxq 88(%1), %%r11;"
+		"  adcx %3, %%rax;"
+		"  adox %3, %%rax;"
+		"  imul %%rdx, %%rax;"
+
+		/* Step 2: Fold the carry back into dst */
+		"  add %%rax, %%r8;"
+		"  adcx %3, %%r9;"
+		"  movq %%r9, 40(%0);"
+		"  adcx %3, %%r10;"
+		"  movq %%r10, 48(%0);"
+		"  adcx %3, %%r11;"
+		"  movq %%r11, 56(%0);"
+
+		/* Step 3: Fold the carry bit back in; guaranteed not to carry at this point */
+		"  mov $0, %%rax;"
+		"  cmovc %%rdx, %%rax;"
+		"  add %%rax, %%r8;"
+		"  movq %%r8, 32(%0);"
+	: "+&r" (tmp), "+&r" (f1), "+&r" (out), "+&r" (f2)
+	:
+	: "%rax", "%rdx", "%r8", "%r9", "%r10", "%r11", "%r12", "%r13", "%r14", "memory", "cc"
+	);
+}
+
+/* Computes the field multiplication of four-element f1 with value in f2 */
+static inline void fmul_scalar(u64 *out, const u64 *f1, u64 f2)
+{
+	register u64 f2_r asm("rdx") = f2;
+
+	asm volatile(
+		/* Compute the raw multiplication of f1*f2 */
+		"  mulxq 0(%2), %%r8, %%rcx;"      /* f1[0]*f2 */
+		"  mulxq 8(%2), %%r9, %%r12;"      /* f1[1]*f2 */
+		"  add %%rcx, %%r9;"
+		"  mov $0, %%rcx;"
+		"  mulxq 16(%2), %%r10, %%r13;"    /* f1[2]*f2 */
+		"  adcx %%r12, %%r10;"
+		"  mulxq 24(%2), %%r11, %%rax;"    /* f1[3]*f2 */
+		"  adcx %%r13, %%r11;"
+		"  adcx %%rcx, %%rax;"
+
+		/* Wrap the result back into the field */
+
+		/* Step 1: Compute carry*38 */
+		"  mov $38, %%rdx;"
+		"  imul %%rdx, %%rax;"
+
+		/* Step 2: Fold the carry back into dst */
+		"  add %%rax, %%r8;"
+		"  adcx %%rcx, %%r9;"
+		"  movq %%r9, 8(%1);"
+		"  adcx %%rcx, %%r10;"
+		"  movq %%r10, 16(%1);"
+		"  adcx %%rcx, %%r11;"
+		"  movq %%r11, 24(%1);"
+
+		/* Step 3: Fold the carry bit back in; guaranteed not to carry at this point */
+		"  mov $0, %%rax;"
+		"  cmovc %%rdx, %%rax;"
+		"  add %%rax, %%r8;"
+		"  movq %%r8, 0(%1);"
+	: "+&r" (f2_r)
+	: "r" (out), "r" (f1)
+	: "%rax", "%rcx", "%r8", "%r9", "%r10", "%r11", "%r12", "%r13", "memory", "cc"
+	);
+}
+
+/* Computes p1 <- bit ? p2 : p1 in constant time */
+static inline void cswap2(u64 bit, const u64 *p1, const u64 *p2)
+{
+	asm volatile(
+		/* Invert the polarity of bit to match cmov expectations */
+		"  add $18446744073709551615, %0;"
+
+		/* cswap p1[0], p2[0] */
+		"  movq 0(%1), %%r8;"
+		"  movq 0(%2), %%r9;"
+		"  mov %%r8, %%r10;"
+		"  cmovc %%r9, %%r8;"
+		"  cmovc %%r10, %%r9;"
+		"  movq %%r8, 0(%1);"
+		"  movq %%r9, 0(%2);"
+
+		/* cswap p1[1], p2[1] */
+		"  movq 8(%1), %%r8;"
+		"  movq 8(%2), %%r9;"
+		"  mov %%r8, %%r10;"
+		"  cmovc %%r9, %%r8;"
+		"  cmovc %%r10, %%r9;"
+		"  movq %%r8, 8(%1);"
+		"  movq %%r9, 8(%2);"
+
+		/* cswap p1[2], p2[2] */
+		"  movq 16(%1), %%r8;"
+		"  movq 16(%2), %%r9;"
+		"  mov %%r8, %%r10;"
+		"  cmovc %%r9, %%r8;"
+		"  cmovc %%r10, %%r9;"
+		"  movq %%r8, 16(%1);"
+		"  movq %%r9, 16(%2);"
+
+		/* cswap p1[3], p2[3] */
+		"  movq 24(%1), %%r8;"
+		"  movq 24(%2), %%r9;"
+		"  mov %%r8, %%r10;"
+		"  cmovc %%r9, %%r8;"
+		"  cmovc %%r10, %%r9;"
+		"  movq %%r8, 24(%1);"
+		"  movq %%r9, 24(%2);"
+
+		/* cswap p1[4], p2[4] */
+		"  movq 32(%1), %%r8;"
+		"  movq 32(%2), %%r9;"
+		"  mov %%r8, %%r10;"
+		"  cmovc %%r9, %%r8;"
+		"  cmovc %%r10, %%r9;"
+		"  movq %%r8, 32(%1);"
+		"  movq %%r9, 32(%2);"
+
+		/* cswap p1[5], p2[5] */
+		"  movq 40(%1), %%r8;"
+		"  movq 40(%2), %%r9;"
+		"  mov %%r8, %%r10;"
+		"  cmovc %%r9, %%r8;"
+		"  cmovc %%r10, %%r9;"
+		"  movq %%r8, 40(%1);"
+		"  movq %%r9, 40(%2);"
+
+		/* cswap p1[6], p2[6] */
+		"  movq 48(%1), %%r8;"
+		"  movq 48(%2), %%r9;"
+		"  mov %%r8, %%r10;"
+		"  cmovc %%r9, %%r8;"
+		"  cmovc %%r10, %%r9;"
+		"  movq %%r8, 48(%1);"
+		"  movq %%r9, 48(%2);"
+
+		/* cswap p1[7], p2[7] */
+		"  movq 56(%1), %%r8;"
+		"  movq 56(%2), %%r9;"
+		"  mov %%r8, %%r10;"
+		"  cmovc %%r9, %%r8;"
+		"  cmovc %%r10, %%r9;"
+		"  movq %%r8, 56(%1);"
+		"  movq %%r9, 56(%2);"
+	: "+&r" (bit)
+	: "r" (p1), "r" (p2)
+	: "%r8", "%r9", "%r10", "memory", "cc"
+	);
+}
+
+/* Computes the square of a field element: out <- f * f
+ * Uses the 8-element buffer tmp for intermediate results */
+static inline void fsqr(u64 *out, const u64 *f, u64 *tmp)
+{
+	asm volatile(
+		/* Compute the raw multiplication: tmp <- f * f */
+
+		/* Step 1: Compute all partial products */
+		"  movq 0(%1), %%rdx;"                                       /* f[0] */
+		"  mulxq 8(%1), %%r8, %%r14;"      "  xor %%r15, %%r15;"     /* f[1]*f[0] */
+		"  mulxq 16(%1), %%r9, %%r10;"     "  adcx %%r14, %%r9;"     /* f[2]*f[0] */
+		"  mulxq 24(%1), %%rax, %%rcx;"    "  adcx %%rax, %%r10;"    /* f[3]*f[0] */
+		"  movq 24(%1), %%rdx;"                                      /* f[3] */
+		"  mulxq 8(%1), %%r11, %%r12;"     "  adcx %%rcx, %%r11;"    /* f[1]*f[3] */
+		"  mulxq 16(%1), %%rax, %%r13;"    "  adcx %%rax, %%r12;"    /* f[2]*f[3] */
+		"  movq 8(%1), %%rdx;"             "  adcx %%r15, %%r13;"    /* f1 */
+		"  mulxq 16(%1), %%rax, %%rcx;"    "  mov $0, %%r14;"        /* f[2]*f[1] */
+
+		/* Step 2: Compute two parallel carry chains */
+		"  xor %%r15, %%r15;"
+		"  adox %%rax, %%r10;"
+		"  adcx %%r8, %%r8;"
+		"  adox %%rcx, %%r11;"
+		"  adcx %%r9, %%r9;"
+		"  adox %%r15, %%r12;"
+		"  adcx %%r10, %%r10;"
+		"  adox %%r15, %%r13;"
+		"  adcx %%r11, %%r11;"
+		"  adox %%r15, %%r14;"
+		"  adcx %%r12, %%r12;"
+		"  adcx %%r13, %%r13;"
+		"  adcx %%r14, %%r14;"
+
+		/* Step 3: Compute intermediate squares */
+		"  movq 0(%1), %%rdx;"     "  mulx %%rdx, %%rax, %%rcx;"    /* f[0]^2 */
+		                           "  movq %%rax, 0(%0);"
+		"  add %%rcx, %%r8;"       "  movq %%r8, 8(%0);"
+		"  movq 8(%1), %%rdx;"     "  mulx %%rdx, %%rax, %%rcx;"    /* f[1]^2 */
+		"  adcx %%rax, %%r9;"      "  movq %%r9, 16(%0);"
+		"  adcx %%rcx, %%r10;"     "  movq %%r10, 24(%0);"
+		"  movq 16(%1), %%rdx;"    "  mulx %%rdx, %%rax, %%rcx;"    /* f[2]^2 */
+		"  adcx %%rax, %%r11;"     "  movq %%r11, 32(%0);"
+		"  adcx %%rcx, %%r12;"     "  movq %%r12, 40(%0);"
+		"  movq 24(%1), %%rdx;"    "  mulx %%rdx, %%rax, %%rcx;"    /* f[3]^2 */
+		"  adcx %%rax, %%r13;"     "  movq %%r13, 48(%0);"
+		"  adcx %%rcx, %%r14;"     "  movq %%r14, 56(%0);"
+
+		/* Line up pointers */
+		"  mov %0, %1;"
+		"  mov %2, %0;"
+
+		/* Wrap the result back into the field */
+
+		/* Step 1: Compute dst + carry == tmp_hi * 38 + tmp_lo */
+		"  mov $38, %%rdx;"
+		"  mulxq 32(%1), %%r8, %%r13;"
+		"  xor %%rcx, %%rcx;"
+		"  adoxq 0(%1), %%r8;"
+		"  mulxq 40(%1), %%r9, %%r12;"
+		"  adcx %%r13, %%r9;"
+		"  adoxq 8(%1), %%r9;"
+		"  mulxq 48(%1), %%r10, %%r13;"
+		"  adcx %%r12, %%r10;"
+		"  adoxq 16(%1), %%r10;"
+		"  mulxq 56(%1), %%r11, %%rax;"
+		"  adcx %%r13, %%r11;"
+		"  adoxq 24(%1), %%r11;"
+		"  adcx %%rcx, %%rax;"
+		"  adox %%rcx, %%rax;"
+		"  imul %%rdx, %%rax;"
+
+		/* Step 2: Fold the carry back into dst */
+		"  add %%rax, %%r8;"
+		"  adcx %%rcx, %%r9;"
+		"  movq %%r9, 8(%0);"
+		"  adcx %%rcx, %%r10;"
+		"  movq %%r10, 16(%0);"
+		"  adcx %%rcx, %%r11;"
+		"  movq %%r11, 24(%0);"
+
+		/* Step 3: Fold the carry bit back in; guaranteed not to carry at this point */
+		"  mov $0, %%rax;"
+		"  cmovc %%rdx, %%rax;"
+		"  add %%rax, %%r8;"
+		"  movq %%r8, 0(%0);"
+	: "+&r" (tmp), "+&r" (f), "+&r" (out)
+	:
+	: "%rax", "%rcx", "%rdx", "%r8", "%r9", "%r10", "%r11", "%r12", "%r13", "%r14", "%r15", "memory", "cc"
+	);
+}
+
+/* Computes two field squarings:
+ * out[0] <- f[0] * f[0]
+ * out[1] <- f[1] * f[1]
+ * Uses the 16-element buffer tmp for intermediate results */
+static inline void fsqr2(u64 *out, const u64 *f, u64 *tmp)
+{
+	asm volatile(
+		/* Step 1: Compute all partial products */
+		"  movq 0(%1), %%rdx;"                                       /* f[0] */
+		"  mulxq 8(%1), %%r8, %%r14;"      "  xor %%r15, %%r15;"     /* f[1]*f[0] */
+		"  mulxq 16(%1), %%r9, %%r10;"     "  adcx %%r14, %%r9;"     /* f[2]*f[0] */
+		"  mulxq 24(%1), %%rax, %%rcx;"    "  adcx %%rax, %%r10;"    /* f[3]*f[0] */
+		"  movq 24(%1), %%rdx;"                                      /* f[3] */
+		"  mulxq 8(%1), %%r11, %%r12;"     "  adcx %%rcx, %%r11;"    /* f[1]*f[3] */
+		"  mulxq 16(%1), %%rax, %%r13;"    "  adcx %%rax, %%r12;"    /* f[2]*f[3] */
+		"  movq 8(%1), %%rdx;"             "  adcx %%r15, %%r13;"    /* f1 */
+		"  mulxq 16(%1), %%rax, %%rcx;"    "  mov $0, %%r14;"        /* f[2]*f[1] */
+
+		/* Step 2: Compute two parallel carry chains */
+		"  xor %%r15, %%r15;"
+		"  adox %%rax, %%r10;"
+		"  adcx %%r8, %%r8;"
+		"  adox %%rcx, %%r11;"
+		"  adcx %%r9, %%r9;"
+		"  adox %%r15, %%r12;"
+		"  adcx %%r10, %%r10;"
+		"  adox %%r15, %%r13;"
+		"  adcx %%r11, %%r11;"
+		"  adox %%r15, %%r14;"
+		"  adcx %%r12, %%r12;"
+		"  adcx %%r13, %%r13;"
+		"  adcx %%r14, %%r14;"
+
+		/* Step 3: Compute intermediate squares */
+		"  movq 0(%1), %%rdx;"     "  mulx %%rdx, %%rax, %%rcx;"    /* f[0]^2 */
+		                           "  movq %%rax, 0(%0);"
+		"  add %%rcx, %%r8;"       "  movq %%r8, 8(%0);"
+		"  movq 8(%1), %%rdx;"     "  mulx %%rdx, %%rax, %%rcx;"    /* f[1]^2 */
+		"  adcx %%rax, %%r9;"      "  movq %%r9, 16(%0);"
+		"  adcx %%rcx, %%r10;"     "  movq %%r10, 24(%0);"
+		"  movq 16(%1), %%rdx;"    "  mulx %%rdx, %%rax, %%rcx;"    /* f[2]^2 */
+		"  adcx %%rax, %%r11;"     "  movq %%r11, 32(%0);"
+		"  adcx %%rcx, %%r12;"     "  movq %%r12, 40(%0);"
+		"  movq 24(%1), %%rdx;"    "  mulx %%rdx, %%rax, %%rcx;"    /* f[3]^2 */
+		"  adcx %%rax, %%r13;"     "  movq %%r13, 48(%0);"
+		"  adcx %%rcx, %%r14;"     "  movq %%r14, 56(%0);"
+
+		/* Step 1: Compute all partial products */
+		"  movq 32(%1), %%rdx;"                                       /* f[0] */
+		"  mulxq 40(%1), %%r8, %%r14;"      "  xor %%r15, %%r15;"     /* f[1]*f[0] */
+		"  mulxq 48(%1), %%r9, %%r10;"     "  adcx %%r14, %%r9;"     /* f[2]*f[0] */
+		"  mulxq 56(%1), %%rax, %%rcx;"    "  adcx %%rax, %%r10;"    /* f[3]*f[0] */
+		"  movq 56(%1), %%rdx;"                                      /* f[3] */
+		"  mulxq 40(%1), %%r11, %%r12;"     "  adcx %%rcx, %%r11;"    /* f[1]*f[3] */
+		"  mulxq 48(%1), %%rax, %%r13;"    "  adcx %%rax, %%r12;"    /* f[2]*f[3] */
+		"  movq 40(%1), %%rdx;"             "  adcx %%r15, %%r13;"    /* f1 */
+		"  mulxq 48(%1), %%rax, %%rcx;"    "  mov $0, %%r14;"        /* f[2]*f[1] */
+
+		/* Step 2: Compute two parallel carry chains */
+		"  xor %%r15, %%r15;"
+		"  adox %%rax, %%r10;"
+		"  adcx %%r8, %%r8;"
+		"  adox %%rcx, %%r11;"
+		"  adcx %%r9, %%r9;"
+		"  adox %%r15, %%r12;"
+		"  adcx %%r10, %%r10;"
+		"  adox %%r15, %%r13;"
+		"  adcx %%r11, %%r11;"
+		"  adox %%r15, %%r14;"
+		"  adcx %%r12, %%r12;"
+		"  adcx %%r13, %%r13;"
+		"  adcx %%r14, %%r14;"
+
+		/* Step 3: Compute intermediate squares */
+		"  movq 32(%1), %%rdx;"     "  mulx %%rdx, %%rax, %%rcx;"    /* f[0]^2 */
+		                           "  movq %%rax, 64(%0);"
+		"  add %%rcx, %%r8;"       "  movq %%r8, 72(%0);"
+		"  movq 40(%1), %%rdx;"     "  mulx %%rdx, %%rax, %%rcx;"    /* f[1]^2 */
+		"  adcx %%rax, %%r9;"      "  movq %%r9, 80(%0);"
+		"  adcx %%rcx, %%r10;"     "  movq %%r10, 88(%0);"
+		"  movq 48(%1), %%rdx;"    "  mulx %%rdx, %%rax, %%rcx;"    /* f[2]^2 */
+		"  adcx %%rax, %%r11;"     "  movq %%r11, 96(%0);"
+		"  adcx %%rcx, %%r12;"     "  movq %%r12, 104(%0);"
+		"  movq 56(%1), %%rdx;"    "  mulx %%rdx, %%rax, %%rcx;"    /* f[3]^2 */
+		"  adcx %%rax, %%r13;"     "  movq %%r13, 112(%0);"
+		"  adcx %%rcx, %%r14;"     "  movq %%r14, 120(%0);"
+
+		/* Line up pointers */
+		"  mov %0, %1;"
+		"  mov %2, %0;"
+
+		/* Step 1: Compute dst + carry == tmp_hi * 38 + tmp_lo */
+		"  mov $38, %%rdx;"
+		"  mulxq 32(%1), %%r8, %%r13;"
+		"  xor %%rcx, %%rcx;"
+		"  adoxq 0(%1), %%r8;"
+		"  mulxq 40(%1), %%r9, %%r12;"
+		"  adcx %%r13, %%r9;"
+		"  adoxq 8(%1), %%r9;"
+		"  mulxq 48(%1), %%r10, %%r13;"
+		"  adcx %%r12, %%r10;"
+		"  adoxq 16(%1), %%r10;"
+		"  mulxq 56(%1), %%r11, %%rax;"
+		"  adcx %%r13, %%r11;"
+		"  adoxq 24(%1), %%r11;"
+		"  adcx %%rcx, %%rax;"
+		"  adox %%rcx, %%rax;"
+		"  imul %%rdx, %%rax;"
+
+		/* Step 2: Fold the carry back into dst */
+		"  add %%rax, %%r8;"
+		"  adcx %%rcx, %%r9;"
+		"  movq %%r9, 8(%0);"
+		"  adcx %%rcx, %%r10;"
+		"  movq %%r10, 16(%0);"
+		"  adcx %%rcx, %%r11;"
+		"  movq %%r11, 24(%0);"
+
+		/* Step 3: Fold the carry bit back in; guaranteed not to carry at this point */
+		"  mov $0, %%rax;"
+		"  cmovc %%rdx, %%rax;"
+		"  add %%rax, %%r8;"
+		"  movq %%r8, 0(%0);"
+
+		/* Step 1: Compute dst + carry == tmp_hi * 38 + tmp_lo */
+		"  mov $38, %%rdx;"
+		"  mulxq 96(%1), %%r8, %%r13;"
+		"  xor %%rcx, %%rcx;"
+		"  adoxq 64(%1), %%r8;"
+		"  mulxq 104(%1), %%r9, %%r12;"
+		"  adcx %%r13, %%r9;"
+		"  adoxq 72(%1), %%r9;"
+		"  mulxq 112(%1), %%r10, %%r13;"
+		"  adcx %%r12, %%r10;"
+		"  adoxq 80(%1), %%r10;"
+		"  mulxq 120(%1), %%r11, %%rax;"
+		"  adcx %%r13, %%r11;"
+		"  adoxq 88(%1), %%r11;"
+		"  adcx %%rcx, %%rax;"
+		"  adox %%rcx, %%rax;"
+		"  imul %%rdx, %%rax;"
+
+		/* Step 2: Fold the carry back into dst */
+		"  add %%rax, %%r8;"
+		"  adcx %%rcx, %%r9;"
+		"  movq %%r9, 40(%0);"
+		"  adcx %%rcx, %%r10;"
+		"  movq %%r10, 48(%0);"
+		"  adcx %%rcx, %%r11;"
+		"  movq %%r11, 56(%0);"
+
+		/* Step 3: Fold the carry bit back in; guaranteed not to carry at this point */
+		"  mov $0, %%rax;"
+		"  cmovc %%rdx, %%rax;"
+		"  add %%rax, %%r8;"
+		"  movq %%r8, 32(%0);"
+	: "+&r" (tmp), "+&r" (f), "+&r" (out)
+	:
+	: "%rax", "%rcx", "%rdx", "%r8", "%r9", "%r10", "%r11", "%r12", "%r13", "%r14", "%r15", "memory", "cc"
+	);
+}
+
+static void point_add_and_double(u64 *q, u64 *p01_tmp1, u64 *tmp2)
+{
+	u64 *nq = p01_tmp1;
+	u64 *nq_p1 = p01_tmp1 + (u32)8U;
+	u64 *tmp1 = p01_tmp1 + (u32)16U;
+	u64 *x1 = q;
+	u64 *x2 = nq;
+	u64 *z2 = nq + (u32)4U;
+	u64 *z3 = nq_p1 + (u32)4U;
+	u64 *a = tmp1;
+	u64 *b = tmp1 + (u32)4U;
+	u64 *ab = tmp1;
+	u64 *dc = tmp1 + (u32)8U;
+	u64 *x3;
+	u64 *z31;
+	u64 *d0;
+	u64 *c0;
+	u64 *a1;
+	u64 *b1;
+	u64 *d;
+	u64 *c;
+	u64 *ab1;
+	u64 *dc1;
+	fadd(a, x2, z2);
+	fsub(b, x2, z2);
+	x3 = nq_p1;
+	z31 = nq_p1 + (u32)4U;
+	d0 = dc;
+	c0 = dc + (u32)4U;
+	fadd(c0, x3, z31);
+	fsub(d0, x3, z31);
+	fmul2(dc, dc, ab, tmp2);
+	fadd(x3, d0, c0);
+	fsub(z31, d0, c0);
+	a1 = tmp1;
+	b1 = tmp1 + (u32)4U;
+	d = tmp1 + (u32)8U;
+	c = tmp1 + (u32)12U;
+	ab1 = tmp1;
+	dc1 = tmp1 + (u32)8U;
+	fsqr2(dc1, ab1, tmp2);
+	fsqr2(nq_p1, nq_p1, tmp2);
+	a1[0U] = c[0U];
+	a1[1U] = c[1U];
+	a1[2U] = c[2U];
+	a1[3U] = c[3U];
+	fsub(c, d, c);
+	fmul_scalar(b1, c, (u64)121665U);
+	fadd(b1, b1, d);
+	fmul2(nq, dc1, ab1, tmp2);
+	fmul(z3, z3, x1, tmp2);
+}
+
+static void point_double(u64 *nq, u64 *tmp1, u64 *tmp2)
+{
+	u64 *x2 = nq;
+	u64 *z2 = nq + (u32)4U;
+	u64 *a = tmp1;
+	u64 *b = tmp1 + (u32)4U;
+	u64 *d = tmp1 + (u32)8U;
+	u64 *c = tmp1 + (u32)12U;
+	u64 *ab = tmp1;
+	u64 *dc = tmp1 + (u32)8U;
+	fadd(a, x2, z2);
+	fsub(b, x2, z2);
+	fsqr2(dc, ab, tmp2);
+	a[0U] = c[0U];
+	a[1U] = c[1U];
+	a[2U] = c[2U];
+	a[3U] = c[3U];
+	fsub(c, d, c);
+	fmul_scalar(b, c, (u64)121665U);
+	fadd(b, b, d);
+	fmul2(nq, dc, ab, tmp2);
+}
+
+static void montgomery_ladder(u64 *out, const u8 *key, u64 *init1)
+{
+	u64 tmp2[16U] = { 0U };
+	u64 p01_tmp1_swap[33U] = { 0U };
+	u64 *p0 = p01_tmp1_swap;
+	u64 *p01 = p01_tmp1_swap;
+	u64 *p03 = p01;
+	u64 *p11 = p01 + (u32)8U;
+	u64 *x0;
+	u64 *z0;
+	u64 *p01_tmp1;
+	u64 *p01_tmp11;
+	u64 *nq10;
+	u64 *nq_p11;
+	u64 *swap1;
+	u64 sw0;
+	u64 *nq1;
+	u64 *tmp1;
+	memcpy(p11, init1, (u32)8U * sizeof(init1[0U]));
+	x0 = p03;
+	z0 = p03 + (u32)4U;
+	x0[0U] = (u64)1U;
+	x0[1U] = (u64)0U;
+	x0[2U] = (u64)0U;
+	x0[3U] = (u64)0U;
+	z0[0U] = (u64)0U;
+	z0[1U] = (u64)0U;
+	z0[2U] = (u64)0U;
+	z0[3U] = (u64)0U;
+	p01_tmp1 = p01_tmp1_swap;
+	p01_tmp11 = p01_tmp1_swap;
+	nq10 = p01_tmp1_swap;
+	nq_p11 = p01_tmp1_swap + (u32)8U;
+	swap1 = p01_tmp1_swap + (u32)32U;
+	cswap2((u64)1U, nq10, nq_p11);
+	point_add_and_double(init1, p01_tmp11, tmp2);
+	swap1[0U] = (u64)1U;
+	{
+		u32 i;
+		for (i = (u32)0U; i < (u32)251U; i = i + (u32)1U) {
+			u64 *p01_tmp12 = p01_tmp1_swap;
+			u64 *swap2 = p01_tmp1_swap + (u32)32U;
+			u64 *nq2 = p01_tmp12;
+			u64 *nq_p12 = p01_tmp12 + (u32)8U;
+			u64 bit = (u64)(key[((u32)253U - i) / (u32)8U] >> ((u32)253U - i) % (u32)8U & (u8)1U);
+			u64 sw = swap2[0U] ^ bit;
+			cswap2(sw, nq2, nq_p12);
+			point_add_and_double(init1, p01_tmp12, tmp2);
+			swap2[0U] = bit;
+		}
+	}
+	sw0 = swap1[0U];
+	cswap2(sw0, nq10, nq_p11);
+	nq1 = p01_tmp1;
+	tmp1 = p01_tmp1 + (u32)16U;
+	point_double(nq1, tmp1, tmp2);
+	point_double(nq1, tmp1, tmp2);
+	point_double(nq1, tmp1, tmp2);
+	memcpy(out, p0, (u32)8U * sizeof(p0[0U]));
+
+	memzero_explicit(tmp2, sizeof(tmp2));
+	memzero_explicit(p01_tmp1_swap, sizeof(p01_tmp1_swap));
+}
+
+static void fsquare_times(u64 *o, const u64 *inp, u64 *tmp, u32 n1)
+{
+	u32 i;
+	fsqr(o, inp, tmp);
+	for (i = (u32)0U; i < n1 - (u32)1U; i = i + (u32)1U)
+		fsqr(o, o, tmp);
+}
+
+static void finv(u64 *o, const u64 *i, u64 *tmp)
+{
+	u64 t1[16U] = { 0U };
+	u64 *a0 = t1;
+	u64 *b = t1 + (u32)4U;
+	u64 *c = t1 + (u32)8U;
+	u64 *t00 = t1 + (u32)12U;
+	u64 *tmp1 = tmp;
+	u64 *a;
+	u64 *t0;
+	fsquare_times(a0, i, tmp1, (u32)1U);
+	fsquare_times(t00, a0, tmp1, (u32)2U);
+	fmul(b, t00, i, tmp);
+	fmul(a0, b, a0, tmp);
+	fsquare_times(t00, a0, tmp1, (u32)1U);
+	fmul(b, t00, b, tmp);
+	fsquare_times(t00, b, tmp1, (u32)5U);
+	fmul(b, t00, b, tmp);
+	fsquare_times(t00, b, tmp1, (u32)10U);
+	fmul(c, t00, b, tmp);
+	fsquare_times(t00, c, tmp1, (u32)20U);
+	fmul(t00, t00, c, tmp);
+	fsquare_times(t00, t00, tmp1, (u32)10U);
+	fmul(b, t00, b, tmp);
+	fsquare_times(t00, b, tmp1, (u32)50U);
+	fmul(c, t00, b, tmp);
+	fsquare_times(t00, c, tmp1, (u32)100U);
+	fmul(t00, t00, c, tmp);
+	fsquare_times(t00, t00, tmp1, (u32)50U);
+	fmul(t00, t00, b, tmp);
+	fsquare_times(t00, t00, tmp1, (u32)5U);
+	a = t1;
+	t0 = t1 + (u32)12U;
+	fmul(o, t0, a, tmp);
+}
+
+static void store_felem(u64 *b, u64 *f)
+{
+	u64 f30 = f[3U];
+	u64 top_bit0 = f30 >> (u32)63U;
+	u64 carry0;
+	u64 f31;
+	u64 top_bit;
+	u64 carry;
+	u64 f0;
+	u64 f1;
+	u64 f2;
+	u64 f3;
+	u64 m0;
+	u64 m1;
+	u64 m2;
+	u64 m3;
+	u64 mask;
+	u64 f0_;
+	u64 f1_;
+	u64 f2_;
+	u64 f3_;
+	u64 o0;
+	u64 o1;
+	u64 o2;
+	u64 o3;
+	f[3U] = f30 & (u64)0x7fffffffffffffffU;
+	carry0 = add_scalar(f, f, (u64)19U * top_bit0);
+	f31 = f[3U];
+	top_bit = f31 >> (u32)63U;
+	f[3U] = f31 & (u64)0x7fffffffffffffffU;
+	carry = add_scalar(f, f, (u64)19U * top_bit);
+	f0 = f[0U];
+	f1 = f[1U];
+	f2 = f[2U];
+	f3 = f[3U];
+	m0 = gte_mask(f0, (u64)0xffffffffffffffedU);
+	m1 = eq_mask(f1, (u64)0xffffffffffffffffU);
+	m2 = eq_mask(f2, (u64)0xffffffffffffffffU);
+	m3 = eq_mask(f3, (u64)0x7fffffffffffffffU);
+	mask = ((m0 & m1) & m2) & m3;
+	f0_ = f0 - (mask & (u64)0xffffffffffffffedU);
+	f1_ = f1 - (mask & (u64)0xffffffffffffffffU);
+	f2_ = f2 - (mask & (u64)0xffffffffffffffffU);
+	f3_ = f3 - (mask & (u64)0x7fffffffffffffffU);
+	o0 = f0_;
+	o1 = f1_;
+	o2 = f2_;
+	o3 = f3_;
+	b[0U] = o0;
+	b[1U] = o1;
+	b[2U] = o2;
+	b[3U] = o3;
+}
+
+static void encode_point(u8 *o, const u64 *i)
+{
+	const u64 *x = i;
+	const u64 *z = i + (u32)4U;
+	u64 tmp[4U] = { 0U };
+	u64 tmp_w[16U] = { 0U };
+	finv(tmp, z, tmp_w);
+	fmul(tmp, tmp, x, tmp_w);
+	store_felem((u64 *)o, tmp);
+}
+
+static void curve25519_ever64(u8 *out, const u8 *priv, const u8 *pub)
+{
+	u64 init1[8U] = { 0U };
+	u64 tmp[4U] = { 0U };
+	u64 tmp3;
+	u64 *x;
+	u64 *z;
+	{
+		u32 i;
+		for (i = (u32)0U; i < (u32)4U; i = i + (u32)1U) {
+			u64 *os = tmp;
+			const u8 *bj = pub + i * (u32)8U;
+			u64 u = *(u64 *)bj;
+			u64 r = u;
+			u64 x0 = r;
+			os[i] = x0;
+		}
+	}
+	tmp3 = tmp[3U];
+	tmp[3U] = tmp3 & (u64)0x7fffffffffffffffU;
+	x = init1;
+	z = init1 + (u32)4U;
+	z[0U] = (u64)1U;
+	z[1U] = (u64)0U;
+	z[2U] = (u64)0U;
+	z[3U] = (u64)0U;
+	x[0U] = tmp[0U];
+	x[1U] = tmp[1U];
+	x[2U] = tmp[2U];
+	x[3U] = tmp[3U];
+	montgomery_ladder(init1, priv, init1);
+	encode_point(out, init1);
+}
+
+/* The below constants were generated using this sage script:
+ *
+ * #!/usr/bin/env sage
+ * import sys
+ * from sage.all import *
+ * def limbs(n):
+ * 	n = int(n)
+ * 	l = ((n >> 0) % 2^64, (n >> 64) % 2^64, (n >> 128) % 2^64, (n >> 192) % 2^64)
+ * 	return "0x%016xULL, 0x%016xULL, 0x%016xULL, 0x%016xULL" % l
+ * ec = EllipticCurve(GF(2^255 - 19), [0, 486662, 0, 1, 0])
+ * p_minus_s = (ec.lift_x(9) - ec.lift_x(1))[0]
+ * print("static const u64 p_minus_s[] = { %s };\n" % limbs(p_minus_s))
+ * print("static const u64 table_ladder[] = {")
+ * p = ec.lift_x(9)
+ * for i in range(252):
+ * 	l = (p[0] + p[2]) / (p[0] - p[2])
+ * 	print(("\t%s" + ("," if i != 251 else "")) % limbs(l))
+ * 	p = p * 2
+ * print("};")
+ *
+ */
+
+static const u64 p_minus_s[] = { 0x816b1e0137d48290ULL, 0x440f6a51eb4d1207ULL, 0x52385f46dca2b71dULL, 0x215132111d8354cbULL };
+
+static const u64 table_ladder[] = {
+	0xfffffffffffffff3ULL, 0xffffffffffffffffULL, 0xffffffffffffffffULL, 0x5fffffffffffffffULL,
+	0x6b8220f416aafe96ULL, 0x82ebeb2b4f566a34ULL, 0xd5a9a5b075a5950fULL, 0x5142b2cf4b2488f4ULL,
+	0x6aaebc750069680cULL, 0x89cf7820a0f99c41ULL, 0x2a58d9183b56d0f4ULL, 0x4b5aca80e36011a4ULL,
+	0x329132348c29745dULL, 0xf4a2e616e1642fd7ULL, 0x1e45bb03ff67bc34ULL, 0x306912d0f42a9b4aULL,
+	0xff886507e6af7154ULL, 0x04f50e13dfeec82fULL, 0xaa512fe82abab5ceULL, 0x174e251a68d5f222ULL,
+	0xcf96700d82028898ULL, 0x1743e3370a2c02c5ULL, 0x379eec98b4e86eaaULL, 0x0c59888a51e0482eULL,
+	0xfbcbf1d699b5d189ULL, 0xacaef0d58e9fdc84ULL, 0xc1c20d06231f7614ULL, 0x2938218da274f972ULL,
+	0xf6af49beff1d7f18ULL, 0xcc541c22387ac9c2ULL, 0x96fcc9ef4015c56bULL, 0x69c1627c690913a9ULL,
+	0x7a86fd2f4733db0eULL, 0xfdb8c4f29e087de9ULL, 0x095e4b1a8ea2a229ULL, 0x1ad7a7c829b37a79ULL,
+	0x342d89cad17ea0c0ULL, 0x67bedda6cced2051ULL, 0x19ca31bf2bb42f74ULL, 0x3df7b4c84980acbbULL,
+	0xa8c6444dc80ad883ULL, 0xb91e440366e3ab85ULL, 0xc215cda00164f6d8ULL, 0x3d867c6ef247e668ULL,
+	0xc7dd582bcc3e658cULL, 0xfd2c4748ee0e5528ULL, 0xa0fd9b95cc9f4f71ULL, 0x7529d871b0675ddfULL,
+	0xb8f568b42d3cbd78ULL, 0x1233011b91f3da82ULL, 0x2dce6ccd4a7c3b62ULL, 0x75e7fc8e9e498603ULL,
+	0x2f4f13f1fcd0b6ecULL, 0xf1a8ca1f29ff7a45ULL, 0xc249c1a72981e29bULL, 0x6ebe0dbb8c83b56aULL,
+	0x7114fa8d170bb222ULL, 0x65a2dcd5bf93935fULL, 0xbdc41f68b59c979aULL, 0x2f0eef79a2ce9289ULL,
+	0x42ecbf0c083c37ceULL, 0x2930bc09ec496322ULL, 0xf294b0c19cfeac0dULL, 0x3780aa4bedfabb80ULL,
+	0x56c17d3e7cead929ULL, 0xe7cb4beb2e5722c5ULL, 0x0ce931732dbfe15aULL, 0x41b883c7621052f8ULL,
+	0xdbf75ca0c3d25350ULL, 0x2936be086eb1e351ULL, 0xc936e03cb4a9b212ULL, 0x1d45bf82322225aaULL,
+	0xe81ab1036a024cc5ULL, 0xe212201c304c9a72ULL, 0xc5d73fba6832b1fcULL, 0x20ffdb5a4d839581ULL,
+	0xa283d367be5d0fadULL, 0x6c2b25ca8b164475ULL, 0x9d4935467caaf22eULL, 0x5166408eee85ff49ULL,
+	0x3c67baa2fab4e361ULL, 0xb3e433c67ef35cefULL, 0x5259729241159b1cULL, 0x6a621892d5b0ab33ULL,
+	0x20b74a387555cdcbULL, 0x532aa10e1208923fULL, 0xeaa17b7762281dd1ULL, 0x61ab3443f05c44bfULL,
+	0x257a6c422324def8ULL, 0x131c6c1017e3cf7fULL, 0x23758739f630a257ULL, 0x295a407a01a78580ULL,
+	0xf8c443246d5da8d9ULL, 0x19d775450c52fa5dULL, 0x2afcfc92731bf83dULL, 0x7d10c8e81b2b4700ULL,
+	0xc8e0271f70baa20bULL, 0x993748867ca63957ULL, 0x5412efb3cb7ed4bbULL, 0x3196d36173e62975ULL,
+	0xde5bcad141c7dffcULL, 0x47cc8cd2b395c848ULL, 0xa34cd942e11af3cbULL, 0x0256dbf2d04ecec2ULL,
+	0x875ab7e94b0e667fULL, 0xcad4dd83c0850d10ULL, 0x47f12e8f4e72c79fULL, 0x5f1a87bb8c85b19bULL,
+	0x7ae9d0b6437f51b8ULL, 0x12c7ce5518879065ULL, 0x2ade09fe5cf77aeeULL, 0x23a05a2f7d2c5627ULL,
+	0x5908e128f17c169aULL, 0xf77498dd8ad0852dULL, 0x74b4c4ceab102f64ULL, 0x183abadd10139845ULL,
+	0xb165ba8daa92aaacULL, 0xd5c5ef9599386705ULL, 0xbe2f8f0cf8fc40d1ULL, 0x2701e635ee204514ULL,
+	0x629fa80020156514ULL, 0xf223868764a8c1ceULL, 0x5b894fff0b3f060eULL, 0x60d9944cf708a3faULL,
+	0xaeea001a1c7a201fULL, 0xebf16a633ee2ce63ULL, 0x6f7709594c7a07e1ULL, 0x79b958150d0208cbULL,
+	0x24b55e5301d410e7ULL, 0xe3a34edff3fdc84dULL, 0xd88768e4904032d8ULL, 0x131384427b3aaeecULL,
+	0x8405e51286234f14ULL, 0x14dc4739adb4c529ULL, 0xb8a2b5b250634ffdULL, 0x2fe2a94ad8a7ff93ULL,
+	0xec5c57efe843faddULL, 0x2843ce40f0bb9918ULL, 0xa4b561d6cf3d6305ULL, 0x743629bde8fb777eULL,
+	0x343edd46bbaf738fULL, 0xed981828b101a651ULL, 0xa401760b882c797aULL, 0x1fc223e28dc88730ULL,
+	0x48604e91fc0fba0eULL, 0xb637f78f052c6fa4ULL, 0x91ccac3d09e9239cULL, 0x23f7eed4437a687cULL,
+	0x5173b1118d9bd800ULL, 0x29d641b63189d4a7ULL, 0xfdbf177988bbc586ULL, 0x2959894fcad81df5ULL,
+	0xaebc8ef3b4bbc899ULL, 0x4148995ab26992b9ULL, 0x24e20b0134f92cfbULL, 0x40d158894a05dee8ULL,
+	0x46b00b1185af76f6ULL, 0x26bac77873187a79ULL, 0x3dc0bf95ab8fff5fULL, 0x2a608bd8945524d7ULL,
+	0x26449588bd446302ULL, 0x7c4bc21c0388439cULL, 0x8e98a4f383bd11b2ULL, 0x26218d7bc9d876b9ULL,
+	0xe3081542997c178aULL, 0x3c2d29a86fb6606fULL, 0x5c217736fa279374ULL, 0x7dde05734afeb1faULL,
+	0x3bf10e3906d42babULL, 0xe4f7803e1980649cULL, 0xe6053bf89595bf7aULL, 0x394faf38da245530ULL,
+	0x7a8efb58896928f4ULL, 0xfbc778e9cc6a113cULL, 0x72670ce330af596fULL, 0x48f222a81d3d6cf7ULL,
+	0xf01fce410d72caa7ULL, 0x5a20ecc7213b5595ULL, 0x7bc21165c1fa1483ULL, 0x07f89ae31da8a741ULL,
+	0x05d2c2b4c6830ff9ULL, 0xd43e330fc6316293ULL, 0xa5a5590a96d3a904ULL, 0x705edb91a65333b6ULL,
+	0x048ee15e0bb9a5f7ULL, 0x3240cfca9e0aaf5dULL, 0x8f4b71ceedc4a40bULL, 0x621c0da3de544a6dULL,
+	0x92872836a08c4091ULL, 0xce8375b010c91445ULL, 0x8a72eb524f276394ULL, 0x2667fcfa7ec83635ULL,
+	0x7f4c173345e8752aULL, 0x061b47feee7079a5ULL, 0x25dd9afa9f86ff34ULL, 0x3780cef5425dc89cULL,
+	0x1a46035a513bb4e9ULL, 0x3e1ef379ac575adaULL, 0xc78c5f1c5fa24b50ULL, 0x321a967634fd9f22ULL,
+	0x946707b8826e27faULL, 0x3dca84d64c506fd0ULL, 0xc189218075e91436ULL, 0x6d9284169b3b8484ULL,
+	0x3a67e840383f2ddfULL, 0x33eec9a30c4f9b75ULL, 0x3ec7c86fa783ef47ULL, 0x26ec449fbac9fbc4ULL,
+	0x5c0f38cba09b9e7dULL, 0x81168cc762a3478cULL, 0x3e23b0d306fc121cULL, 0x5a238aa0a5efdcddULL,
+	0x1ba26121c4ea43ffULL, 0x36f8c77f7c8832b5ULL, 0x88fbea0b0adcf99aULL, 0x5ca9938ec25bebf9ULL,
+	0xd5436a5e51fccda0ULL, 0x1dbc4797c2cd893bULL, 0x19346a65d3224a08ULL, 0x0f5034e49b9af466ULL,
+	0xf23c3967a1e0b96eULL, 0xe58b08fa867a4d88ULL, 0xfb2fabc6a7341679ULL, 0x2a75381eb6026946ULL,
+	0xc80a3be4c19420acULL, 0x66b1f6c681f2b6dcULL, 0x7cf7036761e93388ULL, 0x25abbbd8a660a4c4ULL,
+	0x91ea12ba14fd5198ULL, 0x684950fc4a3cffa9ULL, 0xf826842130f5ad28ULL, 0x3ea988f75301a441ULL,
+	0xc978109a695f8c6fULL, 0x1746eb4a0530c3f3ULL, 0x444d6d77b4459995ULL, 0x75952b8c054e5cc7ULL,
+	0xa3703f7915f4d6aaULL, 0x66c346202f2647d8ULL, 0xd01469df811d644bULL, 0x77fea47d81a5d71fULL,
+	0xc5e9529ef57ca381ULL, 0x6eeeb4b9ce2f881aULL, 0xb6e91a28e8009bd6ULL, 0x4b80be3e9afc3fecULL,
+	0x7e3773c526aed2c5ULL, 0x1b4afcb453c9a49dULL, 0xa920bdd7baffb24dULL, 0x7c54699f122d400eULL,
+	0xef46c8e14fa94bc8ULL, 0xe0b074ce2952ed5eULL, 0xbea450e1dbd885d5ULL, 0x61b68649320f712cULL,
+	0x8a485f7309ccbdd1ULL, 0xbd06320d7d4d1a2dULL, 0x25232973322dbef4ULL, 0x445dc4758c17f770ULL,
+	0xdb0434177cc8933cULL, 0xed6fe82175ea059fULL, 0x1efebefdc053db34ULL, 0x4adbe867c65daf99ULL,
+	0x3acd71a2a90609dfULL, 0xe5e991856dd04050ULL, 0x1ec69b688157c23cULL, 0x697427f6885cfe4dULL,
+	0xd7be7b9b65e1a851ULL, 0xa03d28d522c536ddULL, 0x28399d658fd2b645ULL, 0x49e5b7e17c2641e1ULL,
+	0x6f8c3a98700457a4ULL, 0x5078f0a25ebb6778ULL, 0xd13c3ccbc382960fULL, 0x2e003258a7df84b1ULL,
+	0x8ad1f39be6296a1cULL, 0xc1eeaa652a5fbfb2ULL, 0x33ee0673fd26f3cbULL, 0x59256173a69d2cccULL,
+	0x41ea07aa4e18fc41ULL, 0xd9fc19527c87a51eULL, 0xbdaacb805831ca6fULL, 0x445b652dc916694fULL,
+	0xce92a3a7f2172315ULL, 0x1edc282de11b9964ULL, 0xa1823aafe04c314aULL, 0x790a2d94437cf586ULL,
+	0x71c447fb93f6e009ULL, 0x8922a56722845276ULL, 0xbf70903b204f5169ULL, 0x2f7a89891ba319feULL,
+	0x02a08eb577e2140cULL, 0xed9a4ed4427bdcf4ULL, 0x5253ec44e4323cd1ULL, 0x3e88363c14e9355bULL,
+	0xaa66c14277110b8cULL, 0x1ae0391610a23390ULL, 0x2030bd12c93fc2a2ULL, 0x3ee141579555c7abULL,
+	0x9214de3a6d6e7d41ULL, 0x3ccdd88607f17efeULL, 0x674f1288f8e11217ULL, 0x5682250f329f93d0ULL,
+	0x6cf00b136d2e396eULL, 0x6e4cf86f1014debfULL, 0x5930b1b5bfcc4e83ULL, 0x047069b48aba16b6ULL,
+	0x0d4ce4ab69b20793ULL, 0xb24db91a97d0fb9eULL, 0xcdfa50f54e00d01dULL, 0x221b1085368bddb5ULL,
+	0xe7e59468b1e3d8d2ULL, 0x53c56563bd122f93ULL, 0xeee8a903e0663f09ULL, 0x61efa662cbbe3d42ULL,
+	0x2cf8ddddde6eab2aULL, 0x9bf80ad51435f231ULL, 0x5deadacec9f04973ULL, 0x29275b5d41d29b27ULL,
+	0xcfde0f0895ebf14fULL, 0xb9aab96b054905a7ULL, 0xcae80dd9a1c420fdULL, 0x0a63bf2f1673bbc7ULL,
+	0x092f6e11958fbc8cULL, 0x672a81e804822fadULL, 0xcac8351560d52517ULL, 0x6f3f7722c8f192f8ULL,
+	0xf8ba90ccc2e894b7ULL, 0x2c7557a438ff9f0dULL, 0x894d1d855ae52359ULL, 0x68e122157b743d69ULL,
+	0xd87e5570cfb919f3ULL, 0x3f2cdecd95798db9ULL, 0x2121154710c0a2ceULL, 0x3c66a115246dc5b2ULL,
+	0xcbedc562294ecb72ULL, 0xba7143c36a280b16ULL, 0x9610c2efd4078b67ULL, 0x6144735d946a4b1eULL,
+	0x536f111ed75b3350ULL, 0x0211db8c2041d81bULL, 0xf93cb1000e10413cULL, 0x149dfd3c039e8876ULL,
+	0xd479dde46b63155bULL, 0xb66e15e93c837976ULL, 0xdafde43b1f13e038ULL, 0x5fafda1a2e4b0b35ULL,
+	0x3600bbdf17197581ULL, 0x3972050bbe3cd2c2ULL, 0x5938906dbdd5be86ULL, 0x34fce5e43f9b860fULL,
+	0x75a8a4cd42d14d02ULL, 0x828dabc53441df65ULL, 0x33dcabedd2e131d3ULL, 0x3ebad76fb814d25fULL,
+	0xd4906f566f70e10fULL, 0x5d12f7aa51690f5aULL, 0x45adb16e76cefcf2ULL, 0x01f768aead232999ULL,
+	0x2b6cc77b6248febdULL, 0x3cd30628ec3aaffdULL, 0xce1c0b80d4ef486aULL, 0x4c3bff2ea6f66c23ULL,
+	0x3f2ec4094aeaeb5fULL, 0x61b19b286e372ca7ULL, 0x5eefa966de2a701dULL, 0x23b20565de55e3efULL,
+	0xe301ca5279d58557ULL, 0x07b2d4ce27c2874fULL, 0xa532cd8a9dcf1d67ULL, 0x2a52fee23f2bff56ULL,
+	0x8624efb37cd8663dULL, 0xbbc7ac20ffbd7594ULL, 0x57b85e9c82d37445ULL, 0x7b3052cb86a6ec66ULL,
+	0x3482f0ad2525e91eULL, 0x2cb68043d28edca0ULL, 0xaf4f6d052e1b003aULL, 0x185f8c2529781b0aULL,
+	0xaa41de5bd80ce0d6ULL, 0x9407b2416853e9d6ULL, 0x563ec36e357f4c3aULL, 0x4cc4b8dd0e297bceULL,
+	0xa2fc1a52ffb8730eULL, 0x1811f16e67058e37ULL, 0x10f9a366cddf4ee1ULL, 0x72f4a0c4a0b9f099ULL,
+	0x8c16c06f663f4ea7ULL, 0x693b3af74e970fbaULL, 0x2102e7f1d69ec345ULL, 0x0ba53cbc968a8089ULL,
+	0xca3d9dc7fea15537ULL, 0x4c6824bb51536493ULL, 0xb9886314844006b1ULL, 0x40d2a72ab454cc60ULL,
+	0x5936a1b712570975ULL, 0x91b9d648debda657ULL, 0x3344094bb64330eaULL, 0x006ba10d12ee51d0ULL,
+	0x19228468f5de5d58ULL, 0x0eb12f4c38cc05b0ULL, 0xa1039f9dd5601990ULL, 0x4502d4ce4fff0e0bULL,
+	0xeb2054106837c189ULL, 0xd0f6544c6dd3b93cULL, 0x40727064c416d74fULL, 0x6e15c6114b502ef0ULL,
+	0x4df2a398cfb1a76bULL, 0x11256c7419f2f6b1ULL, 0x4a497962066e6043ULL, 0x705b3aab41355b44ULL,
+	0x365ef536d797b1d8ULL, 0x00076bd622ddf0dbULL, 0x3bbf33b0e0575a88ULL, 0x3777aa05c8e4ca4dULL,
+	0x392745c85578db5fULL, 0x6fda4149dbae5ae2ULL, 0xb1f0b00b8adc9867ULL, 0x09963437d36f1da3ULL,
+	0x7e824e90a5dc3853ULL, 0xccb5f6641f135cbdULL, 0x6736d86c87ce8fccULL, 0x625f3ce26604249fULL,
+	0xaf8ac8059502f63fULL, 0x0c05e70a2e351469ULL, 0x35292e9c764b6305ULL, 0x1a394360c7e23ac3ULL,
+	0xd5c6d53251183264ULL, 0x62065abd43c2b74fULL, 0xb5fbf5d03b973f9bULL, 0x13a3da3661206e5eULL,
+	0xc6bd5837725d94e5ULL, 0x18e30912205016c5ULL, 0x2088ce1570033c68ULL, 0x7fba1f495c837987ULL,
+	0x5a8c7423f2f9079dULL, 0x1735157b34023fc5ULL, 0xe4f9b49ad2fab351ULL, 0x6691ff72c878e33cULL,
+	0x122c2adedc5eff3eULL, 0xf8dd4bf1d8956cf4ULL, 0xeb86205d9e9e5bdaULL, 0x049b92b9d975c743ULL,
+	0xa5379730b0f6c05aULL, 0x72a0ffacc6f3a553ULL, 0xb0032c34b20dcd6dULL, 0x470e9dbc88d5164aULL,
+	0xb19cf10ca237c047ULL, 0xb65466711f6c81a2ULL, 0xb3321bd16dd80b43ULL, 0x48c14f600c5fbe8eULL,
+	0x66451c264aa6c803ULL, 0xb66e3904a4fa7da6ULL, 0xd45f19b0b3128395ULL, 0x31602627c3c9bc10ULL,
+	0x3120dc4832e4e10dULL, 0xeb20c46756c717f7ULL, 0x00f52e3f67280294ULL, 0x566d4fc14730c509ULL,
+	0x7e3a5d40fd837206ULL, 0xc1e926dc7159547aULL, 0x216730fba68d6095ULL, 0x22e8c3843f69cea7ULL,
+	0x33d074e8930e4b2bULL, 0xb6e4350e84d15816ULL, 0x5534c26ad6ba2365ULL, 0x7773c12f89f1f3f3ULL,
+	0x8cba404da57962aaULL, 0x5b9897a81999ce56ULL, 0x508e862f121692fcULL, 0x3a81907fa093c291ULL,
+	0x0dded0ff4725a510ULL, 0x10d8cc10673fc503ULL, 0x5b9d151c9f1f4e89ULL, 0x32a5c1d5cb09a44cULL,
+	0x1e0aa442b90541fbULL, 0x5f85eb7cc1b485dbULL, 0xbee595ce8a9df2e5ULL, 0x25e496c722422236ULL,
+	0x5edf3c46cd0fe5b9ULL, 0x34e75a7ed2a43388ULL, 0xe488de11d761e352ULL, 0x0e878a01a085545cULL,
+	0xba493c77e021bb04ULL, 0x2b4d1843c7df899aULL, 0x9ea37a487ae80d67ULL, 0x67a9958011e41794ULL,
+	0x4b58051a6697b065ULL, 0x47e33f7d8d6ba6d4ULL, 0xbb4da8d483ca46c1ULL, 0x68becaa181c2db0dULL,
+	0x8d8980e90b989aa5ULL, 0xf95eb14a2c93c99bULL, 0x51c6c7c4796e73a2ULL, 0x6e228363b5efb569ULL,
+	0xc6bbc0b02dd624c8ULL, 0x777eb47dec8170eeULL, 0x3cde15a004cfafa9ULL, 0x1dc6bc087160bf9bULL,
+	0x2e07e043eec34002ULL, 0x18e9fc677a68dc7fULL, 0xd8da03188bd15b9aULL, 0x48fbc3bb00568253ULL,
+	0x57547d4cfb654ce1ULL, 0xd3565b82a058e2adULL, 0xf63eaf0bbf154478ULL, 0x47531ef114dfbb18ULL,
+	0xe1ec630a4278c587ULL, 0x5507d546ca8e83f3ULL, 0x85e135c63adc0c2bULL, 0x0aa7efa85682844eULL,
+	0x72691ba8b3e1f615ULL, 0x32b4e9701fbe3ffaULL, 0x97b6d92e39bb7868ULL, 0x2cfe53dea02e39e8ULL,
+	0x687392cd85cd52b0ULL, 0x27ff66c910e29831ULL, 0x97134556a9832d06ULL, 0x269bb0360a84f8a0ULL,
+	0x706e55457643f85cULL, 0x3734a48c9b597d1bULL, 0x7aee91e8c6efa472ULL, 0x5cd6abc198a9d9e0ULL,
+	0x0e04de06cb3ce41aULL, 0xd8c6eb893402e138ULL, 0x904659bb686e3772ULL, 0x7215c371746ba8c8ULL,
+	0xfd12a97eeae4a2d9ULL, 0x9514b7516394f2c5ULL, 0x266fd5809208f294ULL, 0x5c847085619a26b9ULL,
+	0x52985410fed694eaULL, 0x3c905b934a2ed254ULL, 0x10bb47692d3be467ULL, 0x063b3d2d69e5e9e1ULL,
+	0x472726eedda57debULL, 0xefb6c4ae10f41891ULL, 0x2b1641917b307614ULL, 0x117c554fc4f45b7cULL,
+	0xc07cf3118f9d8812ULL, 0x01dbd82050017939ULL, 0xd7e803f4171b2827ULL, 0x1015e87487d225eaULL,
+	0xc58de3fed23acc4dULL, 0x50db91c294a7be2dULL, 0x0b94d43d1c9cf457ULL, 0x6b1640fa6e37524aULL,
+	0x692f346c5fda0d09ULL, 0x200b1c59fa4d3151ULL, 0xb8c46f760777a296ULL, 0x4b38395f3ffdfbcfULL,
+	0x18d25e00be54d671ULL, 0x60d50582bec8aba6ULL, 0x87ad8f263b78b982ULL, 0x50fdf64e9cda0432ULL,
+	0x90f567aac578dcf0ULL, 0xef1e9b0ef2a3133bULL, 0x0eebba9242d9de71ULL, 0x15473c9bf03101c7ULL,
+	0x7c77e8ae56b78095ULL, 0xb678e7666e6f078eULL, 0x2da0b9615348ba1fULL, 0x7cf931c1ff733f0bULL,
+	0x26b357f50a0a366cULL, 0xe9708cf42b87d732ULL, 0xc13aeea5f91cb2c0ULL, 0x35d90c991143bb4cULL,
+	0x47c1c404a9a0d9dcULL, 0x659e58451972d251ULL, 0x3875a8c473b38c31ULL, 0x1fbd9ed379561f24ULL,
+	0x11fabc6fd41ec28dULL, 0x7ef8dfe3cd2a2dcaULL, 0x72e73b5d8c404595ULL, 0x6135fa4954b72f27ULL,
+	0xccfc32a2de24b69cULL, 0x3f55698c1f095d88ULL, 0xbe3350ed5ac3f929ULL, 0x5e9bf806ca477eebULL,
+	0xe9ce8fb63c309f68ULL, 0x5376f63565e1f9f4ULL, 0xd1afcfb35a6393f1ULL, 0x6632a1ede5623506ULL,
+	0x0b7d6c390c2ded4cULL, 0x56cb3281df04cb1fULL, 0x66305a1249ecc3c7ULL, 0x5d588b60a38ca72aULL,
+	0xa6ecbf78e8e5f42dULL, 0x86eeb44b3c8a3eecULL, 0xec219c48fbd21604ULL, 0x1aaf1af517c36731ULL,
+	0xc306a2836769bde7ULL, 0x208280622b1e2adbULL, 0x8027f51ffbff94a6ULL, 0x76cfa1ce1124f26bULL,
+	0x18eb00562422abb6ULL, 0xf377c4d58f8c29c3ULL, 0x4dbbc207f531561aULL, 0x0253b7f082128a27ULL,
+	0x3d1f091cb62c17e0ULL, 0x4860e1abd64628a9ULL, 0x52d17436309d4253ULL, 0x356f97e13efae576ULL,
+	0xd351e11aa150535bULL, 0x3e6b45bb1dd878ccULL, 0x0c776128bed92c98ULL, 0x1d34ae93032885b8ULL,
+	0x4ba0488ca85ba4c3ULL, 0x985348c33c9ce6ceULL, 0x66124c6f97bda770ULL, 0x0f81a0290654124aULL,
+	0x9ed09ca6569b86fdULL, 0x811009fd18af9a2dULL, 0xff08d03f93d8c20aULL, 0x52a148199faef26bULL,
+	0x3e03f9dc2d8d1b73ULL, 0x4205801873961a70ULL, 0xc0d987f041a35970ULL, 0x07aa1f15a1c0d549ULL,
+	0xdfd46ce08cd27224ULL, 0x6d0a024f934e4239ULL, 0x808a7a6399897b59ULL, 0x0a4556e9e13d95a2ULL,
+	0xd21a991fe9c13045ULL, 0x9b0e8548fe7751b8ULL, 0x5da643cb4bf30035ULL, 0x77db28d63940f721ULL,
+	0xfc5eeb614adc9011ULL, 0x5229419ae8c411ebULL, 0x9ec3e7787d1dcf74ULL, 0x340d053e216e4cb5ULL,
+	0xcac7af39b48df2b4ULL, 0xc0faec2871a10a94ULL, 0x140a69245ca575edULL, 0x0cf1c37134273a4cULL,
+	0xc8ee306ac224b8a5ULL, 0x57eaee7ccb4930b0ULL, 0xa1e806bdaacbe74fULL, 0x7d9a62742eeb657dULL,
+	0x9eb6b6ef546c4830ULL, 0x885cca1fddb36e2eULL, 0xe6b9f383ef0d7105ULL, 0x58654fef9d2e0412ULL,
+	0xa905c4ffbe0e8e26ULL, 0x942de5df9b31816eULL, 0x497d723f802e88e1ULL, 0x30684dea602f408dULL,
+	0x21e5a278a3e6cb34ULL, 0xaefb6e6f5b151dc4ULL, 0xb30b8e049d77ca15ULL, 0x28c3c9cf53b98981ULL,
+	0x287fb721556cdd2aULL, 0x0d317ca897022274ULL, 0x7468c7423a543258ULL, 0x4a7f11464eb5642fULL,
+	0xa237a4774d193aa6ULL, 0xd865986ea92129a1ULL, 0x24c515ecf87c1a88ULL, 0x604003575f39f5ebULL,
+	0x47b9f189570a9b27ULL, 0x2b98cede465e4b78ULL, 0x026df551dbb85c20ULL, 0x74fcd91047e21901ULL,
+	0x13e2a90a23c1bfa3ULL, 0x0cb0074e478519f6ULL, 0x5ff1cbbe3af6cf44ULL, 0x67fe5438be812dbeULL,
+	0xd13cf64fa40f05b0ULL, 0x054dfb2f32283787ULL, 0x4173915b7f0d2aeaULL, 0x482f144f1f610d4eULL,
+	0xf6210201b47f8234ULL, 0x5d0ae1929e70b990ULL, 0xdcd7f455b049567cULL, 0x7e93d0f1f0916f01ULL,
+	0xdd79cbf18a7db4faULL, 0xbe8391bf6f74c62fULL, 0x027145d14b8291bdULL, 0x585a73ea2cbf1705ULL,
+	0x485ca03e928a0db2ULL, 0x10fc01a5742857e7ULL, 0x2f482edbd6d551a7ULL, 0x0f0433b5048fdb8aULL,
+	0x60da2e8dd7dc6247ULL, 0x88b4c9d38cd4819aULL, 0x13033ac001f66697ULL, 0x273b24fe3b367d75ULL,
+	0xc6e8f66a31b3b9d4ULL, 0x281514a494df49d5ULL, 0xd1726fdfc8b23da7ULL, 0x4b3ae7d103dee548ULL,
+	0xc6256e19ce4b9d7eULL, 0xff5c5cf186e3c61cULL, 0xacc63ca34b8ec145ULL, 0x74621888fee66574ULL,
+	0x956f409645290a1eULL, 0xef0bf8e3263a962eULL, 0xed6a50eb5ec2647bULL, 0x0694283a9dca7502ULL,
+	0x769b963643a2dcd1ULL, 0x42b7c8ea09fc5353ULL, 0x4f002aee13397eabULL, 0x63005e2c19b7d63aULL,
+	0xca6736da63023beaULL, 0x966c7f6db12a99b7ULL, 0xace09390c537c5e1ULL, 0x0b696063a1aa89eeULL,
+	0xebb03e97288c56e5ULL, 0x432a9f9f938c8be8ULL, 0xa6a5a93d5b717f71ULL, 0x1a5fb4c3e18f9d97ULL,
+	0x1c94e7ad1c60cdceULL, 0xee202a43fc02c4a0ULL, 0x8dafe4d867c46a20ULL, 0x0a10263c8ac27b58ULL,
+	0xd0dea9dfe4432a4aULL, 0x856af87bbe9277c5ULL, 0xce8472acc212c71aULL, 0x6f151b6d9bbb1e91ULL,
+	0x26776c527ceed56aULL, 0x7d211cb7fbf8faecULL, 0x37ae66a6fd4609ccULL, 0x1f81b702d2770c42ULL,
+	0x2fb0b057eac58392ULL, 0xe1dd89fe29744e9dULL, 0xc964f8eb17beb4f8ULL, 0x29571073c9a2d41eULL,
+	0xa948a18981c0e254ULL, 0x2df6369b65b22830ULL, 0xa33eb2d75fcfd3c6ULL, 0x078cd6ec4199a01fULL,
+	0x4a584a41ad900d2fULL, 0x32142b78e2c74c52ULL, 0x68c4e8338431c978ULL, 0x7f69ea9008689fc2ULL,
+	0x52f2c81e46a38265ULL, 0xfd78072d04a832fdULL, 0x8cd7d5fa25359e94ULL, 0x4de71b7454cc29d2ULL,
+	0x42eb60ad1eda6ac9ULL, 0x0aad37dfdbc09c3aULL, 0x81004b71e33cc191ULL, 0x44e6be345122803cULL,
+	0x03fe8388ba1920dbULL, 0xf5d57c32150db008ULL, 0x49c8c4281af60c29ULL, 0x21edb518de701aeeULL,
+	0x7fb63e418f06dc99ULL, 0xa4460d99c166d7b8ULL, 0x24dd5248ce520a83ULL, 0x5ec3ad712b928358ULL,
+	0x15022a5fbd17930fULL, 0xa4f64a77d82570e3ULL, 0x12bc8d6915783712ULL, 0x498194c0fc620abbULL,
+	0x38a2d9d255686c82ULL, 0x785c6bd9193e21f0ULL, 0xe4d5c81ab24a5484ULL, 0x56307860b2e20989ULL,
+	0x429d55f78b4d74c4ULL, 0x22f1834643350131ULL, 0x1e60c24598c71fffULL, 0x59f2f014979983efULL,
+	0x46a47d56eb494a44ULL, 0x3e22a854d636a18eULL, 0xb346e15274491c3bULL, 0x2ceafd4e5390cde7ULL,
+	0xba8a8538be0d6675ULL, 0x4b9074bb50818e23ULL, 0xcbdab89085d304c3ULL, 0x61a24fe0e56192c4ULL,
+	0xcb7615e6db525bcbULL, 0xdd7d8c35a567e4caULL, 0xe6b4153acafcdd69ULL, 0x2d668e097f3c9766ULL,
+	0xa57e7e265ce55ef0ULL, 0x5d9f4e527cd4b967ULL, 0xfbc83606492fd1e5ULL, 0x090d52beb7c3f7aeULL,
+	0x09b9515a1e7b4d7cULL, 0x1f266a2599da44c0ULL, 0xa1c49548e2c55504ULL, 0x7ef04287126f15ccULL,
+	0xfed1659dbd30ef15ULL, 0x8b4ab9eec4e0277bULL, 0x884d6236a5df3291ULL, 0x1fd96ea6bf5cf788ULL,
+	0x42a161981f190d9aULL, 0x61d849507e6052c1ULL, 0x9fe113bf285a2cd5ULL, 0x7c22d676dbad85d8ULL,
+	0x82e770ed2bfbd27dULL, 0x4c05b2ece996f5a5ULL, 0xcd40a9c2b0900150ULL, 0x5895319213d9bf64ULL,
+	0xe7cc5d703fea2e08ULL, 0xb50c491258e2188cULL, 0xcce30baa48205bf0ULL, 0x537c659ccfa32d62ULL,
+	0x37b6623a98cfc088ULL, 0xfe9bed1fa4d6aca4ULL, 0x04d29b8e56a8d1b0ULL, 0x725f71c40b519575ULL,
+	0x28c7f89cd0339ce6ULL, 0x8367b14469ddc18bULL, 0x883ada83a6a1652cULL, 0x585f1974034d6c17ULL,
+	0x89cfb266f1b19188ULL, 0xe63b4863e7c35217ULL, 0xd88c9da6b4c0526aULL, 0x3e035c9df0954635ULL,
+	0xdd9d5412fb45de9dULL, 0xdd684532e4cff40dULL, 0x4b5c999b151d671cULL, 0x2d8c2cc811e7f690ULL,
+	0x7f54be1d90055d40ULL, 0xa464c5df464aaf40ULL, 0x33979624f0e917beULL, 0x2c018dc527356b30ULL,
+	0xa5415024e330b3d4ULL, 0x73ff3d96691652d3ULL, 0x94ec42c4ef9b59f1ULL, 0x0747201618d08e5aULL,
+	0x4d6ca48aca411c53ULL, 0x66415f2fcfa66119ULL, 0x9c4dd40051e227ffULL, 0x59810bc09a02f7ebULL,
+	0x2a7eb171b3dc101dULL, 0x441c5ab99ffef68eULL, 0x32025c9b93b359eaULL, 0x5e8ce0a71e9d112fULL,
+	0xbfcccb92429503fdULL, 0xd271ba752f095d55ULL, 0x345ead5e972d091eULL, 0x18c8df11a83103baULL,
+	0x90cd949a9aed0f4cULL, 0xc5d1f4cb6660e37eULL, 0xb8cac52d56c52e0bULL, 0x6e42e400c5808e0dULL,
+	0xa3b46966eeaefd23ULL, 0x0c4f1f0be39ecdcaULL, 0x189dc8c9d683a51dULL, 0x51f27f054c09351bULL,
+	0x4c487ccd2a320682ULL, 0x587ea95bb3df1c96ULL, 0xc8ccf79e555cb8e8ULL, 0x547dc829a206d73dULL,
+	0xb822a6cd80c39b06ULL, 0xe96d54732000d4c6ULL, 0x28535b6f91463b4dULL, 0x228f4660e2486e1dULL,
+	0x98799538de8d3abfULL, 0x8cd8330045ebca6eULL, 0x79952a008221e738ULL, 0x4322e1a7535cd2bbULL,
+	0xb114c11819d1801cULL, 0x2016e4d84f3f5ec7ULL, 0xdd0e2df409260f4cULL, 0x5ec362c0ae5f7266ULL,
+	0xc0462b18b8b2b4eeULL, 0x7cc8d950274d1afbULL, 0xf25f7105436b02d2ULL, 0x43bbf8dcbff9ccd3ULL,
+	0xb6ad1767a039e9dfULL, 0xb0714da8f69d3583ULL, 0x5e55fa18b42931f5ULL, 0x4ed5558f33c60961ULL,
+	0x1fe37901c647a5ddULL, 0x593ddf1f8081d357ULL, 0x0249a4fd813fd7a6ULL, 0x69acca274e9caf61ULL,
+	0x047ba3ea330721c9ULL, 0x83423fc20e7e1ea0ULL, 0x1df4c0af01314a60ULL, 0x09a62dab89289527ULL,
+	0xa5b325a49cc6cb00ULL, 0xe94b5dc654b56cb6ULL, 0x3be28779adc994a0ULL, 0x4296e8f8ba3a4aadULL,
+	0x328689761e451eabULL, 0x2e4d598bff59594aULL, 0x49b96853d7a7084aULL, 0x4980a319601420a8ULL,
+	0x9565b9e12f552c42ULL, 0x8a5318db7100fe96ULL, 0x05c90b4d43add0d7ULL, 0x538b4cd66a5d4edaULL,
+	0xf4e94fc3e89f039fULL, 0x592c9af26f618045ULL, 0x08a36eb5fd4b9550ULL, 0x25fffaf6c2ed1419ULL,
+	0x34434459cc79d354ULL, 0xeeecbfb4b1d5476bULL, 0xddeb34a061615d99ULL, 0x5129cecceb64b773ULL,
+	0xee43215894993520ULL, 0x772f9c7cf14c0b3bULL, 0xd2e2fce306bedad5ULL, 0x715f42b546f06a97ULL,
+	0x434ecdceda5b5f1aULL, 0x0da17115a49741a9ULL, 0x680bd77c73edad2eULL, 0x487c02354edd9041ULL,
+	0xb8efeff3a70ed9c4ULL, 0x56a32aa3e857e302ULL, 0xdf3a68bd48a2a5a0ULL, 0x07f650b73176c444ULL,
+	0xe38b9b1626e0ccb1ULL, 0x79e053c18b09fb36ULL, 0x56d90319c9f94964ULL, 0x1ca941e7ac9ff5c4ULL,
+	0x49c4df29162fa0bbULL, 0x8488cf3282b33305ULL, 0x95dfda14cabb437dULL, 0x3391f78264d5ad86ULL,
+	0x729ae06ae2b5095dULL, 0xd58a58d73259a946ULL, 0xe9834262d13921edULL, 0x27fedafaa54bb592ULL,
+	0xa99dc5b829ad48bbULL, 0x5f025742499ee260ULL, 0x802c8ecd5d7513fdULL, 0x78ceb3ef3f6dd938ULL,
+	0xc342f44f8a135d94ULL, 0x7b9edb44828cdda3ULL, 0x9436d11a0537cfe7ULL, 0x5064b164ec1ab4c8ULL,
+	0x7020eccfd37eb2fcULL, 0x1f31ea3ed90d25fcULL, 0x1b930d7bdfa1bb34ULL, 0x5344467a48113044ULL,
+	0x70073170f25e6dfbULL, 0xe385dc1a50114cc8ULL, 0x2348698ac8fc4f00ULL, 0x2a77a55284dd40d8ULL,
+	0xfe06afe0c98c6ce4ULL, 0xc235df96dddfd6e4ULL, 0x1428d01e33bf1ed3ULL, 0x785768ec9300bdafULL,
+	0x9702e57a91deb63bULL, 0x61bdb8bfe5ce8b80ULL, 0x645b426f3d1d58acULL, 0x4804a82227a557bcULL,
+	0x8e57048ab44d2601ULL, 0x68d6501a4b3a6935ULL, 0xc39c9ec3f9e1c293ULL, 0x4172f257d4de63e2ULL,
+	0xd368b450330c6401ULL, 0x040d3017418f2391ULL, 0x2c34bb6090b7d90dULL, 0x16f649228fdfd51fULL,
+	0xbea6818e2b928ef5ULL, 0xe28ccf91cdc11e72ULL, 0x594aaa68e77a36cdULL, 0x313034806c7ffd0fULL,
+	0x8a9d27ac2249bd65ULL, 0x19a3b464018e9512ULL, 0xc26ccff352b37ec7ULL, 0x056f68341d797b21ULL,
+	0x5e79d6757efd2327ULL, 0xfabdbcb6553afe15ULL, 0xd3e7222c6eaf5a60ULL, 0x7046c76d4dae743bULL,
+	0x660be872b18d4a55ULL, 0x19992518574e1496ULL, 0xc103053a302bdcbbULL, 0x3ed8e9800b218e8eULL,
+	0x7b0b9239fa75e03eULL, 0xefe9fb684633c083ULL, 0x98a35fbe391a7793ULL, 0x6065510fe2d0fe34ULL,
+	0x55cb668548abad0cULL, 0xb4584548da87e527ULL, 0x2c43ecea0107c1ddULL, 0x526028809372de35ULL,
+	0x3415c56af9213b1fULL, 0x5bee1a4d017e98dbULL, 0x13f6b105b5cf709bULL, 0x5ff20e3482b29ab6ULL,
+	0x0aa29c75cc2e6c90ULL, 0xfc7d73ca3a70e206ULL, 0x899fc38fc4b5c515ULL, 0x250386b124ffc207ULL,
+	0x54ea28d5ae3d2b56ULL, 0x9913149dd6de60ceULL, 0x16694fc58f06d6c1ULL, 0x46b23975eb018fc7ULL,
+	0x470a6a0fb4b7b4e2ULL, 0x5d92475a8f7253deULL, 0xabeee5b52fbd3adbULL, 0x7fa20801a0806968ULL,
+	0x76f3faf19f7714d2ULL, 0xb3e840c12f4660c3ULL, 0x0fb4cd8df212744eULL, 0x4b065a251d3a2dd2ULL,
+	0x5cebde383d77cd4aULL, 0x6adf39df882c9cb1ULL, 0xa2dd242eb09af759ULL, 0x3147c0e50e5f6422ULL,
+	0x164ca5101d1350dbULL, 0xf8d13479c33fc962ULL, 0xe640ce4d13e5da08ULL, 0x4bdee0c45061f8baULL,
+	0xd7c46dc1a4edb1c9ULL, 0x5514d7b6437fd98aULL, 0x58942f6bb2a1c00bULL, 0x2dffb2ab1d70710eULL,
+	0xccdfcf2fc18b6d68ULL, 0xa8ebcba8b7806167ULL, 0x980697f95e2937e3ULL, 0x02fbba1cd0126e8cULL
+};
+
+static void curve25519_ever64_base(u8 *out, const u8 *priv)
+{
+	u64 swap = 1;
+	int i, j, k;
+	u64 tmp[16 + 32 + 4];
+	u64 *x1 = &tmp[0];
+	u64 *z1 = &tmp[4];
+	u64 *x2 = &tmp[8];
+	u64 *z2 = &tmp[12];
+	u64 *xz1 = &tmp[0];
+	u64 *xz2 = &tmp[8];
+	u64 *a = &tmp[0 + 16];
+	u64 *b = &tmp[4 + 16];
+	u64 *c = &tmp[8 + 16];
+	u64 *ab = &tmp[0 + 16];
+	u64 *abcd = &tmp[0 + 16];
+	u64 *ef = &tmp[16 + 16];
+	u64 *efgh = &tmp[16 + 16];
+	u64 *key = &tmp[0 + 16 + 32];
+
+	memcpy(key, priv, 32);
+	((u8 *)key)[0] &= 248;
+	((u8 *)key)[31] = (((u8 *)key)[31] & 127) | 64;
+
+	x1[0] = 1, x1[1] = x1[2] = x1[3] = 0;
+	z1[0] = 1, z1[1] = z1[2] = z1[3] = 0;
+	z2[0] = 1, z2[1] = z2[2] = z2[3] = 0;
+	memcpy(x2, p_minus_s, sizeof(p_minus_s));
+
+	j = 3;
+	for (i = 0; i < 4; ++i) {
+		while (j < (const int[]){ 64, 64, 64, 63 }[i]) {
+			u64 bit = (key[i] >> j) & 1;
+			k = (64 * i + j - 3);
+			swap = swap ^ bit;
+			cswap2(swap, xz1, xz2);
+			swap = bit;
+			fsub(b, x1, z1);
+			fadd(a, x1, z1);
+			fmul(c, &table_ladder[4 * k], b, ef);
+			fsub(b, a, c);
+			fadd(a, a, c);
+			fsqr2(ab, ab, efgh);
+			fmul2(xz1, xz2, ab, efgh);
+			++j;
+		}
+		j = 0;
+	}
+
+	point_double(xz1, abcd, efgh);
+	point_double(xz1, abcd, efgh);
+	point_double(xz1, abcd, efgh);
+	encode_point(out, xz1);
+
+	memzero_explicit(tmp, sizeof(tmp));
+}
diff --git a/net/wireguard/device.h b/net/wireguard/device.h
new file mode 100644
index 000000000000..b15a8be9d816
--- /dev/null
+++ b/net/wireguard/device.h
@@ -0,0 +1,65 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (C) 2015-2019 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved.
+ */
+
+#ifndef _WG_DEVICE_H
+#define _WG_DEVICE_H
+
+#include "noise.h"
+#include "allowedips.h"
+#include "peerlookup.h"
+#include "cookie.h"
+
+#include <linux/types.h>
+#include <linux/netdevice.h>
+#include <linux/workqueue.h>
+#include <linux/mutex.h>
+#include <linux/net.h>
+#include <linux/ptr_ring.h>
+
+struct wg_device;
+
+struct multicore_worker {
+	void *ptr;
+	struct work_struct work;
+};
+
+struct crypt_queue {
+	struct ptr_ring ring;
+	union {
+		struct {
+			struct multicore_worker __percpu *worker;
+			int last_cpu;
+		};
+		struct work_struct work;
+	};
+};
+
+struct wg_device {
+	struct net_device *dev;
+	struct crypt_queue encrypt_queue, decrypt_queue;
+	struct sock __rcu *sock4, *sock6;
+	struct net *creating_net;
+	struct noise_static_identity static_identity;
+	struct workqueue_struct *handshake_receive_wq, *handshake_send_wq;
+	struct workqueue_struct *packet_crypt_wq;
+	struct sk_buff_head incoming_handshakes;
+	int incoming_handshake_cpu;
+	struct multicore_worker __percpu *incoming_handshakes_worker;
+	struct cookie_checker cookie_checker;
+	struct pubkey_hashtable *peer_hashtable;
+	struct index_hashtable *index_hashtable;
+	struct allowedips peer_allowedips;
+	struct mutex device_update_lock, socket_update_lock;
+	struct list_head device_list, peer_list;
+	unsigned int num_peers, device_update_gen;
+	u32 fwmark;
+	u16 incoming_port;
+	bool have_creating_net_ref;
+};
+
+int wg_device_init(void);
+void wg_device_uninit(void);
+
+#endif /* _WG_DEVICE_H */
diff --git a/net/wireguard/dkms.conf b/net/wireguard/dkms.conf
new file mode 100644
index 000000000000..b094a8766f07
--- /dev/null
+++ b/net/wireguard/dkms.conf
@@ -0,0 +1,9 @@
+PACKAGE_NAME="wireguard"
+PACKAGE_VERSION="0.0.20200215"
+AUTOINSTALL=yes
+
+BUILT_MODULE_NAME="wireguard"
+DEST_MODULE_LOCATION="/kernel/net"
+
+# requires kernel 3.10 - 5.5, inclusive:
+BUILD_EXCLUSIVE_KERNEL="^((5\.[0-5]($|[.-]))|(4\.)|(3\.1[0-9]))"
diff --git a/net/wireguard/socket.c b/net/wireguard/socket.c
new file mode 100644
index 000000000000..b0d6541582d3
--- /dev/null
+++ b/net/wireguard/socket.c
@@ -0,0 +1,437 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2015-2019 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved.
+ */
+
+#include "device.h"
+#include "peer.h"
+#include "socket.h"
+#include "queueing.h"
+#include "messages.h"
+
+#include <linux/ctype.h>
+#include <linux/net.h>
+#include <linux/if_vlan.h>
+#include <linux/if_ether.h>
+#include <linux/inetdevice.h>
+#include <net/udp_tunnel.h>
+#include <net/ipv6.h>
+
+static int send4(struct wg_device *wg, struct sk_buff *skb,
+		 struct endpoint *endpoint, u8 ds, struct dst_cache *cache)
+{
+	struct flowi4 fl = {
+		.saddr = endpoint->src4.s_addr,
+		.daddr = endpoint->addr4.sin_addr.s_addr,
+		.fl4_dport = endpoint->addr4.sin_port,
+		.flowi4_mark = wg->fwmark,
+		.flowi4_proto = IPPROTO_UDP
+	};
+	struct rtable *rt = NULL;
+	struct sock *sock;
+	int ret = 0;
+
+	skb_mark_not_on_list(skb);
+	skb->dev = wg->dev;
+	skb->mark = wg->fwmark;
+
+	rcu_read_lock_bh();
+	sock = rcu_dereference_bh(wg->sock4);
+
+	if (unlikely(!sock)) {
+		ret = -ENONET;
+		goto err;
+	}
+
+	fl.fl4_sport = inet_sk(sock)->inet_sport;
+
+	if (cache)
+		rt = dst_cache_get_ip4(cache, &fl.saddr);
+
+	if (!rt) {
+		security_sk_classify_flow(sock, flowi4_to_flowi(&fl));
+		if (unlikely(!inet_confirm_addr(sock_net(sock), NULL, 0,
+						fl.saddr, RT_SCOPE_HOST))) {
+			endpoint->src4.s_addr = 0;
+			*(__force __be32 *)&endpoint->src_if4 = 0;
+			fl.saddr = 0;
+			if (cache)
+				dst_cache_reset(cache);
+		}
+		rt = ip_route_output_flow(sock_net(sock), &fl, sock);
+		if (unlikely(endpoint->src_if4 && ((IS_ERR(rt) &&
+			     PTR_ERR(rt) == -EINVAL) || (!IS_ERR(rt) &&
+			     rt->dst.dev->ifindex != endpoint->src_if4)))) {
+			endpoint->src4.s_addr = 0;
+			*(__force __be32 *)&endpoint->src_if4 = 0;
+			fl.saddr = 0;
+			if (cache)
+				dst_cache_reset(cache);
+			if (!IS_ERR(rt))
+				ip_rt_put(rt);
+			rt = ip_route_output_flow(sock_net(sock), &fl, sock);
+		}
+		if (unlikely(IS_ERR(rt))) {
+			ret = PTR_ERR(rt);
+			net_dbg_ratelimited("%s: No route to %pISpfsc, error %d\n",
+					    wg->dev->name, &endpoint->addr, ret);
+			goto err;
+		} else if (unlikely(rt->dst.dev == skb->dev)) {
+			ip_rt_put(rt);
+			ret = -ELOOP;
+			net_dbg_ratelimited("%s: Avoiding routing loop to %pISpfsc\n",
+					    wg->dev->name, &endpoint->addr);
+			goto err;
+		}
+		if (cache)
+			dst_cache_set_ip4(cache, &rt->dst, fl.saddr);
+	}
+
+	skb->ignore_df = 1;
+	udp_tunnel_xmit_skb(rt, sock, skb, fl.saddr, fl.daddr, ds,
+			    ip4_dst_hoplimit(&rt->dst), 0, fl.fl4_sport,
+			    fl.fl4_dport, false, false);
+	goto out;
+
+err:
+	kfree_skb(skb);
+out:
+	rcu_read_unlock_bh();
+	return ret;
+}
+
+static int send6(struct wg_device *wg, struct sk_buff *skb,
+		 struct endpoint *endpoint, u8 ds, struct dst_cache *cache)
+{
+#if IS_ENABLED(CONFIG_IPV6)
+	struct flowi6 fl = {
+		.saddr = endpoint->src6,
+		.daddr = endpoint->addr6.sin6_addr,
+		.fl6_dport = endpoint->addr6.sin6_port,
+		.flowi6_mark = wg->fwmark,
+		.flowi6_oif = endpoint->addr6.sin6_scope_id,
+		.flowi6_proto = IPPROTO_UDP
+		/* TODO: addr->sin6_flowinfo */
+	};
+	struct dst_entry *dst = NULL;
+	struct sock *sock;
+	int ret = 0;
+
+	skb_mark_not_on_list(skb);
+	skb->dev = wg->dev;
+	skb->mark = wg->fwmark;
+
+	rcu_read_lock_bh();
+	sock = rcu_dereference_bh(wg->sock6);
+
+	if (unlikely(!sock)) {
+		ret = -ENONET;
+		goto err;
+	}
+
+	fl.fl6_sport = inet_sk(sock)->inet_sport;
+
+	if (cache)
+		dst = dst_cache_get_ip6(cache, &fl.saddr);
+
+	if (!dst) {
+		security_sk_classify_flow(sock, flowi6_to_flowi(&fl));
+		if (unlikely(!ipv6_addr_any(&fl.saddr) &&
+			     !ipv6_chk_addr(sock_net(sock), &fl.saddr, NULL, 0))) {
+			endpoint->src6 = fl.saddr = in6addr_any;
+			if (cache)
+				dst_cache_reset(cache);
+		}
+		dst = ipv6_stub->ipv6_dst_lookup_flow(sock_net(sock), sock, &fl,
+						      NULL);
+		if (unlikely(IS_ERR(dst))) {
+			ret = PTR_ERR(dst);
+			net_dbg_ratelimited("%s: No route to %pISpfsc, error %d\n",
+					    wg->dev->name, &endpoint->addr, ret);
+			goto err;
+		} else if (unlikely(dst->dev == skb->dev)) {
+			dst_release(dst);
+			ret = -ELOOP;
+			net_dbg_ratelimited("%s: Avoiding routing loop to %pISpfsc\n",
+					    wg->dev->name, &endpoint->addr);
+			goto err;
+		}
+		if (cache)
+			dst_cache_set_ip6(cache, dst, &fl.saddr);
+	}
+
+	skb->ignore_df = 1;
+	udp_tunnel6_xmit_skb(dst, sock, skb, skb->dev, &fl.saddr, &fl.daddr, ds,
+			     ip6_dst_hoplimit(dst), 0, fl.fl6_sport,
+			     fl.fl6_dport, false);
+	goto out;
+
+err:
+	kfree_skb(skb);
+out:
+	rcu_read_unlock_bh();
+	return ret;
+#else
+	return -EAFNOSUPPORT;
+#endif
+}
+
+int wg_socket_send_skb_to_peer(struct wg_peer *peer, struct sk_buff *skb, u8 ds)
+{
+	size_t skb_len = skb->len;
+	int ret = -EAFNOSUPPORT;
+
+	read_lock_bh(&peer->endpoint_lock);
+	if (peer->endpoint.addr.sa_family == AF_INET)
+		ret = send4(peer->device, skb, &peer->endpoint, ds,
+			    &peer->endpoint_cache);
+	else if (peer->endpoint.addr.sa_family == AF_INET6)
+		ret = send6(peer->device, skb, &peer->endpoint, ds,
+			    &peer->endpoint_cache);
+	else
+		dev_kfree_skb(skb);
+	if (likely(!ret))
+		peer->tx_bytes += skb_len;
+	read_unlock_bh(&peer->endpoint_lock);
+
+	return ret;
+}
+
+int wg_socket_send_buffer_to_peer(struct wg_peer *peer, void *buffer,
+				  size_t len, u8 ds)
+{
+	struct sk_buff *skb = alloc_skb(len + SKB_HEADER_LEN, GFP_ATOMIC);
+
+	if (unlikely(!skb))
+		return -ENOMEM;
+
+	skb_reserve(skb, SKB_HEADER_LEN);
+	skb_set_inner_network_header(skb, 0);
+	skb_put_data(skb, buffer, len);
+	return wg_socket_send_skb_to_peer(peer, skb, ds);
+}
+
+int wg_socket_send_buffer_as_reply_to_skb(struct wg_device *wg,
+					  struct sk_buff *in_skb, void *buffer,
+					  size_t len)
+{
+	int ret = 0;
+	struct sk_buff *skb;
+	struct endpoint endpoint;
+
+	if (unlikely(!in_skb))
+		return -EINVAL;
+	ret = wg_socket_endpoint_from_skb(&endpoint, in_skb);
+	if (unlikely(ret < 0))
+		return ret;
+
+	skb = alloc_skb(len + SKB_HEADER_LEN, GFP_ATOMIC);
+	if (unlikely(!skb))
+		return -ENOMEM;
+	skb_reserve(skb, SKB_HEADER_LEN);
+	skb_set_inner_network_header(skb, 0);
+	skb_put_data(skb, buffer, len);
+
+	if (endpoint.addr.sa_family == AF_INET)
+		ret = send4(wg, skb, &endpoint, 0, NULL);
+	else if (endpoint.addr.sa_family == AF_INET6)
+		ret = send6(wg, skb, &endpoint, 0, NULL);
+	/* No other possibilities if the endpoint is valid, which it is,
+	 * as we checked above.
+	 */
+
+	return ret;
+}
+
+int wg_socket_endpoint_from_skb(struct endpoint *endpoint,
+				const struct sk_buff *skb)
+{
+	memset(endpoint, 0, sizeof(*endpoint));
+	if (skb->protocol == htons(ETH_P_IP)) {
+		endpoint->addr4.sin_family = AF_INET;
+		endpoint->addr4.sin_port = udp_hdr(skb)->source;
+		endpoint->addr4.sin_addr.s_addr = ip_hdr(skb)->saddr;
+		endpoint->src4.s_addr = ip_hdr(skb)->daddr;
+		endpoint->src_if4 = skb->skb_iif;
+	} else if (skb->protocol == htons(ETH_P_IPV6)) {
+		endpoint->addr6.sin6_family = AF_INET6;
+		endpoint->addr6.sin6_port = udp_hdr(skb)->source;
+		endpoint->addr6.sin6_addr = ipv6_hdr(skb)->saddr;
+		endpoint->addr6.sin6_scope_id = ipv6_iface_scope_id(
+			&ipv6_hdr(skb)->saddr, skb->skb_iif);
+		endpoint->src6 = ipv6_hdr(skb)->daddr;
+	} else {
+		return -EINVAL;
+	}
+	return 0;
+}
+
+static bool endpoint_eq(const struct endpoint *a, const struct endpoint *b)
+{
+	return (a->addr.sa_family == AF_INET && b->addr.sa_family == AF_INET &&
+		a->addr4.sin_port == b->addr4.sin_port &&
+		a->addr4.sin_addr.s_addr == b->addr4.sin_addr.s_addr &&
+		a->src4.s_addr == b->src4.s_addr && a->src_if4 == b->src_if4) ||
+	       (a->addr.sa_family == AF_INET6 &&
+		b->addr.sa_family == AF_INET6 &&
+		a->addr6.sin6_port == b->addr6.sin6_port &&
+		ipv6_addr_equal(&a->addr6.sin6_addr, &b->addr6.sin6_addr) &&
+		a->addr6.sin6_scope_id == b->addr6.sin6_scope_id &&
+		ipv6_addr_equal(&a->src6, &b->src6)) ||
+	       unlikely(!a->addr.sa_family && !b->addr.sa_family);
+}
+
+void wg_socket_set_peer_endpoint(struct wg_peer *peer,
+				 const struct endpoint *endpoint)
+{
+	/* First we check unlocked, in order to optimize, since it's pretty rare
+	 * that an endpoint will change. If we happen to be mid-write, and two
+	 * CPUs wind up writing the same thing or something slightly different,
+	 * it doesn't really matter much either.
+	 */
+	if (endpoint_eq(endpoint, &peer->endpoint))
+		return;
+	write_lock_bh(&peer->endpoint_lock);
+	if (endpoint->addr.sa_family == AF_INET) {
+		peer->endpoint.addr4 = endpoint->addr4;
+		peer->endpoint.src4 = endpoint->src4;
+		peer->endpoint.src_if4 = endpoint->src_if4;
+	} else if (endpoint->addr.sa_family == AF_INET6) {
+		peer->endpoint.addr6 = endpoint->addr6;
+		peer->endpoint.src6 = endpoint->src6;
+	} else {
+		goto out;
+	}
+	dst_cache_reset(&peer->endpoint_cache);
+out:
+	write_unlock_bh(&peer->endpoint_lock);
+}
+
+void wg_socket_set_peer_endpoint_from_skb(struct wg_peer *peer,
+					  const struct sk_buff *skb)
+{
+	struct endpoint endpoint;
+
+	if (!wg_socket_endpoint_from_skb(&endpoint, skb))
+		wg_socket_set_peer_endpoint(peer, &endpoint);
+}
+
+void wg_socket_clear_peer_endpoint_src(struct wg_peer *peer)
+{
+	write_lock_bh(&peer->endpoint_lock);
+	memset(&peer->endpoint.src6, 0, sizeof(peer->endpoint.src6));
+	dst_cache_reset(&peer->endpoint_cache);
+	write_unlock_bh(&peer->endpoint_lock);
+}
+
+static int wg_receive(struct sock *sk, struct sk_buff *skb)
+{
+	struct wg_device *wg;
+
+	if (unlikely(!sk))
+		goto err;
+	wg = sk->sk_user_data;
+	if (unlikely(!wg))
+		goto err;
+	skb_mark_not_on_list(skb);
+	wg_packet_receive(wg, skb);
+	return 0;
+
+err:
+	kfree_skb(skb);
+	return 0;
+}
+
+static void sock_free(struct sock *sock)
+{
+	if (unlikely(!sock))
+		return;
+	sk_clear_memalloc(sock);
+	udp_tunnel_sock_release(sock->sk_socket);
+}
+
+static void set_sock_opts(struct socket *sock)
+{
+	sock->sk->sk_allocation = GFP_ATOMIC;
+	sock->sk->sk_sndbuf = INT_MAX;
+	sk_set_memalloc(sock->sk);
+}
+
+int wg_socket_init(struct wg_device *wg, u16 port)
+{
+	int ret;
+	struct udp_tunnel_sock_cfg cfg = {
+		.sk_user_data = wg,
+		.encap_type = 1,
+		.encap_rcv = wg_receive
+	};
+	struct socket *new4 = NULL, *new6 = NULL;
+	struct udp_port_cfg port4 = {
+		.family = AF_INET,
+		.local_ip.s_addr = htonl(INADDR_ANY),
+		.local_udp_port = htons(port),
+		.use_udp_checksums = true
+	};
+#if IS_ENABLED(CONFIG_IPV6)
+	int retries = 0;
+	struct udp_port_cfg port6 = {
+		.family = AF_INET6,
+		.local_ip6 = IN6ADDR_ANY_INIT,
+		.use_udp6_tx_checksums = true,
+		.use_udp6_rx_checksums = true,
+		.ipv6_v6only = true
+	};
+#endif
+
+#if IS_ENABLED(CONFIG_IPV6)
+retry:
+#endif
+
+	ret = udp_sock_create(wg->creating_net, &port4, &new4);
+	if (ret < 0) {
+		pr_err("%s: Could not create IPv4 socket\n", wg->dev->name);
+		return ret;
+	}
+	set_sock_opts(new4);
+	setup_udp_tunnel_sock(wg->creating_net, new4, &cfg);
+
+#if IS_ENABLED(CONFIG_IPV6)
+	if (ipv6_mod_enabled()) {
+		port6.local_udp_port = inet_sk(new4->sk)->inet_sport;
+		ret = udp_sock_create(wg->creating_net, &port6, &new6);
+		if (ret < 0) {
+			udp_tunnel_sock_release(new4);
+			if (ret == -EADDRINUSE && !port && retries++ < 100)
+				goto retry;
+			pr_err("%s: Could not create IPv6 socket\n",
+			       wg->dev->name);
+			return ret;
+		}
+		set_sock_opts(new6);
+		setup_udp_tunnel_sock(wg->creating_net, new6, &cfg);
+	}
+#endif
+
+	wg_socket_reinit(wg, new4->sk, new6 ? new6->sk : NULL);
+	return 0;
+}
+
+void wg_socket_reinit(struct wg_device *wg, struct sock *new4,
+		      struct sock *new6)
+{
+	struct sock *old4, *old6;
+
+	mutex_lock(&wg->socket_update_lock);
+	old4 = rcu_dereference_protected(wg->sock4,
+				lockdep_is_held(&wg->socket_update_lock));
+	old6 = rcu_dereference_protected(wg->sock6,
+				lockdep_is_held(&wg->socket_update_lock));
+	rcu_assign_pointer(wg->sock4, new4);
+	rcu_assign_pointer(wg->sock6, new6);
+	if (new4)
+		wg->incoming_port = ntohs(inet_sk(new4)->inet_sport);
+	mutex_unlock(&wg->socket_update_lock);
+	synchronize_rcu();
+	sock_free(old4);
+	sock_free(old6);
+}
diff --git a/net/wireguard/tests/netns.sh b/net/wireguard/tests/netns.sh
new file mode 100755
index 000000000000..ba5f83a91c68
--- /dev/null
+++ b/net/wireguard/tests/netns.sh
@@ -0,0 +1,553 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0
+#
+# Copyright (C) 2015-2019 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved.
+#
+# This script tests the below topology:
+#
+# ┌─────────────────────┐   ┌──────────────────────────────────┐   ┌─────────────────────┐
+# │   $ns1 namespace    │   │          $ns0 namespace          │   │   $ns2 namespace    │
+# │                     │   │                                  │   │                     │
+# │┌────────┐           │   │            ┌────────┐            │   │           ┌────────┐│
+# ││  wg0   │───────────┼───┼────────────│   lo   │────────────┼───┼───────────│  wg0   ││
+# │├────────┴──────────┐│   │    ┌───────┴────────┴────────┐   │   │┌──────────┴────────┤│
+# ││192.168.241.1/24   ││   │    │(ns1)         (ns2)      │   │   ││192.168.241.2/24   ││
+# ││fd00::1/24         ││   │    │127.0.0.1:1   127.0.0.1:2│   │   ││fd00::2/24         ││
+# │└───────────────────┘│   │    │[::]:1        [::]:2     │   │   │└───────────────────┘│
+# └─────────────────────┘   │    └─────────────────────────┘   │   └─────────────────────┘
+#                           └──────────────────────────────────┘
+#
+# After the topology is prepared we run a series of TCP/UDP iperf3 tests between the
+# wireguard peers in $ns1 and $ns2. Note that $ns0 is the endpoint for the wg0
+# interfaces in $ns1 and $ns2. See https://www.wireguard.com/netns/ for further
+# details on how this is accomplished.
+set -e
+
+exec 3>&1
+export LANG=C
+export WG_HIDE_KEYS=never
+netns0="wg-test-$$-0"
+netns1="wg-test-$$-1"
+netns2="wg-test-$$-2"
+pretty() { echo -e "\x1b[32m\x1b[1m[+] ${1:+NS$1: }${2}\x1b[0m" >&3; }
+pp() { pretty "" "$*"; "$@"; }
+maybe_exec() { if [[ $BASHPID -eq $$ ]]; then "$@"; else exec "$@"; fi; }
+n0() { pretty 0 "$*"; maybe_exec ip netns exec $netns0 "$@"; }
+n1() { pretty 1 "$*"; maybe_exec ip netns exec $netns1 "$@"; }
+n2() { pretty 2 "$*"; maybe_exec ip netns exec $netns2 "$@"; }
+ip0() { pretty 0 "ip $*"; ip -n $netns0 "$@"; }
+ip1() { pretty 1 "ip $*"; ip -n $netns1 "$@"; }
+ip2() { pretty 2 "ip $*"; ip -n $netns2 "$@"; }
+sleep() { read -t "$1" -N 1 || true; }
+waitiperf() { pretty "${1//*-}" "wait for iperf:5201 pid $2"; while [[ $(ss -N "$1" -tlpH 'sport = 5201') != *\"iperf3\",pid=$2,fd=* ]]; do sleep 0.1; done; }
+waitncatudp() { pretty "${1//*-}" "wait for udp:1111 pid $2"; while [[ $(ss -N "$1" -ulpH 'sport = 1111') != *\"ncat\",pid=$2,fd=* ]]; do sleep 0.1; done; }
+waitiface() { pretty "${1//*-}" "wait for $2 to come up"; ip netns exec "$1" bash -c "while [[ \$(< \"/sys/class/net/$2/operstate\") != up ]]; do read -t .1 -N 0 || true; done;"; }
+
+cleanup() {
+	set +e
+	exec 2>/dev/null
+	printf "$orig_message_cost" > /proc/sys/net/core/message_cost
+	ip0 link del dev wg0
+	ip1 link del dev wg0
+	ip2 link del dev wg0
+	local to_kill="$(ip netns pids $netns0) $(ip netns pids $netns1) $(ip netns pids $netns2)"
+	[[ -n $to_kill ]] && kill $to_kill
+	pp ip netns del $netns1
+	pp ip netns del $netns2
+	pp ip netns del $netns0
+	exit
+}
+
+orig_message_cost="$(< /proc/sys/net/core/message_cost)"
+trap cleanup EXIT
+printf 0 > /proc/sys/net/core/message_cost
+
+ip netns del $netns0 2>/dev/null || true
+ip netns del $netns1 2>/dev/null || true
+ip netns del $netns2 2>/dev/null || true
+pp ip netns add $netns0
+pp ip netns add $netns1
+pp ip netns add $netns2
+ip0 link set up dev lo
+
+ip0 link add dev wg0 type wireguard
+ip0 link set wg0 netns $netns1
+ip0 link add dev wg0 type wireguard
+ip0 link set wg0 netns $netns2
+key1="$(pp wg genkey)"
+key2="$(pp wg genkey)"
+key3="$(pp wg genkey)"
+pub1="$(pp wg pubkey <<<"$key1")"
+pub2="$(pp wg pubkey <<<"$key2")"
+pub3="$(pp wg pubkey <<<"$key3")"
+psk="$(pp wg genpsk)"
+[[ -n $key1 && -n $key2 && -n $psk ]]
+
+configure_peers() {
+	ip1 addr add 192.168.241.1/24 dev wg0
+	ip1 addr add fd00::1/24 dev wg0
+
+	ip2 addr add 192.168.241.2/24 dev wg0
+	ip2 addr add fd00::2/24 dev wg0
+
+	n1 wg set wg0 \
+		private-key <(echo "$key1") \
+		listen-port 1 \
+		peer "$pub2" \
+			preshared-key <(echo "$psk") \
+			allowed-ips 192.168.241.2/32,fd00::2/128
+	n2 wg set wg0 \
+		private-key <(echo "$key2") \
+		listen-port 2 \
+		peer "$pub1" \
+			preshared-key <(echo "$psk") \
+			allowed-ips 192.168.241.1/32,fd00::1/128
+
+	ip1 link set up dev wg0
+	ip2 link set up dev wg0
+}
+configure_peers
+
+tests() {
+	# Ping over IPv4
+	n2 ping -c 10 -f -W 1 192.168.241.1
+	n1 ping -c 10 -f -W 1 192.168.241.2
+
+	# Ping over IPv6
+	n2 ping6 -c 10 -f -W 1 fd00::1
+	n1 ping6 -c 10 -f -W 1 fd00::2
+
+	# TCP over IPv4
+	n2 iperf3 -s -1 -B 192.168.241.2 &
+	waitiperf $netns2 $!
+	n1 iperf3 -Z -t 3 -c 192.168.241.2
+
+	# TCP over IPv6
+	n1 iperf3 -s -1 -B fd00::1 &
+	waitiperf $netns1 $!
+	n2 iperf3 -Z -t 3 -c fd00::1
+
+	# UDP over IPv4
+	n1 iperf3 -s -1 -B 192.168.241.1 &
+	waitiperf $netns1 $!
+	n2 iperf3 -Z -t 3 -b 0 -u -c 192.168.241.1
+
+	# UDP over IPv6
+	n2 iperf3 -s -1 -B fd00::2 &
+	waitiperf $netns2 $!
+	n1 iperf3 -Z -t 3 -b 0 -u -c fd00::2
+}
+
+[[ $(ip1 link show dev wg0) =~ mtu\ ([0-9]+) ]] && orig_mtu="${BASH_REMATCH[1]}"
+big_mtu=$(( 34816 - 1500 + $orig_mtu ))
+
+# Test using IPv4 as outer transport
+n1 wg set wg0 peer "$pub2" endpoint 127.0.0.1:2
+n2 wg set wg0 peer "$pub1" endpoint 127.0.0.1:1
+# Before calling tests, we first make sure that the stats counters and timestamper are working
+n2 ping -c 10 -f -W 1 192.168.241.1
+{ read _; read _; read _; read rx_bytes _; read _; read tx_bytes _; } < <(ip2 -stats link show dev wg0)
+(( rx_bytes == 1372 && (tx_bytes == 1428 || tx_bytes == 1460) ))
+{ read _; read _; read _; read rx_bytes _; read _; read tx_bytes _; } < <(ip1 -stats link show dev wg0)
+(( tx_bytes == 1372 && (rx_bytes == 1428 || rx_bytes == 1460) ))
+read _ rx_bytes tx_bytes < <(n2 wg show wg0 transfer)
+(( rx_bytes == 1372 && (tx_bytes == 1428 || tx_bytes == 1460) ))
+read _ rx_bytes tx_bytes < <(n1 wg show wg0 transfer)
+(( tx_bytes == 1372 && (rx_bytes == 1428 || rx_bytes == 1460) ))
+read _ timestamp < <(n1 wg show wg0 latest-handshakes)
+(( timestamp != 0 ))
+
+tests
+ip1 link set wg0 mtu $big_mtu
+ip2 link set wg0 mtu $big_mtu
+tests
+
+ip1 link set wg0 mtu $orig_mtu
+ip2 link set wg0 mtu $orig_mtu
+
+# Test using IPv6 as outer transport
+n1 wg set wg0 peer "$pub2" endpoint [::1]:2
+n2 wg set wg0 peer "$pub1" endpoint [::1]:1
+tests
+ip1 link set wg0 mtu $big_mtu
+ip2 link set wg0 mtu $big_mtu
+tests
+
+# Test that route MTUs work with the padding
+ip1 link set wg0 mtu 1300
+ip2 link set wg0 mtu 1300
+n1 wg set wg0 peer "$pub2" endpoint 127.0.0.1:2
+n2 wg set wg0 peer "$pub1" endpoint 127.0.0.1:1
+n0 iptables -A INPUT -m length --length 1360 -j DROP
+n1 ip route add 192.168.241.2/32 dev wg0 mtu 1299
+n2 ip route add 192.168.241.1/32 dev wg0 mtu 1299
+n2 ping -c 1 -W 1 -s 1269 192.168.241.1
+n2 ip route delete 192.168.241.1/32 dev wg0 mtu 1299
+n1 ip route delete 192.168.241.2/32 dev wg0 mtu 1299
+n0 iptables -F INPUT
+
+ip1 link set wg0 mtu $orig_mtu
+ip2 link set wg0 mtu $orig_mtu
+
+# Test using IPv4 that roaming works
+ip0 -4 addr del 127.0.0.1/8 dev lo
+ip0 -4 addr add 127.212.121.99/8 dev lo
+n1 wg set wg0 listen-port 9999
+n1 wg set wg0 peer "$pub2" endpoint 127.0.0.1:2
+n1 ping6 -W 1 -c 1 fd00::2
+[[ $(n2 wg show wg0 endpoints) == "$pub1	127.212.121.99:9999" ]]
+
+# Test using IPv6 that roaming works
+n1 wg set wg0 listen-port 9998
+n1 wg set wg0 peer "$pub2" endpoint [::1]:2
+n1 ping -W 1 -c 1 192.168.241.2
+[[ $(n2 wg show wg0 endpoints) == "$pub1	[::1]:9998" ]]
+
+# Test that crypto-RP filter works
+n1 wg set wg0 peer "$pub2" allowed-ips 192.168.241.0/24
+exec 4< <(n1 ncat -l -u -p 1111)
+ncat_pid=$!
+waitncatudp $netns1 $ncat_pid
+n2 ncat -u 192.168.241.1 1111 <<<"X"
+read -r -N 1 -t 1 out <&4 && [[ $out == "X" ]]
+kill $ncat_pid
+more_specific_key="$(pp wg genkey | pp wg pubkey)"
+n1 wg set wg0 peer "$more_specific_key" allowed-ips 192.168.241.2/32
+n2 wg set wg0 listen-port 9997
+exec 4< <(n1 ncat -l -u -p 1111)
+ncat_pid=$!
+waitncatudp $netns1 $ncat_pid
+n2 ncat -u 192.168.241.1 1111 <<<"X"
+! read -r -N 1 -t 1 out <&4 || false
+kill $ncat_pid
+n1 wg set wg0 peer "$more_specific_key" remove
+[[ $(n1 wg show wg0 endpoints) == "$pub2	[::1]:9997" ]]
+
+# Test that we can change private keys keys and immediately handshake
+n1 wg set wg0 private-key <(echo "$key1") peer "$pub2" preshared-key <(echo "$psk") allowed-ips 192.168.241.2/32 endpoint 127.0.0.1:2
+n2 wg set wg0 private-key <(echo "$key2") listen-port 2 peer "$pub1" preshared-key <(echo "$psk") allowed-ips 192.168.241.1/32
+n1 ping -W 1 -c 1 192.168.241.2
+n1 wg set wg0 private-key <(echo "$key3")
+n2 wg set wg0 peer "$pub3" preshared-key <(echo "$psk") allowed-ips 192.168.241.1/32 peer "$pub1" remove
+n1 ping -W 1 -c 1 192.168.241.2
+
+ip1 link del wg0
+ip2 link del wg0
+
+# Test using NAT. We now change the topology to this:
+# ┌────────────────────────────────────────┐    ┌────────────────────────────────────────────────┐     ┌────────────────────────────────────────┐
+# │             $ns1 namespace             │    │                 $ns0 namespace                 │     │             $ns2 namespace             │
+# │                                        │    │                                                │     │                                        │
+# │  ┌─────┐             ┌─────┐           │    │    ┌──────┐              ┌──────┐              │     │  ┌─────┐            ┌─────┐            │
+# │  │ wg0 │─────────────│vethc│───────────┼────┼────│vethrc│              │vethrs│──────────────┼─────┼──│veths│────────────│ wg0 │            │
+# │  ├─────┴──────────┐  ├─────┴──────────┐│    │    ├──────┴─────────┐    ├──────┴────────────┐ │     │  ├─────┴──────────┐ ├─────┴──────────┐ │
+# │  │192.168.241.1/24│  │192.168.1.100/24││    │    │192.168.1.1/24  │    │10.0.0.1/24        │ │     │  │10.0.0.100/24   │ │192.168.241.2/24│ │
+# │  │fd00::1/24      │  │                ││    │    │                │    │SNAT:192.168.1.0/24│ │     │  │                │ │fd00::2/24      │ │
+# │  └────────────────┘  └────────────────┘│    │    └────────────────┘    └───────────────────┘ │     │  └────────────────┘ └────────────────┘ │
+# └────────────────────────────────────────┘    └────────────────────────────────────────────────┘     └────────────────────────────────────────┘
+
+ip1 link add dev wg0 type wireguard
+ip2 link add dev wg0 type wireguard
+configure_peers
+
+ip0 link add vethrc type veth peer name vethc
+ip0 link add vethrs type veth peer name veths
+ip0 link set vethc netns $netns1
+ip0 link set veths netns $netns2
+ip0 link set vethrc up
+ip0 link set vethrs up
+ip0 addr add 192.168.1.1/24 dev vethrc
+ip0 addr add 10.0.0.1/24 dev vethrs
+ip1 addr add 192.168.1.100/24 dev vethc
+ip1 link set vethc up
+ip1 route add default via 192.168.1.1
+ip2 addr add 10.0.0.100/24 dev veths
+ip2 link set veths up
+waitiface $netns0 vethrc
+waitiface $netns0 vethrs
+waitiface $netns1 vethc
+waitiface $netns2 veths
+
+n0 bash -c 'printf 1 > /proc/sys/net/ipv4/ip_forward'
+n0 bash -c 'printf 2 > /proc/sys/net/netfilter/nf_conntrack_udp_timeout'
+n0 bash -c 'printf 2 > /proc/sys/net/netfilter/nf_conntrack_udp_timeout_stream'
+n0 iptables -t nat -A POSTROUTING -s 192.168.1.0/24 -d 10.0.0.0/24 -j SNAT --to 10.0.0.1
+
+n1 wg set wg0 peer "$pub2" endpoint 10.0.0.100:2 persistent-keepalive 1
+n1 ping -W 1 -c 1 192.168.241.2
+n2 ping -W 1 -c 1 192.168.241.1
+[[ $(n2 wg show wg0 endpoints) == "$pub1	10.0.0.1:1" ]]
+# Demonstrate n2 can still send packets to n1, since persistent-keepalive will prevent connection tracking entry from expiring (to see entries: `n0 conntrack -L`).
+pp sleep 3
+n2 ping -W 1 -c 1 192.168.241.1
+n1 wg set wg0 peer "$pub2" persistent-keepalive 0
+
+# Do a wg-quick(8)-style policy routing for the default route, making sure vethc has a v6 address to tease out bugs.
+ip1 -6 addr add fc00::9/96 dev vethc
+ip1 -6 route add default via fc00::1
+ip2 -4 addr add 192.168.99.7/32 dev wg0
+ip2 -6 addr add abab::1111/128 dev wg0
+n1 wg set wg0 fwmark 51820 peer "$pub2" allowed-ips 192.168.99.7,abab::1111
+ip1 -6 route add default dev wg0 table 51820
+ip1 -6 rule add not fwmark 51820 table 51820
+ip1 -6 rule add table main suppress_prefixlength 0
+ip1 -4 route add default dev wg0 table 51820
+ip1 -4 rule add not fwmark 51820 table 51820
+ip1 -4 rule add table main suppress_prefixlength 0
+# suppress_prefixlength only got added in 3.12, and we want to support 3.10+.
+if [[ $(ip1 -4 rule show all) == *suppress_prefixlength* ]]; then
+	# Flood the pings instead of sending just one, to trigger routing table reference counting bugs.
+	n1 ping -W 1 -c 100 -f 192.168.99.7
+	n1 ping -W 1 -c 100 -f abab::1111
+fi
+
+# Have ns2 NAT into wg0 packets from ns0, but return an icmp error along the right route.
+n2 iptables -t nat -A POSTROUTING -s 10.0.0.0/24 -d 192.168.241.0/24 -j SNAT --to 192.168.241.2
+n0 iptables -t filter -A INPUT \! -s 10.0.0.0/24 -i vethrs -j DROP # Manual rpfilter just to be explicit.
+n2 bash -c 'printf 1 > /proc/sys/net/ipv4/ip_forward'
+ip0 -4 route add 192.168.241.1 via 10.0.0.100
+n2 wg set wg0 peer "$pub1" remove
+[[ $(! n0 ping -W 1 -c 1 192.168.241.1 || false) == *"From 10.0.0.100 icmp_seq=1 Destination Host Unreachable"* ]]
+
+n0 iptables -t nat -F
+n0 iptables -t filter -F
+n2 iptables -t nat -F
+ip0 link del vethrc
+ip0 link del vethrs
+ip1 link del wg0
+ip2 link del wg0
+
+# Test that saddr routing is sticky but not too sticky, changing to this topology:
+# ┌────────────────────────────────────────┐    ┌────────────────────────────────────────┐
+# │             $ns1 namespace             │    │             $ns2 namespace             │
+# │                                        │    │                                        │
+# │  ┌─────┐             ┌─────┐           │    │  ┌─────┐            ┌─────┐            │
+# │  │ wg0 │─────────────│veth1│───────────┼────┼──│veth2│────────────│ wg0 │            │
+# │  ├─────┴──────────┐  ├─────┴──────────┐│    │  ├─────┴──────────┐ ├─────┴──────────┐ │
+# │  │192.168.241.1/24│  │10.0.0.1/24     ││    │  │10.0.0.2/24     │ │192.168.241.2/24│ │
+# │  │fd00::1/24      │  │fd00:aa::1/96   ││    │  │fd00:aa::2/96   │ │fd00::2/24      │ │
+# │  └────────────────┘  └────────────────┘│    │  └────────────────┘ └────────────────┘ │
+# └────────────────────────────────────────┘    └────────────────────────────────────────┘
+
+ip1 link add dev wg0 type wireguard
+ip2 link add dev wg0 type wireguard
+configure_peers
+ip1 link add veth1 type veth peer name veth2
+ip1 link set veth2 netns $netns2
+n1 bash -c 'printf 0 > /proc/sys/net/ipv6/conf/all/accept_dad'
+n2 bash -c 'printf 0 > /proc/sys/net/ipv6/conf/all/accept_dad'
+n1 bash -c 'printf 0 > /proc/sys/net/ipv6/conf/veth1/accept_dad'
+n2 bash -c 'printf 0 > /proc/sys/net/ipv6/conf/veth2/accept_dad'
+n1 bash -c 'printf 1 > /proc/sys/net/ipv4/conf/veth1/promote_secondaries'
+
+# First we check that we aren't overly sticky and can fall over to new IPs when old ones are removed
+ip1 addr add 10.0.0.1/24 dev veth1
+ip1 addr add fd00:aa::1/96 dev veth1
+ip2 addr add 10.0.0.2/24 dev veth2
+ip2 addr add fd00:aa::2/96 dev veth2
+ip1 link set veth1 up
+ip2 link set veth2 up
+waitiface $netns1 veth1
+waitiface $netns2 veth2
+n1 wg set wg0 peer "$pub2" endpoint 10.0.0.2:2
+n1 ping -W 1 -c 1 192.168.241.2
+ip1 addr add 10.0.0.10/24 dev veth1
+ip1 addr del 10.0.0.1/24 dev veth1
+n1 ping -W 1 -c 1 192.168.241.2
+n1 wg set wg0 peer "$pub2" endpoint [fd00:aa::2]:2
+n1 ping -W 1 -c 1 192.168.241.2
+ip1 addr add fd00:aa::10/96 dev veth1
+ip1 addr del fd00:aa::1/96 dev veth1
+n1 ping -W 1 -c 1 192.168.241.2
+
+# Now we show that we can successfully do reply to sender routing
+ip1 link set veth1 down
+ip2 link set veth2 down
+ip1 addr flush dev veth1
+ip2 addr flush dev veth2
+ip1 addr add 10.0.0.1/24 dev veth1
+ip1 addr add 10.0.0.2/24 dev veth1
+ip1 addr add fd00:aa::1/96 dev veth1
+ip1 addr add fd00:aa::2/96 dev veth1
+ip2 addr add 10.0.0.3/24 dev veth2
+ip2 addr add fd00:aa::3/96 dev veth2
+ip1 link set veth1 up
+ip2 link set veth2 up
+waitiface $netns1 veth1
+waitiface $netns2 veth2
+n2 wg set wg0 peer "$pub1" endpoint 10.0.0.1:1
+n2 ping -W 1 -c 1 192.168.241.1
+[[ $(n2 wg show wg0 endpoints) == "$pub1	10.0.0.1:1" ]]
+n2 wg set wg0 peer "$pub1" endpoint [fd00:aa::1]:1
+n2 ping -W 1 -c 1 192.168.241.1
+[[ $(n2 wg show wg0 endpoints) == "$pub1	[fd00:aa::1]:1" ]]
+n2 wg set wg0 peer "$pub1" endpoint 10.0.0.2:1
+n2 ping -W 1 -c 1 192.168.241.1
+[[ $(n2 wg show wg0 endpoints) == "$pub1	10.0.0.2:1" ]]
+n2 wg set wg0 peer "$pub1" endpoint [fd00:aa::2]:1
+n2 ping -W 1 -c 1 192.168.241.1
+[[ $(n2 wg show wg0 endpoints) == "$pub1	[fd00:aa::2]:1" ]]
+
+# What happens if the inbound destination address belongs to a different interface as the default route?
+ip1 link add dummy0 type dummy
+ip1 addr add 10.50.0.1/24 dev dummy0
+ip1 link set dummy0 up
+ip2 route add 10.50.0.0/24 dev veth2
+n2 wg set wg0 peer "$pub1" endpoint 10.50.0.1:1
+n2 ping -W 1 -c 1 192.168.241.1
+[[ $(n2 wg show wg0 endpoints) == "$pub1	10.50.0.1:1" ]]
+
+ip1 link del dummy0
+ip1 addr flush dev veth1
+ip2 addr flush dev veth2
+ip1 route flush dev veth1
+ip2 route flush dev veth2
+
+# Now we see what happens if another interface route takes precedence over an ongoing one
+ip1 link add veth3 type veth peer name veth4
+ip1 link set veth4 netns $netns2
+ip1 addr add 10.0.0.1/24 dev veth1
+ip2 addr add 10.0.0.2/24 dev veth2
+ip1 addr add 10.0.0.3/24 dev veth3
+ip1 link set veth1 up
+ip2 link set veth2 up
+ip1 link set veth3 up
+ip2 link set veth4 up
+waitiface $netns1 veth1
+waitiface $netns2 veth2
+waitiface $netns1 veth3
+waitiface $netns2 veth4
+ip1 route flush dev veth1
+ip1 route flush dev veth3
+ip1 route add 10.0.0.0/24 dev veth1 src 10.0.0.1 metric 2
+n1 wg set wg0 peer "$pub2" endpoint 10.0.0.2:2
+n1 ping -W 1 -c 1 192.168.241.2
+[[ $(n2 wg show wg0 endpoints) == "$pub1	10.0.0.1:1" ]]
+ip1 route add 10.0.0.0/24 dev veth3 src 10.0.0.3 metric 1
+n1 bash -c 'printf 0 > /proc/sys/net/ipv4/conf/veth1/rp_filter'
+n2 bash -c 'printf 0 > /proc/sys/net/ipv4/conf/veth4/rp_filter'
+n1 bash -c 'printf 0 > /proc/sys/net/ipv4/conf/all/rp_filter'
+n2 bash -c 'printf 0 > /proc/sys/net/ipv4/conf/all/rp_filter'
+n1 ping -W 1 -c 1 192.168.241.2
+[[ $(n2 wg show wg0 endpoints) == "$pub1	10.0.0.3:1" ]]
+
+ip1 link del veth1
+ip1 link del veth3
+ip1 link del wg0
+ip2 link del wg0
+
+# We test that Netlink/IPC is working properly by doing things that usually cause split responses
+ip0 link add dev wg0 type wireguard
+config=( "[Interface]" "PrivateKey=$(wg genkey)" "[Peer]" "PublicKey=$(wg genkey)" )
+for a in {1..255}; do
+	for b in {0..255}; do
+		config+=( "AllowedIPs=$a.$b.0.0/16,$a::$b/128" )
+	done
+done
+n0 wg setconf wg0 <(printf '%s\n' "${config[@]}")
+i=0
+for ip in $(n0 wg show wg0 allowed-ips); do
+	((++i))
+done
+((i == 255*256*2+1))
+ip0 link del wg0
+ip0 link add dev wg0 type wireguard
+config=( "[Interface]" "PrivateKey=$(wg genkey)" )
+for a in {1..40}; do
+	config+=( "[Peer]" "PublicKey=$(wg genkey)" )
+	for b in {1..52}; do
+		config+=( "AllowedIPs=$a.$b.0.0/16" )
+	done
+done
+n0 wg setconf wg0 <(printf '%s\n' "${config[@]}")
+i=0
+while read -r line; do
+	j=0
+	for ip in $line; do
+		((++j))
+	done
+	((j == 53))
+	((++i))
+done < <(n0 wg show wg0 allowed-ips)
+((i == 40))
+ip0 link del wg0
+ip0 link add wg0 type wireguard
+config=( )
+for i in {1..29}; do
+	config+=( "[Peer]" "PublicKey=$(wg genkey)" )
+done
+config+=( "[Peer]" "PublicKey=$(wg genkey)" "AllowedIPs=255.2.3.4/32,abcd::255/128" )
+n0 wg setconf wg0 <(printf '%s\n' "${config[@]}")
+n0 wg showconf wg0 > /dev/null
+ip0 link del wg0
+
+allowedips=( )
+for i in {1..197}; do
+        allowedips+=( abcd::$i )
+done
+saved_ifs="$IFS"
+IFS=,
+allowedips="${allowedips[*]}"
+IFS="$saved_ifs"
+ip0 link add wg0 type wireguard
+n0 wg set wg0 peer "$pub1"
+n0 wg set wg0 peer "$pub2" allowed-ips "$allowedips"
+{
+	read -r pub allowedips
+	[[ $pub == "$pub1" && $allowedips == "(none)" ]]
+	read -r pub allowedips
+	[[ $pub == "$pub2" ]]
+	i=0
+	for _ in $allowedips; do
+		((++i))
+	done
+	((i == 197))
+} < <(n0 wg show wg0 allowed-ips)
+ip0 link del wg0
+
+! n0 wg show doesnotexist || false
+
+ip0 link add wg0 type wireguard
+n0 wg set wg0 private-key <(echo "$key1") peer "$pub2" preshared-key <(echo "$psk")
+[[ $(n0 wg show wg0 private-key) == "$key1" ]]
+[[ $(n0 wg show wg0 preshared-keys) == "$pub2	$psk" ]]
+n0 wg set wg0 private-key /dev/null peer "$pub2" preshared-key /dev/null
+[[ $(n0 wg show wg0 private-key) == "(none)" ]]
+[[ $(n0 wg show wg0 preshared-keys) == "$pub2	(none)" ]]
+n0 wg set wg0 peer "$pub2"
+n0 wg set wg0 private-key <(echo "$key2")
+[[ $(n0 wg show wg0 public-key) == "$pub2" ]]
+[[ -z $(n0 wg show wg0 peers) ]]
+n0 wg set wg0 peer "$pub2"
+[[ -z $(n0 wg show wg0 peers) ]]
+n0 wg set wg0 private-key <(echo "$key1")
+n0 wg set wg0 peer "$pub2"
+[[ $(n0 wg show wg0 peers) == "$pub2" ]]
+n0 wg set wg0 private-key <(echo "/${key1:1}")
+[[ $(n0 wg show wg0 private-key) == "+${key1:1}" ]]
+n0 wg set wg0 peer "$pub2" allowed-ips 0.0.0.0/0,10.0.0.0/8,100.0.0.0/10,172.16.0.0/12,192.168.0.0/16
+n0 wg set wg0 peer "$pub2" allowed-ips 0.0.0.0/0
+n0 wg set wg0 peer "$pub2" allowed-ips ::/0,1700::/111,5000::/4,e000::/37,9000::/75
+n0 wg set wg0 peer "$pub2" allowed-ips ::/0
+n0 wg set wg0 peer "$pub2" remove
+low_order_points=( AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA= AQAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA= 4Ot6fDtBuK4WVuP68Z/EatoJjeucMrH9hmIFFl9JuAA= X5yVvKNQjCSx0LFVnIPvWwREXMRYHI6G2CJO3dCfEVc= 7P///////////////////////////////////////38= 7f///////////////////////////////////////38= 7v///////////////////////////////////////38= )
+n0 wg set wg0 private-key /dev/null ${low_order_points[@]/#/peer }
+[[ -z $(n0 wg show wg0 peers) ]]
+n0 wg set wg0 private-key <(echo "$key1") ${low_order_points[@]/#/peer }
+[[ -z $(n0 wg show wg0 peers) ]]
+ip0 link del wg0
+
+declare -A objects
+while read -t 0.1 -r line 2>/dev/null || [[ $? -ne 142 ]]; do
+	[[ $line =~ .*(wg[0-9]+:\ [A-Z][a-z]+\ [0-9]+)\ .*(created|destroyed).* ]] || continue
+	objects["${BASH_REMATCH[1]}"]+="${BASH_REMATCH[2]}"
+done < /dev/kmsg
+alldeleted=1
+for object in "${!objects[@]}"; do
+	if [[ ${objects["$object"]} != *createddestroyed ]]; then
+		echo "Error: $object: merely ${objects["$object"]}" >&3
+		alldeleted=0
+	fi
+done
+[[ $alldeleted -eq 1 ]]
+pretty "" "Objects that were created were also destroyed."
diff --git a/net/wireguard/tests/qemu/Makefile b/net/wireguard/tests/qemu/Makefile
new file mode 100644
index 000000000000..76dc450460e3
--- /dev/null
+++ b/net/wireguard/tests/qemu/Makefile
@@ -0,0 +1,410 @@
+# SPDX-License-Identifier: GPL-2.0
+#
+# Copyright (C) 2015-2019 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved.
+
+PWD := $(shell pwd)
+
+CHOST := $(shell gcc -dumpmachine)
+HOST_ARCH := $(firstword $(subst -, ,$(CHOST)))
+ifneq (,$(ARCH))
+CBUILD := $(subst -gcc,,$(lastword $(subst /, ,$(firstword $(filter-out android,$(wildcard $(foreach bindir,$(subst :, ,$(PATH)),$(bindir)/$(ARCH)-*-gcc)))))))
+endif
+ifeq (,$(CBUILD))
+CBUILD := $(CHOST)
+endif
+ARCH := $(firstword $(subst -, ,$(CBUILD)))
+
+# Set these from the environment to override
+KERNEL_VERSION ?= 5.5
+KERNEL_VERSION := $(KERNEL_VERSION)$(if $(DEBUG_KERNEL),$(if $(findstring -debug,$(KERNEL_VERSION)),,-debug),)
+BUILD_PATH ?= $(PWD)/../../../qemu-build/$(ARCH)
+DISTFILES_PATH ?= $(PWD)/distfiles
+NR_CPUS ?= 4
+
+MIRROR := https://download.wireguard.com/qemu-test/distfiles/
+
+rwildcard=$(foreach d,$(wildcard $1*),$(call rwildcard,$d/,$2) $(filter $(subst *,%,$2),$d))
+WIREGUARD_SOURCES := ../../Kbuild ../../Kconfig $(filter-out ../../tests/%,$(call rwildcard,../../,*.c *.h *.S *.pl *.include))
+
+default: qemu
+
+# variable name, tarball project name, version, tarball extension, default URI base
+define tar_download =
+$(1)_VERSION := $(3)
+$(1)_NAME := $(2)-$$($(1)_VERSION)
+$(1)_TAR := $(DISTFILES_PATH)/$$($(1)_NAME:-debug=)$(4)
+$(1)_PATH := $(BUILD_PATH)/$$($(1)_NAME)
+$(call file_download,$$($(1)_NAME:-debug=)$(4),$(5))
+endef
+
+define file_download =
+$(DISTFILES_PATH)/$(1):
+	mkdir -p $(DISTFILES_PATH)
+	flock -x $$@.lock -c '[ -f $$@ ] && exit 0; wget -O $$@ $(MIRROR)$(1) || wget -t inf --retry-on-http-error=404 -O $$@ $(2)$(1) || rm -f $$@'
+endef
+
+ifeq ($(findstring -rc,$(KERNEL_VERSION)),)
+KERNEL_URL_DIRECTORY := https://cdn.kernel.org/pub/linux/kernel/v$(firstword $(subst ., ,$(KERNEL_VERSION:-debug=))).x/
+else
+KERNEL_URL_DIRECTORY := https://git.kernel.org/torvalds/t/
+endif
+
+$(eval $(call tar_download,KERNEL,linux,$(KERNEL_VERSION),.tar.gz,$(KERNEL_URL_DIRECTORY)))
+$(eval $(call tar_download,MUSL,musl,1.1.24,.tar.gz,https://www.musl-libc.org/releases/))
+$(eval $(call tar_download,LIBMNL,libmnl,1.0.4,.tar.bz2,https://www.netfilter.org/projects/libmnl/files/))
+$(eval $(call tar_download,IPERF,iperf,3.7,.tar.gz,https://downloads.es.net/pub/iperf/))
+$(eval $(call tar_download,BASH,bash,5.0,.tar.gz,https://ftp.gnu.org/gnu/bash/))
+$(eval $(call tar_download,IPROUTE2,iproute2,5.4.0,.tar.xz,https://www.kernel.org/pub/linux/utils/net/iproute2/))
+$(eval $(call tar_download,IPTABLES,iptables,1.8.4,.tar.bz2,https://www.netfilter.org/projects/iptables/files/))
+$(eval $(call tar_download,NMAP,nmap,7.80,.tar.bz2,https://nmap.org/dist/))
+$(eval $(call tar_download,IPUTILS,iputils,s20161105,.tar.gz,https://github.com/iputils/iputils/archive/s20161105.tar.gz/#))
+$(eval $(call tar_download,WIREGUARD_TOOLS,wireguard-tools,1.0.20191226,.tar.xz,https://git.zx2c4.com/wireguard-tools/snapshot/))
+
+export CFLAGS ?= -O3 -pipe
+export LDFLAGS ?=
+export CPPFLAGS := -I$(BUILD_PATH)/include
+
+ifeq ($(HOST_ARCH),$(ARCH))
+CROSS_COMPILE_FLAG := --host=$(CHOST)
+NOPIE_GCC := gcc -fno-PIE
+CFLAGS += -march=native
+OMIT_AVX512 := $(shell gcc -march=native -Q --help=target | sed -n 's/.*\-m\(avx512[^ ]*\).*\[enabled\].*/-mno-\1/p')
+CFLAGS += $(OMIT_AVX512)
+STRIP := strip
+else
+$(info Cross compilation: building for $(CBUILD) using $(CHOST))
+CROSS_COMPILE_FLAG := --build=$(CBUILD) --host=$(CHOST)
+export CROSS_COMPILE=$(CBUILD)-
+NOPIE_GCC := $(CBUILD)-gcc -fno-PIE
+STRIP := $(CBUILD)-strip
+endif
+ifeq ($(ARCH),aarch64)
+QEMU_ARCH := aarch64
+KERNEL_ARCH := arm64
+KERNEL_BZIMAGE := $(KERNEL_PATH)/arch/arm64/boot/Image
+ifeq ($(HOST_ARCH),$(ARCH))
+QEMU_MACHINE := -cpu host -machine virt,gic_version=host,accel=kvm
+else
+QEMU_MACHINE := -cpu cortex-a53 -machine virt
+CFLAGS += -march=armv8-a -mtune=cortex-a53
+endif
+else ifeq ($(ARCH),aarch64_be)
+QEMU_ARCH := aarch64
+KERNEL_ARCH := arm64
+KERNEL_BZIMAGE := $(KERNEL_PATH)/arch/arm64/boot/Image
+ifeq ($(HOST_ARCH),$(ARCH))
+QEMU_MACHINE := -cpu host -machine virt,gic_version=host,accel=kvm
+else
+QEMU_MACHINE := -cpu cortex-a53 -machine virt
+CFLAGS += -march=armv8-a -mtune=cortex-a53
+endif
+else ifeq ($(ARCH),arm)
+QEMU_ARCH := arm
+KERNEL_ARCH := arm
+KERNEL_BZIMAGE := $(KERNEL_PATH)/arch/arm/boot/zImage
+ifeq ($(HOST_ARCH),$(ARCH))
+QEMU_MACHINE := -cpu host -machine virt,gic_version=host,accel=kvm
+else
+QEMU_MACHINE := -cpu cortex-a15 -machine virt
+CFLAGS += -march=armv7-a -mtune=cortex-a15 -mabi=aapcs-linux
+endif
+else ifeq ($(ARCH),armeb)
+QEMU_ARCH := arm
+KERNEL_ARCH := arm
+KERNEL_BZIMAGE := $(KERNEL_PATH)/arch/arm/boot/zImage
+ifeq ($(HOST_ARCH),$(ARCH))
+QEMU_MACHINE := -cpu host -machine virt,gic_version=host,accel=kvm
+else
+QEMU_MACHINE := -cpu cortex-a15 -machine virt
+CFLAGS += -march=armv7-a -mabi=aapcs-linux # We don't pass -mtune=cortex-a15 due to a compiler bug on big endian.
+LDFLAGS += -Wl,--be8
+endif
+else ifeq ($(ARCH),x86_64)
+QEMU_ARCH := x86_64
+KERNEL_ARCH := x86_64
+KERNEL_BZIMAGE := $(KERNEL_PATH)/arch/x86/boot/bzImage
+ifeq ($(HOST_ARCH),$(ARCH))
+QEMU_MACHINE := -cpu host -machine q35,accel=kvm
+else
+QEMU_MACHINE := -cpu Skylake-Server -machine q35
+CFLAGS += -march=skylake-avx512
+endif
+else ifeq ($(ARCH),i686)
+QEMU_ARCH := i386
+KERNEL_ARCH := x86
+KERNEL_BZIMAGE := $(KERNEL_PATH)/arch/x86/boot/bzImage
+ifeq ($(subst x86_64,i686,$(HOST_ARCH)),$(ARCH))
+QEMU_MACHINE := -cpu host -machine q35,accel=kvm
+else
+QEMU_MACHINE := -cpu coreduo -machine q35
+CFLAGS += -march=prescott
+endif
+else ifeq ($(ARCH),mips64)
+QEMU_ARCH := mips64
+KERNEL_ARCH := mips
+KERNEL_BZIMAGE := $(KERNEL_PATH)/vmlinux
+ifeq ($(HOST_ARCH),$(ARCH))
+QEMU_MACHINE := -cpu host -machine malta,accel=kvm
+CFLAGS += -EB
+else
+QEMU_MACHINE := -cpu MIPS64R2-generic -machine malta -smp 1
+CFLAGS += -march=mips64r2 -EB
+endif
+else ifeq ($(ARCH),mips64el)
+QEMU_ARCH := mips64el
+KERNEL_ARCH := mips
+KERNEL_BZIMAGE := $(KERNEL_PATH)/vmlinux
+ifeq ($(HOST_ARCH),$(ARCH))
+QEMU_MACHINE := -cpu host -machine malta,accel=kvm
+CFLAGS += -EL
+else
+QEMU_MACHINE := -cpu MIPS64R2-generic -machine malta -smp 1
+CFLAGS += -march=mips64r2 -EL
+endif
+else ifeq ($(ARCH),mips)
+QEMU_ARCH := mips
+KERNEL_ARCH := mips
+KERNEL_BZIMAGE := $(KERNEL_PATH)/vmlinux
+ifeq ($(HOST_ARCH),$(ARCH))
+QEMU_MACHINE := -cpu host -machine malta,accel=kvm
+CFLAGS += -EB
+else
+QEMU_MACHINE := -cpu 24Kf -machine malta -smp 1
+CFLAGS += -march=mips32r2 -EB
+endif
+else ifeq ($(ARCH),mipsel)
+QEMU_ARCH := mipsel
+KERNEL_ARCH := mips
+KERNEL_BZIMAGE := $(KERNEL_PATH)/vmlinux
+ifeq ($(HOST_ARCH),$(ARCH))
+QEMU_MACHINE := -cpu host -machine malta,accel=kvm
+CFLAGS += -EL
+else
+QEMU_MACHINE := -cpu 24Kf -machine malta -smp 1
+CFLAGS += -march=mips32r2 -EL
+endif
+else ifeq ($(ARCH),powerpc64le)
+QEMU_ARCH := ppc64
+KERNEL_ARCH := powerpc
+KERNEL_BZIMAGE := $(KERNEL_PATH)/vmlinux
+ifeq ($(HOST_ARCH),$(ARCH))
+QEMU_MACHINE := -cpu host,accel=kvm -machine pseries
+else
+QEMU_MACHINE := -machine pseries
+endif
+CFLAGS += -mcpu=powerpc64le -mlong-double-64
+else ifeq ($(ARCH),powerpc)
+QEMU_ARCH := ppc
+KERNEL_ARCH := powerpc
+KERNEL_BZIMAGE := $(KERNEL_PATH)/arch/powerpc/boot/uImage
+ifeq ($(HOST_ARCH),$(ARCH))
+QEMU_MACHINE := -cpu host,accel=kvm -machine ppce500
+else
+QEMU_MACHINE := -machine ppce500
+endif
+CFLAGS += -mcpu=powerpc -mlong-double-64 -msecure-plt
+else ifeq ($(ARCH),m68k)
+QEMU_ARCH := m68k
+KERNEL_ARCH := m68k
+KERNEL_BZIMAGE := $(KERNEL_PATH)/vmlinux
+KERNEL_CMDLINE := $(shell sed -n 's/CONFIG_CMDLINE=\(.*\)/\1/p' arch/m68k.config)
+ifeq ($(HOST_ARCH),$(ARCH))
+QEMU_MACHINE := -cpu host,accel=kvm -machine q800 -smp 1 -append $(KERNEL_CMDLINE)
+else
+QEMU_MACHINE := -machine q800 -smp 1 -append $(KERNEL_CMDLINE)
+endif
+else
+$(error I only build: x86_64, i686, arm, armeb, aarch64, aarch64_be, mips, mipsel, mips64, mips64el, powerpc64le, powerpc, m68k)
+endif
+
+REAL_CC := $(CBUILD)-gcc
+MUSL_CC := $(BUILD_PATH)/musl-gcc
+export CC := $(MUSL_CC)
+USERSPACE_DEPS := $(MUSL_CC) $(BUILD_PATH)/include/.installed $(BUILD_PATH)/include/linux/.installed
+
+build: $(KERNEL_BZIMAGE)
+qemu: $(KERNEL_BZIMAGE)
+	rm -f $(BUILD_PATH)/result
+	timeout --foreground 20m qemu-system-$(QEMU_ARCH) \
+		-nodefaults \
+		-nographic \
+		-smp $(NR_CPUS) \
+		$(QEMU_MACHINE) \
+		-m $$(grep -q CONFIG_DEBUG_KMEMLEAK=y $(KERNEL_PATH)/.config && echo 1G || echo 256M) \
+		-serial stdio \
+		-serial file:$(BUILD_PATH)/result \
+		-no-reboot \
+		-monitor none \
+		-kernel $<
+	grep -Fq success $(BUILD_PATH)/result
+
+$(BUILD_PATH)/init-cpio-spec.txt:
+	mkdir -p $(BUILD_PATH)
+	echo "file /init $(BUILD_PATH)/init 755 0 0" > $@
+	echo "file /init.sh $(PWD)/../netns.sh 755 0 0" >> $@
+	echo "dir /dev 755 0 0" >> $@
+	echo "nod /dev/console 644 0 0 c 5 1" >> $@
+	echo "dir /bin 755 0 0" >> $@
+	echo "file /bin/iperf3 $(IPERF_PATH)/src/iperf3 755 0 0" >> $@
+	echo "file /bin/wg $(WIREGUARD_TOOLS_PATH)/src/wg 755 0 0" >> $@
+	echo "file /bin/bash $(BASH_PATH)/bash 755 0 0" >> $@
+	echo "file /bin/ip $(IPROUTE2_PATH)/ip/ip 755 0 0" >> $@
+	echo "file /bin/ss $(IPROUTE2_PATH)/misc/ss 755 0 0" >> $@
+	echo "file /bin/ping $(IPUTILS_PATH)/ping 755 0 0" >> $@
+	echo "file /bin/ncat $(NMAP_PATH)/ncat/ncat 755 0 0" >> $@
+	echo "file /bin/xtables-legacy-multi $(IPTABLES_PATH)/iptables/xtables-legacy-multi 755 0 0" >> $@
+	echo "slink /bin/iptables xtables-legacy-multi 777 0 0" >> $@
+	echo "slink /bin/ping6 ping 777 0 0" >> $@
+	echo "dir /lib 755 0 0" >> $@
+	echo "file /lib/libc.so $(MUSL_PATH)/lib/libc.so 755 0 0" >> $@
+	echo "slink /lib/ld-linux.so.1 libc.so 777 0 0" >> $@
+
+ifeq ($(findstring -git,$(KERNEL_VERSION)),)
+$(KERNEL_PATH)/.installed: $(KERNEL_TAR)
+	mkdir -p $(KERNEL_PATH)
+	flock -s $<.lock tar --strip-components=1 -C $(KERNEL_PATH) -xf $<
+	sed -i "/^if INET\$$/a source \"net/wireguard/Kconfig\"" $(KERNEL_PATH)/net/Kconfig
+	sed -i "/^obj-\$$(CONFIG_NETFILTER).*+=/a obj-\$$(CONFIG_WIREGUARD) += wireguard/" $(KERNEL_PATH)/net/Makefile
+	ln -sfT $(shell readlink -f ../..) $(KERNEL_PATH)/net/wireguard
+	touch $@
+else
+$(KERNEL_PATH)/.installed:
+	mkdir -p $(dir $(KERNEL_PATH))
+	flock -s $(dir $(KERNEL_TAR))/$(KERNEL_VERSION).lock git clone $(GIT_URI_$(patsubst %-git,%,$(patsubst %-debug,%,$(KERNEL_VERSION)))) $(KERNEL_PATH)
+	touch $@
+always-pull: $(KERNEL_PATH)/.installed
+	flock -s $(dir $(KERNEL_TAR))/$(KERNEL_VERSION).lock git -C $(KERNEL_PATH) fetch
+	flock -s $(dir $(KERNEL_TAR))/$(KERNEL_VERSION).lock git -C $(KERNEL_PATH) reset --hard FETCH_HEAD
+.PHONY: always-pull
+$(KERNEL_BZIMAGE): always-pull
+endif
+
+$(KERNEL_PATH)/.config: kernel.config arch/$(ARCH).config | $(KERNEL_PATH)/.installed
+	cp kernel.config $(KERNEL_PATH)/minimal.config
+	printf 'CONFIG_NR_CPUS=$(NR_CPUS)\nCONFIG_INITRAMFS_SOURCE="$(BUILD_PATH)/init-cpio-spec.txt"\n' >> $(KERNEL_PATH)/minimal.config
+	cat arch/$(ARCH).config >> $(KERNEL_PATH)/minimal.config
+	$(MAKE) -C $(KERNEL_PATH) ARCH=$(KERNEL_ARCH) allnoconfig
+	cd $(KERNEL_PATH) && ARCH=$(KERNEL_ARCH) scripts/kconfig/merge_config.sh -n .config minimal.config
+	$(if $(findstring -debug,$(KERNEL_VERSION)),cd $(KERNEL_PATH) && sed -i 's/^EXTRAVERSION =.*/EXTRAVERSION = -debug/' Makefile && ARCH=$(KERNEL_ARCH) scripts/kconfig/merge_config.sh -n .config $(PWD)/debug.config,)
+
+$(KERNEL_BZIMAGE): $(KERNEL_PATH)/.config $(BUILD_PATH)/init-cpio-spec.txt $(MUSL_PATH)/lib/libc.so $(IPERF_PATH)/src/iperf3 $(IPUTILS_PATH)/ping $(BASH_PATH)/bash $(IPROUTE2_PATH)/misc/ss $(IPROUTE2_PATH)/ip/ip $(IPTABLES_PATH)/iptables/xtables-legacy-multi $(NMAP_PATH)/ncat/ncat $(WIREGUARD_TOOLS_PATH)/src/wg $(BUILD_PATH)/init ../netns.sh $(WIREGUARD_SOURCES)
+	LOCALVERSION="" $(MAKE) -C $(KERNEL_PATH) ARCH=$(KERNEL_ARCH) CROSS_COMPILE=$(CROSS_COMPILE) CC="$(NOPIE_GCC)"
+
+$(BUILD_PATH)/include/linux/.installed: | $(KERNEL_PATH)/.config
+	LOCALVERSION="" $(MAKE) -C $(KERNEL_PATH) INSTALL_HDR_PATH=$(BUILD_PATH) ARCH=$(KERNEL_ARCH) CROSS_COMPILE=$(CROSS_COMPILE) headers_install
+	touch $@
+
+$(MUSL_PATH)/lib/libc.so: $(MUSL_TAR)
+	mkdir -p $(BUILD_PATH)
+	flock -s $<.lock tar -C $(BUILD_PATH) -xf $<
+	cd $(MUSL_PATH) && CC=$(REAL_CC) ./configure --prefix=/ --disable-static --build=$(CBUILD)
+	$(MAKE) -C $(MUSL_PATH)
+	$(STRIP) -s $@
+
+$(BUILD_PATH)/include/.installed: $(MUSL_PATH)/lib/libc.so
+	$(MAKE) -C $(MUSL_PATH) DESTDIR=$(BUILD_PATH) install-headers
+	touch $@
+
+$(MUSL_CC): $(MUSL_PATH)/lib/libc.so
+	sh $(MUSL_PATH)/tools/musl-gcc.specs.sh $(BUILD_PATH)/include $(MUSL_PATH)/lib /lib/ld-linux.so.1 > $(BUILD_PATH)/musl-gcc.specs
+	printf '#!/bin/sh\nexec "$(REAL_CC)" --specs="$(BUILD_PATH)/musl-gcc.specs" "$$@"\n' > $(BUILD_PATH)/musl-gcc
+	chmod +x $(BUILD_PATH)/musl-gcc
+
+$(IPERF_PATH)/.installed: $(IPERF_TAR)
+	mkdir -p $(BUILD_PATH)
+	flock -s $<.lock tar -C $(BUILD_PATH) -xf $<
+	sed -i '1s/^/#include <stdint.h>/' $(IPERF_PATH)/src/cjson.h $(IPERF_PATH)/src/timer.h
+	sed -i -r 's/-p?g//g' $(IPERF_PATH)/src/Makefile*
+	touch $@
+
+$(IPERF_PATH)/src/iperf3: | $(IPERF_PATH)/.installed $(USERSPACE_DEPS)
+	cd $(IPERF_PATH) && CFLAGS="$(CFLAGS) -D_GNU_SOURCE" ./configure --prefix=/ $(CROSS_COMPILE_FLAG) --enable-static --disable-shared --with-openssl=no
+	$(MAKE) -C $(IPERF_PATH)
+	$(STRIP) -s $@
+
+$(LIBMNL_PATH)/.installed: $(LIBMNL_TAR)
+	flock -s $<.lock tar -C $(BUILD_PATH) -xf $<
+	touch $@
+
+$(LIBMNL_PATH)/src/.libs/libmnl.a: | $(LIBMNL_PATH)/.installed $(USERSPACE_DEPS)
+	cd $(LIBMNL_PATH) && ./configure --prefix=/ $(CROSS_COMPILE_FLAG) --enable-static --disable-shared
+	$(MAKE) -C $(LIBMNL_PATH)
+	sed -i 's:prefix=.*:prefix=$(LIBMNL_PATH):' $(LIBMNL_PATH)/libmnl.pc
+
+$(WIREGUARD_TOOLS_PATH)/.installed: $(WIREGUARD_TOOLS_TAR)
+	flock -s $<.lock tar -C $(BUILD_PATH) -xf $<
+	touch $@
+
+$(WIREGUARD_TOOLS_PATH)/src/wg: | $(WIREGUARD_TOOLS_PATH)/.installed $(LIBMNL_PATH)/src/.libs/libmnl.a $(USERSPACE_DEPS)
+	LDFLAGS="$(LDFLAGS) -L$(LIBMNL_PATH)/src/.libs" $(MAKE) -C $(WIREGUARD_TOOLS_PATH)/src LIBMNL_CFLAGS="-I$(LIBMNL_PATH)/include" LIBMNL_LDLIBS="-lmnl" wg
+	$(STRIP) -s $@
+
+$(BUILD_PATH)/init: init.c | $(USERSPACE_DEPS)
+	mkdir -p $(BUILD_PATH)
+	$(MUSL_CC) -o $@ $(CFLAGS) $(LDFLAGS) -std=gnu11 $<
+	$(STRIP) -s $@
+
+$(IPUTILS_PATH)/.installed: $(IPUTILS_TAR)
+	mkdir -p $(BUILD_PATH)
+	flock -s $<.lock tar -C $(BUILD_PATH) -xf $<
+	touch $@
+
+$(IPUTILS_PATH)/ping: | $(IPUTILS_PATH)/.installed $(USERSPACE_DEPS)
+	$(MAKE) -C $(IPUTILS_PATH) USE_CAP=no USE_IDN=no USE_NETTLE=no USE_CRYPTO=no ping
+	$(STRIP) -s $@
+
+$(BASH_PATH)/.installed: $(BASH_TAR)
+	mkdir -p $(BUILD_PATH)
+	flock -s $<.lock tar -C $(BUILD_PATH) -xf $<
+	touch $@
+
+$(BASH_PATH)/bash: | $(BASH_PATH)/.installed $(USERSPACE_DEPS)
+	cd $(BASH_PATH) && ./configure --prefix=/ $(CROSS_COMPILE_FLAG) --without-bash-malloc --disable-debugger --disable-help-builtin --disable-history --disable-multibyte --disable-progcomp --disable-readline --disable-mem-scramble
+	$(MAKE) -C $(BASH_PATH)
+	$(STRIP) -s $@
+
+$(IPROUTE2_PATH)/.installed: $(IPROUTE2_TAR)
+	mkdir -p $(BUILD_PATH)
+	flock -s $<.lock tar -C $(BUILD_PATH) -xf $<
+	printf 'CC:=$(CC)\nPKG_CONFIG:=pkg-config\nTC_CONFIG_XT:=n\nTC_CONFIG_ATM:=n\nTC_CONFIG_IPSET:=n\nIP_CONFIG_SETNS:=y\nHAVE_ELF:=n\nHAVE_MNL:=y\nHAVE_BERKELEY_DB:=n\nHAVE_LATEX:=n\nHAVE_PDFLATEX:=n\nCFLAGS+=-DHAVE_SETNS -DHAVE_LIBMNL -I$(LIBMNL_PATH)/include\nLDLIBS+=-lmnl' > $(IPROUTE2_PATH)/config.mk
+	printf 'lib: snapshot\n\t$$(MAKE) -C lib\nip/ip: lib\n\t$$(MAKE) -C ip ip\nmisc/ss: lib\n\t$$(MAKE) -C misc ss\n' >> $(IPROUTE2_PATH)/Makefile
+	touch $@
+
+$(IPROUTE2_PATH)/ip/ip: | $(IPROUTE2_PATH)/.installed $(LIBMNL_PATH)/src/.libs/libmnl.a $(USERSPACE_DEPS)
+	LDFLAGS="$(LDFLAGS) -L$(LIBMNL_PATH)/src/.libs" PKG_CONFIG_LIBDIR="$(LIBMNL_PATH)" $(MAKE) -C $(IPROUTE2_PATH) PREFIX=/ ip/ip
+	$(STRIP) -s $(IPROUTE2_PATH)/ip/ip
+
+$(IPROUTE2_PATH)/misc/ss: | $(IPROUTE2_PATH)/.installed $(LIBMNL_PATH)/src/.libs/libmnl.a $(USERSPACE_DEPS)
+	LDFLAGS="$(LDFLAGS) -L$(LIBMNL_PATH)/src/.libs" PKG_CONFIG_LIBDIR="$(LIBMNL_PATH)" $(MAKE) -C $(IPROUTE2_PATH) PREFIX=/ misc/ss
+	$(STRIP) -s $(IPROUTE2_PATH)/misc/ss
+
+$(IPTABLES_PATH)/.installed: $(IPTABLES_TAR)
+	mkdir -p $(BUILD_PATH)
+	flock -s $<.lock tar -C $(BUILD_PATH) -xf $<
+	sed -i -e "/nfnetlink=[01]/s:=[01]:=0:" -e "/nfconntrack=[01]/s:=[01]:=0:" $(IPTABLES_PATH)/configure
+	touch $@
+
+$(IPTABLES_PATH)/iptables/xtables-legacy-multi: | $(IPTABLES_PATH)/.installed $(LIBMNL_PATH)/src/.libs/libmnl.a $(USERSPACE_DEPS)
+	cd $(IPTABLES_PATH) && PKG_CONFIG_LIBDIR="$(LIBMNL_PATH)" ./configure --prefix=/ $(CROSS_COMPILE_FLAG) --enable-static --disable-shared --disable-nftables --disable-bpf-compiler --disable-nfsynproxy --disable-libipq --with-kernel=$(BUILD_PATH)/include
+	$(MAKE) -C $(IPTABLES_PATH)
+	$(STRIP) -s $@
+
+$(NMAP_PATH)/.installed: $(NMAP_TAR)
+	mkdir -p $(BUILD_PATH)
+	flock -s $<.lock tar -C $(BUILD_PATH) -xf $<
+	touch $@
+
+$(NMAP_PATH)/ncat/ncat: | $(NMAP_PATH)/.installed $(USERSPACE_DEPS)
+	cd $(NMAP_PATH) && ./configure --prefix=/ $(CROSS_COMPILE_FLAG) --enable-static --disable-shared --without-ndiff --without-zenmap --without-nping --with-libpcap=included --with-libpcre=included --with-libdnet=included --without-liblua --with-liblinear=included --without-nmap-update --without-openssl --with-pcap=linux --without-libssh
+	$(MAKE) -C $(NMAP_PATH)/libpcap
+	$(MAKE) -C $(NMAP_PATH)/ncat
+	$(STRIP) -s $@
+
+clean:
+	rm -rf $(BUILD_PATH)
+
+distclean: clean
+	rm -rf $(DISTFILES_PATH)
+
+.PHONY: qemu build clean distclean
+.DELETE_ON_ERROR:
diff --git a/net/wireguard/tests/qemu/arch/m68k.config b/net/wireguard/tests/qemu/arch/m68k.config
new file mode 100644
index 000000000000..62a15bdb877e
--- /dev/null
+++ b/net/wireguard/tests/qemu/arch/m68k.config
@@ -0,0 +1,9 @@
+CONFIG_MMU=y
+CONFIG_M68KCLASSIC=y
+CONFIG_M68040=y
+CONFIG_MAC=y
+CONFIG_SERIAL_PMACZILOG=y
+CONFIG_SERIAL_PMACZILOG_TTYS=y
+CONFIG_SERIAL_PMACZILOG_CONSOLE=y
+CONFIG_CMDLINE="console=ttyS0 wg.success=ttyS1"
+CONFIG_FRAME_WARN=1024
diff --git a/net/wireguard/tests/qemu/init.c b/net/wireguard/tests/qemu/init.c
new file mode 100644
index 000000000000..90bc9813cadc
--- /dev/null
+++ b/net/wireguard/tests/qemu/init.c
@@ -0,0 +1,285 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2015-2019 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved.
+ */
+
+#define _GNU_SOURCE
+#include <unistd.h>
+#include <errno.h>
+#include <string.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdbool.h>
+#include <fcntl.h>
+#include <sys/wait.h>
+#include <sys/mount.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <sys/types.h>
+#include <sys/io.h>
+#include <sys/ioctl.h>
+#include <sys/reboot.h>
+#include <sys/utsname.h>
+#include <sys/sendfile.h>
+#include <sys/sysmacros.h>
+#include <linux/random.h>
+#include <linux/version.h>
+
+__attribute__((noreturn)) static void poweroff(void)
+{
+	fflush(stdout);
+	fflush(stderr);
+	reboot(RB_AUTOBOOT);
+	sleep(30);
+	fprintf(stderr, "\x1b[37m\x1b[41m\x1b[1mFailed to power off!!!\x1b[0m\n");
+	exit(1);
+}
+
+static void panic(const char *what)
+{
+	fprintf(stderr, "\n\n\x1b[37m\x1b[41m\x1b[1mSOMETHING WENT HORRIBLY WRONG\x1b[0m\n\n    \x1b[31m\x1b[1m%s: %s\x1b[0m\n\n\x1b[37m\x1b[44m\x1b[1mPower off...\x1b[0m\n\n", what, strerror(errno));
+	poweroff();
+}
+
+#define pretty_message(msg) puts("\x1b[32m\x1b[1m" msg "\x1b[0m")
+
+static void print_banner(void)
+{
+	struct utsname utsname;
+	int len;
+
+	if (uname(&utsname) < 0)
+		panic("uname");
+
+	len = strlen("    WireGuard Test Suite on       ") + strlen(utsname.sysname) + strlen(utsname.release) + strlen(utsname.machine);
+	printf("\x1b[45m\x1b[33m\x1b[1m%*.s\x1b[0m\n\x1b[45m\x1b[33m\x1b[1m    WireGuard Test Suite on %s %s %s    \x1b[0m\n\x1b[45m\x1b[33m\x1b[1m%*.s\x1b[0m\n\n", len, "", utsname.sysname, utsname.release, utsname.machine, len, "");
+}
+
+static void seed_rng(void)
+{
+	int fd;
+	struct {
+		int entropy_count;
+		int buffer_size;
+		unsigned char buffer[256];
+	} entropy = {
+		.entropy_count = sizeof(entropy.buffer) * 8,
+		.buffer_size = sizeof(entropy.buffer),
+		.buffer = "Adding real entropy is not actually important for these tests. Don't try this at home, kids!"
+	};
+
+	if (mknod("/dev/urandom", S_IFCHR | 0644, makedev(1, 9)))
+		panic("mknod(/dev/urandom)");
+	fd = open("/dev/urandom", O_WRONLY);
+	if (fd < 0)
+		panic("open(urandom)");
+	for (int i = 0; i < 256; ++i) {
+		if (ioctl(fd, RNDADDENTROPY, &entropy) < 0)
+			panic("ioctl(urandom)");
+	}
+	close(fd);
+}
+
+static void mount_filesystems(void)
+{
+	pretty_message("[+] Mounting filesystems...");
+	mkdir("/dev", 0755);
+	mkdir("/proc", 0755);
+	mkdir("/sys", 0755);
+	mkdir("/tmp", 0755);
+	mkdir("/run", 0755);
+	mkdir("/var", 0755);
+	if (mount("none", "/dev", "devtmpfs", 0, NULL))
+		panic("devtmpfs mount");
+	if (mount("none", "/proc", "proc", 0, NULL))
+		panic("procfs mount");
+	if (mount("none", "/sys", "sysfs", 0, NULL))
+		panic("sysfs mount");
+	if (mount("none", "/tmp", "tmpfs", 0, NULL))
+		panic("tmpfs mount");
+	if (mount("none", "/run", "tmpfs", 0, NULL))
+		panic("tmpfs mount");
+	if (mount("none", "/sys/kernel/debug", "debugfs", 0, NULL))
+		; /* Not a problem if it fails.*/
+	if (symlink("/run", "/var/run"))
+		panic("run symlink");
+	if (symlink("/proc/self/fd", "/dev/fd"))
+		panic("fd symlink");
+}
+
+static void enable_logging(void)
+{
+	int fd;
+	pretty_message("[+] Enabling logging...");
+	fd = open("/proc/sys/kernel/printk", O_WRONLY);
+	if (fd >= 0) {
+		if (write(fd, "9\n", 2) != 2)
+			panic("write(printk)");
+		close(fd);
+	}
+	fd = open("/proc/sys/debug/exception-trace", O_WRONLY);
+	if (fd >= 0) {
+		if (write(fd, "1\n", 2) != 2)
+			panic("write(exception-trace)");
+		close(fd);
+	}
+	fd = open("/proc/sys/kernel/panic_on_warn", O_WRONLY);
+	if (fd >= 0) {
+		if (write(fd, "1\n", 2) != 2)
+			panic("write(panic_on_warn)");
+		close(fd);
+	}
+}
+
+static void kmod_selftests(void)
+{
+	FILE *file;
+	char line[2048], *start, *pass;
+	bool success = true;
+	pretty_message("[+] Module self-tests:");
+	file = fopen("/proc/kmsg", "r");
+	if (!file)
+		panic("fopen(kmsg)");
+	if (fcntl(fileno(file), F_SETFL, O_NONBLOCK) < 0)
+		panic("fcntl(kmsg, nonblock)");
+	while (fgets(line, sizeof(line), file)) {
+		start = strstr(line, "wireguard: ");
+		if (!start)
+			continue;
+		start += 11;
+		*strchrnul(start, '\n') = '\0';
+		if (strstr(start, "www.wireguard.com"))
+			break;
+		pass = strstr(start, ": pass");
+		if (!pass || pass[6] != '\0') {
+			success = false;
+			printf(" \x1b[31m*  %s\x1b[0m\n", start);
+		} else
+			printf(" \x1b[32m*  %s\x1b[0m\n", start);
+	}
+	fclose(file);
+	if (!success) {
+		puts("\x1b[31m\x1b[1m[-] Tests failed! \u2639\x1b[0m");
+		poweroff();
+	}
+}
+
+static void launch_tests(void)
+{
+	char cmdline[4096], *success_dev;
+	int status, fd;
+	pid_t pid;
+
+	pretty_message("[+] Launching tests...");
+	pid = fork();
+	if (pid == -1)
+		panic("fork");
+	else if (pid == 0) {
+		execl("/init.sh", "init", NULL);
+		panic("exec");
+	}
+	if (waitpid(pid, &status, 0) < 0)
+		panic("waitpid");
+	if (WIFEXITED(status) && WEXITSTATUS(status) == 0) {
+		pretty_message("[+] Tests successful! :-)");
+		fd = open("/proc/cmdline", O_RDONLY);
+		if (fd < 0)
+			panic("open(/proc/cmdline)");
+		if (read(fd, cmdline, sizeof(cmdline) - 1) <= 0)
+			panic("read(/proc/cmdline)");
+		cmdline[sizeof(cmdline) - 1] = '\0';
+		for (success_dev = strtok(cmdline, " \n"); success_dev; success_dev = strtok(NULL, " \n")) {
+			if (strncmp(success_dev, "wg.success=", 11))
+				continue;
+			memcpy(success_dev + 11 - 5, "/dev/", 5);
+			success_dev += 11 - 5;
+			break;
+		}
+		if (!success_dev || !strlen(success_dev))
+			panic("Unable to find success device");
+
+		fd = open(success_dev, O_WRONLY);
+		if (fd < 0)
+			panic("open(success_dev)");
+		if (write(fd, "success\n", 8) != 8)
+			panic("write(success_dev)");
+		close(fd);
+	} else {
+		const char *why = "unknown cause";
+		int what = -1;
+
+		if (WIFEXITED(status)) {
+			why = "exit code";
+			what = WEXITSTATUS(status);
+		} else if (WIFSIGNALED(status)) {
+			why = "signal";
+			what = WTERMSIG(status);
+		}
+		printf("\x1b[31m\x1b[1m[-] Tests failed with %s %d! \u2639\x1b[0m\n", why, what);
+	}
+}
+
+static void ensure_console(void)
+{
+	for (unsigned int i = 0; i < 1000; ++i) {
+		int fd = open("/dev/console", O_RDWR);
+		if (fd < 0) {
+			usleep(50000);
+			continue;
+		}
+		dup2(fd, 0);
+		dup2(fd, 1);
+		dup2(fd, 2);
+		close(fd);
+		if (write(1, "\0\0\0\0\n", 5) == 5)
+			return;
+	}
+	panic("Unable to open console device");
+}
+
+static void clear_leaks(void)
+{
+	int fd;
+
+	fd = open("/sys/kernel/debug/kmemleak", O_WRONLY);
+	if (fd < 0)
+		return;
+	pretty_message("[+] Starting memory leak detection...");
+	write(fd, "clear\n", 5);
+	close(fd);
+}
+
+static void check_leaks(void)
+{
+	int fd;
+
+	fd = open("/sys/kernel/debug/kmemleak", O_WRONLY);
+	if (fd < 0)
+		return;
+	pretty_message("[+] Scanning for memory leaks...");
+	sleep(2); /* Wait for any grace periods. */
+	write(fd, "scan\n", 5);
+	close(fd);
+
+	fd = open("/sys/kernel/debug/kmemleak", O_RDONLY);
+	if (fd < 0)
+		return;
+	if (sendfile(1, fd, NULL, 0x7ffff000) > 0)
+		panic("Memory leaks encountered");
+	close(fd);
+}
+
+int main(int argc, char *argv[])
+{
+	seed_rng();
+	ensure_console();
+	print_banner();
+	mount_filesystems();
+	kmod_selftests();
+	enable_logging();
+	clear_leaks();
+	launch_tests();
+	check_leaks();
+	poweroff();
+	return 1;
+}
diff --git a/net/wireguard/tests/qemu/kernel.config b/net/wireguard/tests/qemu/kernel.config
new file mode 100644
index 000000000000..071a43f5dfff
--- /dev/null
+++ b/net/wireguard/tests/qemu/kernel.config
@@ -0,0 +1,87 @@
+CONFIG_NET=y
+CONFIG_NETDEVICES=y
+CONFIG_NET_CORE=y
+CONFIG_NET_IPIP=y
+CONFIG_DUMMY=y
+CONFIG_VETH=y
+CONFIG_MULTIUSER=y
+CONFIG_NAMESPACES=y
+CONFIG_NET_NS=y
+CONFIG_UNIX=y
+CONFIG_INET=y
+CONFIG_IPV6=y
+CONFIG_NETFILTER=y
+CONFIG_NETFILTER_ADVANCED=y
+CONFIG_NF_CONNTRACK=y
+CONFIG_NF_NAT=y
+CONFIG_NETFILTER_XTABLES=y
+CONFIG_NETFILTER_XT_NAT=y
+CONFIG_NETFILTER_XT_MATCH_LENGTH=y
+CONFIG_NF_CONNTRACK_IPV4=y
+CONFIG_NF_NAT_IPV4=y
+CONFIG_IP_NF_IPTABLES=y
+CONFIG_IP_NF_FILTER=y
+CONFIG_IP_NF_NAT=y
+CONFIG_IP_ADVANCED_ROUTER=y
+CONFIG_IP_MULTIPLE_TABLES=y
+CONFIG_IPV6_MULTIPLE_TABLES=y
+CONFIG_TTY=y
+CONFIG_BINFMT_ELF=y
+CONFIG_BINFMT_SCRIPT=y
+CONFIG_VDSO=y
+CONFIG_VIRTUALIZATION=y
+CONFIG_HYPERVISOR_GUEST=y
+CONFIG_PARAVIRT=y
+CONFIG_KVM_GUEST=y
+CONFIG_PARAVIRT_SPINLOCKS=y
+CONFIG_PRINTK=y
+CONFIG_KALLSYMS=y
+CONFIG_BUG=y
+CONFIG_CC_OPTIMIZE_FOR_PERFORMANCE=y
+CONFIG_JUMP_LABEL=y
+CONFIG_EMBEDDED=n
+CONFIG_BASE_FULL=y
+CONFIG_FUTEX=y
+CONFIG_SHMEM=y
+CONFIG_SLUB=y
+CONFIG_SPARSEMEM_VMEMMAP=y
+CONFIG_SMP=y
+CONFIG_SCHED_SMT=y
+CONFIG_SCHED_MC=y
+CONFIG_NUMA=y
+CONFIG_PREEMPT=y
+CONFIG_NO_HZ=y
+CONFIG_NO_HZ_IDLE=y
+CONFIG_NO_HZ_FULL=n
+CONFIG_HZ_PERIODIC=n
+CONFIG_HIGH_RES_TIMERS=y
+CONFIG_COMPAT_32BIT_TIME=y
+CONFIG_ARCH_RANDOM=y
+CONFIG_FILE_LOCKING=y
+CONFIG_POSIX_TIMERS=y
+CONFIG_DEVTMPFS=y
+CONFIG_PROC_FS=y
+CONFIG_PROC_SYSCTL=y
+CONFIG_SYSFS=y
+CONFIG_TMPFS=y
+CONFIG_CONSOLE_LOGLEVEL_DEFAULT=15
+CONFIG_PRINTK_TIME=y
+CONFIG_BLK_DEV_INITRD=y
+CONFIG_LEGACY_VSYSCALL_NONE=y
+CONFIG_KERNEL_GZIP=y
+CONFIG_PANIC_ON_OOPS=y
+CONFIG_BUG_ON_DATA_CORRUPTION=y
+CONFIG_LOCKUP_DETECTOR=y
+CONFIG_SOFTLOCKUP_DETECTOR=y
+CONFIG_HARDLOCKUP_DETECTOR=y
+CONFIG_WQ_WATCHDOG=y
+CONFIG_DETECT_HUNG_TASK=y
+CONFIG_BOOTPARAM_HARDLOCKUP_PANIC=y
+CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC=y
+CONFIG_BOOTPARAM_HUNG_TASK_PANIC=y
+CONFIG_PANIC_TIMEOUT=-1
+CONFIG_STACKTRACE=y
+CONFIG_EARLY_PRINTK=y
+CONFIG_GDB_SCRIPTS=y
+CONFIG_WIREGUARD=y
+CONFIG_WIREGUARD_DEBUG=y
diff --git a/net/wireguard/version.h b/net/wireguard/version.h
new file mode 100644
index 000000000000..1f89f50912ea
--- /dev/null
+++ b/net/wireguard/version.h
@@ -0,0 +1,3 @@
+#ifndef WIREGUARD_VERSION
+#define WIREGUARD_VERSION "0.0.20200215"
+#endif