httpserver/all: Clean up and standardize request URL handling (#1633)

* httpserver/all: Clean up and standardize request URL handling

The HTTP server now always creates a context value on the request which
is a copy of the request's URL struct. It should not be modified by
middlewares, but it is safe to get the value out of the request and make
changes to it locally-scoped. Thus, the value in the context always
stores the original request URL information as it was received. Any
rewrites that happen will be to the request's URL field directly.

The HTTP server no longer cleans /sanitizes the request URL. It made too
many strong assumptions and ended up making a lot of middleware more
complicated, including upstream proxying (and fastcgi). To alleviate
this complexity, we no longer change the request URL. Middlewares are
responsible to access the disk safely by using http.Dir or, if not
actually opening files, they can use httpserver.SafePath().

I'm hoping this will address issues with #1624, #1584, #1582, and others.

* staticfiles: Fix test on Windows

@abiosoft: I still can't figure out exactly what this is for. 😅

* Use (potentially) changed URL for browse redirects, as before

* Use filepath.ToSlash, clean up a couple proxy test cases

* Oops, fix variable name
This commit is contained in:
Matt Holt
2017-05-01 23:11:10 -06:00
committed by GitHub
parent 5685a16449
commit d5371aff22
23 changed files with 366 additions and 574 deletions

View File

@@ -1,6 +1,7 @@
package httpserver
import (
"context"
"fmt"
"net/http"
"strings"
@@ -94,16 +95,21 @@ func TestConditions(t *testing.T) {
for i, test := range replaceTests {
r, err := http.NewRequest("GET", test.url, nil)
if err != nil {
t.Error(err)
t.Errorf("Test %d: failed to create request: %v", i, err)
continue
}
ctx := context.WithValue(r.Context(), OriginalURLCtxKey, *r.URL)
r = r.WithContext(ctx)
str := strings.Fields(test.condition)
ifCond, err := newIfCond(str[0], str[1], str[2])
if err != nil {
t.Error(err)
t.Errorf("Test %d: failed to create 'if' condition %v", i, err)
continue
}
isTrue := ifCond.True(r)
if isTrue != test.isTrue {
t.Errorf("Test %v: expected %v found %v", i, test.isTrue, isTrue)
continue
}
}
}

View File

@@ -198,15 +198,11 @@ func SameNext(next1, next2 Handler) bool {
return fmt.Sprintf("%v", next1) == fmt.Sprintf("%v", next2)
}
// Context key constants
// Context key constants.
const (
// URIxRewriteCtxKey is a context key used to store original unrewritten
// URI in context.WithValue
URIxRewriteCtxKey caddy.CtxKey = "caddy_rewrite_original_uri"
// RemoteUserCtxKey is a context key used to store remote user for request
// RemoteUserCtxKey is the key for the remote user of the request, if any (basicauth).
RemoteUserCtxKey caddy.CtxKey = "remote_user"
// MitmCtxKey stores Mitm result
// MitmCtxKey is the key for the result of MITM detection
MitmCtxKey caddy.CtxKey = "mitm"
)

View File

@@ -1,76 +0,0 @@
package httpserver
import (
"math/rand"
"path"
"strings"
"time"
)
// CleanMaskedPath prevents one or more of the path cleanup operations:
// - collapse multiple slashes into one
// - eliminate "/." (current directory)
// - eliminate "<parent_directory>/.."
// by masking certain patterns in the path with a temporary random string.
// This could be helpful when certain patterns in the path are desired to be preserved
// that would otherwise be changed by path.Clean().
// One such use case is the presence of the double slashes as protocol separator
// (e.g., /api/endpoint/http://example.com).
// This is a common pattern in many applications to allow passing URIs as path argument.
func CleanMaskedPath(reqPath string, masks ...string) string {
var replacerVal string
maskMap := make(map[string]string)
// Iterate over supplied masks and create temporary replacement strings
// only for the masks that are present in the path, then replace all occurrences
for _, mask := range masks {
if strings.Index(reqPath, mask) >= 0 {
replacerVal = "/_caddy" + generateRandomString() + "__"
maskMap[mask] = replacerVal
reqPath = strings.Replace(reqPath, mask, replacerVal, -1)
}
}
reqPath = path.Clean(reqPath)
// Revert the replaced masks after path cleanup
for mask, replacerVal := range maskMap {
reqPath = strings.Replace(reqPath, replacerVal, mask, -1)
}
return reqPath
}
// CleanPath calls CleanMaskedPath() with the default mask of "://"
// to preserve double slashes of protocols
// such as "http://", "https://", and "ftp://" etc.
func CleanPath(reqPath string) string {
return CleanMaskedPath(reqPath, "://")
}
// An efficient and fast method for random string generation.
// Inspired by http://stackoverflow.com/a/31832326.
const randomStringLength = 4
const letterBytes = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ"
const (
letterIdxBits = 6
letterIdxMask = 1<<letterIdxBits - 1
letterIdxMax = 63 / letterIdxBits
)
var src = rand.NewSource(time.Now().UnixNano())
func generateRandomString() string {
b := make([]byte, randomStringLength)
for i, cache, remain := randomStringLength-1, src.Int63(), letterIdxMax; i >= 0; {
if remain == 0 {
cache, remain = src.Int63(), letterIdxMax
}
if idx := int(cache & letterIdxMask); idx < len(letterBytes) {
b[i] = letterBytes[idx]
i--
}
cache >>= letterIdxBits
remain--
}
return string(b)
}

View File

@@ -1,120 +0,0 @@
package httpserver
import (
"path"
"testing"
)
var paths = map[string]map[string]string{
"/../a/b/../././/c": {
"preserve_all": "/../a/b/../././/c",
"preserve_protocol": "/a/c",
"preserve_slashes": "/a//c",
"preserve_dots": "/../a/b/../././c",
"clean_all": "/a/c",
},
"/path/https://www.google.com": {
"preserve_all": "/path/https://www.google.com",
"preserve_protocol": "/path/https://www.google.com",
"preserve_slashes": "/path/https://www.google.com",
"preserve_dots": "/path/https:/www.google.com",
"clean_all": "/path/https:/www.google.com",
},
"/a/b/../././/c/http://example.com/foo//bar/../blah": {
"preserve_all": "/a/b/../././/c/http://example.com/foo//bar/../blah",
"preserve_protocol": "/a/c/http://example.com/foo/blah",
"preserve_slashes": "/a//c/http://example.com/foo/blah",
"preserve_dots": "/a/b/../././c/http:/example.com/foo/bar/../blah",
"clean_all": "/a/c/http:/example.com/foo/blah",
},
}
func assertEqual(t *testing.T, expected, received string) {
if expected != received {
t.Errorf("\tExpected: %s\n\t\t\tReceived: %s", expected, received)
}
}
func maskedTestRunner(t *testing.T, variation string, masks ...string) {
for reqPath, transformation := range paths {
assertEqual(t, transformation[variation], CleanMaskedPath(reqPath, masks...))
}
}
// No need to test the built-in path.Clean() function.
// However, it could be useful to cross-examine the test dataset.
func TestPathClean(t *testing.T) {
for reqPath, transformation := range paths {
assertEqual(t, transformation["clean_all"], path.Clean(reqPath))
}
}
func TestCleanAll(t *testing.T) {
maskedTestRunner(t, "clean_all")
}
func TestPreserveAll(t *testing.T) {
maskedTestRunner(t, "preserve_all", "//", "/..", "/.")
}
func TestPreserveProtocol(t *testing.T) {
maskedTestRunner(t, "preserve_protocol", "://")
}
func TestPreserveSlashes(t *testing.T) {
maskedTestRunner(t, "preserve_slashes", "//")
}
func TestPreserveDots(t *testing.T) {
maskedTestRunner(t, "preserve_dots", "/..", "/.")
}
func TestDefaultMask(t *testing.T) {
for reqPath, transformation := range paths {
assertEqual(t, transformation["preserve_protocol"], CleanPath(reqPath))
}
}
func maskedBenchmarkRunner(b *testing.B, masks ...string) {
for n := 0; n < b.N; n++ {
for reqPath := range paths {
CleanMaskedPath(reqPath, masks...)
}
}
}
func BenchmarkPathClean(b *testing.B) {
for n := 0; n < b.N; n++ {
for reqPath := range paths {
path.Clean(reqPath)
}
}
}
func BenchmarkCleanAll(b *testing.B) {
maskedBenchmarkRunner(b)
}
func BenchmarkPreserveAll(b *testing.B) {
maskedBenchmarkRunner(b, "//", "/..", "/.")
}
func BenchmarkPreserveProtocol(b *testing.B) {
maskedBenchmarkRunner(b, "://")
}
func BenchmarkPreserveSlashes(b *testing.B) {
maskedBenchmarkRunner(b, "//")
}
func BenchmarkPreserveDots(b *testing.B) {
maskedBenchmarkRunner(b, "/..", "/.")
}
func BenchmarkDefaultMask(b *testing.B) {
for n := 0; n < b.N; n++ {
for reqPath := range paths {
CleanPath(reqPath)
}
}
}

View File

@@ -238,37 +238,24 @@ func (r *replacer) getSubstitution(key string) string {
}
return host
case "{path}":
// if a rewrite has happened, the original URI should be used as the path
// rather than the rewritten URI
var path string
origpath, _ := r.request.Context().Value(URIxRewriteCtxKey).(string)
if origpath == "" {
path = r.request.URL.Path
} else {
parsedURL, _ := url.Parse(origpath)
path = parsedURL.Path
}
return path
u, _ := r.request.Context().Value(OriginalURLCtxKey).(url.URL)
return u.Path
case "{path_escaped}":
var path string
origpath, _ := r.request.Context().Value(URIxRewriteCtxKey).(string)
if origpath == "" {
path = r.request.URL.Path
} else {
parsedURL, _ := url.Parse(origpath)
path = parsedURL.Path
}
return url.QueryEscape(path)
u, _ := r.request.Context().Value(OriginalURLCtxKey).(url.URL)
return url.QueryEscape(u.Path)
case "{rewrite_path}":
return r.request.URL.Path
case "{rewrite_path_escaped}":
return url.QueryEscape(r.request.URL.Path)
case "{query}":
return r.request.URL.RawQuery
u, _ := r.request.Context().Value(OriginalURLCtxKey).(url.URL)
return u.RawQuery
case "{query_escaped}":
return url.QueryEscape(r.request.URL.RawQuery)
u, _ := r.request.Context().Value(OriginalURLCtxKey).(url.URL)
return url.QueryEscape(u.RawQuery)
case "{fragment}":
return r.request.URL.Fragment
u, _ := r.request.Context().Value(OriginalURLCtxKey).(url.URL)
return u.Fragment
case "{proto}":
return r.request.Proto
case "{remote}":
@@ -284,17 +271,11 @@ func (r *replacer) getSubstitution(key string) string {
}
return port
case "{uri}":
uri, _ := r.request.Context().Value(URIxRewriteCtxKey).(string)
if uri == "" {
uri = r.request.URL.RequestURI()
}
return uri
u, _ := r.request.Context().Value(OriginalURLCtxKey).(url.URL)
return u.RequestURI()
case "{uri_escaped}":
uri, _ := r.request.Context().Value(URIxRewriteCtxKey).(string)
if uri == "" {
uri = r.request.URL.RequestURI()
}
return url.QueryEscape(uri)
u, _ := r.request.Context().Value(OriginalURLCtxKey).(url.URL)
return url.QueryEscape(u.RequestURI())
case "{rewrite_uri}":
return r.request.URL.RequestURI()
case "{rewrite_uri_escaped}":

View File

@@ -41,8 +41,11 @@ func TestReplace(t *testing.T) {
request, err := http.NewRequest("POST", "http://localhost/?foo=bar", reader)
if err != nil {
t.Fatal("Request Formation Failed\n")
t.Fatalf("Failed to make request: %v", err)
}
ctx := context.WithValue(request.Context(), OriginalURLCtxKey, *request.URL)
request = request.WithContext(ctx)
request.Header.Set("Custom", "foobarbaz")
request.Header.Set("ShorterVal", "1")
repl := NewReplacer(request, recordRequest, "-")
@@ -52,7 +55,7 @@ func TestReplace(t *testing.T) {
hostname, err := os.Hostname()
if err != nil {
t.Fatal("Failed to determine hostname\n")
t.Fatalf("Failed to determine hostname: %v", err)
}
old := now
@@ -161,25 +164,26 @@ func TestPathRewrite(t *testing.T) {
if err != nil {
t.Fatalf("Request Formation Failed: %s\n", err.Error())
}
ctx := context.WithValue(request.Context(), URIxRewriteCtxKey, "a/custom/path.php?key=value")
urlCopy := *request.URL
urlCopy.Path = "a/custom/path.php"
ctx := context.WithValue(request.Context(), OriginalURLCtxKey, urlCopy)
request = request.WithContext(ctx)
repl := NewReplacer(request, recordRequest, "")
if repl.Replace("This path is '{path}'") != "This path is 'a/custom/path.php'" {
t.Error("Expected host {path} replacement failed (" + repl.Replace("This path is '{path}'") + ")")
if got, want := repl.Replace("This path is '{path}'"), "This path is 'a/custom/path.php'"; got != want {
t.Errorf("{path} replacement failed; got '%s', want '%s'", got, want)
}
if repl.Replace("This path is {rewrite_path}") != "This path is /index.php" {
t.Error("Expected host {rewrite_path} replacement failed (" + repl.Replace("This path is {rewrite_path}") + ")")
if got, want := repl.Replace("This path is {rewrite_path}"), "This path is /index.php"; got != want {
t.Errorf("{rewrite_path} replacement failed; got '%s', want '%s'", got, want)
}
if repl.Replace("This path is '{uri}'") != "This path is 'a/custom/path.php?key=value'" {
t.Error("Expected host {uri} replacement failed (" + repl.Replace("This path is '{uri}'") + ")")
if got, want := repl.Replace("This path is '{uri}'"), "This path is 'a/custom/path.php?key=value'"; got != want {
t.Errorf("{uri} replacement failed; got '%s', want '%s'", got, want)
}
if repl.Replace("This path is {rewrite_uri}") != "This path is /index.php?key=value" {
t.Error("Expected host {rewrite_uri} replacement failed (" + repl.Replace("This path is {rewrite_uri}") + ")")
if got, want := repl.Replace("This path is {rewrite_uri}"), "This path is /index.php?key=value"; got != want {
t.Errorf("{rewrite_uri} replacement failed; got '%s', want '%s'", got, want)
}
}

View File

@@ -9,7 +9,10 @@ import (
"log"
"net"
"net/http"
"net/url"
"os"
"path"
"path/filepath"
"runtime"
"strings"
"sync"
@@ -290,11 +293,18 @@ func (s *Server) ServeHTTP(w http.ResponseWriter, r *http.Request) {
}
}()
w.Header().Set("Server", "Caddy")
c := context.WithValue(r.Context(), staticfiles.URLPathCtxKey, r.URL.Path)
// copy the original, unchanged URL into the context
// so it can be referenced by middlewares
urlCopy := *r.URL
if r.URL.User != nil {
userInfo := new(url.Userinfo)
*userInfo = *r.URL.User
urlCopy.User = userInfo
}
c := context.WithValue(r.Context(), OriginalURLCtxKey, urlCopy)
r = r.WithContext(c)
sanitizePath(r)
w.Header().Set("Server", "Caddy")
status, _ := s.serveHTTP(w, r)
@@ -353,6 +363,7 @@ func (s *Server) serveHTTP(w http.ResponseWriter, r *http.Request) (int, error)
// The error returned by MaxBytesReader is meant to be handled
// by whichever middleware/plugin that receives it when calling
// .Read() or a similar method on the request body
// TODO: Make this middleware instead?
if r.Body != nil {
for _, pathlimit := range vhost.MaxRequestBodySizes {
if Path(r.URL.Path).Matches(pathlimit.Path) {
@@ -407,28 +418,6 @@ func (s *Server) Stop() error {
return nil
}
// sanitizePath collapses any ./ ../ /// madness which helps prevent
// path traversal attacks. Note to middleware: use the value within the
// request's context at key caddy.URLPathContextKey to access the
// "original" URL.Path value.
func sanitizePath(r *http.Request) {
if r.URL.Path == "/" {
return
}
cleanedPath := CleanPath(r.URL.Path)
if cleanedPath == "." {
r.URL.Path = "/"
} else {
if !strings.HasPrefix(cleanedPath, "/") {
cleanedPath = "/" + cleanedPath
}
if strings.HasSuffix(r.URL.Path, "/") && !strings.HasSuffix(cleanedPath, "/") {
cleanedPath = cleanedPath + "/"
}
r.URL.Path = cleanedPath
}
}
// OnStartupComplete lists the sites served by this server
// and any relevant information, assuming caddy.Quiet == false.
func (s *Server) OnStartupComplete() {
@@ -558,3 +547,20 @@ func WriteTextResponse(w http.ResponseWriter, status int, body string) {
w.WriteHeader(status)
w.Write([]byte(body))
}
// SafePath joins siteRoot and reqPath and converts it to a path that can
// be used to access a path on the local disk. It ensures the path does
// not traverse outside of the site root.
//
// If opening a file, use http.Dir instead.
func SafePath(siteRoot, reqPath string) string {
reqPath = filepath.ToSlash(reqPath)
reqPath = strings.Replace(reqPath, "\x00", "", -1) // NOTE: Go 1.9 checks for null bytes in the syscall package
if siteRoot == "" {
siteRoot = "."
}
return filepath.Join(siteRoot, filepath.FromSlash(path.Clean("/"+reqPath)))
}
// OriginalURLCtxKey is the key for accessing the original, incoming URL on an HTTP request.
const OriginalURLCtxKey = caddy.CtxKey("original_url")