Run Format

Source file src/pkg/net/http/fs.go

     1	// Copyright 2009 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	// HTTP file system request handler
     6	
     7	package http
     8	
     9	import (
    10		"errors"
    11		"fmt"
    12		"io"
    13		"mime"
    14		"mime/multipart"
    15		"net/textproto"
    16		"os"
    17		"path"
    18		"path/filepath"
    19		"strconv"
    20		"strings"
    21		"time"
    22	)
    23	
    24	// A Dir implements http.FileSystem using the native file
    25	// system restricted to a specific directory tree.
    26	//
    27	// An empty Dir is treated as ".".
    28	type Dir string
    29	
    30	func (d Dir) Open(name string) (File, error) {
    31		if filepath.Separator != '/' && strings.IndexRune(name, filepath.Separator) >= 0 ||
    32			strings.Contains(name, "\x00") {
    33			return nil, errors.New("http: invalid character in file path")
    34		}
    35		dir := string(d)
    36		if dir == "" {
    37			dir = "."
    38		}
    39		f, err := os.Open(filepath.Join(dir, filepath.FromSlash(path.Clean("/"+name))))
    40		if err != nil {
    41			return nil, err
    42		}
    43		return f, nil
    44	}
    45	
    46	// A FileSystem implements access to a collection of named files.
    47	// The elements in a file path are separated by slash ('/', U+002F)
    48	// characters, regardless of host operating system convention.
    49	type FileSystem interface {
    50		Open(name string) (File, error)
    51	}
    52	
    53	// A File is returned by a FileSystem's Open method and can be
    54	// served by the FileServer implementation.
    55	type File interface {
    56		Close() error
    57		Stat() (os.FileInfo, error)
    58		Readdir(count int) ([]os.FileInfo, error)
    59		Read([]byte) (int, error)
    60		Seek(offset int64, whence int) (int64, error)
    61	}
    62	
    63	func dirList(w ResponseWriter, f File) {
    64		w.Header().Set("Content-Type", "text/html; charset=utf-8")
    65		fmt.Fprintf(w, "<pre>\n")
    66		for {
    67			dirs, err := f.Readdir(100)
    68			if err != nil || len(dirs) == 0 {
    69				break
    70			}
    71			for _, d := range dirs {
    72				name := d.Name()
    73				if d.IsDir() {
    74					name += "/"
    75				}
    76				// TODO htmlescape
    77				fmt.Fprintf(w, "<a href=\"%s\">%s</a>\n", name, name)
    78			}
    79		}
    80		fmt.Fprintf(w, "</pre>\n")
    81	}
    82	
    83	// ServeContent replies to the request using the content in the
    84	// provided ReadSeeker.  The main benefit of ServeContent over io.Copy
    85	// is that it handles Range requests properly, sets the MIME type, and
    86	// handles If-Modified-Since requests.
    87	//
    88	// If the response's Content-Type header is not set, ServeContent
    89	// first tries to deduce the type from name's file extension and,
    90	// if that fails, falls back to reading the first block of the content
    91	// and passing it to DetectContentType.
    92	// The name is otherwise unused; in particular it can be empty and is
    93	// never sent in the response.
    94	//
    95	// If modtime is not the zero time, ServeContent includes it in a
    96	// Last-Modified header in the response.  If the request includes an
    97	// If-Modified-Since header, ServeContent uses modtime to decide
    98	// whether the content needs to be sent at all.
    99	//
   100	// The content's Seek method must work: ServeContent uses
   101	// a seek to the end of the content to determine its size.
   102	//
   103	// If the caller has set w's ETag header, ServeContent uses it to
   104	// handle requests using If-Range and If-None-Match.
   105	//
   106	// Note that *os.File implements the io.ReadSeeker interface.
   107	func ServeContent(w ResponseWriter, req *Request, name string, modtime time.Time, content io.ReadSeeker) {
   108		sizeFunc := func() (int64, error) {
   109			size, err := content.Seek(0, os.SEEK_END)
   110			if err != nil {
   111				return 0, errSeeker
   112			}
   113			_, err = content.Seek(0, os.SEEK_SET)
   114			if err != nil {
   115				return 0, errSeeker
   116			}
   117			return size, nil
   118		}
   119		serveContent(w, req, name, modtime, sizeFunc, content)
   120	}
   121	
   122	// errSeeker is returned by ServeContent's sizeFunc when the content
   123	// doesn't seek properly. The underlying Seeker's error text isn't
   124	// included in the sizeFunc reply so it's not sent over HTTP to end
   125	// users.
   126	var errSeeker = errors.New("seeker can't seek")
   127	
   128	// if name is empty, filename is unknown. (used for mime type, before sniffing)
   129	// if modtime.IsZero(), modtime is unknown.
   130	// content must be seeked to the beginning of the file.
   131	// The sizeFunc is called at most once. Its error, if any, is sent in the HTTP response.
   132	func serveContent(w ResponseWriter, r *Request, name string, modtime time.Time, sizeFunc func() (int64, error), content io.ReadSeeker) {
   133		if checkLastModified(w, r, modtime) {
   134			return
   135		}
   136		rangeReq, done := checkETag(w, r)
   137		if done {
   138			return
   139		}
   140	
   141		code := StatusOK
   142	
   143		// If Content-Type isn't set, use the file's extension to find it, but
   144		// if the Content-Type is unset explicitly, do not sniff the type.
   145		ctypes, haveType := w.Header()["Content-Type"]
   146		var ctype string
   147		if !haveType {
   148			ctype = mime.TypeByExtension(filepath.Ext(name))
   149			if ctype == "" {
   150				// read a chunk to decide between utf-8 text and binary
   151				var buf [sniffLen]byte
   152				n, _ := io.ReadFull(content, buf[:])
   153				ctype = DetectContentType(buf[:n])
   154				_, err := content.Seek(0, os.SEEK_SET) // rewind to output whole file
   155				if err != nil {
   156					Error(w, "seeker can't seek", StatusInternalServerError)
   157					return
   158				}
   159			}
   160			w.Header().Set("Content-Type", ctype)
   161		} else if len(ctypes) > 0 {
   162			ctype = ctypes[0]
   163		}
   164	
   165		size, err := sizeFunc()
   166		if err != nil {
   167			Error(w, err.Error(), StatusInternalServerError)
   168			return
   169		}
   170	
   171		// handle Content-Range header.
   172		sendSize := size
   173		var sendContent io.Reader = content
   174		if size >= 0 {
   175			ranges, err := parseRange(rangeReq, size)
   176			if err != nil {
   177				Error(w, err.Error(), StatusRequestedRangeNotSatisfiable)
   178				return
   179			}
   180			if sumRangesSize(ranges) > size {
   181				// The total number of bytes in all the ranges
   182				// is larger than the size of the file by
   183				// itself, so this is probably an attack, or a
   184				// dumb client.  Ignore the range request.
   185				ranges = nil
   186			}
   187			switch {
   188			case len(ranges) == 1:
   189				// RFC 2616, Section 14.16:
   190				// "When an HTTP message includes the content of a single
   191				// range (for example, a response to a request for a
   192				// single range, or to a request for a set of ranges
   193				// that overlap without any holes), this content is
   194				// transmitted with a Content-Range header, and a
   195				// Content-Length header showing the number of bytes
   196				// actually transferred.
   197				// ...
   198				// A response to a request for a single range MUST NOT
   199				// be sent using the multipart/byteranges media type."
   200				ra := ranges[0]
   201				if _, err := content.Seek(ra.start, os.SEEK_SET); err != nil {
   202					Error(w, err.Error(), StatusRequestedRangeNotSatisfiable)
   203					return
   204				}
   205				sendSize = ra.length
   206				code = StatusPartialContent
   207				w.Header().Set("Content-Range", ra.contentRange(size))
   208			case len(ranges) > 1:
   209				for _, ra := range ranges {
   210					if ra.start > size {
   211						Error(w, err.Error(), StatusRequestedRangeNotSatisfiable)
   212						return
   213					}
   214				}
   215				sendSize = rangesMIMESize(ranges, ctype, size)
   216				code = StatusPartialContent
   217	
   218				pr, pw := io.Pipe()
   219				mw := multipart.NewWriter(pw)
   220				w.Header().Set("Content-Type", "multipart/byteranges; boundary="+mw.Boundary())
   221				sendContent = pr
   222				defer pr.Close() // cause writing goroutine to fail and exit if CopyN doesn't finish.
   223				go func() {
   224					for _, ra := range ranges {
   225						part, err := mw.CreatePart(ra.mimeHeader(ctype, size))
   226						if err != nil {
   227							pw.CloseWithError(err)
   228							return
   229						}
   230						if _, err := content.Seek(ra.start, os.SEEK_SET); err != nil {
   231							pw.CloseWithError(err)
   232							return
   233						}
   234						if _, err := io.CopyN(part, content, ra.length); err != nil {
   235							pw.CloseWithError(err)
   236							return
   237						}
   238					}
   239					mw.Close()
   240					pw.Close()
   241				}()
   242			}
   243	
   244			w.Header().Set("Accept-Ranges", "bytes")
   245			if w.Header().Get("Content-Encoding") == "" {
   246				w.Header().Set("Content-Length", strconv.FormatInt(sendSize, 10))
   247			}
   248		}
   249	
   250		w.WriteHeader(code)
   251	
   252		if r.Method != "HEAD" {
   253			io.CopyN(w, sendContent, sendSize)
   254		}
   255	}
   256	
   257	// modtime is the modification time of the resource to be served, or IsZero().
   258	// return value is whether this request is now complete.
   259	func checkLastModified(w ResponseWriter, r *Request, modtime time.Time) bool {
   260		if modtime.IsZero() {
   261			return false
   262		}
   263	
   264		// The Date-Modified header truncates sub-second precision, so
   265		// use mtime < t+1s instead of mtime <= t to check for unmodified.
   266		if t, err := time.Parse(TimeFormat, r.Header.Get("If-Modified-Since")); err == nil && modtime.Before(t.Add(1*time.Second)) {
   267			h := w.Header()
   268			delete(h, "Content-Type")
   269			delete(h, "Content-Length")
   270			w.WriteHeader(StatusNotModified)
   271			return true
   272		}
   273		w.Header().Set("Last-Modified", modtime.UTC().Format(TimeFormat))
   274		return false
   275	}
   276	
   277	// checkETag implements If-None-Match and If-Range checks.
   278	// The ETag must have been previously set in the ResponseWriter's headers.
   279	//
   280	// The return value is the effective request "Range" header to use and
   281	// whether this request is now considered done.
   282	func checkETag(w ResponseWriter, r *Request) (rangeReq string, done bool) {
   283		etag := w.Header().get("Etag")
   284		rangeReq = r.Header.get("Range")
   285	
   286		// Invalidate the range request if the entity doesn't match the one
   287		// the client was expecting.
   288		// "If-Range: version" means "ignore the Range: header unless version matches the
   289		// current file."
   290		// We only support ETag versions.
   291		// The caller must have set the ETag on the response already.
   292		if ir := r.Header.get("If-Range"); ir != "" && ir != etag {
   293			// TODO(bradfitz): handle If-Range requests with Last-Modified
   294			// times instead of ETags? I'd rather not, at least for
   295			// now. That seems like a bug/compromise in the RFC 2616, and
   296			// I've never heard of anybody caring about that (yet).
   297			rangeReq = ""
   298		}
   299	
   300		if inm := r.Header.get("If-None-Match"); inm != "" {
   301			// Must know ETag.
   302			if etag == "" {
   303				return rangeReq, false
   304			}
   305	
   306			// TODO(bradfitz): non-GET/HEAD requests require more work:
   307			// sending a different status code on matches, and
   308			// also can't use weak cache validators (those with a "W/
   309			// prefix).  But most users of ServeContent will be using
   310			// it on GET or HEAD, so only support those for now.
   311			if r.Method != "GET" && r.Method != "HEAD" {
   312				return rangeReq, false
   313			}
   314	
   315			// TODO(bradfitz): deal with comma-separated or multiple-valued
   316			// list of If-None-match values.  For now just handle the common
   317			// case of a single item.
   318			if inm == etag || inm == "*" {
   319				h := w.Header()
   320				delete(h, "Content-Type")
   321				delete(h, "Content-Length")
   322				w.WriteHeader(StatusNotModified)
   323				return "", true
   324			}
   325		}
   326		return rangeReq, false
   327	}
   328	
   329	// name is '/'-separated, not filepath.Separator.
   330	func serveFile(w ResponseWriter, r *Request, fs FileSystem, name string, redirect bool) {
   331		const indexPage = "/index.html"
   332	
   333		// redirect .../index.html to .../
   334		// can't use Redirect() because that would make the path absolute,
   335		// which would be a problem running under StripPrefix
   336		if strings.HasSuffix(r.URL.Path, indexPage) {
   337			localRedirect(w, r, "./")
   338			return
   339		}
   340	
   341		f, err := fs.Open(name)
   342		if err != nil {
   343			// TODO expose actual error?
   344			NotFound(w, r)
   345			return
   346		}
   347		defer f.Close()
   348	
   349		d, err1 := f.Stat()
   350		if err1 != nil {
   351			// TODO expose actual error?
   352			NotFound(w, r)
   353			return
   354		}
   355	
   356		if redirect {
   357			// redirect to canonical path: / at end of directory url
   358			// r.URL.Path always begins with /
   359			url := r.URL.Path
   360			if d.IsDir() {
   361				if url[len(url)-1] != '/' {
   362					localRedirect(w, r, path.Base(url)+"/")
   363					return
   364				}
   365			} else {
   366				if url[len(url)-1] == '/' {
   367					localRedirect(w, r, "../"+path.Base(url))
   368					return
   369				}
   370			}
   371		}
   372	
   373		// use contents of index.html for directory, if present
   374		if d.IsDir() {
   375			index := name + indexPage
   376			ff, err := fs.Open(index)
   377			if err == nil {
   378				defer ff.Close()
   379				dd, err := ff.Stat()
   380				if err == nil {
   381					name = index
   382					d = dd
   383					f = ff
   384				}
   385			}
   386		}
   387	
   388		// Still a directory? (we didn't find an index.html file)
   389		if d.IsDir() {
   390			if checkLastModified(w, r, d.ModTime()) {
   391				return
   392			}
   393			dirList(w, f)
   394			return
   395		}
   396	
   397		// serverContent will check modification time
   398		sizeFunc := func() (int64, error) { return d.Size(), nil }
   399		serveContent(w, r, d.Name(), d.ModTime(), sizeFunc, f)
   400	}
   401	
   402	// localRedirect gives a Moved Permanently response.
   403	// It does not convert relative paths to absolute paths like Redirect does.
   404	func localRedirect(w ResponseWriter, r *Request, newPath string) {
   405		if q := r.URL.RawQuery; q != "" {
   406			newPath += "?" + q
   407		}
   408		w.Header().Set("Location", newPath)
   409		w.WriteHeader(StatusMovedPermanently)
   410	}
   411	
   412	// ServeFile replies to the request with the contents of the named file or directory.
   413	func ServeFile(w ResponseWriter, r *Request, name string) {
   414		dir, file := filepath.Split(name)
   415		serveFile(w, r, Dir(dir), file, false)
   416	}
   417	
   418	type fileHandler struct {
   419		root FileSystem
   420	}
   421	
   422	// FileServer returns a handler that serves HTTP requests
   423	// with the contents of the file system rooted at root.
   424	//
   425	// To use the operating system's file system implementation,
   426	// use http.Dir:
   427	//
   428	//     http.Handle("/", http.FileServer(http.Dir("/tmp")))
   429	func FileServer(root FileSystem) Handler {
   430		return &fileHandler{root}
   431	}
   432	
   433	func (f *fileHandler) ServeHTTP(w ResponseWriter, r *Request) {
   434		upath := r.URL.Path
   435		if !strings.HasPrefix(upath, "/") {
   436			upath = "/" + upath
   437			r.URL.Path = upath
   438		}
   439		serveFile(w, r, f.root, path.Clean(upath), true)
   440	}
   441	
   442	// httpRange specifies the byte range to be sent to the client.
   443	type httpRange struct {
   444		start, length int64
   445	}
   446	
   447	func (r httpRange) contentRange(size int64) string {
   448		return fmt.Sprintf("bytes %d-%d/%d", r.start, r.start+r.length-1, size)
   449	}
   450	
   451	func (r httpRange) mimeHeader(contentType string, size int64) textproto.MIMEHeader {
   452		return textproto.MIMEHeader{
   453			"Content-Range": {r.contentRange(size)},
   454			"Content-Type":  {contentType},
   455		}
   456	}
   457	
   458	// parseRange parses a Range header string as per RFC 2616.
   459	func parseRange(s string, size int64) ([]httpRange, error) {
   460		if s == "" {
   461			return nil, nil // header not present
   462		}
   463		const b = "bytes="
   464		if !strings.HasPrefix(s, b) {
   465			return nil, errors.New("invalid range")
   466		}
   467		var ranges []httpRange
   468		for _, ra := range strings.Split(s[len(b):], ",") {
   469			ra = strings.TrimSpace(ra)
   470			if ra == "" {
   471				continue
   472			}
   473			i := strings.Index(ra, "-")
   474			if i < 0 {
   475				return nil, errors.New("invalid range")
   476			}
   477			start, end := strings.TrimSpace(ra[:i]), strings.TrimSpace(ra[i+1:])
   478			var r httpRange
   479			if start == "" {
   480				// If no start is specified, end specifies the
   481				// range start relative to the end of the file.
   482				i, err := strconv.ParseInt(end, 10, 64)
   483				if err != nil {
   484					return nil, errors.New("invalid range")
   485				}
   486				if i > size {
   487					i = size
   488				}
   489				r.start = size - i
   490				r.length = size - r.start
   491			} else {
   492				i, err := strconv.ParseInt(start, 10, 64)
   493				if err != nil || i > size || i < 0 {
   494					return nil, errors.New("invalid range")
   495				}
   496				r.start = i
   497				if end == "" {
   498					// If no end is specified, range extends to end of the file.
   499					r.length = size - r.start
   500				} else {
   501					i, err := strconv.ParseInt(end, 10, 64)
   502					if err != nil || r.start > i {
   503						return nil, errors.New("invalid range")
   504					}
   505					if i >= size {
   506						i = size - 1
   507					}
   508					r.length = i - r.start + 1
   509				}
   510			}
   511			ranges = append(ranges, r)
   512		}
   513		return ranges, nil
   514	}
   515	
   516	// countingWriter counts how many bytes have been written to it.
   517	type countingWriter int64
   518	
   519	func (w *countingWriter) Write(p []byte) (n int, err error) {
   520		*w += countingWriter(len(p))
   521		return len(p), nil
   522	}
   523	
   524	// rangesMIMESize returns the nunber of bytes it takes to encode the
   525	// provided ranges as a multipart response.
   526	func rangesMIMESize(ranges []httpRange, contentType string, contentSize int64) (encSize int64) {
   527		var w countingWriter
   528		mw := multipart.NewWriter(&w)
   529		for _, ra := range ranges {
   530			mw.CreatePart(ra.mimeHeader(contentType, contentSize))
   531			encSize += ra.length
   532		}
   533		mw.Close()
   534		encSize += int64(w)
   535		return
   536	}
   537	
   538	func sumRangesSize(ranges []httpRange) (size int64) {
   539		for _, ra := range ranges {
   540			size += ra.length
   541		}
   542		return
   543	}

View as plain text