net: WireGuard secure network tunnel
WireGuard is a layer 3 secure networking tunnel made specifically for
the kernel, that aims to be much simpler and easier to audit than IPsec.
Extensive documentation and description of the protocol and
considerations, along with formal proofs of the cryptography, are
available at:
* https://www.wireguard.com/
* https://www.wireguard.com/papers/wireguard.pdf
This commit implements WireGuard as a simple network device driver,
accessible in the usual RTNL way used by virtual network drivers. It
makes use of the udp_tunnel APIs, GRO, GSO, NAPI, and the usual set of
networking subsystem APIs. It has a somewhat novel multicore queueing
system designed for maximum throughput and minimal latency of encryption
operations, but it is implemented modestly using workqueues and NAPI.
Configuration is done via generic Netlink, and following a review from
the Netlink maintainer a year ago, several high profile userspace tools
have already implemented the API.
This commit also comes with several different tests, both in-kernel
tests and out-of-kernel tests based on network namespaces, taking profit
of the fact that sockets used by WireGuard intentionally stay in the
namespace the WireGuard interface was originally created, exactly like
the semantics of userspace tun devices. See wireguard.com/netns/ for
pictures and examples.
The source code is fairly short, but rather than combining everything
into a single file, WireGuard is developed as cleanly separable files,
making auditing and comprehension easier. Things are laid out as
follows:
* noise.[ch], cookie.[ch], messages.h: These implement the bulk of the
cryptographic aspects of the protocol, and are mostly data-only in
nature, taking in buffers of bytes and spitting out buffers of
bytes. They also handle reference counting for their various shared
pieces of data, like keys and key lists.
* ratelimiter.[ch]: Used as an integral part of cookie.[ch] for
ratelimiting certain types of cryptographic operations in accordance
with particular WireGuard semantics.
* allowedips.[ch], peerlookup.[ch]: The main lookup structures of
WireGuard, the former being trie-like with particular semantics, an
integral part of the design of the protocol, and the latter just
being nice helper functions around the various hashtables we use.
* device.[ch]: Implementation of functions for the netdevice and for
rtnl, responsible for maintaining the life of a given interface and
wiring it up to the rest of WireGuard.
* peer.[ch]: Each interface has a list of peers, with helper functions
available here for creation, destruction, and reference counting.
* socket.[ch]: Implementation of functions related to udp_socket and
the general set of kernel socket APIs, for sending and receiving
ciphertext UDP packets, and taking care of WireGuard-specific sticky
socket routing semantics for the automatic roaming.
* netlink.[ch]: Userspace API entry point for configuring WireGuard
peers and devices. The API has been implemented by several userspace
tools and network management utility, and the WireGuard project
distributes the basic wg(8) tool.
* queueing.[ch]: Shared function on the rx and tx path for handling
the various queues used in the multicore algorithms.
* send.c: Handles encrypting outgoing packets in parallel on
multiple cores, before sending them in order on a single core, via
workqueues and ring buffers. Also handles sending handshake and cookie
messages as part of the protocol, in parallel.
* receive.c: Handles decrypting incoming packets in parallel on
multiple cores, before passing them off in order to be ingested via
the rest of the networking subsystem with GRO via the typical NAPI
poll function. Also handles receiving handshake and cookie messages
as part of the protocol, in parallel.
* timers.[ch]: Uses the timer wheel to implement protocol particular
event timeouts, and gives a set of very simple event-driven entry
point functions for callers.
* main.c, version.h: Initialization and deinitialization of the module.
* selftest/*.h: Runtime unit tests for some of the most security
sensitive functions.
* tools/testing/selftests/wireguard/netns.sh: Aforementioned testing
script using network namespaces.
This commit aims to be as self-contained as possible, implementing
WireGuard as a standalone module not needing much special handling or
coordination from the network subsystem. I expect for future
optimizations to the network stack to positively improve WireGuard, and
vice-versa, but for the time being, this exists as intentionally
standalone.
We introduce a menu option for CONFIG_WIREGUARD, as well as providing a
verbose debug log and self-tests via CONFIG_WIREGUARD_DEBUG.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: David Miller <davem@davemloft.net>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: linux-crypto@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-09 07:27:34 +08:00
#!/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
2020-02-12 03:47:08 +08:00
export LANG = C
net: WireGuard secure network tunnel
WireGuard is a layer 3 secure networking tunnel made specifically for
the kernel, that aims to be much simpler and easier to audit than IPsec.
Extensive documentation and description of the protocol and
considerations, along with formal proofs of the cryptography, are
available at:
* https://www.wireguard.com/
* https://www.wireguard.com/papers/wireguard.pdf
This commit implements WireGuard as a simple network device driver,
accessible in the usual RTNL way used by virtual network drivers. It
makes use of the udp_tunnel APIs, GRO, GSO, NAPI, and the usual set of
networking subsystem APIs. It has a somewhat novel multicore queueing
system designed for maximum throughput and minimal latency of encryption
operations, but it is implemented modestly using workqueues and NAPI.
Configuration is done via generic Netlink, and following a review from
the Netlink maintainer a year ago, several high profile userspace tools
have already implemented the API.
This commit also comes with several different tests, both in-kernel
tests and out-of-kernel tests based on network namespaces, taking profit
of the fact that sockets used by WireGuard intentionally stay in the
namespace the WireGuard interface was originally created, exactly like
the semantics of userspace tun devices. See wireguard.com/netns/ for
pictures and examples.
The source code is fairly short, but rather than combining everything
into a single file, WireGuard is developed as cleanly separable files,
making auditing and comprehension easier. Things are laid out as
follows:
* noise.[ch], cookie.[ch], messages.h: These implement the bulk of the
cryptographic aspects of the protocol, and are mostly data-only in
nature, taking in buffers of bytes and spitting out buffers of
bytes. They also handle reference counting for their various shared
pieces of data, like keys and key lists.
* ratelimiter.[ch]: Used as an integral part of cookie.[ch] for
ratelimiting certain types of cryptographic operations in accordance
with particular WireGuard semantics.
* allowedips.[ch], peerlookup.[ch]: The main lookup structures of
WireGuard, the former being trie-like with particular semantics, an
integral part of the design of the protocol, and the latter just
being nice helper functions around the various hashtables we use.
* device.[ch]: Implementation of functions for the netdevice and for
rtnl, responsible for maintaining the life of a given interface and
wiring it up to the rest of WireGuard.
* peer.[ch]: Each interface has a list of peers, with helper functions
available here for creation, destruction, and reference counting.
* socket.[ch]: Implementation of functions related to udp_socket and
the general set of kernel socket APIs, for sending and receiving
ciphertext UDP packets, and taking care of WireGuard-specific sticky
socket routing semantics for the automatic roaming.
* netlink.[ch]: Userspace API entry point for configuring WireGuard
peers and devices. The API has been implemented by several userspace
tools and network management utility, and the WireGuard project
distributes the basic wg(8) tool.
* queueing.[ch]: Shared function on the rx and tx path for handling
the various queues used in the multicore algorithms.
* send.c: Handles encrypting outgoing packets in parallel on
multiple cores, before sending them in order on a single core, via
workqueues and ring buffers. Also handles sending handshake and cookie
messages as part of the protocol, in parallel.
* receive.c: Handles decrypting incoming packets in parallel on
multiple cores, before passing them off in order to be ingested via
the rest of the networking subsystem with GRO via the typical NAPI
poll function. Also handles receiving handshake and cookie messages
as part of the protocol, in parallel.
* timers.[ch]: Uses the timer wheel to implement protocol particular
event timeouts, and gives a set of very simple event-driven entry
point functions for callers.
* main.c, version.h: Initialization and deinitialization of the module.
* selftest/*.h: Runtime unit tests for some of the most security
sensitive functions.
* tools/testing/selftests/wireguard/netns.sh: Aforementioned testing
script using network namespaces.
This commit aims to be as self-contained as possible, implementing
WireGuard as a standalone module not needing much special handling or
coordination from the network subsystem. I expect for future
optimizations to the network stack to positively improve WireGuard, and
vice-versa, but for the time being, this exists as intentionally
standalone.
We introduce a menu option for CONFIG_WIREGUARD, as well as providing a
verbose debug log and self-tests via CONFIG_WIREGUARD_DEBUG.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: David Miller <davem@davemloft.net>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: linux-crypto@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-09 07:27:34 +08:00
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 " $@ " ; }
2020-01-03 00:47:49 +08:00
sleep( ) { read -t " $1 " -N 1 || true; }
2020-02-05 05:17:29 +08:00
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 ; }
net: WireGuard secure network tunnel
WireGuard is a layer 3 secure networking tunnel made specifically for
the kernel, that aims to be much simpler and easier to audit than IPsec.
Extensive documentation and description of the protocol and
considerations, along with formal proofs of the cryptography, are
available at:
* https://www.wireguard.com/
* https://www.wireguard.com/papers/wireguard.pdf
This commit implements WireGuard as a simple network device driver,
accessible in the usual RTNL way used by virtual network drivers. It
makes use of the udp_tunnel APIs, GRO, GSO, NAPI, and the usual set of
networking subsystem APIs. It has a somewhat novel multicore queueing
system designed for maximum throughput and minimal latency of encryption
operations, but it is implemented modestly using workqueues and NAPI.
Configuration is done via generic Netlink, and following a review from
the Netlink maintainer a year ago, several high profile userspace tools
have already implemented the API.
This commit also comes with several different tests, both in-kernel
tests and out-of-kernel tests based on network namespaces, taking profit
of the fact that sockets used by WireGuard intentionally stay in the
namespace the WireGuard interface was originally created, exactly like
the semantics of userspace tun devices. See wireguard.com/netns/ for
pictures and examples.
The source code is fairly short, but rather than combining everything
into a single file, WireGuard is developed as cleanly separable files,
making auditing and comprehension easier. Things are laid out as
follows:
* noise.[ch], cookie.[ch], messages.h: These implement the bulk of the
cryptographic aspects of the protocol, and are mostly data-only in
nature, taking in buffers of bytes and spitting out buffers of
bytes. They also handle reference counting for their various shared
pieces of data, like keys and key lists.
* ratelimiter.[ch]: Used as an integral part of cookie.[ch] for
ratelimiting certain types of cryptographic operations in accordance
with particular WireGuard semantics.
* allowedips.[ch], peerlookup.[ch]: The main lookup structures of
WireGuard, the former being trie-like with particular semantics, an
integral part of the design of the protocol, and the latter just
being nice helper functions around the various hashtables we use.
* device.[ch]: Implementation of functions for the netdevice and for
rtnl, responsible for maintaining the life of a given interface and
wiring it up to the rest of WireGuard.
* peer.[ch]: Each interface has a list of peers, with helper functions
available here for creation, destruction, and reference counting.
* socket.[ch]: Implementation of functions related to udp_socket and
the general set of kernel socket APIs, for sending and receiving
ciphertext UDP packets, and taking care of WireGuard-specific sticky
socket routing semantics for the automatic roaming.
* netlink.[ch]: Userspace API entry point for configuring WireGuard
peers and devices. The API has been implemented by several userspace
tools and network management utility, and the WireGuard project
distributes the basic wg(8) tool.
* queueing.[ch]: Shared function on the rx and tx path for handling
the various queues used in the multicore algorithms.
* send.c: Handles encrypting outgoing packets in parallel on
multiple cores, before sending them in order on a single core, via
workqueues and ring buffers. Also handles sending handshake and cookie
messages as part of the protocol, in parallel.
* receive.c: Handles decrypting incoming packets in parallel on
multiple cores, before passing them off in order to be ingested via
the rest of the networking subsystem with GRO via the typical NAPI
poll function. Also handles receiving handshake and cookie messages
as part of the protocol, in parallel.
* timers.[ch]: Uses the timer wheel to implement protocol particular
event timeouts, and gives a set of very simple event-driven entry
point functions for callers.
* main.c, version.h: Initialization and deinitialization of the module.
* selftest/*.h: Runtime unit tests for some of the most security
sensitive functions.
* tools/testing/selftests/wireguard/netns.sh: Aforementioned testing
script using network namespaces.
This commit aims to be as self-contained as possible, implementing
WireGuard as a standalone module not needing much special handling or
coordination from the network subsystem. I expect for future
optimizations to the network stack to positively improve WireGuard, and
vice-versa, but for the time being, this exists as intentionally
standalone.
We introduce a menu option for CONFIG_WIREGUARD, as well as providing a
verbose debug log and self-tests via CONFIG_WIREGUARD_DEBUG.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: David Miller <davem@davemloft.net>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: linux-crypto@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-09 07:27:34 +08:00
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
wireguard: socket: remove errant restriction on looping to self
It's already possible to create two different interfaces and loop
packets between them. This has always been possible with tunnels in the
kernel, and isn't specific to wireguard. Therefore, the networking stack
already needs to deal with that. At the very least, the packet winds up
exceeding the MTU and is discarded at that point. So, since this is
already something that happens, there's no need to forbid the not very
exceptional case of routing a packet back to the same interface; this
loop is no different than others, and we shouldn't special case it, but
rather rely on generic handling of loops in general. This also makes it
easier to do interesting things with wireguard such as onion routing.
At the same time, we add a selftest for this, ensuring that both onion
routing works and infinite routing loops do not crash the kernel. We
also add a test case for wireguard interfaces nesting packets and
sending traffic between each other, as well as the loop in this case
too. We make sure to send some throughput-heavy traffic for this use
case, to stress out any possible recursion issues with the locks around
workqueues.
Fixes: e7096c131e51 ("net: WireGuard secure network tunnel")
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2020-05-07 05:33:03 +08:00
ip0 link del dev wg1
net: WireGuard secure network tunnel
WireGuard is a layer 3 secure networking tunnel made specifically for
the kernel, that aims to be much simpler and easier to audit than IPsec.
Extensive documentation and description of the protocol and
considerations, along with formal proofs of the cryptography, are
available at:
* https://www.wireguard.com/
* https://www.wireguard.com/papers/wireguard.pdf
This commit implements WireGuard as a simple network device driver,
accessible in the usual RTNL way used by virtual network drivers. It
makes use of the udp_tunnel APIs, GRO, GSO, NAPI, and the usual set of
networking subsystem APIs. It has a somewhat novel multicore queueing
system designed for maximum throughput and minimal latency of encryption
operations, but it is implemented modestly using workqueues and NAPI.
Configuration is done via generic Netlink, and following a review from
the Netlink maintainer a year ago, several high profile userspace tools
have already implemented the API.
This commit also comes with several different tests, both in-kernel
tests and out-of-kernel tests based on network namespaces, taking profit
of the fact that sockets used by WireGuard intentionally stay in the
namespace the WireGuard interface was originally created, exactly like
the semantics of userspace tun devices. See wireguard.com/netns/ for
pictures and examples.
The source code is fairly short, but rather than combining everything
into a single file, WireGuard is developed as cleanly separable files,
making auditing and comprehension easier. Things are laid out as
follows:
* noise.[ch], cookie.[ch], messages.h: These implement the bulk of the
cryptographic aspects of the protocol, and are mostly data-only in
nature, taking in buffers of bytes and spitting out buffers of
bytes. They also handle reference counting for their various shared
pieces of data, like keys and key lists.
* ratelimiter.[ch]: Used as an integral part of cookie.[ch] for
ratelimiting certain types of cryptographic operations in accordance
with particular WireGuard semantics.
* allowedips.[ch], peerlookup.[ch]: The main lookup structures of
WireGuard, the former being trie-like with particular semantics, an
integral part of the design of the protocol, and the latter just
being nice helper functions around the various hashtables we use.
* device.[ch]: Implementation of functions for the netdevice and for
rtnl, responsible for maintaining the life of a given interface and
wiring it up to the rest of WireGuard.
* peer.[ch]: Each interface has a list of peers, with helper functions
available here for creation, destruction, and reference counting.
* socket.[ch]: Implementation of functions related to udp_socket and
the general set of kernel socket APIs, for sending and receiving
ciphertext UDP packets, and taking care of WireGuard-specific sticky
socket routing semantics for the automatic roaming.
* netlink.[ch]: Userspace API entry point for configuring WireGuard
peers and devices. The API has been implemented by several userspace
tools and network management utility, and the WireGuard project
distributes the basic wg(8) tool.
* queueing.[ch]: Shared function on the rx and tx path for handling
the various queues used in the multicore algorithms.
* send.c: Handles encrypting outgoing packets in parallel on
multiple cores, before sending them in order on a single core, via
workqueues and ring buffers. Also handles sending handshake and cookie
messages as part of the protocol, in parallel.
* receive.c: Handles decrypting incoming packets in parallel on
multiple cores, before passing them off in order to be ingested via
the rest of the networking subsystem with GRO via the typical NAPI
poll function. Also handles receiving handshake and cookie messages
as part of the protocol, in parallel.
* timers.[ch]: Uses the timer wheel to implement protocol particular
event timeouts, and gives a set of very simple event-driven entry
point functions for callers.
* main.c, version.h: Initialization and deinitialization of the module.
* selftest/*.h: Runtime unit tests for some of the most security
sensitive functions.
* tools/testing/selftests/wireguard/netns.sh: Aforementioned testing
script using network namespaces.
This commit aims to be as self-contained as possible, implementing
WireGuard as a standalone module not needing much special handling or
coordination from the network subsystem. I expect for future
optimizations to the network stack to positively improve WireGuard, and
vice-versa, but for the time being, this exists as intentionally
standalone.
We introduce a menu option for CONFIG_WIREGUARD, as well as providing a
verbose debug log and self-tests via CONFIG_WIREGUARD_DEBUG.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: David Miller <davem@davemloft.net>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: linux-crypto@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-09 07:27:34 +08:00
ip1 link del dev wg0
wireguard: socket: remove errant restriction on looping to self
It's already possible to create two different interfaces and loop
packets between them. This has always been possible with tunnels in the
kernel, and isn't specific to wireguard. Therefore, the networking stack
already needs to deal with that. At the very least, the packet winds up
exceeding the MTU and is discarded at that point. So, since this is
already something that happens, there's no need to forbid the not very
exceptional case of routing a packet back to the same interface; this
loop is no different than others, and we shouldn't special case it, but
rather rely on generic handling of loops in general. This also makes it
easier to do interesting things with wireguard such as onion routing.
At the same time, we add a selftest for this, ensuring that both onion
routing works and infinite routing loops do not crash the kernel. We
also add a test case for wireguard interfaces nesting packets and
sending traffic between each other, as well as the loop in this case
too. We make sure to send some throughput-heavy traffic for this use
case, to stress out any possible recursion issues with the locks around
workqueues.
Fixes: e7096c131e51 ("net: WireGuard secure network tunnel")
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2020-05-07 05:33:03 +08:00
ip1 link del dev wg1
net: WireGuard secure network tunnel
WireGuard is a layer 3 secure networking tunnel made specifically for
the kernel, that aims to be much simpler and easier to audit than IPsec.
Extensive documentation and description of the protocol and
considerations, along with formal proofs of the cryptography, are
available at:
* https://www.wireguard.com/
* https://www.wireguard.com/papers/wireguard.pdf
This commit implements WireGuard as a simple network device driver,
accessible in the usual RTNL way used by virtual network drivers. It
makes use of the udp_tunnel APIs, GRO, GSO, NAPI, and the usual set of
networking subsystem APIs. It has a somewhat novel multicore queueing
system designed for maximum throughput and minimal latency of encryption
operations, but it is implemented modestly using workqueues and NAPI.
Configuration is done via generic Netlink, and following a review from
the Netlink maintainer a year ago, several high profile userspace tools
have already implemented the API.
This commit also comes with several different tests, both in-kernel
tests and out-of-kernel tests based on network namespaces, taking profit
of the fact that sockets used by WireGuard intentionally stay in the
namespace the WireGuard interface was originally created, exactly like
the semantics of userspace tun devices. See wireguard.com/netns/ for
pictures and examples.
The source code is fairly short, but rather than combining everything
into a single file, WireGuard is developed as cleanly separable files,
making auditing and comprehension easier. Things are laid out as
follows:
* noise.[ch], cookie.[ch], messages.h: These implement the bulk of the
cryptographic aspects of the protocol, and are mostly data-only in
nature, taking in buffers of bytes and spitting out buffers of
bytes. They also handle reference counting for their various shared
pieces of data, like keys and key lists.
* ratelimiter.[ch]: Used as an integral part of cookie.[ch] for
ratelimiting certain types of cryptographic operations in accordance
with particular WireGuard semantics.
* allowedips.[ch], peerlookup.[ch]: The main lookup structures of
WireGuard, the former being trie-like with particular semantics, an
integral part of the design of the protocol, and the latter just
being nice helper functions around the various hashtables we use.
* device.[ch]: Implementation of functions for the netdevice and for
rtnl, responsible for maintaining the life of a given interface and
wiring it up to the rest of WireGuard.
* peer.[ch]: Each interface has a list of peers, with helper functions
available here for creation, destruction, and reference counting.
* socket.[ch]: Implementation of functions related to udp_socket and
the general set of kernel socket APIs, for sending and receiving
ciphertext UDP packets, and taking care of WireGuard-specific sticky
socket routing semantics for the automatic roaming.
* netlink.[ch]: Userspace API entry point for configuring WireGuard
peers and devices. The API has been implemented by several userspace
tools and network management utility, and the WireGuard project
distributes the basic wg(8) tool.
* queueing.[ch]: Shared function on the rx and tx path for handling
the various queues used in the multicore algorithms.
* send.c: Handles encrypting outgoing packets in parallel on
multiple cores, before sending them in order on a single core, via
workqueues and ring buffers. Also handles sending handshake and cookie
messages as part of the protocol, in parallel.
* receive.c: Handles decrypting incoming packets in parallel on
multiple cores, before passing them off in order to be ingested via
the rest of the networking subsystem with GRO via the typical NAPI
poll function. Also handles receiving handshake and cookie messages
as part of the protocol, in parallel.
* timers.[ch]: Uses the timer wheel to implement protocol particular
event timeouts, and gives a set of very simple event-driven entry
point functions for callers.
* main.c, version.h: Initialization and deinitialization of the module.
* selftest/*.h: Runtime unit tests for some of the most security
sensitive functions.
* tools/testing/selftests/wireguard/netns.sh: Aforementioned testing
script using network namespaces.
This commit aims to be as self-contained as possible, implementing
WireGuard as a standalone module not needing much special handling or
coordination from the network subsystem. I expect for future
optimizations to the network stack to positively improve WireGuard, and
vice-versa, but for the time being, this exists as intentionally
standalone.
We introduce a menu option for CONFIG_WIREGUARD, as well as providing a
verbose debug log and self-tests via CONFIG_WIREGUARD_DEBUG.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: David Miller <davem@davemloft.net>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: linux-crypto@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-09 07:27:34 +08:00
ip2 link del dev wg0
wireguard: socket: remove errant restriction on looping to self
It's already possible to create two different interfaces and loop
packets between them. This has always been possible with tunnels in the
kernel, and isn't specific to wireguard. Therefore, the networking stack
already needs to deal with that. At the very least, the packet winds up
exceeding the MTU and is discarded at that point. So, since this is
already something that happens, there's no need to forbid the not very
exceptional case of routing a packet back to the same interface; this
loop is no different than others, and we shouldn't special case it, but
rather rely on generic handling of loops in general. This also makes it
easier to do interesting things with wireguard such as onion routing.
At the same time, we add a selftest for this, ensuring that both onion
routing works and infinite routing loops do not crash the kernel. We
also add a test case for wireguard interfaces nesting packets and
sending traffic between each other, as well as the loop in this case
too. We make sure to send some throughput-heavy traffic for this use
case, to stress out any possible recursion issues with the locks around
workqueues.
Fixes: e7096c131e51 ("net: WireGuard secure network tunnel")
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2020-05-07 05:33:03 +08:00
ip2 link del dev wg1
net: WireGuard secure network tunnel
WireGuard is a layer 3 secure networking tunnel made specifically for
the kernel, that aims to be much simpler and easier to audit than IPsec.
Extensive documentation and description of the protocol and
considerations, along with formal proofs of the cryptography, are
available at:
* https://www.wireguard.com/
* https://www.wireguard.com/papers/wireguard.pdf
This commit implements WireGuard as a simple network device driver,
accessible in the usual RTNL way used by virtual network drivers. It
makes use of the udp_tunnel APIs, GRO, GSO, NAPI, and the usual set of
networking subsystem APIs. It has a somewhat novel multicore queueing
system designed for maximum throughput and minimal latency of encryption
operations, but it is implemented modestly using workqueues and NAPI.
Configuration is done via generic Netlink, and following a review from
the Netlink maintainer a year ago, several high profile userspace tools
have already implemented the API.
This commit also comes with several different tests, both in-kernel
tests and out-of-kernel tests based on network namespaces, taking profit
of the fact that sockets used by WireGuard intentionally stay in the
namespace the WireGuard interface was originally created, exactly like
the semantics of userspace tun devices. See wireguard.com/netns/ for
pictures and examples.
The source code is fairly short, but rather than combining everything
into a single file, WireGuard is developed as cleanly separable files,
making auditing and comprehension easier. Things are laid out as
follows:
* noise.[ch], cookie.[ch], messages.h: These implement the bulk of the
cryptographic aspects of the protocol, and are mostly data-only in
nature, taking in buffers of bytes and spitting out buffers of
bytes. They also handle reference counting for their various shared
pieces of data, like keys and key lists.
* ratelimiter.[ch]: Used as an integral part of cookie.[ch] for
ratelimiting certain types of cryptographic operations in accordance
with particular WireGuard semantics.
* allowedips.[ch], peerlookup.[ch]: The main lookup structures of
WireGuard, the former being trie-like with particular semantics, an
integral part of the design of the protocol, and the latter just
being nice helper functions around the various hashtables we use.
* device.[ch]: Implementation of functions for the netdevice and for
rtnl, responsible for maintaining the life of a given interface and
wiring it up to the rest of WireGuard.
* peer.[ch]: Each interface has a list of peers, with helper functions
available here for creation, destruction, and reference counting.
* socket.[ch]: Implementation of functions related to udp_socket and
the general set of kernel socket APIs, for sending and receiving
ciphertext UDP packets, and taking care of WireGuard-specific sticky
socket routing semantics for the automatic roaming.
* netlink.[ch]: Userspace API entry point for configuring WireGuard
peers and devices. The API has been implemented by several userspace
tools and network management utility, and the WireGuard project
distributes the basic wg(8) tool.
* queueing.[ch]: Shared function on the rx and tx path for handling
the various queues used in the multicore algorithms.
* send.c: Handles encrypting outgoing packets in parallel on
multiple cores, before sending them in order on a single core, via
workqueues and ring buffers. Also handles sending handshake and cookie
messages as part of the protocol, in parallel.
* receive.c: Handles decrypting incoming packets in parallel on
multiple cores, before passing them off in order to be ingested via
the rest of the networking subsystem with GRO via the typical NAPI
poll function. Also handles receiving handshake and cookie messages
as part of the protocol, in parallel.
* timers.[ch]: Uses the timer wheel to implement protocol particular
event timeouts, and gives a set of very simple event-driven entry
point functions for callers.
* main.c, version.h: Initialization and deinitialization of the module.
* selftest/*.h: Runtime unit tests for some of the most security
sensitive functions.
* tools/testing/selftests/wireguard/netns.sh: Aforementioned testing
script using network namespaces.
This commit aims to be as self-contained as possible, implementing
WireGuard as a standalone module not needing much special handling or
coordination from the network subsystem. I expect for future
optimizations to the network stack to positively improve WireGuard, and
vice-versa, but for the time being, this exists as intentionally
standalone.
We introduce a menu option for CONFIG_WIREGUARD, as well as providing a
verbose debug log and self-tests via CONFIG_WIREGUARD_DEBUG.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: David Miller <davem@davemloft.net>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: linux-crypto@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-09 07:27:34 +08:00
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) "
wireguard: socket: remove errant restriction on looping to self
It's already possible to create two different interfaces and loop
packets between them. This has always been possible with tunnels in the
kernel, and isn't specific to wireguard. Therefore, the networking stack
already needs to deal with that. At the very least, the packet winds up
exceeding the MTU and is discarded at that point. So, since this is
already something that happens, there's no need to forbid the not very
exceptional case of routing a packet back to the same interface; this
loop is no different than others, and we shouldn't special case it, but
rather rely on generic handling of loops in general. This also makes it
easier to do interesting things with wireguard such as onion routing.
At the same time, we add a selftest for this, ensuring that both onion
routing works and infinite routing loops do not crash the kernel. We
also add a test case for wireguard interfaces nesting packets and
sending traffic between each other, as well as the loop in this case
too. We make sure to send some throughput-heavy traffic for this use
case, to stress out any possible recursion issues with the locks around
workqueues.
Fixes: e7096c131e51 ("net: WireGuard secure network tunnel")
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2020-05-07 05:33:03 +08:00
key4 = " $( pp wg genkey) "
net: WireGuard secure network tunnel
WireGuard is a layer 3 secure networking tunnel made specifically for
the kernel, that aims to be much simpler and easier to audit than IPsec.
Extensive documentation and description of the protocol and
considerations, along with formal proofs of the cryptography, are
available at:
* https://www.wireguard.com/
* https://www.wireguard.com/papers/wireguard.pdf
This commit implements WireGuard as a simple network device driver,
accessible in the usual RTNL way used by virtual network drivers. It
makes use of the udp_tunnel APIs, GRO, GSO, NAPI, and the usual set of
networking subsystem APIs. It has a somewhat novel multicore queueing
system designed for maximum throughput and minimal latency of encryption
operations, but it is implemented modestly using workqueues and NAPI.
Configuration is done via generic Netlink, and following a review from
the Netlink maintainer a year ago, several high profile userspace tools
have already implemented the API.
This commit also comes with several different tests, both in-kernel
tests and out-of-kernel tests based on network namespaces, taking profit
of the fact that sockets used by WireGuard intentionally stay in the
namespace the WireGuard interface was originally created, exactly like
the semantics of userspace tun devices. See wireguard.com/netns/ for
pictures and examples.
The source code is fairly short, but rather than combining everything
into a single file, WireGuard is developed as cleanly separable files,
making auditing and comprehension easier. Things are laid out as
follows:
* noise.[ch], cookie.[ch], messages.h: These implement the bulk of the
cryptographic aspects of the protocol, and are mostly data-only in
nature, taking in buffers of bytes and spitting out buffers of
bytes. They also handle reference counting for their various shared
pieces of data, like keys and key lists.
* ratelimiter.[ch]: Used as an integral part of cookie.[ch] for
ratelimiting certain types of cryptographic operations in accordance
with particular WireGuard semantics.
* allowedips.[ch], peerlookup.[ch]: The main lookup structures of
WireGuard, the former being trie-like with particular semantics, an
integral part of the design of the protocol, and the latter just
being nice helper functions around the various hashtables we use.
* device.[ch]: Implementation of functions for the netdevice and for
rtnl, responsible for maintaining the life of a given interface and
wiring it up to the rest of WireGuard.
* peer.[ch]: Each interface has a list of peers, with helper functions
available here for creation, destruction, and reference counting.
* socket.[ch]: Implementation of functions related to udp_socket and
the general set of kernel socket APIs, for sending and receiving
ciphertext UDP packets, and taking care of WireGuard-specific sticky
socket routing semantics for the automatic roaming.
* netlink.[ch]: Userspace API entry point for configuring WireGuard
peers and devices. The API has been implemented by several userspace
tools and network management utility, and the WireGuard project
distributes the basic wg(8) tool.
* queueing.[ch]: Shared function on the rx and tx path for handling
the various queues used in the multicore algorithms.
* send.c: Handles encrypting outgoing packets in parallel on
multiple cores, before sending them in order on a single core, via
workqueues and ring buffers. Also handles sending handshake and cookie
messages as part of the protocol, in parallel.
* receive.c: Handles decrypting incoming packets in parallel on
multiple cores, before passing them off in order to be ingested via
the rest of the networking subsystem with GRO via the typical NAPI
poll function. Also handles receiving handshake and cookie messages
as part of the protocol, in parallel.
* timers.[ch]: Uses the timer wheel to implement protocol particular
event timeouts, and gives a set of very simple event-driven entry
point functions for callers.
* main.c, version.h: Initialization and deinitialization of the module.
* selftest/*.h: Runtime unit tests for some of the most security
sensitive functions.
* tools/testing/selftests/wireguard/netns.sh: Aforementioned testing
script using network namespaces.
This commit aims to be as self-contained as possible, implementing
WireGuard as a standalone module not needing much special handling or
coordination from the network subsystem. I expect for future
optimizations to the network stack to positively improve WireGuard, and
vice-versa, but for the time being, this exists as intentionally
standalone.
We introduce a menu option for CONFIG_WIREGUARD, as well as providing a
verbose debug log and self-tests via CONFIG_WIREGUARD_DEBUG.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: David Miller <davem@davemloft.net>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: linux-crypto@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-09 07:27:34 +08:00
pub1 = " $( pp wg pubkey <<< " $key1 " ) "
pub2 = " $( pp wg pubkey <<< " $key2 " ) "
pub3 = " $( pp wg pubkey <<< " $key3 " ) "
wireguard: socket: remove errant restriction on looping to self
It's already possible to create two different interfaces and loop
packets between them. This has always been possible with tunnels in the
kernel, and isn't specific to wireguard. Therefore, the networking stack
already needs to deal with that. At the very least, the packet winds up
exceeding the MTU and is discarded at that point. So, since this is
already something that happens, there's no need to forbid the not very
exceptional case of routing a packet back to the same interface; this
loop is no different than others, and we shouldn't special case it, but
rather rely on generic handling of loops in general. This also makes it
easier to do interesting things with wireguard such as onion routing.
At the same time, we add a selftest for this, ensuring that both onion
routing works and infinite routing loops do not crash the kernel. We
also add a test case for wireguard interfaces nesting packets and
sending traffic between each other, as well as the loop in this case
too. We make sure to send some throughput-heavy traffic for this use
case, to stress out any possible recursion issues with the locks around
workqueues.
Fixes: e7096c131e51 ("net: WireGuard secure network tunnel")
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2020-05-07 05:33:03 +08:00
pub4 = " $( pp wg pubkey <<< " $key4 " ) "
net: WireGuard secure network tunnel
WireGuard is a layer 3 secure networking tunnel made specifically for
the kernel, that aims to be much simpler and easier to audit than IPsec.
Extensive documentation and description of the protocol and
considerations, along with formal proofs of the cryptography, are
available at:
* https://www.wireguard.com/
* https://www.wireguard.com/papers/wireguard.pdf
This commit implements WireGuard as a simple network device driver,
accessible in the usual RTNL way used by virtual network drivers. It
makes use of the udp_tunnel APIs, GRO, GSO, NAPI, and the usual set of
networking subsystem APIs. It has a somewhat novel multicore queueing
system designed for maximum throughput and minimal latency of encryption
operations, but it is implemented modestly using workqueues and NAPI.
Configuration is done via generic Netlink, and following a review from
the Netlink maintainer a year ago, several high profile userspace tools
have already implemented the API.
This commit also comes with several different tests, both in-kernel
tests and out-of-kernel tests based on network namespaces, taking profit
of the fact that sockets used by WireGuard intentionally stay in the
namespace the WireGuard interface was originally created, exactly like
the semantics of userspace tun devices. See wireguard.com/netns/ for
pictures and examples.
The source code is fairly short, but rather than combining everything
into a single file, WireGuard is developed as cleanly separable files,
making auditing and comprehension easier. Things are laid out as
follows:
* noise.[ch], cookie.[ch], messages.h: These implement the bulk of the
cryptographic aspects of the protocol, and are mostly data-only in
nature, taking in buffers of bytes and spitting out buffers of
bytes. They also handle reference counting for their various shared
pieces of data, like keys and key lists.
* ratelimiter.[ch]: Used as an integral part of cookie.[ch] for
ratelimiting certain types of cryptographic operations in accordance
with particular WireGuard semantics.
* allowedips.[ch], peerlookup.[ch]: The main lookup structures of
WireGuard, the former being trie-like with particular semantics, an
integral part of the design of the protocol, and the latter just
being nice helper functions around the various hashtables we use.
* device.[ch]: Implementation of functions for the netdevice and for
rtnl, responsible for maintaining the life of a given interface and
wiring it up to the rest of WireGuard.
* peer.[ch]: Each interface has a list of peers, with helper functions
available here for creation, destruction, and reference counting.
* socket.[ch]: Implementation of functions related to udp_socket and
the general set of kernel socket APIs, for sending and receiving
ciphertext UDP packets, and taking care of WireGuard-specific sticky
socket routing semantics for the automatic roaming.
* netlink.[ch]: Userspace API entry point for configuring WireGuard
peers and devices. The API has been implemented by several userspace
tools and network management utility, and the WireGuard project
distributes the basic wg(8) tool.
* queueing.[ch]: Shared function on the rx and tx path for handling
the various queues used in the multicore algorithms.
* send.c: Handles encrypting outgoing packets in parallel on
multiple cores, before sending them in order on a single core, via
workqueues and ring buffers. Also handles sending handshake and cookie
messages as part of the protocol, in parallel.
* receive.c: Handles decrypting incoming packets in parallel on
multiple cores, before passing them off in order to be ingested via
the rest of the networking subsystem with GRO via the typical NAPI
poll function. Also handles receiving handshake and cookie messages
as part of the protocol, in parallel.
* timers.[ch]: Uses the timer wheel to implement protocol particular
event timeouts, and gives a set of very simple event-driven entry
point functions for callers.
* main.c, version.h: Initialization and deinitialization of the module.
* selftest/*.h: Runtime unit tests for some of the most security
sensitive functions.
* tools/testing/selftests/wireguard/netns.sh: Aforementioned testing
script using network namespaces.
This commit aims to be as self-contained as possible, implementing
WireGuard as a standalone module not needing much special handling or
coordination from the network subsystem. I expect for future
optimizations to the network stack to positively improve WireGuard, and
vice-versa, but for the time being, this exists as intentionally
standalone.
We introduce a menu option for CONFIG_WIREGUARD, as well as providing a
verbose debug log and self-tests via CONFIG_WIREGUARD_DEBUG.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: David Miller <davem@davemloft.net>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: linux-crypto@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-09 07:27:34 +08:00
psk = " $( pp wg genpsk) "
[ [ -n $key1 && -n $key2 && -n $psk ] ]
configure_peers( ) {
ip1 addr add 192.168.241.1/24 dev wg0
wireguard: socket: remove errant restriction on looping to self
It's already possible to create two different interfaces and loop
packets between them. This has always been possible with tunnels in the
kernel, and isn't specific to wireguard. Therefore, the networking stack
already needs to deal with that. At the very least, the packet winds up
exceeding the MTU and is discarded at that point. So, since this is
already something that happens, there's no need to forbid the not very
exceptional case of routing a packet back to the same interface; this
loop is no different than others, and we shouldn't special case it, but
rather rely on generic handling of loops in general. This also makes it
easier to do interesting things with wireguard such as onion routing.
At the same time, we add a selftest for this, ensuring that both onion
routing works and infinite routing loops do not crash the kernel. We
also add a test case for wireguard interfaces nesting packets and
sending traffic between each other, as well as the loop in this case
too. We make sure to send some throughput-heavy traffic for this use
case, to stress out any possible recursion issues with the locks around
workqueues.
Fixes: e7096c131e51 ("net: WireGuard secure network tunnel")
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2020-05-07 05:33:03 +08:00
ip1 addr add fd00::1/112 dev wg0
net: WireGuard secure network tunnel
WireGuard is a layer 3 secure networking tunnel made specifically for
the kernel, that aims to be much simpler and easier to audit than IPsec.
Extensive documentation and description of the protocol and
considerations, along with formal proofs of the cryptography, are
available at:
* https://www.wireguard.com/
* https://www.wireguard.com/papers/wireguard.pdf
This commit implements WireGuard as a simple network device driver,
accessible in the usual RTNL way used by virtual network drivers. It
makes use of the udp_tunnel APIs, GRO, GSO, NAPI, and the usual set of
networking subsystem APIs. It has a somewhat novel multicore queueing
system designed for maximum throughput and minimal latency of encryption
operations, but it is implemented modestly using workqueues and NAPI.
Configuration is done via generic Netlink, and following a review from
the Netlink maintainer a year ago, several high profile userspace tools
have already implemented the API.
This commit also comes with several different tests, both in-kernel
tests and out-of-kernel tests based on network namespaces, taking profit
of the fact that sockets used by WireGuard intentionally stay in the
namespace the WireGuard interface was originally created, exactly like
the semantics of userspace tun devices. See wireguard.com/netns/ for
pictures and examples.
The source code is fairly short, but rather than combining everything
into a single file, WireGuard is developed as cleanly separable files,
making auditing and comprehension easier. Things are laid out as
follows:
* noise.[ch], cookie.[ch], messages.h: These implement the bulk of the
cryptographic aspects of the protocol, and are mostly data-only in
nature, taking in buffers of bytes and spitting out buffers of
bytes. They also handle reference counting for their various shared
pieces of data, like keys and key lists.
* ratelimiter.[ch]: Used as an integral part of cookie.[ch] for
ratelimiting certain types of cryptographic operations in accordance
with particular WireGuard semantics.
* allowedips.[ch], peerlookup.[ch]: The main lookup structures of
WireGuard, the former being trie-like with particular semantics, an
integral part of the design of the protocol, and the latter just
being nice helper functions around the various hashtables we use.
* device.[ch]: Implementation of functions for the netdevice and for
rtnl, responsible for maintaining the life of a given interface and
wiring it up to the rest of WireGuard.
* peer.[ch]: Each interface has a list of peers, with helper functions
available here for creation, destruction, and reference counting.
* socket.[ch]: Implementation of functions related to udp_socket and
the general set of kernel socket APIs, for sending and receiving
ciphertext UDP packets, and taking care of WireGuard-specific sticky
socket routing semantics for the automatic roaming.
* netlink.[ch]: Userspace API entry point for configuring WireGuard
peers and devices. The API has been implemented by several userspace
tools and network management utility, and the WireGuard project
distributes the basic wg(8) tool.
* queueing.[ch]: Shared function on the rx and tx path for handling
the various queues used in the multicore algorithms.
* send.c: Handles encrypting outgoing packets in parallel on
multiple cores, before sending them in order on a single core, via
workqueues and ring buffers. Also handles sending handshake and cookie
messages as part of the protocol, in parallel.
* receive.c: Handles decrypting incoming packets in parallel on
multiple cores, before passing them off in order to be ingested via
the rest of the networking subsystem with GRO via the typical NAPI
poll function. Also handles receiving handshake and cookie messages
as part of the protocol, in parallel.
* timers.[ch]: Uses the timer wheel to implement protocol particular
event timeouts, and gives a set of very simple event-driven entry
point functions for callers.
* main.c, version.h: Initialization and deinitialization of the module.
* selftest/*.h: Runtime unit tests for some of the most security
sensitive functions.
* tools/testing/selftests/wireguard/netns.sh: Aforementioned testing
script using network namespaces.
This commit aims to be as self-contained as possible, implementing
WireGuard as a standalone module not needing much special handling or
coordination from the network subsystem. I expect for future
optimizations to the network stack to positively improve WireGuard, and
vice-versa, but for the time being, this exists as intentionally
standalone.
We introduce a menu option for CONFIG_WIREGUARD, as well as providing a
verbose debug log and self-tests via CONFIG_WIREGUARD_DEBUG.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: David Miller <davem@davemloft.net>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: linux-crypto@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-09 07:27:34 +08:00
ip2 addr add 192.168.241.2/24 dev wg0
wireguard: socket: remove errant restriction on looping to self
It's already possible to create two different interfaces and loop
packets between them. This has always been possible with tunnels in the
kernel, and isn't specific to wireguard. Therefore, the networking stack
already needs to deal with that. At the very least, the packet winds up
exceeding the MTU and is discarded at that point. So, since this is
already something that happens, there's no need to forbid the not very
exceptional case of routing a packet back to the same interface; this
loop is no different than others, and we shouldn't special case it, but
rather rely on generic handling of loops in general. This also makes it
easier to do interesting things with wireguard such as onion routing.
At the same time, we add a selftest for this, ensuring that both onion
routing works and infinite routing loops do not crash the kernel. We
also add a test case for wireguard interfaces nesting packets and
sending traffic between each other, as well as the loop in this case
too. We make sure to send some throughput-heavy traffic for this use
case, to stress out any possible recursion issues with the locks around
workqueues.
Fixes: e7096c131e51 ("net: WireGuard secure network tunnel")
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2020-05-07 05:33:03 +08:00
ip2 addr add fd00::2/112 dev wg0
net: WireGuard secure network tunnel
WireGuard is a layer 3 secure networking tunnel made specifically for
the kernel, that aims to be much simpler and easier to audit than IPsec.
Extensive documentation and description of the protocol and
considerations, along with formal proofs of the cryptography, are
available at:
* https://www.wireguard.com/
* https://www.wireguard.com/papers/wireguard.pdf
This commit implements WireGuard as a simple network device driver,
accessible in the usual RTNL way used by virtual network drivers. It
makes use of the udp_tunnel APIs, GRO, GSO, NAPI, and the usual set of
networking subsystem APIs. It has a somewhat novel multicore queueing
system designed for maximum throughput and minimal latency of encryption
operations, but it is implemented modestly using workqueues and NAPI.
Configuration is done via generic Netlink, and following a review from
the Netlink maintainer a year ago, several high profile userspace tools
have already implemented the API.
This commit also comes with several different tests, both in-kernel
tests and out-of-kernel tests based on network namespaces, taking profit
of the fact that sockets used by WireGuard intentionally stay in the
namespace the WireGuard interface was originally created, exactly like
the semantics of userspace tun devices. See wireguard.com/netns/ for
pictures and examples.
The source code is fairly short, but rather than combining everything
into a single file, WireGuard is developed as cleanly separable files,
making auditing and comprehension easier. Things are laid out as
follows:
* noise.[ch], cookie.[ch], messages.h: These implement the bulk of the
cryptographic aspects of the protocol, and are mostly data-only in
nature, taking in buffers of bytes and spitting out buffers of
bytes. They also handle reference counting for their various shared
pieces of data, like keys and key lists.
* ratelimiter.[ch]: Used as an integral part of cookie.[ch] for
ratelimiting certain types of cryptographic operations in accordance
with particular WireGuard semantics.
* allowedips.[ch], peerlookup.[ch]: The main lookup structures of
WireGuard, the former being trie-like with particular semantics, an
integral part of the design of the protocol, and the latter just
being nice helper functions around the various hashtables we use.
* device.[ch]: Implementation of functions for the netdevice and for
rtnl, responsible for maintaining the life of a given interface and
wiring it up to the rest of WireGuard.
* peer.[ch]: Each interface has a list of peers, with helper functions
available here for creation, destruction, and reference counting.
* socket.[ch]: Implementation of functions related to udp_socket and
the general set of kernel socket APIs, for sending and receiving
ciphertext UDP packets, and taking care of WireGuard-specific sticky
socket routing semantics for the automatic roaming.
* netlink.[ch]: Userspace API entry point for configuring WireGuard
peers and devices. The API has been implemented by several userspace
tools and network management utility, and the WireGuard project
distributes the basic wg(8) tool.
* queueing.[ch]: Shared function on the rx and tx path for handling
the various queues used in the multicore algorithms.
* send.c: Handles encrypting outgoing packets in parallel on
multiple cores, before sending them in order on a single core, via
workqueues and ring buffers. Also handles sending handshake and cookie
messages as part of the protocol, in parallel.
* receive.c: Handles decrypting incoming packets in parallel on
multiple cores, before passing them off in order to be ingested via
the rest of the networking subsystem with GRO via the typical NAPI
poll function. Also handles receiving handshake and cookie messages
as part of the protocol, in parallel.
* timers.[ch]: Uses the timer wheel to implement protocol particular
event timeouts, and gives a set of very simple event-driven entry
point functions for callers.
* main.c, version.h: Initialization and deinitialization of the module.
* selftest/*.h: Runtime unit tests for some of the most security
sensitive functions.
* tools/testing/selftests/wireguard/netns.sh: Aforementioned testing
script using network namespaces.
This commit aims to be as self-contained as possible, implementing
WireGuard as a standalone module not needing much special handling or
coordination from the network subsystem. I expect for future
optimizations to the network stack to positively improve WireGuard, and
vice-versa, but for the time being, this exists as intentionally
standalone.
We introduce a menu option for CONFIG_WIREGUARD, as well as providing a
verbose debug log and self-tests via CONFIG_WIREGUARD_DEBUG.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: David Miller <davem@davemloft.net>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: linux-crypto@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-09 07:27:34 +08:00
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 &
2020-02-05 05:17:29 +08:00
waitiperf $netns2 $!
net: WireGuard secure network tunnel
WireGuard is a layer 3 secure networking tunnel made specifically for
the kernel, that aims to be much simpler and easier to audit than IPsec.
Extensive documentation and description of the protocol and
considerations, along with formal proofs of the cryptography, are
available at:
* https://www.wireguard.com/
* https://www.wireguard.com/papers/wireguard.pdf
This commit implements WireGuard as a simple network device driver,
accessible in the usual RTNL way used by virtual network drivers. It
makes use of the udp_tunnel APIs, GRO, GSO, NAPI, and the usual set of
networking subsystem APIs. It has a somewhat novel multicore queueing
system designed for maximum throughput and minimal latency of encryption
operations, but it is implemented modestly using workqueues and NAPI.
Configuration is done via generic Netlink, and following a review from
the Netlink maintainer a year ago, several high profile userspace tools
have already implemented the API.
This commit also comes with several different tests, both in-kernel
tests and out-of-kernel tests based on network namespaces, taking profit
of the fact that sockets used by WireGuard intentionally stay in the
namespace the WireGuard interface was originally created, exactly like
the semantics of userspace tun devices. See wireguard.com/netns/ for
pictures and examples.
The source code is fairly short, but rather than combining everything
into a single file, WireGuard is developed as cleanly separable files,
making auditing and comprehension easier. Things are laid out as
follows:
* noise.[ch], cookie.[ch], messages.h: These implement the bulk of the
cryptographic aspects of the protocol, and are mostly data-only in
nature, taking in buffers of bytes and spitting out buffers of
bytes. They also handle reference counting for their various shared
pieces of data, like keys and key lists.
* ratelimiter.[ch]: Used as an integral part of cookie.[ch] for
ratelimiting certain types of cryptographic operations in accordance
with particular WireGuard semantics.
* allowedips.[ch], peerlookup.[ch]: The main lookup structures of
WireGuard, the former being trie-like with particular semantics, an
integral part of the design of the protocol, and the latter just
being nice helper functions around the various hashtables we use.
* device.[ch]: Implementation of functions for the netdevice and for
rtnl, responsible for maintaining the life of a given interface and
wiring it up to the rest of WireGuard.
* peer.[ch]: Each interface has a list of peers, with helper functions
available here for creation, destruction, and reference counting.
* socket.[ch]: Implementation of functions related to udp_socket and
the general set of kernel socket APIs, for sending and receiving
ciphertext UDP packets, and taking care of WireGuard-specific sticky
socket routing semantics for the automatic roaming.
* netlink.[ch]: Userspace API entry point for configuring WireGuard
peers and devices. The API has been implemented by several userspace
tools and network management utility, and the WireGuard project
distributes the basic wg(8) tool.
* queueing.[ch]: Shared function on the rx and tx path for handling
the various queues used in the multicore algorithms.
* send.c: Handles encrypting outgoing packets in parallel on
multiple cores, before sending them in order on a single core, via
workqueues and ring buffers. Also handles sending handshake and cookie
messages as part of the protocol, in parallel.
* receive.c: Handles decrypting incoming packets in parallel on
multiple cores, before passing them off in order to be ingested via
the rest of the networking subsystem with GRO via the typical NAPI
poll function. Also handles receiving handshake and cookie messages
as part of the protocol, in parallel.
* timers.[ch]: Uses the timer wheel to implement protocol particular
event timeouts, and gives a set of very simple event-driven entry
point functions for callers.
* main.c, version.h: Initialization and deinitialization of the module.
* selftest/*.h: Runtime unit tests for some of the most security
sensitive functions.
* tools/testing/selftests/wireguard/netns.sh: Aforementioned testing
script using network namespaces.
This commit aims to be as self-contained as possible, implementing
WireGuard as a standalone module not needing much special handling or
coordination from the network subsystem. I expect for future
optimizations to the network stack to positively improve WireGuard, and
vice-versa, but for the time being, this exists as intentionally
standalone.
We introduce a menu option for CONFIG_WIREGUARD, as well as providing a
verbose debug log and self-tests via CONFIG_WIREGUARD_DEBUG.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: David Miller <davem@davemloft.net>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: linux-crypto@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-09 07:27:34 +08:00
n1 iperf3 -Z -t 3 -c 192.168.241.2
# TCP over IPv6
n1 iperf3 -s -1 -B fd00::1 &
2020-02-05 05:17:29 +08:00
waitiperf $netns1 $!
net: WireGuard secure network tunnel
WireGuard is a layer 3 secure networking tunnel made specifically for
the kernel, that aims to be much simpler and easier to audit than IPsec.
Extensive documentation and description of the protocol and
considerations, along with formal proofs of the cryptography, are
available at:
* https://www.wireguard.com/
* https://www.wireguard.com/papers/wireguard.pdf
This commit implements WireGuard as a simple network device driver,
accessible in the usual RTNL way used by virtual network drivers. It
makes use of the udp_tunnel APIs, GRO, GSO, NAPI, and the usual set of
networking subsystem APIs. It has a somewhat novel multicore queueing
system designed for maximum throughput and minimal latency of encryption
operations, but it is implemented modestly using workqueues and NAPI.
Configuration is done via generic Netlink, and following a review from
the Netlink maintainer a year ago, several high profile userspace tools
have already implemented the API.
This commit also comes with several different tests, both in-kernel
tests and out-of-kernel tests based on network namespaces, taking profit
of the fact that sockets used by WireGuard intentionally stay in the
namespace the WireGuard interface was originally created, exactly like
the semantics of userspace tun devices. See wireguard.com/netns/ for
pictures and examples.
The source code is fairly short, but rather than combining everything
into a single file, WireGuard is developed as cleanly separable files,
making auditing and comprehension easier. Things are laid out as
follows:
* noise.[ch], cookie.[ch], messages.h: These implement the bulk of the
cryptographic aspects of the protocol, and are mostly data-only in
nature, taking in buffers of bytes and spitting out buffers of
bytes. They also handle reference counting for their various shared
pieces of data, like keys and key lists.
* ratelimiter.[ch]: Used as an integral part of cookie.[ch] for
ratelimiting certain types of cryptographic operations in accordance
with particular WireGuard semantics.
* allowedips.[ch], peerlookup.[ch]: The main lookup structures of
WireGuard, the former being trie-like with particular semantics, an
integral part of the design of the protocol, and the latter just
being nice helper functions around the various hashtables we use.
* device.[ch]: Implementation of functions for the netdevice and for
rtnl, responsible for maintaining the life of a given interface and
wiring it up to the rest of WireGuard.
* peer.[ch]: Each interface has a list of peers, with helper functions
available here for creation, destruction, and reference counting.
* socket.[ch]: Implementation of functions related to udp_socket and
the general set of kernel socket APIs, for sending and receiving
ciphertext UDP packets, and taking care of WireGuard-specific sticky
socket routing semantics for the automatic roaming.
* netlink.[ch]: Userspace API entry point for configuring WireGuard
peers and devices. The API has been implemented by several userspace
tools and network management utility, and the WireGuard project
distributes the basic wg(8) tool.
* queueing.[ch]: Shared function on the rx and tx path for handling
the various queues used in the multicore algorithms.
* send.c: Handles encrypting outgoing packets in parallel on
multiple cores, before sending them in order on a single core, via
workqueues and ring buffers. Also handles sending handshake and cookie
messages as part of the protocol, in parallel.
* receive.c: Handles decrypting incoming packets in parallel on
multiple cores, before passing them off in order to be ingested via
the rest of the networking subsystem with GRO via the typical NAPI
poll function. Also handles receiving handshake and cookie messages
as part of the protocol, in parallel.
* timers.[ch]: Uses the timer wheel to implement protocol particular
event timeouts, and gives a set of very simple event-driven entry
point functions for callers.
* main.c, version.h: Initialization and deinitialization of the module.
* selftest/*.h: Runtime unit tests for some of the most security
sensitive functions.
* tools/testing/selftests/wireguard/netns.sh: Aforementioned testing
script using network namespaces.
This commit aims to be as self-contained as possible, implementing
WireGuard as a standalone module not needing much special handling or
coordination from the network subsystem. I expect for future
optimizations to the network stack to positively improve WireGuard, and
vice-versa, but for the time being, this exists as intentionally
standalone.
We introduce a menu option for CONFIG_WIREGUARD, as well as providing a
verbose debug log and self-tests via CONFIG_WIREGUARD_DEBUG.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: David Miller <davem@davemloft.net>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: linux-crypto@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-09 07:27:34 +08:00
n2 iperf3 -Z -t 3 -c fd00::1
# UDP over IPv4
n1 iperf3 -s -1 -B 192.168.241.1 &
2020-02-05 05:17:29 +08:00
waitiperf $netns1 $!
net: WireGuard secure network tunnel
WireGuard is a layer 3 secure networking tunnel made specifically for
the kernel, that aims to be much simpler and easier to audit than IPsec.
Extensive documentation and description of the protocol and
considerations, along with formal proofs of the cryptography, are
available at:
* https://www.wireguard.com/
* https://www.wireguard.com/papers/wireguard.pdf
This commit implements WireGuard as a simple network device driver,
accessible in the usual RTNL way used by virtual network drivers. It
makes use of the udp_tunnel APIs, GRO, GSO, NAPI, and the usual set of
networking subsystem APIs. It has a somewhat novel multicore queueing
system designed for maximum throughput and minimal latency of encryption
operations, but it is implemented modestly using workqueues and NAPI.
Configuration is done via generic Netlink, and following a review from
the Netlink maintainer a year ago, several high profile userspace tools
have already implemented the API.
This commit also comes with several different tests, both in-kernel
tests and out-of-kernel tests based on network namespaces, taking profit
of the fact that sockets used by WireGuard intentionally stay in the
namespace the WireGuard interface was originally created, exactly like
the semantics of userspace tun devices. See wireguard.com/netns/ for
pictures and examples.
The source code is fairly short, but rather than combining everything
into a single file, WireGuard is developed as cleanly separable files,
making auditing and comprehension easier. Things are laid out as
follows:
* noise.[ch], cookie.[ch], messages.h: These implement the bulk of the
cryptographic aspects of the protocol, and are mostly data-only in
nature, taking in buffers of bytes and spitting out buffers of
bytes. They also handle reference counting for their various shared
pieces of data, like keys and key lists.
* ratelimiter.[ch]: Used as an integral part of cookie.[ch] for
ratelimiting certain types of cryptographic operations in accordance
with particular WireGuard semantics.
* allowedips.[ch], peerlookup.[ch]: The main lookup structures of
WireGuard, the former being trie-like with particular semantics, an
integral part of the design of the protocol, and the latter just
being nice helper functions around the various hashtables we use.
* device.[ch]: Implementation of functions for the netdevice and for
rtnl, responsible for maintaining the life of a given interface and
wiring it up to the rest of WireGuard.
* peer.[ch]: Each interface has a list of peers, with helper functions
available here for creation, destruction, and reference counting.
* socket.[ch]: Implementation of functions related to udp_socket and
the general set of kernel socket APIs, for sending and receiving
ciphertext UDP packets, and taking care of WireGuard-specific sticky
socket routing semantics for the automatic roaming.
* netlink.[ch]: Userspace API entry point for configuring WireGuard
peers and devices. The API has been implemented by several userspace
tools and network management utility, and the WireGuard project
distributes the basic wg(8) tool.
* queueing.[ch]: Shared function on the rx and tx path for handling
the various queues used in the multicore algorithms.
* send.c: Handles encrypting outgoing packets in parallel on
multiple cores, before sending them in order on a single core, via
workqueues and ring buffers. Also handles sending handshake and cookie
messages as part of the protocol, in parallel.
* receive.c: Handles decrypting incoming packets in parallel on
multiple cores, before passing them off in order to be ingested via
the rest of the networking subsystem with GRO via the typical NAPI
poll function. Also handles receiving handshake and cookie messages
as part of the protocol, in parallel.
* timers.[ch]: Uses the timer wheel to implement protocol particular
event timeouts, and gives a set of very simple event-driven entry
point functions for callers.
* main.c, version.h: Initialization and deinitialization of the module.
* selftest/*.h: Runtime unit tests for some of the most security
sensitive functions.
* tools/testing/selftests/wireguard/netns.sh: Aforementioned testing
script using network namespaces.
This commit aims to be as self-contained as possible, implementing
WireGuard as a standalone module not needing much special handling or
coordination from the network subsystem. I expect for future
optimizations to the network stack to positively improve WireGuard, and
vice-versa, but for the time being, this exists as intentionally
standalone.
We introduce a menu option for CONFIG_WIREGUARD, as well as providing a
verbose debug log and self-tests via CONFIG_WIREGUARD_DEBUG.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: David Miller <davem@davemloft.net>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: linux-crypto@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-09 07:27:34 +08:00
n2 iperf3 -Z -t 3 -b 0 -u -c 192.168.241.1
# UDP over IPv6
n2 iperf3 -s -1 -B fd00::2 &
2020-02-05 05:17:29 +08:00
waitiperf $netns2 $!
net: WireGuard secure network tunnel
WireGuard is a layer 3 secure networking tunnel made specifically for
the kernel, that aims to be much simpler and easier to audit than IPsec.
Extensive documentation and description of the protocol and
considerations, along with formal proofs of the cryptography, are
available at:
* https://www.wireguard.com/
* https://www.wireguard.com/papers/wireguard.pdf
This commit implements WireGuard as a simple network device driver,
accessible in the usual RTNL way used by virtual network drivers. It
makes use of the udp_tunnel APIs, GRO, GSO, NAPI, and the usual set of
networking subsystem APIs. It has a somewhat novel multicore queueing
system designed for maximum throughput and minimal latency of encryption
operations, but it is implemented modestly using workqueues and NAPI.
Configuration is done via generic Netlink, and following a review from
the Netlink maintainer a year ago, several high profile userspace tools
have already implemented the API.
This commit also comes with several different tests, both in-kernel
tests and out-of-kernel tests based on network namespaces, taking profit
of the fact that sockets used by WireGuard intentionally stay in the
namespace the WireGuard interface was originally created, exactly like
the semantics of userspace tun devices. See wireguard.com/netns/ for
pictures and examples.
The source code is fairly short, but rather than combining everything
into a single file, WireGuard is developed as cleanly separable files,
making auditing and comprehension easier. Things are laid out as
follows:
* noise.[ch], cookie.[ch], messages.h: These implement the bulk of the
cryptographic aspects of the protocol, and are mostly data-only in
nature, taking in buffers of bytes and spitting out buffers of
bytes. They also handle reference counting for their various shared
pieces of data, like keys and key lists.
* ratelimiter.[ch]: Used as an integral part of cookie.[ch] for
ratelimiting certain types of cryptographic operations in accordance
with particular WireGuard semantics.
* allowedips.[ch], peerlookup.[ch]: The main lookup structures of
WireGuard, the former being trie-like with particular semantics, an
integral part of the design of the protocol, and the latter just
being nice helper functions around the various hashtables we use.
* device.[ch]: Implementation of functions for the netdevice and for
rtnl, responsible for maintaining the life of a given interface and
wiring it up to the rest of WireGuard.
* peer.[ch]: Each interface has a list of peers, with helper functions
available here for creation, destruction, and reference counting.
* socket.[ch]: Implementation of functions related to udp_socket and
the general set of kernel socket APIs, for sending and receiving
ciphertext UDP packets, and taking care of WireGuard-specific sticky
socket routing semantics for the automatic roaming.
* netlink.[ch]: Userspace API entry point for configuring WireGuard
peers and devices. The API has been implemented by several userspace
tools and network management utility, and the WireGuard project
distributes the basic wg(8) tool.
* queueing.[ch]: Shared function on the rx and tx path for handling
the various queues used in the multicore algorithms.
* send.c: Handles encrypting outgoing packets in parallel on
multiple cores, before sending them in order on a single core, via
workqueues and ring buffers. Also handles sending handshake and cookie
messages as part of the protocol, in parallel.
* receive.c: Handles decrypting incoming packets in parallel on
multiple cores, before passing them off in order to be ingested via
the rest of the networking subsystem with GRO via the typical NAPI
poll function. Also handles receiving handshake and cookie messages
as part of the protocol, in parallel.
* timers.[ch]: Uses the timer wheel to implement protocol particular
event timeouts, and gives a set of very simple event-driven entry
point functions for callers.
* main.c, version.h: Initialization and deinitialization of the module.
* selftest/*.h: Runtime unit tests for some of the most security
sensitive functions.
* tools/testing/selftests/wireguard/netns.sh: Aforementioned testing
script using network namespaces.
This commit aims to be as self-contained as possible, implementing
WireGuard as a standalone module not needing much special handling or
coordination from the network subsystem. I expect for future
optimizations to the network stack to positively improve WireGuard, and
vice-versa, but for the time being, this exists as intentionally
standalone.
We introduce a menu option for CONFIG_WIREGUARD, as well as providing a
verbose debug log and self-tests via CONFIG_WIREGUARD_DEBUG.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: David Miller <davem@davemloft.net>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: linux-crypto@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-09 07:27:34 +08:00
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 = $!
2020-02-05 05:17:29 +08:00
waitncatudp $netns1 $ncat_pid
net: WireGuard secure network tunnel
WireGuard is a layer 3 secure networking tunnel made specifically for
the kernel, that aims to be much simpler and easier to audit than IPsec.
Extensive documentation and description of the protocol and
considerations, along with formal proofs of the cryptography, are
available at:
* https://www.wireguard.com/
* https://www.wireguard.com/papers/wireguard.pdf
This commit implements WireGuard as a simple network device driver,
accessible in the usual RTNL way used by virtual network drivers. It
makes use of the udp_tunnel APIs, GRO, GSO, NAPI, and the usual set of
networking subsystem APIs. It has a somewhat novel multicore queueing
system designed for maximum throughput and minimal latency of encryption
operations, but it is implemented modestly using workqueues and NAPI.
Configuration is done via generic Netlink, and following a review from
the Netlink maintainer a year ago, several high profile userspace tools
have already implemented the API.
This commit also comes with several different tests, both in-kernel
tests and out-of-kernel tests based on network namespaces, taking profit
of the fact that sockets used by WireGuard intentionally stay in the
namespace the WireGuard interface was originally created, exactly like
the semantics of userspace tun devices. See wireguard.com/netns/ for
pictures and examples.
The source code is fairly short, but rather than combining everything
into a single file, WireGuard is developed as cleanly separable files,
making auditing and comprehension easier. Things are laid out as
follows:
* noise.[ch], cookie.[ch], messages.h: These implement the bulk of the
cryptographic aspects of the protocol, and are mostly data-only in
nature, taking in buffers of bytes and spitting out buffers of
bytes. They also handle reference counting for their various shared
pieces of data, like keys and key lists.
* ratelimiter.[ch]: Used as an integral part of cookie.[ch] for
ratelimiting certain types of cryptographic operations in accordance
with particular WireGuard semantics.
* allowedips.[ch], peerlookup.[ch]: The main lookup structures of
WireGuard, the former being trie-like with particular semantics, an
integral part of the design of the protocol, and the latter just
being nice helper functions around the various hashtables we use.
* device.[ch]: Implementation of functions for the netdevice and for
rtnl, responsible for maintaining the life of a given interface and
wiring it up to the rest of WireGuard.
* peer.[ch]: Each interface has a list of peers, with helper functions
available here for creation, destruction, and reference counting.
* socket.[ch]: Implementation of functions related to udp_socket and
the general set of kernel socket APIs, for sending and receiving
ciphertext UDP packets, and taking care of WireGuard-specific sticky
socket routing semantics for the automatic roaming.
* netlink.[ch]: Userspace API entry point for configuring WireGuard
peers and devices. The API has been implemented by several userspace
tools and network management utility, and the WireGuard project
distributes the basic wg(8) tool.
* queueing.[ch]: Shared function on the rx and tx path for handling
the various queues used in the multicore algorithms.
* send.c: Handles encrypting outgoing packets in parallel on
multiple cores, before sending them in order on a single core, via
workqueues and ring buffers. Also handles sending handshake and cookie
messages as part of the protocol, in parallel.
* receive.c: Handles decrypting incoming packets in parallel on
multiple cores, before passing them off in order to be ingested via
the rest of the networking subsystem with GRO via the typical NAPI
poll function. Also handles receiving handshake and cookie messages
as part of the protocol, in parallel.
* timers.[ch]: Uses the timer wheel to implement protocol particular
event timeouts, and gives a set of very simple event-driven entry
point functions for callers.
* main.c, version.h: Initialization and deinitialization of the module.
* selftest/*.h: Runtime unit tests for some of the most security
sensitive functions.
* tools/testing/selftests/wireguard/netns.sh: Aforementioned testing
script using network namespaces.
This commit aims to be as self-contained as possible, implementing
WireGuard as a standalone module not needing much special handling or
coordination from the network subsystem. I expect for future
optimizations to the network stack to positively improve WireGuard, and
vice-versa, but for the time being, this exists as intentionally
standalone.
We introduce a menu option for CONFIG_WIREGUARD, as well as providing a
verbose debug log and self-tests via CONFIG_WIREGUARD_DEBUG.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: David Miller <davem@davemloft.net>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: linux-crypto@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-09 07:27:34 +08:00
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 = $!
2020-02-05 05:17:29 +08:00
waitncatudp $netns1 $ncat_pid
net: WireGuard secure network tunnel
WireGuard is a layer 3 secure networking tunnel made specifically for
the kernel, that aims to be much simpler and easier to audit than IPsec.
Extensive documentation and description of the protocol and
considerations, along with formal proofs of the cryptography, are
available at:
* https://www.wireguard.com/
* https://www.wireguard.com/papers/wireguard.pdf
This commit implements WireGuard as a simple network device driver,
accessible in the usual RTNL way used by virtual network drivers. It
makes use of the udp_tunnel APIs, GRO, GSO, NAPI, and the usual set of
networking subsystem APIs. It has a somewhat novel multicore queueing
system designed for maximum throughput and minimal latency of encryption
operations, but it is implemented modestly using workqueues and NAPI.
Configuration is done via generic Netlink, and following a review from
the Netlink maintainer a year ago, several high profile userspace tools
have already implemented the API.
This commit also comes with several different tests, both in-kernel
tests and out-of-kernel tests based on network namespaces, taking profit
of the fact that sockets used by WireGuard intentionally stay in the
namespace the WireGuard interface was originally created, exactly like
the semantics of userspace tun devices. See wireguard.com/netns/ for
pictures and examples.
The source code is fairly short, but rather than combining everything
into a single file, WireGuard is developed as cleanly separable files,
making auditing and comprehension easier. Things are laid out as
follows:
* noise.[ch], cookie.[ch], messages.h: These implement the bulk of the
cryptographic aspects of the protocol, and are mostly data-only in
nature, taking in buffers of bytes and spitting out buffers of
bytes. They also handle reference counting for their various shared
pieces of data, like keys and key lists.
* ratelimiter.[ch]: Used as an integral part of cookie.[ch] for
ratelimiting certain types of cryptographic operations in accordance
with particular WireGuard semantics.
* allowedips.[ch], peerlookup.[ch]: The main lookup structures of
WireGuard, the former being trie-like with particular semantics, an
integral part of the design of the protocol, and the latter just
being nice helper functions around the various hashtables we use.
* device.[ch]: Implementation of functions for the netdevice and for
rtnl, responsible for maintaining the life of a given interface and
wiring it up to the rest of WireGuard.
* peer.[ch]: Each interface has a list of peers, with helper functions
available here for creation, destruction, and reference counting.
* socket.[ch]: Implementation of functions related to udp_socket and
the general set of kernel socket APIs, for sending and receiving
ciphertext UDP packets, and taking care of WireGuard-specific sticky
socket routing semantics for the automatic roaming.
* netlink.[ch]: Userspace API entry point for configuring WireGuard
peers and devices. The API has been implemented by several userspace
tools and network management utility, and the WireGuard project
distributes the basic wg(8) tool.
* queueing.[ch]: Shared function on the rx and tx path for handling
the various queues used in the multicore algorithms.
* send.c: Handles encrypting outgoing packets in parallel on
multiple cores, before sending them in order on a single core, via
workqueues and ring buffers. Also handles sending handshake and cookie
messages as part of the protocol, in parallel.
* receive.c: Handles decrypting incoming packets in parallel on
multiple cores, before passing them off in order to be ingested via
the rest of the networking subsystem with GRO via the typical NAPI
poll function. Also handles receiving handshake and cookie messages
as part of the protocol, in parallel.
* timers.[ch]: Uses the timer wheel to implement protocol particular
event timeouts, and gives a set of very simple event-driven entry
point functions for callers.
* main.c, version.h: Initialization and deinitialization of the module.
* selftest/*.h: Runtime unit tests for some of the most security
sensitive functions.
* tools/testing/selftests/wireguard/netns.sh: Aforementioned testing
script using network namespaces.
This commit aims to be as self-contained as possible, implementing
WireGuard as a standalone module not needing much special handling or
coordination from the network subsystem. I expect for future
optimizations to the network stack to positively improve WireGuard, and
vice-versa, but for the time being, this exists as intentionally
standalone.
We introduce a menu option for CONFIG_WIREGUARD, as well as providing a
verbose debug log and self-tests via CONFIG_WIREGUARD_DEBUG.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: David Miller <davem@davemloft.net>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: linux-crypto@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-09 07:27:34 +08:00
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
wireguard: socket: remove errant restriction on looping to self
It's already possible to create two different interfaces and loop
packets between them. This has always been possible with tunnels in the
kernel, and isn't specific to wireguard. Therefore, the networking stack
already needs to deal with that. At the very least, the packet winds up
exceeding the MTU and is discarded at that point. So, since this is
already something that happens, there's no need to forbid the not very
exceptional case of routing a packet back to the same interface; this
loop is no different than others, and we shouldn't special case it, but
rather rely on generic handling of loops in general. This also makes it
easier to do interesting things with wireguard such as onion routing.
At the same time, we add a selftest for this, ensuring that both onion
routing works and infinite routing loops do not crash the kernel. We
also add a test case for wireguard interfaces nesting packets and
sending traffic between each other, as well as the loop in this case
too. We make sure to send some throughput-heavy traffic for this use
case, to stress out any possible recursion issues with the locks around
workqueues.
Fixes: e7096c131e51 ("net: WireGuard secure network tunnel")
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2020-05-07 05:33:03 +08:00
n2 wg set wg0 peer " $pub3 " remove
# Test that we can route wg through wg
ip1 addr flush dev wg0
ip2 addr flush dev wg0
ip1 addr add fd00::5:1/112 dev wg0
ip2 addr add fd00::5:2/112 dev wg0
n1 wg set wg0 private-key <( echo " $key1 " ) peer " $pub2 " preshared-key <( echo " $psk " ) allowed-ips fd00::5:2/128 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 fd00::5:1/128 endpoint 127.212.121.99:9998
ip1 link add wg1 type wireguard
ip2 link add wg1 type wireguard
ip1 addr add 192.168.241.1/24 dev wg1
ip1 addr add fd00::1/112 dev wg1
ip2 addr add 192.168.241.2/24 dev wg1
ip2 addr add fd00::2/112 dev wg1
ip1 link set mtu 1340 up dev wg1
ip2 link set mtu 1340 up dev wg1
n1 wg set wg1 listen-port 5 private-key <( echo " $key3 " ) peer " $pub4 " allowed-ips 192.168.241.2/32,fd00::2/128 endpoint [ fd00::5:2] :5
n2 wg set wg1 listen-port 5 private-key <( echo " $key4 " ) peer " $pub3 " allowed-ips 192.168.241.1/32,fd00::1/128 endpoint [ fd00::5:1] :5
tests
# Try to set up a routing loop between the two namespaces
ip1 link set netns $netns0 dev wg1
ip0 addr add 192.168.241.1/24 dev wg1
ip0 link set up dev wg1
n0 ping -W 1 -c 1 192.168.241.2
n1 wg set wg0 peer " $pub2 " endpoint 192.168.241.2:7
ip2 link del wg0
ip2 link del wg1
! n0 ping -W 1 -c 10 -f 192.168.241.2 || false # Should not crash kernel
net: WireGuard secure network tunnel
WireGuard is a layer 3 secure networking tunnel made specifically for
the kernel, that aims to be much simpler and easier to audit than IPsec.
Extensive documentation and description of the protocol and
considerations, along with formal proofs of the cryptography, are
available at:
* https://www.wireguard.com/
* https://www.wireguard.com/papers/wireguard.pdf
This commit implements WireGuard as a simple network device driver,
accessible in the usual RTNL way used by virtual network drivers. It
makes use of the udp_tunnel APIs, GRO, GSO, NAPI, and the usual set of
networking subsystem APIs. It has a somewhat novel multicore queueing
system designed for maximum throughput and minimal latency of encryption
operations, but it is implemented modestly using workqueues and NAPI.
Configuration is done via generic Netlink, and following a review from
the Netlink maintainer a year ago, several high profile userspace tools
have already implemented the API.
This commit also comes with several different tests, both in-kernel
tests and out-of-kernel tests based on network namespaces, taking profit
of the fact that sockets used by WireGuard intentionally stay in the
namespace the WireGuard interface was originally created, exactly like
the semantics of userspace tun devices. See wireguard.com/netns/ for
pictures and examples.
The source code is fairly short, but rather than combining everything
into a single file, WireGuard is developed as cleanly separable files,
making auditing and comprehension easier. Things are laid out as
follows:
* noise.[ch], cookie.[ch], messages.h: These implement the bulk of the
cryptographic aspects of the protocol, and are mostly data-only in
nature, taking in buffers of bytes and spitting out buffers of
bytes. They also handle reference counting for their various shared
pieces of data, like keys and key lists.
* ratelimiter.[ch]: Used as an integral part of cookie.[ch] for
ratelimiting certain types of cryptographic operations in accordance
with particular WireGuard semantics.
* allowedips.[ch], peerlookup.[ch]: The main lookup structures of
WireGuard, the former being trie-like with particular semantics, an
integral part of the design of the protocol, and the latter just
being nice helper functions around the various hashtables we use.
* device.[ch]: Implementation of functions for the netdevice and for
rtnl, responsible for maintaining the life of a given interface and
wiring it up to the rest of WireGuard.
* peer.[ch]: Each interface has a list of peers, with helper functions
available here for creation, destruction, and reference counting.
* socket.[ch]: Implementation of functions related to udp_socket and
the general set of kernel socket APIs, for sending and receiving
ciphertext UDP packets, and taking care of WireGuard-specific sticky
socket routing semantics for the automatic roaming.
* netlink.[ch]: Userspace API entry point for configuring WireGuard
peers and devices. The API has been implemented by several userspace
tools and network management utility, and the WireGuard project
distributes the basic wg(8) tool.
* queueing.[ch]: Shared function on the rx and tx path for handling
the various queues used in the multicore algorithms.
* send.c: Handles encrypting outgoing packets in parallel on
multiple cores, before sending them in order on a single core, via
workqueues and ring buffers. Also handles sending handshake and cookie
messages as part of the protocol, in parallel.
* receive.c: Handles decrypting incoming packets in parallel on
multiple cores, before passing them off in order to be ingested via
the rest of the networking subsystem with GRO via the typical NAPI
poll function. Also handles receiving handshake and cookie messages
as part of the protocol, in parallel.
* timers.[ch]: Uses the timer wheel to implement protocol particular
event timeouts, and gives a set of very simple event-driven entry
point functions for callers.
* main.c, version.h: Initialization and deinitialization of the module.
* selftest/*.h: Runtime unit tests for some of the most security
sensitive functions.
* tools/testing/selftests/wireguard/netns.sh: Aforementioned testing
script using network namespaces.
This commit aims to be as self-contained as possible, implementing
WireGuard as a standalone module not needing much special handling or
coordination from the network subsystem. I expect for future
optimizations to the network stack to positively improve WireGuard, and
vice-versa, but for the time being, this exists as intentionally
standalone.
We introduce a menu option for CONFIG_WIREGUARD, as well as providing a
verbose debug log and self-tests via CONFIG_WIREGUARD_DEBUG.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: David Miller <davem@davemloft.net>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: linux-crypto@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-09 07:27:34 +08:00
wireguard: socket: remove errant restriction on looping to self
It's already possible to create two different interfaces and loop
packets between them. This has always been possible with tunnels in the
kernel, and isn't specific to wireguard. Therefore, the networking stack
already needs to deal with that. At the very least, the packet winds up
exceeding the MTU and is discarded at that point. So, since this is
already something that happens, there's no need to forbid the not very
exceptional case of routing a packet back to the same interface; this
loop is no different than others, and we shouldn't special case it, but
rather rely on generic handling of loops in general. This also makes it
easier to do interesting things with wireguard such as onion routing.
At the same time, we add a selftest for this, ensuring that both onion
routing works and infinite routing loops do not crash the kernel. We
also add a test case for wireguard interfaces nesting packets and
sending traffic between each other, as well as the loop in this case
too. We make sure to send some throughput-heavy traffic for this use
case, to stress out any possible recursion issues with the locks around
workqueues.
Fixes: e7096c131e51 ("net: WireGuard secure network tunnel")
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2020-05-07 05:33:03 +08:00
ip0 link del wg1
net: WireGuard secure network tunnel
WireGuard is a layer 3 secure networking tunnel made specifically for
the kernel, that aims to be much simpler and easier to audit than IPsec.
Extensive documentation and description of the protocol and
considerations, along with formal proofs of the cryptography, are
available at:
* https://www.wireguard.com/
* https://www.wireguard.com/papers/wireguard.pdf
This commit implements WireGuard as a simple network device driver,
accessible in the usual RTNL way used by virtual network drivers. It
makes use of the udp_tunnel APIs, GRO, GSO, NAPI, and the usual set of
networking subsystem APIs. It has a somewhat novel multicore queueing
system designed for maximum throughput and minimal latency of encryption
operations, but it is implemented modestly using workqueues and NAPI.
Configuration is done via generic Netlink, and following a review from
the Netlink maintainer a year ago, several high profile userspace tools
have already implemented the API.
This commit also comes with several different tests, both in-kernel
tests and out-of-kernel tests based on network namespaces, taking profit
of the fact that sockets used by WireGuard intentionally stay in the
namespace the WireGuard interface was originally created, exactly like
the semantics of userspace tun devices. See wireguard.com/netns/ for
pictures and examples.
The source code is fairly short, but rather than combining everything
into a single file, WireGuard is developed as cleanly separable files,
making auditing and comprehension easier. Things are laid out as
follows:
* noise.[ch], cookie.[ch], messages.h: These implement the bulk of the
cryptographic aspects of the protocol, and are mostly data-only in
nature, taking in buffers of bytes and spitting out buffers of
bytes. They also handle reference counting for their various shared
pieces of data, like keys and key lists.
* ratelimiter.[ch]: Used as an integral part of cookie.[ch] for
ratelimiting certain types of cryptographic operations in accordance
with particular WireGuard semantics.
* allowedips.[ch], peerlookup.[ch]: The main lookup structures of
WireGuard, the former being trie-like with particular semantics, an
integral part of the design of the protocol, and the latter just
being nice helper functions around the various hashtables we use.
* device.[ch]: Implementation of functions for the netdevice and for
rtnl, responsible for maintaining the life of a given interface and
wiring it up to the rest of WireGuard.
* peer.[ch]: Each interface has a list of peers, with helper functions
available here for creation, destruction, and reference counting.
* socket.[ch]: Implementation of functions related to udp_socket and
the general set of kernel socket APIs, for sending and receiving
ciphertext UDP packets, and taking care of WireGuard-specific sticky
socket routing semantics for the automatic roaming.
* netlink.[ch]: Userspace API entry point for configuring WireGuard
peers and devices. The API has been implemented by several userspace
tools and network management utility, and the WireGuard project
distributes the basic wg(8) tool.
* queueing.[ch]: Shared function on the rx and tx path for handling
the various queues used in the multicore algorithms.
* send.c: Handles encrypting outgoing packets in parallel on
multiple cores, before sending them in order on a single core, via
workqueues and ring buffers. Also handles sending handshake and cookie
messages as part of the protocol, in parallel.
* receive.c: Handles decrypting incoming packets in parallel on
multiple cores, before passing them off in order to be ingested via
the rest of the networking subsystem with GRO via the typical NAPI
poll function. Also handles receiving handshake and cookie messages
as part of the protocol, in parallel.
* timers.[ch]: Uses the timer wheel to implement protocol particular
event timeouts, and gives a set of very simple event-driven entry
point functions for callers.
* main.c, version.h: Initialization and deinitialization of the module.
* selftest/*.h: Runtime unit tests for some of the most security
sensitive functions.
* tools/testing/selftests/wireguard/netns.sh: Aforementioned testing
script using network namespaces.
This commit aims to be as self-contained as possible, implementing
WireGuard as a standalone module not needing much special handling or
coordination from the network subsystem. I expect for future
optimizations to the network stack to positively improve WireGuard, and
vice-versa, but for the time being, this exists as intentionally
standalone.
We introduce a menu option for CONFIG_WIREGUARD, as well as providing a
verbose debug log and self-tests via CONFIG_WIREGUARD_DEBUG.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: David Miller <davem@davemloft.net>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: linux-crypto@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-09 07:27:34 +08:00
ip1 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
wireguard: socket: remove errant restriction on looping to self
It's already possible to create two different interfaces and loop
packets between them. This has always been possible with tunnels in the
kernel, and isn't specific to wireguard. Therefore, the networking stack
already needs to deal with that. At the very least, the packet winds up
exceeding the MTU and is discarded at that point. So, since this is
already something that happens, there's no need to forbid the not very
exceptional case of routing a packet back to the same interface; this
loop is no different than others, and we shouldn't special case it, but
rather rely on generic handling of loops in general. This also makes it
easier to do interesting things with wireguard such as onion routing.
At the same time, we add a selftest for this, ensuring that both onion
routing works and infinite routing loops do not crash the kernel. We
also add a test case for wireguard interfaces nesting packets and
sending traffic between each other, as well as the loop in this case
too. We make sure to send some throughput-heavy traffic for this use
case, to stress out any possible recursion issues with the locks around
workqueues.
Fixes: e7096c131e51 ("net: WireGuard secure network tunnel")
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2020-05-07 05:33:03 +08:00
2020-10-29 10:56:05 +08:00
# Test that sk_bound_dev_if works
n1 ping -I wg0 -c 1 -W 1 192.168.241.2
# What about when the mark changes and the packet must be rerouted?
n1 iptables -t mangle -I OUTPUT -j MARK --set-xmark 1
n1 ping -c 1 -W 1 192.168.241.2 # First the boring case
n1 ping -I wg0 -c 1 -W 1 192.168.241.2 # Then the sk_bound_dev_if case
n1 iptables -t mangle -D OUTPUT -j MARK --set-xmark 1
wireguard: socket: remove errant restriction on looping to self
It's already possible to create two different interfaces and loop
packets between them. This has always been possible with tunnels in the
kernel, and isn't specific to wireguard. Therefore, the networking stack
already needs to deal with that. At the very least, the packet winds up
exceeding the MTU and is discarded at that point. So, since this is
already something that happens, there's no need to forbid the not very
exceptional case of routing a packet back to the same interface; this
loop is no different than others, and we shouldn't special case it, but
rather rely on generic handling of loops in general. This also makes it
easier to do interesting things with wireguard such as onion routing.
At the same time, we add a selftest for this, ensuring that both onion
routing works and infinite routing loops do not crash the kernel. We
also add a test case for wireguard interfaces nesting packets and
sending traffic between each other, as well as the loop in this case
too. We make sure to send some throughput-heavy traffic for this use
case, to stress out any possible recursion issues with the locks around
workqueues.
Fixes: e7096c131e51 ("net: WireGuard secure network tunnel")
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2020-05-07 05:33:03 +08:00
# Test that onion routing works, even when it loops
n1 wg set wg0 peer " $pub3 " allowed-ips 192.168.242.2/32 endpoint 192.168.241.2:5
ip1 addr add 192.168.242.1/24 dev wg0
ip2 link add wg1 type wireguard
ip2 addr add 192.168.242.2/24 dev wg1
n2 wg set wg1 private-key <( echo " $key3 " ) listen-port 5 peer " $pub1 " allowed-ips 192.168.242.1/32
ip2 link set wg1 up
n1 ping -W 1 -c 1 192.168.242.2
ip2 link del wg1
n1 wg set wg0 peer " $pub3 " endpoint 192.168.242.2:5
! n1 ping -W 1 -c 1 192.168.242.2 || false # Should not crash kernel
n1 wg set wg0 peer " $pub3 " remove
ip1 addr del 192.168.242.1/24 dev wg0
net: WireGuard secure network tunnel
WireGuard is a layer 3 secure networking tunnel made specifically for
the kernel, that aims to be much simpler and easier to audit than IPsec.
Extensive documentation and description of the protocol and
considerations, along with formal proofs of the cryptography, are
available at:
* https://www.wireguard.com/
* https://www.wireguard.com/papers/wireguard.pdf
This commit implements WireGuard as a simple network device driver,
accessible in the usual RTNL way used by virtual network drivers. It
makes use of the udp_tunnel APIs, GRO, GSO, NAPI, and the usual set of
networking subsystem APIs. It has a somewhat novel multicore queueing
system designed for maximum throughput and minimal latency of encryption
operations, but it is implemented modestly using workqueues and NAPI.
Configuration is done via generic Netlink, and following a review from
the Netlink maintainer a year ago, several high profile userspace tools
have already implemented the API.
This commit also comes with several different tests, both in-kernel
tests and out-of-kernel tests based on network namespaces, taking profit
of the fact that sockets used by WireGuard intentionally stay in the
namespace the WireGuard interface was originally created, exactly like
the semantics of userspace tun devices. See wireguard.com/netns/ for
pictures and examples.
The source code is fairly short, but rather than combining everything
into a single file, WireGuard is developed as cleanly separable files,
making auditing and comprehension easier. Things are laid out as
follows:
* noise.[ch], cookie.[ch], messages.h: These implement the bulk of the
cryptographic aspects of the protocol, and are mostly data-only in
nature, taking in buffers of bytes and spitting out buffers of
bytes. They also handle reference counting for their various shared
pieces of data, like keys and key lists.
* ratelimiter.[ch]: Used as an integral part of cookie.[ch] for
ratelimiting certain types of cryptographic operations in accordance
with particular WireGuard semantics.
* allowedips.[ch], peerlookup.[ch]: The main lookup structures of
WireGuard, the former being trie-like with particular semantics, an
integral part of the design of the protocol, and the latter just
being nice helper functions around the various hashtables we use.
* device.[ch]: Implementation of functions for the netdevice and for
rtnl, responsible for maintaining the life of a given interface and
wiring it up to the rest of WireGuard.
* peer.[ch]: Each interface has a list of peers, with helper functions
available here for creation, destruction, and reference counting.
* socket.[ch]: Implementation of functions related to udp_socket and
the general set of kernel socket APIs, for sending and receiving
ciphertext UDP packets, and taking care of WireGuard-specific sticky
socket routing semantics for the automatic roaming.
* netlink.[ch]: Userspace API entry point for configuring WireGuard
peers and devices. The API has been implemented by several userspace
tools and network management utility, and the WireGuard project
distributes the basic wg(8) tool.
* queueing.[ch]: Shared function on the rx and tx path for handling
the various queues used in the multicore algorithms.
* send.c: Handles encrypting outgoing packets in parallel on
multiple cores, before sending them in order on a single core, via
workqueues and ring buffers. Also handles sending handshake and cookie
messages as part of the protocol, in parallel.
* receive.c: Handles decrypting incoming packets in parallel on
multiple cores, before passing them off in order to be ingested via
the rest of the networking subsystem with GRO via the typical NAPI
poll function. Also handles receiving handshake and cookie messages
as part of the protocol, in parallel.
* timers.[ch]: Uses the timer wheel to implement protocol particular
event timeouts, and gives a set of very simple event-driven entry
point functions for callers.
* main.c, version.h: Initialization and deinitialization of the module.
* selftest/*.h: Runtime unit tests for some of the most security
sensitive functions.
* tools/testing/selftests/wireguard/netns.sh: Aforementioned testing
script using network namespaces.
This commit aims to be as self-contained as possible, implementing
WireGuard as a standalone module not needing much special handling or
coordination from the network subsystem. I expect for future
optimizations to the network stack to positively improve WireGuard, and
vice-versa, but for the time being, this exists as intentionally
standalone.
We introduce a menu option for CONFIG_WIREGUARD, as well as providing a
verbose debug log and self-tests via CONFIG_WIREGUARD_DEBUG.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: David Miller <davem@davemloft.net>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: linux-crypto@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-09 07:27:34 +08:00
# 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
2020-01-03 00:47:49 +08:00
# 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
net: WireGuard secure network tunnel
WireGuard is a layer 3 secure networking tunnel made specifically for
the kernel, that aims to be much simpler and easier to audit than IPsec.
Extensive documentation and description of the protocol and
considerations, along with formal proofs of the cryptography, are
available at:
* https://www.wireguard.com/
* https://www.wireguard.com/papers/wireguard.pdf
This commit implements WireGuard as a simple network device driver,
accessible in the usual RTNL way used by virtual network drivers. It
makes use of the udp_tunnel APIs, GRO, GSO, NAPI, and the usual set of
networking subsystem APIs. It has a somewhat novel multicore queueing
system designed for maximum throughput and minimal latency of encryption
operations, but it is implemented modestly using workqueues and NAPI.
Configuration is done via generic Netlink, and following a review from
the Netlink maintainer a year ago, several high profile userspace tools
have already implemented the API.
This commit also comes with several different tests, both in-kernel
tests and out-of-kernel tests based on network namespaces, taking profit
of the fact that sockets used by WireGuard intentionally stay in the
namespace the WireGuard interface was originally created, exactly like
the semantics of userspace tun devices. See wireguard.com/netns/ for
pictures and examples.
The source code is fairly short, but rather than combining everything
into a single file, WireGuard is developed as cleanly separable files,
making auditing and comprehension easier. Things are laid out as
follows:
* noise.[ch], cookie.[ch], messages.h: These implement the bulk of the
cryptographic aspects of the protocol, and are mostly data-only in
nature, taking in buffers of bytes and spitting out buffers of
bytes. They also handle reference counting for their various shared
pieces of data, like keys and key lists.
* ratelimiter.[ch]: Used as an integral part of cookie.[ch] for
ratelimiting certain types of cryptographic operations in accordance
with particular WireGuard semantics.
* allowedips.[ch], peerlookup.[ch]: The main lookup structures of
WireGuard, the former being trie-like with particular semantics, an
integral part of the design of the protocol, and the latter just
being nice helper functions around the various hashtables we use.
* device.[ch]: Implementation of functions for the netdevice and for
rtnl, responsible for maintaining the life of a given interface and
wiring it up to the rest of WireGuard.
* peer.[ch]: Each interface has a list of peers, with helper functions
available here for creation, destruction, and reference counting.
* socket.[ch]: Implementation of functions related to udp_socket and
the general set of kernel socket APIs, for sending and receiving
ciphertext UDP packets, and taking care of WireGuard-specific sticky
socket routing semantics for the automatic roaming.
* netlink.[ch]: Userspace API entry point for configuring WireGuard
peers and devices. The API has been implemented by several userspace
tools and network management utility, and the WireGuard project
distributes the basic wg(8) tool.
* queueing.[ch]: Shared function on the rx and tx path for handling
the various queues used in the multicore algorithms.
* send.c: Handles encrypting outgoing packets in parallel on
multiple cores, before sending them in order on a single core, via
workqueues and ring buffers. Also handles sending handshake and cookie
messages as part of the protocol, in parallel.
* receive.c: Handles decrypting incoming packets in parallel on
multiple cores, before passing them off in order to be ingested via
the rest of the networking subsystem with GRO via the typical NAPI
poll function. Also handles receiving handshake and cookie messages
as part of the protocol, in parallel.
* timers.[ch]: Uses the timer wheel to implement protocol particular
event timeouts, and gives a set of very simple event-driven entry
point functions for callers.
* main.c, version.h: Initialization and deinitialization of the module.
* selftest/*.h: Runtime unit tests for some of the most security
sensitive functions.
* tools/testing/selftests/wireguard/netns.sh: Aforementioned testing
script using network namespaces.
This commit aims to be as self-contained as possible, implementing
WireGuard as a standalone module not needing much special handling or
coordination from the network subsystem. I expect for future
optimizations to the network stack to positively improve WireGuard, and
vice-versa, but for the time being, this exists as intentionally
standalone.
We introduce a menu option for CONFIG_WIREGUARD, as well as providing a
verbose debug log and self-tests via CONFIG_WIREGUARD_DEBUG.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: David Miller <davem@davemloft.net>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: linux-crypto@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-09 07:27:34 +08:00
2020-02-12 03:47:08 +08:00
# 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" * ] ]
net: WireGuard secure network tunnel
WireGuard is a layer 3 secure networking tunnel made specifically for
the kernel, that aims to be much simpler and easier to audit than IPsec.
Extensive documentation and description of the protocol and
considerations, along with formal proofs of the cryptography, are
available at:
* https://www.wireguard.com/
* https://www.wireguard.com/papers/wireguard.pdf
This commit implements WireGuard as a simple network device driver,
accessible in the usual RTNL way used by virtual network drivers. It
makes use of the udp_tunnel APIs, GRO, GSO, NAPI, and the usual set of
networking subsystem APIs. It has a somewhat novel multicore queueing
system designed for maximum throughput and minimal latency of encryption
operations, but it is implemented modestly using workqueues and NAPI.
Configuration is done via generic Netlink, and following a review from
the Netlink maintainer a year ago, several high profile userspace tools
have already implemented the API.
This commit also comes with several different tests, both in-kernel
tests and out-of-kernel tests based on network namespaces, taking profit
of the fact that sockets used by WireGuard intentionally stay in the
namespace the WireGuard interface was originally created, exactly like
the semantics of userspace tun devices. See wireguard.com/netns/ for
pictures and examples.
The source code is fairly short, but rather than combining everything
into a single file, WireGuard is developed as cleanly separable files,
making auditing and comprehension easier. Things are laid out as
follows:
* noise.[ch], cookie.[ch], messages.h: These implement the bulk of the
cryptographic aspects of the protocol, and are mostly data-only in
nature, taking in buffers of bytes and spitting out buffers of
bytes. They also handle reference counting for their various shared
pieces of data, like keys and key lists.
* ratelimiter.[ch]: Used as an integral part of cookie.[ch] for
ratelimiting certain types of cryptographic operations in accordance
with particular WireGuard semantics.
* allowedips.[ch], peerlookup.[ch]: The main lookup structures of
WireGuard, the former being trie-like with particular semantics, an
integral part of the design of the protocol, and the latter just
being nice helper functions around the various hashtables we use.
* device.[ch]: Implementation of functions for the netdevice and for
rtnl, responsible for maintaining the life of a given interface and
wiring it up to the rest of WireGuard.
* peer.[ch]: Each interface has a list of peers, with helper functions
available here for creation, destruction, and reference counting.
* socket.[ch]: Implementation of functions related to udp_socket and
the general set of kernel socket APIs, for sending and receiving
ciphertext UDP packets, and taking care of WireGuard-specific sticky
socket routing semantics for the automatic roaming.
* netlink.[ch]: Userspace API entry point for configuring WireGuard
peers and devices. The API has been implemented by several userspace
tools and network management utility, and the WireGuard project
distributes the basic wg(8) tool.
* queueing.[ch]: Shared function on the rx and tx path for handling
the various queues used in the multicore algorithms.
* send.c: Handles encrypting outgoing packets in parallel on
multiple cores, before sending them in order on a single core, via
workqueues and ring buffers. Also handles sending handshake and cookie
messages as part of the protocol, in parallel.
* receive.c: Handles decrypting incoming packets in parallel on
multiple cores, before passing them off in order to be ingested via
the rest of the networking subsystem with GRO via the typical NAPI
poll function. Also handles receiving handshake and cookie messages
as part of the protocol, in parallel.
* timers.[ch]: Uses the timer wheel to implement protocol particular
event timeouts, and gives a set of very simple event-driven entry
point functions for callers.
* main.c, version.h: Initialization and deinitialization of the module.
* selftest/*.h: Runtime unit tests for some of the most security
sensitive functions.
* tools/testing/selftests/wireguard/netns.sh: Aforementioned testing
script using network namespaces.
This commit aims to be as self-contained as possible, implementing
WireGuard as a standalone module not needing much special handling or
coordination from the network subsystem. I expect for future
optimizations to the network stack to positively improve WireGuard, and
vice-versa, but for the time being, this exists as intentionally
standalone.
We introduce a menu option for CONFIG_WIREGUARD, as well as providing a
verbose debug log and self-tests via CONFIG_WIREGUARD_DEBUG.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: David Miller <davem@davemloft.net>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: linux-crypto@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-09 07:27:34 +08:00
n0 iptables -t nat -F
2020-02-12 03:47:08 +08:00
n0 iptables -t filter -F
n2 iptables -t nat -F
net: WireGuard secure network tunnel
WireGuard is a layer 3 secure networking tunnel made specifically for
the kernel, that aims to be much simpler and easier to audit than IPsec.
Extensive documentation and description of the protocol and
considerations, along with formal proofs of the cryptography, are
available at:
* https://www.wireguard.com/
* https://www.wireguard.com/papers/wireguard.pdf
This commit implements WireGuard as a simple network device driver,
accessible in the usual RTNL way used by virtual network drivers. It
makes use of the udp_tunnel APIs, GRO, GSO, NAPI, and the usual set of
networking subsystem APIs. It has a somewhat novel multicore queueing
system designed for maximum throughput and minimal latency of encryption
operations, but it is implemented modestly using workqueues and NAPI.
Configuration is done via generic Netlink, and following a review from
the Netlink maintainer a year ago, several high profile userspace tools
have already implemented the API.
This commit also comes with several different tests, both in-kernel
tests and out-of-kernel tests based on network namespaces, taking profit
of the fact that sockets used by WireGuard intentionally stay in the
namespace the WireGuard interface was originally created, exactly like
the semantics of userspace tun devices. See wireguard.com/netns/ for
pictures and examples.
The source code is fairly short, but rather than combining everything
into a single file, WireGuard is developed as cleanly separable files,
making auditing and comprehension easier. Things are laid out as
follows:
* noise.[ch], cookie.[ch], messages.h: These implement the bulk of the
cryptographic aspects of the protocol, and are mostly data-only in
nature, taking in buffers of bytes and spitting out buffers of
bytes. They also handle reference counting for their various shared
pieces of data, like keys and key lists.
* ratelimiter.[ch]: Used as an integral part of cookie.[ch] for
ratelimiting certain types of cryptographic operations in accordance
with particular WireGuard semantics.
* allowedips.[ch], peerlookup.[ch]: The main lookup structures of
WireGuard, the former being trie-like with particular semantics, an
integral part of the design of the protocol, and the latter just
being nice helper functions around the various hashtables we use.
* device.[ch]: Implementation of functions for the netdevice and for
rtnl, responsible for maintaining the life of a given interface and
wiring it up to the rest of WireGuard.
* peer.[ch]: Each interface has a list of peers, with helper functions
available here for creation, destruction, and reference counting.
* socket.[ch]: Implementation of functions related to udp_socket and
the general set of kernel socket APIs, for sending and receiving
ciphertext UDP packets, and taking care of WireGuard-specific sticky
socket routing semantics for the automatic roaming.
* netlink.[ch]: Userspace API entry point for configuring WireGuard
peers and devices. The API has been implemented by several userspace
tools and network management utility, and the WireGuard project
distributes the basic wg(8) tool.
* queueing.[ch]: Shared function on the rx and tx path for handling
the various queues used in the multicore algorithms.
* send.c: Handles encrypting outgoing packets in parallel on
multiple cores, before sending them in order on a single core, via
workqueues and ring buffers. Also handles sending handshake and cookie
messages as part of the protocol, in parallel.
* receive.c: Handles decrypting incoming packets in parallel on
multiple cores, before passing them off in order to be ingested via
the rest of the networking subsystem with GRO via the typical NAPI
poll function. Also handles receiving handshake and cookie messages
as part of the protocol, in parallel.
* timers.[ch]: Uses the timer wheel to implement protocol particular
event timeouts, and gives a set of very simple event-driven entry
point functions for callers.
* main.c, version.h: Initialization and deinitialization of the module.
* selftest/*.h: Runtime unit tests for some of the most security
sensitive functions.
* tools/testing/selftests/wireguard/netns.sh: Aforementioned testing
script using network namespaces.
This commit aims to be as self-contained as possible, implementing
WireGuard as a standalone module not needing much special handling or
coordination from the network subsystem. I expect for future
optimizations to the network stack to positively improve WireGuard, and
vice-versa, but for the time being, this exists as intentionally
standalone.
We introduce a menu option for CONFIG_WIREGUARD, as well as providing a
verbose debug log and self-tests via CONFIG_WIREGUARD_DEBUG.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: David Miller <davem@davemloft.net>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: linux-crypto@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-09 07:27:34 +08:00
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
2020-02-05 05:17:27 +08:00
n0 wg set wg0 peer " $pub2 " remove
wireguard: noise: error out precomputed DH during handshake rather than config
We precompute the static-static ECDH during configuration time, in order
to save an expensive computation later when receiving network packets.
However, not all ECDH computations yield a contributory result. Prior,
we were just not letting those peers be added to the interface. However,
this creates a strange inconsistency, since it was still possible to add
other weird points, like a valid public key plus a low-order point, and,
like points that result in zeros, a handshake would not complete. In
order to make the behavior more uniform and less surprising, simply
allow all peers to be added. Then, we'll error out later when doing the
crypto if there's an issue. This also adds more separation between the
crypto layer and the configuration layer.
Discussed-with: Mathias Hall-Andersen <mathias@hall-andersen.dk>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2020-03-19 08:30:47 +08:00
for low_order_point in AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA = AQAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA = 4Ot6fDtBuK4WVuP68Z/EatoJjeucMrH9hmIFFl9JuAA= X5yVvKNQjCSx0LFVnIPvWwREXMRYHI6G2CJO3dCfEVc = 7P///////////////////////////////////////38= 7f///////////////////////////////////////38= 7v///////////////////////////////////////38= ; do
n0 wg set wg0 peer " $low_order_point " persistent-keepalive 1 endpoint 127.0.0.1:1111
done
[ [ -n $( n0 wg show wg0 peers) ] ]
exec 4< <( n0 ncat -l -u -p 1111)
ncat_pid = $!
waitncatudp $netns0 $ncat_pid
ip0 link set wg0 up
! read -r -n 1 -t 2 <& 4 || false
kill $ncat_pid
net: WireGuard secure network tunnel
WireGuard is a layer 3 secure networking tunnel made specifically for
the kernel, that aims to be much simpler and easier to audit than IPsec.
Extensive documentation and description of the protocol and
considerations, along with formal proofs of the cryptography, are
available at:
* https://www.wireguard.com/
* https://www.wireguard.com/papers/wireguard.pdf
This commit implements WireGuard as a simple network device driver,
accessible in the usual RTNL way used by virtual network drivers. It
makes use of the udp_tunnel APIs, GRO, GSO, NAPI, and the usual set of
networking subsystem APIs. It has a somewhat novel multicore queueing
system designed for maximum throughput and minimal latency of encryption
operations, but it is implemented modestly using workqueues and NAPI.
Configuration is done via generic Netlink, and following a review from
the Netlink maintainer a year ago, several high profile userspace tools
have already implemented the API.
This commit also comes with several different tests, both in-kernel
tests and out-of-kernel tests based on network namespaces, taking profit
of the fact that sockets used by WireGuard intentionally stay in the
namespace the WireGuard interface was originally created, exactly like
the semantics of userspace tun devices. See wireguard.com/netns/ for
pictures and examples.
The source code is fairly short, but rather than combining everything
into a single file, WireGuard is developed as cleanly separable files,
making auditing and comprehension easier. Things are laid out as
follows:
* noise.[ch], cookie.[ch], messages.h: These implement the bulk of the
cryptographic aspects of the protocol, and are mostly data-only in
nature, taking in buffers of bytes and spitting out buffers of
bytes. They also handle reference counting for their various shared
pieces of data, like keys and key lists.
* ratelimiter.[ch]: Used as an integral part of cookie.[ch] for
ratelimiting certain types of cryptographic operations in accordance
with particular WireGuard semantics.
* allowedips.[ch], peerlookup.[ch]: The main lookup structures of
WireGuard, the former being trie-like with particular semantics, an
integral part of the design of the protocol, and the latter just
being nice helper functions around the various hashtables we use.
* device.[ch]: Implementation of functions for the netdevice and for
rtnl, responsible for maintaining the life of a given interface and
wiring it up to the rest of WireGuard.
* peer.[ch]: Each interface has a list of peers, with helper functions
available here for creation, destruction, and reference counting.
* socket.[ch]: Implementation of functions related to udp_socket and
the general set of kernel socket APIs, for sending and receiving
ciphertext UDP packets, and taking care of WireGuard-specific sticky
socket routing semantics for the automatic roaming.
* netlink.[ch]: Userspace API entry point for configuring WireGuard
peers and devices. The API has been implemented by several userspace
tools and network management utility, and the WireGuard project
distributes the basic wg(8) tool.
* queueing.[ch]: Shared function on the rx and tx path for handling
the various queues used in the multicore algorithms.
* send.c: Handles encrypting outgoing packets in parallel on
multiple cores, before sending them in order on a single core, via
workqueues and ring buffers. Also handles sending handshake and cookie
messages as part of the protocol, in parallel.
* receive.c: Handles decrypting incoming packets in parallel on
multiple cores, before passing them off in order to be ingested via
the rest of the networking subsystem with GRO via the typical NAPI
poll function. Also handles receiving handshake and cookie messages
as part of the protocol, in parallel.
* timers.[ch]: Uses the timer wheel to implement protocol particular
event timeouts, and gives a set of very simple event-driven entry
point functions for callers.
* main.c, version.h: Initialization and deinitialization of the module.
* selftest/*.h: Runtime unit tests for some of the most security
sensitive functions.
* tools/testing/selftests/wireguard/netns.sh: Aforementioned testing
script using network namespaces.
This commit aims to be as self-contained as possible, implementing
WireGuard as a standalone module not needing much special handling or
coordination from the network subsystem. I expect for future
optimizations to the network stack to positively improve WireGuard, and
vice-versa, but for the time being, this exists as intentionally
standalone.
We introduce a menu option for CONFIG_WIREGUARD, as well as providing a
verbose debug log and self-tests via CONFIG_WIREGUARD_DEBUG.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: David Miller <davem@davemloft.net>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: linux-crypto@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-09 07:27:34 +08:00
ip0 link del wg0
2020-06-23 17:59:45 +08:00
# Ensure there aren't circular reference loops
ip1 link add wg1 type wireguard
ip2 link add wg2 type wireguard
ip1 link set wg1 netns $netns2
ip2 link set wg2 netns $netns1
pp ip netns delete $netns1
pp ip netns delete $netns2
pp ip netns add $netns1
pp ip netns add $netns2
sleep 2 # Wait for cleanup and grace periods
net: WireGuard secure network tunnel
WireGuard is a layer 3 secure networking tunnel made specifically for
the kernel, that aims to be much simpler and easier to audit than IPsec.
Extensive documentation and description of the protocol and
considerations, along with formal proofs of the cryptography, are
available at:
* https://www.wireguard.com/
* https://www.wireguard.com/papers/wireguard.pdf
This commit implements WireGuard as a simple network device driver,
accessible in the usual RTNL way used by virtual network drivers. It
makes use of the udp_tunnel APIs, GRO, GSO, NAPI, and the usual set of
networking subsystem APIs. It has a somewhat novel multicore queueing
system designed for maximum throughput and minimal latency of encryption
operations, but it is implemented modestly using workqueues and NAPI.
Configuration is done via generic Netlink, and following a review from
the Netlink maintainer a year ago, several high profile userspace tools
have already implemented the API.
This commit also comes with several different tests, both in-kernel
tests and out-of-kernel tests based on network namespaces, taking profit
of the fact that sockets used by WireGuard intentionally stay in the
namespace the WireGuard interface was originally created, exactly like
the semantics of userspace tun devices. See wireguard.com/netns/ for
pictures and examples.
The source code is fairly short, but rather than combining everything
into a single file, WireGuard is developed as cleanly separable files,
making auditing and comprehension easier. Things are laid out as
follows:
* noise.[ch], cookie.[ch], messages.h: These implement the bulk of the
cryptographic aspects of the protocol, and are mostly data-only in
nature, taking in buffers of bytes and spitting out buffers of
bytes. They also handle reference counting for their various shared
pieces of data, like keys and key lists.
* ratelimiter.[ch]: Used as an integral part of cookie.[ch] for
ratelimiting certain types of cryptographic operations in accordance
with particular WireGuard semantics.
* allowedips.[ch], peerlookup.[ch]: The main lookup structures of
WireGuard, the former being trie-like with particular semantics, an
integral part of the design of the protocol, and the latter just
being nice helper functions around the various hashtables we use.
* device.[ch]: Implementation of functions for the netdevice and for
rtnl, responsible for maintaining the life of a given interface and
wiring it up to the rest of WireGuard.
* peer.[ch]: Each interface has a list of peers, with helper functions
available here for creation, destruction, and reference counting.
* socket.[ch]: Implementation of functions related to udp_socket and
the general set of kernel socket APIs, for sending and receiving
ciphertext UDP packets, and taking care of WireGuard-specific sticky
socket routing semantics for the automatic roaming.
* netlink.[ch]: Userspace API entry point for configuring WireGuard
peers and devices. The API has been implemented by several userspace
tools and network management utility, and the WireGuard project
distributes the basic wg(8) tool.
* queueing.[ch]: Shared function on the rx and tx path for handling
the various queues used in the multicore algorithms.
* send.c: Handles encrypting outgoing packets in parallel on
multiple cores, before sending them in order on a single core, via
workqueues and ring buffers. Also handles sending handshake and cookie
messages as part of the protocol, in parallel.
* receive.c: Handles decrypting incoming packets in parallel on
multiple cores, before passing them off in order to be ingested via
the rest of the networking subsystem with GRO via the typical NAPI
poll function. Also handles receiving handshake and cookie messages
as part of the protocol, in parallel.
* timers.[ch]: Uses the timer wheel to implement protocol particular
event timeouts, and gives a set of very simple event-driven entry
point functions for callers.
* main.c, version.h: Initialization and deinitialization of the module.
* selftest/*.h: Runtime unit tests for some of the most security
sensitive functions.
* tools/testing/selftests/wireguard/netns.sh: Aforementioned testing
script using network namespaces.
This commit aims to be as self-contained as possible, implementing
WireGuard as a standalone module not needing much special handling or
coordination from the network subsystem. I expect for future
optimizations to the network stack to positively improve WireGuard, and
vice-versa, but for the time being, this exists as intentionally
standalone.
We introduce a menu option for CONFIG_WIREGUARD, as well as providing a
verbose debug log and self-tests via CONFIG_WIREGUARD_DEBUG.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: David Miller <davem@davemloft.net>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: linux-crypto@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-09 07:27:34 +08:00
declare -A objects
while read -t 0.1 -r line 2>/dev/null || [ [ $? -ne 142 ] ] ; do
2020-06-23 17:59:45 +08:00
[ [ $line = ~ .*( wg[ 0-9] +:\ [ A-Z] [ a-z] +\ ?[ 0-9] *) \ .*( created| destroyed) .* ] ] || continue
net: WireGuard secure network tunnel
WireGuard is a layer 3 secure networking tunnel made specifically for
the kernel, that aims to be much simpler and easier to audit than IPsec.
Extensive documentation and description of the protocol and
considerations, along with formal proofs of the cryptography, are
available at:
* https://www.wireguard.com/
* https://www.wireguard.com/papers/wireguard.pdf
This commit implements WireGuard as a simple network device driver,
accessible in the usual RTNL way used by virtual network drivers. It
makes use of the udp_tunnel APIs, GRO, GSO, NAPI, and the usual set of
networking subsystem APIs. It has a somewhat novel multicore queueing
system designed for maximum throughput and minimal latency of encryption
operations, but it is implemented modestly using workqueues and NAPI.
Configuration is done via generic Netlink, and following a review from
the Netlink maintainer a year ago, several high profile userspace tools
have already implemented the API.
This commit also comes with several different tests, both in-kernel
tests and out-of-kernel tests based on network namespaces, taking profit
of the fact that sockets used by WireGuard intentionally stay in the
namespace the WireGuard interface was originally created, exactly like
the semantics of userspace tun devices. See wireguard.com/netns/ for
pictures and examples.
The source code is fairly short, but rather than combining everything
into a single file, WireGuard is developed as cleanly separable files,
making auditing and comprehension easier. Things are laid out as
follows:
* noise.[ch], cookie.[ch], messages.h: These implement the bulk of the
cryptographic aspects of the protocol, and are mostly data-only in
nature, taking in buffers of bytes and spitting out buffers of
bytes. They also handle reference counting for their various shared
pieces of data, like keys and key lists.
* ratelimiter.[ch]: Used as an integral part of cookie.[ch] for
ratelimiting certain types of cryptographic operations in accordance
with particular WireGuard semantics.
* allowedips.[ch], peerlookup.[ch]: The main lookup structures of
WireGuard, the former being trie-like with particular semantics, an
integral part of the design of the protocol, and the latter just
being nice helper functions around the various hashtables we use.
* device.[ch]: Implementation of functions for the netdevice and for
rtnl, responsible for maintaining the life of a given interface and
wiring it up to the rest of WireGuard.
* peer.[ch]: Each interface has a list of peers, with helper functions
available here for creation, destruction, and reference counting.
* socket.[ch]: Implementation of functions related to udp_socket and
the general set of kernel socket APIs, for sending and receiving
ciphertext UDP packets, and taking care of WireGuard-specific sticky
socket routing semantics for the automatic roaming.
* netlink.[ch]: Userspace API entry point for configuring WireGuard
peers and devices. The API has been implemented by several userspace
tools and network management utility, and the WireGuard project
distributes the basic wg(8) tool.
* queueing.[ch]: Shared function on the rx and tx path for handling
the various queues used in the multicore algorithms.
* send.c: Handles encrypting outgoing packets in parallel on
multiple cores, before sending them in order on a single core, via
workqueues and ring buffers. Also handles sending handshake and cookie
messages as part of the protocol, in parallel.
* receive.c: Handles decrypting incoming packets in parallel on
multiple cores, before passing them off in order to be ingested via
the rest of the networking subsystem with GRO via the typical NAPI
poll function. Also handles receiving handshake and cookie messages
as part of the protocol, in parallel.
* timers.[ch]: Uses the timer wheel to implement protocol particular
event timeouts, and gives a set of very simple event-driven entry
point functions for callers.
* main.c, version.h: Initialization and deinitialization of the module.
* selftest/*.h: Runtime unit tests for some of the most security
sensitive functions.
* tools/testing/selftests/wireguard/netns.sh: Aforementioned testing
script using network namespaces.
This commit aims to be as self-contained as possible, implementing
WireGuard as a standalone module not needing much special handling or
coordination from the network subsystem. I expect for future
optimizations to the network stack to positively improve WireGuard, and
vice-versa, but for the time being, this exists as intentionally
standalone.
We introduce a menu option for CONFIG_WIREGUARD, as well as providing a
verbose debug log and self-tests via CONFIG_WIREGUARD_DEBUG.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: David Miller <davem@davemloft.net>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: linux-crypto@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-09 07:27:34 +08:00
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."
