+ dnsfilter: use AG DNS server for SB/PC services

* move SS/SB/PC services to security.go
* remove old useless code (HTTP client)
This commit is contained in:
Simon Zolin
2019-10-16 12:57:49 +03:00
parent 2f1e631c66
commit 7f69848084
6 changed files with 308 additions and 602 deletions

View File

@@ -1,13 +1,7 @@
package dnsfilter
import (
"bufio"
"bytes"
"context"
"crypto/sha256"
"encoding/binary"
"encoding/gob"
"encoding/json"
"fmt"
"io/ioutil"
"net"
@@ -16,30 +10,14 @@ import (
"runtime"
"strings"
"sync"
"sync/atomic"
"time"
"github.com/joomcode/errorx"
"github.com/AdguardTeam/dnsproxy/upstream"
"github.com/AdguardTeam/golibs/cache"
"github.com/AdguardTeam/golibs/log"
"github.com/AdguardTeam/urlfilter"
"github.com/bluele/gcache"
"github.com/miekg/dns"
"golang.org/x/net/publicsuffix"
)
const defaultHTTPTimeout = 5 * time.Minute
const defaultHTTPMaxIdleConnections = 100
const defaultSafebrowsingServer = "sb.adtidy.org"
const defaultSafebrowsingURL = "%s://%s/safebrowsing-lookup-hash.html?prefixes=%s"
const defaultParentalServer = "pctrl.adguard.com"
const defaultParentalURL = "%s://%s/check-parental-control-hash?prefixes=%s&sensitivity=%d"
const defaultParentalSensitivity = 13 // use "TEEN" by default
const maxDialCacheSize = 2 // the number of host names for safebrowsing and parental control
// ServiceEntry - blocked service array element
type ServiceEntry struct {
Name string
@@ -65,7 +43,6 @@ type RewriteEntry struct {
type Config struct {
ParentalSensitivity int `yaml:"parental_sensitivity"` // must be either 3, 10, 13 or 17
ParentalEnabled bool `yaml:"parental_enabled"`
UsePlainHTTP bool `yaml:"-"` // use plain HTTP for requests to parental and safe browsing servers
SafeSearchEnabled bool `yaml:"safesearch_enabled"`
SafeBrowsingEnabled bool `yaml:"safebrowsing_enabled"`
ResolverAddress string `yaml:"-"` // DNS server address
@@ -110,12 +87,10 @@ type Dnsfilter struct {
filteringEngine *urlfilter.DNSEngine
engineLock sync.RWMutex
// HTTP lookups for safebrowsing and parental
client http.Client // handle for http client -- single instance as recommended by docs
transport *http.Transport // handle for http transport used by http client
parentalServer string // access via methods
safeBrowsingServer string // access via methods
parentalServer string // access via methods
safeBrowsingServer string // access via methods
parentalUpstream upstream.Upstream
safeBrowsingUpstream upstream.Upstream
Config // for direct access by library users, even a = assignment
confLock sync.RWMutex
@@ -251,9 +226,6 @@ func (d *Dnsfilter) filtersInitializer() {
// Close - close the object
func (d *Dnsfilter) Close() {
if d != nil && d.transport != nil {
d.transport.CloseIdleConnections()
}
if d.rulesStorage != nil {
d.rulesStorage.Close()
}
@@ -261,7 +233,6 @@ func (d *Dnsfilter) Close() {
type dnsFilterContext struct {
stats Stats
dialCache gcache.Cache // "host" -> "IP" cache for safebrowsing and parental control servers
safebrowsingCache cache.Cache
parentalCache cache.Cache
safeSearchCache cache.Cache
@@ -328,11 +299,10 @@ func (d *Dnsfilter) CheckHost(host string, qtype uint16, setts *RequestFiltering
}
}
// check safeSearch if no match
if setts.SafeSearchEnabled {
result, err = d.checkSafeSearch(host)
if err != nil {
log.Printf("Failed to safesearch HTTP lookup, ignoring check: %v", err)
log.Info("SafeSearch: failed: %v", err)
return Result{}, nil
}
@@ -341,12 +311,10 @@ func (d *Dnsfilter) CheckHost(host string, qtype uint16, setts *RequestFiltering
}
}
// check safebrowsing if no match
if setts.SafeBrowsingEnabled {
result, err = d.checkSafeBrowsing(host)
if err != nil {
// failed to do HTTP lookup -- treat it as if we got empty response, but don't save cache
log.Printf("Failed to do safebrowsing HTTP lookup, ignoring check: %v", err)
log.Info("SafeBrowsing: failed: %v", err)
return Result{}, nil
}
if result.Reason.Matched() {
@@ -354,12 +322,10 @@ func (d *Dnsfilter) CheckHost(host string, qtype uint16, setts *RequestFiltering
}
}
// check parental if no match
if setts.ParentalEnabled {
result, err = d.checkParental(host)
if err != nil {
// failed to do HTTP lookup -- treat it as if we got empty response, but don't save cache
log.Printf("Failed to do parental HTTP lookup, ignoring check: %v", err)
log.Printf("Parental: failed: %v", err)
return Result{}, nil
}
if result.Reason.Matched() {
@@ -367,7 +333,6 @@ func (d *Dnsfilter) CheckHost(host string, qtype uint16, setts *RequestFiltering
}
}
// nothing matched, return nothing
return Result{}, nil
}
@@ -445,311 +410,6 @@ func matchBlockedServicesRules(host string, svcs []ServiceEntry) Result {
return res
}
/*
expire byte[4]
res Result
*/
func (d *Dnsfilter) setCacheResult(cache cache.Cache, host string, res Result) {
var buf bytes.Buffer
expire := uint(time.Now().Unix()) + d.Config.CacheTime*60
var exp []byte
exp = make([]byte, 4)
binary.BigEndian.PutUint32(exp, uint32(expire))
_, _ = buf.Write(exp)
enc := gob.NewEncoder(&buf)
err := enc.Encode(res)
if err != nil {
log.Error("gob.Encode(): %s", err)
return
}
_ = cache.Set([]byte(host), buf.Bytes())
log.Debug("Stored in cache %p: %s", cache, host)
}
func getCachedResult(cache cache.Cache, host string) (Result, bool) {
data := cache.Get([]byte(host))
if data == nil {
return Result{}, false
}
exp := int(binary.BigEndian.Uint32(data[:4]))
if exp <= int(time.Now().Unix()) {
cache.Del([]byte(host))
return Result{}, false
}
var buf bytes.Buffer
buf.Write(data[4:])
dec := gob.NewDecoder(&buf)
r := Result{}
err := dec.Decode(&r)
if err != nil {
log.Debug("gob.Decode(): %s", err)
return Result{}, false
}
return r, true
}
// for each dot, hash it and add it to string
func hostnameToHashParam(host string, addslash bool) (string, map[string]bool) {
var hashparam bytes.Buffer
hashes := map[string]bool{}
tld, icann := publicsuffix.PublicSuffix(host)
if !icann {
// private suffixes like cloudfront.net
tld = ""
}
curhost := host
for {
if curhost == "" {
// we've reached end of string
break
}
if tld != "" && curhost == tld {
// we've reached the TLD, don't hash it
break
}
tohash := []byte(curhost)
if addslash {
tohash = append(tohash, '/')
}
sum := sha256.Sum256(tohash)
hexhash := fmt.Sprintf("%X", sum)
hashes[hexhash] = true
hashparam.WriteString(fmt.Sprintf("%02X%02X%02X%02X/", sum[0], sum[1], sum[2], sum[3]))
pos := strings.IndexByte(curhost, byte('.'))
if pos < 0 {
break
}
curhost = curhost[pos+1:]
}
return hashparam.String(), hashes
}
func (d *Dnsfilter) checkSafeSearch(host string) (Result, error) {
if log.GetLevel() >= log.DEBUG {
timer := log.StartTimer()
defer timer.LogElapsed("SafeSearch HTTP lookup for %s", host)
}
// Check cache. Return cached result if it was found
cachedValue, isFound := getCachedResult(gctx.safeSearchCache, host)
if isFound {
atomic.AddUint64(&gctx.stats.Safesearch.CacheHits, 1)
log.Tracef("%s: found in SafeSearch cache", host)
return cachedValue, nil
}
safeHost, ok := d.SafeSearchDomain(host)
if !ok {
return Result{}, nil
}
res := Result{IsFiltered: true, Reason: FilteredSafeSearch}
if ip := net.ParseIP(safeHost); ip != nil {
res.IP = ip
d.setCacheResult(gctx.safeSearchCache, host, res)
return res, nil
}
// TODO this address should be resolved with upstream that was configured in dnsforward
addrs, err := net.LookupIP(safeHost)
if err != nil {
log.Tracef("SafeSearchDomain for %s was found but failed to lookup for %s cause %s", host, safeHost, err)
return Result{}, err
}
for _, i := range addrs {
if ipv4 := i.To4(); ipv4 != nil {
res.IP = ipv4
break
}
}
if len(res.IP) == 0 {
return Result{}, fmt.Errorf("no ipv4 addresses in safe search response for %s", safeHost)
}
// Cache result
d.setCacheResult(gctx.safeSearchCache, host, res)
return res, nil
}
func (d *Dnsfilter) checkSafeBrowsing(host string) (Result, error) {
if log.GetLevel() >= log.DEBUG {
timer := log.StartTimer()
defer timer.LogElapsed("SafeBrowsing HTTP lookup for %s", host)
}
format := func(hashparam string) string {
schema := "https"
if d.UsePlainHTTP {
schema = "http"
}
url := fmt.Sprintf(defaultSafebrowsingURL, schema, d.safeBrowsingServer, hashparam)
return url
}
handleBody := func(body []byte, hashes map[string]bool) (Result, error) {
result := Result{}
scanner := bufio.NewScanner(strings.NewReader(string(body)))
for scanner.Scan() {
line := scanner.Text()
splitted := strings.Split(line, ":")
if len(splitted) < 3 {
continue
}
hash := splitted[2]
if _, ok := hashes[hash]; ok {
// it's in the hash
result.IsFiltered = true
result.Reason = FilteredSafeBrowsing
result.Rule = splitted[0]
break
}
}
if err := scanner.Err(); err != nil {
// error, don't save cache
return Result{}, err
}
return result, nil
}
// check cache
cachedValue, isFound := getCachedResult(gctx.safebrowsingCache, host)
if isFound {
atomic.AddUint64(&gctx.stats.Safebrowsing.CacheHits, 1)
log.Tracef("%s: found in the lookup cache %p", host, gctx.safebrowsingCache)
return cachedValue, nil
}
result, err := d.lookupCommon(host, &gctx.stats.Safebrowsing, true, format, handleBody)
if err == nil {
d.setCacheResult(gctx.safebrowsingCache, host, result)
}
return result, err
}
func (d *Dnsfilter) checkParental(host string) (Result, error) {
if log.GetLevel() >= log.DEBUG {
timer := log.StartTimer()
defer timer.LogElapsed("Parental HTTP lookup for %s", host)
}
format := func(hashparam string) string {
schema := "https"
if d.UsePlainHTTP {
schema = "http"
}
sensitivity := d.ParentalSensitivity
if sensitivity == 0 {
sensitivity = defaultParentalSensitivity
}
url := fmt.Sprintf(defaultParentalURL, schema, d.parentalServer, hashparam, sensitivity)
return url
}
handleBody := func(body []byte, hashes map[string]bool) (Result, error) {
// parse json
var m []struct {
Blocked bool `json:"blocked"`
ClientTTL int `json:"clientTtl"`
Reason string `json:"reason"`
Hash string `json:"hash"`
}
err := json.Unmarshal(body, &m)
if err != nil {
// error, don't save cache
log.Printf("Couldn't parse json '%s': %s", body, err)
return Result{}, err
}
result := Result{}
for i := range m {
if !hashes[m[i].Hash] {
continue
}
if m[i].Blocked {
result.IsFiltered = true
result.Reason = FilteredParental
result.Rule = fmt.Sprintf("parental %s", m[i].Reason)
break
}
}
return result, nil
}
// check cache
cachedValue, isFound := getCachedResult(gctx.parentalCache, host)
if isFound {
atomic.AddUint64(&gctx.stats.Parental.CacheHits, 1)
log.Tracef("%s: found in the lookup cache %p", host, gctx.parentalCache)
return cachedValue, nil
}
result, err := d.lookupCommon(host, &gctx.stats.Parental, false, format, handleBody)
if err == nil {
d.setCacheResult(gctx.parentalCache, host, result)
}
return result, err
}
type formatHandler func(hashparam string) string
type bodyHandler func(body []byte, hashes map[string]bool) (Result, error)
// real implementation of lookup/check
func (d *Dnsfilter) lookupCommon(host string, lookupstats *LookupStats, hashparamNeedSlash bool, format formatHandler, handleBody bodyHandler) (Result, error) {
// convert hostname to hash parameters
hashparam, hashes := hostnameToHashParam(host, hashparamNeedSlash)
// format URL with our hashes
url := format(hashparam)
// do HTTP request
atomic.AddUint64(&lookupstats.Requests, 1)
atomic.AddInt64(&lookupstats.Pending, 1)
updateMax(&lookupstats.Pending, &lookupstats.PendingMax)
resp, err := d.client.Get(url)
atomic.AddInt64(&lookupstats.Pending, -1)
if resp != nil && resp.Body != nil {
defer resp.Body.Close()
}
if err != nil {
// error, don't save cache
return Result{}, err
}
// get body text
body, err := ioutil.ReadAll(resp.Body)
if err != nil {
// error, don't save cache
return Result{}, err
}
// handle status code
switch {
case resp.StatusCode == 204:
// empty result, save cache
return Result{}, nil
case resp.StatusCode != 200:
return Result{}, fmt.Errorf("HTTP status code: %d", resp.StatusCode)
}
result, err := handleBody(body, hashes)
if err != nil {
return Result{}, err
}
return result, nil
}
//
// Adding rule and matching against the rules
//
@@ -887,97 +547,6 @@ func (d *Dnsfilter) matchHost(host string, qtype uint16) (Result, error) {
return Result{}, nil
}
//
// lifecycle helper functions
//
// Return TRUE if this host's IP should be cached
func (d *Dnsfilter) shouldBeInDialCache(host string) bool {
return host == d.safeBrowsingServer ||
host == d.parentalServer
}
// Search for an IP address by host name
func searchInDialCache(host string) string {
rawValue, err := gctx.dialCache.Get(host)
if err != nil {
return ""
}
ip, _ := rawValue.(string)
log.Debug("Found in cache: %s -> %s", host, ip)
return ip
}
// Add "hostname" -> "IP address" entry to cache
func addToDialCache(host, ip string) {
err := gctx.dialCache.Set(host, ip)
if err != nil {
log.Debug("dialCache.Set: %s", err)
}
log.Debug("Added to cache: %s -> %s", host, ip)
}
type dialFunctionType func(ctx context.Context, network, addr string) (net.Conn, error)
// Connect to a remote server resolving hostname using our own DNS server
func (d *Dnsfilter) createCustomDialContext(resolverAddr string) dialFunctionType {
return func(ctx context.Context, network, addr string) (net.Conn, error) {
log.Tracef("network:%v addr:%v", network, addr)
host, port, err := net.SplitHostPort(addr)
if err != nil {
return nil, err
}
dialer := &net.Dialer{
Timeout: time.Minute * 5,
}
if net.ParseIP(host) != nil {
con, err := dialer.DialContext(ctx, network, addr)
return con, err
}
cache := d.shouldBeInDialCache(host)
if cache {
ip := searchInDialCache(host)
if len(ip) != 0 {
addr = fmt.Sprintf("%s:%s", ip, port)
return dialer.DialContext(ctx, network, addr)
}
}
r := upstream.NewResolver(resolverAddr, 30*time.Second)
addrs, e := r.LookupIPAddr(ctx, host)
log.Tracef("LookupIPAddr: %s: %v", host, addrs)
if e != nil {
return nil, e
}
if len(addrs) == 0 {
return nil, fmt.Errorf("couldn't lookup host: %s", host)
}
var dialErrs []error
for _, a := range addrs {
addr = fmt.Sprintf("%s:%s", a.String(), port)
con, err := dialer.DialContext(ctx, network, addr)
if err != nil {
dialErrs = append(dialErrs, err)
continue
}
if cache {
addToDialCache(host, a.String())
}
return con, err
}
return nil, errorx.DecorateMany(fmt.Sprintf("couldn't dial to %s", addr), dialErrs...)
}
}
// New creates properly initialized DNS Filter that is ready to be used
func New(c *Config, filters map[int]string) *Dnsfilter {
@@ -1002,34 +571,16 @@ func New(c *Config, filters map[int]string) *Dnsfilter {
cacheConf.MaxSize = c.ParentalCacheSize
gctx.parentalCache = cache.New(cacheConf)
}
if len(c.ResolverAddress) != 0 && gctx.dialCache == nil {
dur := time.Duration(c.CacheTime) * time.Minute
gctx.dialCache = gcache.New(maxDialCacheSize).LRU().Expiration(dur).Build()
}
}
d := new(Dnsfilter)
// Customize the Transport to have larger connection pool,
// We are not (re)using http.DefaultTransport because of race conditions found by tests
d.transport = &http.Transport{
Proxy: http.ProxyFromEnvironment,
MaxIdleConns: defaultHTTPMaxIdleConnections, // default 100
MaxIdleConnsPerHost: defaultHTTPMaxIdleConnections, // default 2
IdleConnTimeout: 90 * time.Second,
TLSHandshakeTimeout: 10 * time.Second,
ExpectContinueTimeout: 1 * time.Second,
err := d.initSecurityServices()
if err != nil {
log.Error("dnsfilter: initialize services: %s", err)
return nil
}
if c != nil && len(c.ResolverAddress) != 0 {
d.transport.DialContext = d.createCustomDialContext(c.ResolverAddress)
}
d.client = http.Client{
Transport: d.transport,
Timeout: defaultHTTPTimeout,
}
d.safeBrowsingServer = defaultSafebrowsingServer
d.parentalServer = defaultParentalServer
if c != nil {
d.Config = *c
}
@@ -1053,38 +604,6 @@ func New(c *Config, filters map[int]string) *Dnsfilter {
return d
}
//
// config manipulation helpers
//
// SetSafeBrowsingServer lets you optionally change hostname of safesearch lookup
func (d *Dnsfilter) SetSafeBrowsingServer(host string) {
if len(host) == 0 {
d.safeBrowsingServer = defaultSafebrowsingServer
} else {
d.safeBrowsingServer = host
}
}
// SetHTTPTimeout lets you optionally change timeout during lookups
func (d *Dnsfilter) SetHTTPTimeout(t time.Duration) {
d.client.Timeout = t
}
// ResetHTTPTimeout resets lookup timeouts
func (d *Dnsfilter) ResetHTTPTimeout() {
d.client.Timeout = defaultHTTPTimeout
}
// SafeSearchDomain returns replacement address for search engine
func (d *Dnsfilter) SafeSearchDomain(host string) (string, bool) {
if d.SafeSearchEnabled {
val, ok := safeSearchDomains[host]
return val, ok
}
return "", false
}
//
// stats
//