source: code/trunk/morty.go@ 32

Last change on this file since 32 was 32, checked in by asciimoo, 9 years ago

Merge pull request #30 from dalf/html

[mod] static HTML
File size: 16.5 KB
Line 
1package main
2
3import (
4 "bytes"
5 "crypto/hmac"
6 "crypto/sha256"
7 "encoding/hex"
8 "errors"
9 "flag"
10 "fmt"
11 "io"
12 "log"
13 "net/url"
14 "regexp"
15 "strings"
16 "time"
17
18 "github.com/valyala/fasthttp"
19 "golang.org/x/net/html"
20 "golang.org/x/text/encoding/charmap"
21)
22
23const (
24 STATE_DEFAULT int = 0
25 STATE_IN_STYLE int = 1
26 STATE_IN_NOSCRIPT int = 2
27)
28
29var CLIENT *fasthttp.Client = &fasthttp.Client{
30 MaxResponseBodySize: 10 * 1024 * 1024, // 10M
31}
32
33var CSS_URL_REGEXP *regexp.Regexp = regexp.MustCompile("url\\((['\"]?)[ \\t\\f]*([\u0009\u0021\u0023-\u0026\u0028\u002a-\u007E]+)(['\"]?)\\)?")
34
35var UNSAFE_ELEMENTS [][]byte = [][]byte{
36 []byte("applet"),
37 []byte("canvas"),
38 []byte("embed"),
39 //[]byte("iframe"),
40 []byte("script"),
41}
42
43var SAFE_ATTRIBUTES [][]byte = [][]byte{
44 []byte("abbr"),
45 []byte("accesskey"),
46 []byte("align"),
47 []byte("alt"),
48 []byte("as"),
49 []byte("autocomplete"),
50 []byte("charset"),
51 []byte("checked"),
52 []byte("class"),
53 []byte("content"),
54 []byte("contenteditable"),
55 []byte("contextmenu"),
56 []byte("dir"),
57 []byte("for"),
58 []byte("height"),
59 []byte("hidden"),
60 []byte("id"),
61 []byte("lang"),
62 []byte("media"),
63 []byte("method"),
64 []byte("name"),
65 []byte("nowrap"),
66 []byte("placeholder"),
67 []byte("property"),
68 []byte("rel"),
69 []byte("spellcheck"),
70 []byte("tabindex"),
71 []byte("target"),
72 []byte("title"),
73 []byte("translate"),
74 []byte("type"),
75 []byte("value"),
76 []byte("width"),
77}
78
79var SELF_CLOSING_ELEMENTS [][]byte = [][]byte{
80 []byte("area"),
81 []byte("base"),
82 []byte("br"),
83 []byte("col"),
84 []byte("embed"),
85 []byte("hr"),
86 []byte("img"),
87 []byte("input"),
88 []byte("keygen"),
89 []byte("link"),
90 []byte("meta"),
91 []byte("param"),
92 []byte("source"),
93 []byte("track"),
94 []byte("wbr"),
95}
96
97type Proxy struct {
98 Key []byte
99 RequestTimeout time.Duration
100}
101
102type RequestConfig struct {
103 Key []byte
104 BaseURL *url.URL
105}
106
107var HTML_FORM_EXTENSION string = `<input type="hidden" name="mortyurl" value="%s" /><input type="hidden" name="mortyhash" value="%s" />`
108
109var HTML_BODY_EXTENSION string = `
110<div id="mortyheader">
111 <input type="checkbox" id="mortytoggle" autocomplete="off" />
112 <div><p>This is a proxified and sanitized view of the page,<br />visit <a href="%s" rel="noreferrer">original site</a>.</p><p><label for="mortytoggle">hide</label></p></div>
113</div>
114<style>
115#mortyheader { position: fixed; padding: 12px 12px 12px 0; margin: 0; box-sizing: content-box; top: 15%%; left: 0; max-width: 140px; color: #444; overflow: hidden; z-index: 110000; font-size: 12px; line-height: normal; }
116#mortyheader a { color: #3498db; font-weight: bold; }
117#mortyheader p { padding: 0 0 0.7em 0; margin: 0; }
118#mortyheader > div { padding: 8px; font-size: 12px !important; font-family: sans !important; border-width: 4px 4px 4px 0; border-style: solid; border-color: #1abc9c; background: #FFF; line-height: 1em; }
119#mortyheader label { text-align: right; cursor: pointer; display: block; color: #444; padding: 0; margin: 0; }
120input[type=checkbox]#mortytoggle { display: none; }
121input[type=checkbox]#mortytoggle:checked ~ div { display: none; }
122</style>
123`
124
125func (p *Proxy) RequestHandler(ctx *fasthttp.RequestCtx) {
126
127 if appRequestHandler(ctx) {
128 return
129 }
130
131 requestHash := popRequestParam(ctx, []byte("mortyhash"))
132
133 requestURI := popRequestParam(ctx, []byte("mortyurl"))
134
135 if requestURI == nil {
136 p.serveMainPage(ctx, nil)
137 return
138 }
139
140 if p.Key != nil {
141 if !verifyRequestURI(requestURI, requestHash, p.Key) {
142 p.serveMainPage(ctx, errors.New(`invalid "mortyhash" parameter`))
143 return
144 }
145 }
146
147 parsedURI, err := url.Parse(string(requestURI))
148
149 if strings.HasSuffix(parsedURI.Host, ".onion") {
150 p.serveMainPage(ctx, errors.New("Tor urls are not supported yet"))
151 return
152 }
153
154 if err != nil {
155 p.serveMainPage(ctx, err)
156 return
157 }
158
159 req := fasthttp.AcquireRequest()
160 defer fasthttp.ReleaseRequest(req)
161 req.SetConnectionClose()
162
163 reqQuery := parsedURI.Query()
164 ctx.QueryArgs().VisitAll(func(key, value []byte) {
165 reqQuery.Add(string(key), string(value))
166 })
167
168 parsedURI.RawQuery = reqQuery.Encode()
169
170 uriStr := parsedURI.String()
171
172 log.Println("getting", uriStr)
173
174 req.SetRequestURI(uriStr)
175 req.Header.SetUserAgentBytes([]byte("Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/53.0.2785.143 Safari/537.36"))
176
177 resp := fasthttp.AcquireResponse()
178 defer fasthttp.ReleaseResponse(resp)
179
180 req.Header.SetMethodBytes(ctx.Method())
181 if ctx.IsPost() || ctx.IsPut() {
182 req.SetBody(ctx.PostBody())
183 }
184
185 err = CLIENT.DoTimeout(req, resp, p.RequestTimeout)
186
187 if err != nil {
188 p.serveMainPage(ctx, err)
189 return
190 }
191
192 if resp.StatusCode() != 200 {
193 switch resp.StatusCode() {
194 case 301, 302, 303, 307, 308:
195 loc := resp.Header.Peek("Location")
196 if loc != nil {
197 rc := &RequestConfig{Key: p.Key, BaseURL: parsedURI}
198 url, err := rc.ProxifyURI(string(loc))
199 if err == nil {
200 ctx.SetStatusCode(resp.StatusCode())
201 ctx.Response.Header.Add("Location", url)
202 log.Println("redirect to", string(loc))
203 return
204 }
205 }
206 }
207 log.Println("invalid request:", resp.StatusCode())
208 return
209 }
210
211 contentType := resp.Header.Peek("Content-Type")
212
213 if contentType == nil {
214 p.serveMainPage(ctx, errors.New("invalid content type"))
215 return
216 }
217
218 if bytes.Contains(bytes.ToLower(contentType), []byte("javascript")) {
219 p.serveMainPage(ctx, errors.New("forbidden content type"))
220 return
221 }
222
223 contentInfo := bytes.SplitN(contentType, []byte(";"), 2)
224
225 var responseBody []byte
226
227 if len(contentInfo) == 2 && bytes.Contains(contentInfo[1], []byte("ISO-8859-2")) && bytes.Contains(contentInfo[0], []byte("text")) {
228 var err error
229 responseBody, err = charmap.ISO8859_2.NewDecoder().Bytes(resp.Body())
230 if err != nil {
231 p.serveMainPage(ctx, err)
232 return
233 }
234 } else {
235 responseBody = resp.Body()
236 }
237
238 ctx.SetContentType(fmt.Sprintf("%s; charset=UTF-8", contentInfo[0]))
239
240 switch {
241 case bytes.Contains(contentType, []byte("css")):
242 sanitizeCSS(&RequestConfig{Key: p.Key, BaseURL: parsedURI}, ctx, responseBody)
243 case bytes.Contains(contentType, []byte("html")):
244 sanitizeHTML(&RequestConfig{Key: p.Key, BaseURL: parsedURI}, ctx, responseBody)
245 default:
246 ctx.Write(responseBody)
247 }
248}
249
250func appRequestHandler(ctx *fasthttp.RequestCtx) bool {
251 // serve robots.txt
252 if bytes.Equal(ctx.Path(), []byte("/robots.txt")) {
253 ctx.SetContentType("text/plain")
254 ctx.Write([]byte("User-Agent: *\nDisallow: /\n"))
255 return true
256 }
257
258 return false
259}
260
261func popRequestParam(ctx *fasthttp.RequestCtx, paramName []byte) []byte {
262 param := ctx.QueryArgs().PeekBytes(paramName)
263
264 if param == nil {
265 param = ctx.PostArgs().PeekBytes(paramName)
266 if param != nil {
267 ctx.PostArgs().DelBytes(paramName)
268 }
269 } else {
270 ctx.QueryArgs().DelBytes(paramName)
271 }
272
273 return param
274}
275
276func sanitizeCSS(rc *RequestConfig, out io.Writer, css []byte) {
277 // TODO
278
279 urlSlices := CSS_URL_REGEXP.FindAllSubmatchIndex(css, -1)
280
281 if urlSlices == nil {
282 out.Write(css)
283 return
284 }
285
286 startIndex := 0
287
288 for _, s := range urlSlices {
289 urlStart := s[4]
290 urlEnd := s[5]
291
292 if uri, err := rc.ProxifyURI(string(css[urlStart:urlEnd])); err == nil {
293 out.Write(css[startIndex:urlStart])
294 out.Write([]byte(uri))
295 startIndex = urlEnd
296 } else {
297 log.Println("cannot proxify css uri:", css[urlStart:urlEnd])
298 }
299 }
300 if startIndex < len(css) {
301 out.Write(css[startIndex:len(css)])
302 }
303}
304
305func sanitizeHTML(rc *RequestConfig, out io.Writer, htmlDoc []byte) {
306 r := bytes.NewReader(htmlDoc)
307 decoder := html.NewTokenizer(r)
308 decoder.AllowCDATA(true)
309
310 unsafeElements := make([][]byte, 0, 8)
311 state := STATE_DEFAULT
312
313 for {
314 token := decoder.Next()
315 if token == html.ErrorToken {
316 err := decoder.Err()
317 if err != io.EOF {
318 log.Println("failed to parse HTML:")
319 }
320 break
321 }
322
323 if len(unsafeElements) == 0 {
324
325 switch token {
326 case html.StartTagToken, html.SelfClosingTagToken:
327 tag, hasAttrs := decoder.TagName()
328 safe := !inArray(tag, UNSAFE_ELEMENTS)
329 if !safe {
330 if !inArray(tag, SELF_CLOSING_ELEMENTS) {
331 var unsafeTag []byte = make([]byte, len(tag))
332 copy(unsafeTag, tag)
333 unsafeElements = append(unsafeElements, unsafeTag)
334 }
335 break
336 }
337 if bytes.Equal(tag, []byte("noscript")) {
338 state = STATE_IN_NOSCRIPT
339 break
340 }
341 var attrs [][][]byte
342 if hasAttrs {
343 for {
344 attrName, attrValue, moreAttr := decoder.TagAttr()
345 attrs = append(attrs, [][]byte{
346 attrName,
347 attrValue,
348 []byte(html.EscapeString(string(attrValue))),
349 })
350 if !moreAttr {
351 break
352 }
353 }
354 }
355 if bytes.Equal(tag, []byte("link")) {
356 sanitizeLinkTag(rc, out, attrs)
357 break
358 }
359
360 fmt.Fprintf(out, "<%s", tag)
361
362 if hasAttrs {
363 if bytes.Equal(tag, []byte("meta")) {
364 sanitizeMetaAttrs(rc, out, attrs)
365 } else {
366 sanitizeAttrs(rc, out, attrs)
367 }
368 }
369
370 if token == html.SelfClosingTagToken {
371 fmt.Fprintf(out, " />")
372 } else {
373 fmt.Fprintf(out, ">")
374 if bytes.Equal(tag, []byte("style")) {
375 state = STATE_IN_STYLE
376 }
377 }
378
379 if bytes.Equal(tag, []byte("form")) {
380 var formURL *url.URL
381 for _, attr := range attrs {
382 if bytes.Equal(attr[0], []byte("action")) {
383 formURL, _ = url.Parse(string(attr[1]))
384 formURL = mergeURIs(rc.BaseURL, formURL)
385 break
386 }
387 }
388 if formURL == nil {
389 formURL = rc.BaseURL
390 }
391 urlStr := formURL.String()
392 var key string
393 if rc.Key != nil {
394 key = hash(urlStr, rc.Key)
395 }
396 fmt.Fprintf(out, HTML_FORM_EXTENSION, urlStr, key)
397
398 }
399
400 case html.EndTagToken:
401 tag, _ := decoder.TagName()
402 writeEndTag := true
403 switch string(tag) {
404 case "body":
405 fmt.Fprintf(out, HTML_BODY_EXTENSION, rc.BaseURL.String())
406 case "style":
407 state = STATE_DEFAULT
408 case "noscript":
409 state = STATE_DEFAULT
410 writeEndTag = false
411 }
412 // skip noscript tags - only the tag, not the content, because javascript is sanitized
413 if writeEndTag {
414 fmt.Fprintf(out, "</%s>", tag)
415 }
416
417 case html.TextToken:
418 switch state {
419 case STATE_DEFAULT:
420 fmt.Fprintf(out, "%s", decoder.Raw())
421 case STATE_IN_STYLE:
422 sanitizeCSS(rc, out, decoder.Raw())
423 case STATE_IN_NOSCRIPT:
424 sanitizeHTML(rc, out, decoder.Raw())
425 }
426
427 case html.DoctypeToken, html.CommentToken:
428 out.Write(decoder.Raw())
429 }
430 } else {
431 switch token {
432 case html.StartTagToken:
433 tag, _ := decoder.TagName()
434 if inArray(tag, UNSAFE_ELEMENTS) {
435 unsafeElements = append(unsafeElements, tag)
436 }
437
438 case html.EndTagToken:
439 tag, _ := decoder.TagName()
440 if bytes.Equal(unsafeElements[len(unsafeElements)-1], tag) {
441 unsafeElements = unsafeElements[:len(unsafeElements)-1]
442 }
443 }
444 }
445 }
446}
447
448func sanitizeLinkTag(rc *RequestConfig, out io.Writer, attrs [][][]byte) {
449 exclude := false
450 for _, attr := range attrs {
451 attrName := attr[0]
452 attrValue := attr[1]
453 if bytes.Equal(attrName, []byte("rel")) {
454 if bytes.Equal(attrValue, []byte("dns-prefetch")) {
455 exclude = true
456 break
457 }
458 }
459 if bytes.Equal(attrName, []byte("as")) {
460 if bytes.Equal(attrValue, []byte("script")) {
461 exclude = true
462 break
463 }
464 }
465 }
466
467 if !exclude {
468 out.Write([]byte("<link"))
469 for _, attr := range attrs {
470 sanitizeAttr(rc, out, attr[0], attr[1], attr[2])
471 }
472 out.Write([]byte(">"))
473 }
474}
475
476func sanitizeMetaAttrs(rc *RequestConfig, out io.Writer, attrs [][][]byte) {
477 var http_equiv []byte
478 var content []byte
479
480 for _, attr := range attrs {
481 attrName := attr[0]
482 attrValue := attr[1]
483 if bytes.Equal(attrName, []byte("http-equiv")) {
484 http_equiv = bytes.ToLower(attrValue)
485 }
486 if bytes.Equal(attrName, []byte("content")) {
487 content = attrValue
488 }
489 }
490
491 urlIndex := bytes.Index(bytes.ToLower(content), []byte("url="))
492 if bytes.Equal(http_equiv, []byte("refresh")) && urlIndex != -1 {
493 contentUrl := content[urlIndex+4:]
494 // special case of <meta http-equiv="refresh" content="0; url='example.com/url.with.quote.outside'">
495 if len(contentUrl)>=2 && (contentUrl[0] == byte('\'') || contentUrl[0] == byte('"')) {
496 if contentUrl[0] == contentUrl[len(contentUrl)-1] {
497 contentUrl=contentUrl[1:len(contentUrl)-1]
498 }
499 }
500 // output proxify result
501 if uri, err := rc.ProxifyURI(string(contentUrl)); err == nil {
502 fmt.Fprintf(out, ` http-equiv="refresh" content="%surl=%s"`, content[:urlIndex], uri)
503 }
504 } else {
505 sanitizeAttrs(rc, out, attrs)
506 }
507
508}
509
510func sanitizeAttrs(rc *RequestConfig, out io.Writer, attrs [][][]byte) {
511 for _, attr := range attrs {
512 sanitizeAttr(rc, out, attr[0], attr[1], attr[2])
513 }
514}
515
516func sanitizeAttr(rc *RequestConfig, out io.Writer, attrName, attrValue, escapedAttrValue []byte) {
517 if inArray(attrName, SAFE_ATTRIBUTES) {
518 fmt.Fprintf(out, " %s=\"%s\"", attrName, escapedAttrValue)
519 return
520 }
521 switch string(attrName) {
522 case "src", "href", "action":
523 if uri, err := rc.ProxifyURI(string(attrValue)); err == nil {
524 fmt.Fprintf(out, " %s=\"%s\"", attrName, uri)
525 } else {
526 log.Println("cannot proxify uri:", attrValue)
527 }
528 case "style":
529 cssAttr := bytes.NewBuffer(nil)
530 sanitizeCSS(rc, cssAttr, attrValue)
531 fmt.Fprintf(out, " %s=\"%s\"", attrName, html.EscapeString(string(cssAttr.Bytes())))
532 }
533}
534
535func mergeURIs(u1, u2 *url.URL) *url.URL {
536 return u1.ResolveReference(u2)
537}
538
539func (rc *RequestConfig) ProxifyURI(uri string) (string, error) {
540 // remove javascript protocol
541 if strings.HasPrefix(uri, "javascript:") {
542 return "", nil
543 }
544 // TODO check malicious data: - e.g. data:script
545 if strings.HasPrefix(uri, "data:") {
546 return uri, nil
547 }
548
549 if len(uri) > 0 && uri[0] == '#' {
550 return uri, nil
551 }
552
553 u, err := url.Parse(uri)
554 if err != nil {
555 return "", err
556 }
557 u = mergeURIs(rc.BaseURL, u)
558
559 uri = u.String()
560
561 if rc.Key == nil {
562 return fmt.Sprintf("./?mortyurl=%s", url.QueryEscape(uri)), nil
563 }
564 return fmt.Sprintf("./?mortyhash=%s&mortyurl=%s", hash(uri, rc.Key), url.QueryEscape(uri)), nil
565}
566
567func inArray(b []byte, a [][]byte) bool {
568 for _, b2 := range a {
569 if bytes.Equal(b, b2) {
570 return true
571 }
572 }
573 return false
574}
575
576func hash(msg string, key []byte) string {
577 mac := hmac.New(sha256.New, key)
578 mac.Write([]byte(msg))
579 return hex.EncodeToString(mac.Sum(nil))
580}
581
582func verifyRequestURI(uri, hashMsg, key []byte) bool {
583 h := make([]byte, hex.DecodedLen(len(hashMsg)))
584 _, err := hex.Decode(h, hashMsg)
585 if err != nil {
586 log.Println("hmac error:", err)
587 return false
588 }
589 mac := hmac.New(sha256.New, key)
590 mac.Write(uri)
591 return hmac.Equal(h, mac.Sum(nil))
592}
593
594func (p *Proxy) serveMainPage(ctx *fasthttp.RequestCtx, err error) {
595 ctx.SetContentType("text/html")
596 ctx.Write([]byte(`<!doctype html>
597<head>
598<title>MortyProxy</title>
599<meta name="viewport" content="width=device-width, initial-scale=1 , maximum-scale=1.0, user-scalable=1" />
600<style>
601html { height: 100%; }
602body { min-height : 100%; display: flex; flex-direction:column; font-family: 'Garamond', 'Georgia', serif; text-align: center; color: #444; background: #FAFAFA; margin: 0; padding: 0; font-size: 1.1em; }
603input { border: 1px solid #888; padding: 0.3em; color: #444; background: #FFF; font-size: 1.1em; }
604input[placeholder] { width:80%; }
605a { text-decoration: none; #2980b9; }
606h1, h2 { font-weight: 200; margin-bottom: 2rem; }
607h1 { font-size: 3em; }
608.container { flex:1; min-height: 100%; margin-bottom: 1em; }
609.footer { margin: 1em; }
610.footer p { font-size: 0.8em; }
611</style>
612</head>
613<body>
614 <div class="container">
615 <h1>MortyProxy</h1>
616`))
617 if err != nil {
618 ctx.SetStatusCode(404)
619 log.Println("error:", err)
620 ctx.Write([]byte("<h2>Error: "))
621 ctx.Write([]byte(html.EscapeString(err.Error())))
622 ctx.Write([]byte("</h2>"))
623 } else {
624 ctx.SetStatusCode(200)
625 }
626 if p.Key == nil {
627 ctx.Write([]byte(`
628 <form action="post">
629 Visit url: <input placeholder="https://url.." name="mortyurl" autofocus />
630 <input type="submit" value="go" />
631 </form>`))
632 } else {
633 ctx.Write([]byte(`<h3>Warning! This instance does not support direct URL opening.</h3>`))
634 }
635 ctx.Write([]byte(`
636 </div>
637 <div class="footer">
638 <p>Morty rewrites web pages to exclude malicious HTML tags and CSS/HTML attributes. It also replaces external resource references to prevent third-party information leaks.<br />
639 <a href="https://github.com/asciimoo/morty">view on github</a>
640 </p>
641 </div>
642</body>
643</html>`))
644}
645
646func main() {
647
648 listen := flag.String("listen", "127.0.0.1:3000", "Listen address")
649 key := flag.String("key", "", "HMAC url validation key (hexadecimal encoded) - leave blank to disable")
650 ipv6 := flag.Bool("ipv6", false, "Allow IPv6 HTTP requests")
651 requestTimeout := flag.Uint("timeout", 2, "Request timeout")
652 flag.Parse()
653
654 if *ipv6 {
655 CLIENT.Dial = fasthttp.DialDualStack
656 }
657
658 p := &Proxy{RequestTimeout: time.Duration(*requestTimeout) * time.Second}
659
660 if *key != "" {
661 p.Key = []byte(*key)
662 }
663
664 log.Println("listening on", *listen)
665
666 if err := fasthttp.ListenAndServe(*listen, p.RequestHandler); err != nil {
667 log.Fatal("Error in ListenAndServe:", err)
668 }
669}
Note: See TracBrowser for help on using the repository browser.