proxy.go

     1  // Copyright 2017 The Go Authors. All rights reserved.
     2  // Use of this source code is governed by a BSD-style
     3  // license that can be found in the LICENSE file.
     4  
     5  // Package httpproxy provides support for HTTP proxy determination
     6  // based on environment variables, as provided by net/http's
     7  // ProxyFromEnvironment function.
     8  //
     9  // The API is not subject to the Go 1 compatibility promise and may change at
    10  // any time.
    11  package httpproxy
    12  
    13  import (
    14  	"errors"
    15  	"fmt"
    16  	"net"
    17  	"net/url"
    18  	"os"
    19  	"strings"
    20  	"unicode/utf8"
    21  
    22  	"golang.org/x/net/idna"
    23  )
    24  
    25  // Config holds configuration for HTTP proxy settings. See
    26  // FromEnvironment for details.
    27  type Config struct {
    28  	// HTTPProxy represents the value of the HTTP_PROXY or
    29  	// http_proxy environment variable. It will be used as the proxy
    30  	// URL for HTTP requests unless overridden by NoProxy.
    31  	HTTPProxy string
    32  
    33  	// HTTPSProxy represents the HTTPS_PROXY or https_proxy
    34  	// environment variable. It will be used as the proxy URL for
    35  	// HTTPS requests unless overridden by NoProxy.
    36  	HTTPSProxy string
    37  
    38  	// NoProxy represents the NO_PROXY or no_proxy environment
    39  	// variable. It specifies a string that contains comma-separated values
    40  	// specifying hosts that should be excluded from proxying. Each value is
    41  	// represented by an IP address prefix (1.2.3.4), an IP address prefix in
    42  	// CIDR notation (1.2.3.4/8), a domain name, or a special DNS label (*).
    43  	// An IP address prefix and domain name can also include a literal port
    44  	// number (1.2.3.4:80).
    45  	// A domain name matches that name and all subdomains. A domain name with
    46  	// a leading "." matches subdomains only. For example "foo.com" matches
    47  	// "foo.com" and "bar.foo.com"; ".y.com" matches "x.y.com" but not "y.com".
    48  	// A single asterisk (*) indicates that no proxying should be done.
    49  	// A best effort is made to parse the string and errors are
    50  	// ignored.
    51  	NoProxy string
    52  
    53  	// CGI holds whether the current process is running
    54  	// as a CGI handler (FromEnvironment infers this from the
    55  	// presence of a REQUEST_METHOD environment variable).
    56  	// When this is set, ProxyForURL will return an error
    57  	// when HTTPProxy applies, because a client could be
    58  	// setting HTTP_PROXY maliciously. See https://golang.org/s/cgihttpproxy.
    59  	CGI bool
    60  }
    61  
    62  // config holds the parsed configuration for HTTP proxy settings.
    63  type config struct {
    64  	// Config represents the original configuration as defined above.
    65  	Config
    66  
    67  	// httpsProxy is the parsed URL of the HTTPSProxy if defined.
    68  	httpsProxy *url.URL
    69  
    70  	// httpProxy is the parsed URL of the HTTPProxy if defined.
    71  	httpProxy *url.URL
    72  
    73  	// ipMatchers represent all values in the NoProxy that are IP address
    74  	// prefixes or an IP address in CIDR notation.
    75  	ipMatchers []matcher
    76  
    77  	// domainMatchers represent all values in the NoProxy that are a domain
    78  	// name or hostname & domain name
    79  	domainMatchers []matcher
    80  }
    81  
    82  // FromEnvironment returns a Config instance populated from the
    83  // environment variables HTTP_PROXY, HTTPS_PROXY and NO_PROXY (or the
    84  // lowercase versions thereof).
    85  //
    86  // The environment values may be either a complete URL or a
    87  // "host[:port]", in which case the "http" scheme is assumed. An error
    88  // is returned if the value is a different form.
    89  func FromEnvironment() *Config {
    90  	return &Config{
    91  		HTTPProxy:  getEnvAny("HTTP_PROXY", "http_proxy"),
    92  		HTTPSProxy: getEnvAny("HTTPS_PROXY", "https_proxy"),
    93  		NoProxy:    getEnvAny("NO_PROXY", "no_proxy"),
    94  		CGI:        os.Getenv("REQUEST_METHOD") != "",
    95  	}
    96  }
    97  
    98  func getEnvAny(names ...string) string {
    99  	for _, n := range names {
   100  		if val := os.Getenv(n); val != "" {
   101  			return val
   102  		}
   103  	}
   104  	return ""
   105  }
   106  
   107  // ProxyFunc returns a function that determines the proxy URL to use for
   108  // a given request URL. Changing the contents of cfg will not affect
   109  // proxy functions created earlier.
   110  //
   111  // A nil URL and nil error are returned if no proxy is defined in the
   112  // environment, or a proxy should not be used for the given request, as
   113  // defined by NO_PROXY.
   114  //
   115  // As a special case, if req.URL.Host is "localhost" or a loopback address
   116  // (with or without a port number), then a nil URL and nil error will be returned.
   117  func (cfg *Config) ProxyFunc() func(reqURL *url.URL) (*url.URL, error) {
   118  	// Preprocess the Config settings for more efficient evaluation.
   119  	cfg1 := &config{
   120  		Config: *cfg,
   121  	}
   122  	cfg1.init()
   123  	return cfg1.proxyForURL
   124  }
   125  
   126  func (cfg *config) proxyForURL(reqURL *url.URL) (*url.URL, error) {
   127  	var proxy *url.URL
   128  	if reqURL.Scheme == "https" {
   129  		proxy = cfg.httpsProxy
   130  	} else if reqURL.Scheme == "http" {
   131  		proxy = cfg.httpProxy
   132  		if proxy != nil && cfg.CGI {
   133  			return nil, errors.New("refusing to use HTTP_PROXY value in CGI environment; see golang.org/s/cgihttpproxy")
   134  		}
   135  	}
   136  	if proxy == nil {
   137  		return nil, nil
   138  	}
   139  	if !cfg.useProxy(canonicalAddr(reqURL)) {
   140  		return nil, nil
   141  	}
   142  
   143  	return proxy, nil
   144  }
   145  
   146  func parseProxy(proxy string) (*url.URL, error) {
   147  	if proxy == "" {
   148  		return nil, nil
   149  	}
   150  
   151  	proxyURL, err := url.Parse(proxy)
   152  	if err != nil || proxyURL.Scheme == "" || proxyURL.Host == "" {
   153  		// proxy was bogus. Try prepending "http://" to it and
   154  		// see if that parses correctly. If not, we fall
   155  		// through and complain about the original one.
   156  		if proxyURL, err := url.Parse("http://" + proxy); err == nil {
   157  			return proxyURL, nil
   158  		}
   159  	}
   160  	if err != nil {
   161  		return nil, fmt.Errorf("invalid proxy address %q: %v", proxy, err)
   162  	}
   163  	return proxyURL, nil
   164  }
   165  
   166  // useProxy reports whether requests to addr should use a proxy,
   167  // according to the NO_PROXY or no_proxy environment variable.
   168  // addr is always a canonicalAddr with a host and port.
   169  func (cfg *config) useProxy(addr string) bool {
   170  	if len(addr) == 0 {
   171  		return true
   172  	}
   173  	host, port, err := net.SplitHostPort(addr)
   174  	if err != nil {
   175  		return false
   176  	}
   177  	if host == "localhost" {
   178  		return false
   179  	}
   180  	ip := net.ParseIP(host)
   181  	if ip != nil {
   182  		if ip.IsLoopback() {
   183  			return false
   184  		}
   185  	}
   186  
   187  	addr = strings.ToLower(strings.TrimSpace(host))
   188  
   189  	if ip != nil {
   190  		for _, m := range cfg.ipMatchers {
   191  			if m.match(addr, port, ip) {
   192  				return false
   193  			}
   194  		}
   195  	}
   196  	for _, m := range cfg.domainMatchers {
   197  		if m.match(addr, port, ip) {
   198  			return false
   199  		}
   200  	}
   201  	return true
   202  }
   203  
   204  func (c *config) init() {
   205  	if parsed, err := parseProxy(c.HTTPProxy); err == nil {
   206  		c.httpProxy = parsed
   207  	}
   208  	if parsed, err := parseProxy(c.HTTPSProxy); err == nil {
   209  		c.httpsProxy = parsed
   210  	}
   211  
   212  	for _, p := range strings.Split(c.NoProxy, ",") {
   213  		p = strings.ToLower(strings.TrimSpace(p))
   214  		if len(p) == 0 {
   215  			continue
   216  		}
   217  
   218  		if p == "*" {
   219  			c.ipMatchers = []matcher{allMatch{}}
   220  			c.domainMatchers = []matcher{allMatch{}}
   221  			return
   222  		}
   223  
   224  		// IPv4/CIDR, IPv6/CIDR
   225  		if _, pnet, err := net.ParseCIDR(p); err == nil {
   226  			c.ipMatchers = append(c.ipMatchers, cidrMatch{cidr: pnet})
   227  			continue
   228  		}
   229  
   230  		// IPv4:port, [IPv6]:port
   231  		phost, pport, err := net.SplitHostPort(p)
   232  		if err == nil {
   233  			if len(phost) == 0 {
   234  				// There is no host part, likely the entry is malformed; ignore.
   235  				continue
   236  			}
   237  			if phost[0] == '[' && phost[len(phost)-1] == ']' {
   238  				phost = phost[1 : len(phost)-1]
   239  			}
   240  		} else {
   241  			phost = p
   242  		}
   243  		// IPv4, IPv6
   244  		if pip := net.ParseIP(phost); pip != nil {
   245  			c.ipMatchers = append(c.ipMatchers, ipMatch{ip: pip, port: pport})
   246  			continue
   247  		}
   248  
   249  		if len(phost) == 0 {
   250  			// There is no host part, likely the entry is malformed; ignore.
   251  			continue
   252  		}
   253  
   254  		// domain.com or domain.com:80
   255  		// foo.com matches bar.foo.com
   256  		// .domain.com or .domain.com:port
   257  		// *.domain.com or *.domain.com:port
   258  		if strings.HasPrefix(phost, "*.") {
   259  			phost = phost[1:]
   260  		}
   261  		matchHost := false
   262  		if phost[0] != '.' {
   263  			matchHost = true
   264  			phost = "." + phost
   265  		}
   266  		if v, err := idnaASCII(phost); err == nil {
   267  			phost = v
   268  		}
   269  		c.domainMatchers = append(c.domainMatchers, domainMatch{host: phost, port: pport, matchHost: matchHost})
   270  	}
   271  }
   272  
   273  var portMap = map[string]string{
   274  	"http":   "80",
   275  	"https":  "443",
   276  	"socks5": "1080",
   277  }
   278  
   279  // canonicalAddr returns url.Host but always with a ":port" suffix
   280  func canonicalAddr(url *url.URL) string {
   281  	addr := url.Hostname()
   282  	if v, err := idnaASCII(addr); err == nil {
   283  		addr = v
   284  	}
   285  	port := url.Port()
   286  	if port == "" {
   287  		port = portMap[url.Scheme]
   288  	}
   289  	return net.JoinHostPort(addr, port)
   290  }
   291  
   292  // Given a string of the form "host", "host:port", or "[ipv6::address]:port",
   293  // return true if the string includes a port.
   294  func hasPort(s string) bool { return strings.LastIndex(s, ":") > strings.LastIndex(s, "]") }
   295  
   296  func idnaASCII(v string) (string, error) {
   297  	// TODO: Consider removing this check after verifying performance is okay.
   298  	// Right now punycode verification, length checks, context checks, and the
   299  	// permissible character tests are all omitted. It also prevents the ToASCII
   300  	// call from salvaging an invalid IDN, when possible. As a result it may be
   301  	// possible to have two IDNs that appear identical to the user where the
   302  	// ASCII-only version causes an error downstream whereas the non-ASCII
   303  	// version does not.
   304  	// Note that for correct ASCII IDNs ToASCII will only do considerably more
   305  	// work, but it will not cause an allocation.
   306  	if isASCII(v) {
   307  		return v, nil
   308  	}
   309  	return idna.Lookup.ToASCII(v)
   310  }
   311  
   312  func isASCII(s string) bool {
   313  	for i := 0; i < len(s); i++ {
   314  		if s[i] >= utf8.RuneSelf {
   315  			return false
   316  		}
   317  	}
   318  	return true
   319  }
   320  
   321  // matcher represents the matching rule for a given value in the NO_PROXY list
   322  type matcher interface {
   323  	// match returns true if the host and optional port or ip and optional port
   324  	// are allowed
   325  	match(host, port string, ip net.IP) bool
   326  }
   327  
   328  // allMatch matches on all possible inputs
   329  type allMatch struct{}
   330  
   331  func (a allMatch) match(host, port string, ip net.IP) bool {
   332  	return true
   333  }
   334  
   335  type cidrMatch struct {
   336  	cidr *net.IPNet
   337  }
   338  
   339  func (m cidrMatch) match(host, port string, ip net.IP) bool {
   340  	return m.cidr.Contains(ip)
   341  }
   342  
   343  type ipMatch struct {
   344  	ip   net.IP
   345  	port string
   346  }
   347  
   348  func (m ipMatch) match(host, port string, ip net.IP) bool {
   349  	if m.ip.Equal(ip) {
   350  		return m.port == "" || m.port == port
   351  	}
   352  	return false
   353  }
   354  
   355  type domainMatch struct {
   356  	host string
   357  	port string
   358  
   359  	matchHost bool
   360  }
   361  
   362  func (m domainMatch) match(host, port string, ip net.IP) bool {
   363  	if strings.HasSuffix(host, m.host) || (m.matchHost && host == m.host[1:]) {
   364  		return m.port == "" || m.port == port
   365  	}
   366  	return false
   367  }
   368
View as plain text