You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
tailscale/util/cloudenv/cloudenv.go

185 lines
5.1 KiB
Go

ipn/ipnlocal, net/dns*, util/cloudenv: specialize DNS config on Google Cloud This does three things: * If you're on GCP, it adds a *.internal DNS split route to the metadata server, so we never break GCP DNS names. This lets people have some Tailscale nodes on GCP and some not (e.g. laptops at home) without having to add a Tailnet-wide *.internal DNS route. If you already have such a route, though, it won't overwrite it. * If the 100.100.100.100 DNS forwarder has nowhere to forward to, it forwards it to the GCP metadata IP, which forwards to 8.8.8.8. This means there are never errNoUpstreams ("upstream nameservers not set") errors on GCP due to e.g. mangled /etc/resolv.conf (GCP default VMs don't have systemd-resolved, so it's likely a DNS supremacy fight) * makes the DNS fallback mechanism use the GCP metadata IP as a fallback before our hosted HTTP-based fallbacks I created a default GCP VM from their web wizard. It has no systemd-resolved. I then made its /etc/resolv.conf be empty and deleted its GCP hostnames in /etc/hosts. I then logged in to a tailnet with no global DNS settings. With this, tailscaled writes /etc/resolv.conf (direct mode, as no systemd-resolved) and sets it to 100.100.100.100, which then has regular DNS via the metadata IP and *.internal DNS via the metadata IP as well. If the tailnet configures explicit DNS servers, those are used instead, except for *.internal. This also adds a new util/cloudenv package based on version/distro where the cloud type is only detected once. We'll likely expand it in the future for other clouds, doing variants of this change for other popular cloud environments. Fixes #4911 RELNOTES=Google Cloud DNS improvements Change-Id: I19f3c2075983669b2b2c0f29a548da8de373c7cf Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
2 years ago
// Copyright (c) 2022 Tailscale Inc & AUTHORS All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// Package cloudenv reports which known cloud environment we're running in.
package cloudenv
import (
"context"
"encoding/json"
"log"
"net"
"net/http"
"os"
"runtime"
"strings"
ipn/ipnlocal, net/dns*, util/cloudenv: specialize DNS config on Google Cloud This does three things: * If you're on GCP, it adds a *.internal DNS split route to the metadata server, so we never break GCP DNS names. This lets people have some Tailscale nodes on GCP and some not (e.g. laptops at home) without having to add a Tailnet-wide *.internal DNS route. If you already have such a route, though, it won't overwrite it. * If the 100.100.100.100 DNS forwarder has nowhere to forward to, it forwards it to the GCP metadata IP, which forwards to 8.8.8.8. This means there are never errNoUpstreams ("upstream nameservers not set") errors on GCP due to e.g. mangled /etc/resolv.conf (GCP default VMs don't have systemd-resolved, so it's likely a DNS supremacy fight) * makes the DNS fallback mechanism use the GCP metadata IP as a fallback before our hosted HTTP-based fallbacks I created a default GCP VM from their web wizard. It has no systemd-resolved. I then made its /etc/resolv.conf be empty and deleted its GCP hostnames in /etc/hosts. I then logged in to a tailnet with no global DNS settings. With this, tailscaled writes /etc/resolv.conf (direct mode, as no systemd-resolved) and sets it to 100.100.100.100, which then has regular DNS via the metadata IP and *.internal DNS via the metadata IP as well. If the tailnet configures explicit DNS servers, those are used instead, except for *.internal. This also adds a new util/cloudenv package based on version/distro where the cloud type is only detected once. We'll likely expand it in the future for other clouds, doing variants of this change for other popular cloud environments. Fixes #4911 RELNOTES=Google Cloud DNS improvements Change-Id: I19f3c2075983669b2b2c0f29a548da8de373c7cf Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
2 years ago
"sync/atomic"
"time"
ipn/ipnlocal, net/dns*, util/cloudenv: specialize DNS config on Google Cloud This does three things: * If you're on GCP, it adds a *.internal DNS split route to the metadata server, so we never break GCP DNS names. This lets people have some Tailscale nodes on GCP and some not (e.g. laptops at home) without having to add a Tailnet-wide *.internal DNS route. If you already have such a route, though, it won't overwrite it. * If the 100.100.100.100 DNS forwarder has nowhere to forward to, it forwards it to the GCP metadata IP, which forwards to 8.8.8.8. This means there are never errNoUpstreams ("upstream nameservers not set") errors on GCP due to e.g. mangled /etc/resolv.conf (GCP default VMs don't have systemd-resolved, so it's likely a DNS supremacy fight) * makes the DNS fallback mechanism use the GCP metadata IP as a fallback before our hosted HTTP-based fallbacks I created a default GCP VM from their web wizard. It has no systemd-resolved. I then made its /etc/resolv.conf be empty and deleted its GCP hostnames in /etc/hosts. I then logged in to a tailnet with no global DNS settings. With this, tailscaled writes /etc/resolv.conf (direct mode, as no systemd-resolved) and sets it to 100.100.100.100, which then has regular DNS via the metadata IP and *.internal DNS via the metadata IP as well. If the tailnet configures explicit DNS servers, those are used instead, except for *.internal. This also adds a new util/cloudenv package based on version/distro where the cloud type is only detected once. We'll likely expand it in the future for other clouds, doing variants of this change for other popular cloud environments. Fixes #4911 RELNOTES=Google Cloud DNS improvements Change-Id: I19f3c2075983669b2b2c0f29a548da8de373c7cf Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
2 years ago
)
// CommonNonRoutableMetadataIP is the IP address of the metadata server
// on Amazon EC2, Google Compute Engine, and Azure. It's not routable.
// (169.254.0.0/16 is a Link Local range: RFC 3927)
const CommonNonRoutableMetadataIP = "169.254.169.254"
ipn/ipnlocal, net/dns*, util/cloudenv: specialize DNS config on Google Cloud This does three things: * If you're on GCP, it adds a *.internal DNS split route to the metadata server, so we never break GCP DNS names. This lets people have some Tailscale nodes on GCP and some not (e.g. laptops at home) without having to add a Tailnet-wide *.internal DNS route. If you already have such a route, though, it won't overwrite it. * If the 100.100.100.100 DNS forwarder has nowhere to forward to, it forwards it to the GCP metadata IP, which forwards to 8.8.8.8. This means there are never errNoUpstreams ("upstream nameservers not set") errors on GCP due to e.g. mangled /etc/resolv.conf (GCP default VMs don't have systemd-resolved, so it's likely a DNS supremacy fight) * makes the DNS fallback mechanism use the GCP metadata IP as a fallback before our hosted HTTP-based fallbacks I created a default GCP VM from their web wizard. It has no systemd-resolved. I then made its /etc/resolv.conf be empty and deleted its GCP hostnames in /etc/hosts. I then logged in to a tailnet with no global DNS settings. With this, tailscaled writes /etc/resolv.conf (direct mode, as no systemd-resolved) and sets it to 100.100.100.100, which then has regular DNS via the metadata IP and *.internal DNS via the metadata IP as well. If the tailnet configures explicit DNS servers, those are used instead, except for *.internal. This also adds a new util/cloudenv package based on version/distro where the cloud type is only detected once. We'll likely expand it in the future for other clouds, doing variants of this change for other popular cloud environments. Fixes #4911 RELNOTES=Google Cloud DNS improvements Change-Id: I19f3c2075983669b2b2c0f29a548da8de373c7cf Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
2 years ago
// GoogleMetadataAndDNSIP is the metadata IP used by Google Cloud.
// It's also the *.internal DNS server, and proxies to 8.8.8.8.
const GoogleMetadataAndDNSIP = "169.254.169.254"
// AWSResolverIP is the IP address of the AWS DNS server.
// See https://docs.aws.amazon.com/vpc/latest/userguide/vpc-dns.html
const AWSResolverIP = "169.254.169.253"
// AzureResolverIP is Azure's DNS resolver IP.
// See https://docs.microsoft.com/en-us/azure/virtual-network/what-is-ip-address-168-63-129-16
const AzureResolverIP = "168.63.129.16"
ipn/ipnlocal, net/dns*, util/cloudenv: specialize DNS config on Google Cloud This does three things: * If you're on GCP, it adds a *.internal DNS split route to the metadata server, so we never break GCP DNS names. This lets people have some Tailscale nodes on GCP and some not (e.g. laptops at home) without having to add a Tailnet-wide *.internal DNS route. If you already have such a route, though, it won't overwrite it. * If the 100.100.100.100 DNS forwarder has nowhere to forward to, it forwards it to the GCP metadata IP, which forwards to 8.8.8.8. This means there are never errNoUpstreams ("upstream nameservers not set") errors on GCP due to e.g. mangled /etc/resolv.conf (GCP default VMs don't have systemd-resolved, so it's likely a DNS supremacy fight) * makes the DNS fallback mechanism use the GCP metadata IP as a fallback before our hosted HTTP-based fallbacks I created a default GCP VM from their web wizard. It has no systemd-resolved. I then made its /etc/resolv.conf be empty and deleted its GCP hostnames in /etc/hosts. I then logged in to a tailnet with no global DNS settings. With this, tailscaled writes /etc/resolv.conf (direct mode, as no systemd-resolved) and sets it to 100.100.100.100, which then has regular DNS via the metadata IP and *.internal DNS via the metadata IP as well. If the tailnet configures explicit DNS servers, those are used instead, except for *.internal. This also adds a new util/cloudenv package based on version/distro where the cloud type is only detected once. We'll likely expand it in the future for other clouds, doing variants of this change for other popular cloud environments. Fixes #4911 RELNOTES=Google Cloud DNS improvements Change-Id: I19f3c2075983669b2b2c0f29a548da8de373c7cf Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
2 years ago
// Cloud is a recognize cloud environment with properties that
// Tailscale can specialize for in places.
type Cloud string
const (
AWS = Cloud("aws") // Amazon Web Services (EC2 in particular)
Azure = Cloud("azure") // Microsoft Azure
GCP = Cloud("gcp") // Google Cloud
ipn/ipnlocal, net/dns*, util/cloudenv: specialize DNS config on Google Cloud This does three things: * If you're on GCP, it adds a *.internal DNS split route to the metadata server, so we never break GCP DNS names. This lets people have some Tailscale nodes on GCP and some not (e.g. laptops at home) without having to add a Tailnet-wide *.internal DNS route. If you already have such a route, though, it won't overwrite it. * If the 100.100.100.100 DNS forwarder has nowhere to forward to, it forwards it to the GCP metadata IP, which forwards to 8.8.8.8. This means there are never errNoUpstreams ("upstream nameservers not set") errors on GCP due to e.g. mangled /etc/resolv.conf (GCP default VMs don't have systemd-resolved, so it's likely a DNS supremacy fight) * makes the DNS fallback mechanism use the GCP metadata IP as a fallback before our hosted HTTP-based fallbacks I created a default GCP VM from their web wizard. It has no systemd-resolved. I then made its /etc/resolv.conf be empty and deleted its GCP hostnames in /etc/hosts. I then logged in to a tailnet with no global DNS settings. With this, tailscaled writes /etc/resolv.conf (direct mode, as no systemd-resolved) and sets it to 100.100.100.100, which then has regular DNS via the metadata IP and *.internal DNS via the metadata IP as well. If the tailnet configures explicit DNS servers, those are used instead, except for *.internal. This also adds a new util/cloudenv package based on version/distro where the cloud type is only detected once. We'll likely expand it in the future for other clouds, doing variants of this change for other popular cloud environments. Fixes #4911 RELNOTES=Google Cloud DNS improvements Change-Id: I19f3c2075983669b2b2c0f29a548da8de373c7cf Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
2 years ago
)
// ResolverIP returns the cloud host's recursive DNS server or the
// empty string if not available.
func (c Cloud) ResolverIP() string {
switch c {
case GCP:
return GoogleMetadataAndDNSIP
case AWS:
return AWSResolverIP
case Azure:
return AzureResolverIP
}
return ""
}
// HasInternalTLD reports whether c is a cloud environment
// whose ResolverIP serves *.internal records.
func (c Cloud) HasInternalTLD() bool {
switch c {
case GCP, AWS:
return true
}
return false
}
ipn/ipnlocal, net/dns*, util/cloudenv: specialize DNS config on Google Cloud This does three things: * If you're on GCP, it adds a *.internal DNS split route to the metadata server, so we never break GCP DNS names. This lets people have some Tailscale nodes on GCP and some not (e.g. laptops at home) without having to add a Tailnet-wide *.internal DNS route. If you already have such a route, though, it won't overwrite it. * If the 100.100.100.100 DNS forwarder has nowhere to forward to, it forwards it to the GCP metadata IP, which forwards to 8.8.8.8. This means there are never errNoUpstreams ("upstream nameservers not set") errors on GCP due to e.g. mangled /etc/resolv.conf (GCP default VMs don't have systemd-resolved, so it's likely a DNS supremacy fight) * makes the DNS fallback mechanism use the GCP metadata IP as a fallback before our hosted HTTP-based fallbacks I created a default GCP VM from their web wizard. It has no systemd-resolved. I then made its /etc/resolv.conf be empty and deleted its GCP hostnames in /etc/hosts. I then logged in to a tailnet with no global DNS settings. With this, tailscaled writes /etc/resolv.conf (direct mode, as no systemd-resolved) and sets it to 100.100.100.100, which then has regular DNS via the metadata IP and *.internal DNS via the metadata IP as well. If the tailnet configures explicit DNS servers, those are used instead, except for *.internal. This also adds a new util/cloudenv package based on version/distro where the cloud type is only detected once. We'll likely expand it in the future for other clouds, doing variants of this change for other popular cloud environments. Fixes #4911 RELNOTES=Google Cloud DNS improvements Change-Id: I19f3c2075983669b2b2c0f29a548da8de373c7cf Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
2 years ago
var cloudAtomic atomic.Value // of Cloud
// Get returns the current cloud, or the empty string if unknown.
func Get() Cloud {
c, ok := cloudAtomic.Load().(Cloud)
if ok {
return c
}
c = getCloud()
cloudAtomic.Store(c) // even if empty
return c
}
func readFileTrimmed(name string) string {
v, _ := os.ReadFile(name)
return strings.TrimSpace(string(v))
}
ipn/ipnlocal, net/dns*, util/cloudenv: specialize DNS config on Google Cloud This does three things: * If you're on GCP, it adds a *.internal DNS split route to the metadata server, so we never break GCP DNS names. This lets people have some Tailscale nodes on GCP and some not (e.g. laptops at home) without having to add a Tailnet-wide *.internal DNS route. If you already have such a route, though, it won't overwrite it. * If the 100.100.100.100 DNS forwarder has nowhere to forward to, it forwards it to the GCP metadata IP, which forwards to 8.8.8.8. This means there are never errNoUpstreams ("upstream nameservers not set") errors on GCP due to e.g. mangled /etc/resolv.conf (GCP default VMs don't have systemd-resolved, so it's likely a DNS supremacy fight) * makes the DNS fallback mechanism use the GCP metadata IP as a fallback before our hosted HTTP-based fallbacks I created a default GCP VM from their web wizard. It has no systemd-resolved. I then made its /etc/resolv.conf be empty and deleted its GCP hostnames in /etc/hosts. I then logged in to a tailnet with no global DNS settings. With this, tailscaled writes /etc/resolv.conf (direct mode, as no systemd-resolved) and sets it to 100.100.100.100, which then has regular DNS via the metadata IP and *.internal DNS via the metadata IP as well. If the tailnet configures explicit DNS servers, those are used instead, except for *.internal. This also adds a new util/cloudenv package based on version/distro where the cloud type is only detected once. We'll likely expand it in the future for other clouds, doing variants of this change for other popular cloud environments. Fixes #4911 RELNOTES=Google Cloud DNS improvements Change-Id: I19f3c2075983669b2b2c0f29a548da8de373c7cf Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
2 years ago
func getCloud() Cloud {
var hitMetadata bool
switch runtime.GOOS {
case "android", "ios", "darwin":
// Assume these aren't running on a cloud.
return ""
case "linux":
biosVendor := readFileTrimmed("/sys/class/dmi/id/bios_vendor")
if biosVendor == "Amazon EC2" || strings.HasSuffix(biosVendor, ".amazon") {
return AWS
}
prod := readFileTrimmed("/sys/class/dmi/id/product_name")
if prod == "Google Compute Engine" {
return GCP
}
if prod == "Google" { // old GCP VMs, it seems
hitMetadata = true
}
if prod == "Virtual Machine" || biosVendor == "Microsoft Corporation" {
// Azure, or maybe all Hyper-V?
hitMetadata = true
}
default:
// TODO(bradfitz): use Win32_SystemEnclosure from WMI or something on
// Windows to see if it's a physical machine and skip the cloud check
// early. Otherwise use similar clues as Linux about whether we should
// burn up to 2 seconds waiting for a metadata server that might not be
// there. And for BSDs, look where the /sys stuff is.
return ""
}
if !hitMetadata {
return ""
}
const maxWait = 2 * time.Second
tr := &http.Transport{
DisableKeepAlives: true,
Dial: (&net.Dialer{
Timeout: maxWait,
}).Dial,
}
ctx, cancel := context.WithTimeout(context.Background(), maxWait)
defer cancel()
// We want to hit CommonNonRoutableMetadataIP to see if we're on AWS, GCP,
// or Azure. All three (and many others) use the same metadata IP.
//
// But to avoid triggering the AWS CloudWatch "MetadataNoToken" metric (for which
// there might be an alert registered?), make our initial request be a token
// request. This only works on AWS, but the failing HTTP response on other clouds gives
// us enough clues about which cloud we're on.
req, err := http.NewRequestWithContext(ctx, "PUT", "http://"+CommonNonRoutableMetadataIP+"/latest/api/token", strings.NewReader(""))
if err != nil {
log.Printf("cloudenv: [unexpected] error creating request: %v", err)
return ""
}
req.Header.Set("X-Aws-Ec2-Metadata-Token-Ttl-Seconds", "5")
res, err := tr.RoundTrip(req)
if err != nil {
return ""
}
res.Body.Close()
if res.Header.Get("Metadata-Flavor") == "Google" {
ipn/ipnlocal, net/dns*, util/cloudenv: specialize DNS config on Google Cloud This does three things: * If you're on GCP, it adds a *.internal DNS split route to the metadata server, so we never break GCP DNS names. This lets people have some Tailscale nodes on GCP and some not (e.g. laptops at home) without having to add a Tailnet-wide *.internal DNS route. If you already have such a route, though, it won't overwrite it. * If the 100.100.100.100 DNS forwarder has nowhere to forward to, it forwards it to the GCP metadata IP, which forwards to 8.8.8.8. This means there are never errNoUpstreams ("upstream nameservers not set") errors on GCP due to e.g. mangled /etc/resolv.conf (GCP default VMs don't have systemd-resolved, so it's likely a DNS supremacy fight) * makes the DNS fallback mechanism use the GCP metadata IP as a fallback before our hosted HTTP-based fallbacks I created a default GCP VM from their web wizard. It has no systemd-resolved. I then made its /etc/resolv.conf be empty and deleted its GCP hostnames in /etc/hosts. I then logged in to a tailnet with no global DNS settings. With this, tailscaled writes /etc/resolv.conf (direct mode, as no systemd-resolved) and sets it to 100.100.100.100, which then has regular DNS via the metadata IP and *.internal DNS via the metadata IP as well. If the tailnet configures explicit DNS servers, those are used instead, except for *.internal. This also adds a new util/cloudenv package based on version/distro where the cloud type is only detected once. We'll likely expand it in the future for other clouds, doing variants of this change for other popular cloud environments. Fixes #4911 RELNOTES=Google Cloud DNS improvements Change-Id: I19f3c2075983669b2b2c0f29a548da8de373c7cf Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
2 years ago
return GCP
}
server := res.Header.Get("Server")
if server == "EC2ws" {
return AWS
}
if strings.HasPrefix(server, "Microsoft") {
// e.g. "Microsoft-IIS/10.0"
req, _ := http.NewRequestWithContext(ctx, "GET", "http://"+CommonNonRoutableMetadataIP+"/metadata/instance/compute?api-version=2021-02-01", nil)
req.Header.Set("Metadata", "true")
res, err := tr.RoundTrip(req)
if err != nil {
return ""
}
defer res.Body.Close()
var meta struct {
AzEnvironment string `json:"azEnvironment"`
}
if err := json.NewDecoder(res.Body).Decode(&meta); err != nil {
return ""
}
if strings.HasPrefix(meta.AzEnvironment, "Azure") {
return Azure
}
return ""
}
ipn/ipnlocal, net/dns*, util/cloudenv: specialize DNS config on Google Cloud This does three things: * If you're on GCP, it adds a *.internal DNS split route to the metadata server, so we never break GCP DNS names. This lets people have some Tailscale nodes on GCP and some not (e.g. laptops at home) without having to add a Tailnet-wide *.internal DNS route. If you already have such a route, though, it won't overwrite it. * If the 100.100.100.100 DNS forwarder has nowhere to forward to, it forwards it to the GCP metadata IP, which forwards to 8.8.8.8. This means there are never errNoUpstreams ("upstream nameservers not set") errors on GCP due to e.g. mangled /etc/resolv.conf (GCP default VMs don't have systemd-resolved, so it's likely a DNS supremacy fight) * makes the DNS fallback mechanism use the GCP metadata IP as a fallback before our hosted HTTP-based fallbacks I created a default GCP VM from their web wizard. It has no systemd-resolved. I then made its /etc/resolv.conf be empty and deleted its GCP hostnames in /etc/hosts. I then logged in to a tailnet with no global DNS settings. With this, tailscaled writes /etc/resolv.conf (direct mode, as no systemd-resolved) and sets it to 100.100.100.100, which then has regular DNS via the metadata IP and *.internal DNS via the metadata IP as well. If the tailnet configures explicit DNS servers, those are used instead, except for *.internal. This also adds a new util/cloudenv package based on version/distro where the cloud type is only detected once. We'll likely expand it in the future for other clouds, doing variants of this change for other popular cloud environments. Fixes #4911 RELNOTES=Google Cloud DNS improvements Change-Id: I19f3c2075983669b2b2c0f29a548da8de373c7cf Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
2 years ago
// TODO: more, as needed.
return ""
}