Add release v1.4.1

Only pass $DOCKER_HOST if it is a Unix socket - fixes #1820

.gitignore

@@ -39,24 +39,31 @@ prog/weavehosts/weavehosts
 prog/weaveproxy/weaveproxy
 prog/sigproxy/sigproxy
 prog/weavewait/weavewait
+prog/weavewait/weavewait_noop
+prog/weavewait/weavewait_nomcast
 prog/netcheck/netcheck
+prog/docker_tls_args/docker_tls_args
+prog/plugin/plugin
 testing/cover/cover
 testing/runner/runner
-tools/bin
-tools/build
 releases
 .weaver.uptodate
 .weavedns.uptodate
 .weaveexec.uptodate
-weave.tar
+.dockerplugin.uptodate
+weave.tar.gz
 prog/weaveexec/weave
 prog/weaveexec/weavehosts
 prog/weaveexec/weaveproxy
 prog/weaveexec/weavewait
+prog/weaveexec/weavewait_noop
+prog/weaveexec/weavewait_nomcast
 prog/weaveexec/sigproxy
 prog/weaveexec/netcheck
+prog/weaveexec/docker_tls_args
 prog/weaveexec/docker*.tgz
 Vagrantfile.local
+test/tls/tls
 test/tls/*.pem
 test/coverage
 test/coverage.*

.gitmodules

@@ -0,0 +1,3 @@
+[submodule "tools"]
+	path = tools
+	url = https://github.com/weaveworks/build-tools

.travis.yml

@@ -21,7 +21,6 @@ install:
   - go get ./...
 
 script:
-  - bin/lint .
   - make tests
 
 after_success:

CHANGELOG.md

@@ -1,3 +1,92 @@
+## Release 1.4.1
+
+This is a bug-fix release to cover a few issues that came up since the
+release of 1.4.0.
+
+* Weave would fail to launch when `$DOCKER_HOST` was set to a TCP socket secured with TLS #1820/#1822
+* Weave would fail to stop when run against Docker pre-1.9 #1815/#1817
+* Issue a warning instead of failing on `--with-dns` option to proxy, which was removed #1810/#1812
+* Make `weave version` show the plugin version #1797/#1813
+* Make `weave launch` show when the a container is restarting #1778/#1814
+* Make `weave launch` fail if the plugin is running, for consistency with router and proxy. #1818/#1819
+
+More details in the [change log](https://github.com/weaveworks/weave/issues?q=milestone%3A1.4.1).
+
+## Release 1.4.0
+
+**Highlights**
+
+* The Docker Network plugin can now operate without a cluster store, so it is now run by default.
+* You can now use the fast datapath over trusted links and Weave encryption over untrusted links, in the same network.
+
+Plus many bug fixes and minor enhancements. More details below and in
+the [change log](https://github.com/weaveworks/weave/issues?q=milestone%3A1.4.0).
+
+## Release 1.3.1
+
+**Highlights**
+
+* The minimum Docker version has been increased to 1.6 due to the
+  [upcoming deprecation of Dockerhub access](https://blog.docker.com/2015/10/docker-hub-deprecation-1-5/)
+  for old clients. From December 7th onwards previous versions of the
+  `weave` script will fail to pull down images from the hub; if you
+  are unable to upgrade to 1.3.1 immediately you can work around this
+  by running `weave --local setup` in conjunction with a compatible
+  Docker client installation
+* Docker networking plugin now works with older kernels and allows you
+  to configure the MTU
+
+More details in the
+[change log](https://github.com/weaveworks/weave/issues?q=milestone%3A1.3.1).
+
+## Release 1.3.0
+
+**Highlights**
+
+This release includes a Docker Plugin, so you have the option to use
+Weave Net that way. More details in the
+[change log](https://github.com/weaveworks/weave/issues?q=milestone%3A1.3.0).
+
+## Release 1.2.1
+
+This release contains a number of bug fixes and minor
+enhancements. More details in the
+[change log](https://github.com/weaveworks/weave/issues?q=milestone%3A1.2.1).
+
+The release is fully compatible with 1.2.0 versions, so existing
+clusters can be upgraded incrementally.
+
+## Release 1.2.0
+
+**Highlights**
+
+This release introduces the 
+[Fast Data Path](http://docs.weave.works/weave/master/head/features.html#fast-data-path),
+which allows Weave networks to operate at near wire-level speeds. This
+new feature is enabled by default.
+
+Other highlights:
+
+- auto-configuration of TLS for the Weave Docker API proxy, making it
+  easier to run Weave on Macs and in conjunction with tools like
+  Docker Swarm
+- support for restart policies on application containers and weave
+  infrastructure containers
+- better compatibility with recent and future Docker versions
+
+More details in the 
+[change log](https://github.com/weaveworks/weave/issues?q=milestone%3A1.2.0).
+
+## Release 1.1.2
+
+This release contains a small number of bug fixes. More details in the
+[change log](https://github.com/weaveworks/weave/issues?q=milestone%3A1.1.2).
+
+The release is fully compatible with 1.1.0 versions, so existing
+clusters can be upgraded incrementally. When upgrading from 1.0.x,
+note the compatibility information in the
+[Installation & Upgrading instructions for Weave 1.1.0](https://github.com/weaveworks/weave/releases/tag/v1.1.0).
+
 ## Release 1.1.1
 
 This release contains a number of bug fixes and minor

Makefile

@@ -1,7 +1,7 @@
 PUBLISH=publish_weave publish_weaveexec
 
 .DEFAULT: all
-.PHONY: all update tests publish $(PUBLISH) clean prerequisites build travis run-smoketests
+.PHONY: all update tests lint publish $(PUBLISH) clean clean-bin prerequisites build travis run-smoketests
 
 # If you can use docker without being root, you can do "make SUDO="
 SUDO=sudo
@@ -13,26 +13,31 @@ WEAVER_EXE=prog/weaver/weaver
 WEAVEPROXY_EXE=prog/weaveproxy/weaveproxy
 SIGPROXY_EXE=prog/sigproxy/sigproxy
 WEAVEWAIT_EXE=prog/weavewait/weavewait
-WEAVEHOSTS_EXE=prog/weavehosts/weavehosts
+WEAVEWAIT_NOOP_EXE=prog/weavewait/weavewait_noop
+WEAVEWAIT_NOMCAST_EXE=prog/weavewait/weavewait_nomcast
 NETCHECK_EXE=prog/netcheck/netcheck
-COVER_EXE=testing/cover/cover
-RUNNER_EXE=testing/runner/runner
+DOCKERTLSARGS_EXE=prog/docker_tls_args/docker_tls_args
+DOCKERPLUGIN_EXE=prog/plugin/plugin
+RUNNER_EXE=tools/runner/runner
+TEST_TLS_EXE=test/tls/tls
 
-EXES=$(WEAVER_EXE) $(SIGPROXY_EXE) $(WEAVEPROXY_EXE) $(WEAVEWAIT_EXE) $(WEAVEHOSTS_EXE) $(NETCHECK_EXE) $(COVER_EXE) $(RUNNER_EXE)
+EXES=$(WEAVER_EXE) $(SIGPROXY_EXE) $(WEAVEPROXY_EXE) $(WEAVEWAIT_EXE) $(WEAVEWAIT_NOOP_EXE) $(WEAVEWAIT_NOMCAST_EXE) $(NETCHECK_EXE) $(DOCKERTLSARGS_EXE) $(DOCKERPLUGIN_EXE) $(TEST_TLS_EXE)
 
 WEAVER_UPTODATE=.weaver.uptodate
 WEAVEEXEC_UPTODATE=.weaveexec.uptodate
+DOCKERPLUGIN_UPTODATE=.dockerplugin.uptodate
 
-IMAGES_UPTODATE=$(WEAVER_UPTODATE) $(WEAVEEXEC_UPTODATE)
+IMAGES_UPTODATE=$(WEAVER_UPTODATE) $(WEAVEEXEC_UPTODATE) $(DOCKERPLUGIN_UPTODATE)
 
 WEAVER_IMAGE=$(DOCKERHUB_USER)/weave
 WEAVEEXEC_IMAGE=$(DOCKERHUB_USER)/weaveexec
+DOCKERPLUGIN_IMAGE=$(DOCKERHUB_USER)/plugin
 
-IMAGES=$(WEAVER_IMAGE) $(WEAVEEXEC_IMAGE)
+IMAGES=$(WEAVER_IMAGE) $(WEAVEEXEC_IMAGE) $(DOCKERPLUGIN_IMAGE)
 
-WEAVE_EXPORT=weave.tar
+WEAVE_EXPORT=weave.tar.gz
 
-WEAVEEXEC_DOCKER_VERSION=1.3.1
+WEAVEEXEC_DOCKER_VERSION=1.6.2
 DOCKER_DISTRIB=prog/weaveexec/docker-$(WEAVEEXEC_DOCKER_VERSION).tgz
 DOCKER_DISTRIB_URL=https://get.docker.com/builds/Linux/x86_64/docker-$(WEAVEEXEC_DOCKER_VERSION).tgz
 NETGO_CHECK=@strings $@ | grep cgo_stub\\\.go >/dev/null || { \
@@ -45,7 +50,9 @@ NETGO_CHECK=@strings $@ | grep cgo_stub\\\.go >/dev/null || { \
 }
 BUILD_FLAGS=-ldflags "-extldflags \"-static\" -X main.version $(WEAVE_VERSION)" -tags netgo
 
-all: $(WEAVE_EXPORT) $(COVER_EXE) $(RUNNER_EXE)
+PACKAGE_BASE=$(shell go list -e ./)
+
+all: $(WEAVE_EXPORT) $(RUNNER_EXE) $(TEST_TLS_EXE)
 
 travis: $(EXES)
 
@@ -54,7 +61,7 @@ update:
 
 $(WEAVER_EXE) $(WEAVEPROXY_EXE): common/*.go common/*/*.go net/*.go
 ifeq ($(COVERAGE),true)
-	$(eval COVERAGE_MODULES := $(shell (go list ./$(@D); go list -f '{{join .Deps "\n"}}' ./$(@D) | grep "weaveworks") | paste -s -d,))
+	$(eval COVERAGE_MODULES := $(shell (go list ./$(@D); go list -f '{{join .Deps "\n"}}' ./$(@D) | grep "^$(PACKAGE_BASE)/") | paste -s -d,))
 	go get -t -tags netgo ./$(@D)
 	go test -c -o ./$@ $(BUILD_FLAGS) -v -covermode=atomic -coverpkg $(COVERAGE_MODULES) ./$(@D)/
 else
@@ -68,59 +75,87 @@ $(NETCHECK_EXE): common/*.go common/*/*.go net/*.go
 	go build $(BUILD_FLAGS) -o $@ ./$(@D)
 	$(NETGO_CHECK)
 
-$(WEAVER_EXE): router/*.go ipam/*.go ipam/*/*.go nameserver/*.go prog/weaver/*.go
+$(WEAVER_EXE): router/*.go mesh/*.go ipam/*.go ipam/*/*.go nameserver/*.go prog/weaver/*.go
 $(WEAVEPROXY_EXE): proxy/*.go prog/weaveproxy/main.go
 $(NETCHECK_EXE): prog/netcheck/netcheck.go
 
-# Sigproxy and weavewait need separate rules as they fail the netgo check in
+# These next programs need separate rules as they fail the netgo check in
 # the main build stanza due to not importing net package
 $(SIGPROXY_EXE): prog/sigproxy/main.go
-$(WEAVEWAIT_EXE): prog/weavewait/main.go net/*.go
-$(WEAVEHOSTS_EXE): prog/weavehosts/weavehosts.go
-$(COVER_EXE): testing/cover/cover.go
-$(RUNNER_EXE): testing/runner/runner.go
+$(DOCKERTLSARGS_EXE): prog/docker_tls_args/*.go
+$(DOCKERPLUGIN_EXE): prog/plugin/*.go plugin/net/*.go plugin/ipam/*.go plugin/skel/*.go api/*.go common/docker/*.go
+$(TEST_TLS_EXE): test/tls/*.go
 
-$(WEAVEWAIT_EXE) $(SIGPROXY_EXE) $(WEAVEHOSTS_EXE) $(COVER_EXE) $(RUNNER_EXE):
-	go get ./$(@D)
-	go build -o $@ ./$(@D)
+$(SIGPROXY_EXE) $(DOCKERTLSARGS_EXE) $(DOCKERPLUGIN_EXE) $(TEST_TLS_EXE):
+	go get -tags netgo ./$(@D)
+	go build $(BUILD_FLAGS) -o $@ ./$(@D)
+
+$(WEAVEWAIT_EXE): prog/weavewait/*.go net/*.go
+	go get -tags netgo ./$(@D)
+	go build $(BUILD_FLAGS) -tags "netgo iface mcast" -o $@ ./$(@D)
+
+$(WEAVEWAIT_NOMCAST_EXE): prog/weavewait/*.go net/*.go
+	go get -tags netgo ./$(@D)
+	go build $(BUILD_FLAGS) -tags "netgo iface" -o $@ ./$(@D)
+
+$(WEAVEWAIT_NOOP_EXE): prog/weavewait/*.go
+	go get -tags netgo ./$(@D)
+	go build $(BUILD_FLAGS) -o $@ ./$(@D)
 
 $(WEAVER_UPTODATE): prog/weaver/Dockerfile $(WEAVER_EXE)
-	$(SUDO) docker build -t $(WEAVER_IMAGE) prog/weaver
+	$(SUDO) DOCKER_HOST=$(DOCKER_HOST) docker build -t $(WEAVER_IMAGE) prog/weaver
 	touch $@
 
-$(WEAVEEXEC_UPTODATE): prog/weaveexec/Dockerfile $(DOCKER_DISTRIB) weave $(SIGPROXY_EXE) $(WEAVEPROXY_EXE) $(WEAVEWAIT_EXE) $(WEAVEHOSTS_EXE) $(NETCHECK_EXE)
+$(WEAVEEXEC_UPTODATE): prog/weaveexec/Dockerfile prog/weaveexec/symlink $(DOCKER_DISTRIB) weave $(SIGPROXY_EXE) $(WEAVEPROXY_EXE) $(WEAVEWAIT_EXE) $(WEAVEWAIT_NOOP_EXE) $(WEAVEWAIT_NOMCAST_EXE) $(NETCHECK_EXE) $(DOCKERTLSARGS_EXE)
 	cp weave prog/weaveexec/weave
 	cp $(SIGPROXY_EXE) prog/weaveexec/sigproxy
 	cp $(WEAVEPROXY_EXE) prog/weaveexec/weaveproxy
 	cp $(WEAVEWAIT_EXE) prog/weaveexec/weavewait
-	cp $(WEAVEHOSTS_EXE) prog/weaveexec/weavehosts
+	cp $(WEAVEWAIT_NOOP_EXE) prog/weaveexec/weavewait_noop
+	cp $(WEAVEWAIT_NOMCAST_EXE) prog/weaveexec/weavewait_nomcast
 	cp $(NETCHECK_EXE) prog/weaveexec/netcheck
+	cp $(DOCKERTLSARGS_EXE) prog/weaveexec/docker_tls_args
 	cp $(DOCKER_DISTRIB) prog/weaveexec/docker.tgz
-	$(SUDO) docker build -t $(WEAVEEXEC_IMAGE) prog/weaveexec
+	$(SUDO) DOCKER_HOST=$(DOCKER_HOST) docker build -t $(WEAVEEXEC_IMAGE) prog/weaveexec
+	touch $@
+
+$(DOCKERPLUGIN_UPTODATE): prog/plugin/Dockerfile $(DOCKERPLUGIN_EXE)
+	$(SUDO) docker build -t $(DOCKERPLUGIN_IMAGE) prog/plugin
 	touch $@
 
 $(WEAVE_EXPORT): $(IMAGES_UPTODATE)
-	$(SUDO) docker save $(addsuffix :latest,$(IMAGES)) > $@
+	$(SUDO) DOCKER_HOST=$(DOCKER_HOST) docker save $(addsuffix :latest,$(IMAGES)) | gzip > $@
 
 $(DOCKER_DISTRIB):
 	curl -o $(DOCKER_DISTRIB) $(DOCKER_DISTRIB_URL)
 
-tests: $(COVER_EXE)
-	@test/units.sh
+tests: tools/.git
+	tools/test
+
+lint: tools/.git
+	tools/lint -nocomment -notestpackage .
+
+tools/.git:
+	git submodule update --init
+
+$(RUNNER_EXE): tools/.git
+	make -C tools/runner
 
 $(PUBLISH): publish_%: $(IMAGES_UPTODATE)
-	$(SUDO) docker tag -f $(DOCKERHUB_USER)/$* $(DOCKERHUB_USER)/$*:$(WEAVE_VERSION)
-	$(SUDO) docker push   $(DOCKERHUB_USER)/$*:$(WEAVE_VERSION)
+	$(SUDO) DOCKER_HOST=$(DOCKER_HOST) docker tag -f $(DOCKERHUB_USER)/$* $(DOCKERHUB_USER)/$*:$(WEAVE_VERSION)
+	$(SUDO) DOCKER_HOST=$(DOCKER_HOST) docker push   $(DOCKERHUB_USER)/$*:$(WEAVE_VERSION)
 ifneq ($(UPDATE_LATEST),false)
-	$(SUDO) docker push   $(DOCKERHUB_USER)/$*:latest
+	$(SUDO) DOCKER_HOST=$(DOCKER_HOST) docker push   $(DOCKERHUB_USER)/$*:latest
 endif
 
 publish: $(PUBLISH)
 
-clean:
-	-$(SUDO) docker rmi $(IMAGES)
-	go clean -r ./...
+clean-bin:
+	-$(SUDO) DOCKER_HOST=$(DOCKER_HOST) docker rmi $(IMAGES)
+	go clean -r $(addprefix ./,$(dir $(EXES)))
 	rm -f $(EXES) $(IMAGES_UPTODATE) $(WEAVE_EXPORT)
+
+clean: clean-bin
 	rm -rf test/tls/*.pem test/coverage.* test/coverage
 
 build:

README.md

@@ -7,11 +7,9 @@
 [Weaveworks](http://weave.works) is the company that develops Weave -
 the most productive way for developers to connect, observe and control
 Docker containers. To learn about our products, including getting
-started tutorials, visit our [website](http://weave.works), read about
-the
-[1.0 release of Weave](http://blog.weave.works/2015/06/18/weave-1-0-is-out-and-it-is-good/),
-visit our [documentation](http://docs.weave.works) or continue to read
-about some of the more technical aspects of Weave in this readme.
+started tutorials, visit our [website](http://weave.works) and
+[documentation](http://docs.weave.works) or continue to read about
+some of the more technical aspects of Weave in this readme.
 
 # Weave
 
@@ -43,19 +41,29 @@ capabilities, so these can continue to be used by containers.
 ## Installation
 
 Ensure you are running Linux (kernel 3.8 or later) and have Docker
-(version 1.3.1 or later) installed. Then install weave with
+(version 1.6.0 or later) installed. Then install weave with
 
     sudo curl -L git.io/weave -o /usr/local/bin/weave
     sudo chmod a+x /usr/local/bin/weave
 
+For usage on OSX (with Docker Machine) you first need to make sure
+that a VM is running and configured, as shown in [the Docker Machine
+documentation](https://docs.docker.com/installation/mac/#from-your-shell).
+Then you can launch weave directly from the OSX host.
+
 CoreOS users see [here](https://github.com/fintanr/weave-gs/blob/master/coreos-simple/user-data) for an example of installing weave using cloud-config.
 
+Amazon ECS users see
+[here](https://github.com/weaveworks/guides/blob/master/aws-ecs/LATESTAMIs.md)
+for the latest Weave AMIs and
+[here](http://weave.works/guides/service-discovery-with-weave-aws-ecs.html) to get started with Weave on ECS.
+
 Weave respects the environment variable `DOCKER_HOST`, so you can run
 it locally to control a weave network on a remote host.
 
 ## Quick Start Screencast
 
-<a href="https://youtu.be/BF9YJSPQUcY" alt="Click to watch the screencast" target="_blank">
+<a href="https://youtu.be/kihQCCT1ykE" alt="Click to watch the screencast" target="_blank">
   <img src="/docs/hello-screencast.png" />
 </a>
 
@@ -99,9 +107,8 @@ The only difference, apart from the name of the application container,
 is that we tell our weave that it should peer with the weave on
 `$HOST1` (specified as the IP address or hostname, and optional
 `:port`, by which `$HOST2` can reach it). NB: if there is a firewall
-between `$HOST1` and `$HOST2`, you must open the weave port (6783 by
-default; this can be overriden by setting `WEAVE_PORT`) for TCP and
-UDP.
+between `$HOST1` and `$HOST2`, you must permit traffic to the weave
+control port (TCP 6783) and data ports (UDP 6783/6784).
 
 Note that we could instead have told the weave on `$HOST1` to connect to
 `$HOST2`, or told both about each other. Order does not matter here;
@@ -148,12 +155,16 @@ to each other.
 
 ## Contact Us
 
-Found a bug, want to suggest a feature, or have a question?
-[File an issue](https://github.com/weaveworks/weave/issues), or email
-help@weave.works. When reporting a bug, please include which version of
-weave you are running, as shown by `weave version`.
+Found a bug, want to suggest a feature, or have a question?  You can
+[File an issue](https://github.com/weaveworks/weave/issues).  And
+there is a [Weave Users Google
+Group](https://groups.google.com/a/weave.works/forum/#!forum/weave-users),
+which you can post to by email at weave-users@weave.works.  Or you can
+contact Weaveworks directly at help@weave.works. When reporting a
+bug, please include which version of weave you are running, as shown
+by `weave version`.
 
-Follow weave on Twitter:
+Follow on Twitter:
 [@weaveworks](https://twitter.com/weaveworks).
 
 Read the Weave blog:

api/dns.go

@@ -0,0 +1,18 @@
+package api
+
+import (
+	"fmt"
+	"net/url"
+)
+
+func (client *Client) RegisterWithDNS(ID string, fqdn string, ip string) error {
+	data := url.Values{}
+	data.Add("fqdn", fqdn)
+	_, err := client.httpVerb("PUT", fmt.Sprintf("/name/%s/%s", ID, ip), data)
+	return err
+}
+
+func (client *Client) DeregisterWithDNS(ID string, ip string) error {
+	_, err := client.httpVerb("DELETE", fmt.Sprintf("/name/%s/%s", ID, ip), nil)
+	return err
+}

api/ipam.go

@@ -0,0 +1,49 @@
+package api
+
+import (
+	"fmt"
+	"net"
+)
+
+func (client *Client) ipamOp(ID string, op string) (*net.IPNet, error) {
+	ip, err := client.httpVerb(op, fmt.Sprintf("/ip/%s", ID), nil)
+	if err != nil {
+		return nil, err
+	}
+	return parseIP(ip)
+}
+
+// returns an IP for the ID given, allocating a fresh one if necessary
+func (client *Client) AllocateIP(ID string) (*net.IPNet, error) {
+	return client.ipamOp(ID, "POST")
+}
+
+// returns an IP for the ID given, or nil if one has not been
+// allocated
+func (client *Client) LookupIP(ID string) (*net.IPNet, error) {
+	return client.ipamOp(ID, "GET")
+}
+
+// release an IP which is no longer needed
+func (client *Client) ReleaseIP(ID string) error {
+	_, err := client.ipamOp(ID, "DELETE")
+	return err
+}
+
+func (client *Client) DefaultSubnet() (*net.IPNet, error) {
+	cidr, err := client.httpVerb("GET", fmt.Sprintf("/ipinfo/defaultsubnet"), nil)
+	if err != nil {
+		return nil, err
+	}
+	_, ipnet, err := net.ParseCIDR(cidr)
+	return ipnet, err
+}
+
+func parseIP(body string) (*net.IPNet, error) {
+	ip, ipnet, err := net.ParseCIDR(string(body))
+	if err != nil {
+		return nil, err
+	}
+	ipnet.IP = ip
+	return ipnet, nil
+}

api/weaveapi.go

@@ -0,0 +1,73 @@
+package api
+
+import (
+	"errors"
+	"fmt"
+	"io"
+	"io/ioutil"
+	"net/http"
+	"net/url"
+	"strings"
+
+	. "github.com/weaveworks/weave/common"
+)
+
+const (
+	WeaveHttpPort = 6784
+)
+
+type Client struct {
+	baseUrl string
+	resolve func() (string, error)
+}
+
+func (client *Client) httpVerb(verb string, url string, values url.Values) (string, error) {
+	baseUrl := client.baseUrl
+	if client.resolve != nil {
+		addr, err := client.resolve()
+		if err != nil {
+			return "", err
+		}
+		baseUrl = fmt.Sprintf("http://%s:%d", addr, WeaveHttpPort)
+	}
+	url = baseUrl + url
+	Log.Debugf("weave %s to %s with %v", verb, url, values)
+	var body io.Reader
+	if values != nil {
+		body = strings.NewReader(values.Encode())
+	}
+	req, err := http.NewRequest(verb, url, body)
+	if err != nil {
+		return "", err
+	}
+	if values != nil {
+		req.Header.Set("Content-Type", "application/x-www-form-urlencoded")
+	}
+	if resp, err := http.DefaultClient.Do(req); err != nil {
+		return "", err
+	} else {
+		defer resp.Body.Close()
+		body, err := ioutil.ReadAll(resp.Body)
+		if err != nil {
+			return "", err
+		}
+		if resp.StatusCode == http.StatusOK || resp.StatusCode == http.StatusNoContent {
+			return string(body), nil
+		} else {
+			return "", errors.New(resp.Status + ": " + string(body))
+		}
+	}
+}
+
+func NewClient(addr string) *Client {
+	return &Client{baseUrl: fmt.Sprintf("http://%s:%d", addr, WeaveHttpPort)}
+}
+
+func NewClientWithResolver(resolver func() (string, error)) *Client {
+	return &Client{resolve: resolver}
+}
+
+func (client *Client) Connect(remote string) error {
+	_, err := client.httpVerb("POST", "/connect", url.Values{"peer": {remote}})
+	return err
+}

bin/lint

@@ -1,86 +0,0 @@
-#!/bin/bash
-# This scipt lints go files for common errors.
-#
-# Its runs gofmt and go vet, and optionally golint and
-# gocyclo, if they are installed.
-#
-# With no arguments, it lints the current files staged
-# for git commit.  Or you can pass it explicit filenames
-# (or directories) and it will lint them.
-#
-# To use this script automatically, run:
-#   ln -s ../../bin/lint .git/hooks/pre-commit
-
-set -eu
-
-function lint_go {
-	filename="$1"
-	result=0
-
-	if [ -n "$(gofmt -l "${filename}")" ]; then
-		result=1
-		echo "${filename}: run gofmt -w ${filename}!"
-	fi
-
-	go tool vet "${filename}" || result=$?
-
-	# golint is completely optional.  If you don't like it
-	# don't have it installed.
-	if type golint >/dev/null 2>&1; then
-		# golint doesn't set an exit code it seems
-		lintoutput=$(golint "${filename}" | grep -vE 'comment|dot imports')
-		if [ "$lintoutput" != "" ]; then
-			result=1
-			echo "$lintoutput"
-		fi
-	fi
-
-	# gocyclo is completely optional.  If you don't like it
-	# don't have it installed.  Also never blocks a commit,
-	# it just warns.
-	if type gocyclo >/dev/null 2>&1; then
-		gocyclo -over 25 "${filename}" | while read line; do
-			echo "${filename}": higher than 25 cyclomatic complexity - "${line}"
-		done
-	fi
-
-	return $result
-}
-
-function lint {
-	filename="$1"
-	ext="${filename##*\.}"
-
-	# Don't lint this script
-	if [ "${filename}" = "bin/lint" ]; then
-		return
-	fi
-
-	# Don't lint files that don't exist
-	if ! [ -a "${filename}" ]; then
-		return
-	fi
-
-	case "$ext" in
-		go) lint_go "${filename}"
-		;;
-	esac
-}
-
-function lint_files {
-	lint_result=0
-	while read filename; do
-		lint "${filename}" || lint_result=1
-	done
-	exit $lint_result
-}
-
-function list_files {
-	if [ $# -gt 0 ]; then
-		find "$@" -type f
-	else
-		git diff --cached --name-only
-	fi
-}
-
-list_files "$@" | lint_files

bin/multiweave

@@ -29,6 +29,7 @@ random_mac() {
 case "$1" in
     launch)
         shift 1
+        echo $(date -R) launching
         for i in $(seq $START $FINISH); do
             WEAVE_CONTAINER_NAME=weave$i WEAVE_PORT=$((5000+i)) $weavedir/weave launch-router --iface '' --name $(random_mac) --nickname weave$i --port $PORT --no-dns "$@"
         done
@@ -36,6 +37,7 @@ case "$1" in
             status=$(docker inspect --format='{{.State.Running}} {{.NetworkSettings.IPAddress}}' weave$((START-1)))
             PREV="${status#true }"
         fi
+        echo $(date -R) connecting
         for i in $(seq $START $FINISH); do
             if [ -n "$PREV" ] ; then
                 echo connecting weave$i to $PREV
@@ -44,6 +46,7 @@ case "$1" in
             status=$(docker inspect --format='{{.State.Running}} {{.NetworkSettings.IPAddress}}' weave$i)
             PREV="${status#true }"
         done
+        echo $(date -R) done
         ;;
     stop)
         for i in $(seq $START $FINISH); do

build/Dockerfile

@@ -3,7 +3,7 @@ FROM ubuntu
 RUN apt-get -y update && apt-get -y install --no-install-recommends ca-certificates apt-transport-https
 RUN apt-key adv --keyserver hkp://keyserver.ubuntu.com:80 --recv-keys 36A1D7869245C8950F966E92D8576A8BA88D21E9
 RUN echo deb https://get.docker.io/ubuntu docker main > /etc/apt/sources.list.d/docker.list
-RUN apt-get -y update && apt-get -y install --no-install-recommends build-essential git lxc-docker-1.3.1 mercurial libpcap-dev curl make pkg-config gcc bison flex python-requests
+RUN apt-get -y update && apt-get -y install --no-install-recommends build-essential git lxc-docker-1.6.2 mercurial libpcap-dev curl make pkg-config gcc bison flex python-requests
 
 # When doing a build in a container, "apt-get update" happens twice,
 # which can be a very significant overhead for incremental builds.

build/build.sh

@@ -36,11 +36,7 @@ EOF
     echo "weave:*:::::::" >>/etc/shadow
     echo "weave	ALL=(ALL)	NOPASSWD: ALL" >>/etc/sudoers
 
-    if [ "$1" = "tests" ] ; then
-        su weave -c "PATH=$PATH make -C $WEAVE_SRC tests"
-    else
-        su weave -c "PATH=$PATH make -C $WEAVE_SRC build"
-    fi
+    su weave -c "PATH=$PATH make -C $WEAVE_SRC build"
 else
     # There are arguments to pass to git-clone
     mkdir -p ${WEAVE_SRC%/*}

circle.yml

@@ -12,41 +12,45 @@ machine:
     SRCDIR: /home/ubuntu/src/github.com/weaveworks/weave
     PATH: $PATH:$HOME/.local/bin:$HOME/google-cloud-sdk/bin/
     CLOUDSDK_CORE_DISABLE_PROMPTS: 1
-    WEAVE_BUILD: $HOME/docker/weave-build.tar
+    WEAVE_TEST: $HOME/docker/weave-test.tar
 
 dependencies:
   cache_directories:
     - "~/docker"
   override:
-    - if [[ -e "$WEAVE_BUILD" ]]; then
-        docker load -i $WEAVE_BUILD;
+    - if [[ -e "$WEAVE_TEST" ]]; then
+        docker load -i $WEAVE_TEST;
       else
         docker build -t weaveworks/weave-build build;
-        mkdir -p $(dirname "$WEAVE_BUILD");
-        docker save weaveworks/weave-build >$WEAVE_BUILD;
+        docker build -t weaveworks/weave-test test/image;
+        mkdir -p $(dirname "$WEAVE_TEST");
+        docker save weaveworks/weave-test >$WEAVE_TEST;
       fi
   post:
-    - sudo apt-get update
-    - sudo apt-get install bc jq
-    - pip install requests
+    - sudo apt-get update && sudo apt-get install bc jq && pip install requests
     - curl https://sdk.cloud.google.com | bash
     - bin/setup-circleci-secrets "$SECRET_PASSWORD"
-    - mkdir -p $(dirname $SRCDIR)
-    - cp -r $(pwd)/ $SRCDIR
+    - mkdir -p $(dirname $SRCDIR) && cp -r $(pwd)/ $SRCDIR
+    - cd $SRCDIR/test; ./gce.sh make_template:
+        parallel: false
 
 test:
   override:
-    - docker run -v /var/run/docker.sock:/run/docker.sock -v /home/ubuntu:/home/go -e COVERDIR=test/coverage -e SLOW=true -e CIRCLECI -e CIRCLE_BUILD_NUM -e CIRCLE_NODE_TOTAL -e CIRCLE_NODE_INDEX weaveworks/weave-build tests:
+    - cd $SRCDIR; make lint:
+        parallel: true
+    - docker run -v /home/ubuntu:/home/go -e COVERDIR=test/coverage -e SLOW=true -e CIRCLECI -e CIRCLE_BUILD_NUM -e CIRCLE_NODE_TOTAL -e CIRCLE_NODE_INDEX weaveworks/weave-test:
         parallel: true
     - docker run -v /var/run/docker.sock:/run/docker.sock -v /home/ubuntu:/home/go -e COVERAGE=true weaveworks/weave-build:
         parallel: true
-    - cd $SRCDIR/test; ./gce.sh setup:
-        parallel: true
-    - cd $SRCDIR/test; eval $(./gce.sh hosts); ./setup.sh:
+    - cd $SRCDIR/test; ./gce.sh setup && eval $(./gce.sh hosts) && ./setup.sh:
         parallel: true
     - cd $SRCDIR/test; eval $(./gce.sh hosts); export COVERAGE=true; ./run_all.sh:
         parallel: true
         timeout: 300
+    - cd $SRCDIR; make clean-bin:
+        parallel: true
+    - docker run -v /var/run/docker.sock:/run/docker.sock -v /home/ubuntu:/home/go weaveworks/weave-build:
+        parallel: true
   post:
     - cd $SRCDIR/test; ./gce.sh destroy:
         parallel: true
@@ -59,3 +63,11 @@ teardown:
         parallel: true
     - test "$CIRCLE_NODE_INDEX" != "0" || (cd $SRCDIR/test; cp coverage.* $CIRCLE_ARTIFACTS):
         parallel: true
+
+deployment:
+  hub:
+    branch: master
+    owner: weaveworks
+    commands:
+      - docker login -e $DOCKER_EMAIL -u $DOCKER_USER -p $DOCKER_PASS
+      - cd $SRCDIR; make publish

common/docker/client.go

@@ -1,12 +1,18 @@
 package docker
 
 import (
-	"github.com/fsouza/go-dockerclient"
+	"errors"
+	"fmt"
+	"strings"
+
+	docker "github.com/fsouza/go-dockerclient"
+
 	. "github.com/weaveworks/weave/common"
 )
 
 // An observer for container events
 type ContainerObserver interface {
+	ContainerStarted(ident string)
 	ContainerDied(ident string)
 }
 
@@ -16,18 +22,52 @@ type Client struct {
 
 // NewClient creates a new Docker client and checks we can talk to Docker
 func NewClient(apiPath string) (*Client, error) {
+	if apiPath != "" && !strings.Contains(apiPath, "://") {
+		apiPath = "tcp://" + apiPath
+	}
 	dc, err := docker.NewClient(apiPath)
 	if err != nil {
 		return nil, err
 	}
 	client := &Client{dc}
 
-	env, err := client.Version()
+	return client, client.checkWorking()
+}
+
+func NewVersionedClient(apiPath string, apiVersionString string) (*Client, error) {
+	if !strings.Contains(apiPath, "://") {
+		apiPath = "tcp://" + apiPath
+	}
+	dc, err := docker.NewVersionedClient(apiPath, apiVersionString)
 	if err != nil {
 		return nil, err
 	}
-	Log.Infof("[docker] Using Docker API on %s: %v", apiPath, env)
-	return client, nil
+	client := &Client{dc}
+
+	return client, client.checkWorking()
+}
+
+func NewVersionedClientFromEnv(apiVersionString string) (*Client, error) {
+	dc, err := docker.NewVersionedClientFromEnv(apiVersionString)
+	if err != nil {
+		return nil, err
+	}
+	client := &Client{dc}
+
+	return client, client.checkWorking()
+}
+
+func (c *Client) checkWorking() error {
+	_, err := c.Version()
+	return err
+}
+
+func (c *Client) Info() string {
+	if env, err := c.Version(); err != nil {
+		return fmt.Sprintf("Docker API error: %s", err)
+	} else {
+		return fmt.Sprintf("Docker API on %s: %v", c.Endpoint(), env)
+	}
 }
 
 // AddObserver adds an observer for docker events
@@ -41,6 +81,9 @@ func (c *Client) AddObserver(ob ContainerObserver) error {
 	go func() {
 		for event := range events {
 			switch event.Status {
+			case "start":
+				id := event.ID
+				ob.ContainerStarted(id)
 			case "die":
 				id := event.ID
 				ob.ContainerDied(id)
@@ -62,3 +105,28 @@ func (c *Client) IsContainerNotRunning(idStr string) bool {
 	Log.Errorf("[docker] Could not check container status: %s", err)
 	return false
 }
+
+// This is intended to find an IP address that we can reach the container on;
+// if it is on the Docker bridge network then that address; if on the host network
+// then localhost
+func (c *Client) GetContainerIP(nameOrID string) (string, error) {
+	Log.Debugf("Getting IP for container %s", nameOrID)
+	info, err := c.InspectContainer(nameOrID)
+	if err != nil {
+		return "", err
+	}
+	if info.NetworkSettings.Networks != nil {
+		Log.Debugln("Networks: ", info.NetworkSettings.Networks)
+		if bridgeNetwork, ok := info.NetworkSettings.Networks["bridge"]; ok {
+			return bridgeNetwork.IPAddress, nil
+		} else if _, ok := info.NetworkSettings.Networks["host"]; ok {
+			return "127.0.0.1", nil
+		}
+	} else if info.HostConfig.NetworkMode == "host" {
+		return "127.0.0.1", nil
+	}
+	if info.NetworkSettings.IPAddress == "" {
+		return "", errors.New("No IP address found for container " + nameOrID)
+	}
+	return info.NetworkSettings.IPAddress, nil
+}

common/logging.go

@@ -52,3 +52,15 @@ func SetLogLevel(levelname string) {
 	}
 	Log.Level = level
 }
+
+func CheckFatal(e error) {
+	if e != nil {
+		Log.Fatal(e)
+	}
+}
+
+func CheckWarn(e error) {
+	if e != nil {
+		Log.Warnln(e)
+	}
+}

common/odp/odp.go

@@ -0,0 +1,93 @@
+package odp
+
+import (
+	"fmt"
+	"net"
+	"syscall"
+
+	"github.com/weaveworks/go-odp/odp"
+)
+
+// ODP admin functionality
+
+func CreateDatapath(dpname string) (err error, supported bool) {
+	dpif, err := odp.NewDpif()
+	if err != nil {
+		if odp.IsKernelLacksODPError(err) {
+			return nil, false
+		}
+
+		return err, true
+	}
+
+	defer dpif.Close()
+
+	dp, err := dpif.CreateDatapath(dpname)
+	if err != nil && !odp.IsDatapathNameAlreadyExistsError(err) {
+		return err, true
+	}
+
+	// Pick an ephemeral port number to use in probing for vxlan
+	// support.
+	udpconn, err := net.ListenUDP("udp4", nil)
+	if err != nil {
+		return err, true
+	}
+
+	// we leave the UDP socket open, so creating a vxlan vport on
+	// the same port number should fail.  But that's fine: It's
+	// still sufficient to probe for support.
+	portno := uint16(udpconn.LocalAddr().(*net.UDPAddr).Port)
+	vpid, err := dp.CreateVport(odp.NewVxlanVportSpec(
+		fmt.Sprintf("vxlan-%d", portno), portno))
+	if nlerr, ok := err.(odp.NetlinkError); ok {
+		if syscall.Errno(nlerr) == syscall.EAFNOSUPPORT {
+			dp.Delete()
+			return fmt.Errorf("kernel does not have Open vSwitch VXLAN support"), false
+		}
+	}
+
+	if err == nil {
+		dp.DeleteVport(vpid)
+	}
+
+	udpconn.Close()
+	return nil, true
+}
+
+func DeleteDatapath(dpname string) error {
+	dpif, err := odp.NewDpif()
+	if err != nil {
+		return err
+	}
+
+	defer dpif.Close()
+
+	dp, err := dpif.LookupDatapath(dpname)
+	if err != nil {
+		if odp.IsNoSuchDatapathError(err) {
+			return nil
+		}
+
+		return err
+	}
+
+	return dp.Delete()
+}
+
+func AddDatapathInterface(dpname string, ifname string) error {
+	dpif, err := odp.NewDpif()
+	if err != nil {
+		return err
+	}
+
+	defer dpif.Close()
+
+	dp, err := dpif.LookupDatapath(dpname)
+	if err != nil {
+		return err
+	}
+
+	_, err = dp.CreateVport(odp.NewNetdevVportSpec(ifname))
+	return err
+}

common/signals.go

@@ -14,7 +14,7 @@ type SignalReceiver interface {
 
 func SignalHandlerLoop(ss ...SignalReceiver) {
 	sigs := make(chan os.Signal, 1)
-	signal.Notify(sigs, syscall.SIGINT, syscall.SIGQUIT, syscall.SIGUSR1, syscall.SIGTERM)
+	signal.Notify(sigs, syscall.SIGINT, syscall.SIGQUIT, syscall.SIGTERM)
 	buf := make([]byte, 1<<20)
 	for {
 		switch <-sigs {

docs/architecture.txt

@@ -66,23 +66,23 @@ Router:
     simply responds to traffic received from the remote peer via TCP
   5. We register this connection with the local peer
   6a. If we initiated this connection then we now start sending fast
-    heartbeats to the remote peer so that the remote peer can determine what
-    address/port it should use to send UDP back to us. To do this, we
-    spawn off two further threads which are the forwarder
-    loops. These receive frames which are to be sent to the remote
-    peer. One of them sends frames without the DF flag set. To do this
-    it just sends the packets out of the UDP Listener socket. The
-    other needs to send frames with the DF flag set and needs its own
-    socket so that it can do PMTU discovery easily. To do this, it
-    uses a Raw IP socket (IP has no ports, so there's no collision
-    issue with the UDP Listener socket) and so it must add UDP headers
-    itself.
-  6b. If we did not initiate this connection then the UDP Listener
+    heartbeats to the remote peer so that the remote peer can
+    determine what address/port it should use to send UDP back to
+    us. To do this, we spawn off a "forwarder" thread to send
+    heartbeats, monitor incoming heartbeats, and some other auxiliary
+    duties.  It also consumes frames to be encapsulated and send via
+    UDP from two channels, for DF and non-DF cases.  In the non-DF
+    case, it can just send the packets out of the UDP Listener
+    socket. In the DF case, it needs its own socket so that it can do
+    PMTU discovery easily. To do this, it uses a Raw IP socket (IP has
+    no ports, so there's no collision issue with the UDP Listener
+    socket) and so it must add UDP headers itself.
+ 6b. If we did not initiate this connection then the UDP Listener
     should start receiving fast heartbeats from the remote peer. From
     those it should be able to identify the local connection via the
     local peer. It will tell the local connection (communicating to
     the actor thread) about the UDP address of the remote peer. The
-    local connection will then start its forwarder threads as
+    local connection will then start its forwarder thread as
     described in 6a, and start sending fast heartbeats. We send to the
     remote peer via TCP a ConnectionEstablished message. The remote
     peer receives this (on the TCP receiver process), tells the

docs/fastdp.md

@@ -0,0 +1,170 @@
+# Overlay Method Selection
+
+Weave automatically selects the best working overlay implementation
+when connecting to a new peer. It does this by initiating both
+forwarders in parallel, eventually settling on the most preferred
+established forwarder at which point the other forwarder is shut down,
+assuming it had not failed already. In the event that the remaining
+forwarder fails at a later time (e.g. because of a heartbeat timeout)
+the control plane TCP connection will recycle allowing the selection
+algorithm to run again. By default fast datapath is preferred over
+Sleeve.
+
+# Local Bridging
+
+Unlike the weave bridge netdev used by Sleeve, the OVS datapath has no
+inherent bridging capability. Consequently fastdp implements an
+ethernet bridge in addition to the vxlan overlay, maintaining its own
+forwarding database by learning MAC addresses and dispatching
+broadcast, unicast & multicast traffic accordingly.
+
+# Short Peer IDs
+
+The router relies on having access to the source and destination peer
+when deciding how to forward packets. When the Sleeve overlay is in
+use this information is conveyed directly within the encapsulation by
+including the names of the peers in question, a solution not
+accomodated directly by the vxlan wire format. Fortunately vxlan does
+have a twenty four bit segment ID field in the header we can use to
+encode this data - the challenge is to identify peers uniquely with a
+twelve bit identifier instead of the seventeen bytes used by Sleeve.
+In practice this is achieved by peers adopting a random 'short ID' and
+resolving ownership collisions of same via the existing gossip
+mechanism.
+
+# Heartbeats via vxlan
+
+Heartbeats are implemented by sending an encapsulated ethernet frame
+with source and destination MAC addresses of `00:00:00:00:00:00` and a
+payload consisting of the connection UID and total ethernet frame
+length. The vxlan vport miss handler on the receiving side detects the
+all-zero addresses and acknowledges the heartbeat via the TCP control
+channel after validating the connection UID and frame length against
+their expected values.
+
+# PMTU Discovery
+
+The length of the heartbeat ethernet frame is set to the MTU of the
+datapath interface. In the event the vxlan packet is dropped or
+truncated, the heartbeat will not be acknowledged; this lack of
+acknowledgement will cause the peers to fall back to the Sleeve
+overlay, which has a more sophisticated dynamic mechanism for coping
+with low path MTUs.
+
+To avoid triggering this fallback in typical deployments, the datapath
+interface is statically configured with an MTU of 1410 bytes allowing
+it to work with most underlay network provider MTUs, including GCE at
+1460 bytes (the fifty byte difference accomodates the vxlan overhead).
+This value can be overriden by setting `WEAVE_MTU` at launch if
+necessary.
+
+# Virtual Ports
+
+There are three kinds of virtual port associated with the weave
+datapath:
+
+* `internal` - one of these is created automatically, named after the
+  datapath. It corresponds to the network interface of the same name
+  that appears on the host when the datapath is created; it is the
+  ingress/egress port for `weave expose`
+* `netdev` - one for each application container, corresponding to the
+  host end of the veth pair
+* `vxlan` - one for each UDP port on which we are listening for vxlan
+  encapsulated packets. Typically there is only one of these, but
+  there may be more in a network in which peers do not all use the same
+  port configuration
+
+# Misses and Flow Creation
+
+As mentioned above, the datapath has no inherent behaviour - any
+ingressing packet which does not match a flow is passed to a userspace
+miss handler. The miss handler is then responsible for a) instructing
+the datapath to take actions for the packet that caused the miss (for
+example by copying it to one or more ports) and optionally b)
+installing flow rules which will allow the datapath to act on similar
+packets in future without invoking the miss handler.
+
+At a high level the fast datapath implementation can be viewed as a
+set of miss handlers that determine what actions the OVS datapath
+should take based on router topology information, together with some
+additional machinery that manages the expiry of resulting flows when
+that state changes.
+
+When a miss handler is invoked, it has two pieces of context: a byte
+array containing the packet that triggered the miss, and a set of
+'flow keys' that have been extracted from the packet by the kernel
+module. The following flow keys are of interest to weave:
+
+* `InPort` - the identifier of the ingress virtual port
+* `Ethernet` - the source and destination MAC addresses of the
+  ethernet frame
+* `Tunnel` - tunnel identifier (see section on short peer IDs above),
+  and source/dest IPv4 addresses. Only present for ingress via a vxlan
+  vport
+
+Crucially, the router must use this information alone to determine
+which actions to take - this allows the specification of a flow which
+matches these keys and instructs the kernel to take action
+automatically in future without further userspace involvement.
+
+Two actions are of interest:
+
+* `Output` - output the packet to the specified vport
+* `SetTunnel` - update the effective vxlan tunnel parameters. Only
+  required prior to output from a vxlan vport
+
+It is possible to have multiple actions in a flow, so the router can
+for example create a single rule that copies broadcast traffic to all
+'local' vports (save the ingress vport obviously) as well as to vxlan
+vports for onward peers as dictated by the routing topology.
+
+Every flow specified by the router has the following characteristics:
+
+* An `InPort` key matching the ingress vport
+* A `Tunnel` key if the ingress vport is of type vxlan
+* An `Ethernet` key matching the source and destination MAC
+* A list of `Output` (for internal and netdev vports) and
+  `SetTunnel`+`Output` (for vxlan vports) actions
+
+Under certain circumstances the miss handler will instruct the
+datapath to execute a set of actions for a packet without creating a
+corresponding flow:
+
+* When broadcasting packets to discover the path to unknown
+  destination MACs. This condition is transient in the overwhelming
+  majority of cases and so the benefit of creating a flow is
+  outweighed by the need to invalidate it once the MAC is learned
+* When the packet also needs to be forwarded to one or more peers via
+  Sleeve. In this case the router needs to handle all subsequent
+  matching packets as there is no OVS flow action to do it without
+  userspace involvement
+
+# Flow Invalidation
+
+Once a flow has been created for a particular combination of keys, the
+miss handler will never again be invoked for matching packets. It is
+therefore extremely important that we detect events which invalidate
+existing flow actions:
+
+* Addition of netdev vports to the datapath
+* Route invalidation (topology change)
+* Short peer ID collision
+
+The response in each case is the same - delete all flows from the
+datapath, allowing them to be recreated taking into account the
+updated state.
+
+In addition to these event based invalidations there is an expiry
+process that executes every five minutes. This process enumerates all
+flows in the datapath, removing any which have not been used since the
+last check; this cleans up:
+
+* Flows referring to netdev vports which have been removed
+* Flows created by forwarders which have been stopped
+* Flows related to MAC addresses which have not communicated recently
+
+Flows keyed on tunnel IPv4 address do not need to be cleared when a
+peer appears to change IP address due to NAT; this will cause a miss
+resulting in a new flow, and the old flow will expire naturally via
+the timer mechanism.
+

ipam/allocate.go

@@ -38,6 +38,11 @@ func (g *allocate) Try(alloc *Allocator) bool {
 	alloc.establishRing()
 
 	if ok, addr := alloc.space.Allocate(g.r); ok {
+		// If caller hasn't supplied a unique ID, file it under the IP address
+		// which lets the caller then release the address using DELETE /ip/address
+		if g.ident == "_" {
+			g.ident = addr.String()
+		}
 		alloc.debugln("Allocated", addr, "for", g.ident, "in", g.r)
 		alloc.addOwned(g.ident, addr)
 		g.resultChan <- allocateResult{addr, nil}

ipam/allocator.go

@@ -11,8 +11,8 @@ import (
 	"github.com/weaveworks/weave/ipam/paxos"
 	"github.com/weaveworks/weave/ipam/ring"
 	"github.com/weaveworks/weave/ipam/space"
+	"github.com/weaveworks/weave/mesh"
 	"github.com/weaveworks/weave/net/address"
-	"github.com/weaveworks/weave/router"
 )
 
 // Kinds of message we can unicast to other peers
@@ -43,37 +43,39 @@ type operation interface {
 // are used around data structures.
 type Allocator struct {
 	actionChan       chan<- func()
-	ourName          router.PeerName
+	ourName          mesh.PeerName
 	universe         address.Range                // superset of all ranges
 	ring             *ring.Ring                   // information on ranges owned by all peers
 	space            space.Space                  // more detail on ranges owned by us
 	owned            map[string][]address.Address // who owns what addresses, indexed by container-ID
-	nicknames        map[router.PeerName]string   // so we can map nicknames for rmpeer
+	nicknames        map[mesh.PeerName]string     // so we can map nicknames for rmpeer
 	pendingAllocates []operation                  // held until we get some free space
 	pendingClaims    []operation                  // held until we know who owns the space
-	gossip           router.Gossip                // our link to the outside world for sending messages
+	gossip           mesh.Gossip                  // our link to the outside world for sending messages
 	paxos            *paxos.Node
 	paxosTicker      *time.Ticker
 	shuttingDown     bool // to avoid doing any requests while trying to shut down
+	isKnownPeer      func(mesh.PeerName) bool
 	now              func() time.Time
 }
 
 // NewAllocator creates and initialises a new Allocator
-func NewAllocator(ourName router.PeerName, ourUID router.PeerUID, ourNickname string, universe address.Range, quorum uint) *Allocator {
+func NewAllocator(ourName mesh.PeerName, ourUID mesh.PeerUID, ourNickname string, universe address.Range, quorum uint, isKnownPeer func(name mesh.PeerName) bool) *Allocator {
 	return &Allocator{
-		ourName:   ourName,
-		universe:  universe,
-		ring:      ring.New(universe.Start, universe.End, ourName),
-		owned:     make(map[string][]address.Address),
-		paxos:     paxos.NewNode(ourName, ourUID, quorum),
-		nicknames: map[router.PeerName]string{ourName: ourNickname},
-		now:       time.Now,
+		ourName:     ourName,
+		universe:    universe,
+		ring:        ring.New(universe.Start, universe.End, ourName),
+		owned:       make(map[string][]address.Address),
+		paxos:       paxos.NewNode(ourName, ourUID, quorum),
+		nicknames:   map[mesh.PeerName]string{ourName: ourNickname},
+		isKnownPeer: isKnownPeer,
+		now:         time.Now,
 	}
 }
 
 // Start runs the allocator goroutine
 func (alloc *Allocator) Start() {
-	actionChan := make(chan func(), router.ChannelSize)
+	actionChan := make(chan func(), mesh.ChannelSize)
 	alloc.actionChan = actionChan
 	go alloc.actorLoop(actionChan)
 }
@@ -163,7 +165,7 @@ func (alloc *Allocator) tryPendingOps() {
 	}
 }
 
-func (alloc *Allocator) spaceRequestDenied(sender router.PeerName, r address.Range) {
+func (alloc *Allocator) spaceRequestDenied(sender mesh.PeerName, r address.Range) {
 	for i := 0; i < len(alloc.pendingClaims); {
 		claim := alloc.pendingClaims[i].(*claim)
 		if r.Contains(claim.addr) {
@@ -175,17 +177,6 @@ func (alloc *Allocator) spaceRequestDenied(sender router.PeerName, r address.Ran
 	}
 }
 
-func hasBeenCancelled(cancelChan <-chan bool) func() bool {
-	return func() bool {
-		select {
-		case <-cancelChan:
-			return true
-		default:
-			return false
-		}
-	}
-}
-
 type errorCancelled struct {
 	kind  string
 	ident string
@@ -199,10 +190,9 @@ func (e *errorCancelled) Error() string {
 
 // Allocate (Sync) - get new IP address for container with given name in range
 // if there isn't any space in that range we block indefinitely
-func (alloc *Allocator) Allocate(ident string, r address.Range, cancelChan <-chan bool) (address.Address, error) {
+func (alloc *Allocator) Allocate(ident string, r address.Range, hasBeenCancelled func() bool) (address.Address, error) {
 	resultChan := make(chan allocateResult)
-	op := &allocate{resultChan: resultChan, ident: ident, r: r,
-		hasBeenCancelled: hasBeenCancelled(cancelChan)}
+	op := &allocate{resultChan: resultChan, ident: ident, r: r, hasBeenCancelled: hasBeenCancelled}
 	alloc.doOperation(op, &alloc.pendingAllocates)
 	result := <-resultChan
 	return result.addr, result.err
@@ -237,6 +227,8 @@ func (alloc *Allocator) ContainerDied(ident string) {
 	}
 }
 
+func (alloc *Allocator) ContainerStarted(ident string) {}
+
 // Delete (Sync) - release all IP addresses for container with given name
 func (alloc *Allocator) Delete(ident string) error {
 	errChan := make(chan error)
@@ -284,6 +276,33 @@ func (alloc *Allocator) Free(ident string, addrToFree address.Address) error {
 	return <-errChan
 }
 
+func (alloc *Allocator) pickPeerFromNicknames(isValid func(mesh.PeerName) bool) mesh.PeerName {
+	for name := range alloc.nicknames {
+		if name != alloc.ourName && isValid(name) {
+			return name
+		}
+	}
+	return mesh.UnknownPeerName
+}
+
+func (alloc *Allocator) pickPeerForTransfer() mesh.PeerName {
+	// first try alive peers that actively participate in IPAM (i.e. have entries)
+	if heir := alloc.ring.PickPeerForTransfer(alloc.isKnownPeer); heir != mesh.UnknownPeerName {
+		return heir
+	}
+	// next try alive peers that have IPAM enabled but have no entries
+	if heir := alloc.pickPeerFromNicknames(alloc.isKnownPeer); heir != mesh.UnknownPeerName {
+		return heir
+	}
+	// next try disappeared peers that still have entries
+	t := func(mesh.PeerName) bool { return true }
+	if heir := alloc.ring.PickPeerForTransfer(t); heir != mesh.UnknownPeerName {
+		return heir
+	}
+	// finally, disappeared peers that that passively participated in IPAM
+	return alloc.pickPeerFromNicknames(t)
+}
+
 // Shutdown (Sync)
 func (alloc *Allocator) Shutdown() {
 	alloc.infof("Shutdown")
@@ -292,7 +311,7 @@ func (alloc *Allocator) Shutdown() {
 		alloc.shuttingDown = true
 		alloc.cancelOps(&alloc.pendingClaims)
 		alloc.cancelOps(&alloc.pendingAllocates)
-		if heir := alloc.ring.PickPeerForTransfer(); heir != router.UnknownPeerName {
+		if heir := alloc.pickPeerForTransfer(); heir != mesh.UnknownPeerName {
 			alloc.ring.Transfer(alloc.ourName, heir)
 			alloc.space.Clear()
 			alloc.gossip.GossipBroadcast(alloc.Gossip())
@@ -320,7 +339,6 @@ func (alloc *Allocator) AdminTakeoverRanges(peerNameOrNickname string) error {
 			return
 		}
 
-		delete(alloc.nicknames, peername)
 		newRanges, err := alloc.ring.Transfer(peername, alloc.ourName)
 		alloc.space.AddRanges(newRanges)
 		resultChan <- err
@@ -332,33 +350,45 @@ func (alloc *Allocator) AdminTakeoverRanges(peerNameOrNickname string) error {
 // call into the router for this because we are interested in peers
 // that have gone away but are still in the ring, which is why we
 // maintain our own nicknames map.
-func (alloc *Allocator) lookupPeername(name string) (router.PeerName, error) {
+func (alloc *Allocator) lookupPeername(name string) (mesh.PeerName, error) {
 	for peername, nickname := range alloc.nicknames {
 		if nickname == name {
 			return peername, nil
 		}
 	}
 
-	return router.PeerNameFromString(name)
+	return mesh.PeerNameFromString(name)
 }
 
-// Restrict the peers in "nicknames" to those in the ring and our own
+// Restrict the peers in "nicknames" to those in the ring plus peers known to the router
 func (alloc *Allocator) pruneNicknames() {
 	ringPeers := alloc.ring.PeerNames()
 	for name := range alloc.nicknames {
-		if _, ok := ringPeers[name]; !ok && name != alloc.ourName {
+		if _, ok := ringPeers[name]; !ok && !alloc.isKnownPeer(name) {
 			delete(alloc.nicknames, name)
 		}
 	}
 }
 
+func (alloc *Allocator) annotatePeernames(names []mesh.PeerName) []string {
+	var res []string
+	for _, name := range names {
+		if nickname, found := alloc.nicknames[name]; found {
+			res = append(res, fmt.Sprint(name, "(", nickname, ")"))
+		} else {
+			res = append(res, name.String())
+		}
+	}
+	return res
+}
+
 func decodeRange(msg []byte) (r address.Range, err error) {
 	decoder := gob.NewDecoder(bytes.NewReader(msg))
 	return r, decoder.Decode(&r)
 }
 
 // OnGossipUnicast (Sync)
-func (alloc *Allocator) OnGossipUnicast(sender router.PeerName, msg []byte) error {
+func (alloc *Allocator) OnGossipUnicast(sender mesh.PeerName, msg []byte) error {
 	alloc.debugln("OnGossipUnicast from", sender, ": ", len(msg), "bytes")
 	resultChan := make(chan error)
 	alloc.actionChan <- func() {
@@ -384,7 +414,7 @@ func (alloc *Allocator) OnGossipUnicast(sender router.PeerName, msg []byte) erro
 }
 
 // OnGossipBroadcast (Sync)
-func (alloc *Allocator) OnGossipBroadcast(sender router.PeerName, msg []byte) (router.GossipData, error) {
+func (alloc *Allocator) OnGossipBroadcast(sender mesh.PeerName, msg []byte) (mesh.GossipData, error) {
 	alloc.debugln("OnGossipBroadcast from", sender, ":", len(msg), "bytes")
 	resultChan := make(chan error)
 	alloc.actionChan <- func() {
@@ -397,7 +427,7 @@ type gossipState struct {
 	// We send a timstamp along with the information to be
 	// gossipped in order to detect skewed clocks
 	Now       int64
-	Nicknames map[router.PeerName]string
+	Nicknames map[mesh.PeerName]string
 
 	Paxos paxos.GossipState
 	Ring  *ring.Ring
@@ -433,11 +463,11 @@ func (alloc *Allocator) Encode() []byte {
 }
 
 // OnGossip (Sync)
-func (alloc *Allocator) OnGossip(msg []byte) (router.GossipData, error) {
+func (alloc *Allocator) OnGossip(msg []byte) (mesh.GossipData, error) {
 	alloc.debugln("Allocator.OnGossip:", len(msg), "bytes")
 	resultChan := make(chan error)
 	alloc.actionChan <- func() {
-		resultChan <- alloc.update(router.UnknownPeerName, msg)
+		resultChan <- alloc.update(mesh.UnknownPeerName, msg)
 	}
 	return nil, <-resultChan // for now, we never propagate updates. TBD
 }
@@ -448,8 +478,8 @@ type ipamGossipData struct {
 	alloc *Allocator
 }
 
-func (d *ipamGossipData) Merge(other router.GossipData) {
-	// no-op
+func (d *ipamGossipData) Merge(other mesh.GossipData) mesh.GossipData {
+	return d // no-op
 }
 
 func (d *ipamGossipData) Encode() [][]byte {
@@ -458,12 +488,12 @@ func (d *ipamGossipData) Encode() [][]byte {
 
 // Gossip returns a GossipData implementation, which in this case always
 // returns the latest ring state (and does nothing on merge)
-func (alloc *Allocator) Gossip() router.GossipData {
+func (alloc *Allocator) Gossip() mesh.GossipData {
 	return &ipamGossipData{alloc}
 }
 
 // SetInterfaces gives the allocator two interfaces for talking to the outside world
-func (alloc *Allocator) SetInterfaces(gossip router.Gossip) {
+func (alloc *Allocator) SetInterfaces(gossip mesh.Gossip) {
 	alloc.gossip = gossip
 }
 
@@ -510,7 +540,7 @@ func (alloc *Allocator) establishRing() {
 	}
 }
 
-func (alloc *Allocator) createRing(peers []router.PeerName) {
+func (alloc *Allocator) createRing(peers []mesh.PeerName) {
 	alloc.debugln("Paxos consensus:", peers)
 	alloc.ring.ClaimForPeers(normalizeConsensus(peers))
 	alloc.gossip.GossipBroadcast(alloc.Gossip())
@@ -533,7 +563,7 @@ func (alloc *Allocator) ringUpdated() {
 }
 
 // For compatibility with sort.Interface
-type peerNames []router.PeerName
+type peerNames []mesh.PeerName
 
 func (a peerNames) Len() int           { return len(a) }
 func (a peerNames) Less(i, j int) bool { return a[i] < a[j] }
@@ -542,7 +572,7 @@ func (a peerNames) Swap(i, j int)      { a[i], a[j] = a[j], a[i] }
 // When we get a consensus from Paxos, the peer names are not in a
 // defined order and may contain duplicates.  This function sorts them
 // and de-dupes.
-func normalizeConsensus(consensus []router.PeerName) []router.PeerName {
+func normalizeConsensus(consensus []mesh.PeerName) []mesh.PeerName {
 	if len(consensus) == 0 {
 		return nil
 	}
@@ -577,22 +607,22 @@ func encodeRange(r address.Range) []byte {
 	return buf.Bytes()
 }
 
-func (alloc *Allocator) sendSpaceRequest(dest router.PeerName, r address.Range) error {
-	msg := router.Concat([]byte{msgSpaceRequest}, encodeRange(r))
+func (alloc *Allocator) sendSpaceRequest(dest mesh.PeerName, r address.Range) error {
+	msg := append([]byte{msgSpaceRequest}, encodeRange(r)...)
 	return alloc.gossip.GossipUnicast(dest, msg)
 }
 
-func (alloc *Allocator) sendSpaceRequestDenied(dest router.PeerName, r address.Range) error {
-	msg := router.Concat([]byte{msgSpaceRequestDenied}, encodeRange(r))
+func (alloc *Allocator) sendSpaceRequestDenied(dest mesh.PeerName, r address.Range) error {
+	msg := append([]byte{msgSpaceRequestDenied}, encodeRange(r)...)
 	return alloc.gossip.GossipUnicast(dest, msg)
 }
 
-func (alloc *Allocator) sendRingUpdate(dest router.PeerName) {
-	msg := router.Concat([]byte{msgRingUpdate}, alloc.encode())
+func (alloc *Allocator) sendRingUpdate(dest mesh.PeerName) {
+	msg := append([]byte{msgRingUpdate}, alloc.encode()...)
 	alloc.gossip.GossipUnicast(dest, msg)
 }
 
-func (alloc *Allocator) update(sender router.PeerName, msg []byte) error {
+func (alloc *Allocator) update(sender mesh.PeerName, msg []byte) error {
 	reader := bytes.NewReader(msg)
 	decoder := gob.NewDecoder(reader)
 	var data gossipState
@@ -616,16 +646,20 @@ func (alloc *Allocator) update(sender router.PeerName, msg []byte) error {
 	// shouldn't get updates for a empty Ring. But tolerate
 	// them just in case.
 	if data.Ring != nil {
-		if data.Ring.Range() != alloc.universe {
-			return fmt.Errorf("Incompatible IP allocation range %s; ours is %s",
-				data.Ring.Range().AsCIDRString(), alloc.universe.AsCIDRString())
+		switch err = alloc.ring.Merge(*data.Ring); err {
+		case ring.ErrDifferentSeeds:
+			return fmt.Errorf("IP allocation was seeded by different peers (received: %v, ours: %v)",
+				alloc.annotatePeernames(data.Ring.Seeds), alloc.annotatePeernames(alloc.ring.Seeds))
+		case ring.ErrDifferentRange:
+			return fmt.Errorf("Incompatible IP allocation ranges (received: %s, ours: %s)",
+				data.Ring.Range().AsCIDRString(), alloc.ring.Range().AsCIDRString())
+		default:
+			if err == nil && !alloc.ring.Empty() {
+				alloc.pruneNicknames()
+				alloc.ringUpdated()
+			}
+			return err
 		}
-		err = alloc.ring.Merge(*data.Ring)
-		if !alloc.ring.Empty() {
-			alloc.pruneNicknames()
-			alloc.ringUpdated()
-		}
-		return err
 	}
 
 	if data.Paxos != nil {
@@ -640,7 +674,7 @@ func (alloc *Allocator) update(sender router.PeerName, msg []byte) error {
 					alloc.createRing(cons.Value)
 				}
 			}
-		} else if sender != router.UnknownPeerName {
+		} else if sender != mesh.UnknownPeerName {
 			// Sender is trying to initialize a ring, but we have one
 			// already - send it straight back
 			alloc.sendRingUpdate(sender)
@@ -650,7 +684,7 @@ func (alloc *Allocator) update(sender router.PeerName, msg []byte) error {
 	return nil
 }
 
-func (alloc *Allocator) donateSpace(r address.Range, to router.PeerName) {
+func (alloc *Allocator) donateSpace(r address.Range, to mesh.PeerName) {
 	// No matter what we do, we'll send a unicast gossip
 	// of our ring back to tha chap who asked for space.
 	// This serves to both tell him of any space we might

ipam/allocator_test.go

@@ -20,6 +20,8 @@ const (
 	testStart3 = "10.0.3.0"
 )
 
+func returnFalse() bool { return false }
+
 func TestAllocFree(t *testing.T) {
 	const (
 		container1 = "abcdef"
@@ -39,31 +41,31 @@ func TestAllocFree(t *testing.T) {
 	_, cidr2, _ := address.ParseCIDR(subnet2)
 
 	alloc.claimRingForTesting()
-	addr1, err := alloc.Allocate(container1, cidr1.HostRange(), nil)
+	addr1, err := alloc.Allocate(container1, cidr1.HostRange(), returnFalse)
 	require.NoError(t, err)
 	require.Equal(t, testAddr1, addr1.String(), "address")
 
-	addr2, err := alloc.Allocate(container1, cidr2.HostRange(), nil)
+	addr2, err := alloc.Allocate(container1, cidr2.HostRange(), returnFalse)
 	require.NoError(t, err)
 	require.Equal(t, testAddr2, addr2.String(), "address")
 
 	// Ask for another address for a different container and check it's different
-	addr1b, _ := alloc.Allocate(container2, cidr1.HostRange(), nil)
+	addr1b, _ := alloc.Allocate(container2, cidr1.HostRange(), returnFalse)
 	if addr1b.String() == testAddr1 {
 		t.Fatalf("Expected different address but got %s", addr1b.String())
 	}
 
 	// Ask for the first container again and we should get the same addresses again
-	addr1a, _ := alloc.Allocate(container1, cidr1.HostRange(), nil)
+	addr1a, _ := alloc.Allocate(container1, cidr1.HostRange(), returnFalse)
 	require.Equal(t, testAddr1, addr1a.String(), "address")
-	addr2a, _ := alloc.Allocate(container1, cidr2.HostRange(), nil)
+	addr2a, _ := alloc.Allocate(container1, cidr2.HostRange(), returnFalse)
 	require.Equal(t, testAddr2, addr2a.String(), "address")
 
 	// Now delete the first container, and we should get its addresses back
 	require.NoError(t, alloc.Delete(container1))
-	addr3, _ := alloc.Allocate(container3, cidr1.HostRange(), nil)
+	addr3, _ := alloc.Allocate(container3, cidr1.HostRange(), returnFalse)
 	require.Equal(t, testAddr1, addr3.String(), "address")
-	addr4, _ := alloc.Allocate(container3, cidr2.HostRange(), nil)
+	addr4, _ := alloc.Allocate(container3, cidr2.HostRange(), returnFalse)
 	require.Equal(t, testAddr2, addr4.String(), "address")
 
 	alloc.ContainerDied(container2)
@@ -93,7 +95,7 @@ func TestBootstrap(t *testing.T) {
 	ExpectBroadcastMessage(alloc1, nil) // alloc1 will try to form consensus
 	done := make(chan bool)
 	go func() {
-		alloc1.Allocate("somecontainer", subnet, nil)
+		alloc1.Allocate("somecontainer", subnet, returnFalse)
 		done <- true
 	}()
 	time.Sleep(100 * time.Millisecond)
@@ -147,15 +149,15 @@ func TestAllocatorClaim(t *testing.T) {
 	require.NoError(t, err)
 
 	// Do one allocate to ensure paxos is all done
-	alloc.Allocate("unused", subnet, nil)
+	alloc.Allocate("unused", subnet, returnFalse)
 	// Do an allocate on the other peer, which we will try to claim later
-	addrx, err := allocs[0].Allocate(container1, subnet, nil)
+	addrx, err := allocs[0].Allocate(container1, subnet, returnFalse)
 
 	// Now try the claim again
 	err = alloc.Claim(container3, addr1, true)
 	require.NoError(t, err)
 	// Check we get this address back if we try an allocate
-	addr3, _ := alloc.Allocate(container3, subnet, nil)
+	addr3, _ := alloc.Allocate(container3, subnet, returnFalse)
 	require.Equal(t, testAddr1, addr3.String(), "address")
 	// one more claim should still work
 	err = alloc.Claim(container3, addr1, true)
@@ -206,9 +208,9 @@ func TestCancel(t *testing.T) {
 	alloc1.OnGossipBroadcast(alloc2.ourName, alloc2.Encode())
 
 	// Get some IPs, so each allocator has some space
-	res1, _ := alloc1.Allocate("foo", subnet, nil)
+	res1, _ := alloc1.Allocate("foo", subnet, returnFalse)
 	common.Log.Debugf("res1 = %s", res1.String())
-	res2, _ := alloc2.Allocate("bar", subnet, nil)
+	res2, _ := alloc2.Allocate("bar", subnet, returnFalse)
 	common.Log.Debugf("res2 = %s", res2.String())
 	if res1 == res2 {
 		require.FailNow(t, "Error: got same ips!")
@@ -220,12 +222,20 @@ func TestCancel(t *testing.T) {
 
 	// Use up all the IPs that alloc1 owns, so the allocation after this will prompt a request to alloc2
 	for i := 0; alloc1.NumFreeAddresses(subnet) > 0; i++ {
-		alloc1.Allocate(fmt.Sprintf("tmp%d", i), subnet, nil)
+		alloc1.Allocate(fmt.Sprintf("tmp%d", i), subnet, returnFalse)
 	}
 	cancelChan := make(chan bool, 1)
 	doneChan := make(chan bool)
 	go func() {
-		_, ok := alloc1.Allocate("baz", subnet, cancelChan)
+		_, ok := alloc1.Allocate("baz", subnet,
+			func() bool {
+				select {
+				case <-cancelChan:
+					return true
+				default:
+					return false
+				}
+			})
 		doneChan <- ok == nil
 	}()
 
@@ -252,7 +262,7 @@ func TestCancelOnDied(t *testing.T) {
 
 	doneChan := make(chan bool)
 	f := func() {
-		_, ok := alloc1.Allocate(container1, subnet, nil)
+		_, ok := alloc1.Allocate(container1, subnet, returnFalse)
 		doneChan <- ok == nil
 	}
 
@@ -283,11 +293,11 @@ func TestGossipShutdown(t *testing.T) {
 	defer alloc.Stop()
 
 	alloc.claimRingForTesting()
-	alloc.Allocate(container1, subnet, nil)
+	alloc.Allocate(container1, subnet, returnFalse)
 
 	alloc.Shutdown()
 
-	_, err := alloc.Allocate(container2, subnet, nil) // trying to allocate after shutdown should fail
+	_, err := alloc.Allocate(container2, subnet, returnFalse) // trying to allocate after shutdown should fail
 	require.False(t, err == nil, "no address")
 
 	CheckAllExpectedMessagesSent(alloc)
@@ -302,10 +312,10 @@ func TestTransfer(t *testing.T) {
 	alloc2 := allocs[1]
 	alloc3 := allocs[2] // This will be 'master' and get the first range
 
-	_, err := alloc2.Allocate("foo", subnet, nil)
+	_, err := alloc2.Allocate("foo", subnet, returnFalse)
 	require.True(t, err == nil, "Failed to get address")
 
-	_, err = alloc3.Allocate("bar", subnet, nil)
+	_, err = alloc3.Allocate("bar", subnet, returnFalse)
 	require.True(t, err == nil, "Failed to get address")
 
 	alloc2.gossip.GossipBroadcast(alloc2.Gossip())
@@ -323,7 +333,7 @@ func TestTransfer(t *testing.T) {
 
 	require.Equal(t, address.Offset(1022), alloc1.NumFreeAddresses(subnet))
 
-	_, err = alloc1.Allocate("foo", subnet, nil)
+	_, err = alloc1.Allocate("foo", subnet, returnFalse)
 	require.True(t, err == nil, "Failed to get address")
 	alloc1.Stop()
 }
@@ -338,7 +348,7 @@ func TestFakeRouterSimple(t *testing.T) {
 	alloc1 := allocs[0]
 	//alloc2 := allocs[1]
 
-	_, err := alloc1.Allocate("foo", subnet, nil)
+	_, err := alloc1.Allocate("foo", subnet, returnFalse)
 	require.NoError(t, err, "Failed to get address")
 }
 
@@ -419,7 +429,7 @@ func TestAllocatorFuzz(t *testing.T) {
 		allocIndex := rand.Int31n(nodes)
 		alloc := allocs[allocIndex]
 		//common.Log.Infof("Allocate: asking allocator %d", allocIndex)
-		addr, err := alloc.Allocate(name, subnet, nil)
+		addr, err := alloc.Allocate(name, subnet, returnFalse)
 
 		if err != nil {
 			panic(fmt.Sprintf("Could not allocate addr"))
@@ -476,7 +486,7 @@ func TestAllocatorFuzz(t *testing.T) {
 
 		//common.Log.Infof("Asking for %s (%s) on allocator %d again", res.name, addr, res.alloc)
 
-		newAddr, _ := alloc.Allocate(res.name, subnet, nil)
+		newAddr, _ := alloc.Allocate(res.name, subnet, returnFalse)
 		oldAddr, _ := address.ParseIP(addr)
 		if newAddr != oldAddr {
 			panic(fmt.Sprintf("Got different address for repeat request for %s: %s != %s", res.name, newAddr, oldAddr))

ipam/claim.go

@@ -4,8 +4,8 @@ import (
 	"fmt"
 
 	"github.com/weaveworks/weave/common"
+	"github.com/weaveworks/weave/mesh"
 	"github.com/weaveworks/weave/net/address"
-	"github.com/weaveworks/weave/router"
 )
 
 type claim struct {
@@ -43,7 +43,7 @@ func (c *claim) Try(alloc *Allocator) bool {
 	switch owner := alloc.ring.Owner(c.addr); owner {
 	case alloc.ourName:
 		// success
-	case router.UnknownPeerName:
+	case mesh.UnknownPeerName:
 		// If our ring doesn't know, it must be empty.
 		if c.noErrorOnUnknown {
 			alloc.infof("Claim %s for %s: address allocator still initializing; will try later.", c.addr, c.ident)
@@ -86,7 +86,7 @@ func (c *claim) Try(alloc *Allocator) bool {
 	return true
 }
 
-func (c *claim) deniedBy(alloc *Allocator, owner router.PeerName) {
+func (c *claim) deniedBy(alloc *Allocator, owner mesh.PeerName) {
 	name, found := alloc.nicknames[owner]
 	if found {
 		name = " (" + name + ")"

ipam/http.go

@@ -31,7 +31,17 @@ func parseCIDR(w http.ResponseWriter, cidrStr string) (address.CIDR, bool) {
 
 func (alloc *Allocator) handleHTTPAllocate(dockerCli *docker.Client, w http.ResponseWriter, ident string, checkAlive bool, subnet address.CIDR) {
 	closedChan := w.(http.CloseNotifier).CloseNotify()
-	addr, err := alloc.Allocate(ident, subnet.HostRange(), closedChan)
+	addr, err := alloc.Allocate(ident, subnet.HostRange(),
+		func() bool {
+			select {
+			case <-closedChan:
+				return true
+			default:
+				res := checkAlive && dockerCli != nil && dockerCli.IsContainerNotRunning(ident)
+				checkAlive = false // we check only once; if the container dies later we learn about that through events
+				return res
+			}
+		})
 	if err != nil {
 		if _, ok := err.(*errorCancelled); ok { // cancellation is not really an error
 			common.Log.Infoln("[allocator]:", err.Error())
@@ -41,18 +51,16 @@ func (alloc *Allocator) handleHTTPAllocate(dockerCli *docker.Client, w http.Resp
 		badRequest(w, err)
 		return
 	}
-	if checkAlive && dockerCli != nil && dockerCli.IsContainerNotRunning(ident) {
-		common.Log.Infof("[allocator] '%s' is not running: freeing %s", ident, addr)
-		alloc.Free(ident, addr)
-		fmt.Fprint(w, "cancelled")
-		return
-	}
 
 	fmt.Fprintf(w, "%s/%d", addr, subnet.PrefixLen)
 }
 
 // HandleHTTP wires up ipams HTTP endpoints to the provided mux.
 func (alloc *Allocator) HandleHTTP(router *mux.Router, defaultSubnet address.CIDR, dockerCli *docker.Client) {
+	router.Methods("GET").Path("/ipinfo/defaultsubnet").HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		fmt.Fprintf(w, "%s", defaultSubnet)
+	})
+
 	router.Methods("PUT").Path("/ip/{id}/{ip}").HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		vars := mux.Vars(r)
 		ident := vars["id"]

ipam/http_test.go

@@ -14,7 +14,6 @@ import (
 	"github.com/stretchr/testify/require"
 	"github.com/weaveworks/weave/common"
 	"github.com/weaveworks/weave/net/address"
-	wt "github.com/weaveworks/weave/testing"
 )
 
 func HTTPPost(t *testing.T, url string) string {
@@ -145,12 +144,6 @@ func TestBadHttp(t *testing.T) {
 }
 
 func TestHTTPCancel(t *testing.T) {
-	wt.RunWithTimeout(t, 2*time.Second, func() {
-		impTestHTTPCancel(t)
-	})
-}
-
-func impTestHTTPCancel(t *testing.T) {
 	var (
 		containerID = "deadbeef"
 		testCIDR1   = "10.0.3.0/29"

ipam/paxos/paxos.go

@@ -1,7 +1,7 @@
 package paxos
 
 import (
-	"github.com/weaveworks/weave/router"
+	"github.com/weaveworks/weave/mesh"
 )
 
 // The node identifier.  The use of the UID here is important: Paxos
@@ -9,8 +9,8 @@ import (
 // node does not restart and lose its Paxos state but claim to have
 // the same ID.
 type NodeID struct {
-	Name router.PeerName
-	UID  router.PeerUID
+	Name mesh.PeerName
+	UID  mesh.PeerUID
 }
 
 // note all fields exported in structs so we can Gob them
@@ -39,7 +39,7 @@ func (a ProposalID) valid() bool {
 }
 
 // For seeding IPAM, the value we want consensus on is a set of peer names
-type Value []router.PeerName
+type Value []mesh.PeerName
 
 // An AcceptedValue is a Value plus the proposal which originated that
 // Value.  The origin is not essential, but makes comparing
@@ -72,7 +72,7 @@ type Node struct {
 	knows  GossipState
 }
 
-func NewNode(name router.PeerName, uid router.PeerUID, quorum uint) *Node {
+func NewNode(name mesh.PeerName, uid mesh.PeerUID, quorum uint) *Node {
 	return &Node{
 		id:     NodeID{name, uid},
 		quorum: quorum,
@@ -223,7 +223,7 @@ func (node *Node) Think() bool {
 // about.  This is not necessarily all peer names, but it is at least
 // a quorum, and so good enough for seeding the ring.
 func (node *Node) pickValue() Value {
-	val := make([]router.PeerName, len(node.knows))
+	val := make([]mesh.PeerName, len(node.knows))
 	i := 0
 	for id := range node.knows {
 		val[i] = id.Name

ipam/paxos/paxos_test.go

@@ -6,7 +6,7 @@ import (
 	"testing"
 	"time"
 
-	"github.com/weaveworks/weave/router"
+	"github.com/weaveworks/weave/mesh"
 )
 
 type TestNode struct {
@@ -97,8 +97,8 @@ func makeRandomModel(params *TestParams, r *rand.Rand, t *testing.T) *Model {
 	}
 
 	for i := range m.nodes {
-		m.nodes[i].Node = NewNode(router.PeerName(i/2+1),
-			router.PeerUID(r.Int63()), m.quorum)
+		m.nodes[i].Node = NewNode(mesh.PeerName(i/2+1),
+			mesh.PeerUID(r.Int63()), m.quorum)
 		m.nodes[i].Propose()
 	}
 
@@ -172,8 +172,8 @@ func (m *Model) isolateNode(node *TestNode) {
 
 // Restart a node
 func (m *Model) restart(node *TestNode) {
-	node.Node = NewNode(router.PeerName(m.nextID),
-		router.PeerUID(m.r.Int63()), m.quorum)
+	node.Node = NewNode(mesh.PeerName(m.nextID),
+		mesh.PeerUID(m.r.Int63()), m.quorum)
 	m.nextID++
 	node.Propose()
 

ipam/ring/entry.go

@@ -4,14 +4,14 @@ import (
 	"sort"
 
 	"github.com/weaveworks/weave/common"
+	"github.com/weaveworks/weave/mesh"
 	"github.com/weaveworks/weave/net/address"
-	"github.com/weaveworks/weave/router"
 )
 
 // Entry represents entries around the ring
 type entry struct {
 	Token   address.Address // The start of this range
-	Peer    router.PeerName // Who owns this range
+	Peer    mesh.PeerName   // Who owns this range
 	Version uint32          // Version of this range
 	Free    address.Offset  // Number of free IPs in this range
 }
@@ -21,7 +21,7 @@ func (e *entry) Equal(e2 *entry) bool {
 		e.Version == e2.Version
 }
 
-func (e *entry) update(peername router.PeerName, free address.Offset) {
+func (e *entry) update(peername mesh.PeerName, free address.Offset) {
 	e.Peer = peername
 	e.Version++
 	e.Free = free

ipam/ring/ring.go

@@ -13,15 +13,16 @@ import (
 	"time"
 
 	"github.com/weaveworks/weave/common"
+	"github.com/weaveworks/weave/mesh"
 	"github.com/weaveworks/weave/net/address"
-	"github.com/weaveworks/weave/router"
 )
 
 // Ring represents the ring itself
 type Ring struct {
 	Start, End address.Address // [min, max) tokens in this ring.  Due to wrapping, min == max (effectively)
-	Peer       router.PeerName // name of peer owning this ring instance
+	Peer       mesh.PeerName   // name of peer owning this ring instance
 	Entries    entries         // list of entries sorted by token
+	Seeds      []mesh.PeerName // peers with which the ring was seeded
 }
 
 func (r *Ring) assertInvariants() {
@@ -36,6 +37,7 @@ var (
 	ErrNotSorted       = errors.New("Ring not sorted")
 	ErrTokenRepeated   = errors.New("Token appears twice in ring")
 	ErrTokenOutOfRange = errors.New("Token is out of range")
+	ErrDifferentSeeds  = errors.New("Received ring was seeded differently from ours!")
 	ErrDifferentRange  = errors.New("Received range differs from ours!")
 	ErrNewerVersion    = errors.New("Received new version for entry I own!")
 	ErrInvalidEntry    = errors.New("Received invalid state update!")
@@ -82,7 +84,7 @@ func (r *Ring) checkInvariants() error {
 }
 
 // New creates an empty ring belonging to peer.
-func New(start, end address.Address, peer router.PeerName) *Ring {
+func New(start, end address.Address, peer mesh.PeerName) *Ring {
 	common.Assert(start < end)
 
 	ring := &Ring{Start: start, End: end, Peer: peer, Entries: make([]*entry, 0)}
@@ -109,7 +111,7 @@ func (r *Ring) distance(start, end address.Address) address.Offset {
 // Preconditions:
 // - start < end
 // - [start, end) must be owned by the calling peer
-func (r *Ring) GrantRangeToHost(start, end address.Address, peer router.PeerName) {
+func (r *Ring) GrantRangeToHost(start, end address.Address, peer mesh.PeerName) {
 	//fmt.Printf("%s GrantRangeToHost [%v,%v) -> %s\n", r.Peer, start, end, peer)
 
 	r.assertInvariants()
@@ -188,6 +190,17 @@ func (r *Ring) Merge(gossip Ring) error {
 		return err
 	}
 
+	if len(gossip.Seeds) > 0 && len(r.Seeds) > 0 {
+		if len(gossip.Seeds) != len(r.Seeds) {
+			return ErrDifferentSeeds
+		}
+		for i, seed := range gossip.Seeds {
+			if seed != r.Seeds[i] {
+				return ErrDifferentSeeds
+			}
+		}
+	}
+
 	if r.Start != gossip.Start || r.End != gossip.End {
 		return ErrDifferentRange
 	}
@@ -197,7 +210,7 @@ func (r *Ring) Merge(gossip Ring) error {
 	addToResult := func(e entry) { result = append(result, &e) }
 
 	var mine, theirs *entry
-	var previousOwner *router.PeerName
+	var previousOwner *mesh.PeerName
 	// i is index into r.Entries; j is index into gossip.Entries
 	var i, j int
 	for i < len(r.Entries) && j < len(gossip.Entries) {
@@ -254,6 +267,9 @@ func (r *Ring) Merge(gossip Ring) error {
 		previousOwner = nil
 	}
 
+	if len(r.Seeds) == 0 {
+		r.Seeds = gossip.Seeds
+	}
 	r.Entries = result
 	return nil
 }
@@ -300,9 +316,28 @@ func (r *Ring) OwnedRanges() (result []address.Range) {
 	return r.splitRangesOverZero(result)
 }
 
+// For printing status
+type RangeInfo struct {
+	Peer mesh.PeerName
+	address.Range
+	Version uint32
+}
+
+func (r *Ring) AllRangeInfo() (result []RangeInfo) {
+	for i, entry := range r.Entries {
+		nextEntry := r.Entries.entry(i + 1)
+		ranges := []address.Range{{Start: entry.Token, End: nextEntry.Token}}
+		ranges = r.splitRangesOverZero(ranges)
+		for _, r := range ranges {
+			result = append(result, RangeInfo{entry.Peer, r, entry.Version})
+		}
+	}
+	return
+}
+
 // ClaimForPeers claims the entire ring for the array of peers passed
 // in.  Only works for empty rings.
-func (r *Ring) ClaimForPeers(peers []router.PeerName) {
+func (r *Ring) ClaimForPeers(peers []mesh.PeerName) {
 	common.Assert(r.Empty())
 	defer r.assertInvariants()
 	defer r.updateExportedVariables()
@@ -330,9 +365,11 @@ func (r *Ring) ClaimForPeers(peers []router.PeerName) {
 	}
 
 	common.Assert(pos == r.End)
+
+	r.Seeds = peers
 }
 
-func (r *Ring) FprintWithNicknames(w io.Writer, m map[router.PeerName]string) {
+func (r *Ring) FprintWithNicknames(w io.Writer, m map[mesh.PeerName]string) {
 	for _, entry := range r.Entries {
 		nickname, found := m[entry.Peer]
 		if found {
@@ -347,7 +384,7 @@ func (r *Ring) FprintWithNicknames(w io.Writer, m map[router.PeerName]string) {
 func (r *Ring) String() string {
 	var buffer bytes.Buffer
 	fmt.Fprintf(&buffer, "Ring [%s, %s)", r.Start, r.End)
-	r.FprintWithNicknames(&buffer, make(map[router.PeerName]string))
+	r.FprintWithNicknames(&buffer, make(map[mesh.PeerName]string))
 	return buffer.String()
 }
 
@@ -397,7 +434,7 @@ func (r *Ring) ReportFree(freespace map[address.Address]address.Offset) {
 
 type weightedPeer struct {
 	weight   float64
-	peername router.PeerName
+	peername mesh.PeerName
 }
 type weightedPeers []weightedPeer
 
@@ -408,10 +445,10 @@ func (ws weightedPeers) Swap(i, j int)      { ws[i], ws[j] = ws[j], ws[i] }
 
 // ChoosePeersToAskForSpace returns all peers we can ask for space in
 // the range [start, end), in weighted-random order.  Assumes start<end.
-func (r *Ring) ChoosePeersToAskForSpace(start, end address.Address) []router.PeerName {
+func (r *Ring) ChoosePeersToAskForSpace(start, end address.Address) []mesh.PeerName {
 	var (
 		sum               address.Offset
-		totalSpacePerPeer = make(map[router.PeerName]address.Offset) // Compute total free space per peer
+		totalSpacePerPeer = make(map[mesh.PeerName]address.Offset) // Compute total free space per peer
 	)
 
 	// iterate through tokens
@@ -445,25 +482,25 @@ func (r *Ring) ChoosePeersToAskForSpace(start, end address.Address) []router.Pee
 		ws = append(ws, weightedPeer{weight: float64(space) * rand.Float64(), peername: peername})
 	}
 	sort.Sort(ws)
-	result := make([]router.PeerName, len(ws))
+	result := make([]mesh.PeerName, len(ws))
 	for i, wp := range ws {
 		result[i] = wp.peername
 	}
 	return result
 }
 
-func (r *Ring) PickPeerForTransfer() router.PeerName {
+func (r *Ring) PickPeerForTransfer(isValid func(mesh.PeerName) bool) mesh.PeerName {
 	for _, entry := range r.Entries {
-		if entry.Peer != r.Peer {
+		if entry.Peer != r.Peer && isValid(entry.Peer) {
 			return entry.Peer
 		}
 	}
-	return router.UnknownPeerName
+	return mesh.UnknownPeerName
 }
 
 // Transfer will mark all entries associated with 'from' peer as owned by 'to' peer
 // and return ranges indicating the new space we picked up
-func (r *Ring) Transfer(from, to router.PeerName) ([]address.Range, error) {
+func (r *Ring) Transfer(from, to mesh.PeerName) ([]address.Range, error) {
 	r.assertInvariants()
 	defer r.assertInvariants()
 	defer r.updateExportedVariables()
@@ -493,13 +530,13 @@ func (r *Ring) Contains(addr address.Address) bool {
 }
 
 // Owner returns the peername which owns the range containing addr
-func (r *Ring) Owner(token address.Address) router.PeerName {
+func (r *Ring) Owner(token address.Address) mesh.PeerName {
 	common.Assert(r.Start <= token && token < r.End)
 
 	r.assertInvariants()
 	// There can be no owners on an empty ring
 	if r.Empty() {
-		return router.UnknownPeerName
+		return mesh.UnknownPeerName
 	}
 
 	// Look for the right-most entry, less than or equal to token
@@ -512,8 +549,8 @@ func (r *Ring) Owner(token address.Address) router.PeerName {
 }
 
 // Get the set of PeerNames mentioned in the ring
-func (r *Ring) PeerNames() map[router.PeerName]struct{} {
-	res := make(map[router.PeerName]struct{})
+func (r *Ring) PeerNames() map[mesh.PeerName]struct{} {
+	res := make(map[mesh.PeerName]struct{})
 
 	for _, entry := range r.Entries {
 		res[entry.Peer] = struct{}{}

ipam/ring/ring_test.go

@@ -9,14 +9,14 @@ import (
 
 	"github.com/stretchr/testify/require"
 	"github.com/weaveworks/weave/common"
+	"github.com/weaveworks/weave/mesh"
 	"github.com/weaveworks/weave/net/address"
-	"github.com/weaveworks/weave/router"
 )
 
 var (
-	peer1name, _ = router.PeerNameFromString("01:00:00:00:02:00")
-	peer2name, _ = router.PeerNameFromString("02:00:00:00:02:00")
-	peer3name, _ = router.PeerNameFromString("03:00:00:00:02:00")
+	peer1name, _ = mesh.PeerNameFromString("01:00:00:00:02:00")
+	peer2name, _ = mesh.PeerNameFromString("02:00:00:00:02:00")
+	peer3name, _ = mesh.PeerNameFromString("03:00:00:00:02:00")
 
 	start, end    = ParseIP("10.0.0.0"), ParseIP("10.0.0.255")
 	dot10, dot245 = ParseIP("10.0.0.10"), ParseIP("10.0.0.245")
@@ -351,7 +351,7 @@ func TestGossip(t *testing.T) {
 	assertRing(ring2, []*entry{{Token: start, Peer: peer1name, Free: 255}})
 }
 
-func assertPeersWithSpace(t *testing.T, ring *Ring, start, end address.Address, expected int) []router.PeerName {
+func assertPeersWithSpace(t *testing.T, ring *Ring, start, end address.Address, expected int) []mesh.PeerName {
 	peers := ring.ChoosePeersToAskForSpace(start, end)
 	require.Equal(t, expected, len(peers))
 	return peers
@@ -375,7 +375,7 @@ func TestFindFree(t *testing.T) {
 	ring1.assertInvariants()
 
 	// We should return others
-	var peers []router.PeerName
+	var peers []mesh.PeerName
 
 	ring1.Entries = []*entry{{Token: start, Peer: peer2name, Free: 1}}
 	peers = assertPeersWithSpace(t, ring1, start, end, 1)
@@ -493,7 +493,7 @@ func TestOwner(t *testing.T) {
 	require.True(t, ring1.Contains(start), "start should be in ring")
 	require.False(t, ring1.Contains(end), "end should not be in ring")
 
-	require.Equal(t, router.UnknownPeerName, ring1.Owner(start))
+	require.Equal(t, mesh.UnknownPeerName, ring1.Owner(start))
 
 	ring1.ClaimItAll()
 	ring1.GrantRangeToHost(middle, end, peer2name)
@@ -517,9 +517,9 @@ func TestFuzzRing(t *testing.T) {
 		iterations = 1000
 	)
 
-	peers := make([]router.PeerName, numPeers)
+	peers := make([]mesh.PeerName, numPeers)
 	for i := 0; i < numPeers; i++ {
-		peer, _ := router.PeerNameFromString(fmt.Sprintf("%02d:00:00:00:02:00", i))
+		peer, _ := mesh.PeerNameFromString(fmt.Sprintf("%02d:00:00:00:02:00", i))
 		peers[i] = peer
 	}
 
@@ -601,13 +601,13 @@ func TestFuzzRingHard(t *testing.T) {
 	var (
 		numPeers   = 100
 		iterations = 3000
-		peers      []router.PeerName
+		peers      []mesh.PeerName
 		rings      []*Ring
 		nextPeerID = 0
 	)
 
 	addPeer := func() {
-		peer, _ := router.PeerNameFromString(fmt.Sprintf("%02d:%02d:00:00:00:00", nextPeerID/10, nextPeerID%10))
+		peer, _ := mesh.PeerNameFromString(fmt.Sprintf("%02d:%02d:00:00:00:00", nextPeerID/10, nextPeerID%10))
 		common.Log.Debugf("%s: Adding peer", peer)
 		nextPeerID++
 		peers = append(peers, peer)
@@ -620,7 +620,7 @@ func TestFuzzRingHard(t *testing.T) {
 
 	rings[0].ClaimItAll()
 
-	randomPeer := func(exclude int) (int, router.PeerName, *Ring) {
+	randomPeer := func(exclude int) (int, mesh.PeerName, *Ring) {
 		var peerIndex int
 		if exclude >= 0 {
 			peerIndex = rand.Intn(len(peers) - 1)
@@ -729,7 +729,7 @@ func TestFuzzRingHard(t *testing.T) {
 }
 
 func (r *Ring) ClaimItAll() {
-	r.ClaimForPeers([]router.PeerName{r.Peer})
+	r.ClaimForPeers([]mesh.PeerName{r.Peer})
 }
 
 func (es entries) String() string {

ipam/status.go

@@ -1,6 +1,7 @@
 package ipam
 
 import (
+	"fmt"
 	"github.com/weaveworks/weave/ipam/paxos"
 	"github.com/weaveworks/weave/net/address"
 )
@@ -8,6 +9,7 @@ import (
 type Status struct {
 	Paxos            *paxos.Status
 	Range            string
+	RangeNumIPs      int
 	DefaultSubnet    string
 	Entries          []EntryStatus
 	PendingClaims    []ClaimStatus
@@ -15,9 +17,12 @@ type Status struct {
 }
 
 type EntryStatus struct {
-	Token   string
-	Peer    string
-	Version uint32
+	Token       string
+	Size        uint32
+	Peer        string
+	Nickname    string
+	IsKnownPeer bool
+	Version     uint32
 }
 
 type ClaimStatus struct {
@@ -40,6 +45,7 @@ func NewStatus(allocator *Allocator, defaultSubnet address.CIDR) *Status {
 		resultChan <- &Status{
 			paxosStatus,
 			allocator.universe.String(),
+			int(allocator.universe.Size()),
 			defaultSubnet.String(),
 			newEntryStatusSlice(allocator),
 			newClaimStatusSlice(allocator),
@@ -56,8 +62,15 @@ func newEntryStatusSlice(allocator *Allocator) []EntryStatus {
 		return slice
 	}
 
-	for _, entry := range allocator.ring.Entries {
-		slice = append(slice, EntryStatus{entry.Token.String(), entry.Peer.String(), entry.Version})
+	for _, r := range allocator.ring.AllRangeInfo() {
+		slice = append(slice, EntryStatus{
+			Token:       r.Start.String(),
+			Size:        uint32(r.Size()),
+			Peer:        r.Peer.String(),
+			Nickname:    allocator.nicknames[r.Peer],
+			IsKnownPeer: allocator.isKnownPeer(r.Peer),
+			Version:     r.Version,
+		})
 	}
 
 	return slice
@@ -76,7 +89,7 @@ func newAllocateIdentSlice(allocator *Allocator) []string {
 	var slice []string
 	for _, op := range allocator.pendingAllocates {
 		allocate := op.(*allocate)
-		slice = append(slice, allocate.ident)
+		slice = append(slice, fmt.Sprintf("%s %s", allocate.ident, allocate.r.String()))
 	}
 	return slice
 }

ipam/testutils_test.go

@@ -8,13 +8,13 @@ import (
 	"time"
 
 	"github.com/stretchr/testify/require"
+	"github.com/weaveworks/weave/mesh"
 	"github.com/weaveworks/weave/net/address"
-	"github.com/weaveworks/weave/router"
 	"github.com/weaveworks/weave/testing/gossip"
 )
 
 type mockMessage struct {
-	dst     router.PeerName
+	dst     mesh.PeerName
 	msgType byte
 	buf     []byte
 }
@@ -48,13 +48,13 @@ func (m *mockGossipComms) String() string {
 // that the contents of messages are never re-ordered.  Which, for instance,
 // requires they are not based off iterating through a map.
 
-func (m *mockGossipComms) GossipBroadcast(update router.GossipData) error {
+func (m *mockGossipComms) GossipBroadcast(update mesh.GossipData) error {
 	m.Lock()
 	defer m.Unlock()
 	buf := []byte{}
 	if len(m.messages) == 0 {
 		require.FailNow(m, fmt.Sprintf("%s: Gossip broadcast message unexpected: \n%x", m.name, buf))
-	} else if msg := m.messages[0]; msg.dst != router.UnknownPeerName {
+	} else if msg := m.messages[0]; msg.dst != mesh.UnknownPeerName {
 		require.FailNow(m, fmt.Sprintf("%s: Expected Gossip message to %s but got broadcast", m.name, msg.dst))
 	} else if msg.buf != nil && !equalByteBuffer(msg.buf, buf) {
 		require.FailNow(m, fmt.Sprintf("%s: Gossip message not sent as expected: \nwant: %x\ngot : %x", m.name, msg.buf, buf))
@@ -77,12 +77,12 @@ func equalByteBuffer(a, b []byte) bool {
 	return true
 }
 
-func (m *mockGossipComms) GossipUnicast(dstPeerName router.PeerName, buf []byte) error {
+func (m *mockGossipComms) GossipUnicast(dstPeerName mesh.PeerName, buf []byte) error {
 	m.Lock()
 	defer m.Unlock()
 	if len(m.messages) == 0 {
 		require.FailNow(m, fmt.Sprintf("%s: Gossip message to %s unexpected: \n%s", m.name, dstPeerName, buf))
-	} else if msg := m.messages[0]; msg.dst == router.UnknownPeerName {
+	} else if msg := m.messages[0]; msg.dst == mesh.UnknownPeerName {
 		require.FailNow(m, fmt.Sprintf("%s: Expected Gossip broadcast message but got dest %s", m.name, dstPeerName))
 	} else if msg.dst != dstPeerName {
 		require.FailNow(m, fmt.Sprintf("%s: Expected Gossip message to %s but got dest %s", m.name, msg.dst, dstPeerName))
@@ -99,7 +99,7 @@ func (m *mockGossipComms) GossipUnicast(dstPeerName router.PeerName, buf []byte)
 
 func ExpectMessage(alloc *Allocator, dst string, msgType byte, buf []byte) {
 	m := alloc.gossip.(*mockGossipComms)
-	dstPeerName, _ := router.PeerNameFromString(dst)
+	dstPeerName, _ := mesh.PeerNameFromString(dst)
 	m.Lock()
 	m.messages = append(m.messages, mockMessage{dstPeerName, msgType, buf})
 	m.Unlock()
@@ -108,7 +108,7 @@ func ExpectMessage(alloc *Allocator, dst string, msgType byte, buf []byte) {
 func ExpectBroadcastMessage(alloc *Allocator, buf []byte) {
 	m := alloc.gossip.(*mockGossipComms)
 	m.Lock()
-	m.messages = append(m.messages, mockMessage{router.UnknownPeerName, 0, buf})
+	m.messages = append(m.messages, mockMessage{mesh.UnknownPeerName, 0, buf})
 	m.Unlock()
 }
 
@@ -124,7 +124,7 @@ func CheckAllExpectedMessagesSent(allocs ...*Allocator) {
 }
 
 func makeAllocator(name string, cidrStr string, quorum uint) (*Allocator, address.Range) {
-	peername, err := router.PeerNameFromString(name)
+	peername, err := mesh.PeerNameFromString(name)
 	if err != nil {
 		panic(err)
 	}
@@ -134,8 +134,8 @@ func makeAllocator(name string, cidrStr string, quorum uint) (*Allocator, addres
 		panic(err)
 	}
 
-	alloc := NewAllocator(peername, router.PeerUID(rand.Int63()),
-		"nick-"+name, cidr.Range(), quorum)
+	alloc := NewAllocator(peername, mesh.PeerUID(rand.Int63()),
+		"nick-"+name, cidr.Range(), quorum, func(mesh.PeerName) bool { return true })
 
 	return alloc, cidr.HostRange()
 }
@@ -149,7 +149,7 @@ func makeAllocatorWithMockGossip(t *testing.T, name string, universeCIDR string,
 }
 
 func (alloc *Allocator) claimRingForTesting(allocs ...*Allocator) {
-	peers := []router.PeerName{alloc.ourName}
+	peers := []mesh.PeerName{alloc.ourName}
 	for _, alloc2 := range allocs {
 		peers = append(peers, alloc2.ourName)
 	}

mesh/connection.go

@@ -0,0 +1,460 @@
+package mesh
+
+import (
+	"fmt"
+	"net"
+	"strconv"
+	"sync"
+	"time"
+)
+
+type Connection interface {
+	Local() *Peer
+	Remote() *Peer
+	RemoteTCPAddr() string
+	Outbound() bool
+	Established() bool
+	BreakTie(Connection) ConnectionTieBreak
+	Shutdown(error)
+	Log(args ...interface{})
+}
+
+type ConnectionTieBreak int
+
+const (
+	TieBreakWon ConnectionTieBreak = iota
+	TieBreakLost
+	TieBreakTied
+)
+
+var ErrConnectToSelf = fmt.Errorf("Cannot connect to ourself")
+
+type RemoteConnection struct {
+	local         *Peer
+	remote        *Peer
+	remoteTCPAddr string
+	outbound      bool
+	established   bool
+}
+
+type LocalConnection struct {
+	sync.RWMutex
+	RemoteConnection
+	TCPConn         *net.TCPConn
+	TrustRemote     bool // is remote on a trusted subnet?
+	TrustedByRemote bool // does remote trust us?
+	version         byte
+	tcpSender       TCPSender
+	SessionKey      *[32]byte
+	heartbeatTCP    *time.Ticker
+	Router          *Router
+	uid             uint64
+	actionChan      chan<- ConnectionAction
+	errorChan       chan<- error
+	finished        <-chan struct{} // closed to signal that actorLoop has finished
+	OverlayConn     OverlayConnection
+}
+
+type ConnectionAction func() error
+
+func NewRemoteConnection(from, to *Peer, tcpAddr string, outbound bool, established bool) *RemoteConnection {
+	return &RemoteConnection{
+		local:         from,
+		remote:        to,
+		remoteTCPAddr: tcpAddr,
+		outbound:      outbound,
+		established:   established,
+	}
+}
+
+func (conn *RemoteConnection) Local() *Peer                           { return conn.local }
+func (conn *RemoteConnection) Remote() *Peer                          { return conn.remote }
+func (conn *RemoteConnection) RemoteTCPAddr() string                  { return conn.remoteTCPAddr }
+func (conn *RemoteConnection) Outbound() bool                         { return conn.outbound }
+func (conn *RemoteConnection) Established() bool                      { return conn.established }
+func (conn *RemoteConnection) BreakTie(Connection) ConnectionTieBreak { return TieBreakTied }
+func (conn *RemoteConnection) Shutdown(error)                         {}
+
+func (conn *RemoteConnection) Log(args ...interface{}) {
+	log.Println(append(append([]interface{}{}, fmt.Sprintf("->[%s|%s]:", conn.remoteTCPAddr, conn.remote)), args...)...)
+}
+
+func (conn *RemoteConnection) ErrorLog(args ...interface{}) {
+	log.Errorln(append(append([]interface{}{}, fmt.Sprintf("->[%s|%s]:", conn.remoteTCPAddr, conn.remote)), args...)...)
+}
+
+// Does not return anything. If the connection is successful, it will
+// end up in the local peer's connections map.
+func StartLocalConnection(connRemote *RemoteConnection, tcpConn *net.TCPConn, router *Router, acceptNewPeer bool) {
+	if connRemote.local != router.Ourself.Peer {
+		log.Fatal("Attempt to create local connection from a peer which is not ourself")
+	}
+	actionChan := make(chan ConnectionAction, ChannelSize)
+	errorChan := make(chan error, 1)
+	finished := make(chan struct{})
+	conn := &LocalConnection{
+		RemoteConnection: *connRemote, // NB, we're taking a copy of connRemote here.
+		Router:           router,
+		TCPConn:          tcpConn,
+		TrustRemote:      router.Trusts(connRemote),
+		uid:              randUint64(),
+		actionChan:       actionChan,
+		errorChan:        errorChan,
+		finished:         finished}
+	go conn.run(actionChan, errorChan, finished, acceptNewPeer)
+}
+
+func (conn *LocalConnection) BreakTie(dupConn Connection) ConnectionTieBreak {
+	dupConnLocal := dupConn.(*LocalConnection)
+	// conn.uid is used as the tie breaker here, in the knowledge that
+	// both sides will make the same decision.
+	if conn.uid < dupConnLocal.uid {
+		return TieBreakWon
+	} else if dupConnLocal.uid < conn.uid {
+		return TieBreakLost
+	} else {
+		return TieBreakTied
+	}
+}
+
+func (conn *LocalConnection) Established() bool {
+	return conn.established
+}
+
+func (conn *LocalConnection) SendProtocolMsg(m ProtocolMsg) {
+	if err := conn.sendProtocolMsg(m); err != nil {
+		conn.Shutdown(err)
+	}
+}
+
+// ACTOR methods
+
+// NB: The conn.* fields are only written by the connection actor
+// process, which is the caller of the ConnectionAction funs. Hence we
+// do not need locks for reading, and only need write locks for fields
+// read by other processes.
+
+// Non-blocking.
+func (conn *LocalConnection) Shutdown(err error) {
+	// err should always be a real error, even if only io.EOF
+	if err == nil {
+		panic("nil error")
+	}
+
+	select {
+	case conn.errorChan <- err:
+	default:
+	}
+}
+
+// Send an actor request to the actorLoop, but don't block if
+// actorLoop has exited - see http://blog.golang.org/pipelines for
+// pattern
+func (conn *LocalConnection) sendAction(action ConnectionAction) {
+	select {
+	case conn.actionChan <- action:
+	case <-conn.finished:
+	}
+}
+
+// ACTOR server
+
+func (conn *LocalConnection) run(actionChan <-chan ConnectionAction, errorChan <-chan error, finished chan<- struct{}, acceptNewPeer bool) {
+	var err error // important to use this var and not create another one with 'err :='
+	defer func() { conn.shutdown(err) }()
+	defer close(finished)
+
+	conn.TCPConn.SetLinger(0)
+	intro, err := ProtocolIntroParams{
+		MinVersion: conn.Router.ProtocolMinVersion,
+		MaxVersion: ProtocolMaxVersion,
+		Features:   conn.makeFeatures(),
+		Conn:       conn.TCPConn,
+		Password:   conn.Router.Password,
+		Outbound:   conn.outbound,
+	}.DoIntro()
+	if err != nil {
+		return
+	}
+
+	conn.SessionKey = intro.SessionKey
+	conn.tcpSender = intro.Sender
+	conn.version = intro.Version
+
+	remote, err := conn.parseFeatures(intro.Features)
+	if err != nil {
+		return
+	}
+
+	if err = conn.registerRemote(remote, acceptNewPeer); err != nil {
+		return
+	}
+
+	conn.Log("connection ready; using protocol version", conn.version)
+
+	// only use negotiated session key for untrusted connections
+	var sessionKey *[32]byte
+	if conn.Untrusted() {
+		sessionKey = conn.SessionKey
+	}
+
+	params := OverlayConnectionParams{
+		RemotePeer:         conn.remote,
+		LocalAddr:          conn.TCPConn.LocalAddr().(*net.TCPAddr),
+		RemoteAddr:         conn.TCPConn.RemoteAddr().(*net.TCPAddr),
+		Outbound:           conn.outbound,
+		ConnUID:            conn.uid,
+		SessionKey:         sessionKey,
+		SendControlMessage: conn.sendOverlayControlMessage,
+		Features:           intro.Features,
+	}
+	if conn.OverlayConn, err = conn.Router.Overlay.PrepareConnection(params); err != nil {
+		return
+	}
+
+	// As soon as we do AddConnection, the new connection becomes
+	// visible to the packet routing logic.  So AddConnection must
+	// come after PrepareConnection
+	if err = conn.Router.Ourself.AddConnection(conn); err != nil {
+		return
+	}
+	conn.Router.ConnectionMaker.ConnectionCreated(conn)
+
+	// OverlayConnection confirmation comes after AddConnection,
+	// because only after that completes do we know the connection is
+	// valid: in particular that it is not a duplicate connection to
+	// the same peer. Overlay communication on a duplicate connection
+	// can cause problems such as tripping up overlay crypto at the
+	// other end due to data being decoded by the other connection. It
+	// is also generally wasteful to engage in any interaction with
+	// the remote on a connection that turns out to be invalid.
+	conn.OverlayConn.Confirm()
+
+	// receiveTCP must follow also AddConnection. In the absence
+	// of any indirect connectivity to the remote peer, the first
+	// we hear about it (and any peers reachable from it) is
+	// through topology gossip it sends us on the connection. We
+	// must ensure that the connection has been added to Ourself
+	// prior to processing any such gossip, otherwise we risk
+	// immediately gc'ing part of that newly received portion of
+	// the topology (though not the remote peer itself, since that
+	// will have a positive ref count), leaving behind dangling
+	// references to peers. Hence we must invoke AddConnection,
+	// which is *synchronous*, first.
+	conn.heartbeatTCP = time.NewTicker(TCPHeartbeat)
+	go conn.receiveTCP(intro.Receiver)
+
+	// AddConnection must precede actorLoop. More precisely, it
+	// must precede shutdown, since that invokes DeleteConnection
+	// and is invoked on termination of this entire
+	// function. Essentially this boils down to a prohibition on
+	// running AddConnection in a separate goroutine, at least not
+	// without some synchronisation. Which in turn requires the
+	// launching of the receiveTCP goroutine to precede actorLoop.
+	err = conn.actorLoop(actionChan, errorChan)
+}
+
+func (conn *LocalConnection) makeFeatures() map[string]string {
+	features := map[string]string{
+		"PeerNameFlavour": PeerNameFlavour,
+		"Name":            conn.local.Name.String(),
+		"NickName":        conn.local.NickName,
+		"ShortID":         fmt.Sprint(conn.local.ShortID),
+		"UID":             fmt.Sprint(conn.local.UID),
+		"ConnID":          fmt.Sprint(conn.uid),
+		"Trusted":         fmt.Sprint(conn.TrustRemote),
+	}
+	conn.Router.Overlay.AddFeaturesTo(features)
+	return features
+}
+
+type features map[string]string
+
+func (features features) MustHave(keys []string) error {
+	for _, key := range keys {
+		if _, ok := features[key]; !ok {
+			return fmt.Errorf("Field %s is missing", key)
+		}
+	}
+	return nil
+}
+
+func (features features) Get(key string) string {
+	return features[key]
+}
+
+func (conn *LocalConnection) parseFeatures(features features) (*Peer, error) {
+	if err := features.MustHave([]string{"PeerNameFlavour", "Name", "NickName", "UID", "ConnID"}); err != nil {
+		return nil, err
+	}
+
+	remotePeerNameFlavour := features.Get("PeerNameFlavour")
+	if remotePeerNameFlavour != PeerNameFlavour {
+		return nil, fmt.Errorf("Peer name flavour mismatch (ours: '%s', theirs: '%s')", PeerNameFlavour, remotePeerNameFlavour)
+	}
+
+	name, err := PeerNameFromString(features.Get("Name"))
+	if err != nil {
+		return nil, err
+	}
+
+	nickName := features.Get("NickName")
+
+	var shortID uint64
+	var hasShortID bool
+	if shortIDStr, present := features["ShortID"]; present {
+		hasShortID = true
+		shortID, err = strconv.ParseUint(shortIDStr, 10, PeerShortIDBits)
+		if err != nil {
+			return nil, err
+		}
+	}
+
+	var trusted bool
+	if trustedStr, present := features["Trusted"]; present {
+		trusted, err = strconv.ParseBool(trustedStr)
+		if err != nil {
+			return nil, err
+		}
+	}
+	conn.TrustedByRemote = trusted
+
+	uid, err := ParsePeerUID(features.Get("UID"))
+	if err != nil {
+		return nil, err
+	}
+
+	remoteConnID, err := strconv.ParseUint(features.Get("ConnID"), 10, 64)
+	if err != nil {
+		return nil, err
+	}
+
+	conn.uid ^= remoteConnID
+	peer := NewPeer(name, nickName, uid, 0, PeerShortID(shortID))
+	peer.HasShortID = hasShortID
+	return peer, nil
+}
+
+func (conn *LocalConnection) registerRemote(remote *Peer, acceptNewPeer bool) error {
+	if acceptNewPeer {
+		conn.remote = conn.Router.Peers.FetchWithDefault(remote)
+	} else {
+		conn.remote = conn.Router.Peers.FetchAndAddRef(remote.Name)
+		if conn.remote == nil {
+			return fmt.Errorf("Found unknown remote name: %s at %s", remote.Name, conn.remoteTCPAddr)
+		}
+	}
+
+	if conn.remote == conn.local {
+		return ErrConnectToSelf
+	}
+
+	return nil
+}
+
+func (conn *LocalConnection) actorLoop(actionChan <-chan ConnectionAction, errorChan <-chan error) (err error) {
+	fwdErrorChan := conn.OverlayConn.ErrorChannel()
+	fwdEstablishedChan := conn.OverlayConn.EstablishedChannel()
+
+	for err == nil {
+		select {
+		case action := <-actionChan:
+			err = action()
+
+		case err = <-errorChan:
+
+		case <-conn.heartbeatTCP.C:
+			err = conn.sendSimpleProtocolMsg(ProtocolHeartbeat)
+
+		case <-fwdEstablishedChan:
+			conn.established = true
+			fwdEstablishedChan = nil
+			conn.Router.Ourself.ConnectionEstablished(conn)
+
+		case err = <-fwdErrorChan:
+		}
+	}
+	return
+}
+
+func (conn *LocalConnection) shutdown(err error) {
+	if conn.remote == nil {
+		log.Errorf("->[%s] connection shutting down due to error during handshake: %v", conn.remoteTCPAddr, err)
+	} else {
+		conn.ErrorLog("connection shutting down due to error:", err)
+	}
+
+	if conn.TCPConn != nil {
+		checkWarn(conn.TCPConn.Close())
+	}
+
+	if conn.remote != nil {
+		conn.Router.Peers.Dereference(conn.remote)
+		conn.Router.Ourself.DeleteConnection(conn)
+	}
+
+	if conn.heartbeatTCP != nil {
+		conn.heartbeatTCP.Stop()
+	}
+
+	if conn.OverlayConn != nil {
+		conn.OverlayConn.Stop()
+	}
+
+	conn.Router.ConnectionMaker.ConnectionTerminated(conn, err)
+}
+
+func (conn *LocalConnection) sendOverlayControlMessage(tag byte, msg []byte) error {
+	return conn.sendProtocolMsg(ProtocolMsg{ProtocolTag(tag), msg})
+}
+
+// Helpers
+
+func (conn *LocalConnection) sendSimpleProtocolMsg(tag ProtocolTag) error {
+	return conn.sendProtocolMsg(ProtocolMsg{tag: tag})
+}
+
+func (conn *LocalConnection) sendProtocolMsg(m ProtocolMsg) error {
+	return conn.tcpSender.Send(append([]byte{byte(m.tag)}, m.msg...))
+}
+
+func (conn *LocalConnection) receiveTCP(receiver TCPReceiver) {
+	var err error
+	for {
+		conn.extendReadDeadline()
+
+		var msg []byte
+		if msg, err = receiver.Receive(); err != nil {
+			break
+		}
+		if len(msg) < 1 {
+			conn.Log("ignoring blank msg")
+			continue
+		}
+		if err = conn.handleProtocolMsg(ProtocolTag(msg[0]), msg[1:]); err != nil {
+			break
+		}
+	}
+	conn.Shutdown(err)
+}
+
+func (conn *LocalConnection) handleProtocolMsg(tag ProtocolTag, payload []byte) error {
+	switch tag {
+	case ProtocolHeartbeat:
+	case ProtocolReserved1, ProtocolReserved2, ProtocolReserved3, ProtocolOverlayControlMsg:
+		conn.OverlayConn.ControlMessage(byte(tag), payload)
+	case ProtocolGossipUnicast, ProtocolGossipBroadcast, ProtocolGossip:
+		return conn.Router.handleGossip(tag, payload)
+	default:
+		conn.Log("ignoring unknown protocol tag:", tag)
+	}
+	return nil
+}
+
+func (conn *LocalConnection) extendReadDeadline() {
+	conn.TCPConn.SetReadDeadline(time.Now().Add(TCPHeartbeat * 2))
+}
+
+func (conn *LocalConnection) Untrusted() bool {
+	return !conn.TrustRemote || !conn.TrustedByRemote
+}

mesh/connection_maker.go

@@ -1,4 +1,4 @@
-package router
+package mesh
 
 import (
 	"fmt"
@@ -10,6 +10,7 @@ import (
 const (
 	InitialInterval = 2 * time.Second
 	MaxInterval     = 6 * time.Minute
+	ResetAfter      = 1 * time.Minute
 )
 
 type peerAddrs map[string]*net.TCPAddr
@@ -20,13 +21,22 @@ type ConnectionMaker struct {
 	port        int
 	discovery   bool
 	targets     map[string]*Target
+	connections map[Connection]struct{}
 	directPeers peerAddrs
 	actionChan  chan<- ConnectionMakerAction
 }
 
+type TargetState int
+
+const (
+	TargetWaiting    TargetState = iota // we are waiting to connect there
+	TargetAttempting                    // we are attempting to connect there
+	TargetConnected                     // we are connected to there
+)
+
 // Information about an address where we may find a peer
 type Target struct {
-	attempting  bool          // are we currently attempting to connect there?
+	state       TargetState
 	lastError   error         // reason for disconnection last time
 	tryAfter    time.Time     // next time to try this address
 	tryInterval time.Duration // retry delay on next failure
@@ -43,6 +53,7 @@ func NewConnectionMaker(ourself *LocalPeer, peers *Peers, port int, discovery bo
 		discovery:   discovery,
 		directPeers: peerAddrs{},
 		targets:     make(map[string]*Target),
+		connections: make(map[Connection]struct{}),
 		actionChan:  actionChan}
 	go cm.queryLoop(actionChan)
 	return cm
@@ -70,8 +81,8 @@ func (cm *ConnectionMaker) InitiateConnections(peers []string, replace bool) []e
 		for peer, addr := range addrs {
 			cm.directPeers[peer] = addr
 			// curtail any existing reconnect interval
-			if target, found := cm.targets[addr.String()]; found {
-				target.tryNow()
+			if target, found := cm.targets[cm.completeAddr(*addr)]; found {
+				target.nextTryNow()
 			}
 		}
 		return true
@@ -88,12 +99,42 @@ func (cm *ConnectionMaker) ForgetConnections(peers []string) {
 	}
 }
 
-func (cm *ConnectionMaker) ConnectionTerminated(address string, err error) {
+func (cm *ConnectionMaker) ConnectionAborted(address string, err error) {
 	cm.actionChan <- func() bool {
-		if target, found := cm.targets[address]; found {
-			target.attempting = false
+		target := cm.targets[address]
+		target.state = TargetWaiting
+		target.lastError = err
+		target.nextTryLater()
+		return true
+	}
+}
+
+func (cm *ConnectionMaker) ConnectionCreated(conn Connection) {
+	cm.actionChan <- func() bool {
+		cm.connections[conn] = void
+		if conn.Outbound() {
+			target := cm.targets[conn.RemoteTCPAddr()]
+			target.state = TargetConnected
+		}
+		return false
+	}
+}
+
+func (cm *ConnectionMaker) ConnectionTerminated(conn Connection, err error) {
+	cm.actionChan <- func() bool {
+		delete(cm.connections, conn)
+		if conn.Outbound() {
+			target := cm.targets[conn.RemoteTCPAddr()]
+			target.state = TargetWaiting
 			target.lastError = err
-			target.retry()
+			switch {
+			case err == ErrConnectToSelf:
+				target.nextTryNever()
+			case time.Now().After(target.tryAfter.Add(ResetAfter)):
+				target.nextTryNow()
+			default:
+				target.nextTryLater()
+			}
 		}
 		return true
 	}
@@ -118,14 +159,18 @@ func (cm *ConnectionMaker) queryLoop(actionChan <-chan ConnectionMakerAction) {
 	}
 }
 
+func (cm *ConnectionMaker) completeAddr(addr net.TCPAddr) string {
+	if addr.Port == 0 {
+		addr.Port = cm.port
+	}
+	return addr.String()
+}
+
 func (cm *ConnectionMaker) checkStateAndAttemptConnections() time.Duration {
 	var (
 		validTarget  = make(map[string]struct{})
 		directTarget = make(map[string]struct{})
 	)
-	// Copy the set of things we are connected to, so we can access
-	// them without locking.  Also clear out any entries in cm.targets
-	// for existing connections.
 	ourConnectedPeers, ourConnectedTargets, ourInboundIPs := cm.ourConnections()
 
 	addTarget := func(address string) {
@@ -136,25 +181,23 @@ func (cm *ConnectionMaker) checkStateAndAttemptConnections() time.Duration {
 		if _, found := cm.targets[address]; found {
 			return
 		}
-		target := &Target{}
-		target.tryNow()
+		target := &Target{state: TargetWaiting}
+		target.nextTryNow()
 		cm.targets[address] = target
 	}
 
 	// Add direct targets that are not connected
 	for _, addr := range cm.directPeers {
-		completeAddr := *addr
 		attempt := true
-		if completeAddr.Port == 0 {
-			completeAddr.Port = cm.port
+		if addr.Port == 0 {
 			// If a peer was specified w/o a port, then we do not
 			// attempt to connect to it if we have any inbound
 			// connections from that IP.
-			if _, connected := ourInboundIPs[completeAddr.IP.String()]; connected {
+			if _, connected := ourInboundIPs[addr.IP.String()]; connected {
 				attempt = false
 			}
 		}
-		address := completeAddr.String()
+		address := cm.completeAddr(*addr)
 		directTarget[address] = void
 		if attempt {
 			addTarget(address)
@@ -176,9 +219,8 @@ func (cm *ConnectionMaker) ourConnections() (PeerNameSet, map[string]struct{}, m
 		ourConnectedTargets = make(map[string]struct{})
 		ourInboundIPs       = make(map[string]struct{})
 	)
-	for conn := range cm.ourself.Connections() {
+	for conn := range cm.connections {
 		address := conn.RemoteTCPAddr()
-		delete(cm.targets, address)
 		ourConnectedPeers[conn.Remote().Name] = void
 		ourConnectedTargets[address] = void
 		if conn.Outbound() {
@@ -212,7 +254,7 @@ func (cm *ConnectionMaker) addPeerTargets(ourConnectedPeers PeerNameSet, addTarg
 			} else if ip, _, err := net.SplitHostPort(address); err == nil {
 				// There is no point connecting to the (likely
 				// ephemeral) remote port of an inbound connection
-				// that some peer has. Let's try to connect to on the
+				// that some peer has. Let's try to connect on the
 				// weave port instead.
 				addTarget(fmt.Sprintf("%s:%d", ip, cm.port))
 			}
@@ -224,7 +266,7 @@ func (cm *ConnectionMaker) connectToTargets(validTarget map[string]struct{}, dir
 	now := time.Now() // make sure we catch items just added
 	after := MaxDuration
 	for address, target := range cm.targets {
-		if target.attempting {
+		if target.state != TargetWaiting {
 			continue
 		}
 		if _, valid := validTarget[address]; !valid {
@@ -236,7 +278,7 @@ func (cm *ConnectionMaker) connectToTargets(validTarget map[string]struct{}, dir
 		}
 		switch duration := target.tryAfter.Sub(now); {
 		case duration <= 0:
-			target.attempting = true
+			target.state = TargetAttempting
 			_, isCmdLineTarget := directTarget[address]
 			go cm.attemptConnection(address, isCmdLineTarget)
 		case duration < after:
@@ -250,23 +292,23 @@ func (cm *ConnectionMaker) attemptConnection(address string, acceptNewPeer bool)
 	log.Printf("->[%s] attempting connection", address)
 	if err := cm.ourself.CreateConnection(address, acceptNewPeer); err != nil {
 		log.Errorf("->[%s] error during connection attempt: %v", address, err)
-		cm.ConnectionTerminated(address, err)
+		cm.ConnectionAborted(address, err)
 	}
 }
 
-func (t *Target) tryNow() {
+func (t *Target) nextTryNever() {
+	t.tryAfter = time.Time{}
+	t.tryInterval = MaxInterval
+}
+
+func (t *Target) nextTryNow() {
 	t.tryAfter = time.Now()
 	t.tryInterval = InitialInterval
 }
 
 // The delay at the nth retry is a random value in the range
 // [i-i/2,i+i/2], where i = InitialInterval * 1.5^(n-1).
-func (t *Target) retry() {
-	if t.lastError == ErrConnectToSelf {
-		t.tryAfter = time.Time{}
-		t.tryInterval = MaxInterval
-		return
-	}
+func (t *Target) nextTryLater() {
 	t.tryAfter = time.Now().Add(t.tryInterval/2 + time.Duration(rand.Int63n(int64(t.tryInterval))))
 	t.tryInterval = t.tryInterval * 3 / 2
 	if t.tryInterval > MaxInterval {

mesh/gossip.go

@@ -1,4 +1,4 @@
-package router
+package mesh
 
 import (
 	"fmt"
@@ -6,7 +6,7 @@ import (
 
 type GossipData interface {
 	Encode() [][]byte
-	Merge(GossipData)
+	Merge(GossipData) GossipData
 }
 
 type Gossip interface {
@@ -79,8 +79,7 @@ func (sender *GossipSender) Send(data GossipData) {
 	// NB: this must not be invoked concurrently
 	select {
 	case pending := <-sender.cell:
-		pending.Merge(data)
-		sender.cell <- pending
+		sender.cell <- pending.Merge(data)
 	default:
 		sender.cell <- data
 	}
@@ -142,7 +141,7 @@ func (router *Router) SendAllGossip() {
 func (router *Router) SendAllGossipDown(conn Connection) {
 	for channel := range router.gossipChannelSet() {
 		if gossip := channel.gossiper.Gossip(); gossip != nil {
-			channel.SendDown(conn, channel.gossiper.Gossip())
+			channel.SendDown(conn, gossip)
 		}
 	}
 }

mesh/gossip_channel.go

@@ -1,4 +1,4 @@
-package router
+package mesh
 
 import (
 	"bytes"
@@ -97,78 +97,6 @@ func (c *GossipChannel) deliver(srcName PeerName, _ []byte, dec *gob.Decoder) er
 	return nil
 }
 
-func (c *GossipChannel) Send(srcName PeerName, data GossipData) {
-	// do this outside the lock below so we avoid lock nesting
-	c.routes.EnsureRecalculated()
-	selectedConnections := make(ConnectionSet)
-	for name := range c.routes.RandomNeighbours(srcName) {
-		if conn, found := c.ourself.ConnectionTo(name); found {
-			selectedConnections[conn] = void
-		}
-	}
-	if len(selectedConnections) == 0 {
-		return
-	}
-	connections := c.ourself.Connections()
-	c.Lock()
-	defer c.Unlock()
-	c.gcSenders(connections)
-	for conn := range selectedConnections {
-		c.sendDown(conn, data)
-	}
-}
-
-func (c *GossipChannel) SendDown(conn Connection, data GossipData) {
-	connections := c.ourself.Connections()
-	c.Lock()
-	defer c.Unlock()
-	c.gcSenders(connections)
-	c.sendDown(conn, data)
-}
-
-// GC - randomly (courtesy of go's map iterator) pick some existing
-// senders and stop&remove them if the associated connection is no
-// longer active.  We stop as soon as we encounter a valid entry; the
-// idea being that when there is little or no garbage then this
-// executes close to O(1)[1], whereas when there is lots of garbage we
-// remove it quickly.
-//
-// [1] TODO Unfortunately, due to the desire to avoid nested locks,
-// instead of simply invoking LocalPeer.ConnectionTo(name), we pass in
-// the result of LocalPeer.Connections(). That is O(n_our_connections)
-// at best.
-func (c *GossipChannel) gcSenders(connections ConnectionSet) {
-	for conn, sender := range c.senders {
-		if _, found := connections[conn]; !found {
-			delete(c.senders, conn)
-			sender.Stop()
-		} else {
-			break
-		}
-	}
-}
-
-// We have seen a couple of failures which suggest a >128GB slice was encountered.
-// 100MB should be enough for anyone.
-const maxFeasibleMessageLen = 100 * 1024 * 1024
-
-func (c *GossipChannel) sendDown(conn Connection, data GossipData) {
-	sender, found := c.senders[conn]
-	if !found {
-		sender = NewGossipSender(func(pending GossipData) {
-			for _, msg := range pending.Encode() {
-				if len(msg) > maxFeasibleMessageLen {
-					panic(fmt.Sprintf("Gossip message too large: len=%d bytes; on channel '%s' from %+v", len(msg), c.name, pending))
-				}
-				protocolMsg := ProtocolMsg{ProtocolGossip, GobEncode(c.name, c.ourself.Name, msg)}
-				conn.(ProtocolSender).SendProtocolMsg(protocolMsg)
-			}
-		})
-		c.senders[conn] = sender
-	}
-	sender.Send(data)
-}
-
 func (c *GossipChannel) GossipUnicast(dstPeerName PeerName, msg []byte) error {
 	return c.relayUnicast(dstPeerName, GobEncode(c.name, c.ourself.Name, dstPeerName, msg))
 }
@@ -236,6 +164,81 @@ func (c *GossipChannel) sendBroadcast(srcName PeerName, update GossipData) {
 	}
 }
 
+func (c *GossipChannel) Send(srcName PeerName, data GossipData) {
+	c.routes.EnsureRecalculated()
+	destinations := c.routes.RandomNeighbours(srcName)
+	if len(destinations) == 0 {
+		return
+	}
+	c.sendDown(c.ourself.ConnectionsTo(destinations), data)
+}
+
+func (c *GossipChannel) SendDown(conn Connection, data GossipData) {
+	c.sendDown([]Connection{conn}, data)
+}
+
+func (c *GossipChannel) sendDown(selectedConnections []Connection, data GossipData) {
+	if len(selectedConnections) == 0 {
+		return
+	}
+	connections := c.ourself.Connections()
+	c.Lock()
+	defer c.Unlock()
+	// GC - randomly (courtesy of go's map iterator) pick some
+	// existing senders and stop&remove them if the associated
+	// connection is no longer active.  We stop as soon as we
+	// encounter a valid entry; the idea being that when there is
+	// little or no garbage then this executes close to O(1)[1],
+	// whereas when there is lots of garbage we remove it quickly.
+	//
+	// [1] TODO Unfortunately, due to the desire to avoid nested
+	// locks, instead of simply invoking LocalPeer.ConnectionTo(name),
+	// we operate on LocalPeer.Connections(). That is
+	// O(n_our_connections) at best.
+	for conn, sender := range c.senders {
+		if _, found := connections[conn]; !found {
+			delete(c.senders, conn)
+			sender.Stop()
+		} else {
+			break
+		}
+	}
+	// start senders, if necessary, and send.
+	for _, conn := range selectedConnections {
+		sender, found := c.senders[conn]
+		if !found {
+			sender = c.makeSender(conn)
+			c.senders[conn] = sender
+		}
+		sender.Send(data)
+	}
+}
+
+// We have seen a couple of failures which suggest a >128GB slice was encountered.
+// 100MB should be enough for anyone.
+const maxFeasibleMessageLen = 100 * 1024 * 1024
+
+func (c *GossipChannel) makeSender(conn Connection) *GossipSender {
+	return NewGossipSender(func(pending GossipData) {
+		for _, msg := range pending.Encode() {
+			if len(msg) > maxFeasibleMessageLen {
+				panic(fmt.Sprintf("Gossip message too large: len=%d bytes; on channel '%s' from %+v", len(msg), c.name, pending))
+			}
+			protocolMsg := ProtocolMsg{ProtocolGossip, GobEncode(c.name, c.ourself.Name, msg)}
+			conn.(ProtocolSender).SendProtocolMsg(protocolMsg)
+		}
+	})
+}
+
 func (c *GossipChannel) log(args ...interface{}) {
 	log.Println(append(append([]interface{}{}, "[gossip "+c.name+"]:"), args...)...)
 }
+
+func GobEncode(items ...interface{}) []byte {
+	buf := new(bytes.Buffer)
+	enc := gob.NewEncoder(buf)
+	for _, i := range items {
+		checkFatal(enc.Encode(i))
+	}
+	return buf.Bytes()
+}

mesh/gossip_test.go

@@ -1,36 +1,40 @@
-package router
+package mesh
 
 import (
 	"fmt"
 	"sync"
 	"testing"
-	"time"
 
 	"github.com/stretchr/testify/require"
-	wt "github.com/weaveworks/weave/testing"
 )
 
 // TODO test gossip unicast; atm we only test topology gossip and
 // surrogates, neither of which employ unicast.
 
-type mockChannelConnection struct {
+type mockGossipConnection struct {
 	RemoteConnection
-	dest *Router
+	dest  *Router
+	start chan struct{}
 }
 
 func NewTestRouter(name string) *Router {
 	peerName, _ := PeerNameFromString(name)
-	router := NewRouter(Config{}, peerName, "nick")
+	router := NewRouter(Config{}, peerName, "nick", nil)
 	router.Start()
 	return router
 }
 
-func (conn *mockChannelConnection) SendProtocolMsg(protocolMsg ProtocolMsg) {
+func (conn *mockGossipConnection) SendProtocolMsg(protocolMsg ProtocolMsg) {
+	<-conn.start
 	if err := conn.dest.handleGossip(protocolMsg.tag, protocolMsg.msg); err != nil {
 		panic(err)
 	}
 }
 
+func (conn *mockGossipConnection) Start() {
+	close(conn.start)
+}
+
 func sendPendingGossip(routers ...*Router) {
 	// Loop until all routers report they didn't send anything
 	for sentSomething := true; sentSomething; {
@@ -41,52 +45,43 @@ func sendPendingGossip(routers ...*Router) {
 	}
 }
 
-func (router *Router) AddTestChannelConnection(r *Router) {
-	fromName := router.Ourself.Peer.Name
-	toName := r.Ourself.Peer.Name
+func AddTestGossipConnection(r1, r2 *Router) {
+	c1 := r1.NewTestGossipConnection(r2)
+	c2 := r2.NewTestGossipConnection(r1)
+	c1.Start()
+	c2.Start()
+}
 
-	fromPeer := NewPeer(fromName, "", router.Ourself.Peer.UID, 0)
-	toPeer := NewPeer(toName, "", r.Ourself.Peer.UID, 0)
+func (router *Router) NewTestGossipConnection(r *Router) *mockGossipConnection {
+	to := r.Ourself.Peer
+	toPeer := NewPeer(to.Name, to.NickName, to.UID, 0, to.ShortID)
+	toPeer = router.Peers.FetchWithDefault(toPeer) // Has side-effect of incrementing refcount
 
-	r.Peers.FetchWithDefault(fromPeer)    // Has side-effect of incrementing refcount
-	router.Peers.FetchWithDefault(toPeer) //
-
-	conn := &mockChannelConnection{RemoteConnection{router.Ourself.Peer, toPeer, "", false, true}, r}
+	conn := &mockGossipConnection{
+		RemoteConnection{router.Ourself.Peer, toPeer, "", false, true}, r, make(chan struct{})}
 	router.Ourself.handleAddConnection(conn)
 	router.Ourself.handleConnectionEstablished(conn)
+	return conn
 }
 
-func (router *Router) DeleteTestChannelConnection(r *Router) {
-	fromName := router.Ourself.Peer.Name
+func (router *Router) DeleteTestGossipConnection(r *Router) {
 	toName := r.Ourself.Peer.Name
-
-	fromPeer := r.Peers.Fetch(fromName)
-	toPeer := router.Peers.Fetch(toName)
-
-	r.Peers.Dereference(fromPeer)
-	router.Peers.Dereference(toPeer)
-
 	conn, _ := router.Ourself.ConnectionTo(toName)
+	router.Peers.Dereference(conn.Remote())
 	router.Ourself.handleDeleteConnection(conn)
 }
 
-func TestGossipTopology(t *testing.T) {
-	wt.RunWithTimeout(t, 5*time.Second, func() {
-		implTestGossipTopology(t)
-	})
-}
-
 // Create a Peer representing the receiver router, with connections to
 // the routers supplied as arguments, carrying across all UID and
 // version information.
 func (router *Router) tp(routers ...*Router) *Peer {
-	peer := NewPeer(router.Ourself.Peer.Name, "", router.Ourself.Peer.UID, 0)
+	peer := NewPeerFrom(router.Ourself.Peer)
 	connections := make(map[PeerName]Connection)
 	for _, r := range routers {
-		p := NewPeer(r.Ourself.Peer.Name, "", r.Ourself.Peer.UID, r.Ourself.Peer.version)
+		p := NewPeerFrom(r.Ourself.Peer)
 		connections[r.Ourself.Peer.Name] = newMockConnection(peer, p)
 	}
-	peer.version = router.Ourself.Peer.version
+	peer.Version = router.Ourself.Peer.Version
 	peer.connections = connections
 	return peer
 }
@@ -98,69 +93,46 @@ func checkTopology(t *testing.T, router *Router, wantedPeers ...*Peer) {
 	router.Peers.RUnlock()
 }
 
-func implTestGossipTopology(t *testing.T) {
+func flushAndCheckTopology(t *testing.T, routers []*Router, wantedPeers ...*Peer) {
+	sendPendingGossip(routers...)
+	for _, r := range routers {
+		checkTopology(t, r, wantedPeers...)
+	}
+}
+
+func TestGossipTopology(t *testing.T) {
 	// Create some peers that will talk to each other
 	r1 := NewTestRouter("01:00:00:01:00:00")
 	r2 := NewTestRouter("02:00:00:02:00:00")
 	r3 := NewTestRouter("03:00:00:03:00:00")
-
+	routers := []*Router{r1, r2, r3}
 	// Check state when they have no connections
 	checkTopology(t, r1, r1.tp())
 	checkTopology(t, r2, r2.tp())
 
 	// Now try adding some connections
-	r1.AddTestChannelConnection(r2)
-	sendPendingGossip(r1, r2)
-	checkTopology(t, r1, r1.tp(r2), r2.tp())
-	checkTopology(t, r2, r1.tp(r2), r2.tp())
-	r2.AddTestChannelConnection(r1)
+	AddTestGossipConnection(r1, r2)
 	sendPendingGossip(r1, r2)
 	checkTopology(t, r1, r1.tp(r2), r2.tp(r1))
 	checkTopology(t, r2, r1.tp(r2), r2.tp(r1))
 
-	// Currently, the connection from 2 to 3 is one-way only
-	r2.AddTestChannelConnection(r3)
-	sendPendingGossip(r1, r2, r3)
-	checkTopology(t, r1, r1.tp(r2), r2.tp(r1, r3), r3.tp())
-	checkTopology(t, r2, r1.tp(r2), r2.tp(r1, r3), r3.tp())
-	// When r2 gossiped to r3, 1 was unreachable from r3 so it got removed from the
-	// list of peers, but remains referenced in the connection from 1 to 3.
-	checkTopology(t, r3, r2.tp(r1, r3), r3.tp())
+	AddTestGossipConnection(r2, r3)
+	flushAndCheckTopology(t, routers, r1.tp(r2), r2.tp(r1, r3), r3.tp(r2))
 
-	// Add a connection from 3 to 1 and now r1 is reachable.
-	r3.AddTestChannelConnection(r1)
-	sendPendingGossip(r1, r2, r3)
-	checkTopology(t, r1, r1.tp(r2), r2.tp(r1, r3), r3.tp(r1))
-	checkTopology(t, r2, r1.tp(r2), r2.tp(r1, r3), r3.tp(r1))
-	checkTopology(t, r3, r1.tp(), r2.tp(r1, r3), r3.tp(r1))
-
-	r1.AddTestChannelConnection(r3)
-	sendPendingGossip(r1, r2, r3)
-	checkTopology(t, r1, r1.tp(r2, r3), r2.tp(r1, r3), r3.tp(r1))
-	checkTopology(t, r2, r1.tp(r2, r3), r2.tp(r1, r3), r3.tp(r1))
-	checkTopology(t, r3, r1.tp(r2, r3), r2.tp(r1, r3), r3.tp(r1))
+	AddTestGossipConnection(r3, r1)
+	flushAndCheckTopology(t, routers, r1.tp(r2, r3), r2.tp(r1, r3), r3.tp(r1, r2))
 
 	// Drop the connection from 2 to 3
-	r2.DeleteTestChannelConnection(r3)
-	sendPendingGossip(r1, r2, r3)
-	checkTopology(t, r1, r1.tp(r2, r3), r2.tp(r1), r3.tp(r1))
-	checkTopology(t, r2, r1.tp(r2, r3), r2.tp(r1))
-	checkTopology(t, r3, r1.tp(r2, r3), r2.tp(r1), r3.tp(r1))
+	r2.DeleteTestGossipConnection(r3)
+	flushAndCheckTopology(t, routers, r1.tp(r2, r3), r2.tp(r1), r3.tp(r1, r2))
 
 	// Drop the connection from 1 to 3
-	r1.DeleteTestChannelConnection(r3)
+	r1.DeleteTestGossipConnection(r3)
 	sendPendingGossip(r1, r2, r3)
-	checkTopology(t, r1, r1.tp(r2), r2.tp(r1), r3.tp(r1))
-
-	checkTopology(t, r1, r1.tp(r2), r2.tp(r1), r3.tp(r1))
+	checkTopology(t, r1, r1.tp(r2), r2.tp(r1))
 	checkTopology(t, r2, r1.tp(r2), r2.tp(r1))
 	// r3 still thinks r1 has a connection to it
-	checkTopology(t, r3, r1.tp(r2, r3), r2.tp(r1), r3.tp(r1))
-
-	// On a timer, r3 will gossip to r1
-	r3.SendAllGossip()
-	sendPendingGossip(r1, r2, r3)
-	checkTopology(t, r1, r1.tp(r2), r2.tp(r1), r3.tp(r1))
+	checkTopology(t, r3, r1.tp(r2, r3), r2.tp(r1), r3.tp(r1, r2))
 }
 
 func TestGossipSurrogate(t *testing.T) {
@@ -168,14 +140,10 @@ func TestGossipSurrogate(t *testing.T) {
 	r1 := NewTestRouter("01:00:00:01:00:00")
 	r2 := NewTestRouter("02:00:00:02:00:00")
 	r3 := NewTestRouter("03:00:00:03:00:00")
-	r1.AddTestChannelConnection(r2)
-	r2.AddTestChannelConnection(r1)
-	r3.AddTestChannelConnection(r2)
-	r2.AddTestChannelConnection(r3)
-	sendPendingGossip(r1, r2, r3)
-	checkTopology(t, r1, r1.tp(r2), r2.tp(r1, r3), r3.tp(r2))
-	checkTopology(t, r2, r1.tp(r2), r2.tp(r1, r3), r3.tp(r2))
-	checkTopology(t, r3, r1.tp(r2), r2.tp(r1, r3), r3.tp(r2))
+	routers := []*Router{r1, r2, r3}
+	AddTestGossipConnection(r1, r2)
+	AddTestGossipConnection(r3, r2)
+	flushAndCheckTopology(t, routers, r1.tp(r2), r2.tp(r1, r3), r3.tp(r2))
 
 	// create a gossiper at either end, but not the middle
 	g1 := newTestGossiper()
@@ -241,10 +209,10 @@ func (g *testGossiper) OnGossip(update []byte) (GossipData, error) {
 			g.state[v] = void
 		}
 	}
-	if len(delta) > 0 {
-		return NewSurrogateGossipData(delta), nil
+	if len(delta) == 0 {
+		return nil, nil
 	}
-	return nil, nil
+	return NewSurrogateGossipData(delta), nil
 }
 
 func (g *testGossiper) checkHas(t *testing.T, vs ...byte) {

mesh/local_peer.go

@@ -1,4 +1,4 @@
-package router
+package mesh
 
 import (
 	"fmt"
@@ -19,7 +19,7 @@ type LocalPeerAction func()
 func NewLocalPeer(name PeerName, nickName string, router *Router) *LocalPeer {
 	actionChan := make(chan LocalPeerAction, ChannelSize)
 	peer := &LocalPeer{
-		Peer:       NewPeer(name, nickName, 0, 0),
+		Peer:       NewPeer(name, nickName, randomPeerUID(), 0, randomPeerShortID()),
 		router:     router,
 		actionChan: actionChan,
 	}
@@ -27,56 +27,6 @@ func NewLocalPeer(name PeerName, nickName string, router *Router) *LocalPeer {
 	return peer
 }
 
-func (peer *LocalPeer) Forward(dstPeer *Peer, frame []byte, dec *EthernetDecoder) error {
-	return peer.Relay(peer.Peer, dstPeer, frame, dec)
-}
-
-func (peer *LocalPeer) Broadcast(frame []byte, dec *EthernetDecoder) {
-	peer.RelayBroadcast(peer.Peer, frame, dec)
-}
-
-func (peer *LocalPeer) Relay(srcPeer, dstPeer *Peer, frame []byte, dec *EthernetDecoder) error {
-	relayPeerName, found := peer.router.Routes.Unicast(dstPeer.Name)
-	if !found {
-		// Not necessarily an error as there could be a race with the
-		// dst disappearing whilst the frame is in flight
-		log.Println("Received packet for unknown destination:", dstPeer)
-		return nil
-	}
-	conn, found := peer.ConnectionTo(relayPeerName)
-	if !found {
-		// Again, could just be a race, not necessarily an error
-		log.Println("Unable to find connection to relay peer", relayPeerName)
-		return nil
-	}
-	return conn.(*LocalConnection).Forward(&ForwardedFrame{
-		srcPeer: srcPeer,
-		dstPeer: dstPeer,
-		frame:   frame},
-		dec)
-}
-
-func (peer *LocalPeer) RelayBroadcast(srcPeer *Peer, frame []byte, dec *EthernetDecoder) {
-	nextHops := peer.router.Routes.Broadcast(srcPeer.Name)
-	if len(nextHops) == 0 {
-		return
-	}
-	for _, conn := range peer.ConnectionsTo(nextHops) {
-		err := conn.(*LocalConnection).Forward(&ForwardedFrame{
-			srcPeer: srcPeer,
-			dstPeer: conn.Remote(),
-			frame:   frame},
-			dec)
-		if err != nil {
-			if ftbe, ok := err.(FrameTooBigError); ok {
-				log.Warningf("dropping too big DF broadcast frame (%v -> %v): PMTU= %v", dec.IP.DstIP, dec.IP.SrcIP, ftbe.EPMTU)
-			} else {
-				log.Errorln(err)
-			}
-		}
-	}
-}
-
 func (peer *LocalPeer) Connections() ConnectionSet {
 	connections := make(ConnectionSet)
 	peer.RLock()
@@ -112,21 +62,16 @@ func (peer *LocalPeer) CreateConnection(peerAddr string, acceptNewPeer bool) err
 	if err := peer.checkConnectionLimit(); err != nil {
 		return err
 	}
-	tcpAddr, tcpErr := net.ResolveTCPAddr("tcp4", peerAddr)
-	udpAddr, udpErr := net.ResolveUDPAddr("udp4", peerAddr)
-	if tcpErr != nil || udpErr != nil {
-		// they really should have the same value, but just in case...
-		if tcpErr == nil {
-			return udpErr
-		}
-		return tcpErr
+	tcpAddr, err := net.ResolveTCPAddr("tcp4", peerAddr)
+	if err != nil {
+		return err
 	}
 	tcpConn, err := net.DialTCP("tcp4", nil, tcpAddr)
 	if err != nil {
 		return err
 	}
 	connRemote := NewRemoteConnection(peer.Peer, nil, tcpConn.RemoteAddr().String(), true, false)
-	StartLocalConnection(connRemote, tcpConn, udpAddr, peer.router, acceptNewPeer)
+	StartLocalConnection(connRemote, tcpConn, peer.router, acceptNewPeer)
 	return nil
 }
 
@@ -210,7 +155,10 @@ func (peer *LocalPeer) handleAddConnection(conn Connection) error {
 		conn.Log("connection added (new peer)")
 		peer.router.SendAllGossipDown(conn)
 	}
+
+	peer.router.Routes.Recalculate()
 	peer.broadcastPeerUpdate(conn.Remote())
+
 	return nil
 }
 
@@ -224,6 +172,8 @@ func (peer *LocalPeer) handleConnectionEstablished(conn Connection) {
 	}
 	peer.connectionEstablished(conn)
 	conn.Log("connection fully established")
+
+	peer.router.Routes.Recalculate()
 	peer.broadcastPeerUpdate()
 }
 
@@ -243,14 +193,20 @@ func (peer *LocalPeer) handleDeleteConnection(conn Connection) {
 	// Must do garbage collection first to ensure we don't send out an
 	// update with unreachable peers (can cause looping)
 	peer.router.Peers.GarbageCollect()
+	peer.router.Routes.Recalculate()
 	peer.broadcastPeerUpdate()
 }
 
 // helpers
 
 func (peer *LocalPeer) broadcastPeerUpdate(peers ...*Peer) {
-	peer.router.Routes.Recalculate()
-	peer.router.TopologyGossip.GossipBroadcast(NewTopologyGossipData(peer.router.Peers, append(peers, peer.Peer)...))
+	// Some tests run without a router.  This should be fixed so
+	// that the relevant part of Router can be easily run in the
+	// context of a test, but that will involve significant
+	// reworking of tests.
+	if peer.router != nil {
+		peer.router.BroadcastTopologyUpdate(append(peers, peer.Peer))
+	}
 }
 
 func (peer *LocalPeer) checkConnectionLimit() error {
@@ -265,20 +221,20 @@ func (peer *LocalPeer) addConnection(conn Connection) {
 	peer.Lock()
 	defer peer.Unlock()
 	peer.connections[conn.Remote().Name] = conn
-	peer.version++
+	peer.Version++
 }
 
 func (peer *LocalPeer) deleteConnection(conn Connection) {
 	peer.Lock()
 	defer peer.Unlock()
 	delete(peer.connections, conn.Remote().Name)
-	peer.version++
+	peer.Version++
 }
 
 func (peer *LocalPeer) connectionEstablished(conn Connection) {
 	peer.Lock()
 	defer peer.Unlock()
-	peer.version++
+	peer.Version++
 }
 
 func (peer *LocalPeer) connectionCount() int {
@@ -286,3 +242,20 @@ func (peer *LocalPeer) connectionCount() int {
 	defer peer.RUnlock()
 	return len(peer.connections)
 }
+
+func (peer *LocalPeer) setShortID(shortID PeerShortID) {
+	peer.Lock()
+	defer peer.Unlock()
+	peer.ShortID = shortID
+	peer.Version++
+}
+
+func (peer *LocalPeer) setVersionBeyond(version uint64) bool {
+	peer.Lock()
+	defer peer.Unlock()
+	if version >= peer.Version {
+		peer.Version = version + 1
+		return true
+	}
+	return false
+}

mesh/mocks_test.go

@@ -3,7 +3,7 @@
 // It supplies some mock implementations to other unit tests, and is
 // named "...test.go" so it is only compiled under `go test`.
 
-package router
+package mesh
 
 import (
 	"fmt"
@@ -14,8 +14,10 @@ import (
 
 // Add to peers a connection from peers.ourself to p
 func (peers *Peers) AddTestConnection(p *Peer) {
-	toPeer := NewPeer(p.Name, "", p.UID, 0)
-	peers.FetchWithDefault(toPeer) // Has side-effect of incrementing refcount
+	summary := p.PeerSummary
+	summary.Version = 0
+	toPeer := NewPeerFromSummary(summary)
+	toPeer = peers.FetchWithDefault(toPeer) // Has side-effect of incrementing refcount
 	conn := newMockConnection(peers.ourself.Peer, toPeer)
 	peers.ourself.addConnection(conn)
 	peers.ourself.connectionEstablished(conn)
@@ -23,11 +25,9 @@ func (peers *Peers) AddTestConnection(p *Peer) {
 
 // Add to peers a connection from p1 to p2
 func (peers *Peers) AddTestRemoteConnection(p1, p2 *Peer) {
-	fromName := p1.Name
-	fromPeer := NewPeer(fromName, "", p1.UID, 0)
+	fromPeer := NewPeerFrom(p1)
 	fromPeer = peers.FetchWithDefault(fromPeer)
-	toName := p2.Name
-	toPeer := NewPeer(toName, "", p2.UID, 0)
+	toPeer := NewPeerFrom(p2)
 	toPeer = peers.FetchWithDefault(toPeer)
 	peers.ourself.addConnection(&RemoteConnection{fromPeer, toPeer, "", false, false})
 }
@@ -70,7 +70,7 @@ func checkEqualConns(t *testing.T, ourName PeerName, got, wanted map[PeerName]Co
 // Get all the peers from a Peers in a slice
 func (peers *Peers) allPeers() []*Peer {
 	var res []*Peer
-	for _, peer := range peers.table {
+	for _, peer := range peers.byName {
 		res = append(res, peer)
 	}
 	return res

mesh/overlay.go

@@ -0,0 +1,116 @@
+package mesh
+
+import (
+	"net"
+)
+
+type Overlay interface {
+	// Enhance a features map with overlay-related features
+	AddFeaturesTo(map[string]string)
+
+	// Prepare on overlay connection. The connection should remain
+	// passive until it has been Confirm()ed.
+	PrepareConnection(OverlayConnectionParams) (OverlayConnection, error)
+
+	// Obtain diagnostic information specific to the overlay
+	Diagnostics() interface{}
+}
+
+type OverlayConnectionParams struct {
+	RemotePeer *Peer
+
+	// The local address of the corresponding TCP connection. Used to
+	// derive the local IP address for sending. May differ for
+	// different overlay connections.
+	LocalAddr *net.TCPAddr
+
+	// The remote address of the corresponding TCP connection. Used to
+	// determine the address to send to, but only if the TCP
+	// connection is outbound. Otherwise the Overlay needs to discover
+	// it (e.g. from incoming datagrams).
+	RemoteAddr *net.TCPAddr
+
+	// Is the corresponding TCP connection outbound?
+	Outbound bool
+
+	// Unique identifier for this connection
+	ConnUID uint64
+
+	// Session key, if connection is encrypted; nil otherwise.
+	//
+	// NB: overlay connections must take care not to use nonces which
+	// may collide with those of the main connection. These nonces are
+	// 192 bits, with the top most bit unspecified, the next bit set
+	// to 1, followed by 126 zero bits, and a message sequence number
+	// in the lowest 64 bits.
+	SessionKey *[32]byte
+
+	// Function to send a control message to the counterpart
+	// overlay connection.
+	SendControlMessage func(tag byte, msg []byte) error
+
+	// Features passed at connection initiation
+	Features map[string]string
+}
+
+// All of the machinery to manage overlay connectivity to a particular
+// peer
+type OverlayConnection interface {
+	// Confirm that the connection is really wanted, and so the
+	// Overlay should begin heartbeats etc. to verify the operation of
+	// the overlay connection.
+	Confirm()
+
+	// A channel indicating that the overlay connection is
+	// established, i.e. its operation has been confirmed.
+	EstablishedChannel() <-chan struct{}
+
+	// A channel indicating an error from the overlay connection.  The
+	// overlay connection is not expected to be operational after the
+	// first error, so the channel only needs to buffer a single
+	// error.
+	ErrorChannel() <-chan error
+
+	Stop()
+
+	// Handle a message from the peer.  'tag' exists for
+	// compatibility, and should always be
+	// ProtocolOverlayControlMessage for non-sleeve overlays.
+	ControlMessage(tag byte, msg []byte)
+
+	// User facing overlay name
+	DisplayName() string
+}
+
+type NullOverlay struct{}
+
+func (NullOverlay) AddFeaturesTo(map[string]string) {
+}
+
+func (NullOverlay) PrepareConnection(OverlayConnectionParams) (OverlayConnection, error) {
+	return NullOverlay{}, nil
+}
+
+func (NullOverlay) Diagnostics() interface{} {
+	return nil
+}
+func (NullOverlay) Confirm() {
+}
+
+func (NullOverlay) EstablishedChannel() <-chan struct{} {
+	return nil
+}
+
+func (NullOverlay) ErrorChannel() <-chan error {
+	return nil
+}
+
+func (NullOverlay) Stop() {
+}
+
+func (NullOverlay) ControlMessage(byte, []byte) {
+}
+
+func (NullOverlay) DisplayName() string {
+	return "null"
+}

mesh/peer.go

@@ -1,41 +1,111 @@
-package router
+package mesh
 
 import (
+	"crypto/rand"
+	"encoding/binary"
 	"fmt"
 	"sort"
 	"strconv"
 )
 
+func randBytes(n int) []byte {
+	buf := make([]byte, n)
+	_, err := rand.Read(buf)
+	checkFatal(err)
+	return buf
+}
+
+func randUint64() (r uint64) {
+	return binary.LittleEndian.Uint64(randBytes(8))
+}
+
+func randUint16() (r uint16) {
+	return binary.LittleEndian.Uint16(randBytes(2))
+}
+
 type PeerUID uint64
 
+func randomPeerUID() PeerUID {
+	for {
+		uid := randUint64()
+		if uid != 0 { // uid 0 is reserved for peer placeholder
+			return PeerUID(uid)
+		}
+	}
+}
+
 func ParsePeerUID(s string) (PeerUID, error) {
 	uid, err := strconv.ParseUint(s, 10, 64)
 	return PeerUID(uid), err
 }
 
+// Short IDs exist for the sake of fast datapath.  They are 12 bits,
+// randomly assigned, but we detect and recover from collisions.  This
+// does limit us to 4096 peers, but that should be sufficient for a
+// while.
+type PeerShortID uint16
+
+const PeerShortIDBits = 12
+
+func randomPeerShortID() PeerShortID {
+	return PeerShortID(randUint16() & (1<<PeerShortIDBits - 1))
+}
+
+type PeerSummary struct {
+	NameByte   []byte
+	NickName   string
+	UID        PeerUID
+	Version    uint64
+	ShortID    PeerShortID
+	HasShortID bool
+}
+
 type Peer struct {
-	Name          PeerName
-	NameByte      []byte
-	NickName      string
-	UID           PeerUID
-	version       uint64
+	Name PeerName
+	PeerSummary
 	localRefCount uint64 // maintained by Peers
 	connections   map[PeerName]Connection
 }
 
+type ListOfPeers []*Peer
+
+func (lop ListOfPeers) Len() int {
+	return len(lop)
+}
+func (lop ListOfPeers) Swap(i, j int) {
+	lop[i], lop[j] = lop[j], lop[i]
+}
+func (lop ListOfPeers) Less(i, j int) bool {
+	return lop[i].Name < lop[j].Name
+}
+
 type ConnectionSet map[Connection]struct{}
 
-func NewPeer(name PeerName, nickName string, uid PeerUID, version uint64) *Peer {
-	if uid == 0 {
-		uid = PeerUID(randUint64())
-	}
+func NewPeerFromSummary(summary PeerSummary) *Peer {
 	return &Peer{
-		Name:        name,
-		NameByte:    name.Bin(),
-		NickName:    nickName,
-		UID:         uid,
-		version:     version,
-		connections: make(map[PeerName]Connection)}
+		Name:        PeerNameFromBin(summary.NameByte),
+		PeerSummary: summary,
+		connections: make(map[PeerName]Connection),
+	}
+}
+
+func NewPeer(name PeerName, nickName string, uid PeerUID, version uint64, shortID PeerShortID) *Peer {
+	return NewPeerFromSummary(PeerSummary{
+		NameByte:   name.Bin(),
+		NickName:   nickName,
+		UID:        uid,
+		Version:    version,
+		ShortID:    shortID,
+		HasShortID: true,
+	})
+}
+
+func NewPeerPlaceholder(name PeerName) *Peer {
+	return NewPeerFromSummary(PeerSummary{NameByte: name.Bin()})
+}
+
+func NewPeerFrom(peer *Peer) *Peer {
+	return NewPeerFromSummary(peer.PeerSummary)
 }
 
 func (peer *Peer) String() string {
@@ -66,7 +136,7 @@ func (peer *Peer) String() string {
 // NB: This function should generally be invoked while holding a read
 // lock on Peers and LocalPeer.
 func (peer *Peer) Routes(stopAt *Peer, establishedAndSymmetric bool) (bool, map[PeerName]PeerName) {
-	routes := make(map[PeerName]PeerName)
+	routes := make(unicastRoutes)
 	routes[peer.Name] = UnknownPeerName
 	nextWorklist := []*Peer{peer}
 	for len(nextWorklist) > 0 {

mesh/peer_name_hash.go

@@ -3,7 +3,7 @@
 // Let peer names be sha256 hashes...of anything (as long as it's
 // unique).
 
-package router
+package mesh
 
 import (
 	"crypto/sha256"

mesh/peer_name_mac.go

@@ -15,7 +15,7 @@
 // name. In particular it doesn't actually have to be the MAC of, say,
 // the network interface the peer is sniffing on.
 
-package router
+package mesh
 
 import (
 	"net"
@@ -56,3 +56,20 @@ func (name PeerName) Bin() []byte {
 func (name PeerName) String() string {
 	return intmac(uint64(name)).String()
 }
+
+func macint(mac net.HardwareAddr) (r uint64) {
+	for _, b := range mac {
+		r <<= 8
+		r |= uint64(b)
+	}
+	return
+}
+
+func intmac(key uint64) (r net.HardwareAddr) {
+	r = make([]byte, 6)
+	for i := 5; i >= 0; i-- {
+		r[i] = byte(key)
+		key >>= 8
+	}
+	return
+}

mesh/peers.go

@@ -0,0 +1,530 @@
+package mesh
+
+import (
+	"bytes"
+	"encoding/gob"
+	"io"
+	"math/rand"
+	"sync"
+)
+
+var void = struct{}{}
+
+type Peers struct {
+	sync.RWMutex
+	ourself   *LocalPeer
+	byName    map[PeerName]*Peer
+	byShortID map[PeerShortID]ShortIDPeers
+	onGC      []func(*Peer)
+
+	// Called when the mapping from short ids to peers changes
+	onInvalidateShortIDs []func()
+}
+
+type ShortIDPeers struct {
+	// If we know about a single peer with the short id, this is
+	// that peer.  If there is a collision, this is the peer with
+	// the lowest Name.
+	peer *Peer
+
+	// In case of a collision, this holds the other peers.
+	others []*Peer
+}
+
+type PeerNameSet map[PeerName]struct{}
+
+type ConnectionSummary struct {
+	NameByte      []byte
+	RemoteTCPAddr string
+	Outbound      bool
+	Established   bool
+}
+
+// Pending notifications due to changes to Peers that need to be sent
+// out once the Peers is unlocked.
+type PeersPendingNotifications struct {
+	// Peers that have been GCed
+	removed []*Peer
+
+	// The mapping from shorts ids to peers changed
+	invalidateShortIDs bool
+
+	// The local short ID needs reassigning due to a collision
+	reassignLocalShortID bool
+
+	// The local peer was modified
+	localPeerModified bool
+}
+
+func NewPeers(ourself *LocalPeer) *Peers {
+	peers := &Peers{
+		ourself:   ourself,
+		byName:    make(map[PeerName]*Peer),
+		byShortID: make(map[PeerShortID]ShortIDPeers),
+	}
+	peers.FetchWithDefault(ourself.Peer)
+	return peers
+}
+
+func (peers *Peers) OnGC(callback func(*Peer)) {
+	peers.Lock()
+	defer peers.Unlock()
+
+	// Although the array underlying peers.onGC might be accessed
+	// without holding the lock in unlockAndNotify, we don't
+	// support removing callbacks, so a simple append here is
+	// safe.
+	peers.onGC = append(peers.onGC, callback)
+}
+
+func (peers *Peers) OnInvalidateShortIDs(callback func()) {
+	peers.Lock()
+	defer peers.Unlock()
+
+	// Safe, as in OnGC
+	peers.onInvalidateShortIDs = append(peers.onInvalidateShortIDs, callback)
+}
+
+func (peers *Peers) unlockAndNotify(pending *PeersPendingNotifications) {
+	broadcastLocalPeer := (pending.reassignLocalShortID && peers.reassignLocalShortID(pending)) ||
+		pending.localPeerModified
+	onGC := peers.onGC
+	onInvalidateShortIDs := peers.onInvalidateShortIDs
+	peers.Unlock()
+
+	if pending.removed != nil {
+		for _, callback := range onGC {
+			for _, peer := range pending.removed {
+				callback(peer)
+			}
+		}
+	}
+
+	if pending.invalidateShortIDs {
+		for _, callback := range onInvalidateShortIDs {
+			callback()
+		}
+	}
+
+	if broadcastLocalPeer {
+		peers.ourself.broadcastPeerUpdate()
+	}
+}
+
+func (peers *Peers) addByShortID(peer *Peer, pending *PeersPendingNotifications) {
+	if !peer.HasShortID {
+		return
+	}
+
+	entry, ok := peers.byShortID[peer.ShortID]
+	if !ok {
+		entry = ShortIDPeers{peer: peer}
+	} else if entry.peer == nil {
+		// This short ID is free, but was used in the past.
+		// Because we are reusing it, it's an invalidation
+		// event.
+		entry.peer = peer
+		pending.invalidateShortIDs = true
+	} else if peer.Name < entry.peer.Name {
+		// Short ID collision, this peer becomes the principal
+		// peer for the short ID, bumping the previous one
+		// into others.
+
+		if entry.peer == peers.ourself.Peer {
+			// The bumped peer is peers.ourself, so we
+			// need to look for a new short id
+			pending.reassignLocalShortID = true
+		}
+
+		entry.others = append(entry.others, entry.peer)
+		entry.peer = peer
+		pending.invalidateShortIDs = true
+	} else {
+		// Short ID collision, this peer is secondary
+		entry.others = append(entry.others, peer)
+	}
+
+	peers.byShortID[peer.ShortID] = entry
+}
+
+func (peers *Peers) deleteByShortID(peer *Peer, pending *PeersPendingNotifications) {
+	if !peer.HasShortID {
+		return
+	}
+
+	entry := peers.byShortID[peer.ShortID]
+	var otherIndex int
+
+	if peer != entry.peer {
+		// peer is secondary, find its index in others
+		otherIndex = -1
+
+		for i, other := range entry.others {
+			if peer == other {
+				otherIndex = i
+				break
+			}
+		}
+
+		if otherIndex < 0 {
+			return
+		}
+	} else if len(entry.others) != 0 {
+		// need to find the peer with the lowest name to
+		// become the new principal one
+		otherIndex = 0
+		minName := entry.others[0].Name
+
+		for i := 1; i < len(entry.others); i++ {
+			otherName := entry.others[i].Name
+			if otherName < minName {
+				minName = otherName
+				otherIndex = i
+			}
+		}
+
+		entry.peer = entry.others[otherIndex]
+		pending.invalidateShortIDs = true
+	} else {
+		// This is the last peer with the short id.  We clear
+		// the entry, don't delete it, in order to detect when
+		// it gets re-used.
+		peers.byShortID[peer.ShortID] = ShortIDPeers{}
+		return
+	}
+
+	entry.others[otherIndex] = entry.others[len(entry.others)-1]
+	entry.others = entry.others[:len(entry.others)-1]
+	peers.byShortID[peer.ShortID] = entry
+}
+
+func (peers *Peers) reassignLocalShortID(pending *PeersPendingNotifications) bool {
+	newShortID, ok := peers.chooseShortID()
+	if ok {
+		peers.setLocalShortID(newShortID, pending)
+		return true
+	}
+
+	// Otherwise we'll try again later on in garbageColleect
+	return false
+}
+
+func (peers *Peers) setLocalShortID(newShortID PeerShortID, pending *PeersPendingNotifications) {
+	peers.deleteByShortID(peers.ourself.Peer, pending)
+	peers.ourself.setShortID(newShortID)
+	peers.addByShortID(peers.ourself.Peer, pending)
+}
+
+// Choose an available short id at random
+func (peers *Peers) chooseShortID() (PeerShortID, bool) {
+	rng := rand.New(rand.NewSource(int64(randUint64())))
+
+	// First, just try picking some short ids at random, and
+	// seeing if they are available:
+	for i := 0; i < 10; i++ {
+		shortID := PeerShortID(rng.Intn(1 << PeerShortIDBits))
+		if peers.byShortID[shortID].peer == nil {
+			return shortID, true
+		}
+	}
+
+	// Looks like most short ids are used.  So count the number of
+	// unused ones, and pick one at random.
+	available := int(1 << PeerShortIDBits)
+	for _, entry := range peers.byShortID {
+		if entry.peer != nil {
+			available--
+		}
+	}
+
+	if available == 0 {
+		// All short ids are used.
+		return 0, false
+	}
+
+	n := rng.Intn(available)
+	var i PeerShortID
+	for {
+		if peers.byShortID[i].peer == nil {
+			if n == 0 {
+				return PeerShortID(i), true
+			}
+
+			n--
+		}
+
+		i++
+	}
+}
+
+func (peers *Peers) FetchWithDefault(peer *Peer) *Peer {
+	peers.Lock()
+	var pending PeersPendingNotifications
+	defer peers.unlockAndNotify(&pending)
+
+	if existingPeer, found := peers.byName[peer.Name]; found {
+		existingPeer.localRefCount++
+		return existingPeer
+	}
+
+	peers.byName[peer.Name] = peer
+	peers.addByShortID(peer, &pending)
+	peer.localRefCount++
+	return peer
+}
+
+func (peers *Peers) Fetch(name PeerName) *Peer {
+	peers.RLock()
+	defer peers.RUnlock()
+	return peers.byName[name]
+}
+
+func (peers *Peers) FetchAndAddRef(name PeerName) *Peer {
+	peers.Lock()
+	defer peers.Unlock()
+	peer := peers.byName[name]
+	if peer != nil {
+		peer.localRefCount++
+	}
+	return peer
+}
+
+func (peers *Peers) FetchByShortID(shortID PeerShortID) *Peer {
+	peers.RLock()
+	defer peers.RUnlock()
+	return peers.byShortID[shortID].peer
+}
+
+func (peers *Peers) Dereference(peer *Peer) {
+	peers.Lock()
+	defer peers.Unlock()
+	peer.localRefCount--
+}
+
+func (peers *Peers) ForEach(fun func(*Peer)) {
+	peers.RLock()
+	defer peers.RUnlock()
+	for _, peer := range peers.byName {
+		fun(peer)
+	}
+}
+
+// Merge an incoming update with our own topology.
+//
+// We add peers hitherto unknown to us, and update peers for which the
+// update contains a more recent version than known to us. The return
+// value is a) a representation of the received update, and b) an
+// "improved" update containing just these new/updated elements.
+func (peers *Peers) ApplyUpdate(update []byte) (PeerNameSet, PeerNameSet, error) {
+	peers.Lock()
+	var pending PeersPendingNotifications
+	defer peers.unlockAndNotify(&pending)
+
+	newPeers, decodedUpdate, decodedConns, err := peers.decodeUpdate(update)
+	if err != nil {
+		return nil, nil, err
+	}
+
+	// Add new peers
+	for name, newPeer := range newPeers {
+		peers.byName[name] = newPeer
+		peers.addByShortID(newPeer, &pending)
+	}
+
+	// Now apply the updates
+	newUpdate := peers.applyUpdate(decodedUpdate, decodedConns, &pending)
+	peers.garbageCollect(&pending)
+	for _, peerRemoved := range pending.removed {
+		delete(newUpdate, peerRemoved.Name)
+	}
+
+	updateNames := make(PeerNameSet)
+	for _, peer := range decodedUpdate {
+		updateNames[peer.Name] = void
+	}
+
+	return updateNames, newUpdate, nil
+}
+
+func (peers *Peers) Names() PeerNameSet {
+	peers.RLock()
+	defer peers.RUnlock()
+
+	names := make(PeerNameSet)
+	for name := range peers.byName {
+		names[name] = void
+	}
+	return names
+}
+
+func (peers *Peers) EncodePeers(names PeerNameSet) []byte {
+	buf := new(bytes.Buffer)
+	enc := gob.NewEncoder(buf)
+	peers.RLock()
+	defer peers.RUnlock()
+	for name := range names {
+		if peer, found := peers.byName[name]; found {
+			if peer == peers.ourself.Peer {
+				peers.ourself.Encode(enc)
+			} else {
+				peer.Encode(enc)
+			}
+		}
+	}
+	return buf.Bytes()
+}
+
+func (peers *Peers) GarbageCollect() {
+	peers.Lock()
+	var pending PeersPendingNotifications
+	defer peers.unlockAndNotify(&pending)
+
+	peers.garbageCollect(&pending)
+}
+
+func (peers *Peers) garbageCollect(pending *PeersPendingNotifications) {
+	peers.ourself.RLock()
+	_, reached := peers.ourself.Routes(nil, false)
+	peers.ourself.RUnlock()
+
+	for name, peer := range peers.byName {
+		if _, found := reached[peer.Name]; !found && peer.localRefCount == 0 {
+			delete(peers.byName, name)
+			peers.deleteByShortID(peer, pending)
+			pending.removed = append(pending.removed, peer)
+		}
+	}
+
+	if peers.byShortID[peers.ourself.ShortID].peer != peers.ourself.Peer {
+		// The local peer doesn't own its short id.  Garbage
+		// collection might have freed some up, so try to
+		// reassign.
+		pending.reassignLocalShortID = true
+	}
+}
+
+func (peers *Peers) decodeUpdate(update []byte) (newPeers map[PeerName]*Peer, decodedUpdate []*Peer, decodedConns [][]ConnectionSummary, err error) {
+	newPeers = make(map[PeerName]*Peer)
+	decodedUpdate = []*Peer{}
+	decodedConns = [][]ConnectionSummary{}
+
+	decoder := gob.NewDecoder(bytes.NewReader(update))
+
+	for {
+		peerSummary, connSummaries, decErr := decodePeer(decoder)
+		if decErr == io.EOF {
+			break
+		} else if decErr != nil {
+			err = decErr
+			return
+		}
+		newPeer := NewPeerFromSummary(peerSummary)
+		decodedUpdate = append(decodedUpdate, newPeer)
+		decodedConns = append(decodedConns, connSummaries)
+		if _, found := peers.byName[newPeer.Name]; !found {
+			newPeers[newPeer.Name] = newPeer
+		}
+	}
+
+	for _, connSummaries := range decodedConns {
+		for _, connSummary := range connSummaries {
+			remoteName := PeerNameFromBin(connSummary.NameByte)
+			if _, found := newPeers[remoteName]; found {
+				continue
+			}
+			if _, found := peers.byName[remoteName]; found {
+				continue
+			}
+			// Update refers to a peer which we have no knowledge of.
+			newPeers[remoteName] = NewPeerPlaceholder(remoteName)
+		}
+	}
+	return
+}
+
+func (peers *Peers) applyUpdate(decodedUpdate []*Peer, decodedConns [][]ConnectionSummary, pending *PeersPendingNotifications) PeerNameSet {
+	newUpdate := make(PeerNameSet)
+	for idx, newPeer := range decodedUpdate {
+		connSummaries := decodedConns[idx]
+		name := newPeer.Name
+		// guaranteed to find peer in the peers.byName
+		switch peer := peers.byName[name]; peer {
+		case peers.ourself.Peer:
+			if newPeer.UID != peer.UID {
+				// The update contains information about an old
+				// incarnation of ourselves. We increase our version
+				// number beyond that which we received, so our
+				// information supersedes the old one when it is
+				// received by other peers.
+				pending.localPeerModified = peers.ourself.setVersionBeyond(newPeer.Version)
+			}
+		case newPeer:
+			peer.connections = makeConnsMap(peer, connSummaries, peers.byName)
+			newUpdate[name] = void
+		default: // existing peer
+			if newPeer.Version < peer.Version ||
+				(newPeer.Version == peer.Version &&
+					(newPeer.UID < peer.UID ||
+						(newPeer.UID == peer.UID &&
+							(!newPeer.HasShortID || peer.HasShortID)))) {
+				continue
+			}
+			peer.Version = newPeer.Version
+			peer.UID = newPeer.UID
+			peer.NickName = newPeer.NickName
+			peer.connections = makeConnsMap(peer, connSummaries, peers.byName)
+
+			if newPeer.ShortID != peer.ShortID || newPeer.HasShortID != peer.HasShortID {
+				peers.deleteByShortID(peer, pending)
+				peer.ShortID = newPeer.ShortID
+				peer.HasShortID = newPeer.HasShortID
+				peers.addByShortID(peer, pending)
+			}
+			newUpdate[name] = void
+		}
+	}
+	return newUpdate
+}
+
+func (peer *Peer) Encode(enc *gob.Encoder) {
+	checkFatal(enc.Encode(peer.PeerSummary))
+
+	connSummaries := []ConnectionSummary{}
+	for _, conn := range peer.connections {
+		connSummaries = append(connSummaries, ConnectionSummary{
+			conn.Remote().NameByte,
+			conn.RemoteTCPAddr(),
+			conn.Outbound(),
+			conn.Established(),
+		})
+	}
+
+	checkFatal(enc.Encode(connSummaries))
+}
+
+func (peer *LocalPeer) Encode(enc *gob.Encoder) {
+	peer.RLock()
+	defer peer.RUnlock()
+	peer.Peer.Encode(enc)
+}
+
+func decodePeer(dec *gob.Decoder) (peerSummary PeerSummary, connSummaries []ConnectionSummary, err error) {
+	if err = dec.Decode(&peerSummary); err != nil {
+		return
+	}
+	if err = dec.Decode(&connSummaries); err != nil {
+		return
+	}
+	return
+}
+
+func makeConnsMap(peer *Peer, connSummaries []ConnectionSummary, byName map[PeerName]*Peer) map[PeerName]Connection {
+	conns := make(map[PeerName]Connection)
+	for _, connSummary := range connSummaries {
+		name := PeerNameFromBin(connSummary.NameByte)
+		remotePeer := byName[name]
+		conn := NewRemoteConnection(peer, remotePeer, connSummary.RemoteTCPAddr, connSummary.Outbound, connSummary.Established)
+		conns[name] = conn
+	}
+	return conns
+}

mesh/peers_test.go

@@ -0,0 +1,376 @@
+package mesh
+
+import (
+	"fmt"
+	"math/rand"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/require"
+)
+
+// TODO we should also test:
+//
+// - applying an incremental update, including the case where that
+//   leads to an UnknownPeerError
+//
+// - the "improved update" calculation
+//
+// - non-gc of peers that are only referenced locally
+
+func newNode(name PeerName) (*Peer, *Peers) {
+	peer := NewLocalPeer(name, "", nil)
+	peers := NewPeers(peer)
+	return peer.Peer, peers
+}
+
+// Check that ApplyUpdate copies the whole topology from peers
+func checkApplyUpdate(t *testing.T, peers *Peers) {
+	dummyName, _ := PeerNameFromString("99:00:00:01:00:00")
+	// We need a new node outside of the network, with a connection
+	// into it.
+	_, testBedPeers := newNode(dummyName)
+	testBedPeers.AddTestConnection(peers.ourself.Peer)
+	testBedPeers.ApplyUpdate(peers.EncodePeers(peers.Names()))
+
+	checkTopologyPeers(t, true, testBedPeers.allPeersExcept(dummyName), peers.allPeers()...)
+}
+
+func TestPeersEncoding(t *testing.T) {
+	const numNodes = 20
+	const numIters = 1000
+	var peer [numNodes]*Peer
+	var ps [numNodes]*Peers
+	for i := 0; i < numNodes; i++ {
+		name, _ := PeerNameFromString(fmt.Sprintf("%02d:00:00:01:00:00", i))
+		peer[i], ps[i] = newNode(name)
+	}
+
+	var conns []struct{ from, to int }
+	for i := 0; i < numIters; i++ {
+		oper := rand.Intn(2)
+		switch oper {
+		case 0:
+			from, to := rand.Intn(numNodes), rand.Intn(numNodes)
+			if from != to {
+				if _, found := peer[from].connections[peer[to].Name]; !found {
+					ps[from].AddTestConnection(peer[to])
+					conns = append(conns, struct{ from, to int }{from, to})
+					checkApplyUpdate(t, ps[from])
+				}
+			}
+		case 1:
+			if len(conns) > 0 {
+				n := rand.Intn(len(conns))
+				c := conns[n]
+				ps[c.from].DeleteTestConnection(peer[c.to])
+				ps[c.from].GarbageCollect()
+				checkApplyUpdate(t, ps[c.from])
+				conns = append(conns[:n], conns[n+1:]...)
+			}
+		}
+	}
+}
+
+func garbageCollect(peers *Peers) []*Peer {
+	var removed []*Peer
+	peers.OnGC(func(peer *Peer) { removed = append(removed, peer) })
+	peers.GarbageCollect()
+	return removed
+}
+
+func TestPeersGarbageCollection(t *testing.T) {
+	const (
+		peer1NameString = "01:00:00:01:00:00"
+		peer2NameString = "02:00:00:02:00:00"
+		peer3NameString = "03:00:00:03:00:00"
+	)
+	var (
+		peer1Name, _ = PeerNameFromString(peer1NameString)
+		peer2Name, _ = PeerNameFromString(peer2NameString)
+		peer3Name, _ = PeerNameFromString(peer3NameString)
+	)
+
+	// Create some peers with some connections to each other
+	p1, ps1 := newNode(peer1Name)
+	p2, ps2 := newNode(peer2Name)
+	p3, ps3 := newNode(peer3Name)
+	ps1.AddTestConnection(p2)
+	ps2.AddTestRemoteConnection(p1, p2)
+	ps2.AddTestConnection(p1)
+	ps2.AddTestConnection(p3)
+	ps3.AddTestConnection(p1)
+	ps1.AddTestConnection(p3)
+	ps2.AddTestRemoteConnection(p1, p3)
+	ps2.AddTestRemoteConnection(p3, p1)
+
+	// Every peer is referenced, so nothing should be dropped
+	require.Empty(t, garbageCollect(ps1), "peers removed")
+	require.Empty(t, garbageCollect(ps2), "peers removed")
+	require.Empty(t, garbageCollect(ps3), "peers removed")
+
+	// Drop the connection from 2 to 3, and 3 isn't garbage-collected
+	// because 1 has a connection to 3
+	ps2.DeleteTestConnection(p3)
+	require.Empty(t, garbageCollect(ps2), "peers removed")
+
+	// Drop the connection from 1 to 3, and 3 will get removed by
+	// garbage-collection
+	ps1.DeleteTestConnection(p3)
+	checkPeerArray(t, garbageCollect(ps1), p3)
+}
+
+func TestShortIDCollisions(t *testing.T) {
+	rng := rand.New(rand.NewSource(time.Now().UnixNano()))
+	_, peers := newNode(PeerName(1 << PeerShortIDBits))
+
+	// Make enough peers that short id collisions are
+	// overwhelmingly likely
+	ps := make([]*Peer, 1<<PeerShortIDBits)
+	for i := 0; i < 1<<PeerShortIDBits; i++ {
+		ps[i] = NewPeer(PeerName(i), "", PeerUID(i), 0,
+			PeerShortID(rng.Intn(1<<PeerShortIDBits)))
+	}
+
+	shuffle := func() {
+		for i := range ps {
+			j := rng.Intn(i + 1)
+			ps[i], ps[j] = ps[j], ps[i]
+		}
+	}
+
+	// Fill peers
+	shuffle()
+	var pending PeersPendingNotifications
+	for _, p := range ps {
+		peers.addByShortID(p, &pending)
+	}
+
+	// Check invariants
+	counts := make([]int, 1<<PeerShortIDBits)
+	saw := func(p *Peer) {
+		if p != peers.ourself.Peer {
+			counts[p.UID]++
+		}
+	}
+
+	for shortID, entry := range peers.byShortID {
+		if entry.peer == nil {
+			// no principal peer for this short id, so
+			// others must be empty
+			require.Empty(t, entry.others)
+			continue
+		}
+
+		require.Equal(t, shortID, entry.peer.ShortID)
+		saw(entry.peer)
+
+		for _, p := range entry.others {
+			saw(p)
+			require.Equal(t, shortID, p.ShortID)
+
+			// the principal peer should have the lowest name
+			require.True(t, p.Name > entry.peer.Name)
+		}
+	}
+
+	// Check that every peer was seen
+	for _, n := range counts {
+		require.Equal(t, 1, n)
+	}
+
+	// Delete all the peers
+	shuffle()
+	for _, p := range ps {
+		peers.deleteByShortID(p, &pending)
+	}
+
+	for _, entry := range peers.byShortID {
+		if entry.peer != peers.ourself.Peer {
+			require.Nil(t, entry.peer)
+		}
+
+		require.Empty(t, entry.others)
+	}
+}
+
+// Test the easy case of short id reassignment, when few short ids are taken
+func TestShortIDReassignmentEasy(t *testing.T) {
+	rng := rand.New(rand.NewSource(time.Now().UnixNano()))
+	_, peers := newNode(PeerName(0))
+
+	for i := 1; i <= 10; i++ {
+		peers.FetchWithDefault(NewPeer(PeerName(i), "", PeerUID(i), 0,
+			PeerShortID(rng.Intn(1<<PeerShortIDBits))))
+	}
+
+	checkShortIDReassignment(t, peers)
+}
+
+// Test the hard case of short id reassignment, when most short ids are taken
+func TestShortIDReassignmentHard(t *testing.T) {
+	rng := rand.New(rand.NewSource(time.Now().UnixNano()))
+	_, peers := newNode(PeerName(1 << PeerShortIDBits))
+
+	// Take all short ids
+	ps := make([]*Peer, 1<<PeerShortIDBits)
+	var pending PeersPendingNotifications
+	for i := 0; i < 1<<PeerShortIDBits; i++ {
+		ps[i] = NewPeer(PeerName(i), "", PeerUID(i), 0,
+			PeerShortID(i))
+		peers.addByShortID(ps[i], &pending)
+	}
+
+	// As all short ids are taken, an attempted reassigment won't
+	// do anything
+	oldShortID := peers.ourself.ShortID
+	require.False(t, peers.reassignLocalShortID(&pending))
+	require.Equal(t, oldShortID, peers.ourself.ShortID)
+
+	// Free up a few ids
+	for i := 0; i < 10; i++ {
+		x := rng.Intn(len(ps))
+		if ps[x] != nil {
+			peers.deleteByShortID(ps[x], &pending)
+			ps[x] = nil
+		}
+	}
+
+	checkShortIDReassignment(t, peers)
+}
+
+func checkShortIDReassignment(t *testing.T, peers *Peers) {
+	oldShortID := peers.ourself.ShortID
+	peers.reassignLocalShortID(&PeersPendingNotifications{})
+	require.NotEqual(t, oldShortID, peers.ourself.ShortID)
+	require.Equal(t, peers.ourself.Peer, peers.byShortID[peers.ourself.ShortID].peer)
+}
+
+func TestShortIDInvalidation(t *testing.T) {
+	_, peers := newNode(PeerName(1 << PeerShortIDBits))
+
+	// need to use a short id that is not the local peer's
+	shortID := peers.ourself.ShortID + 1
+
+	var pending PeersPendingNotifications
+
+	requireInvalidateShortIDs := func(expect bool) {
+		require.Equal(t, expect, pending.invalidateShortIDs)
+		pending.invalidateShortIDs = false
+	}
+
+	// The use of a fresh short id does not cause invalidation
+	a := NewPeer(PeerName(1), "", PeerUID(1), 0, shortID)
+	peers.addByShortID(a, &pending)
+	requireInvalidateShortIDs(false)
+
+	// An addition which does not change the mapping
+	b := NewPeer(PeerName(2), "", PeerUID(2), 0, shortID)
+	peers.addByShortID(b, &pending)
+	requireInvalidateShortIDs(false)
+
+	// An addition which does change the mapping
+	c := NewPeer(PeerName(0), "", PeerUID(0), 0, shortID)
+	peers.addByShortID(c, &pending)
+	requireInvalidateShortIDs(true)
+
+	// A deletion which does not change the mapping
+	peers.deleteByShortID(b, &pending)
+	requireInvalidateShortIDs(false)
+
+	// A deletion which does change the mapping
+	peers.deleteByShortID(c, &pending)
+	requireInvalidateShortIDs(true)
+
+	// Deleting the last peer with a short id does not cause invalidation
+	peers.deleteByShortID(a, &pending)
+	requireInvalidateShortIDs(false)
+
+	// .. but subsequent reuse of that short id does cause invalidation
+	peers.addByShortID(a, &pending)
+	requireInvalidateShortIDs(true)
+}
+
+func TestShortIDPropagation(t *testing.T) {
+	_, peers1 := newNode(PeerName(1))
+	_, peers2 := newNode(PeerName(2))
+
+	peers1.AddTestConnection(peers2.ourself.Peer)
+	peers1.ApplyUpdate(peers2.EncodePeers(peers2.Names()))
+	peers12 := peers1.Fetch(PeerName(2))
+	old := peers12.PeerSummary
+
+	require.True(t,
+		peers2.reassignLocalShortID(&PeersPendingNotifications{}))
+	peers1.ApplyUpdate(peers2.EncodePeers(peers2.Names()))
+	require.NotEqual(t, old.Version, peers12.Version)
+	require.NotEqual(t, old.ShortID, peers12.ShortID)
+}
+
+func TestShortIDCollision(t *testing.T) {
+	// Create 3 peers
+	_, peers1 := newNode(PeerName(1))
+	_, peers2 := newNode(PeerName(2))
+	_, peers3 := newNode(PeerName(3))
+
+	var pending PeersPendingNotifications
+	peers1.setLocalShortID(1, &pending)
+	peers2.setLocalShortID(2, &pending)
+	peers3.setLocalShortID(3, &pending)
+
+	peers2.AddTestConnection(peers1.ourself.Peer)
+	peers3.AddTestConnection(peers2.ourself.Peer)
+
+	// Propogate from 1 to 2 to 3
+	peers2.ApplyUpdate(peers1.EncodePeers(peers1.Names()))
+	peers3.ApplyUpdate(peers2.EncodePeers(peers2.Names()))
+
+	// Force the short id of peer 1 to collide with peer 2.  Peer
+	// 1 has the lowest name, so it gets to keep the short id
+	peers1.setLocalShortID(2, &pending)
+
+	oldShortID := peers2.ourself.ShortID
+	_, updated, _ := peers2.ApplyUpdate(peers1.EncodePeers(peers1.Names()))
+
+	// peer 2 should have noticed the collision and resolved it
+	require.NotEqual(t, oldShortID, peers2.ourself.ShortID)
+
+	// The Peers do not have a Router, so broadcastPeerUpdate does
+	// nothing in the context of this test.  So we fake what it
+	// would do.
+	updated[PeerName(2)] = void
+
+	// the update from peer 2 should include its short id change
+	peers3.ApplyUpdate(peers2.EncodePeers(updated))
+	require.Equal(t, peers2.ourself.ShortID,
+		peers3.Fetch(PeerName(2)).ShortID)
+}
+
+// Test the case where all short ids are taken, but then some peers go
+// away, so the local peer reassigns
+func TestDeferredShortIDReassignment(t *testing.T) {
+	rng := rand.New(rand.NewSource(time.Now().UnixNano()))
+	_, us := newNode(PeerName(1 << PeerShortIDBits))
+
+	// Connect us to other peers occupying all short ids
+	others := make([]*Peers, 1<<PeerShortIDBits)
+	var pending PeersPendingNotifications
+	for i := range others {
+		_, others[i] = newNode(PeerName(i))
+		others[i].setLocalShortID(PeerShortID(i), &pending)
+		us.AddTestConnection(others[i].ourself.Peer)
+	}
+
+	// Check that, as expected, the local peer does not own its
+	// short id
+	require.NotEqual(t, us.ourself.Peer,
+		us.byShortID[us.ourself.ShortID].peer)
+
+	// Disconnect one peer, and we should now be able to claim its
+	// short id
+	other := others[rng.Intn(1<<PeerShortIDBits)]
+	us.DeleteTestConnection(other.ourself.Peer)
+	us.GarbageCollect()
+
+	require.Equal(t, us.ourself.Peer, us.byShortID[us.ourself.ShortID].peer)
+}

mesh/protocol.go

@@ -1,4 +1,4 @@
-package router
+package mesh
 
 import (
 	"bytes"
@@ -328,13 +328,14 @@ func (res *ProtocolIntroResults) setupCrypto(params ProtocolIntroParams, remoteP
 type ProtocolTag byte
 
 const (
-	ProtocolHeartbeat ProtocolTag = iota
-	ProtocolConnectionEstablished
-	ProtocolFragmentationReceived
-	ProtocolPMTUVerified
+	ProtocolHeartbeat = iota
+	ProtocolReserved1 // legacy overlay control messages
+	ProtocolReserved2 //
+	ProtocolReserved3 //
 	ProtocolGossip
 	ProtocolGossipUnicast
 	ProtocolGossipBroadcast
+	ProtocolOverlayControlMsg
 )
 
 type ProtocolMsg struct {

mesh/protocol_crypto.go

@@ -0,0 +1,180 @@
+package mesh
+
+import (
+	"crypto/rand"
+	"crypto/sha256"
+	"encoding/binary"
+	"encoding/gob"
+	"fmt"
+	"io"
+	"sync"
+
+	"golang.org/x/crypto/nacl/box"
+	"golang.org/x/crypto/nacl/secretbox"
+)
+
+const MaxTCPMsgSize = 10 * 1024 * 1024
+
+func GenerateKeyPair() (publicKey, privateKey *[32]byte, err error) {
+	return box.GenerateKey(rand.Reader)
+}
+
+func FormSessionKey(remotePublicKey, localPrivateKey *[32]byte, secretKey []byte) *[32]byte {
+	var sharedKey [32]byte
+	box.Precompute(&sharedKey, remotePublicKey, localPrivateKey)
+	sharedKeySlice := sharedKey[:]
+	sharedKeySlice = append(sharedKeySlice, secretKey...)
+	sessionKey := sha256.Sum256(sharedKeySlice)
+	return &sessionKey
+}
+
+// TCP Senders/Receivers
+
+// The lowest 64 bits of the nonce contain the message sequence
+// number. The top most bit indicates the connection polarity at the
+// sender - '1' for outbound; the next indicates protocol type - '1'
+// for TCP. The remaining 126 bits are zero. The polarity is needed so
+// that the two ends of a connection do not use the same nonces; the
+// protocol type so that the TCP connection nonces are distinct from
+// nonces used by overlay connections, if they share the session key.
+// This is a requirement of the NaCl Security Model; see
+// http://nacl.cr.yp.to/box.html.
+type TCPCryptoState struct {
+	sessionKey *[32]byte
+	nonce      [24]byte
+	seqNo      uint64
+}
+
+func NewTCPCryptoState(sessionKey *[32]byte, outbound bool) *TCPCryptoState {
+	s := &TCPCryptoState{sessionKey: sessionKey}
+	if outbound {
+		s.nonce[0] |= (1 << 7)
+	}
+	s.nonce[0] |= (1 << 6)
+	return s
+}
+
+func (s *TCPCryptoState) advance() {
+	s.seqNo++
+	binary.BigEndian.PutUint64(s.nonce[16:24], s.seqNo)
+}
+
+type TCPSender interface {
+	Send([]byte) error
+}
+
+type GobTCPSender struct {
+	encoder *gob.Encoder
+}
+
+type LengthPrefixTCPSender struct {
+	writer io.Writer
+}
+
+type EncryptedTCPSender struct {
+	sync.RWMutex
+	sender TCPSender
+	state  *TCPCryptoState
+}
+
+func NewGobTCPSender(encoder *gob.Encoder) *GobTCPSender {
+	return &GobTCPSender{encoder: encoder}
+}
+
+func (sender *GobTCPSender) Send(msg []byte) error {
+	return sender.encoder.Encode(msg)
+}
+
+func NewLengthPrefixTCPSender(writer io.Writer) *LengthPrefixTCPSender {
+	return &LengthPrefixTCPSender{writer: writer}
+}
+
+func (sender *LengthPrefixTCPSender) Send(msg []byte) error {
+	l := len(msg)
+	if l > MaxTCPMsgSize {
+		return fmt.Errorf("outgoing message exceeds maximum size: %d > %d", l, MaxTCPMsgSize)
+	}
+	// We copy the message so we can send it in a single Write
+	// operation, thus making this thread-safe without locking.
+	prefixedMsg := make([]byte, 4+l)
+	binary.BigEndian.PutUint32(prefixedMsg, uint32(l))
+	copy(prefixedMsg[4:], msg)
+	_, err := sender.writer.Write(prefixedMsg)
+	return err
+}
+
+func NewEncryptedTCPSender(sender TCPSender, sessionKey *[32]byte, outbound bool) *EncryptedTCPSender {
+	return &EncryptedTCPSender{sender: sender, state: NewTCPCryptoState(sessionKey, outbound)}
+}
+
+func (sender *EncryptedTCPSender) Send(msg []byte) error {
+	sender.Lock()
+	defer sender.Unlock()
+	encodedMsg := secretbox.Seal(nil, msg, &sender.state.nonce, sender.state.sessionKey)
+	sender.state.advance()
+	return sender.sender.Send(encodedMsg)
+}
+
+type TCPReceiver interface {
+	Receive() ([]byte, error)
+}
+
+type GobTCPReceiver struct {
+	decoder *gob.Decoder
+}
+
+type LengthPrefixTCPReceiver struct {
+	reader io.Reader
+}
+
+type EncryptedTCPReceiver struct {
+	receiver TCPReceiver
+	state    *TCPCryptoState
+}
+
+func NewGobTCPReceiver(decoder *gob.Decoder) *GobTCPReceiver {
+	return &GobTCPReceiver{decoder: decoder}
+}
+
+func (receiver *GobTCPReceiver) Receive() ([]byte, error) {
+	var msg []byte
+	err := receiver.decoder.Decode(&msg)
+	return msg, err
+}
+
+func NewLengthPrefixTCPReceiver(reader io.Reader) *LengthPrefixTCPReceiver {
+	return &LengthPrefixTCPReceiver{reader: reader}
+}
+
+func (receiver *LengthPrefixTCPReceiver) Receive() ([]byte, error) {
+	lenPrefix := make([]byte, 4)
+	if _, err := io.ReadFull(receiver.reader, lenPrefix); err != nil {
+		return nil, err
+	}
+	l := binary.BigEndian.Uint32(lenPrefix)
+	if l > MaxTCPMsgSize {
+		return nil, fmt.Errorf("incoming message exceeds maximum size: %d > %d", l, MaxTCPMsgSize)
+	}
+	msg := make([]byte, l)
+	_, err := io.ReadFull(receiver.reader, msg)
+	return msg, err
+}
+
+func NewEncryptedTCPReceiver(receiver TCPReceiver, sessionKey *[32]byte, outbound bool) *EncryptedTCPReceiver {
+	return &EncryptedTCPReceiver{receiver: receiver, state: NewTCPCryptoState(sessionKey, !outbound)}
+}
+
+func (receiver *EncryptedTCPReceiver) Receive() ([]byte, error) {
+	msg, err := receiver.receiver.Receive()
+	if err != nil {
+		return nil, err
+	}
+
+	decodedMsg, success := secretbox.Open(nil, msg, &receiver.state.nonce, receiver.state.sessionKey)
+	if !success {
+		return nil, fmt.Errorf("Unable to decrypt TCP msg")
+	}
+
+	receiver.state.advance()
+	return decodedMsg, nil
+}

mesh/protocol_test.go

@@ -1,4 +1,4 @@
-package router
+package mesh
 
 import (
 	"io"

mesh/router.go

@@ -0,0 +1,198 @@
+package mesh
+
+import (
+	"fmt"
+	"math"
+	"net"
+	"sync"
+	"time"
+
+	"github.com/weaveworks/weave/common"
+)
+
+const (
+	Port           = 6783
+	ChannelSize    = 16
+	TCPHeartbeat   = 30 * time.Second
+	GossipInterval = 30 * time.Second
+	MaxDuration    = time.Duration(math.MaxInt64)
+
+	acceptMaxTokens  = 100                    // [1]
+	acceptTokenDelay = 100 * time.Millisecond // [2]
+)
+
+// [1] capacity of token bucket for rate limiting accepts
+
+// [2] control rate at which new tokens are added to the bucket
+
+var (
+	log        = common.Log
+	checkFatal = common.CheckFatal
+	checkWarn  = common.CheckWarn
+)
+
+type Config struct {
+	Port               int
+	ProtocolMinVersion byte
+	Password           []byte
+	ConnLimit          int
+	PeerDiscovery      bool
+	TrustedSubnets     []*net.IPNet
+}
+
+type Router struct {
+	Config
+	Overlay         Overlay
+	Ourself         *LocalPeer
+	Peers           *Peers
+	Routes          *Routes
+	ConnectionMaker *ConnectionMaker
+	gossipLock      sync.RWMutex
+	gossipChannels  GossipChannels
+	TopologyGossip  Gossip
+	acceptLimiter   *TokenBucket
+}
+
+func NewRouter(config Config, name PeerName, nickName string, overlay Overlay) *Router {
+	router := &Router{Config: config, gossipChannels: make(GossipChannels)}
+
+	if overlay == nil {
+		overlay = NullOverlay{}
+	}
+
+	router.Overlay = overlay
+	router.Ourself = NewLocalPeer(name, nickName, router)
+	router.Peers = NewPeers(router.Ourself)
+	router.Peers.OnGC(func(peer *Peer) {
+		log.Println("Removed unreachable peer", peer)
+	})
+	router.Routes = NewRoutes(router.Ourself, router.Peers)
+	router.ConnectionMaker = NewConnectionMaker(router.Ourself, router.Peers, router.Port, router.PeerDiscovery)
+	router.TopologyGossip = router.NewGossip("topology", router)
+	router.acceptLimiter = NewTokenBucket(acceptMaxTokens, acceptTokenDelay)
+
+	return router
+}
+
+// Start listening for TCP connections. This is separate from
+// NewRouter so that gossipers can register before we start forming
+// connections.
+func (router *Router) Start() {
+	router.listenTCP(router.Port)
+}
+
+func (router *Router) Stop() error {
+	// TODO: perform graceful shutdown...
+	return nil
+}
+
+func (router *Router) UsingPassword() bool {
+	return router.Password != nil
+}
+
+func (router *Router) listenTCP(localPort int) {
+	localAddr, err := net.ResolveTCPAddr("tcp4", fmt.Sprint(":", localPort))
+	checkFatal(err)
+	ln, err := net.ListenTCP("tcp4", localAddr)
+	checkFatal(err)
+	go func() {
+		defer ln.Close()
+		for {
+			tcpConn, err := ln.AcceptTCP()
+			if err != nil {
+				log.Errorln(err)
+				continue
+			}
+			router.acceptTCP(tcpConn)
+			router.acceptLimiter.Wait()
+		}
+	}()
+}
+
+func (router *Router) acceptTCP(tcpConn *net.TCPConn) {
+	remoteAddrStr := tcpConn.RemoteAddr().String()
+	log.Printf("->[%s] connection accepted", remoteAddrStr)
+	connRemote := NewRemoteConnection(router.Ourself.Peer, nil, remoteAddrStr, false, false)
+	StartLocalConnection(connRemote, tcpConn, router, true)
+}
+
+// Gossiper methods - the Router is the topology Gossiper
+
+type TopologyGossipData struct {
+	peers  *Peers
+	update PeerNameSet
+}
+
+func (d *TopologyGossipData) Merge(other GossipData) GossipData {
+	names := make(PeerNameSet)
+	for name := range d.update {
+		names[name] = void
+	}
+	for name := range other.(*TopologyGossipData).update {
+		names[name] = void
+	}
+	return &TopologyGossipData{peers: d.peers, update: names}
+}
+
+func (d *TopologyGossipData) Encode() [][]byte {
+	return [][]byte{d.peers.EncodePeers(d.update)}
+}
+
+func (router *Router) BroadcastTopologyUpdate(update []*Peer) {
+	names := make(PeerNameSet)
+	for _, p := range update {
+		names[p.Name] = void
+	}
+	router.TopologyGossip.GossipBroadcast(
+		&TopologyGossipData{peers: router.Peers, update: names})
+}
+
+func (router *Router) OnGossipUnicast(sender PeerName, msg []byte) error {
+	return fmt.Errorf("unexpected topology gossip unicast: %v", msg)
+}
+
+func (router *Router) OnGossipBroadcast(_ PeerName, update []byte) (GossipData, error) {
+	origUpdate, _, err := router.applyTopologyUpdate(update)
+	if err != nil || len(origUpdate) == 0 {
+		return nil, err
+	}
+	return &TopologyGossipData{peers: router.Peers, update: origUpdate}, nil
+}
+
+func (router *Router) Gossip() GossipData {
+	return &TopologyGossipData{peers: router.Peers, update: router.Peers.Names()}
+}
+
+func (router *Router) OnGossip(update []byte) (GossipData, error) {
+	_, newUpdate, err := router.applyTopologyUpdate(update)
+	if err != nil || len(newUpdate) == 0 {
+		return nil, err
+	}
+	return &TopologyGossipData{peers: router.Peers, update: newUpdate}, nil
+}
+
+func (router *Router) applyTopologyUpdate(update []byte) (PeerNameSet, PeerNameSet, error) {
+	origUpdate, newUpdate, err := router.Peers.ApplyUpdate(update)
+	if err != nil {
+		return nil, nil, err
+	}
+	if len(newUpdate) > 0 {
+		router.ConnectionMaker.Refresh()
+		router.Routes.Recalculate()
+	}
+	return origUpdate, newUpdate, nil
+}
+
+func (router *Router) Trusts(remote *RemoteConnection) bool {
+	if tcpAddr, err := net.ResolveTCPAddr("tcp4", remote.remoteTCPAddr); err == nil {
+		for _, trustedSubnet := range router.TrustedSubnets {
+			if trustedSubnet.Contains(tcpAddr.IP) {
+				return true
+			}
+		}
+	} else {
+		// Should not happen as remoteTCPAddr was obtained from TCPConn
+		log.Errorf("Unable to parse remote TCP addr: %s", err)
+	}
+	return false
+}

mesh/routes.go

@@ -1,18 +1,22 @@
-package router
+package mesh
 
 import (
 	"math"
 	"sync"
 )
 
+type unicastRoutes map[PeerName]PeerName
+type broadcastRoutes map[PeerName][]PeerName
+
 type Routes struct {
 	sync.RWMutex
 	ourself      *LocalPeer
 	peers        *Peers
-	unicast      map[PeerName]PeerName
-	unicastAll   map[PeerName]PeerName // [1]
-	broadcast    map[PeerName][]PeerName
-	broadcastAll map[PeerName][]PeerName // [1]
+	onChange     []func()
+	unicast      unicastRoutes
+	unicastAll   unicastRoutes // [1]
+	broadcast    broadcastRoutes
+	broadcastAll broadcastRoutes // [1]
 	recalculate  chan<- *struct{}
 	wait         chan<- chan struct{}
 	// [1] based on *all* connections, not just established &
@@ -25,10 +29,10 @@ func NewRoutes(ourself *LocalPeer, peers *Peers) *Routes {
 	routes := &Routes{
 		ourself:      ourself,
 		peers:        peers,
-		unicast:      make(map[PeerName]PeerName),
-		unicastAll:   make(map[PeerName]PeerName),
-		broadcast:    make(map[PeerName][]PeerName),
-		broadcastAll: make(map[PeerName][]PeerName),
+		unicast:      make(unicastRoutes),
+		unicastAll:   make(unicastRoutes),
+		broadcast:    make(broadcastRoutes),
+		broadcastAll: make(broadcastRoutes),
 		recalculate:  recalculate,
 		wait:         wait}
 	routes.unicast[ourself.Name] = UnknownPeerName
@@ -39,6 +43,12 @@ func NewRoutes(ourself *LocalPeer, peers *Peers) *Routes {
 	return routes
 }
 
+func (routes *Routes) OnChange(callback func()) {
+	routes.Lock()
+	defer routes.Unlock()
+	routes.onChange = append(routes.onChange, callback)
+}
+
 func (routes *Routes) PeerNames() PeerNameSet {
 	return routes.peers.Names()
 }
@@ -60,21 +70,13 @@ func (routes *Routes) UnicastAll(name PeerName) (PeerName, bool) {
 func (routes *Routes) Broadcast(name PeerName) []PeerName {
 	routes.RLock()
 	defer routes.RUnlock()
-	hops, found := routes.broadcast[name]
-	if !found {
-		return []PeerName{}
-	}
-	return hops
+	return routes.broadcast[name]
 }
 
 func (routes *Routes) BroadcastAll(name PeerName) []PeerName {
 	routes.RLock()
 	defer routes.RUnlock()
-	hops, found := routes.broadcastAll[name]
-	if !found {
-		return []PeerName{}
-	}
-	return hops
+	return routes.broadcastAll[name]
 }
 
 // Choose min(log2(n_peers), n_neighbouring_peers) neighbours, with a
@@ -90,20 +92,24 @@ func (routes *Routes) BroadcastAll(name PeerName) []PeerName {
 // returns a higher proportion of neighbours than elsewhere. In
 // extremis, on peers with fewer than log2(n_peers) neighbours, all
 // neighbours are returned.
-func (routes *Routes) RandomNeighbours(except PeerName) PeerNameSet {
-	res := make(PeerNameSet)
+func (routes *Routes) RandomNeighbours(except PeerName) []PeerName {
+	destinations := make(PeerNameSet)
 	routes.RLock()
 	defer routes.RUnlock()
 	count := int(math.Log2(float64(len(routes.unicastAll))))
 	// depends on go's random map iteration
 	for _, dst := range routes.unicastAll {
 		if dst != UnknownPeerName && dst != except {
-			res[dst] = void
-			if len(res) >= count {
+			destinations[dst] = void
+			if len(destinations) >= count {
 				break
 			}
 		}
 	}
+	res := make([]PeerName, 0, len(destinations))
+	for dst := range destinations {
+		res = append(res, dst)
+	}
 	return res
 }
 
@@ -147,6 +153,8 @@ func (routes *Routes) calculate() {
 	routes.peers.RLock()
 	routes.ourself.RLock()
 	var (
+		oldUnicast   = routes.unicast
+		oldBroadcast = routes.broadcast
 		unicast      = routes.calculateUnicast(true)
 		unicastAll   = routes.calculateUnicast(false)
 		broadcast    = routes.calculateBroadcast(true)
@@ -160,7 +168,14 @@ func (routes *Routes) calculate() {
 	routes.unicastAll = unicastAll
 	routes.broadcast = broadcast
 	routes.broadcastAll = broadcastAll
+	onChange := routes.onChange
 	routes.Unlock()
+
+	if !unicast.equals(oldUnicast) || !broadcast.equals(oldBroadcast) {
+		for _, callback := range onChange {
+			callback()
+		}
+	}
 }
 
 // Calculate all the routes for the question: if *we* want to send a
@@ -172,7 +187,7 @@ func (routes *Routes) calculate() {
 // any knowledge of the MAC address at all. Thus there's no need
 // to exchange knowledge of MAC addresses, nor any constraints on
 // the routes that we construct.
-func (routes *Routes) calculateUnicast(establishedAndSymmetric bool) map[PeerName]PeerName {
+func (routes *Routes) calculateUnicast(establishedAndSymmetric bool) unicastRoutes {
 	_, unicast := routes.ourself.Routes(nil, establishedAndSymmetric)
 	return unicast
 }
@@ -195,9 +210,9 @@ func (routes *Routes) calculateUnicast(establishedAndSymmetric bool) map[PeerNam
 //     Y =/= Z /\ X.Routes(Y) <= X.Routes(Z) =>
 //     X.Routes(Y) u [P | Y.HasSymmetricConnectionTo(P)] <= X.Routes(Z)
 // where <= is the subset relationship on keys of the returned map.
-func (routes *Routes) calculateBroadcast(establishedAndSymmetric bool) map[PeerName][]PeerName {
-	broadcast := make(map[PeerName][]PeerName)
-	for _, peer := range routes.peers.table {
+func (routes *Routes) calculateBroadcast(establishedAndSymmetric bool) broadcastRoutes {
+	broadcast := make(broadcastRoutes)
+	for _, peer := range routes.peers.byName {
 		hops := []PeerName{}
 		if found, reached := peer.Routes(routes.ourself.Peer, establishedAndSymmetric); found {
 			routes.ourself.ForEachConnectedPeer(establishedAndSymmetric, reached,
@@ -207,3 +222,55 @@ func (routes *Routes) calculateBroadcast(establishedAndSymmetric bool) map[PeerN
 	}
 	return broadcast
 }
+
+func (a unicastRoutes) equals(b unicastRoutes) bool {
+	for key, aval := range a {
+		if bval, ok := b[key]; !ok || bval != aval {
+			return false
+		}
+	}
+
+	for key := range b {
+		if _, ok := a[key]; !ok {
+			return false
+		}
+	}
+
+	return true
+}
+
+func (a broadcastRoutes) equals(b broadcastRoutes) bool {
+	set := make(map[PeerName]struct{})
+
+	for key, aval := range a {
+		bval, ok := b[key]
+		if !ok {
+			return false
+		}
+
+		for _, peer := range aval {
+			set[peer] = struct{}{}
+		}
+
+		for _, peer := range bval {
+			if _, ok := set[peer]; !ok {
+				return false
+			}
+
+			delete(set, peer)
+		}
+
+		if len(set) != 0 {
+			return false
+		}
+
+	}
+
+	for key := range b {
+		if _, ok := a[key]; !ok {
+			return false
+		}
+	}
+
+	return true
+}

mesh/status.go

@@ -1,7 +1,8 @@
-package router
+package mesh
 
 import (
-	"time"
+	"fmt"
+	"net"
 )
 
 type Status struct {
@@ -13,27 +14,20 @@ type Status struct {
 	Name               string
 	NickName           string
 	Port               int
-	Interface          string
-	CaptureStats       map[string]int
-	MACs               []MACStatus
 	Peers              []PeerStatus
 	UnicastRoutes      []UnicastRouteStatus
 	BroadcastRoutes    []BroadcastRouteStatus
 	Connections        []LocalConnectionStatus
 	Targets            []string
-}
-
-type MACStatus struct {
-	Mac      string
-	Name     string
-	NickName string
-	LastSeen time.Time
+	OverlayDiagnostics interface{}
+	TrustedSubnets     []string
 }
 
 type PeerStatus struct {
 	Name        string
 	NickName    string
 	UID         PeerUID
+	ShortID     PeerShortID
 	Version     uint64
 	Connections []ConnectionStatus
 }
@@ -63,14 +57,6 @@ type LocalConnectionStatus struct {
 }
 
 func NewStatus(router *Router) *Status {
-	var ifaceName string
-	if router.Iface != nil {
-		ifaceName = router.Iface.Name
-	}
-	var captureStats map[string]int
-	if router.PacketSource != nil {
-		captureStats = router.PacketSource.Stats()
-	}
 	return &Status{
 		Protocol,
 		ProtocolMinVersion,
@@ -80,30 +66,13 @@ func NewStatus(router *Router) *Status {
 		router.Ourself.Name.String(),
 		router.Ourself.NickName,
 		router.Port,
-		ifaceName,
-		captureStats,
-		NewMACStatusSlice(router.Macs),
 		NewPeerStatusSlice(router.Peers),
 		NewUnicastRouteStatusSlice(router.Routes),
 		NewBroadcastRouteStatusSlice(router.Routes),
 		NewLocalConnectionStatusSlice(router.ConnectionMaker),
-		NewTargetSlice(router.ConnectionMaker)}
-}
-
-func NewMACStatusSlice(cache *MacCache) []MACStatus {
-	cache.RLock()
-	defer cache.RUnlock()
-
-	var slice []MACStatus
-	for key, entry := range cache.table {
-		slice = append(slice, MACStatus{
-			intmac(key).String(),
-			entry.peer.Name.String(),
-			entry.peer.NickName,
-			entry.lastSeen})
-	}
-
-	return slice
+		NewTargetSlice(router.ConnectionMaker),
+		router.Overlay.Diagnostics(),
+		NewTrustedSubnetsSlice(router.TrustedSubnets)}
 }
 
 func NewPeerStatusSlice(peers *Peers) []PeerStatus {
@@ -127,7 +96,8 @@ func NewPeerStatusSlice(peers *Peers) []PeerStatus {
 			peer.Name.String(),
 			peer.NickName,
 			peer.UID,
-			peer.version,
+			peer.ShortID,
+			peer.Version,
 			connections})
 	})
 
@@ -170,40 +140,48 @@ func NewBroadcastRouteStatusSlice(routes *Routes) []BroadcastRouteStatus {
 }
 
 func NewLocalConnectionStatusSlice(cm *ConnectionMaker) []LocalConnectionStatus {
-	// We need to Refresh first in order to clear out any 'attempting'
-	// connections from cm.targets that have been established since
-	// the last run of cm.checkStateAndAttemptConnections. These
-	// entries are harmless but do represent stale state that we do
-	// not want to report.
-	cm.Refresh()
 	resultChan := make(chan []LocalConnectionStatus, 0)
 	cm.actionChan <- func() bool {
 		var slice []LocalConnectionStatus
-		for conn := range cm.ourself.Connections() {
+		for conn := range cm.connections {
 			state := "pending"
 			if conn.Established() {
 				state = "established"
 			}
-			slice = append(slice, LocalConnectionStatus{conn.RemoteTCPAddr(), conn.Outbound(), state, conn.Remote().String()})
+			lc, _ := conn.(*LocalConnection)
+			info := fmt.Sprintf("%-6v %v", lc.OverlayConn.DisplayName(), conn.Remote())
+			if lc.Router.UsingPassword() {
+				if lc.Untrusted() {
+					info = fmt.Sprintf("%-11v %v", "encrypted", info)
+				} else {
+					info = fmt.Sprintf("%-11v %v", "unencrypted", info)
+				}
+			}
+			slice = append(slice, LocalConnectionStatus{conn.RemoteTCPAddr(), conn.Outbound(), state, info})
 		}
 		for address, target := range cm.targets {
-			var state, info string
-			switch {
-			case target.lastError == nil:
-				state = "connecting"
-				info = ""
-			case target.attempting:
-				state = "retrying"
-				info = target.lastError.Error()
-			default:
-				state = "failed"
-				retry := "never"
+			add := func(state, info string) {
+				slice = append(slice, LocalConnectionStatus{address, true, state, info})
+			}
+			switch target.state {
+			case TargetWaiting:
+				until := "never"
 				if !target.tryAfter.IsZero() {
-					retry = target.tryAfter.String()
+					until = target.tryAfter.String()
 				}
-				info = target.lastError.Error() + ", retry: " + retry
+				if target.lastError == nil { // shouldn't happen
+					add("waiting", "until: "+until)
+				} else {
+					add("failed", target.lastError.Error()+", retry: "+until)
+				}
+			case TargetAttempting:
+				if target.lastError == nil {
+					add("connecting", "")
+				} else {
+					add("retrying", target.lastError.Error())
+				}
+			case TargetConnected:
 			}
-			slice = append(slice, LocalConnectionStatus{address, true, state, info})
 		}
 		resultChan <- slice
 		return false
@@ -223,3 +201,11 @@ func NewTargetSlice(cm *ConnectionMaker) []string {
 	}
 	return <-resultChan
 }
+
+func NewTrustedSubnetsSlice(trustedSubnets []*net.IPNet) []string {
+	trustedSubnetStrs := []string{}
+	for _, trustedSubnet := range trustedSubnets {
+		trustedSubnetStrs = append(trustedSubnetStrs, trustedSubnet.String())
+	}
+	return trustedSubnetStrs
+}

mesh/surrogate_gossiper.go

@@ -1,4 +1,4 @@
-package router
+package mesh
 
 type SurrogateGossipData struct {
 	messages [][]byte
@@ -12,8 +12,12 @@ func (d *SurrogateGossipData) Encode() [][]byte {
 	return d.messages
 }
 
-func (d *SurrogateGossipData) Merge(other GossipData) {
-	d.messages = append(d.messages, other.(*SurrogateGossipData).messages...)
+func (d *SurrogateGossipData) Merge(other GossipData) GossipData {
+	o := other.(*SurrogateGossipData)
+	messages := make([][]byte, 0, len(d.messages)+len(o.messages))
+	messages = append(messages, d.messages...)
+	messages = append(messages, o.messages...)
+	return &SurrogateGossipData{messages: messages}
 }
 
 // SurrogateGossiper ignores unicasts and relays broadcasts and gossips.

mesh/token_bucket.go

@@ -1,4 +1,4 @@
-package router
+package mesh
 
 import (
 	"time"

nameserver/dns.go

@@ -216,7 +216,7 @@ func (h *handler) handleRecursive(w dns.ResponseWriter, req *dns.Msg) {
 		reqCopy := req.Copy()
 		reqCopy.Id = dns.Id()
 		response, _, err := h.client.Exchange(reqCopy, fmt.Sprintf("%s:%s", server, h.upstream.Port))
-		if err != nil || response == nil {
+		if (err != nil && err != dns.ErrTruncated) || response == nil {
 			h.ns.debugf("error trying %s: %v", server, err)
 			continue
 		}

nameserver/dns_test.go

@@ -11,14 +11,14 @@ import (
 	"github.com/miekg/dns"
 	"github.com/stretchr/testify/require"
 
+	"github.com/weaveworks/weave/mesh"
 	"github.com/weaveworks/weave/net/address"
-	"github.com/weaveworks/weave/router"
 )
 
 func startServer(t *testing.T, upstream *dns.ClientConfig) (*DNSServer, *Nameserver, int, int) {
-	peername, err := router.PeerNameFromString("00:00:00:02:00:00")
+	peername, err := mesh.PeerNameFromString("00:00:00:02:00:00")
 	require.Nil(t, err)
-	nameserver := New(peername, nil, nil, "")
+	nameserver := New(peername, "", func(mesh.PeerName) bool { return true })
 	dnsserver, err := NewDNSServer(nameserver, "weave.local.", "0.0.0.0:0", "", 30, 5*time.Second)
 	require.Nil(t, err)
 	udpPort := dnsserver.servers[0].PacketConn.LocalAddr().(*net.UDPAddr).Port
@@ -39,20 +39,20 @@ func TestTruncation(t *testing.T) {
 	addrs := []address.Address{}
 	for i := address.Address(0); i < 100; i++ {
 		addrs = append(addrs, i)
-		nameserver.AddEntry("foo.weave.local.", "", router.UnknownPeerName, i)
+		nameserver.AddEntry("foo.weave.local.", "", mesh.UnknownPeerName, i)
 	}
 
-	doRequest := func(client *dns.Client, request *dns.Msg, port int) *dns.Msg {
+	doRequest := func(client *dns.Client, request *dns.Msg, port int, expectedErr error) *dns.Msg {
 		request.SetQuestion("foo.weave.local.", dns.TypeA)
 		response, _, err := client.Exchange(request, fmt.Sprintf("127.0.0.1:%d", port))
-		require.Nil(t, err)
+		require.Equal(t, expectedErr, err)
 		return response
 	}
 
 	// do a udp query, ensure we get a truncated response
 	{
 		udpClient := dns.Client{Net: "udp", UDPSize: minUDPSize}
-		response := doRequest(&udpClient, &dns.Msg{}, udpPort)
+		response := doRequest(&udpClient, &dns.Msg{}, udpPort, dns.ErrTruncated)
 		require.True(t, response.MsgHdr.Truncated)
 		require.True(t, len(response.Answer) < 100)
 	}
@@ -62,7 +62,7 @@ func TestTruncation(t *testing.T) {
 		udpClient := dns.Client{Net: "udp", UDPSize: 65535}
 		request := &dns.Msg{}
 		request.SetEdns0(65535, false)
-		response := doRequest(&udpClient, request, udpPort)
+		response := doRequest(&udpClient, request, udpPort, nil)
 		require.False(t, response.MsgHdr.Truncated)
 		require.Equal(t, len(response.Answer), 100)
 	}
@@ -70,7 +70,7 @@ func TestTruncation(t *testing.T) {
 	// do a tcp query, ensure we don't get a truncated response
 	{
 		tcpClient := dns.Client{Net: "tcp"}
-		response := doRequest(&tcpClient, &dns.Msg{}, tcpPort)
+		response := doRequest(&tcpClient, &dns.Msg{}, tcpPort, nil)
 		require.False(t, response.MsgHdr.Truncated)
 		require.Equal(t, len(response.Answer), 100)
 	}

nameserver/entry.go

@@ -5,26 +5,46 @@ import (
 	"encoding/gob"
 	"fmt"
 	"sort"
+	"strings"
 	"time"
 
+	"github.com/weaveworks/weave/mesh"
 	"github.com/weaveworks/weave/net/address"
-	"github.com/weaveworks/weave/router"
 )
 
 var now = func() int64 { return time.Now().Unix() }
 
 type Entry struct {
 	ContainerID string
-	Origin      router.PeerName
+	Origin      mesh.PeerName
 	Addr        address.Address
-	Hostname    string
+	Hostname    string // as supplied
+	lHostname   string // lowercased (not exported, so not encoded by gob)
 	Version     int
 	Tombstone   int64 // timestamp of when it was deleted
 }
 
 type Entries []Entry
+type CaseSensitive Entries
+type CaseInsensitive Entries
+type SortableEntries interface {
+	sort.Interface
+	Get(i int) Entry
+}
 
-func (e1 *Entry) equal(e2 *Entry) bool {
+// Gossip messages are sorted in a case sensitive order...
+func (es CaseSensitive) Len() int           { return len(es) }
+func (es CaseSensitive) Swap(i, j int)      { es[i], es[j] = es[j], es[i] }
+func (es CaseSensitive) Get(i int) Entry    { return es[i] }
+func (es CaseSensitive) Less(i, j int) bool { return es[i].less(&es[j]) }
+
+// ... but we store entries in a case insensitive order.
+func (es CaseInsensitive) Len() int           { return len(es) }
+func (es CaseInsensitive) Swap(i, j int)      { es[i], es[j] = es[j], es[i] }
+func (es CaseInsensitive) Get(i int) Entry    { return es[i] }
+func (es CaseInsensitive) Less(i, j int) bool { return es[i].insensitiveLess(&es[j]) }
+
+func (e1 Entry) equal(e2 Entry) bool {
 	return e1.ContainerID == e2.ContainerID &&
 		e1.Origin == e2.Origin &&
 		e1.Addr == e2.Addr &&
@@ -48,6 +68,24 @@ func (e1 *Entry) less(e2 *Entry) bool {
 	}
 }
 
+func (e1 *Entry) insensitiveLess(e2 *Entry) bool {
+	// Entries are kept sorted by Hostname, Origin, ContainerID then address
+	e1Hostname, e2Hostname := e1.lHostname, e2.lHostname
+	switch {
+	case e1Hostname != e2Hostname:
+		return e1Hostname < e2Hostname
+
+	case e1.Origin != e2.Origin:
+		return e1.Origin < e2.Origin
+
+	case e1.ContainerID != e2.ContainerID:
+		return e1.ContainerID < e2.ContainerID
+
+	default:
+		return e1.Addr < e2.Addr
+	}
+}
+
 // returns true to indicate a change
 func (e1 *Entry) merge(e2 *Entry) bool {
 	// we know container id, origin, add and hostname are equal
@@ -63,37 +101,42 @@ func (e1 *Entry) String() string {
 	return fmt.Sprintf("%s -> %s", e1.Hostname, e1.Addr.String())
 }
 
-func (es Entries) Len() int           { return len(es) }
-func (es Entries) Swap(i, j int)      { panic("Swap") }
-func (es Entries) Less(i, j int) bool { return es[i].less(&es[j]) }
+func (e1 *Entry) addLowercase() {
+	e1.lHostname = strings.ToLower(e1.Hostname)
+}
 
-func (es Entries) check() error {
+func check(es SortableEntries) error {
 	if !sort.IsSorted(es) {
 		return fmt.Errorf("Not sorted!")
 	}
-	for i := 1; i < len(es); i++ {
-		if es[i].equal(&es[i-1]) {
-			return fmt.Errorf("Duplicate entry: %d:%v and %d:%v", i-1, es[i-1], i, es[i])
+	for i := 1; i < es.Len(); i++ {
+		if es.Get(i).equal(es.Get(i - 1)) {
+			return fmt.Errorf("Duplicate entry: %d:%v and %d:%v", i-1, es.Get(i-1), i, es.Get(i))
 		}
 	}
 	return nil
 }
 
-func (es *Entries) checkAndPanic() {
-	if err := es.check(); err != nil {
+func checkAndPanic(es SortableEntries) {
+	if err := check(es); err != nil {
 		panic(err)
 	}
 }
 
-func (es *Entries) add(hostname, containerid string, origin router.PeerName, addr address.Address) Entry {
-	es.checkAndPanic()
-	defer es.checkAndPanic()
+func (es *Entries) checkAndPanic() *Entries {
+	checkAndPanic(CaseInsensitive(*es))
+	return es
+}
 
-	entry := Entry{Hostname: hostname, Origin: origin, ContainerID: containerid, Addr: addr}
+func (es *Entries) add(hostname, containerid string, origin mesh.PeerName, addr address.Address) Entry {
+	defer es.checkAndPanic().checkAndPanic()
+
+	entry := Entry{Hostname: hostname, lHostname: strings.ToLower(hostname),
+		Origin: origin, ContainerID: containerid, Addr: addr}
 	i := sort.Search(len(*es), func(i int) bool {
-		return !(*es)[i].less(&entry)
+		return !(*es)[i].insensitiveLess(&entry)
 	})
-	if i < len(*es) && (*es)[i].equal(&entry) {
+	if i < len(*es) && (*es)[i].equal(entry) {
 		if (*es)[i].Tombstone > 0 {
 			(*es)[i].Tombstone = 0
 			(*es)[i].Version++
@@ -107,16 +150,17 @@ func (es *Entries) add(hostname, containerid string, origin router.PeerName, add
 }
 
 func (es *Entries) merge(incoming Entries) Entries {
-	es.checkAndPanic()
-	defer es.checkAndPanic()
+	defer es.checkAndPanic().checkAndPanic()
+	incoming.checkAndPanic()
+
 	newEntries := Entries{}
 	i := 0
 
 	for _, entry := range incoming {
-		for i < len(*es) && (*es)[i].less(&entry) {
+		for i < len(*es) && (*es)[i].insensitiveLess(&entry) {
 			i++
 		}
-		if i < len(*es) && (*es)[i].equal(&entry) {
+		if i < len(*es) && (*es)[i].equal(entry) {
 			if (*es)[i].merge(&entry) {
 				newEntries = append(newEntries, entry)
 			}
@@ -132,9 +176,9 @@ func (es *Entries) merge(incoming Entries) Entries {
 }
 
 // f returning true means keep the entry.
-func (es *Entries) tombstone(ourname router.PeerName, f func(*Entry) bool) Entries {
-	es.checkAndPanic()
-	defer es.checkAndPanic()
+func (es *Entries) tombstone(ourname mesh.PeerName, f func(*Entry) bool) Entries {
+	defer es.checkAndPanic().checkAndPanic()
+
 	tombstoned := Entries{}
 	for i, e := range *es {
 		if f(&e) && e.Origin == ourname {
@@ -148,8 +192,8 @@ func (es *Entries) tombstone(ourname router.PeerName, f func(*Entry) bool) Entri
 }
 
 func (es *Entries) filter(f func(*Entry) bool) {
-	es.checkAndPanic()
-	defer es.checkAndPanic()
+	defer es.checkAndPanic().checkAndPanic()
+
 	i := 0
 	for _, e := range *es {
 		if !f(&e) {
@@ -163,23 +207,26 @@ func (es *Entries) filter(f func(*Entry) bool) {
 
 func (es Entries) lookup(hostname string) Entries {
 	es.checkAndPanic()
+
+	lowerHostname := strings.ToLower(hostname)
 	i := sort.Search(len(es), func(i int) bool {
-		return es[i].Hostname >= hostname
+		return es[i].lHostname >= lowerHostname
 	})
-	if i >= len(es) || es[i].Hostname != hostname {
+	if i >= len(es) || es[i].lHostname != lowerHostname {
 		return Entries{}
 	}
 
 	j := sort.Search(len(es)-i, func(j int) bool {
-		return es[i+j].Hostname > hostname
+		return es[i+j].lHostname > lowerHostname
 	})
 
 	return es[i : i+j]
 }
 
-func (es *Entries) first(f func(*Entry) bool) (*Entry, error) {
+func (es Entries) first(f func(*Entry) bool) (*Entry, error) {
 	es.checkAndPanic()
-	for _, e := range *es {
+
+	for _, e := range es {
 		if f(&e) {
 			return &e, nil
 		}
@@ -187,23 +234,49 @@ func (es *Entries) first(f func(*Entry) bool) (*Entry, error) {
 	return nil, fmt.Errorf("Not found")
 }
 
+func (es Entries) addLowercase() {
+	for i := range es {
+		es[i].addLowercase()
+	}
+}
+
 type GossipData struct {
 	Timestamp int64
 	Entries
 }
 
-func (g *GossipData) Merge(o router.GossipData) {
+func (g *GossipData) Merge(o mesh.GossipData) mesh.GossipData {
 	other := o.(*GossipData)
-	g.Entries.merge(other.Entries)
-	if g.Timestamp < other.Timestamp {
-		g.Timestamp = other.Timestamp
+	gossip := g.copy()
+	gossip.Entries.merge(other.Entries)
+	if gossip.Timestamp < other.Timestamp {
+		gossip.Timestamp = other.Timestamp
 	}
+	return gossip
+}
+
+func (g *GossipData) Decode(msg []byte) error {
+	if err := gob.NewDecoder(bytes.NewReader(msg)).Decode(g); err != nil {
+		return err
+	}
+
+	g.Entries.addLowercase() // lowercase strings not sent on the wire
+	sort.Sort(CaseInsensitive(g.Entries))
+	return nil
 }
 
 func (g *GossipData) Encode() [][]byte {
+	g2 := g.copy()
+	sort.Sort(CaseSensitive(g2.Entries))
 	buf := &bytes.Buffer{}
-	if err := gob.NewEncoder(buf).Encode(g); err != nil {
+	if err := gob.NewEncoder(buf).Encode(g2); err != nil {
 		panic(err)
 	}
 	return [][]byte{buf.Bytes()}
 }
+
+func (g *GossipData) copy() *GossipData {
+	g2 := &GossipData{Timestamp: g.Timestamp, Entries: make(Entries, len(g.Entries))}
+	copy(g2.Entries, g.Entries)
+	return g2
+}

nameserver/entry_test.go

@@ -5,79 +5,70 @@ import (
 
 	"github.com/stretchr/testify/require"
 
+	"github.com/weaveworks/weave/mesh"
 	"github.com/weaveworks/weave/net/address"
-	"github.com/weaveworks/weave/router"
 )
 
+func l(es Entries) Entries {
+	es.addLowercase()
+	return es
+}
+
+func makeEntries(values string) Entries {
+	entries := make(Entries, len(values))
+	for i, c := range values {
+		entries[i] = Entry{Hostname: string(c)}
+	}
+	return l(entries)
+}
+
 func TestAdd(t *testing.T) {
 	oldNow := now
 	defer func() { now = oldNow }()
 	now = func() int64 { return 1234 }
 
 	entries := Entries{}
-	entries.add("A", "", router.UnknownPeerName, address.Address(0))
-	expected := Entries{
-		Entry{Hostname: "A", Origin: router.UnknownPeerName, Addr: address.Address(0)},
-	}
+	entries.add("A", "", mesh.UnknownPeerName, address.Address(0))
+	expected := l(Entries{
+		Entry{Hostname: "A", Origin: mesh.UnknownPeerName, Addr: address.Address(0)},
+	})
 	require.Equal(t, entries, expected)
 
-	entries.tombstone(router.UnknownPeerName, func(e *Entry) bool { return e.Hostname == "A" })
-	expected = Entries{
-		Entry{Hostname: "A", Origin: router.UnknownPeerName, Addr: address.Address(0), Version: 1, Tombstone: 1234},
-	}
+	entries.tombstone(mesh.UnknownPeerName, func(e *Entry) bool { return e.Hostname == "A" })
+	expected = l(Entries{
+		Entry{Hostname: "A", Origin: mesh.UnknownPeerName, Addr: address.Address(0), Version: 1, Tombstone: 1234},
+	})
 	require.Equal(t, entries, expected)
 
-	entries.add("A", "", router.UnknownPeerName, address.Address(0))
-	expected = Entries{
-		Entry{Hostname: "A", Origin: router.UnknownPeerName, Addr: address.Address(0), Version: 2},
-	}
+	entries.add("A", "", mesh.UnknownPeerName, address.Address(0))
+	expected = l(Entries{
+		Entry{Hostname: "A", Origin: mesh.UnknownPeerName, Addr: address.Address(0), Version: 2},
+	})
 	require.Equal(t, entries, expected)
 }
 
 func TestMerge(t *testing.T) {
-	e1 := Entries{
-		Entry{Hostname: "A"},
-		Entry{Hostname: "C"},
-		Entry{Hostname: "D"},
-		Entry{Hostname: "F"},
-	}
-
-	e2 := Entries{
-		Entry{Hostname: "B"},
-		Entry{Hostname: "E"},
-		Entry{Hostname: "F"},
-	}
+	e1 := makeEntries("ACDF")
+	e2 := makeEntries("BEF")
 
 	diff := e1.merge(e2)
-	expectedDiff := Entries{
-		Entry{Hostname: "B"},
-		Entry{Hostname: "E"},
-	}
-	require.Equal(t, expectedDiff, diff)
 
-	expected := Entries{
-		Entry{Hostname: "A"},
-		Entry{Hostname: "B"},
-		Entry{Hostname: "C"},
-		Entry{Hostname: "D"},
-		Entry{Hostname: "E"},
-		Entry{Hostname: "F"},
-	}
-	require.Equal(t, expected, e1)
+	require.Equal(t, makeEntries("BE"), diff)
+	require.Equal(t, makeEntries("ABCDEF"), e1)
 
 	diff = e1.merge(e1)
 	require.Equal(t, Entries{}, diff)
 }
 
 func TestOldMerge(t *testing.T) {
-	e1 := Entries{Entry{Hostname: "A", Version: 0}}
-	diff := e1.merge(Entries{Entry{Hostname: "A", Version: 1}})
-	require.Equal(t, Entries{Entry{Hostname: "A", Version: 1}}, diff)
-	require.Equal(t, Entries{Entry{Hostname: "A", Version: 1}}, e1)
+	e1 := l(Entries{Entry{Hostname: "A", Version: 0}})
+	diff := e1.merge(l(Entries{Entry{Hostname: "A", Version: 1}}))
+	require.Equal(t, l(Entries{Entry{Hostname: "A", Version: 1}}), diff)
+	require.Equal(t, l(Entries{Entry{Hostname: "A", Version: 1}}), e1)
 
-	diff = e1.merge(Entries{Entry{Hostname: "A", Version: 0}})
+	diff = e1.merge(l(Entries{Entry{Hostname: "A", Version: 0}}))
 	require.Equal(t, Entries{}, diff)
-	require.Equal(t, Entries{Entry{Hostname: "A", Version: 1}}, e1)
+	require.Equal(t, l(Entries{Entry{Hostname: "A", Version: 1}}), e1)
 }
 
 func TestTombstone(t *testing.T) {
@@ -85,48 +76,48 @@ func TestTombstone(t *testing.T) {
 	defer func() { now = oldNow }()
 	now = func() int64 { return 1234 }
 
-	es := Entries{
-		Entry{Hostname: "A"},
-		Entry{Hostname: "B"},
-	}
+	es := makeEntries("AB")
 
-	es.tombstone(router.UnknownPeerName, func(e *Entry) bool {
+	es.tombstone(mesh.UnknownPeerName, func(e *Entry) bool {
 		return e.Hostname == "B"
 	})
-	expected := Entries{
+	expected := l(Entries{
 		Entry{Hostname: "A"},
 		Entry{Hostname: "B", Version: 1, Tombstone: 1234},
-	}
+	})
 	require.Equal(t, expected, es)
 }
 
 func TestDelete(t *testing.T) {
-	es := Entries{
-		Entry{Hostname: "A"},
-		Entry{Hostname: "B"},
-	}
+	es := makeEntries("AB")
 
 	es.filter(func(e *Entry) bool {
 		return e.Hostname != "A"
 	})
-	expected := Entries{
-		Entry{Hostname: "B"},
-	}
-	require.Equal(t, expected, es)
+	require.Equal(t, makeEntries("B"), es)
 }
 
 func TestLookup(t *testing.T) {
-	es := Entries{
+	es := l(Entries{
 		Entry{Hostname: "A"},
 		Entry{Hostname: "B", ContainerID: "bar"},
 		Entry{Hostname: "B", ContainerID: "foo"},
 		Entry{Hostname: "C"},
-	}
+	})
 
 	have := es.lookup("B")
-	want := Entries{
+	want := l(Entries{
 		Entry{Hostname: "B", ContainerID: "bar"},
 		Entry{Hostname: "B", ContainerID: "foo"},
-	}
+	})
 	require.Equal(t, have, want)
 }
+
+func TestGossipDataMerge(t *testing.T) {
+	g1 := GossipData{Entries: makeEntries("AcDf")}
+	g2 := GossipData{Entries: makeEntries("BEf")}
+
+	g3 := g1.Merge(&g2).(*GossipData)
+
+	require.Equal(t, GossipData{Entries: makeEntries("ABcDEf")}, *g3)
+}

nameserver/http.go

@@ -8,6 +8,7 @@ import (
 	"github.com/gorilla/mux"
 	"github.com/miekg/dns"
 
+	"github.com/weaveworks/weave/common/docker"
 	"github.com/weaveworks/weave/net/address"
 )
 
@@ -16,7 +17,7 @@ func (n *Nameserver) badRequest(w http.ResponseWriter, err error) {
 	n.infof("%v", err)
 }
 
-func (n *Nameserver) HandleHTTP(router *mux.Router) {
+func (n *Nameserver) HandleHTTP(router *mux.Router, dockerCli *docker.Client) {
 	router.Methods("GET").Path("/domain").HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		fmt.Fprint(w, n.domain)
 	})
@@ -34,12 +35,17 @@ func (n *Nameserver) HandleHTTP(router *mux.Router) {
 			return
 		}
 
+		if !dns.IsSubDomain(n.domain, hostname) {
+			n.infof("Ignoring registration %s %s %s (not a subdomain of %s)", hostname, ipStr, container, n.domain)
+			return
+		}
+
 		if err := n.AddEntry(hostname, container, n.ourName, ip); err != nil {
 			n.badRequest(w, fmt.Errorf("Unable to add entry: %v", err))
 			return
 		}
 
-		if r.FormValue("check-alive") == "true" && n.docker.IsContainerNotRunning(container) {
+		if r.FormValue("check-alive") == "true" && dockerCli != nil && dockerCli.IsContainerNotRunning(container) {
 			n.infof("container '%s' is not running: removing", container)
 			if err := n.Delete(hostname, container, ipStr, ip); err != nil {
 				n.infof("failed to remove: %v", err)

nameserver/nameserver.go

@@ -1,8 +1,6 @@
 package nameserver
 
 import (
-	"bytes"
-	"encoding/gob"
 	"fmt"
 	"sync"
 	"time"
@@ -10,9 +8,8 @@ import (
 	"github.com/miekg/dns"
 
 	. "github.com/weaveworks/weave/common"
-	"github.com/weaveworks/weave/common/docker"
+	"github.com/weaveworks/weave/mesh"
 	"github.com/weaveworks/weave/net/address"
-	"github.com/weaveworks/weave/router"
 )
 
 const (
@@ -37,30 +34,24 @@ const (
 // - Update is O(n) for now
 type Nameserver struct {
 	sync.RWMutex
-	ourName router.PeerName
-	domain  string
-	gossip  router.Gossip
-	docker  *docker.Client
-	entries Entries
-	peers   *router.Peers
-	quit    chan struct{}
+	ourName     mesh.PeerName
+	domain      string
+	gossip      mesh.Gossip
+	entries     Entries
+	isKnownPeer func(mesh.PeerName) bool
+	quit        chan struct{}
 }
 
-func New(ourName router.PeerName, peers *router.Peers, docker *docker.Client, domain string) *Nameserver {
-	ns := &Nameserver{
-		ourName: ourName,
-		domain:  dns.Fqdn(domain),
-		peers:   peers,
-		docker:  docker,
-		quit:    make(chan struct{}),
+func New(ourName mesh.PeerName, domain string, isKnownPeer func(mesh.PeerName) bool) *Nameserver {
+	return &Nameserver{
+		ourName:     ourName,
+		domain:      dns.Fqdn(domain),
+		isKnownPeer: isKnownPeer,
+		quit:        make(chan struct{}),
 	}
-	if peers != nil {
-		peers.OnGC(ns.PeerGone)
-	}
-	return ns
 }
 
-func (n *Nameserver) SetGossip(gossip router.Gossip) {
+func (n *Nameserver) SetGossip(gossip mesh.Gossip) {
 	n.gossip = gossip
 }
 
@@ -83,16 +74,16 @@ func (n *Nameserver) Stop() {
 }
 
 func (n *Nameserver) broadcastEntries(es ...Entry) error {
-	if n.gossip != nil {
-		return n.gossip.GossipBroadcast(&GossipData{
-			Entries:   Entries(es),
-			Timestamp: now(),
-		})
+	if n.gossip == nil || len(es) == 0 {
+		return nil
 	}
-	return nil
+	return n.gossip.GossipBroadcast(&GossipData{
+		Entries:   Entries(es),
+		Timestamp: now(),
+	})
 }
 
-func (n *Nameserver) AddEntry(hostname, containerid string, origin router.PeerName, addr address.Address) error {
+func (n *Nameserver) AddEntry(hostname, containerid string, origin mesh.PeerName, addr address.Address) error {
 	n.infof("adding entry %s -> %s", hostname, addr.String())
 	n.Lock()
 	entry := n.entries.add(hostname, containerid, origin, addr)
@@ -130,6 +121,8 @@ func (n *Nameserver) ReverseLookup(ip address.Address) (string, error) {
 	return match.Hostname, nil
 }
 
+func (n *Nameserver) ContainerStarted(ident string) {}
+
 func (n *Nameserver) ContainerDied(ident string) {
 	n.Lock()
 	entries := n.entries.tombstone(n.ourName, func(e *Entry) bool {
@@ -140,19 +133,17 @@ func (n *Nameserver) ContainerDied(ident string) {
 		return false
 	})
 	n.Unlock()
-	if len(entries) > 0 {
-		if err := n.broadcastEntries(entries...); err != nil {
-			n.errorf("failed to broadcast container %s death: %v", ident, err)
-		}
+	if err := n.broadcastEntries(entries...); err != nil {
+		n.errorf("failed to broadcast container %s death: %v", ident, err)
 	}
 }
 
-func (n *Nameserver) PeerGone(peer *router.Peer) {
+func (n *Nameserver) PeerGone(peer mesh.PeerName) {
 	n.infof("peer %s gone", peer.String())
 	n.Lock()
 	defer n.Unlock()
 	n.entries.filter(func(e *Entry) bool {
-		return e.Origin != peer.Name
+		return e.Origin != peer
 	})
 }
 
@@ -188,7 +179,7 @@ func (n *Nameserver) deleteTombstones() {
 	})
 }
 
-func (n *Nameserver) Gossip() router.GossipData {
+func (n *Nameserver) Gossip() mesh.GossipData {
 	n.RLock()
 	defer n.RUnlock()
 	gossip := &GossipData{
@@ -199,32 +190,25 @@ func (n *Nameserver) Gossip() router.GossipData {
 	return gossip
 }
 
-func (n *Nameserver) OnGossipUnicast(sender router.PeerName, msg []byte) error {
+func (n *Nameserver) OnGossipUnicast(sender mesh.PeerName, msg []byte) error {
 	return nil
 }
 
-func (n *Nameserver) receiveGossip(msg []byte) (router.GossipData, router.GossipData, error) {
+func (n *Nameserver) receiveGossip(msg []byte) (mesh.GossipData, mesh.GossipData, error) {
 	var gossip GossipData
-	if err := gob.NewDecoder(bytes.NewReader(msg)).Decode(&gossip); err != nil {
+	if err := gossip.Decode(msg); err != nil {
 		return nil, nil, err
 	}
-
 	if delta := gossip.Timestamp - now(); delta > gossipWindow || delta < -gossipWindow {
 		return nil, nil, fmt.Errorf("clock skew of %d detected", delta)
 	}
 
-	if err := gossip.Entries.check(); err != nil {
-		return nil, nil, err
-	}
-
 	n.Lock()
 	defer n.Unlock()
 
-	if n.peers != nil {
-		gossip.Entries.filter(func(e *Entry) bool {
-			return n.peers.Fetch(e.Origin) != nil
-		})
-	}
+	gossip.Entries.filter(func(e *Entry) bool {
+		return n.isKnownPeer(e.Origin)
+	})
 
 	newEntries := n.entries.merge(gossip.Entries)
 	if len(newEntries) > 0 {
@@ -235,14 +219,14 @@ func (n *Nameserver) receiveGossip(msg []byte) (router.GossipData, router.Gossip
 
 // merge received data into state and return "everything new I've
 // just learnt", or nil if nothing in the received data was new
-func (n *Nameserver) OnGossip(msg []byte) (router.GossipData, error) {
+func (n *Nameserver) OnGossip(msg []byte) (mesh.GossipData, error) {
 	newEntries, _, err := n.receiveGossip(msg)
 	return newEntries, err
 }
 
 // merge received data into state and return a representation of
 // the received data, for further propagation
-func (n *Nameserver) OnGossipBroadcast(_ router.PeerName, msg []byte) (router.GossipData, error) {
+func (n *Nameserver) OnGossipBroadcast(_ mesh.PeerName, msg []byte) (mesh.GossipData, error) {
 	_, entries, err := n.receiveGossip(msg)
 	return entries, err
 }

nameserver/nameserver_test.go

@@ -11,19 +11,22 @@ import (
 
 	"github.com/stretchr/testify/require"
 
+	"github.com/weaveworks/weave/mesh"
 	"github.com/weaveworks/weave/net/address"
-	"github.com/weaveworks/weave/router"
-	wt "github.com/weaveworks/weave/testing"
 	"github.com/weaveworks/weave/testing/gossip"
 )
 
+func makeNameserver(name mesh.PeerName) *Nameserver {
+	return New(name, "", func(mesh.PeerName) bool { return true })
+}
+
 func makeNetwork(size int) ([]*Nameserver, *gossip.TestRouter) {
 	gossipRouter := gossip.NewTestRouter(0.0)
 	nameservers := make([]*Nameserver, size)
 
 	for i := 0; i < size; i++ {
-		name, _ := router.PeerNameFromString(fmt.Sprintf("%02d:00:00:02:00:00", i))
-		nameserver := New(name, nil, nil, "")
+		name, _ := mesh.PeerNameFromString(fmt.Sprintf("%02d:00:00:02:00:00", i))
+		nameserver := makeNameserver(name)
 		nameserver.SetGossip(gossipRouter.Connect(nameserver.ourName, nameserver))
 		nameserver.Start()
 		nameservers[i] = nameserver
@@ -40,7 +43,7 @@ func stopNetwork(nameservers []*Nameserver, grouter *gossip.TestRouter) {
 }
 
 type pair struct {
-	origin router.PeerName
+	origin mesh.PeerName
 	addr   address.Address
 }
 
@@ -72,18 +75,16 @@ func (m mapping) Addrs() []address.Address {
 }
 
 func TestNameservers(t *testing.T) {
-	wt.RunWithTimeout(t, time.Minute, func() {
-		testNameservers(t)
-	})
-}
-
-func testNameservers(t *testing.T) {
 	//common.SetLogLevel("debug")
 
 	lookupTimeout := 10 // ms
-	nameservers, grouter := makeNetwork(50)
+	nameservers, grouter := makeNetwork(30)
 	defer stopNetwork(nameservers, grouter)
-	nameserversByName := map[router.PeerName]*Nameserver{}
+	// This subset will sometimes lose touch with some of the others
+	badNameservers := nameservers[25:]
+	// This subset will remain well-connected, and we will deal mainly with them
+	nameservers = nameservers[:25]
+	nameserversByName := map[mesh.PeerName]*Nameserver{}
 	for _, n := range nameservers {
 		nameserversByName[n.ourName] = n
 	}
@@ -106,7 +107,13 @@ func testNameservers(t *testing.T) {
 	addMapping := func() {
 		nameserver := nameservers[rand.Intn(len(nameservers))]
 		addr := address.Address(rand.Int31())
-		hostname := fmt.Sprintf("hostname%d", rand.Int63())
+		// Create a hostname which has some upper and lowercase letters,
+		// and a unique number so we don't have to check if we allocated it already
+		randomBits := rand.Int63()
+		firstLetter := 'H' + (randomBits&1)*32
+		secondLetter := 'O' + (randomBits&2)*16
+		randomBits = randomBits >> 2
+		hostname := fmt.Sprintf("%c%cstname%d", firstLetter, secondLetter, randomBits)
 		mapping := mapping{hostname, []pair{{nameserver.ourName, addr}}}
 		mappings = append(mappings, mapping)
 
@@ -129,6 +136,12 @@ func testNameservers(t *testing.T) {
 		check(nameserver, mapping)
 	}
 
+	loseConnection := func() {
+		nameserver1 := badNameservers[rand.Intn(len(badNameservers))]
+		nameserver2 := nameservers[rand.Intn(len(nameservers))]
+		nameserver1.PeerGone(nameserver2.ourName)
+	}
+
 	deleteMapping := func() {
 		if len(mappings) <= 0 {
 			return
@@ -179,7 +192,7 @@ func testNameservers(t *testing.T) {
 		require.Equal(t, mapping.hostname, hostname)
 	}
 
-	for i := 0; i < 1000; i++ {
+	for i := 0; i < 800; i++ {
 		r := rand.Float32()
 		switch {
 		case r < 0.1:
@@ -191,7 +204,10 @@ func testNameservers(t *testing.T) {
 		case 0.2 <= r && r < 0.3:
 			deleteMapping()
 
-		case 0.3 <= r && r < 0.9:
+		case 0.3 <= r && r < 0.35:
+			loseConnection()
+
+		case 0.35 <= r && r < 0.9:
 			doLookup()
 
 		case 0.9 <= r:
@@ -203,9 +219,9 @@ func testNameservers(t *testing.T) {
 }
 
 func TestContainerAndPeerDeath(t *testing.T) {
-	peername, err := router.PeerNameFromString("00:00:00:02:00:00")
+	peername, err := mesh.PeerNameFromString("00:00:00:02:00:00")
 	require.Nil(t, err)
-	nameserver := New(peername, nil, nil, "")
+	nameserver := makeNameserver(peername)
 
 	err = nameserver.AddEntry("hostname", "containerid", peername, address.Address(0))
 	require.Nil(t, err)
@@ -218,7 +234,7 @@ func TestContainerAndPeerDeath(t *testing.T) {
 	require.Nil(t, err)
 	require.Equal(t, []address.Address{0}, nameserver.Lookup("hostname"))
 
-	nameserver.PeerGone(&router.Peer{Name: peername})
+	nameserver.PeerGone(peername)
 	require.Equal(t, []address.Address{}, nameserver.Lookup("hostname"))
 }
 
@@ -227,9 +243,9 @@ func TestTombstoneDeletion(t *testing.T) {
 	defer func() { now = oldNow }()
 	now = func() int64 { return 1234 }
 
-	peername, err := router.PeerNameFromString("00:00:00:02:00:00")
+	peername, err := mesh.PeerNameFromString("00:00:00:02:00:00")
 	require.Nil(t, err)
-	nameserver := New(peername, nil, nil, "")
+	nameserver := makeNameserver(peername)
 
 	err = nameserver.AddEntry("hostname", "containerid", peername, address.Address(0))
 	require.Nil(t, err)
@@ -241,14 +257,14 @@ func TestTombstoneDeletion(t *testing.T) {
 	err = nameserver.Delete("hostname", "containerid", "", address.Address(0))
 	require.Nil(t, err)
 	require.Equal(t, []address.Address{}, nameserver.Lookup("hostname"))
-	require.Equal(t, Entries{{
+	require.Equal(t, l(Entries{Entry{
 		ContainerID: "containerid",
 		Origin:      peername,
 		Addr:        address.Address(0),
 		Hostname:    "hostname",
 		Version:     1,
 		Tombstone:   1234,
-	}}, nameserver.entries)
+	}}), nameserver.entries)
 
 	now = func() int64 { return 1234 + int64(tombstoneTimeout/time.Second) + 1 }
 	nameserver.deleteTombstones()

nameserver/status.go

@@ -1,10 +1,11 @@
 package nameserver
 
 type Status struct {
-	Domain  string
-	Address string
-	TTL     uint32
-	Entries []EntryStatus
+	Domain   string
+	Upstream []string
+	Address  string
+	TTL      uint32
+	Entries  []EntryStatus
 }
 
 type EntryStatus struct {
@@ -37,6 +38,7 @@ func NewStatus(ns *Nameserver, dnsServer *DNSServer) *Status {
 
 	return &Status{
 		dnsServer.domain,
+		dnsServer.upstream.Servers,
 		dnsServer.address,
 		dnsServer.ttl,
 		entryStatusSlice}

net/address/address.go

@@ -19,7 +19,7 @@ func NewRange(start Address, size Offset) Range {
 	return Range{Start: start, End: Add(start, size)}
 }
 func (r Range) Size() Offset               { return Subtract(r.End, r.Start) }
-func (r Range) String() string             { return fmt.Sprintf("[%s-%s)", r.Start, r.End) }
+func (r Range) String() string             { return fmt.Sprintf("%s-%s", r.Start, r.End-1) }
 func (r Range) Overlaps(or Range) bool     { return !(r.Start >= or.End || r.End <= or.Start) }
 func (r Range) Contains(addr Address) bool { return addr >= r.Start && addr < r.End }
 

net/if.go

@@ -5,28 +5,33 @@ import (
 	"net"
 	"syscall"
 
-	"github.com/vishvananda/netlink/nl"
+	"github.com/vishvananda/netlink"
 )
 
 // Wait for an interface to come up.
 func EnsureInterface(ifaceName string) (*net.Interface, error) {
-	s, err := nl.Subscribe(syscall.NETLINK_ROUTE, syscall.RTNLGRP_LINK)
-	if err != nil {
-		return nil, err
-	}
-	defer s.Close()
-	iface, err := ensureInterface(s, ifaceName)
+	iface, err := ensureInterface(ifaceName)
 	if err != nil {
 		return nil, err
 	}
 	return iface, err
 }
 
-func ensureInterface(s *nl.NetlinkSocket, ifaceName string) (*net.Interface, error) {
+func ensureInterface(ifaceName string) (*net.Interface, error) {
 	if iface, err := findInterface(ifaceName); err == nil {
 		return iface, nil
 	}
-	netlinkReceiveUntil(s, matchIfName(ifaceName))
+	ch := make(chan netlink.LinkUpdate)
+	done := make(chan struct{})
+	defer close(done)
+	if err := netlink.LinkSubscribe(ch, done); err != nil {
+		return nil, err
+	}
+	for update := range ch {
+		if ifaceName == update.Link.Attrs().Name && update.IfInfomsg.Flags&syscall.IFF_UP != 0 {
+			break
+		}
+	}
 	iface, err := findInterface(ifaceName)
 	return iface, err
 }
@@ -35,12 +40,7 @@ func ensureInterface(s *nl.NetlinkSocket, ifaceName string) (*net.Interface, err
 // This matches the behaviour in 'weave attach', which is the only context in which
 // we expect this to be called.  If you change one, change the other to match.
 func EnsureInterfaceAndMcastRoute(ifaceName string) (*net.Interface, error) {
-	s, err := nl.Subscribe(syscall.NETLINK_ROUTE, syscall.RTNLGRP_LINK, syscall.RTNLGRP_IPV4_ROUTE)
-	if err != nil {
-		return nil, err
-	}
-	defer s.Close()
-	iface, err := ensureInterface(s, ifaceName)
+	iface, err := ensureInterface(ifaceName)
 	if err != nil {
 		return nil, err
 	}
@@ -48,8 +48,19 @@ func EnsureInterfaceAndMcastRoute(ifaceName string) (*net.Interface, error) {
 	if CheckRouteExists(ifaceName, dest) {
 		return iface, err
 	}
-	netlinkReceiveUntil(s, matchRoute(ifaceName, dest))
-	return iface, err
+	ch := make(chan netlink.RouteUpdate)
+	done := make(chan struct{})
+	defer close(done)
+	if err := netlink.RouteSubscribe(ch, done); err != nil {
+		return nil, err
+	}
+	for update := range ch {
+		link, _ := netlink.LinkByIndex(update.Route.LinkIndex)
+		if link.Attrs().Name == ifaceName && update.Route.Dst.IP.Equal(dest) {
+			break
+		}
+	}
+	return iface, nil
 }
 
 func findInterface(ifaceName string) (iface *net.Interface, err error) {
@@ -61,40 +72,3 @@ func findInterface(ifaceName string) (iface *net.Interface, err error) {
 	}
 	return
 }
-
-func netlinkReceiveUntil(s *nl.NetlinkSocket, f func(syscall.NetlinkMessage) (bool, error)) error {
-	for {
-		msgs, err := s.Receive()
-		if err != nil {
-			return err
-		}
-		for _, m := range msgs {
-			if done, err := f(m); done {
-				return err
-			}
-		}
-	}
-}
-
-func matchIfName(ifaceName string) func(m syscall.NetlinkMessage) (bool, error) {
-	return func(m syscall.NetlinkMessage) (bool, error) {
-		switch m.Header.Type {
-		case syscall.RTM_NEWLINK: // receive this type for link 'up'
-			ifmsg := nl.DeserializeIfInfomsg(m.Data)
-			attrs, err := syscall.ParseNetlinkRouteAttr(&m)
-			if err != nil {
-				return true, err
-			}
-			name := ""
-			for _, attr := range attrs {
-				if attr.Attr.Type == syscall.IFA_LABEL {
-					name = string(attr.Value[:len(attr.Value)-1])
-				}
-			}
-			if ifaceName == name && ifmsg.Flags&syscall.IFF_UP != 0 {
-				return true, nil
-			}
-		}
-		return false, nil
-	}
-}

net/route.go

@@ -3,10 +3,8 @@ package net
 import (
 	"fmt"
 	"net"
-	"syscall"
 
 	"github.com/vishvananda/netlink"
-	"github.com/vishvananda/netlink/nl"
 )
 
 // A network is considered free if it does not overlap any existing
@@ -65,51 +63,3 @@ func CheckRouteExists(ifaceName string, dest net.IP) bool {
 	})
 	return found
 }
-
-func matchRoute(ifaceName string, dest net.IP) func(m syscall.NetlinkMessage) (bool, error) {
-	return func(m syscall.NetlinkMessage) (bool, error) {
-		switch m.Header.Type {
-		case syscall.RTM_NEWROUTE:
-			route, err := deserializeRoute(m.Data)
-			if err != nil {
-				return true, err
-			}
-			link, _ := netlink.LinkByIndex(route.LinkIndex)
-			if link.Attrs().Name == ifaceName && route.Dst.IP.Equal(dest) {
-				return true, nil
-			}
-		}
-		return false, nil
-	}
-}
-
-// This code proposed for addition to the vishvananda/netlink library
-// deserializeRoute decodes a binary netlink message into a Route struct
-func deserializeRoute(m []byte) (netlink.Route, error) {
-	route := netlink.Route{}
-	msg := nl.DeserializeRtMsg(m)
-	attrs, err := nl.ParseRouteAttr(m[msg.Len():])
-	if err != nil {
-		return route, err
-	}
-	route.Scope = netlink.Scope(msg.Scope)
-
-	native := nl.NativeEndian()
-	for _, attr := range attrs {
-		switch attr.Attr.Type {
-		case syscall.RTA_GATEWAY:
-			route.Gw = net.IP(attr.Value)
-		case syscall.RTA_PREFSRC:
-			route.Src = net.IP(attr.Value)
-		case syscall.RTA_DST:
-			route.Dst = &net.IPNet{
-				IP:   attr.Value,
-				Mask: net.CIDRMask(int(msg.Dst_len), 8*len(attr.Value)),
-			}
-		case syscall.RTA_OIF:
-			routeIndex := int(native.Uint32(attr.Value[0:4]))
-			route.LinkIndex = routeIndex
-		}
-	}
-	return route, nil
-}

plugin/README.md

@@ -0,0 +1,9 @@
+# Weave network driver extension for Docker
+
+A [plugin](http://docs.docker.com/engine/extend/plugin_api/) to
+integrate [weave Net](http://weave.works/net/) with Docker.
+
+The plugin runs automatically when you `weave launch`, provided your
+Docker daemon is version 1.9 or newer.
+
+More detail [here](http://docs.weave.works/weave/latest_release/plugin.html).

plugin/ipam/driver.go

@@ -0,0 +1,58 @@
+package ipamplugin
+
+import (
+	"net"
+
+	"github.com/docker/libnetwork/ipamapi"
+	"github.com/docker/libnetwork/netlabel"
+	"github.com/weaveworks/weave/api"
+	. "github.com/weaveworks/weave/common"
+	"github.com/weaveworks/weave/common/docker"
+)
+
+const (
+	WeaveContainer = "weave"
+)
+
+type ipam struct {
+	client *docker.Client
+	weave  *api.Client
+}
+
+func NewIpam(client *docker.Client, version string) (ipamapi.Ipam, error) {
+	resolver := func() (string, error) { return client.GetContainerIP(WeaveContainer) }
+	return &ipam{client: client, weave: api.NewClientWithResolver(resolver)}, nil
+}
+
+func (i *ipam) GetDefaultAddressSpaces() (string, string, error) {
+	Log.Debugln("GetDefaultAddressSpaces")
+	return "weavelocal", "weaveglobal", nil
+}
+
+func (i *ipam) RequestPool(addressSpace, pool, subPool string, options map[string]string, v6 bool) (string, *net.IPNet, map[string]string, error) {
+	Log.Debugln("RequestPool", addressSpace, pool, subPool, options)
+	cidr, err := i.weave.DefaultSubnet()
+	Log.Debugln("RequestPool returning ", cidr, err)
+	// Pass back a fake "gateway address"; we don't actually use it,
+	// so just give the network address.
+	data := map[string]string{netlabel.Gateway: cidr.String()}
+	return "weavepool", cidr, data, err
+}
+
+func (i *ipam) ReleasePool(poolID string) error {
+	Log.Debugln("ReleasePool", poolID)
+	return nil
+}
+
+func (i *ipam) RequestAddress(poolID string, address net.IP, options map[string]string) (*net.IPNet, map[string]string, error) {
+	Log.Debugln("RequestAddress", poolID, address, options)
+	// Pass magic string to weave IPAM, which then stores the address under its own string
+	ip, err := i.weave.AllocateIP("_")
+	Log.Debugln("allocateIP returned", ip, err)
+	return ip, nil, err
+}
+
+func (i *ipam) ReleaseAddress(poolID string, address net.IP) error {
+	Log.Debugln("ReleaseAddress", poolID, address)
+	return i.weave.ReleaseIP(address.String())
+}

plugin/net/driver.go

@@ -0,0 +1,206 @@
+package plugin
+
+import (
+	"fmt"
+	"sync"
+
+	"github.com/docker/libnetwork/drivers/remote/api"
+	"github.com/docker/libnetwork/types"
+
+	. "github.com/weaveworks/weave/common"
+	"github.com/weaveworks/weave/common/docker"
+	"github.com/weaveworks/weave/common/odp"
+	"github.com/weaveworks/weave/plugin/skel"
+
+	"github.com/vishvananda/netlink"
+)
+
+const (
+	WeaveContainer = "weave"
+	WeaveBridge    = "weave"
+)
+
+type driver struct {
+	version          string
+	nameserver       string
+	scope            string
+	noMulticastRoute bool
+	sync.RWMutex
+	endpoints map[string]struct{}
+}
+
+func New(client *docker.Client, version string, nameserver string, scope string, noMulticastRoute bool) (skel.Driver, error) {
+	driver := &driver{
+		nameserver:       nameserver,
+		noMulticastRoute: noMulticastRoute,
+		version:          version,
+		scope:            scope,
+		endpoints:        make(map[string]struct{}),
+	}
+
+	_, err := NewWatcher(client, driver)
+	if err != nil {
+		return nil, err
+	}
+	return driver, nil
+}
+
+func errorf(format string, a ...interface{}) error {
+	Log.Errorf(format, a...)
+	return fmt.Errorf(format, a...)
+}
+
+// === protocol handlers
+
+func (driver *driver) GetCapabilities() (*api.GetCapabilityResponse, error) {
+	var caps = &api.GetCapabilityResponse{
+		Scope: driver.scope,
+	}
+	Log.Debugf("Get capabilities: responded with %+v", caps)
+	return caps, nil
+}
+
+func (driver *driver) CreateNetwork(create *api.CreateNetworkRequest) error {
+	Log.Debugf("Create network request %+v", create)
+	Log.Infof("Create network %s", create.NetworkID)
+	return nil
+}
+
+func (driver *driver) DeleteNetwork(delete *api.DeleteNetworkRequest) error {
+	Log.Debugf("Delete network request: %+v", delete)
+	Log.Infof("Destroy network %s", delete.NetworkID)
+	return nil
+}
+
+func (driver *driver) CreateEndpoint(create *api.CreateEndpointRequest) (*api.CreateEndpointResponse, error) {
+	Log.Debugf("Create endpoint request %+v", create)
+	endID := create.EndpointID
+
+	if create.Interface == nil {
+		return nil, fmt.Errorf("Not supported: creating an interface from within CreateEndpoint")
+	}
+	driver.Lock()
+	driver.endpoints[endID] = struct{}{}
+	driver.Unlock()
+	resp := &api.CreateEndpointResponse{}
+
+	Log.Infof("Create endpoint %s %+v", endID, resp)
+	return resp, nil
+}
+
+func (driver *driver) DeleteEndpoint(deleteReq *api.DeleteEndpointRequest) error {
+	Log.Debugf("Delete endpoint request: %+v", deleteReq)
+	Log.Infof("Delete endpoint %s", deleteReq.EndpointID)
+	driver.Lock()
+	delete(driver.endpoints, deleteReq.EndpointID)
+	driver.Unlock()
+	return nil
+}
+
+func (driver *driver) HasEndpoint(endpointID string) bool {
+	driver.Lock()
+	_, found := driver.endpoints[endpointID]
+	driver.Unlock()
+	return found
+}
+
+func (driver *driver) EndpointInfo(req *api.EndpointInfoRequest) (*api.EndpointInfoResponse, error) {
+	Log.Debugf("Endpoint info request: %+v", req)
+	Log.Infof("Endpoint info %s", req.EndpointID)
+	return &api.EndpointInfoResponse{Value: map[string]interface{}{}}, nil
+}
+
+func (driver *driver) JoinEndpoint(j *api.JoinRequest) (*api.JoinResponse, error) {
+	endID := j.EndpointID
+
+	maybeBridge, err := netlink.LinkByName(WeaveBridge)
+	if err != nil {
+		return nil, errorf(`bridge "%s" not present; did you launch weave?`, WeaveBridge)
+	}
+
+	// create and attach local name to the bridge
+	local := vethPair(endID[:5])
+	local.Attrs().MTU = maybeBridge.Attrs().MTU
+	if err := netlink.LinkAdd(local); err != nil {
+		return nil, errorf("could not create veth pair: %s", err)
+	}
+
+	switch maybeBridge.(type) {
+	case *netlink.Bridge:
+		if err := netlink.LinkSetMasterByIndex(local, maybeBridge.Attrs().Index); err != nil {
+			return nil, errorf(`unable to set master: %s`, err)
+		}
+	case *netlink.GenericLink:
+		if maybeBridge.Type() != "openvswitch" {
+			Log.Errorf("device %s is %+v", WeaveBridge, maybeBridge)
+			return nil, errorf(`device "%s" is of type "%s"`, WeaveBridge, maybeBridge.Type())
+		}
+		odp.AddDatapathInterface(WeaveBridge, local.Name)
+	case *netlink.Device:
+		Log.Warnf("kernel does not report what kind of device %s is, just %+v", WeaveBridge, maybeBridge)
+		// Assume it's our openvswitch device, and the kernel has not been updated to report the kind.
+		odp.AddDatapathInterface(WeaveBridge, local.Name)
+	default:
+		Log.Errorf("device %s is %+v", WeaveBridge, maybeBridge)
+		return nil, errorf(`device "%s" not a bridge`, WeaveBridge)
+	}
+	if err := netlink.LinkSetUp(local); err != nil {
+		return nil, errorf(`unable to bring veth up: %s`, err)
+	}
+
+	ifname := &api.InterfaceName{
+		SrcName:   local.PeerName,
+		DstPrefix: "ethwe",
+	}
+
+	response := &api.JoinResponse{
+		InterfaceName: ifname,
+	}
+	if driver.nameserver != "" {
+		routeToDNS := api.StaticRoute{
+			Destination: driver.nameserver + "/32",
+			RouteType:   types.CONNECTED,
+			NextHop:     "",
+		}
+		response.StaticRoutes = []api.StaticRoute{routeToDNS}
+	}
+	if !driver.noMulticastRoute {
+		multicastRoute := api.StaticRoute{
+			Destination: "224.0.0.0/4",
+			RouteType:   types.CONNECTED,
+		}
+		response.StaticRoutes = append(response.StaticRoutes, multicastRoute)
+	}
+	Log.Infof("Join endpoint %s:%s to %s", j.NetworkID, j.EndpointID, j.SandboxKey)
+	return response, nil
+}
+
+func (driver *driver) LeaveEndpoint(leave *api.LeaveRequest) error {
+	Log.Debugf("Leave request: %+v", leave)
+
+	local := vethPair(leave.EndpointID[:5])
+	if err := netlink.LinkDel(local); err != nil {
+		Log.Warningf("unable to delete veth on leave: %s", err)
+	}
+	Log.Infof("Leave %s:%s", leave.NetworkID, leave.EndpointID)
+	return nil
+}
+
+func (driver *driver) DiscoverNew(disco *api.DiscoveryNotification) error {
+	Log.Debugf("Dicovery new notification: %+v", disco)
+	return nil
+}
+
+func (driver *driver) DiscoverDelete(disco *api.DiscoveryNotification) error {
+	Log.Debugf("Dicovery delete notification: %+v", disco)
+	return nil
+}
+
+// ===
+
+func vethPair(suffix string) *netlink.Veth {
+	return &netlink.Veth{
+		LinkAttrs: netlink.LinkAttrs{Name: "vethwl" + suffix},
+		PeerName:  "vethwg" + suffix,
+	}
+}

plugin/net/watcher.go

@@ -0,0 +1,57 @@
+package plugin
+
+import (
+	"fmt"
+
+	weaveapi "github.com/weaveworks/weave/api"
+	. "github.com/weaveworks/weave/common"
+	"github.com/weaveworks/weave/common/docker"
+)
+
+const (
+	WeaveDomain = "weave.local"
+)
+
+type watcher struct {
+	client *docker.Client
+	weave  *weaveapi.Client
+	driver *driver
+}
+
+type Watcher interface {
+}
+
+func NewWatcher(client *docker.Client, driver *driver) (Watcher, error) {
+	resolver := func() (string, error) { return client.GetContainerIP(WeaveContainer) }
+	w := &watcher{client: client, driver: driver,
+		weave: weaveapi.NewClientWithResolver(resolver)}
+	err := client.AddObserver(w)
+	if err != nil {
+		return nil, err
+	}
+
+	return w, nil
+}
+
+func (w *watcher) ContainerStarted(id string) {
+	Log.Debugf("Container started %s", id)
+	info, err := w.client.InspectContainer(id)
+	if err != nil {
+		Log.Warningf("error inspecting container: %s", err)
+		return
+	}
+	// check that it's on our network, via the endpointID
+	for _, net := range info.NetworkSettings.Networks {
+		if w.driver.HasEndpoint(net.EndpointID) {
+			fqdn := fmt.Sprintf("%s.%s", info.Config.Hostname, info.Config.Domainname)
+			if err := w.weave.RegisterWithDNS(id, fqdn, net.IPAddress); err != nil {
+				Log.Warningf("unable to register with weaveDNS: %s", err)
+			}
+		}
+	}
+}
+
+func (w *watcher) ContainerDied(id string) {
+	// don't need to do this as WeaveDNS removes names on container died anyway
+	// (note by the time we get this event we can't see the EndpointID)
+}

plugin/skel/listener.go

@@ -0,0 +1,298 @@
+package skel
+
+import (
+	"encoding/json"
+	"fmt"
+	"net"
+	"net/http"
+
+	"github.com/gorilla/mux"
+
+	"github.com/docker/libnetwork/drivers/remote/api"
+	"github.com/docker/libnetwork/ipamapi"
+	iapi "github.com/docker/libnetwork/ipams/remote/api"
+)
+
+const (
+	networkReceiver = "NetworkDriver"
+	ipamReceiver    = ipamapi.PluginEndpointType
+)
+
+type Driver interface {
+	GetCapabilities() (*api.GetCapabilityResponse, error)
+	CreateNetwork(create *api.CreateNetworkRequest) error
+	DeleteNetwork(delete *api.DeleteNetworkRequest) error
+	CreateEndpoint(create *api.CreateEndpointRequest) (*api.CreateEndpointResponse, error)
+	DeleteEndpoint(delete *api.DeleteEndpointRequest) error
+	EndpointInfo(req *api.EndpointInfoRequest) (*api.EndpointInfoResponse, error)
+	JoinEndpoint(j *api.JoinRequest) (response *api.JoinResponse, error error)
+	LeaveEndpoint(leave *api.LeaveRequest) error
+	DiscoverNew(discover *api.DiscoveryNotification) error
+	DiscoverDelete(delete *api.DiscoveryNotification) error
+}
+
+type listener struct {
+	d Driver
+	i ipamapi.Ipam
+}
+
+func Listen(socket net.Listener, driver Driver, ipamDriver ipamapi.Ipam) error {
+	router := mux.NewRouter()
+	router.NotFoundHandler = http.HandlerFunc(notFound)
+
+	listener := &listener{driver, ipamDriver}
+
+	router.Methods("POST").Path("/Plugin.Activate").HandlerFunc(listener.handshake)
+
+	handleMethod := func(receiver, method string, h http.HandlerFunc) {
+		router.Methods("POST").Path(fmt.Sprintf("/%s.%s", receiver, method)).HandlerFunc(h)
+	}
+
+	handleMethod(networkReceiver, "GetCapabilities", listener.getCapabilities)
+
+	if driver != nil {
+		handleMethod(networkReceiver, "CreateNetwork", listener.createNetwork)
+		handleMethod(networkReceiver, "DeleteNetwork", listener.deleteNetwork)
+		handleMethod(networkReceiver, "CreateEndpoint", listener.createEndpoint)
+		handleMethod(networkReceiver, "DeleteEndpoint", listener.deleteEndpoint)
+		handleMethod(networkReceiver, "EndpointOperInfo", listener.infoEndpoint)
+		handleMethod(networkReceiver, "Join", listener.joinEndpoint)
+		handleMethod(networkReceiver, "Leave", listener.leaveEndpoint)
+	}
+
+	if ipamDriver != nil {
+		handleMethod(ipamReceiver, "GetDefaultAddressSpaces", listener.getDefaultAddressSpaces)
+		handleMethod(ipamReceiver, "RequestPool", listener.requestPool)
+		handleMethod(ipamReceiver, "ReleasePool", listener.releasePool)
+		handleMethod(ipamReceiver, "RequestAddress", listener.requestAddress)
+		handleMethod(ipamReceiver, "ReleaseAddress", listener.releaseAddress)
+	}
+
+	return http.Serve(socket, router)
+}
+
+func decode(w http.ResponseWriter, r *http.Request, v interface{}) error {
+	err := json.NewDecoder(r.Body).Decode(v)
+	if err != nil {
+		sendError(w, "Unable to decode JSON payload: "+err.Error(), http.StatusBadRequest)
+	}
+	return err
+}
+
+// === protocol handlers
+
+type handshakeResp struct {
+	Implements []string
+}
+
+func (listener *listener) handshake(w http.ResponseWriter, r *http.Request) {
+	var resp handshakeResp
+	if listener.d != nil {
+		resp.Implements = append(resp.Implements, "NetworkDriver")
+	}
+	if listener.i != nil {
+		resp.Implements = append(resp.Implements, "IpamDriver")
+	}
+	err := json.NewEncoder(w).Encode(&resp)
+	if err != nil {
+		sendError(w, "encode error", http.StatusInternalServerError)
+		return
+	}
+}
+
+func (listener *listener) getCapabilities(w http.ResponseWriter, r *http.Request) {
+	caps, err := listener.d.GetCapabilities()
+	objectOrErrorResponse(w, caps, err)
+}
+
+func (listener *listener) createNetwork(w http.ResponseWriter, r *http.Request) {
+	var create api.CreateNetworkRequest
+	err := json.NewDecoder(r.Body).Decode(&create)
+	if err != nil {
+		sendError(w, "Unable to decode JSON payload: "+err.Error(), http.StatusBadRequest)
+		return
+	}
+	emptyOrErrorResponse(w, listener.d.CreateNetwork(&create))
+}
+
+func (listener *listener) deleteNetwork(w http.ResponseWriter, r *http.Request) {
+	var delete api.DeleteNetworkRequest
+	if err := json.NewDecoder(r.Body).Decode(&delete); err != nil {
+		sendError(w, "Unable to decode JSON payload: "+err.Error(), http.StatusBadRequest)
+		return
+	}
+	emptyOrErrorResponse(w, listener.d.DeleteNetwork(&delete))
+}
+
+func (listener *listener) createEndpoint(w http.ResponseWriter, r *http.Request) {
+	var create api.CreateEndpointRequest
+	if err := json.NewDecoder(r.Body).Decode(&create); err != nil {
+		sendError(w, "unable to decode JSON payload: "+err.Error(), http.StatusBadRequest)
+		return
+	}
+	res, err := listener.d.CreateEndpoint(&create)
+	objectOrErrorResponse(w, res, err)
+}
+
+func (listener *listener) deleteEndpoint(w http.ResponseWriter, r *http.Request) {
+	var delete api.DeleteEndpointRequest
+	if err := json.NewDecoder(r.Body).Decode(&delete); err != nil {
+		sendError(w, "Could not decode JSON encode payload", http.StatusBadRequest)
+		return
+	}
+	emptyOrErrorResponse(w, listener.d.DeleteEndpoint(&delete))
+}
+
+func (listener *listener) infoEndpoint(w http.ResponseWriter, r *http.Request) {
+	var req api.EndpointInfoRequest
+	if err := json.NewDecoder(r.Body).Decode(&req); err != nil {
+		sendError(w, "Could not decode JSON encode payload", http.StatusBadRequest)
+		return
+	}
+	info, err := listener.d.EndpointInfo(&req)
+	objectOrErrorResponse(w, info, err)
+}
+
+func (listener *listener) joinEndpoint(w http.ResponseWriter, r *http.Request) {
+	var join api.JoinRequest
+	if err := json.NewDecoder(r.Body).Decode(&join); err != nil {
+		sendError(w, "Could not decode JSON encode payload", http.StatusBadRequest)
+		return
+	}
+	res, err := listener.d.JoinEndpoint(&join)
+	objectOrErrorResponse(w, res, err)
+}
+
+func (listener *listener) leaveEndpoint(w http.ResponseWriter, r *http.Request) {
+	var l api.LeaveRequest
+	if err := json.NewDecoder(r.Body).Decode(&l); err != nil {
+		sendError(w, "Could not decode JSON encode payload", http.StatusBadRequest)
+		return
+	}
+	emptyOrErrorResponse(w, listener.d.LeaveEndpoint(&l))
+}
+
+func (listener *listener) discoverNew(w http.ResponseWriter, r *http.Request) {
+	var disco api.DiscoveryNotification
+	if err := json.NewDecoder(r.Body).Decode(&disco); err != nil {
+		sendError(w, "Could not decode JSON encode payload", http.StatusBadRequest)
+		return
+	}
+	emptyOrErrorResponse(w, listener.d.DiscoverNew(&disco))
+}
+
+func (listener *listener) discoverDelete(w http.ResponseWriter, r *http.Request) {
+	var disco api.DiscoveryNotification
+	if err := json.NewDecoder(r.Body).Decode(&disco); err != nil {
+		sendError(w, "Could not decode JSON encode payload", http.StatusBadRequest)
+		return
+	}
+	emptyOrErrorResponse(w, listener.d.DiscoverDelete(&disco))
+}
+
+// ===
+
+func (listener *listener) getDefaultAddressSpaces(w http.ResponseWriter, r *http.Request) {
+	local, global, err := listener.i.GetDefaultAddressSpaces()
+	response := &iapi.GetAddressSpacesResponse{
+		LocalDefaultAddressSpace:  local,
+		GlobalDefaultAddressSpace: global,
+	}
+	objectOrErrorResponse(w, response, err)
+}
+
+func (listener *listener) requestPool(w http.ResponseWriter, r *http.Request) {
+	var rq iapi.RequestPoolRequest
+	if err := decode(w, r, &rq); err != nil {
+		return
+	}
+	poolID, pool, data, err := listener.i.RequestPool(rq.AddressSpace, rq.Pool, rq.SubPool, rq.Options, rq.V6)
+	if err != nil {
+		errorResponse(w, err.Error())
+		return
+	}
+	response := &iapi.RequestPoolResponse{
+		PoolID: poolID,
+		Pool:   pool.String(),
+		Data:   data,
+	}
+	objectResponse(w, response)
+}
+
+func (listener *listener) releasePool(w http.ResponseWriter, r *http.Request) {
+	var rq iapi.ReleasePoolRequest
+	if err := decode(w, r, &rq); err != nil {
+		return
+	}
+	err := listener.i.ReleasePool(rq.PoolID)
+	emptyOrErrorResponse(w, err)
+}
+
+func (listener *listener) requestAddress(w http.ResponseWriter, r *http.Request) {
+	var rq iapi.RequestAddressRequest
+	if err := decode(w, r, &rq); err != nil {
+		return
+	}
+	address, data, err := listener.i.RequestAddress(rq.PoolID, net.ParseIP(rq.Address), rq.Options)
+	if err != nil {
+		errorResponse(w, err.Error())
+		return
+	}
+	response := &iapi.RequestAddressResponse{
+		Address: address.String(),
+		Data:    data,
+	}
+	objectResponse(w, response)
+}
+
+func (listener *listener) releaseAddress(w http.ResponseWriter, r *http.Request) {
+	var rq iapi.ReleaseAddressRequest
+	if err := decode(w, r, &rq); err != nil {
+		return
+	}
+	err := listener.i.ReleaseAddress(rq.PoolID, net.ParseIP(rq.Address))
+	emptyOrErrorResponse(w, err)
+}
+
+// ===
+
+func notFound(w http.ResponseWriter, r *http.Request) {
+	http.NotFound(w, r)
+}
+
+func sendError(w http.ResponseWriter, msg string, code int) {
+	http.Error(w, msg, code)
+}
+
+func errorResponse(w http.ResponseWriter, fmtString string, item ...interface{}) {
+	json.NewEncoder(w).Encode(map[string]string{
+		"Err": fmt.Sprintf(fmtString, item...),
+	})
+}
+
+func objectResponse(w http.ResponseWriter, obj interface{}) {
+	if err := json.NewEncoder(w).Encode(obj); err != nil {
+		sendError(w, "Could not JSON encode response", http.StatusInternalServerError)
+		return
+	}
+}
+
+func emptyResponse(w http.ResponseWriter) {
+	json.NewEncoder(w).Encode(map[string]string{})
+}
+
+func objectOrErrorResponse(w http.ResponseWriter, obj interface{}, err error) {
+	if err != nil {
+		errorResponse(w, err.Error())
+		return
+	}
+	objectResponse(w, obj)
+}
+
+func emptyOrErrorResponse(w http.ResponseWriter, err error) {
+	if err != nil {
+		errorResponse(w, err.Error())
+		return
+	}
+	emptyResponse(w)
+}

prog/docker_tls_args/main.go

@@ -0,0 +1,77 @@
+// docker_tls_args: find the docker daemon's tls args
+// This reimplements pgrep, because pgrep will only look in /proc, but not in
+// $PROCFS.
+package main
+
+import (
+	"bytes"
+	"fmt"
+	"io/ioutil"
+	"os"
+	"path/filepath"
+	"strconv"
+	"strings"
+)
+
+func main() {
+	procRoot := os.Getenv("PROCFS")
+	if procRoot == "" {
+		procRoot = "/proc"
+	}
+
+	dirEntries, err := ioutil.ReadDir(procRoot)
+	checkErr(err)
+
+	for _, dirEntry := range dirEntries {
+		dirName := dirEntry.Name()
+		if _, err := strconv.Atoi(dirName); err != nil {
+			continue
+		}
+
+		if comm, err := ioutil.ReadFile(filepath.Join(procRoot, dirName, "comm")); err != nil || string(comm) != "docker\n" {
+			continue
+		}
+
+		cmdline, err := ioutil.ReadFile(filepath.Join(procRoot, dirName, "cmdline"))
+		if err != nil {
+			continue
+		}
+
+		isDaemon := false
+		tlsArgs := []string{}
+		args := bytes.Split(cmdline, []byte{'\000'})
+		for i := 0; i < len(args); i++ {
+			arg := string(args[i])
+			switch {
+			case arg == "-d" || arg == "daemon":
+				isDaemon = true
+				break
+			case arg == "--tls", arg == "--tlsverify":
+				tlsArgs = append(tlsArgs, arg)
+			case strings.HasPrefix(arg, "--tls"):
+				tlsArgs = append(tlsArgs, arg)
+				if len(args) > i+1 &&
+					!strings.Contains(arg, "=") &&
+					!strings.HasPrefix(string(args[i+1]), "-") {
+					tlsArgs = append(tlsArgs, string(args[i+1]))
+					i++
+				}
+			}
+		}
+		if !isDaemon {
+			continue
+		}
+
+		fmt.Println(strings.Join(tlsArgs, " "))
+		os.Exit(0)
+	}
+
+	os.Exit(1)
+}
+
+func checkErr(err error) {
+	if err != nil {
+		fmt.Fprintln(os.Stderr, err)
+		os.Exit(1)
+	}
+}

prog/plugin/Dockerfile

@@ -0,0 +1,3 @@
+FROM weaveworks/weaveexec
+ADD ./plugin /home/weave/
+ENTRYPOINT ["/home/weave/plugin"]

prog/plugin/main.go

@@ -0,0 +1,163 @@
+package main
+
+import (
+	"flag"
+	"fmt"
+	"net"
+	"os"
+	"os/signal"
+	"strings"
+	"syscall"
+
+	"github.com/docker/libnetwork/ipamapi"
+	go_docker "github.com/fsouza/go-dockerclient"
+	. "github.com/weaveworks/weave/common"
+	"github.com/weaveworks/weave/common/docker"
+	ipamplugin "github.com/weaveworks/weave/plugin/ipam"
+	netplugin "github.com/weaveworks/weave/plugin/net"
+	"github.com/weaveworks/weave/plugin/skel"
+)
+
+var version = "(unreleased version)"
+
+func main() {
+	var (
+		justVersion      bool
+		address          string
+		nameserver       string
+		meshAddress      string
+		logLevel         string
+		meshNetworkName  string
+		noMulticastRoute bool
+		removeNetwork    bool
+	)
+
+	flag.BoolVar(&justVersion, "version", false, "print version and exit")
+	flag.StringVar(&logLevel, "log-level", "info", "logging level (debug, info, warning, error)")
+	flag.StringVar(&address, "socket", "/run/docker/plugins/weave.sock", "socket on which to listen")
+	flag.StringVar(&nameserver, "nameserver", "", "nameserver to provide to containers")
+	flag.StringVar(&meshAddress, "meshsocket", "/run/docker/plugins/weavemesh.sock", "socket on which to listen in mesh mode")
+	flag.StringVar(&meshNetworkName, "mesh-network-name", "weave", "network name to create in mesh mode")
+	flag.BoolVar(&noMulticastRoute, "no-multicast-route", false, "do not add a multicast route to network endpoints")
+	flag.BoolVar(&removeNetwork, "remove-network", false, "remove mesh network and exit")
+
+	flag.Parse()
+
+	if justVersion {
+		fmt.Printf("weave plugin %s\n", version)
+		os.Exit(0)
+	}
+
+	SetLogLevel(logLevel)
+
+	// API 1.21 is the first version that supports docker network commands
+	dockerClient, err := docker.NewVersionedClientFromEnv("1.21")
+	if err != nil {
+		Log.Fatalf("unable to connect to docker: %s", err)
+	}
+
+	if removeNetwork {
+		if _, err = dockerClient.Client.NetworkInfo(meshNetworkName); err == nil {
+			err = dockerClient.Client.RemoveNetwork(meshNetworkName)
+			if err != nil {
+				Log.Fatalf("unable to remove network: %s", err)
+			}
+		}
+		os.Exit(0)
+	}
+
+	Log.Println("Weave plugin", version, "Command line options:", os.Args[1:])
+	Log.Info(dockerClient.Info())
+
+	var globalListener, meshListener net.Listener
+	endChan := make(chan error, 1)
+	if address != "" {
+		globalListener, err := listenAndServe(dockerClient, address, nameserver, noMulticastRoute, endChan, "global", false)
+		if err != nil {
+			Log.Fatalf("unable to create driver: %s", err)
+		}
+		defer globalListener.Close()
+	}
+	if meshAddress != "" {
+		meshListener, err := listenAndServe(dockerClient, meshAddress, nameserver, noMulticastRoute, endChan, "local", true)
+		if err != nil {
+			Log.Fatalf("unable to create driver: %s", err)
+		}
+		defer meshListener.Close()
+	}
+
+	if meshNetworkName != "" {
+		createNetwork(dockerClient, meshNetworkName, meshAddress)
+	}
+
+	sigChan := make(chan os.Signal, 1)
+	signal.Notify(sigChan, os.Interrupt, os.Kill, syscall.SIGTERM)
+
+	select {
+	case sig := <-sigChan:
+		Log.Debugf("Caught signal %s; shutting down", sig)
+		if err := dockerClient.Client.RemoveNetwork(meshNetworkName); err != nil {
+			Log.Fatal(err)
+		}
+	case err := <-endChan:
+		if err != nil {
+			Log.Errorf("Error from listener: %s", err)
+			globalListener.Close()
+			meshListener.Close()
+			os.Exit(1)
+		}
+	}
+}
+
+func listenAndServe(dockerClient *docker.Client, address, nameserver string, noMulticastRoute bool, endChan chan<- error, scope string, withIpam bool) (net.Listener, error) {
+	d, err := netplugin.New(dockerClient, version, nameserver, scope, noMulticastRoute)
+	if err != nil {
+		return nil, err
+	}
+
+	var i ipamapi.Ipam
+	if withIpam {
+		if i, err = ipamplugin.NewIpam(dockerClient, version); err != nil {
+			return nil, err
+		}
+	}
+
+	var listener net.Listener
+
+	// remove sockets from last invocation
+	if err := os.Remove(address); err != nil && !os.IsNotExist(err) {
+		return nil, err
+	}
+	listener, err = net.Listen("unix", address)
+	if err != nil {
+		return nil, err
+	}
+	Log.Printf("Listening on %s for %s scope", address, scope)
+
+	go func() {
+		endChan <- skel.Listen(listener, d, i)
+	}()
+
+	return listener, nil
+}
+
+func createNetwork(dockerClient *docker.Client, networkName, address string) {
+	if _, err := dockerClient.Client.NetworkInfo(networkName); err == nil {
+		Log.Printf("Docker network '%s' already exists", networkName)
+	} else if _, ok := err.(*go_docker.NoSuchNetwork); ok {
+		driverName := strings.TrimSuffix(address, ".sock")
+		if i := strings.LastIndex(driverName, "/"); i >= 0 {
+			driverName = driverName[i+1:]
+		}
+		options := go_docker.CreateNetworkOptions{
+			Name:           networkName,
+			CheckDuplicate: true,
+			Driver:         driverName,
+			IPAM:           go_docker.IPAMOptions{Driver: driverName},
+		}
+		_, err := dockerClient.Client.CreateNetwork(options)
+		if err != nil {
+			Log.Fatalf("Error creating network: %s", err)
+		}
+	}
+}

prog/weaveexec/Dockerfile

@@ -1,10 +1,12 @@
-FROM gliderlabs/alpine
+FROM alpine
 
 MAINTAINER Weaveworks Inc <help@weave.works>
 LABEL works.weave.role=system
 
 WORKDIR /home/weave
 VOLUME /w
+VOLUME /w-noop
+VOLUME /w-nomcast
 ENTRYPOINT ["/home/weave/sigproxy", "/home/weave/weave"]
 
 RUN apk add --update \
@@ -17,7 +19,9 @@ RUN apk add --update \
     bind-tools \
   && rm -rf /var/cache/apk/*
 
-ADD ./weave ./sigproxy ./weaveproxy /home/weave/
-ADD ./netcheck ./weavehosts /usr/bin/
+ADD ./weave ./sigproxy ./weaveproxy ./symlink /home/weave/
+ADD ./netcheck ./docker_tls_args /usr/bin/
 ADD ./weavewait /w/w
+ADD ./weavewait_noop /w-noop/w
+ADD ./weavewait_nomcast /w-nomcast/w
 ADD ./docker.tgz /

prog/weaveexec/symlink

@@ -0,0 +1,18 @@
+#!/bin/sh
+# This lets us force-symlink a bunch of stuff all at once, so we can minimize
+# running containers.
+
+set -e
+
+[ -z "$WEAVE_DEBUG" ] || set -x
+
+target="$1"
+user="$(stat -c '%u:%g' "$target")"
+mode="$(stat -c '%a' "$target")"
+shift 1
+
+for link in "$@" ; do
+  ln -sfT "$target" "$link"
+done
+chown -h "$user" "$@"
+chmod "$mode" "$@"

prog/weavehosts/weavehosts.go

@@ -1,120 +0,0 @@
-/* weavehosts: rewrite /etc/hosts as necessary to put the weave address(es) at the top */
-package main
-
-import (
-	"bufio"
-	"bytes"
-	"fmt"
-	"io/ioutil"
-	"net"
-	"os"
-	"sort"
-	"strings"
-)
-
-func main() {
-	args := os.Args
-	if len(args) < 4 {
-		fmt.Fprintln(os.Stderr, "usage: weavehosts hostpath hostname cidr [cidr ...]")
-		os.Exit(1)
-	}
-
-	checkErr(updateHosts(args[1], args[2], args[3:]))
-
-	os.Exit(0)
-}
-
-func checkErr(err error) {
-	if err != nil {
-		fmt.Fprintln(os.Stderr, err)
-		os.Exit(1)
-	}
-}
-
-func updateHosts(path, hostname string, cidrs []string) error {
-	hosts, err := parseHosts(path)
-	if err != nil {
-		return err
-	}
-
-	// Remove existing ips pointing to our hostname
-	toRemove := []string{}
-	for ip, addrs := range hosts {
-		for _, addr := range addrs {
-			if addr == hostname {
-				toRemove = append(toRemove, ip)
-				break
-			}
-		}
-	}
-	for _, ip := range toRemove {
-		delete(hosts, ip)
-	}
-
-	// Add the weave ip(s)
-	for _, cidr := range cidrs {
-		ip, _, err := net.ParseCIDR(cidr)
-		checkErr(err)
-		ipStr := ip.String()
-		hosts[ipStr] = append(hosts[ipStr], hostname)
-	}
-
-	return writeHosts(path, hosts)
-}
-
-func parseHosts(path string) (map[string][]string, error) {
-	f, err := os.Open(path)
-	if err != nil {
-		return nil, err
-	}
-	defer f.Close()
-	ips := map[string][]string{}
-	scanner := bufio.NewScanner(f)
-	for scanner.Scan() {
-		line := scanner.Text()
-
-		// Remove any comments
-		if i := strings.IndexByte(line, '#'); i != -1 {
-			line = line[:i]
-		}
-
-		fields := strings.Fields(line)
-		if len(fields) > 0 {
-			ips[fields[0]] = append(ips[fields[0]], fields[1:]...)
-		}
-	}
-	if scanner.Err() != nil {
-		return nil, scanner.Err()
-	}
-	return ips, nil
-}
-
-func writeHosts(path string, contents map[string][]string) error {
-	ips := []string{}
-	for ip := range contents {
-		ips = append(ips, ip)
-	}
-	sort.Strings(ips)
-
-	buf := &bytes.Buffer{}
-	fmt.Fprintln(buf, "# modified by weave")
-	for _, ip := range ips {
-		if addrs := contents[ip]; len(addrs) > 0 {
-			fmt.Fprintf(buf, "%s\t%s\n", ip, strings.Join(uniqueStrs(addrs), " "))
-		}
-	}
-	return ioutil.WriteFile(path, buf.Bytes(), 644)
-}
-
-func uniqueStrs(s []string) []string {
-	m := map[string]struct{}{}
-	result := []string{}
-	for _, str := range s {
-		if _, ok := m[str]; !ok {
-			m[str] = struct{}{}
-			result = append(result, str)
-		}
-	}
-	sort.Strings(result)
-	return result
-}

prog/weaveproxy/main.go

@@ -12,14 +12,16 @@ import (
 )
 
 var (
-	version = "(unreleased version)"
+	version           = "(unreleased version)"
+	defaultDockerHost = "unix:///var/run/docker.sock"
 )
 
 func main() {
 	var (
 		justVersion bool
 		logLevel    = "info"
-		c           = proxy.Config{ListenAddrs: []string{}}
+		c           = proxy.Config{Image: "weaveworks/weaveexec"}
+		withDNS     bool
 	)
 
 	c.Version = version
@@ -35,32 +37,45 @@ func main() {
 	mflag.BoolVar(&c.NoRewriteHosts, []string{"-no-rewrite-hosts"}, false, "do not automatically rewrite /etc/hosts. Use if you need the docker IP to remain in /etc/hosts")
 	mflag.StringVar(&c.TLSConfig.CACert, []string{"#tlscacert", "-tlscacert"}, "", "Trust certs signed only by this CA")
 	mflag.StringVar(&c.TLSConfig.Cert, []string{"#tlscert", "-tlscert"}, "", "Path to TLS certificate file")
-	mflag.BoolVar(&c.TLSConfig.Enabled, []string{"#tls", "-tls"}, false, "Use TLS; implied by --tls-verify")
+	mflag.BoolVar(&c.TLSConfig.Enabled, []string{"#tls", "-tls"}, false, "Use TLS; implied by --tlsverify")
 	mflag.StringVar(&c.TLSConfig.Key, []string{"#tlskey", "-tlskey"}, "", "Path to TLS key file")
 	mflag.BoolVar(&c.TLSConfig.Verify, []string{"#tlsverify", "-tlsverify"}, false, "Use TLS and verify the remote")
-	mflag.BoolVar(&c.WithDNS, []string{"-with-dns", "w"}, false, "instruct created containers to always use weaveDNS as their nameserver")
+	mflag.BoolVar(&withDNS, []string{"#-with-dns", "#w"}, false, "option removed")
 	mflag.BoolVar(&c.WithoutDNS, []string{"-without-dns"}, false, "instruct created containers to never use weaveDNS as their nameserver")
+	mflag.BoolVar(&c.NoMulticastRoute, []string{"-no-multicast-route"}, false, "do not add a multicast route via the weave interface when attaching containers")
 	mflag.Parse()
 
 	if justVersion {
-		fmt.Printf("weave proxy %s\n", version)
+		fmt.Printf("weave proxy  %s\n", version)
 		os.Exit(0)
 	}
 
-	if c.WithDNS && c.WithoutDNS {
-		Log.Fatalf("Cannot use both '--with-dns' and '--without-dns' flags")
-	}
-
 	SetLogLevel(logLevel)
 
+	if image := os.Getenv("EXEC_IMAGE"); image != "" {
+		c.Image = image
+	}
+
 	Log.Infoln("weave proxy", version)
 	Log.Infoln("Command line arguments:", strings.Join(os.Args[1:], " "))
 
+	if withDNS {
+		Log.Warning("--with-dns option has been removed; DNS is on by default")
+	}
+
+	c.DockerHost = defaultDockerHost
+	if dockerHost := os.Getenv("DOCKER_HOST"); dockerHost != "" {
+		c.DockerHost = dockerHost
+	}
+
 	p, err := proxy.NewProxy(c)
 	if err != nil {
 		Log.Fatalf("Could not start proxy: %s", err)
 	}
 
-	go p.ListenAndServe()
+	listeners := p.Listen()
+	p.AttachExistingContainers()
+	go p.Serve(listeners)
+	go p.ListenAndServeStatus("/home/weave/status.sock")
 	SignalHandlerLoop()
 }

prog/weaver/http.go

@@ -1,17 +1,20 @@
 package main
 
 import (
+	"bytes"
 	"encoding/json"
 	"fmt"
-	"github.com/gorilla/mux"
-	. "github.com/weaveworks/weave/common"
-	"github.com/weaveworks/weave/ipam"
-	"github.com/weaveworks/weave/nameserver"
-	"github.com/weaveworks/weave/net/address"
-	weave "github.com/weaveworks/weave/router"
 	"net/http"
 	"strings"
 	"text/template"
+
+	"github.com/gorilla/mux"
+	. "github.com/weaveworks/weave/common"
+	"github.com/weaveworks/weave/ipam"
+	"github.com/weaveworks/weave/mesh"
+	"github.com/weaveworks/weave/nameserver"
+	"github.com/weaveworks/weave/net/address"
+	weave "github.com/weaveworks/weave/router"
 )
 
 var rootTemplate = template.New("root").Funcs(map[string]interface{}{
@@ -24,14 +27,74 @@ var rootTemplate = template.New("root").Funcs(map[string]interface{}{
 		}
 		return count
 	},
-	"printConnectionCounts": func(conns []weave.LocalConnectionStatus) string {
+	"printList": func(list []string) string {
+		if len(list) == 0 {
+			return "none"
+		}
+		return strings.Join(list, ", ")
+	},
+	"printIPAMRanges": func(router weave.NetworkRouterStatus, status ipam.Status) string {
+		var buffer bytes.Buffer
+
+		type stats struct {
+			ips       uint32
+			nickname  string
+			reachable bool
+		}
+
+		peerStats := make(map[string]*stats)
+
+		for _, entry := range status.Entries {
+			s, found := peerStats[entry.Peer]
+			if !found {
+				s = &stats{nickname: entry.Nickname, reachable: entry.IsKnownPeer}
+				peerStats[entry.Peer] = s
+			}
+			s.ips += entry.Size
+		}
+
+		printOwned := func(name string, nickName string, reachable bool, ips uint32) {
+			reachableStr := ""
+			if !reachable {
+				reachableStr = "- unreachable!"
+			}
+			percentageRanges := float32(ips) * 100.0 / float32(status.RangeNumIPs)
+
+			displayName := name + "(" + nickName + ")"
+			fmt.Fprintf(&buffer, "%-37v %8d IPs (%04.1f%% of total) %s\n",
+				displayName, ips, percentageRanges, reachableStr)
+		}
+
+		// print the local info first
+		if ourStats := peerStats[router.Name]; ourStats != nil {
+			printOwned(router.Name, ourStats.nickname, true, ourStats.ips)
+		}
+
+		// and then the rest
+		for peer, stats := range peerStats {
+			if peer != router.Name {
+				printOwned(peer, stats.nickname, stats.reachable, stats.ips)
+			}
+		}
+
+		return buffer.String()
+	},
+	"allIPAMOwnersUnreachable": func(status ipam.Status) bool {
+		for _, entry := range status.Entries {
+			if entry.Size > 0 && entry.IsKnownPeer {
+				return false
+			}
+		}
+		return true
+	},
+	"printConnectionCounts": func(conns []mesh.LocalConnectionStatus) string {
 		counts := make(map[string]int)
 		for _, conn := range conns {
 			counts[conn.State]++
 		}
 		return printCounts(counts, []string{"established", "pending", "retrying", "failed", "connecting"})
 	},
-	"printPeerConnectionCounts": func(peers []weave.PeerStatus) string {
+	"printPeerConnectionCounts": func(peers []mesh.PeerStatus) string {
 		counts := make(map[string]int)
 		for _, peer := range peers {
 			for _, conn := range peer.Connections {
@@ -75,40 +138,48 @@ func defTemplate(name string, text string) *template.Template {
 }
 
 var statusTemplate = defTemplate("status", `\
-       Version: {{.Version}}
+        Version: {{.Version}}
 
-       Service: router
-      Protocol: {{.Router.Protocol}} \
+        Service: router
+       Protocol: {{.Router.Protocol}} \
 {{if eq .Router.ProtocolMinVersion .Router.ProtocolMaxVersion}}\
 {{.Router.ProtocolMaxVersion}}\
 {{else}}\
 {{.Router.ProtocolMinVersion}}..{{.Router.ProtocolMaxVersion}}\
 {{end}}
-          Name: {{.Router.Name}}({{.Router.NickName}})
-    Encryption: {{printState .Router.Encryption}}
- PeerDiscovery: {{printState .Router.PeerDiscovery}}
-       Targets: {{len .Router.Targets}}
-   Connections: {{len .Router.Connections}}{{with printConnectionCounts .Router.Connections}} ({{.}}){{end}}
-         Peers: {{len .Router.Peers}}{{with printPeerConnectionCounts .Router.Peers}} (with {{.}} connections between them){{end}}
+           Name: {{.Router.Name}}({{.Router.NickName}})
+     Encryption: {{printState .Router.Encryption}}
+  PeerDiscovery: {{printState .Router.PeerDiscovery}}
+        Targets: {{len .Router.Targets}}
+    Connections: {{len .Router.Connections}}{{with printConnectionCounts .Router.Connections}} ({{.}}){{end}}
+          Peers: {{len .Router.Peers}}{{with printPeerConnectionCounts .Router.Peers}} (with {{.}} connections){{end}}
+ TrustedSubnets: {{printList .Router.TrustedSubnets}}
 {{if .IPAM}}\
 
-       Service: ipam
+        Service: ipam
 {{if .IPAM.Entries}}\
-     Consensus: achieved
-{{else if .IPAM.Paxos}}\
-     Consensus: waiting (quorum: {{.IPAM.Paxos.Quorum}}, known: {{.IPAM.Paxos.KnownNodes}})
+{{if allIPAMOwnersUnreachable .IPAM}}\
+         Status: all IP ranges owned by unreachable peers - use 'rmpeer' if they are dead
+{{else if len .IPAM.PendingAllocates}}\
+         Status: waiting for IP range grant from peers
 {{else}}\
-     Consensus: deferred
+         Status: ready
 {{end}}\
-         Range: {{.IPAM.Range}}
- DefaultSubnet: {{.IPAM.DefaultSubnet}}
+{{else if .IPAM.Paxos}}\
+         Status: awaiting consensus (quorum: {{.IPAM.Paxos.Quorum}}, known: {{.IPAM.Paxos.KnownNodes}})
+{{else}}\
+         Status: idle
+{{end}}\
+          Range: {{.IPAM.Range}}
+  DefaultSubnet: {{.IPAM.DefaultSubnet}}
 {{end}}\
 {{if .DNS}}\
 
-       Service: dns
-        Domain: {{.DNS.Domain}}
-           TTL: {{.DNS.TTL}}
-       Entries: {{countDNSEntries .DNS.Entries}}
+        Service: dns
+         Domain: {{.DNS.Domain}}
+       Upstream: {{printList .DNS.Upstream}}
+            TTL: {{.DNS.TTL}}
+        Entries: {{countDNSEntries .DNS.Entries}}
 {{end}}\
 `)
 
@@ -128,7 +199,7 @@ var peersTemplate = defTemplate("peers", `\
 {{.Name}}({{.NickName}})
 {{range .Connections}}\
    {{if .Outbound}}->{{else}}<-{{end}} {{printf "%-21v" .Address}} \
-{{$nameNickName := printf "%v(%v)" .Name .NickName}}{{printf "%-32v" $nameNickName}} \
+{{$nameNickName := printf "%v(%v)" .Name .NickName}}{{printf "%-37v" $nameNickName}} \
 {{if .Established}}established{{else}}pending{{end}}
 {{end}}\
 {{end}}\
@@ -144,31 +215,44 @@ var dnsEntriesTemplate = defTemplate("dnsEntries", `\
 {{end}}\
 `)
 
+var ipamTemplate = defTemplate("ipamTemplate", `{{printIPAMRanges .Router .IPAM}}`)
+
 type WeaveStatus struct {
 	Version string
-	Router  *weave.Status      `json:"Router,omitempty"`
-	IPAM    *ipam.Status       `json:"IPAM,omitempty"`
-	DNS     *nameserver.Status `json:"DNS,omitempty"`
+	Router  *weave.NetworkRouterStatus `json:"Router,omitempty"`
+	IPAM    *ipam.Status               `json:"IPAM,omitempty"`
+	DNS     *nameserver.Status         `json:"DNS,omitempty"`
 }
 
-func HandleHTTP(muxRouter *mux.Router, version string, router *weave.Router, allocator *ipam.Allocator, defaultSubnet address.CIDR, ns *nameserver.Nameserver, dnsserver *nameserver.DNSServer) {
+func HandleHTTP(muxRouter *mux.Router, version string, router *weave.NetworkRouter, allocator *ipam.Allocator, defaultSubnet address.CIDR, ns *nameserver.Nameserver, dnsserver *nameserver.DNSServer) {
 	status := func() WeaveStatus {
 		return WeaveStatus{
 			version,
-			weave.NewStatus(router),
+			weave.NewNetworkRouterStatus(router),
 			ipam.NewStatus(allocator, defaultSubnet),
 			nameserver.NewStatus(ns, dnsserver)}
 	}
 	muxRouter.Methods("GET").Path("/report").Headers("Accept", "application/json").HandlerFunc(
 		func(w http.ResponseWriter, r *http.Request) {
-			json, _ := json.MarshalIndent(status(), "", "    ")
+			json, err := json.MarshalIndent(status(), "", "    ")
+			if err != nil {
+				http.Error(w, err.Error(), http.StatusInternalServerError)
+				Log.Error("Error during report marshalling: ", err)
+				return
+			}
 			w.Header().Set("Content-Type", "application/json")
 			w.Write(json)
 		})
 
 	muxRouter.Methods("GET").Path("/report").Queries("format", "{format}").HandlerFunc(
 		func(w http.ResponseWriter, r *http.Request) {
-			formatTemplate, err := template.New("format").Parse(mux.Vars(r)["format"])
+			funcs := template.FuncMap{
+				"json": func(v interface{}) string {
+					a, _ := json.Marshal(v)
+					return string(a)
+				},
+			}
+			formatTemplate, err := template.New("format").Funcs(funcs).Parse(mux.Vars(r)["format"])
 			if err != nil {
 				http.Error(w, err.Error(), http.StatusBadRequest)
 				return
@@ -194,5 +278,6 @@ func HandleHTTP(muxRouter *mux.Router, version string, router *weave.Router, all
 	defHandler("/status/connections", connectionsTemplate)
 	defHandler("/status/peers", peersTemplate)
 	defHandler("/status/dns", dnsEntriesTemplate)
+	defHandler("/status/ipam", ipamTemplate)
 
 }

prog/weaver/main.go

@@ -1,7 +1,6 @@
 package main
 
 import (
-	"crypto/sha256"
 	"fmt"
 	"net"
 	"net/http"
@@ -17,7 +16,9 @@ import (
 
 	. "github.com/weaveworks/weave/common"
 	"github.com/weaveworks/weave/common/docker"
+	"github.com/weaveworks/weave/common/odp"
 	"github.com/weaveworks/weave/ipam"
+	"github.com/weaveworks/weave/mesh"
 	"github.com/weaveworks/weave/nameserver"
 	weavenet "github.com/weaveworks/weave/net"
 	"github.com/weaveworks/weave/net/address"
@@ -36,36 +37,54 @@ func main() {
 	runtime.GOMAXPROCS(procs)
 
 	var (
-		config             weave.Config
-		justVersion        bool
-		protocolMinVersion int
-		ifaceName          string
-		routerName         string
-		nickName           string
-		password           string
-		pktdebug           bool
-		logLevel           string
-		prof               string
-		bufSzMB            int
-		noDiscovery        bool
-		httpAddr           string
-		iprangeCIDR        string
-		ipsubnetCIDR       string
-		peerCount          int
-		apiPath            string
-		peers              []string
+		// flags that cause immediate exit
+		justVersion          bool
+		createDatapath       bool
+		deleteDatapath       bool
+		addDatapathInterface string
 
+		config                    mesh.Config
+		networkConfig             weave.NetworkConfig
+		protocolMinVersion        int
+		ifaceName                 string
+		routerName                string
+		nickName                  string
+		password                  string
+		pktdebug                  bool
+		logLevel                  string
+		prof                      string
+		bufSzMB                   int
+		noDiscovery               bool
+		httpAddr                  string
+		iprangeCIDR               string
+		ipsubnetCIDR              string
+		peerCount                 int
+		dockerAPI                 string
+		peers                     []string
 		noDNS                     bool
 		dnsDomain                 string
 		dnsListenAddress          string
 		dnsTTL                    int
 		dnsClientTimeout          time.Duration
 		dnsEffectiveListenAddress string
+		iface                     *net.Interface
+		datapathName              string
+		trustedSubnetStr          string
+
+		defaultDockerHost = "unix:///var/run/docker.sock"
 	)
 
+	if val := os.Getenv("DOCKER_HOST"); val != "" {
+		defaultDockerHost = val
+	}
+
 	mflag.BoolVar(&justVersion, []string{"#version", "-version"}, false, "print version and exit")
-	mflag.IntVar(&config.Port, []string{"#port", "-port"}, weave.Port, "router port")
-	mflag.IntVar(&protocolMinVersion, []string{"-min-protocol-version"}, weave.ProtocolMinVersion, "minimum weave protocol version")
+	mflag.BoolVar(&createDatapath, []string{"-create-datapath"}, false, "create ODP datapath and exit")
+	mflag.BoolVar(&deleteDatapath, []string{"-delete-datapath"}, false, "delete ODP datapath and exit")
+	mflag.StringVar(&addDatapathInterface, []string{"-add-datapath-iface"}, "", "add a network interface to the ODP datapath and exit")
+
+	mflag.IntVar(&config.Port, []string{"#port", "-port"}, mesh.Port, "router port")
+	mflag.IntVar(&protocolMinVersion, []string{"-min-protocol-version"}, mesh.ProtocolMinVersion, "minimum weave protocol version")
 	mflag.StringVar(&ifaceName, []string{"#iface", "-iface"}, "", "name of interface to capture/inject from (disabled if blank)")
 	mflag.StringVar(&routerName, []string{"#name", "-name"}, "", "name of router (defaults to MAC of interface)")
 	mflag.StringVar(&nickName, []string{"#nickname", "-nickname"}, "", "nickname of peer (defaults to hostname)")
@@ -76,17 +95,20 @@ func main() {
 	mflag.IntVar(&config.ConnLimit, []string{"#connlimit", "#-connlimit", "-conn-limit"}, 30, "connection limit (0 for unlimited)")
 	mflag.BoolVar(&noDiscovery, []string{"#nodiscovery", "#-nodiscovery", "-no-discovery"}, false, "disable peer discovery")
 	mflag.IntVar(&bufSzMB, []string{"#bufsz", "-bufsz"}, 8, "capture buffer size in MB")
-	mflag.StringVar(&httpAddr, []string{"#httpaddr", "#-httpaddr", "-http-addr"}, fmt.Sprintf(":%d", weave.HTTPPort), "address to bind HTTP interface to (disabled if blank, absolute path indicates unix domain socket)")
+	mflag.StringVar(&httpAddr, []string{"#httpaddr", "#-httpaddr", "-http-addr"}, "", "address to bind HTTP interface to (disabled if blank, absolute path indicates unix domain socket)")
 	mflag.StringVar(&iprangeCIDR, []string{"#iprange", "#-iprange", "-ipalloc-range"}, "", "IP address range reserved for automatic allocation, in CIDR notation")
 	mflag.StringVar(&ipsubnetCIDR, []string{"#ipsubnet", "#-ipsubnet", "-ipalloc-default-subnet"}, "", "subnet to allocate within by default, in CIDR notation")
 	mflag.IntVar(&peerCount, []string{"#initpeercount", "#-initpeercount", "-init-peer-count"}, 0, "number of peers in network (for IP address allocation)")
-	mflag.StringVar(&apiPath, []string{"#api", "-api"}, "unix:///var/run/docker.sock", "Path to Docker API socket")
+	mflag.StringVar(&dockerAPI, []string{"#api", "#-api", "-docker-api"}, defaultDockerHost, "Docker API endpoint")
 	mflag.BoolVar(&noDNS, []string{"-no-dns"}, false, "disable DNS server")
 	mflag.StringVar(&dnsDomain, []string{"-dns-domain"}, nameserver.DefaultDomain, "local domain to server requests for")
 	mflag.StringVar(&dnsListenAddress, []string{"-dns-listen-address"}, nameserver.DefaultListenAddress, "address to listen on for DNS requests")
 	mflag.IntVar(&dnsTTL, []string{"-dns-ttl"}, nameserver.DefaultTTL, "TTL for DNS request from our domain")
 	mflag.DurationVar(&dnsClientTimeout, []string{"-dns-fallback-timeout"}, nameserver.DefaultClientTimeout, "timeout for fallback DNS requests")
 	mflag.StringVar(&dnsEffectiveListenAddress, []string{"-dns-effective-listen-address"}, "", "address DNS will actually be listening, after Docker port mapping")
+	mflag.StringVar(&datapathName, []string{"-datapath"}, "", "ODP datapath name")
+
+	mflag.StringVar(&trustedSubnetStr, []string{"-trusted-subnets"}, "", "Command separated list of trusted subnets in CIDR notation")
 
 	// crude way of detecting that we probably have been started in a
 	// container, with `weave launch` --> suppress misleading paths in
@@ -95,59 +117,115 @@ func main() {
 		os.Args[0] = "weave"
 		mflag.CommandLine.Init("weave", mflag.ExitOnError)
 	}
+
 	mflag.Parse()
 
 	peers = mflag.Args()
 
 	SetLogLevel(logLevel)
-	if justVersion {
+
+	switch {
+	case justVersion:
 		fmt.Printf("weave router %s\n", version)
 		os.Exit(0)
+
+	case createDatapath:
+		err, odp_supported := odp.CreateDatapath(datapathName)
+		if !odp_supported {
+			if err != nil {
+				Log.Error(err)
+			}
+
+			// When the kernel lacks ODP support, exit
+			// with a special status to distinguish it for
+			// the weave script.
+			os.Exit(17)
+		}
+
+		checkFatal(err)
+		os.Exit(0)
+
+	case deleteDatapath:
+		checkFatal(odp.DeleteDatapath(datapathName))
+		os.Exit(0)
+
+	case addDatapathInterface != "":
+		checkFatal(odp.AddDatapathInterface(datapathName, addDatapathInterface))
+		os.Exit(0)
 	}
 
 	Log.Println("Command line options:", options())
 	Log.Println("Command line peers:", peers)
 
-	if protocolMinVersion < weave.ProtocolMinVersion || protocolMinVersion > weave.ProtocolMaxVersion {
-		Log.Fatalf("--min-protocol-version must be in range [%d,%d]", weave.ProtocolMinVersion, weave.ProtocolMaxVersion)
+	if protocolMinVersion < mesh.ProtocolMinVersion || protocolMinVersion > mesh.ProtocolMaxVersion {
+		Log.Fatalf("--min-protocol-version must be in range [%d,%d]", mesh.ProtocolMinVersion, mesh.ProtocolMaxVersion)
 	}
 	config.ProtocolMinVersion = byte(protocolMinVersion)
 
-	var err error
+	var fastDPOverlay weave.NetworkOverlay
+	if datapathName != "" {
+		// A datapath name implies that "Bridge" and "Overlay"
+		// packet handling use fast datapath, although other
+		// options can override that below.  Even if both
+		// things are overridden, we might need bridging on
+		// the datapath.
+		fastdp, err := weave.NewFastDatapath(weave.FastDatapathConfig{
+			DatapathName: datapathName,
+			Port:         config.Port,
+		})
+
+		checkFatal(err)
+		networkConfig.Bridge = fastdp.Bridge()
+		fastDPOverlay = fastdp.Overlay()
+	}
 
 	if ifaceName != "" {
-		config.Iface, err = weavenet.EnsureInterface(ifaceName)
-		if err != nil {
-			Log.Fatal(err)
-		}
-	}
+		// -iface can coexist with -datapath, because
+		// pcap-based packet capture is a bit more efficient
+		// than capture via ODP misses, even when using an
+		// ODP-based bridge.  So when using weave encyption,
+		// it's preferable to use -iface.
+		var err error
+		iface, err = weavenet.EnsureInterface(ifaceName)
+		checkFatal(err)
 
-	if routerName == "" {
-		if config.Iface == nil {
-			Log.Fatal("Either an interface must be specified with --iface or a name with -name")
-		}
-		routerName = config.Iface.HardwareAddr.String()
-	}
-	name, err := weave.PeerNameFromUserInput(routerName)
-	if err != nil {
-		Log.Fatal(err)
-	}
-
-	if nickName == "" {
-		nickName, err = os.Hostname()
-		if err != nil {
-			Log.Fatal(err)
-		}
+		// bufsz flag is in MB
+		networkConfig.Bridge, err = weave.NewPcap(iface, bufSzMB*1024*1024)
+		checkFatal(err)
 	}
 
 	if password == "" {
 		password = os.Getenv("WEAVE_PASSWORD")
 	}
+
 	if password == "" {
 		Log.Println("Communication between peers is unencrypted.")
 	} else {
 		config.Password = []byte(password)
-		Log.Println("Communication between peers is encrypted.")
+		Log.Println("Communication between peers via untrusted networks is encrypted.")
+	}
+
+	overlays := weave.NewOverlaySwitch()
+	if fastDPOverlay != nil {
+		overlays.Add("fastdp", fastDPOverlay)
+	}
+	sleeve := weave.NewSleeveOverlay(config.Port)
+	overlays.Add("sleeve", sleeve)
+	overlays.SetCompatOverlay(sleeve)
+
+	if routerName == "" {
+		if iface == nil {
+			Log.Fatal("Either an interface must be specified with --iface or a name with -name")
+		}
+		routerName = iface.HardwareAddr.String()
+	}
+
+	name, err := mesh.PeerNameFromUserInput(routerName)
+	checkFatal(err)
+
+	if nickName == "" {
+		nickName, err = os.Hostname()
+		checkFatal(err)
 	}
 
 	if prof != "" {
@@ -157,25 +235,46 @@ func main() {
 		defer profile.Start(&p).Stop()
 	}
 
-	config.BufSz = bufSzMB * 1024 * 1024
-	config.LogFrame = logFrameFunc(pktdebug)
 	config.PeerDiscovery = !noDiscovery
 
-	router := weave.NewRouter(config, name, nickName)
-	Log.Println("Our name is", router.Ourself)
-
-	dockerCli, err := docker.NewClient(apiPath)
-	if err != nil {
-		Log.Fatal("Unable to start docker client: ", err)
+	if pktdebug {
+		networkConfig.PacketLogging = packetLogging{}
+	} else {
+		networkConfig.PacketLogging = nopPacketLogging{}
 	}
 
+	if config.TrustedSubnets, err = parseTrustedSubnets(trustedSubnetStr); err != nil {
+		Log.Fatal("Unable to parse trusted subnets: ", err)
+	}
+
+	router := weave.NewNetworkRouter(config, networkConfig, name, nickName, overlays)
+	Log.Println("Our name is", router.Ourself)
+
+	var dockerCli *docker.Client
+	if dockerAPI != "" {
+		dc, err := docker.NewClient(dockerAPI)
+		if err != nil {
+			Log.Fatal("Unable to start docker client: ", err)
+		} else {
+			Log.Info(dc.Info())
+		}
+		dockerCli = dc
+	}
+	observeContainers := func(o docker.ContainerObserver) {
+		if dockerCli != nil {
+			if err = dockerCli.AddObserver(o); err != nil {
+				Log.Fatal("Unable to start watcher", err)
+			}
+		}
+	}
+	isKnownPeer := func(name mesh.PeerName) bool {
+		return router.Peers.Fetch(name) != nil
+	}
 	var allocator *ipam.Allocator
 	var defaultSubnet address.CIDR
 	if iprangeCIDR != "" {
-		allocator, defaultSubnet = createAllocator(router, iprangeCIDR, ipsubnetCIDR, determineQuorum(peerCount, peers))
-		if err = dockerCli.AddObserver(allocator); err != nil {
-			Log.Fatal("Unable to start watcher", err)
-		}
+		allocator, defaultSubnet = createAllocator(router.Router, iprangeCIDR, ipsubnetCIDR, determineQuorum(peerCount, peers), isKnownPeer)
+		observeContainers(allocator)
 	} else if peerCount > 0 {
 		Log.Fatal("--init-peer-count flag specified without --ipalloc-range")
 	}
@@ -185,11 +284,10 @@ func main() {
 		dnsserver *nameserver.DNSServer
 	)
 	if !noDNS {
-		ns = nameserver.New(router.Ourself.Peer.Name, router.Peers, dockerCli, dnsDomain)
+		ns = nameserver.New(router.Ourself.Peer.Name, dnsDomain, isKnownPeer)
+		router.Peers.OnGC(func(peer *mesh.Peer) { ns.PeerGone(peer.Name) })
 		ns.SetGossip(router.NewGossip("nameserver", ns))
-		if err = dockerCli.AddObserver(ns); err != nil {
-			Log.Fatal("Unable to start watcher", err)
-		}
+		observeContainers(ns)
 		ns.Start()
 		defer ns.Stop()
 		dnsserver, err = nameserver.NewDNSServer(ns, dnsDomain, dnsListenAddress,
@@ -197,6 +295,11 @@ func main() {
 		if err != nil {
 			Log.Fatal("Unable to start dns server: ", err)
 		}
+		listenAddr := dnsListenAddress
+		if dnsEffectiveListenAddress != "" {
+			listenAddr = dnsEffectiveListenAddress
+		}
+		Log.Println("Listening for DNS queries on", listenAddr)
 		dnsserver.ActivateAndServe()
 		defer dnsserver.Stop()
 	}
@@ -215,11 +318,12 @@ func main() {
 			allocator.HandleHTTP(muxRouter, defaultSubnet, dockerCli)
 		}
 		if ns != nil {
-			ns.HandleHTTP(muxRouter)
+			ns.HandleHTTP(muxRouter, dockerCli)
 		}
 		router.HandleHTTP(muxRouter)
 		HandleHTTP(muxRouter, version, router, allocator, defaultSubnet, ns, dnsserver)
 		http.Handle("/", muxRouter)
+		Log.Println("Listening for HTTP control messages on", httpAddr)
 		go listenAndServeHTTP(httpAddr, muxRouter)
 	}
 
@@ -248,38 +352,35 @@ func canonicalName(f *mflag.Flag) string {
 	return ""
 }
 
-func logFrameFunc(debug bool) weave.LogFrameFunc {
-	if !debug {
-		return func(prefix string, frame []byte, dec *weave.EthernetDecoder) {}
-	}
-	return func(prefix string, frame []byte, dec *weave.EthernetDecoder) {
-		h := fmt.Sprintf("%x", sha256.Sum256(frame))
-		parts := []interface{}{prefix, len(frame), "bytes (", h, ")"}
+type packetLogging struct{}
 
-		if dec != nil {
-			parts = append(parts, dec.Eth.SrcMAC, "->", dec.Eth.DstMAC)
+func (packetLogging) LogPacket(msg string, key weave.PacketKey) {
+	Log.Println(msg, key.SrcMAC, "->", key.DstMAC)
+}
 
-			if dec.DF() {
-				parts = append(parts, "(DF)")
-			}
-		}
+func (packetLogging) LogForwardPacket(msg string, key weave.ForwardPacketKey) {
+	Log.Println(msg, key.SrcPeer, key.SrcMAC, "->", key.DstPeer, key.DstMAC)
+}
 
-		Log.Println(parts...)
-	}
+type nopPacketLogging struct{}
+
+func (nopPacketLogging) LogPacket(string, weave.PacketKey) {
+}
+
+func (nopPacketLogging) LogForwardPacket(string, weave.ForwardPacketKey) {
 }
 
 func parseAndCheckCIDR(cidrStr string) address.CIDR {
 	_, cidr, err := address.ParseCIDR(cidrStr)
-	if err != nil {
-		Log.Fatal(err)
-	}
+	checkFatal(err)
+
 	if cidr.Size() < ipam.MinSubnetSize {
 		Log.Fatalf("Allocation range smaller than minimum size %d: %s", ipam.MinSubnetSize, cidrStr)
 	}
 	return cidr
 }
 
-func createAllocator(router *weave.Router, ipRangeStr string, defaultSubnetStr string, quorum uint) (*ipam.Allocator, address.CIDR) {
+func createAllocator(router *mesh.Router, ipRangeStr string, defaultSubnetStr string, quorum uint, isKnownPeer func(mesh.PeerName) bool) (*ipam.Allocator, address.CIDR) {
 	ipRange := parseAndCheckCIDR(ipRangeStr)
 	defaultSubnet := ipRange
 	if defaultSubnetStr != "" {
@@ -288,7 +389,7 @@ func createAllocator(router *weave.Router, ipRangeStr string, defaultSubnetStr s
 			Log.Fatalf("IP address allocation default subnet %s does not overlap with allocation range %s", defaultSubnet, ipRange)
 		}
 	}
-	allocator := ipam.NewAllocator(router.Ourself.Peer.Name, router.Ourself.Peer.UID, router.Ourself.Peer.NickName, ipRange.Range(), quorum)
+	allocator := ipam.NewAllocator(router.Ourself.Peer.Name, router.Ourself.Peer.UID, router.Ourself.Peer.NickName, ipRange.Range(), quorum, isKnownPeer)
 
 	allocator.SetInterfaces(router.NewGossip("IPallocation", allocator))
 	allocator.Start()
@@ -316,6 +417,24 @@ func determineQuorum(initPeerCountFlag int, peers []string) uint {
 	return quorum
 }
 
+func parseTrustedSubnets(trustedSubnetStr string) ([]*net.IPNet, error) {
+	trustedSubnets := []*net.IPNet{}
+
+	if trustedSubnetStr == "" {
+		return trustedSubnets, nil
+	}
+
+	for _, subnetStr := range strings.Split(trustedSubnetStr, ",") {
+		_, subnet, err := net.ParseCIDR(subnetStr)
+		if err != nil {
+			return nil, err
+		}
+		trustedSubnets = append(trustedSubnets, subnet)
+	}
+
+	return trustedSubnets, nil
+}
+
 func listenAndServeHTTP(httpAddr string, muxRouter *mux.Router) {
 	protocol := "tcp"
 	if strings.HasPrefix(httpAddr, "/") {
@@ -331,3 +450,9 @@ func listenAndServeHTTP(httpAddr string, muxRouter *mux.Router) {
 		Log.Fatal("Unable to create http server", err)
 	}
 }
+
+func checkFatal(e error) {
+	if e != nil {
+		Log.Fatal(e)
+	}
+}

prog/weavewait/check_network.go

@@ -0,0 +1,7 @@
+// +build !iface,!mcast
+
+package main
+
+func checkNetwork() error {
+	return nil
+}

prog/weavewait/check_network_iface.go

@@ -0,0 +1,12 @@
+// +build iface,!mcast
+
+package main
+
+import (
+	weavenet "github.com/weaveworks/weave/net"
+)
+
+func checkNetwork() error {
+	_, err := weavenet.EnsureInterface("ethwe")
+	return err
+}

prog/weavewait/check_network_iface_mcast.go

@@ -0,0 +1,12 @@
+// +build iface,mcast
+
+package main
+
+import (
+	weavenet "github.com/weaveworks/weave/net"
+)
+
+func checkNetwork() error {
+	_, err := weavenet.EnsureInterfaceAndMcastRoute("ethwe")
+	return err
+}

prog/weavewait/main.go

@@ -6,8 +6,6 @@ import (
 	"os"
 	"os/exec"
 	"syscall"
-
-	weavenet "github.com/weaveworks/weave/net"
 )
 
 var (
@@ -19,8 +17,7 @@ func main() {
 		args = os.Args[1:]
 	)
 
-	_, err := weavenet.EnsureInterfaceAndMcastRoute("ethwe")
-	checkErr(err)
+	checkErr(checkNetwork())
 
 	if len(args) == 0 {
 		checkErr(ErrNoCommandSpecified)

proxy/common.go

@@ -20,19 +20,35 @@ import (
 var (
 	containerIDRegexp   = regexp.MustCompile("^(/v[0-9\\.]*)?/containers/([^/]*)/.*")
 	weaveWaitEntrypoint = []string{"/w/w"}
+	weaveEntrypoint     = "/home/weave/weaver"
+	weaveContainerName  = "/weave"
 )
 
 func callWeave(args ...string) ([]byte, []byte, error) {
 	args = append([]string{"--local"}, args...)
-	Log.Debug("Calling weave", args)
 	cmd := exec.Command("./weave", args...)
 	cmd.Env = []string{
 		"PATH=/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin",
 		"PROCFS=/hostproc",
 	}
-	if bridge := os.Getenv("DOCKER_BRIDGE"); bridge != "" {
-		cmd.Env = append(cmd.Env, fmt.Sprintf("DOCKER_BRIDGE=%s", bridge))
+
+	propagateEnv := func(key string) {
+		if val := os.Getenv(key); val != "" {
+			cmd.Env = append(cmd.Env, fmt.Sprintf("%s=%s", key, val))
+		}
 	}
+
+	propagateEnv("DOCKER_BRIDGE")
+	propagateEnv("DOCKER_HOST")
+
+	// Propogage WEAVE_DEBUG, to make debugging easier.
+	propagateEnv("WEAVE_DEBUG")
+
+	// This prevents the code coverage contortions in our
+	// integration test suite breaking things.
+	propagateEnv("COVERAGE")
+
+	Log.Debug("Calling weave args: ", args, "env: ", cmd.Env)
 	var stdout, stderr bytes.Buffer
 	cmd.Stdout = &stdout
 	cmd.Stderr = &stderr
@@ -45,6 +61,7 @@ func unmarshalRequestBody(r *http.Request, target interface{}) error {
 	if err != nil {
 		return err
 	}
+	Log.Debugf("->requestBody: %s", body)
 	if err := r.Body.Close(); err != nil {
 		return err
 	}
@@ -60,6 +77,7 @@ func marshalRequestBody(r *http.Request, body interface{}) error {
 	if err != nil {
 		return err
 	}
+	Log.Debugf("<-requestBody: %s", newBody)
 	r.Body = ioutil.NopCloser(bytes.NewReader(newBody))
 	r.ContentLength = int64(len(newBody))
 	return nil
@@ -70,6 +88,7 @@ func unmarshalResponseBody(r *http.Response, target interface{}) error {
 	if err != nil {
 		return err
 	}
+	Log.Debugf("->responseBody: %s", body)
 	if err := r.Body.Close(); err != nil {
 		return err
 	}
@@ -85,12 +104,12 @@ func marshalResponseBody(r *http.Response, body interface{}) error {
 	if err != nil {
 		return err
 	}
+	Log.Debugf("<-responseBody: %s", newBody)
 	r.Body = ioutil.NopCloser(bytes.NewReader(newBody))
 	r.ContentLength = int64(len(newBody))
 	// Stop it being chunked, because that hangs
 	r.TransferEncoding = nil
 	return nil
-
 }
 
 func inspectContainerInPath(client *docker.Client, path string) (*docker.Container, error) {
@@ -137,3 +156,25 @@ func weaveContainerIPs(containerID string) (mac string, ips []net.IP, nets []*ne
 	}
 	return
 }
+
+func addVolume(hostConfig jsonObject, source, target, mode string) error {
+	configBinds, err := hostConfig.StringArray("Binds")
+	if err != nil {
+		return err
+	}
+
+	var binds []string
+	for _, bind := range configBinds {
+		s := strings.Split(bind, ":")
+		if len(s) >= 2 && s[1] == target {
+			continue
+		}
+		binds = append(binds, bind)
+	}
+	bind := source + ":" + target
+	if mode != "" {
+		bind += ":" + mode
+	}
+	hostConfig["Binds"] = append(binds, bind)
+	return nil
+}

proxy/create_container_interceptor.go

@@ -2,16 +2,11 @@ package proxy
 
 import (
 	"errors"
-	"fmt"
-	"io/ioutil"
 	"net/http"
 	"strings"
 
 	"github.com/fsouza/go-dockerclient"
-
 	. "github.com/weaveworks/weave/common"
-	"github.com/weaveworks/weave/nameserver"
-	"github.com/weaveworks/weave/router"
 )
 
 const MaxDockerHostname = 64
@@ -22,12 +17,6 @@ var (
 
 type createContainerInterceptor struct{ proxy *Proxy }
 
-type createContainerRequestBody struct {
-	*docker.Config
-	HostConfig *docker.HostConfig `json:"HostConfig,omitempty" yaml:"HostConfig,omitempty"`
-	MacAddress string             `json:"MacAddress,omitempty" yaml:"MacAddress,omitempty"`
-}
-
 // ErrNoSuchImage replaces docker.NoSuchImage, which does not contain the image
 // name, which in turn breaks docker clients post 1.7.0 since they expect the
 // image name to be present in errors.
@@ -40,35 +29,54 @@ func (err *ErrNoSuchImage) Error() string {
 }
 
 func (i *createContainerInterceptor) InterceptRequest(r *http.Request) error {
-	container := createContainerRequestBody{}
+	container := jsonObject{}
 	if err := unmarshalRequestBody(r, &container); err != nil {
 		return err
 	}
 
-	if cidrs, err := i.proxy.weaveCIDRsFromConfig(container.Config, container.HostConfig); err != nil {
+	hostConfig, err := container.Object("HostConfig")
+	if err != nil {
+		return err
+	}
+
+	networkMode, err := hostConfig.String("NetworkMode")
+	if err != nil {
+		return err
+	}
+
+	env, err := container.StringArray("Env")
+	if err != nil {
+		return err
+	}
+
+	if cidrs, err := i.proxy.weaveCIDRs(networkMode, env); err != nil {
 		Log.Infof("Leaving container alone because %s", err)
 	} else {
 		Log.Infof("Creating container with WEAVE_CIDR \"%s\"", strings.Join(cidrs, " "))
-		if container.HostConfig == nil {
-			container.HostConfig = &docker.HostConfig{}
-		}
-		if container.Config == nil {
-			container.Config = &docker.Config{}
-		}
-		i.addWeaveWaitVolume(container.HostConfig)
-		if err := i.setWeaveWaitEntrypoint(container.Config); err != nil {
-			return err
-		}
-		hostname := r.URL.Query().Get("name")
-		if i.proxy.Config.HostnameFromLabel != "" {
-			if labelValue, ok := container.Labels[i.proxy.Config.HostnameFromLabel]; ok {
-				hostname = labelValue
+		if i.proxy.NoMulticastRoute {
+			if err := addVolume(hostConfig, i.proxy.weaveWaitNomcastVolume, "/w", "ro"); err != nil {
+				return err
+			}
+		} else {
+			if err := addVolume(hostConfig, i.proxy.weaveWaitVolume, "/w", "ro"); err != nil {
+				return err
 			}
 		}
-		hostname = i.proxy.hostnameMatchRegexp.ReplaceAllString(hostname, i.proxy.HostnameReplacement)
-		if err := i.setWeaveDNS(&container, hostname); err != nil {
+		if err := i.setWeaveWaitEntrypoint(container); err != nil {
 			return err
 		}
+		hostname, err := i.containerHostname(r, container)
+		if err != nil {
+			return err
+		}
+		if dnsDomain := i.proxy.getDNSDomain(); dnsDomain != "" {
+			if err := i.setHostname(container, hostname, dnsDomain); err != nil {
+				return err
+			}
+			if err := i.proxy.setWeaveDNS(hostConfig, hostname, dnsDomain); err != nil {
+				return err
+			}
+		}
 
 		return marshalRequestBody(r, container)
 	}
@@ -76,105 +84,97 @@ func (i *createContainerInterceptor) InterceptRequest(r *http.Request) error {
 	return nil
 }
 
-func (i *createContainerInterceptor) addWeaveWaitVolume(hostConfig *docker.HostConfig) {
-	var binds []string
-	for _, bind := range hostConfig.Binds {
-		s := strings.Split(bind, ":")
-		if len(s) >= 2 && s[1] == "/w" {
-			continue
-		}
-		binds = append(binds, bind)
+func (i *createContainerInterceptor) setWeaveWaitEntrypoint(container jsonObject) error {
+	var entrypoint []string
+	entrypoint, err := container.StringArray("Entrypoint")
+	if err != nil {
+		return err
 	}
-	hostConfig.Binds = append(binds, fmt.Sprintf("%s:/w:ro", i.proxy.weaveWaitVolume))
-}
 
-func (i *createContainerInterceptor) setWeaveWaitEntrypoint(container *docker.Config) error {
-	if len(container.Entrypoint) == 0 {
-		image, err := i.proxy.client.InspectImage(container.Image)
+	cmd, err := container.StringArray("Cmd")
+	if err != nil {
+		return err
+	}
+
+	if len(entrypoint) == 0 {
+		containerImage, err := container.String("Image")
+		if err != nil {
+			return err
+		}
+
+		image, err := i.proxy.client.InspectImage(containerImage)
 		if err == docker.ErrNoSuchImage {
-			return &ErrNoSuchImage{container.Image}
+			return &ErrNoSuchImage{containerImage}
 		} else if err != nil {
 			return err
 		}
 
-		if len(container.Cmd) == 0 {
-			container.Cmd = image.Config.Cmd
+		if len(cmd) == 0 {
+			cmd = image.Config.Cmd
+			container["Cmd"] = cmd
 		}
 
-		if container.Entrypoint == nil {
-			container.Entrypoint = image.Config.Entrypoint
+		if entrypoint == nil {
+			entrypoint = image.Config.Entrypoint
+			container["Entrypoint"] = entrypoint
 		}
 	}
 
-	if len(container.Entrypoint) == 0 && len(container.Cmd) == 0 {
+	if len(entrypoint) == 0 && len(cmd) == 0 {
 		return ErrNoCommandSpecified
 	}
 
-	if len(container.Entrypoint) == 0 || container.Entrypoint[0] != weaveWaitEntrypoint[0] {
-		container.Entrypoint = append(weaveWaitEntrypoint, container.Entrypoint...)
+	if len(entrypoint) == 0 || entrypoint[0] != weaveWaitEntrypoint[0] {
+		container["Entrypoint"] = append(weaveWaitEntrypoint, entrypoint...)
 	}
 
 	return nil
 }
 
-func (i *createContainerInterceptor) setWeaveDNS(container *createContainerRequestBody, name string) error {
-	if i.proxy.WithoutDNS {
-		return nil
+func (i *createContainerInterceptor) setHostname(container jsonObject, name, dnsDomain string) error {
+	hostname, err := container.String("Hostname")
+	if err != nil {
+		return err
 	}
-
-	dnsDomain, dnsRunning := i.getDNSDomain()
-	if !(dnsRunning || i.proxy.WithDNS) {
-		return nil
-	}
-
-	container.HostConfig.DNS = append(container.HostConfig.DNS, i.proxy.dockerBridgeIP)
-
-	if container.Hostname == "" && name != "" {
+	if hostname == "" && name != "" {
 		// Strip trailing period because it's unusual to see it used on the end of a host name
 		trimmedDNSDomain := strings.TrimSuffix(dnsDomain, ".")
 		if len(name)+1+len(trimmedDNSDomain) > MaxDockerHostname {
 			Log.Warningf("Container name [%s] too long to be used as hostname", name)
 		} else {
-			container.Hostname = name
-			container.Domainname = trimmedDNSDomain
-		}
-	}
-
-	if len(container.HostConfig.DNSSearch) == 0 {
-		if container.Hostname == "" {
-			container.HostConfig.DNSSearch = []string{dnsDomain}
-		} else {
-			container.HostConfig.DNSSearch = []string{"."}
+			container["Hostname"] = name
+			container["Domainname"] = trimmedDNSDomain
 		}
 	}
 
 	return nil
 }
 
-func (i *createContainerInterceptor) getDNSDomain() (domain string, running bool) {
-	domain = nameserver.DefaultDomain
-	weaveContainer, err := i.proxy.client.InspectContainer("weave")
-	if err != nil ||
-		weaveContainer.NetworkSettings == nil ||
-		weaveContainer.NetworkSettings.IPAddress == "" {
-		return
-	}
-
-	url := fmt.Sprintf("http://%s:%d/domain", weaveContainer.NetworkSettings.IPAddress, router.HTTPPort)
-	resp, err := http.Get(url)
-	if err != nil || resp.StatusCode != http.StatusOK {
-		return
-	}
-	defer resp.Body.Close()
-
-	b, err := ioutil.ReadAll(resp.Body)
-	if err != nil {
-		return
-	}
-
-	return string(b), true
-}
-
 func (i *createContainerInterceptor) InterceptResponse(r *http.Response) error {
 	return nil
 }
+
+func (i *createContainerInterceptor) containerHostname(r *http.Request, container jsonObject) (hostname string, err error) {
+	hostname = r.URL.Query().Get("name")
+	if i.proxy.Config.HostnameFromLabel != "" {
+		hostname, err = i.hostnameFromLabel(hostname, container)
+	}
+	hostname = i.proxy.hostnameMatchRegexp.ReplaceAllString(hostname, i.proxy.HostnameReplacement)
+	return
+}
+
+func (i *createContainerInterceptor) hostnameFromLabel(hostname string, container jsonObject) (string, error) {
+	labels, err := container.Object("Labels")
+	if err != nil {
+		return "", err
+	}
+	label, err := labels.String(i.proxy.Config.HostnameFromLabel)
+	if err != nil {
+		return "", err
+	}
+	if label == "" {
+		return hostname, nil
+	}
+
+	return label, nil
+}

proxy/create_exec_interceptor.go

@@ -4,14 +4,13 @@ import (
 	"net/http"
 	"strings"
 
-	"github.com/fsouza/go-dockerclient"
 	. "github.com/weaveworks/weave/common"
 )
 
 type createExecInterceptor struct{ proxy *Proxy }
 
 func (i *createExecInterceptor) InterceptRequest(r *http.Request) error {
-	options := docker.CreateExecOptions{}
+	options := jsonObject{}
 	if err := unmarshalRequestBody(r, &options); err != nil {
 		return err
 	}
@@ -25,14 +24,19 @@ func (i *createExecInterceptor) InterceptRequest(r *http.Request) error {
 		return nil
 	}
 
-	cidrs, err := i.proxy.weaveCIDRsFromConfig(container.Config, container.HostConfig)
+	cidrs, err := i.proxy.weaveCIDRs(container.HostConfig.NetworkMode, container.Config.Env)
 	if err != nil {
 		Log.Infof("Leaving container %s alone because %s", container.ID, err)
 		return nil
 	}
 
+	cmd, err := options.StringArray("Cmd")
+	if err != nil {
+		return err
+	}
+
 	Log.Infof("Exec in container %s with WEAVE_CIDR \"%s\"", container.ID, strings.Join(cidrs, " "))
-	options.Cmd = append(weaveWaitEntrypoint, options.Cmd...)
+	options["Cmd"] = append(weaveWaitEntrypoint, cmd...)
 
 	return marshalRequestBody(r, options)
 }

proxy/inspect_container_interceptor.go

@@ -13,39 +13,18 @@ func (i *inspectContainerInterceptor) InterceptRequest(r *http.Request) error {
 }
 
 func (i *inspectContainerInterceptor) InterceptResponse(r *http.Response) error {
-	if !i.proxy.RewriteInspect {
+	if !i.proxy.RewriteInspect || r.StatusCode != 200 {
 		return nil
 	}
 
-	container := map[string]interface{}{}
+	container := jsonObject{}
 	if err := unmarshalResponseBody(r, &container); err != nil {
 		return err
 	}
 
-	if err := updateContainerNetworkSettings(container); err != nil {
+	if err := i.proxy.updateContainerNetworkSettings(container); err != nil {
 		Log.Warningf("Inspecting container %s failed: %s", container["Id"], err)
 	}
 
 	return marshalResponseBody(r, container)
 }
-
-func updateContainerNetworkSettings(container map[string]interface{}) error {
-	containerID, err := lookupString(container, "Id")
-	if err != nil {
-		return err
-	}
-
-	mac, ips, nets, err := weaveContainerIPs(containerID)
-	if err != nil || len(ips) == 0 {
-		return err
-	}
-
-	networkSettings, err := lookupObject(container, "NetworkSettings")
-	if err != nil {
-		return err
-	}
-	networkSettings["MacAddress"] = mac
-	networkSettings["IPAddress"] = ips[0].String()
-	networkSettings["IPPrefixLen"], _ = nets[0].Mask.Size()
-	return nil
-}

proxy/inspect_exec_interceptor.go

@@ -17,17 +17,17 @@ func (i *inspectExecInterceptor) InterceptResponse(r *http.Response) error {
 		return nil
 	}
 
-	exec := map[string]interface{}{}
+	exec := jsonObject{}
 	if err := unmarshalResponseBody(r, &exec); err != nil {
 		return err
 	}
 
-	container, err := lookupObject(exec, "Container")
+	container, err := exec.Object("Container")
 	if err != nil {
 		return err
 	}
 
-	if err := updateContainerNetworkSettings(container); err != nil {
+	if err := i.proxy.updateContainerNetworkSettings(container); err != nil {
 		Log.Warningf("Inspecting exec %s failed: %s", exec["Id"], err)
 	}
 

proxy/json.go

@@ -13,12 +13,13 @@ func (e *UnmarshalWrongTypeError) Error() string {
 	return fmt.Sprintf("Wrong type for %s field, expected %s, but got %T", e.Field, e.Expected, e.Got)
 }
 
-// For parsing json objects. Look up (or create) a child object in the given map.
-func lookupObject(m map[string]interface{}, key string) (map[string]interface{}, error) {
-	iface, ok := m[key]
+type jsonObject map[string]interface{}
+
+func (j jsonObject) Object(key string) (jsonObject, error) {
+	iface, ok := j[key]
 	if !ok || iface == nil {
-		result := map[string]interface{}{}
-		m[key] = result
+		result := jsonObject{}
+		j[key] = result
 		return result, nil
 	}
 
@@ -27,12 +28,11 @@ func lookupObject(m map[string]interface{}, key string) (map[string]interface{},
 		return nil, &UnmarshalWrongTypeError{key, "object", iface}
 	}
 
-	return result, nil
+	return jsonObject(result), nil
 }
 
-// For parsing json objects. Look up (or create) a string in the given map.
-func lookupString(m map[string]interface{}, key string) (string, error) {
-	iface, ok := m[key]
+func (j jsonObject) String(key string) (string, error) {
+	iface, ok := j[key]
 	if !ok || iface == nil {
 		return "", nil
 	}
@@ -44,3 +44,29 @@ func lookupString(m map[string]interface{}, key string) (string, error) {
 
 	return result, nil
 }
+
+func (j jsonObject) StringArray(key string) ([]string, error) {
+	iface, ok := j[key]
+	if !ok || iface == nil {
+		return nil, nil
+	}
+
+	switch o := iface.(type) {
+	case string:
+		return []string{o}, nil
+	case []string:
+		return o, nil
+	case []interface{}:
+		result := []string{}
+		for _, s := range o {
+			if s, ok := s.(string); ok {
+				result = append(result, s)
+			} else {
+				return nil, &UnmarshalWrongTypeError{key, "string or array of strings", iface}
+			}
+		}
+		return result, nil
+	}
+
+	return nil, &UnmarshalWrongTypeError{key, "string or array of strings", iface}
+}

proxy/json_test.go

@@ -0,0 +1,116 @@
+package proxy
+
+import (
+	"fmt"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestLookupObject(t *testing.T) {
+	tests := []struct {
+		root   jsonObject
+		key    string
+		result jsonObject
+		err    error
+	}{
+		{
+			jsonObject{},
+			"a",
+			jsonObject{},
+			nil,
+		},
+		{
+			jsonObject{"a": map[string]interface{}{"b": int(1)}},
+			"a",
+			jsonObject{"b": int(1)},
+			nil,
+		},
+		{
+			jsonObject{"nonObject": int(1)},
+			"nonObject",
+			nil,
+			&UnmarshalWrongTypeError{Field: "nonObject", Expected: "object", Got: 1},
+		},
+	}
+	for _, test := range tests {
+		gotResult, gotErr := test.root.Object(test.key)
+		msg := fmt.Sprintf("%q.Object(%q) => %q, %q", test.root, test.key, gotResult, gotErr)
+		assert.Equal(t, test.result, gotResult, msg)
+		assert.Equal(t, test.err, gotErr, msg)
+	}
+}
+
+func TestLookupString(t *testing.T) {
+	tests := []struct {
+		root   jsonObject
+		key    string
+		result string
+		err    error
+	}{
+		{
+			jsonObject{},
+			"a",
+			"",
+			nil,
+		},
+		{
+			jsonObject{"nonString": int(1)},
+			"nonString",
+			"",
+			&UnmarshalWrongTypeError{Field: "nonString", Expected: "string", Got: 1},
+		},
+	}
+	for _, test := range tests {
+		gotResult, gotErr := test.root.String(test.key)
+		msg := fmt.Sprintf("%q.String(%q) => %q, %q", test.root, test.key, gotResult, gotErr)
+		assert.Equal(t, test.result, gotResult, msg)
+		assert.Equal(t, test.err, gotErr, msg)
+	}
+}
+
+func TestLookupStringArray(t *testing.T) {
+	tests := []struct {
+		root   jsonObject
+		key    string
+		result []string
+		err    error
+	}{
+		{
+			jsonObject{},
+			"a",
+			nil,
+			nil,
+		},
+		{
+			jsonObject{"a": []string{"foo"}},
+			"a",
+			[]string{"foo"},
+			nil,
+		},
+		{
+			jsonObject{"a": []string{}},
+			"a",
+			[]string{},
+			nil,
+		},
+		{
+			jsonObject{"a": "foo"},
+			"a",
+			[]string{"foo"},
+			nil,
+		},
+		{
+			jsonObject{"int": 5},
+			"int",
+			nil,
+			&UnmarshalWrongTypeError{Field: "int", Expected: "string or array of strings", Got: 5},
+		},
+	}
+	for _, test := range tests {
+		gotResult, gotErr := test.root.StringArray(test.key)
+		msg := fmt.Sprintf("%q.String(%q) => %q, %q", test.root, test.key, gotResult, gotErr)
+		assert.Equal(t, test.result, gotResult, msg)
+		assert.Equal(t, test.err, gotErr, msg)
+	}
+}

proxy/proxy.go

@@ -1,26 +1,32 @@
 package proxy
 
 import (
+	"bytes"
 	"crypto/tls"
 	"errors"
 	"fmt"
+	"io/ioutil"
 	"net"
 	"net/http"
 	"os"
+	"path/filepath"
 	"regexp"
 	"strings"
+	"sync"
 	"syscall"
 
-	"github.com/fsouza/go-dockerclient"
+	docker "github.com/fsouza/go-dockerclient"
 	. "github.com/weaveworks/weave/common"
+	weavedocker "github.com/weaveworks/weave/common/docker"
 )
 
 const (
 	defaultCaFile   = "ca.pem"
 	defaultKeyFile  = "key.pem"
 	defaultCertFile = "cert.pem"
-	dockerSock      = "/var/run/docker.sock"
-	dockerSockUnix  = "unix://" + dockerSock
+
+	weaveSock     = "/var/run/weave/weave.sock"
+	weaveSockUnix = "unix://" + weaveSock
 )
 
 var (
@@ -42,41 +48,56 @@ type Config struct {
 	HostnameFromLabel   string
 	HostnameMatch       string
 	HostnameReplacement string
+	Image               string
 	ListenAddrs         []string
 	RewriteInspect      bool
 	NoDefaultIPAM       bool
 	NoRewriteHosts      bool
 	TLSConfig           TLSConfig
 	Version             string
-	WithDNS             bool
 	WithoutDNS          bool
+	NoMulticastRoute    bool
+	DockerHost          string
+}
+
+type wait struct {
+	ident string
+	ch    chan error
+	done  bool
 }
 
 type Proxy struct {
+	sync.Mutex
 	Config
-	client              *docker.Client
-	dockerBridgeIP      string
-	hostnameMatchRegexp *regexp.Regexp
-	weaveWaitVolume     string
+	client                 *docker.Client
+	dockerBridgeIP         string
+	hostnameMatchRegexp    *regexp.Regexp
+	weaveWaitVolume        string
+	weaveWaitNoopVolume    string
+	weaveWaitNomcastVolume string
+	normalisedAddrs        []string
+	waiters                map[*http.Request]*wait
 }
 
 func NewProxy(c Config) (*Proxy, error) {
-	p := &Proxy{Config: c}
+	p := &Proxy{Config: c, waiters: make(map[*http.Request]*wait)}
 
-	if err := p.TLSConfig.loadCerts(); err != nil {
+	if err := p.TLSConfig.LoadCerts(); err != nil {
 		Log.Fatalf("Could not configure tls for proxy: %s", err)
 	}
 
-	// We pin the protocol version to 1.15 (which corresponds to
-	// Docker 1.3.x; the earliest version supported by weave) in order
+	// We pin the protocol version to 1.18 (which corresponds to
+	// Docker 1.6.x; the earliest version supported by weave) in order
 	// to insulate ourselves from breaking changes to the API, as
 	// happened in 1.20 (Docker 1.8.0) when the presentation of
 	// volumes changed in `inspect`.
-	client, err := docker.NewVersionedClient(dockerSockUnix, "1.15")
+	client, err := weavedocker.NewVersionedClient(c.DockerHost, "1.18")
 	if err != nil {
 		return nil, err
+	} else {
+		Log.Info(client.Info())
 	}
-	p.client = client
+	p.client = client.Client
 
 	if !p.WithoutDNS {
 		dockerBridgeIP, stderr, err := callWeave("docker-bridge-ip")
@@ -92,34 +113,72 @@ func NewProxy(c Config) (*Proxy, error) {
 		return nil, err
 	}
 
-	if err = p.findWeaveWaitVolume(); err != nil {
+	if err = p.findWeaveWaitVolumes(); err != nil {
 		return nil, err
 	}
 
+	client.AddObserver(p)
+
 	return p, nil
 }
 
-func (proxy *Proxy) Dial() (net.Conn, error) {
-	return net.Dial("unix", dockerSock)
+func (proxy *Proxy) AttachExistingContainers() {
+	containers, _ := proxy.client.ListContainers(docker.ListContainersOptions{})
+	for _, c := range containers {
+		container, err := proxy.client.InspectContainer(c.ID)
+		if err != nil {
+			if _, ok := err.(*docker.NoSuchContainer); !ok {
+				Log.Warningf("unable to attach existing container %s since inspecting it failed: %v", c.ID, err)
+			}
+			continue
+		}
+		if containerShouldAttach(container) && (container.State.Running || container.State.Paused) {
+			proxy.attach(container, false)
+		}
+	}
 }
 
-func (proxy *Proxy) findWeaveWaitVolume() error {
+func (proxy *Proxy) Dial() (net.Conn, error) {
+	proto := "tcp"
+	addr := proxy.Config.DockerHost
+	switch {
+	case strings.HasPrefix(addr, "unix://"):
+		proto = "unix"
+		addr = strings.TrimPrefix(addr, "unix://")
+	case strings.HasPrefix(addr, "tcp://"):
+		addr = strings.TrimPrefix(addr, "tcp://")
+	}
+	return net.Dial(proto, addr)
+}
+
+func (proxy *Proxy) findWeaveWaitVolumes() error {
+	var err error
+	if proxy.weaveWaitVolume, err = proxy.findVolume("/w"); err != nil {
+		return err
+	}
+	if proxy.weaveWaitNoopVolume, err = proxy.findVolume("/w-noop"); err != nil {
+		return err
+	}
+	proxy.weaveWaitNomcastVolume, err = proxy.findVolume("/w-nomcast")
+	return err
+}
+
+func (proxy *Proxy) findVolume(v string) (string, error) {
 	container, err := proxy.client.InspectContainer("weaveproxy")
 	if err != nil {
-		return fmt.Errorf("Could not find the weavewait volume: %s", err)
+		return "", fmt.Errorf("Could not find the weavewait volume: %s", err)
 	}
 
 	if container.Volumes == nil {
-		return fmt.Errorf("Could not find the weavewait volume")
+		return "", fmt.Errorf("Could not find the weavewait volume")
 	}
 
-	volume, ok := container.Volumes["/w"]
+	volume, ok := container.Volumes[v]
 	if !ok {
-		return fmt.Errorf("Could not find the weavewait volume")
+		return "", fmt.Errorf("Could not find the weavewait volume")
 	}
 
-	proxy.weaveWaitVolume = volume
-	return nil
+	return volume, nil
 }
 
 func (proxy *Proxy) ServeHTTP(w http.ResponseWriter, r *http.Request) {
@@ -143,22 +202,44 @@ func (proxy *Proxy) ServeHTTP(w http.ResponseWriter, r *http.Request) {
 	proxy.Intercept(i, w, r)
 }
 
-func (proxy *Proxy) ListenAndServe() {
+func (proxy *Proxy) Listen() []net.Listener {
 	listeners := []net.Listener{}
-	addrs := []string{}
+	proxy.normalisedAddrs = []string{}
+	unixAddrs := []string{}
 	for _, addr := range proxy.ListenAddrs {
+		if strings.HasPrefix(addr, "unix://") || strings.HasPrefix(addr, "/") {
+			unixAddrs = append(unixAddrs, addr)
+			continue
+		}
 		listener, normalisedAddr, err := proxy.listen(addr)
 		if err != nil {
 			Log.Fatalf("Cannot listen on %s: %s", addr, err)
 		}
 		listeners = append(listeners, listener)
-		addrs = append(addrs, normalisedAddr)
+		proxy.normalisedAddrs = append(proxy.normalisedAddrs, normalisedAddr)
 	}
 
-	for _, addr := range addrs {
+	if len(unixAddrs) > 0 {
+		listener, _, err := proxy.listen(weaveSockUnix)
+		if err != nil {
+			Log.Fatalf("Cannot listen on %s: %s", weaveSockUnix, err)
+		}
+		listeners = append(listeners, listener)
+
+		if err := proxy.symlink(unixAddrs); err != nil {
+			Log.Fatalf("Cannot listen on unix sockets: %s", err)
+		}
+
+		proxy.normalisedAddrs = append(proxy.normalisedAddrs, weaveSockUnix)
+	}
+
+	for _, addr := range proxy.normalisedAddrs {
 		Log.Infoln("proxy listening on", addr)
 	}
+	return listeners
+}
 
+func (proxy *Proxy) Serve(listeners []net.Listener) {
 	errs := make(chan error)
 	for _, listener := range listeners {
 		go func(listener net.Listener) {
@@ -173,6 +254,26 @@ func (proxy *Proxy) ListenAndServe() {
 	}
 }
 
+func (proxy *Proxy) ListenAndServeStatus(socket string) {
+	if err := os.Remove(socket); err != nil && !os.IsNotExist(err) {
+		Log.Fatalf("Error removing existing status socket: %s", err)
+	}
+	listener, err := net.Listen("unix", socket)
+	if err != nil {
+		Log.Fatalf("ListenAndServeStatus failed: %s", err)
+	}
+	handler := http.HandlerFunc(proxy.StatusHTTP)
+	if err := (&http.Server{Handler: handler}).Serve(listener); err != nil {
+		Log.Fatalf("ListenAndServeStatus failed: %s", err)
+	}
+}
+
+func (proxy *Proxy) StatusHTTP(w http.ResponseWriter, r *http.Request) {
+	for _, addr := range proxy.normalisedAddrs {
+		fmt.Fprintln(w, addr)
+	}
+}
+
 func copyOwnerAndPermissions(from, to string) error {
 	stat, err := os.Stat(from)
 	if err != nil {
@@ -215,7 +316,7 @@ func (proxy *Proxy) listen(protoAndAddr string) (net.Listener, string, error) {
 		if err != nil {
 			return nil, "", err
 		}
-		if proxy.TLSConfig.enabled() {
+		if proxy.TLSConfig.IsEnabled() {
 			listener = tls.NewListener(listener, proxy.TLSConfig.Config)
 		}
 
@@ -228,8 +329,10 @@ func (proxy *Proxy) listen(protoAndAddr string) (net.Listener, string, error) {
 		if err != nil {
 			return nil, "", err
 		}
-		if err = copyOwnerAndPermissions(dockerSock, addr); err != nil {
-			return nil, "", err
+		if strings.HasPrefix(proxy.Config.DockerHost, "unix://") {
+			if err = copyOwnerAndPermissions(strings.TrimPrefix(proxy.Config.DockerHost, "unix://"), addr); err != nil {
+				return nil, "", err
+			}
 		}
 
 	default:
@@ -239,15 +342,143 @@ func (proxy *Proxy) listen(protoAndAddr string) (net.Listener, string, error) {
 	return listener, fmt.Sprintf("%s://%s", proto, addr), nil
 }
 
-func (proxy *Proxy) weaveCIDRsFromConfig(config *docker.Config, hostConfig *docker.HostConfig) ([]string, error) {
-	netMode := ""
-	if hostConfig != nil {
-		netMode = hostConfig.NetworkMode
+// weavedocker.ContainerObserver interface
+func (proxy *Proxy) ContainerStarted(ident string) {
+	container, err := proxy.client.InspectContainer(ident)
+	if err == nil {
+		if containerShouldAttach(container) {
+			err = proxy.attach(container, true)
+		} else if containerIsWeaveRouter(container) {
+			err = proxy.attachRouter(container)
+		}
 	}
-	if netMode == "host" || strings.HasPrefix(netMode, "container:") {
-		return nil, fmt.Errorf("the container was created with the '--net=%s'", netMode)
+	if _, ok := err.(*docker.NoSuchContainer); err != nil && !ok {
+		Log.Warningf("unable to attach new container %s since inspecting it failed: %v", ident, err)
 	}
-	for _, e := range config.Env {
+	proxy.notifyWaiters(ident, err)
+}
+
+func containerShouldAttach(container *docker.Container) bool {
+	return len(container.Config.Entrypoint) > 0 && container.Config.Entrypoint[0] == weaveWaitEntrypoint[0]
+}
+
+func containerIsWeaveRouter(container *docker.Container) bool {
+	return container.Name == weaveContainerName &&
+		len(container.Config.Entrypoint) > 0 && container.Config.Entrypoint[0] == weaveEntrypoint
+}
+
+func (proxy *Proxy) createWait(r *http.Request, ident string) {
+	proxy.Lock()
+	proxy.waiters[r] = &wait{ident: ident, ch: make(chan error, 1)}
+	proxy.Unlock()
+}
+
+func (proxy *Proxy) removeWait(r *http.Request) {
+	proxy.Lock()
+	delete(proxy.waiters, r)
+	proxy.Unlock()
+}
+
+func (proxy *Proxy) notifyWaiters(ident string, err error) {
+	proxy.Lock()
+	for _, wait := range proxy.waiters {
+		if ident == wait.ident && !wait.done {
+			wait.ch <- err
+			close(wait.ch)
+			wait.done = true
+		}
+	}
+	proxy.Unlock()
+}
+
+func (proxy *Proxy) waitForStart(r *http.Request) error {
+	var ch chan error
+	proxy.Lock()
+	wait, found := proxy.waiters[r]
+	if found {
+		ch = wait.ch
+	}
+	proxy.Unlock()
+	if ch != nil {
+		Log.Debugf("Wait for start of container %s", wait.ident)
+		return <-ch
+	}
+	return nil
+}
+
+// If some other operation is waiting for a container to start, join in the wait
+func (proxy *Proxy) waitForStartByIdent(ident string) error {
+	var ch chan error
+	proxy.Lock()
+	for _, wait := range proxy.waiters {
+		if ident == wait.ident && !wait.done {
+			ch = wait.ch
+			break
+		}
+	}
+	proxy.Unlock()
+	if ch != nil {
+		Log.Debugf("Wait for start of container %s", ident)
+		return <-ch
+	}
+	return nil
+}
+
+func (proxy *Proxy) ContainerDied(ident string) {
+}
+
+func (proxy *Proxy) attach(container *docker.Container, orDie bool) error {
+	cidrs, err := proxy.weaveCIDRs(container.HostConfig.NetworkMode, container.Config.Env)
+	if err != nil {
+		Log.Infof("Leaving container %s alone because %s", container.ID, err)
+		return nil
+	}
+	Log.Infof("Attaching container %s with WEAVE_CIDR \"%s\" to weave network", container.ID, strings.Join(cidrs, " "))
+	args := []string{"attach"}
+	args = append(args, cidrs...)
+	if !proxy.NoRewriteHosts {
+		args = append(args, "--rewrite-hosts")
+
+		if container.HostConfig != nil {
+			for _, eh := range container.HostConfig.ExtraHosts {
+				args = append(args, fmt.Sprintf("--add-host=%s", eh))
+			}
+		}
+	}
+	if proxy.NoMulticastRoute {
+		args = append(args, "--no-multicast-route")
+	}
+	if orDie {
+		args = append(args, "--or-die")
+	}
+	args = append(args, container.ID)
+	if _, stderr, err := callWeave(args...); err != nil {
+		Log.Warningf("Attaching container %s to weave network failed: %s", container.ID, string(stderr))
+		return errors.New(string(stderr))
+	} else if len(stderr) > 0 {
+		Log.Warningf("Attaching container %s to weave network: %s", container.ID, string(stderr))
+	}
+
+	return nil
+}
+
+func (proxy *Proxy) attachRouter(container *docker.Container) error {
+	Log.Infof("Attaching weave router container: %s", container.ID)
+	args := []string{"attach-router"}
+	if _, stderr, err := callWeave(args...); err != nil {
+		Log.Warningf("Attaching container %s to weave network failed: %s", container.ID, string(stderr))
+		return errors.New(string(stderr))
+	} else if len(stderr) > 0 {
+		Log.Warningf("Attaching container %s to weave network: %s", container.ID, string(stderr))
+	}
+	return nil
+}
+
+func (proxy *Proxy) weaveCIDRs(networkMode string, env []string) ([]string, error) {
+	if networkMode == "host" || strings.HasPrefix(networkMode, "container:") {
+		return nil, fmt.Errorf("the container has '--net=%s'", networkMode)
+	}
+	for _, e := range env {
 		if strings.HasPrefix(e, "WEAVE_CIDR=") {
 			if e[11:] == "none" {
 				return nil, ErrWeaveCIDRNone
@@ -260,3 +491,151 @@ func (proxy *Proxy) weaveCIDRsFromConfig(config *docker.Config, hostConfig *dock
 	}
 	return nil, nil
 }
+
+func (proxy *Proxy) setWeaveDNS(hostConfig jsonObject, hostname, dnsDomain string) error {
+	dns, err := hostConfig.StringArray("Dns")
+	if err != nil {
+		return err
+	}
+	hostConfig["Dns"] = append(dns, proxy.dockerBridgeIP)
+
+	dnsSearch, err := hostConfig.StringArray("DnsSearch")
+	if err != nil {
+		return err
+	}
+	if len(dnsSearch) == 0 {
+		if hostname == "" {
+			hostConfig["DnsSearch"] = []string{dnsDomain}
+		} else {
+			hostConfig["DnsSearch"] = []string{"."}
+		}
+	}
+
+	return nil
+}
+
+func (proxy *Proxy) getDNSDomain() (domain string) {
+	if proxy.WithoutDNS {
+		return ""
+	}
+
+	weaveContainer, err := proxy.client.InspectContainer("weave")
+	var weaveIP string
+	if err == nil && weaveContainer.NetworkSettings != nil {
+		weaveIP = weaveContainer.NetworkSettings.IPAddress
+	}
+	if weaveIP == "" {
+		weaveIP = "127.0.0.1"
+	}
+
+	url := fmt.Sprintf("http://%s:6784/domain", weaveIP)
+	resp, err := http.Get(url)
+	if err != nil || resp.StatusCode != http.StatusOK {
+		return
+	}
+	defer resp.Body.Close()
+
+	b, err := ioutil.ReadAll(resp.Body)
+	if err != nil {
+		return
+	}
+
+	return string(b)
+}
+
+func (proxy *Proxy) updateContainerNetworkSettings(container jsonObject) error {
+	containerID, err := container.String("Id")
+	if err != nil {
+		return err
+	}
+
+	if err := proxy.waitForStartByIdent(containerID); err != nil {
+		return err
+	}
+	mac, ips, nets, err := weaveContainerIPs(containerID)
+	if err != nil || len(ips) == 0 {
+		return err
+	}
+
+	networkSettings, err := container.Object("NetworkSettings")
+	if err != nil {
+		return err
+	}
+	networkSettings["MacAddress"] = mac
+	networkSettings["IPAddress"] = ips[0].String()
+	networkSettings["IPPrefixLen"], _ = nets[0].Mask.Size()
+	return nil
+}
+
+func (proxy *Proxy) symlink(unixAddrs []string) (err error) {
+	var container *docker.Container
+	binds := []string{"/var/run/weave:/var/run/weave"}
+	froms := []string{}
+	for _, addr := range unixAddrs {
+		if addr == weaveSockUnix {
+			continue
+		}
+		from := strings.TrimPrefix(addr, "unix://")
+		dir := filepath.Dir(from)
+		binds = append(binds, dir+":"+filepath.Join("/host", dir))
+		froms = append(froms, filepath.Join("/host", from))
+		proxy.normalisedAddrs = append(proxy.normalisedAddrs, addr)
+	}
+	if len(froms) == 0 {
+		return
+	}
+
+	env := []string{
+		"PATH=/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin",
+	}
+
+	if val := os.Getenv("WEAVE_DEBUG"); val != "" {
+		env = append(env, fmt.Sprintf("%s=%s", "WEAVE_DEBUG", val))
+	}
+
+	container, err = proxy.client.CreateContainer(docker.CreateContainerOptions{
+		Config: &docker.Config{
+			Image:      proxy.Image,
+			Entrypoint: []string{"/home/weave/symlink", weaveSock},
+			Cmd:        froms,
+			Env:        env,
+		},
+		HostConfig: &docker.HostConfig{Binds: binds},
+	})
+	if err != nil {
+		return
+	}
+
+	defer func() {
+		err2 := proxy.client.RemoveContainer(docker.RemoveContainerOptions{ID: container.ID})
+		if err == nil {
+			err = err2
+		}
+	}()
+
+	err = proxy.client.StartContainer(container.ID, nil)
+	if err != nil {
+		return
+	}
+
+	var buf bytes.Buffer
+	err = proxy.client.AttachToContainer(docker.AttachToContainerOptions{
+		Container:   container.ID,
+		ErrorStream: &buf,
+		Logs:        true,
+		Stderr:      true,
+	})
+	if err != nil {
+		return
+	}
+
+	var rc int
+	rc, err = proxy.client.WaitContainer(container.ID)
+	if err != nil {
+		return
+	}
+	if rc != 0 {
+		err = errors.New(buf.String())
+	}
+	return
+}

proxy/start_container_interceptor.go

@@ -1,47 +1,76 @@
 package proxy
 
 import (
-	"errors"
 	"net/http"
 	"strings"
-
-	. "github.com/weaveworks/weave/common"
 )
 
 type startContainerInterceptor struct{ proxy *Proxy }
 
 func (i *startContainerInterceptor) InterceptRequest(r *http.Request) error {
-	return nil
-}
-
-func (i *startContainerInterceptor) InterceptResponse(r *http.Response) error {
-	if r.StatusCode < 200 || r.StatusCode >= 300 { // Docker didn't do the start
-		return nil
-	}
-
-	container, err := inspectContainerInPath(i.proxy.client, r.Request.URL.Path)
+	container, err := inspectContainerInPath(i.proxy.client, r.URL.Path)
 	if err != nil {
 		return err
 	}
 
-	cidrs, err := i.proxy.weaveCIDRsFromConfig(container.Config, container.HostConfig)
-	if err != nil {
-		Log.Infof("Leaving container %s alone because %s", container.ID, err)
-		return nil
-	}
-	Log.Infof("Attaching container %s with WEAVE_CIDR \"%s\" to weave network", container.ID, strings.Join(cidrs, " "))
-	args := []string{"attach"}
-	args = append(args, cidrs...)
-	if !i.proxy.NoRewriteHosts {
-		args = append(args, "--rewrite-hosts")
-	}
-	args = append(args, "--or-die", container.ID)
-	if _, stderr, err := callWeave(args...); err != nil {
-		Log.Warningf("Attaching container %s to weave network failed: %s", container.ID, string(stderr))
-		return errors.New(string(stderr))
-	} else if len(stderr) > 0 {
-		Log.Warningf("Attaching container %s to weave network: %s", container.ID, string(stderr))
-	}
+	// If the client has sent some JSON which might be a HostConfig, add our
+	// parameters back into it, otherwise Docker will consider them overwritten
+	if containerShouldAttach(container) && r.Header.Get("Content-Type") == "application/json" {
+		params := map[string]interface{}{}
+		if err := unmarshalRequestBody(r, &params); err != nil {
+			return err
+		}
+		// HostConfig can be sent either as a struct named HostConfig, or unnamed at top level
+		var hostConfig map[string]interface{}
+		if subParam, found := params["HostConfig"]; found {
+			if typecast, ok := subParam.(map[string]interface{}); ok {
+				hostConfig = typecast
+			}
+			// We can't reliably detect what Docker will see as a top-level HostConfig,
+			// so just assume any parameter at top level indicates there is one.
+		} else if len(params) > 0 {
+			hostConfig = params
+		}
+		if hostConfig != nil {
+			networkMode, err := jsonObject(hostConfig).String("NetworkMode")
+			if err != nil {
+				return err
+			}
+			if strings.HasPrefix(networkMode, "container:") || networkMode == "host" {
+				if err := addVolume(hostConfig, i.proxy.weaveWaitNoopVolume, "/w", "ro"); err != nil {
+					return err
+				}
+			} else {
+				if i.proxy.NoMulticastRoute {
+					if err := addVolume(hostConfig, i.proxy.weaveWaitNomcastVolume, "/w", "ro"); err != nil {
+						return err
+					}
+				} else {
+					if err := addVolume(hostConfig, i.proxy.weaveWaitVolume, "/w", "ro"); err != nil {
+						return err
+					}
+				}
+				if dnsDomain := i.proxy.getDNSDomain(); dnsDomain != "" {
+					if err := i.proxy.setWeaveDNS(hostConfig, container.Config.Hostname, dnsDomain); err != nil {
+						return err
+					}
+				}
+			}
 
+			// Note we marshal the original top-level dictionary to avoid disturbing anything else
+			if err := marshalRequestBody(r, params); err != nil {
+				return err
+			}
+		}
+	}
+	i.proxy.createWait(r, container.ID)
 	return nil
 }
+
+func (i *startContainerInterceptor) InterceptResponse(r *http.Response) error {
+	defer i.proxy.removeWait(r.Request)
+	if r.StatusCode < 200 || r.StatusCode >= 300 { // Docker didn't do the start
+		return nil
+	}
+	return i.proxy.waitForStart(r.Request)
+}

proxy/tls.go

@@ -17,15 +17,17 @@ type TLSConfig struct {
 	*tls.Config
 }
 
-func (c *TLSConfig) enabled() bool {
+// IsEnabled returns true if TLS is enable, according to the config.
+func (c *TLSConfig) IsEnabled() bool {
 	if c == nil {
 		return false
 	}
 	return c.Enabled || c.Verify
 }
 
-func (c *TLSConfig) loadCerts() error {
-	if !c.enabled() {
+// LoadCerts loads the certificates into c.Config, if TLS is enabled.
+func (c *TLSConfig) LoadCerts() error {
+	if !c.IsEnabled() {
 		return nil
 	}
 

router/bridge.go

@@ -0,0 +1,35 @@
+package router
+
+// Interface to packet handling on the local virtual bridge
+type Bridge interface {
+	// Inject a packet to be delivered locally
+	InjectPacket(PacketKey) FlowOp
+
+	// Start consuming packets from the bridge.  Injected packets
+	// should not be included.
+	StartConsumingPackets(BridgeConsumer) error
+
+	String() string
+	Stats() map[string]int
+}
+
+// A function that determines how to handle locally captured packets.
+type BridgeConsumer func(PacketKey) FlowOp
+
+type NullBridge struct{}
+
+func (NullBridge) InjectPacket(PacketKey) FlowOp {
+	return nil
+}
+
+func (NullBridge) StartConsumingPackets(BridgeConsumer) error {
+	return nil
+}
+
+func (NullBridge) String() string {
+	return "<no bridge networking>"
+}
+
+func (NullBridge) Stats() map[string]int {
+	return nil
+}

router/connection.go

@@ -1,589 +0,0 @@
-package router
-
-import (
-	"encoding/binary"
-	"errors"
-	"fmt"
-	"net"
-	"strconv"
-	"sync"
-	"time"
-)
-
-type Connection interface {
-	Local() *Peer
-	Remote() *Peer
-	RemoteTCPAddr() string
-	Outbound() bool
-	Established() bool
-	BreakTie(Connection) ConnectionTieBreak
-	Shutdown(error)
-	Log(args ...interface{})
-}
-
-type ConnectionTieBreak int
-
-const (
-	TieBreakWon ConnectionTieBreak = iota
-	TieBreakLost
-	TieBreakTied
-)
-
-var ErrConnectToSelf = errors.New("Cannot connect to ourself")
-
-type RemoteConnection struct {
-	local         *Peer
-	remote        *Peer
-	remoteTCPAddr string
-	outbound      bool
-	established   bool
-}
-
-type LocalConnection struct {
-	sync.RWMutex
-	RemoteConnection
-	TCPConn           *net.TCPConn
-	version           byte
-	tcpSender         TCPSender
-	remoteUDPAddr     *net.UDPAddr
-	receivedHeartbeat bool
-	stackFrag         bool
-	effectivePMTU     int
-	SessionKey        *[32]byte
-	heartbeatTCP      *time.Ticker
-	heartbeatTimeout  *time.Timer
-	heartbeatFrame    []byte
-	heartbeat         *time.Ticker
-	fragTest          *time.Ticker
-	forwarder         *Forwarder
-	forwarderDF       *ForwarderDF
-	Decryptor         Decryptor
-	Router            *Router
-	uid               uint64
-	actionChan        chan<- ConnectionAction
-	finished          <-chan struct{} // closed to signal that actorLoop has finished
-}
-
-type ConnectionAction func() error
-
-func NewRemoteConnection(from, to *Peer, tcpAddr string, outbound bool, established bool) *RemoteConnection {
-	return &RemoteConnection{
-		local:         from,
-		remote:        to,
-		remoteTCPAddr: tcpAddr,
-		outbound:      outbound,
-		established:   established,
-	}
-}
-
-func (conn *RemoteConnection) Local() *Peer                           { return conn.local }
-func (conn *RemoteConnection) Remote() *Peer                          { return conn.remote }
-func (conn *RemoteConnection) RemoteTCPAddr() string                  { return conn.remoteTCPAddr }
-func (conn *RemoteConnection) Outbound() bool                         { return conn.outbound }
-func (conn *RemoteConnection) Established() bool                      { return conn.established }
-func (conn *RemoteConnection) BreakTie(Connection) ConnectionTieBreak { return TieBreakTied }
-func (conn *RemoteConnection) Shutdown(error)                         {}
-
-func (conn *RemoteConnection) Log(args ...interface{}) {
-	log.Println(append(append([]interface{}{}, fmt.Sprintf("->[%s|%s]:", conn.remoteTCPAddr, conn.remote)), args...)...)
-}
-
-func (conn *RemoteConnection) ErrorLog(args ...interface{}) {
-	log.Errorln(append(append([]interface{}{}, fmt.Sprintf("->[%s|%s]:", conn.remoteTCPAddr, conn.remote)), args...)...)
-}
-
-// Does not return anything. If the connection is successful, it will
-// end up in the local peer's connections map.
-func StartLocalConnection(connRemote *RemoteConnection, tcpConn *net.TCPConn, udpAddr *net.UDPAddr, router *Router, acceptNewPeer bool) {
-	if connRemote.local != router.Ourself.Peer {
-		log.Fatal("Attempt to create local connection from a peer which is not ourself")
-	}
-	// NB, we're taking a copy of connRemote here.
-	actionChan := make(chan ConnectionAction, ChannelSize)
-	finished := make(chan struct{})
-	conn := &LocalConnection{
-		RemoteConnection: *connRemote,
-		Router:           router,
-		TCPConn:          tcpConn,
-		remoteUDPAddr:    udpAddr,
-		effectivePMTU:    DefaultPMTU,
-		uid:              randUint64(),
-		actionChan:       actionChan,
-		finished:         finished}
-	go conn.run(actionChan, finished, acceptNewPeer)
-}
-
-func (conn *LocalConnection) BreakTie(dupConn Connection) ConnectionTieBreak {
-	dupConnLocal := dupConn.(*LocalConnection)
-	// conn.uid is used as the tie breaker here, in the knowledge that
-	// both sides will make the same decision.
-	if conn.uid < dupConnLocal.uid {
-		return TieBreakWon
-	} else if dupConnLocal.uid < conn.uid {
-		return TieBreakLost
-	} else {
-		return TieBreakTied
-	}
-}
-
-// Read by the forwarder processes when in the UDP senders
-func (conn *LocalConnection) RemoteUDPAddr() *net.UDPAddr {
-	conn.RLock()
-	defer conn.RUnlock()
-	return conn.remoteUDPAddr
-}
-
-func (conn *LocalConnection) Established() bool {
-	conn.RLock()
-	defer conn.RUnlock()
-	return conn.established
-}
-
-// Called by forwarder processes, read in Forward (by sniffer and udp
-// listener process in router).
-func (conn *LocalConnection) setEffectivePMTU(pmtu int) {
-	conn.Lock()
-	defer conn.Unlock()
-	if conn.effectivePMTU != pmtu {
-		conn.effectivePMTU = pmtu
-		conn.Log("Effective PMTU set to", pmtu)
-	}
-}
-
-// Called by the connection's actor process, and by the connection's
-// TCP receiver process. StackFrag is read in conn.forward
-func (conn *LocalConnection) setStackFrag(frag bool) {
-	conn.Lock()
-	defer conn.Unlock()
-	conn.stackFrag = frag
-}
-
-// Called by the connection's TCP receiver process.
-func (conn *LocalConnection) pmtuVerified(pmtu int) {
-	conn.RLock()
-	fwd := conn.forwarderDF
-	conn.RUnlock()
-	if fwd != nil {
-		fwd.PMTUVerified(pmtu)
-	}
-}
-
-// Send directly, not via the Actor.  If it goes via the Actor we can
-// get a deadlock where LocalConnection is blocked talking to
-// LocalPeer and LocalPeer is blocked trying send a ProtocolMsg via
-// LocalConnection, and the channels are full in both directions so
-// nothing can proceed.
-func (conn *LocalConnection) SendProtocolMsg(m ProtocolMsg) {
-	if err := conn.sendProtocolMsg(m); err != nil {
-		conn.Shutdown(err)
-	}
-}
-
-// ACTOR methods
-
-// NB: The conn.* fields are only written by the connection actor
-// process, which is the caller of the ConnectionAction funs. Hence we
-// do not need locks for reading, and only need write locks for fields
-// read by other processes.
-
-// Async
-func (conn *LocalConnection) Shutdown(err error) {
-	// err should always be a real error, even if only io.EOF
-	if err == nil {
-		panic("nil error")
-	}
-
-	// Run on its own goroutine in case the channel is backed up
-	go func() { conn.sendAction(func() error { return err }) }()
-}
-
-// Async
-//
-// Heartbeating serves two purposes: a) keeping NAT paths alive, and
-// b) updating a remote peer's knowledge of our address, in the event
-// it changes (e.g. because NAT paths expired).
-func (conn *LocalConnection) ReceivedHeartbeat(remoteUDPAddr *net.UDPAddr, connUID uint64) {
-	if remoteUDPAddr == nil || connUID != conn.uid {
-		return
-	}
-	conn.sendAction(func() error {
-		oldRemoteUDPAddr := conn.remoteUDPAddr
-		old := conn.receivedHeartbeat
-		conn.Lock()
-		conn.remoteUDPAddr = remoteUDPAddr
-		conn.receivedHeartbeat = true
-		conn.Unlock()
-		conn.heartbeatTimeout.Reset(HeartbeatTimeout)
-		if !old {
-			if err := conn.sendSimpleProtocolMsg(ProtocolConnectionEstablished); err != nil {
-				return err
-			}
-		}
-		if oldRemoteUDPAddr == nil {
-			return conn.sendFastHeartbeats()
-		} else if oldRemoteUDPAddr.String() != remoteUDPAddr.String() {
-			log.Println("Peer", conn.remote, "moved from", oldRemoteUDPAddr, "to", remoteUDPAddr)
-		}
-		return nil
-	})
-}
-
-// Async
-func (conn *LocalConnection) SetEstablished() {
-	conn.sendAction(func() error {
-		stopTicker(conn.heartbeat)
-		old := conn.established
-		conn.Lock()
-		conn.established = true
-		conn.Unlock()
-		if old {
-			return nil
-		}
-		conn.Router.Ourself.ConnectionEstablished(conn)
-		if err := conn.ensureForwarders(); err != nil {
-			return err
-		}
-		// Send a large frame down the DF channel in order to prompt
-		// PMTU discovery to start.
-		conn.Send(true, PMTUDiscovery)
-		conn.heartbeat = time.NewTicker(SlowHeartbeat)
-		conn.fragTest = time.NewTicker(FragTestInterval)
-		// avoid initial waits for timers to fire
-		conn.Send(true, conn.heartbeatFrame)
-		conn.performFragTest()
-		return nil
-	})
-}
-
-// Send an actor request to the actorLoop, but don't block if
-// actorLoop has exited - see http://blog.golang.org/pipelines for
-// pattern
-func (conn *LocalConnection) sendAction(action ConnectionAction) {
-	select {
-	case conn.actionChan <- action:
-	case <-conn.finished:
-	}
-}
-
-// ACTOR server
-
-func (conn *LocalConnection) run(actionChan <-chan ConnectionAction, finished chan<- struct{}, acceptNewPeer bool) {
-	var err error // important to use this var and not create another one with 'err :='
-	defer func() { conn.shutdown(err) }()
-	defer close(finished)
-
-	conn.TCPConn.SetLinger(0)
-	intro, err := ProtocolIntroParams{
-		MinVersion: conn.Router.ProtocolMinVersion,
-		MaxVersion: ProtocolMaxVersion,
-		Features:   conn.makeFeatures(),
-		Conn:       conn.TCPConn,
-		Password:   conn.Router.Password,
-		Outbound:   conn.outbound,
-	}.DoIntro()
-	if err != nil {
-		return
-	}
-
-	conn.SessionKey = intro.SessionKey
-	conn.tcpSender = intro.Sender
-	conn.version = intro.Version
-
-	remote, err := conn.parseFeatures(intro.Features)
-	if err != nil {
-		return
-	}
-
-	if err = conn.registerRemote(remote, acceptNewPeer); err != nil {
-		return
-	}
-
-	if conn.SessionKey == nil {
-		conn.Decryptor = NewNonDecryptor()
-	} else {
-		conn.Decryptor = NewNaClDecryptor(conn.SessionKey, conn.outbound)
-	}
-
-	conn.Log("connection ready; using protocol version", conn.version)
-
-	// The ordering of the following is very important. [1]
-
-	if conn.remoteUDPAddr != nil {
-		if err = conn.ensureForwarders(); err != nil {
-			return
-		}
-	}
-	if err = conn.Router.Ourself.AddConnection(conn); err != nil {
-		return
-	}
-	if err = conn.initHeartbeats(); err != nil {
-		return
-	}
-
-	go conn.receiveTCP(intro.Receiver)
-	err = conn.actorLoop(actionChan)
-}
-
-// [1] Ordering constraints:
-//
-// (a) AddConnections must precede initHeartbeats. It is only after
-// the former completes that we know the connection is valid, in
-// particular is not a duplicate connection to the same peer. Sending
-// heartbeats on a duplicate connection can trip up crypto at the
-// other end, since the associated UDP packets may get decoded by the
-// other connection. It is also generally wasteful to engage in any
-// interaction with the remote on a connection that turns out to be
-// invald.
-//
-// (b) AddConnection must precede receiveTCP. In the absence of any
-// indirect connectivity to the remote peer, the first we hear about
-// it (and any peers reachable from it) is through topology gossip it
-// sends us on the connection. We must ensure that the connection has
-// been added to Ourself prior to processing any such gossip,
-// otherwise we risk immediately gc'ing part of that newly received
-// portion of the topology (though not the remote peer itself, since
-// that will have a positive ref count), leaving behind dangling
-// references to peers. Hence we must invoke AddConnection, which is
-// *synchronous*, first.
-//
-// (c) AddConnection must precede actorLoop. More precisely, it must
-// precede shutdown, since that invokes DeleteConnection and is
-// invoked on termination of this entire function. Essentially this
-// boils down to a prohibition on running AddConnection in a separate
-// goroutine, at least not without some synchronisation. Which in turn
-// requires us the launching of the receiveTCP goroutine to precede
-// actorLoop.
-//
-// (d) AddConnection should precede receiveTCP. There is no point
-// starting the latter if the former fails.
-//
-// (e) initHeartbeats should precede actorLoop. The former is setting
-// LocalConnection fields accessed by the latter. Since the latter
-// runs in a separate goroutine, we'd have to add some synchronisation
-// if initHeartbeats isn't run first.
-//
-// (f) ensureForwarders should precede AddConnection. As soon as a
-// connection has been added to LocalPeer by the latter, it becomes
-// visible to the packet routing logic, which will end up dropping
-// packets if the forwarders haven't been created yet. We cannot
-// prevent that completely, since, for example, forwarder can only be
-// created when we know the remote UDP address, but it helps to try.
-
-func (conn *LocalConnection) makeFeatures() map[string]string {
-	return map[string]string{
-		"PeerNameFlavour": PeerNameFlavour,
-		"Name":            conn.local.Name.String(),
-		"NickName":        conn.local.NickName,
-		"UID":             fmt.Sprint(conn.local.UID),
-		"ConnID":          fmt.Sprint(conn.uid),
-	}
-}
-
-type features map[string]string
-
-func (features features) MustHave(keys []string) error {
-	for _, key := range keys {
-		if _, ok := features[key]; !ok {
-			return fmt.Errorf("Field %s is missing", key)
-		}
-	}
-	return nil
-}
-
-func (features features) Get(key string) string {
-	return features[key]
-}
-
-func (conn *LocalConnection) parseFeatures(features features) (*Peer, error) {
-	if err := features.MustHave([]string{"PeerNameFlavour", "Name", "NickName", "UID", "ConnID"}); err != nil {
-		return nil, err
-	}
-
-	remotePeerNameFlavour := features.Get("PeerNameFlavour")
-	if remotePeerNameFlavour != PeerNameFlavour {
-		return nil, fmt.Errorf("Peer name flavour mismatch (ours: '%s', theirs: '%s')", PeerNameFlavour, remotePeerNameFlavour)
-	}
-
-	name, err := PeerNameFromString(features.Get("Name"))
-	if err != nil {
-		return nil, err
-	}
-
-	nickName := features.Get("NickName")
-
-	uid, err := ParsePeerUID(features.Get("UID"))
-	if err != nil {
-		return nil, err
-	}
-
-	remoteConnID, err := strconv.ParseUint(features.Get("ConnID"), 10, 64)
-	if err != nil {
-		return nil, err
-	}
-
-	conn.uid ^= remoteConnID
-	return NewPeer(name, nickName, uid, 0), nil
-}
-
-func (conn *LocalConnection) registerRemote(remote *Peer, acceptNewPeer bool) error {
-	if acceptNewPeer {
-		conn.remote = conn.Router.Peers.FetchWithDefault(remote)
-	} else {
-		conn.remote = conn.Router.Peers.FetchAndAddRef(remote.Name)
-		if conn.remote == nil {
-			return fmt.Errorf("Found unknown remote name: %s at %s", remote.Name, conn.remoteTCPAddr)
-		}
-	}
-
-	if conn.remote.UID != remote.UID {
-		return fmt.Errorf("Connection appears to be with different version of a peer we already know of")
-	}
-
-	if conn.remote == conn.local {
-		return ErrConnectToSelf
-	}
-
-	return nil
-}
-
-func (conn *LocalConnection) initHeartbeats() error {
-	conn.heartbeatTCP = time.NewTicker(TCPHeartbeat)
-	conn.heartbeatTimeout = time.NewTimer(HeartbeatTimeout)
-	conn.heartbeatFrame = make([]byte, EthernetOverhead+8)
-	binary.BigEndian.PutUint64(conn.heartbeatFrame[EthernetOverhead:], conn.uid)
-	if conn.remoteUDPAddr == nil {
-		return nil
-	}
-	return conn.sendFastHeartbeats()
-}
-
-func (conn *LocalConnection) actorLoop(actionChan <-chan ConnectionAction) (err error) {
-	for err == nil {
-		select {
-		case action := <-actionChan:
-			err = action()
-		case <-conn.heartbeatTCP.C:
-			err = conn.sendSimpleProtocolMsg(ProtocolHeartbeat)
-		case <-conn.heartbeatTimeout.C:
-			err = fmt.Errorf("timed out waiting for UDP heartbeat")
-		case <-tickerChan(conn.heartbeat):
-			conn.Send(true, conn.heartbeatFrame)
-		case <-tickerChan(conn.fragTest):
-			conn.performFragTest()
-		}
-	}
-	return
-}
-
-func (conn *LocalConnection) shutdown(err error) {
-	if conn.remote == nil {
-		log.Errorf("->[%s] connection shutting down due to error during handshake: %v", conn.remoteTCPAddr, err)
-	} else {
-		conn.ErrorLog("connection shutting down due to error:", err)
-	}
-
-	if conn.TCPConn != nil {
-		checkWarn(conn.TCPConn.Close())
-	}
-
-	if conn.remote != nil {
-		conn.Router.Peers.Dereference(conn.remote)
-		conn.Router.Ourself.DeleteConnection(conn)
-	}
-
-	if conn.heartbeatTimeout != nil {
-		conn.heartbeatTimeout.Stop()
-	}
-
-	stopTicker(conn.heartbeatTCP)
-	stopTicker(conn.heartbeat)
-	stopTicker(conn.fragTest)
-
-	// blank out the forwardChan so that the router processes don't
-	// try to send any more
-	conn.stopForwarders()
-
-	conn.Router.ConnectionMaker.ConnectionTerminated(conn.remoteTCPAddr, err)
-}
-
-// Helpers
-
-func (conn *LocalConnection) sendSimpleProtocolMsg(tag ProtocolTag) error {
-	return conn.sendProtocolMsg(ProtocolMsg{tag: tag})
-}
-
-func (conn *LocalConnection) sendProtocolMsg(m ProtocolMsg) error {
-	return conn.tcpSender.Send(Concat([]byte{byte(m.tag)}, m.msg))
-}
-
-func (conn *LocalConnection) receiveTCP(receiver TCPReceiver) {
-	var err error
-	for {
-		conn.extendReadDeadline()
-
-		var msg []byte
-		if msg, err = receiver.Receive(); err != nil {
-			break
-		}
-		if len(msg) < 1 {
-			conn.Log("ignoring blank msg")
-			continue
-		}
-		if err = conn.handleProtocolMsg(ProtocolTag(msg[0]), msg[1:]); err != nil {
-			break
-		}
-	}
-	conn.Shutdown(err)
-}
-
-func (conn *LocalConnection) handleProtocolMsg(tag ProtocolTag, payload []byte) error {
-	switch tag {
-	case ProtocolHeartbeat:
-	case ProtocolConnectionEstablished:
-		// We sent fast heartbeats to the remote peer, which has now
-		// received at least one of them and told us via this message.
-		// We can now consider the connection as established from our
-		// end.
-		conn.SetEstablished()
-	case ProtocolFragmentationReceived:
-		conn.setStackFrag(true)
-	case ProtocolPMTUVerified:
-		conn.pmtuVerified(int(binary.BigEndian.Uint16(payload)))
-	case ProtocolGossipUnicast, ProtocolGossipBroadcast, ProtocolGossip:
-		return conn.Router.handleGossip(tag, payload)
-	default:
-		conn.Log("ignoring unknown protocol tag:", tag)
-	}
-	return nil
-}
-
-func (conn *LocalConnection) extendReadDeadline() {
-	conn.TCPConn.SetReadDeadline(time.Now().Add(TCPHeartbeat * 2))
-}
-
-func (conn *LocalConnection) sendFastHeartbeats() error {
-	err := conn.ensureForwarders()
-	if err == nil {
-		conn.heartbeat = time.NewTicker(FastHeartbeat)
-		conn.Send(true, conn.heartbeatFrame) // avoid initial wait
-	}
-	return err
-}
-
-func (conn *LocalConnection) performFragTest() {
-	conn.setStackFrag(false)
-	conn.Send(false, FragTest)
-}
-
-func tickerChan(ticker *time.Ticker) <-chan time.Time {
-	if ticker != nil {
-		return ticker.C
-	}
-	return nil
-}
-
-func stopTicker(ticker *time.Ticker) {
-	if ticker != nil {
-		ticker.Stop()
-	}
-}

router/consts.go

@@ -1,34 +0,0 @@
-package router
-
-import (
-	"math"
-	"time"
-)
-
-const (
-	EthernetOverhead    = 14
-	UDPOverhead         = 28 // 20 bytes for IPv4, 8 bytes for UDP
-	Port                = 6783
-	HTTPPort            = Port + 1
-	DefaultPMTU         = 65535
-	MaxUDPPacketSize    = 65535
-	ChannelSize         = 16
-	FragTestSize        = 60001
-	PMTUDiscoverySize   = 60000
-	TCPHeartbeat        = 30 * time.Second
-	FastHeartbeat       = 500 * time.Millisecond
-	SlowHeartbeat       = 10 * time.Second
-	FragTestInterval    = 5 * time.Minute
-	GossipInterval      = 30 * time.Second
-	PMTUVerifyAttempts  = 8
-	PMTUVerifyTimeout   = 10 * time.Millisecond // gets doubled with every attempt
-	MaxDuration         = time.Duration(math.MaxInt64)
-	MaxMissedHeartbeats = 6
-	HeartbeatTimeout    = MaxMissedHeartbeats * SlowHeartbeat
-	MaxTCPMsgSize       = 10 * 1024 * 1024
-)
-
-var (
-	FragTest      = make([]byte, FragTestSize)
-	PMTUDiscovery = make([]byte, PMTUDiscoverySize)
-)

router/ethernet_decoder.go

@@ -29,7 +29,13 @@ func (dec *EthernetDecoder) DecodeLayers(data []byte) {
 	dec.parser.DecodeLayers(data, &dec.decoded)
 }
 
-func (dec *EthernetDecoder) sendICMPFragNeeded(mtu int, sendFrame func([]byte) error) error {
+func (dec *EthernetDecoder) PacketKey() (key PacketKey) {
+	copy(key.SrcMAC[:], dec.Eth.SrcMAC)
+	copy(key.DstMAC[:], dec.Eth.DstMAC)
+	return
+}
+
+func (dec *EthernetDecoder) makeICMPFragNeeded(mtu int) ([]byte, error) {
 	buf := gopacket.NewSerializeBuffer()
 	opts := gopacket.SerializeOptions{
 		FixLengths:       true,
@@ -57,23 +63,17 @@ func (dec *EthernetDecoder) sendICMPFragNeeded(mtu int, sendFrame func([]byte) e
 			Seq:      uint16(mtu)},
 		&payload)
 	if err != nil {
-		return err
+		return nil, err
 	}
 
-	log.Printf("Sending ICMP 3,4 (%v -> %v): PMTU= %v", dec.IP.DstIP, dec.IP.SrcIP, mtu)
-	return sendFrame(buf.Bytes())
+	log.Printf("Sending ICMP 3,4 (%v -> %v): PMTU=%v", dec.IP.DstIP, dec.IP.SrcIP, mtu)
+	return buf.Bytes(), nil
 }
 
 var (
-	// see http://en.wikipedia.org/wiki/Multicast_address#Ethernet
-	stpMACPrefix = []byte{0x01, 0x80, 0xC2, 0x00, 0x00}
-	zeroMAC, _   = net.ParseMAC("00:00:00:00:00:00")
+	zeroMAC, _ = net.ParseMAC("00:00:00:00:00:00")
 )
 
-func (dec *EthernetDecoder) DropFrame() bool {
-	return bytes.Equal(stpMACPrefix, dec.Eth.DstMAC[:len(stpMACPrefix)])
-}
-
 func (dec *EthernetDecoder) IsSpecial() bool {
 	return dec.Eth.Length == 0 && dec.Eth.EthernetType == layers.EthernetTypeLLC &&
 		bytes.Equal(zeroMAC, dec.Eth.SrcMAC) && bytes.Equal(zeroMAC, dec.Eth.DstMAC)

router/flow.go

@@ -0,0 +1,105 @@
+package router
+
+import (
+	"net"
+
+	"github.com/weaveworks/weave/mesh"
+)
+
+// Just enough flow machinery for the weave router
+
+type MAC [6]byte
+
+func (mac MAC) String() string {
+	return net.HardwareAddr(mac[:]).String()
+}
+
+type PacketKey struct {
+	SrcMAC MAC
+	DstMAC MAC
+}
+
+type ForwardPacketKey struct {
+	SrcPeer *mesh.Peer
+	DstPeer *mesh.Peer
+	PacketKey
+}
+
+type FlowOp interface {
+	// The caller must supply an EthernetDecoder specific to this
+	// thread, which has already been used to decode the frame.
+	// The broadcast parameter is a hint whether the packet is
+	// being broadcast.
+	Process(frame []byte, dec *EthernetDecoder, broadcast bool)
+
+	// Does the FlowOp discard the packet?
+	Discards() bool
+}
+
+type DiscardingFlowOp struct{}
+
+func (DiscardingFlowOp) Process([]byte, *EthernetDecoder, bool) {
+}
+
+func (DiscardingFlowOp) Discards() bool {
+	return true
+}
+
+type NonDiscardingFlowOp struct{}
+
+func (NonDiscardingFlowOp) Discards() bool {
+	return false
+}
+
+type MultiFlowOp struct {
+	broadcast bool
+	ops       []FlowOp
+}
+
+func NewMultiFlowOp(broadcast bool, ops ...FlowOp) *MultiFlowOp {
+	mfop := &MultiFlowOp{broadcast: broadcast}
+	for _, op := range ops {
+		mfop.Add(op)
+	}
+	return mfop
+}
+
+func (mfop *MultiFlowOp) Add(op FlowOp) {
+	mfop.ops = append(mfop.ops, op)
+}
+
+func (mfop *MultiFlowOp) Process(frame []byte, dec *EthernetDecoder, broadcast bool) {
+	for _, op := range mfop.ops {
+		op.Process(frame, dec, mfop.broadcast)
+	}
+}
+
+func (mfop *MultiFlowOp) Discards() bool {
+	for _, op := range mfop.ops {
+		if !op.Discards() {
+			return false
+		}
+	}
+
+	return true
+}
+
+func FlattenFlowOp(fop FlowOp) []FlowOp {
+	return collectFlowOps(nil, fop)
+}
+
+func collectFlowOps(into []FlowOp, fop FlowOp) []FlowOp {
+	if fop == nil {
+		return into
+	}
+
+	if mfop, ok := fop.(*MultiFlowOp); ok {
+		for _, op := range mfop.ops {
+			into = collectFlowOps(into, op)
+		}
+
+		return into
+	}
+
+	return append(into, fop)
+}