Use a more general (and faster) method to sanitize URLs with credentials (#19239) Use a more general method to sanitize URLs with credentials: Simple and intuitive / Faster / Remove all credentials in all URLs

Changes

12 changed files (+103/-194)

• models/migrations/v180.go
  • @@ -112,7 +112,7 @@ func removeCredentials(payload string) (string, error) {
    opts.AuthPassword = "" opts.AuthToken = "" - opts.CloneAddr = util.NewStringURLSanitizer(opts.CloneAddr, true).Replace(opts.CloneAddr) + opts.CloneAddr = util.SanitizeCredentialURLs(opts.CloneAddr) confBytes, err := json.Marshal(opts) if err != nil {
• models/task.go
  • @@ -245,7 +245,7 @@ func FinishMigrateTask(task *Task) error {
    } conf.AuthPassword = "" conf.AuthToken = "" - conf.CloneAddr = util.NewStringURLSanitizer(conf.CloneAddr, true).Replace(conf.CloneAddr) + conf.CloneAddr = util.SanitizeCredentialURLs(conf.CloneAddr) conf.AuthPasswordEncrypted = "" conf.AuthTokenEncrypted = "" conf.CloneAddrEncrypted = ""
• modules/git/command.go
  • @@ -154,7 +154,7 @@ func (c *Command) RunWithContext(rc *RunContext) error {
    args = make([]string, len(c.args)) copy(args, c.args) for _, urlArgIndex := range argSensitiveURLIndexes { - args[urlArgIndex] = util.NewStringURLSanitizer(args[urlArgIndex], true).Replace(args[urlArgIndex]) + args[urlArgIndex] = util.SanitizeCredentialURLs(args[urlArgIndex]) } } desc = fmt.Sprintf("%s %s [repo_path: %s]", c.name, strings.Join(args, " "), rc.Dir)
• modules/git/repo.go
  • @@ -156,7 +156,7 @@ func CloneWithArgs(ctx context.Context, from, to string, args []string, opts Clo
    cmd.AddArguments("--", from, to) if strings.Contains(from, "://") && strings.Contains(from, "@") { - cmd.SetDescription(fmt.Sprintf("clone branch %s from %s to %s (shared: %t, mirror: %t, depth: %d)", opts.Branch, util.NewStringURLSanitizer(from, true).Replace(from), to, opts.Shared, opts.Mirror, opts.Depth)) + cmd.SetDescription(fmt.Sprintf("clone branch %s from %s to %s (shared: %t, mirror: %t, depth: %d)", opts.Branch, util.SanitizeCredentialURLs(from), to, opts.Shared, opts.Mirror, opts.Depth)) } else { cmd.SetDescription(fmt.Sprintf("clone branch %s from %s to %s (shared: %t, mirror: %t, depth: %d)", opts.Branch, from, to, opts.Shared, opts.Mirror, opts.Depth)) }
  • @@ -209,7 +209,7 @@ func Push(ctx context.Context, repoPath string, opts PushOptions) error {
    cmd.AddArguments(opts.Branch) } if strings.Contains(opts.Remote, "://") && strings.Contains(opts.Remote, "@") { - cmd.SetDescription(fmt.Sprintf("push branch %s to %s (force: %t, mirror: %t)", opts.Branch, util.NewStringURLSanitizer(opts.Remote, true).Replace(opts.Remote), opts.Force, opts.Mirror)) + cmd.SetDescription(fmt.Sprintf("push branch %s to %s (force: %t, mirror: %t)", opts.Branch, util.SanitizeCredentialURLs(opts.Remote), opts.Force, opts.Mirror)) } else { cmd.SetDescription(fmt.Sprintf("push branch %s to %s (force: %t, mirror: %t)", opts.Branch, opts.Remote, opts.Force, opts.Mirror)) }
• modules/util/sanitize.go
  • @@ -5,59 +5,71 @@
    package util import ( - "net/url" - "strings" - ) + "bytes" + "unicode" - const ( - userPlaceholder = "sanitized-credential" - unparsableURL = "(unparsable url)" + "github.com/yuin/goldmark/util" ) type sanitizedError struct { - err error - replacer *strings.Replacer + err error } func (err sanitizedError) Error() string { - return err.replacer.Replace(err.err.Error()) + return SanitizeCredentialURLs(err.err.Error()) } - // NewSanitizedError wraps an error and replaces all old, new string pairs in the message text. - func NewSanitizedError(err error, oldnew ...string) error { - return sanitizedError{err: err, replacer: strings.NewReplacer(oldnew...)} + func (err sanitizedError) Unwrap() error { + return err.err } - // NewURLSanitizedError wraps an error and replaces the url credential or removes them. - func NewURLSanitizedError(err error, u *url.URL, usePlaceholder bool) error { - return sanitizedError{err: err, replacer: NewURLSanitizer(u, usePlaceholder)} + // SanitizeErrorCredentialURLs wraps the error and make sure the returned error message doesn't contain sensitive credentials in URLs + func SanitizeErrorCredentialURLs(err error) error { + return sanitizedError{err: err} } - // NewStringURLSanitizedError wraps an error and replaces the url credential or removes them. - // If the url can't get parsed it gets replaced with a placeholder string. - func NewStringURLSanitizedError(err error, unsanitizedURL string, usePlaceholder bool) error { - return sanitizedError{err: err, replacer: NewStringURLSanitizer(unsanitizedURL, usePlaceholder)} - } + const userPlaceholder = "sanitized-credential" - // NewURLSanitizer creates a replacer for the url with the credential sanitized or removed. - func NewURLSanitizer(u *url.URL, usePlaceholder bool) *strings.Replacer { - old := u.String() + var schemeSep = []byte("://") - if u.User != nil && usePlaceholder { - u.User = url.User(userPlaceholder) - } else { - u.User = nil + // SanitizeCredentialURLs remove all credentials in URLs (starting with "scheme://") for the input string: "https://user:pass@domain.com" => "https://sanitized-credential@domain.com" + func SanitizeCredentialURLs(s string) string { + bs := util.StringToReadOnlyBytes(s) + schemeSepPos := bytes.Index(bs, schemeSep) + if schemeSepPos == -1 || bytes.IndexByte(bs[schemeSepPos:], '@') == -1 { + return s // fast return if there is no URL scheme or no userinfo } - return strings.NewReplacer(old, u.String()) - } - - // NewStringURLSanitizer creates a replacer for the url with the credential sanitized or removed. - // If the url can't get parsed it gets replaced with a placeholder string - func NewStringURLSanitizer(unsanitizedURL string, usePlaceholder bool) *strings.Replacer { - u, err := url.Parse(unsanitizedURL) - if err != nil { - // don't log the error, since it might contain unsanitized URL. - return strings.NewReplacer(unsanitizedURL, unparsableURL) + out := make([]byte, 0, len(bs)+len(userPlaceholder)) + for schemeSepPos != -1 { + schemeSepPos += 3 // skip the "://" + sepAtPos := -1 // the possible '@' position: "https://foo@[^here]host" + sepEndPos := schemeSepPos // the possible end position: "The https://host[^here] in log for test" + sepLoop: + for ; sepEndPos < len(bs); sepEndPos++ { + c := bs[sepEndPos] + if ('A' <= c && c <= 'Z') || ('a' <= c && c <= 'z') || ('0' <= c && c <= '9') { + continue + } + switch c { + case '@': + sepAtPos = sepEndPos + case '-', '.', '_', '~', '!', '$', '&', '\'', '(', ')', '*', '+', ',', ';', '=', ':', '%': + continue // due to RFC 3986, userinfo can contain - . _ ~ ! $ & ' ( ) * + , ; = : and any percent-encoded chars + default: + break sepLoop // if it is an invalid char for URL (eg: space, '/', and others), stop the loop + } + } + // if there is '@', and the string is like "s://u@h", then hide the "u" part + if sepAtPos != -1 && (schemeSepPos >= 4 && unicode.IsLetter(rune(bs[schemeSepPos-4]))) && sepAtPos-schemeSepPos > 0 && sepEndPos-sepAtPos > 0 { + out = append(out, bs[:schemeSepPos]...) + out = append(out, userPlaceholder...) + out = append(out, bs[sepAtPos:sepEndPos]...) + } else { + out = append(out, bs[:sepEndPos]...) + } + bs = bs[sepEndPos:] + schemeSepPos = bytes.Index(bs, schemeSep) } - return NewURLSanitizer(u, usePlaceholder) + out = append(out, bs...) + return util.BytesToReadOnlyString(out) }
• modules/util/sanitize_test.go
  • @@ -11,154 +11,65 @@ import (
    "github.com/stretchr/testify/assert" ) - func TestNewSanitizedError(t *testing.T) { - err := errors.New("error while secret on test") - err2 := NewSanitizedError(err) - assert.Equal(t, err.Error(), err2.Error()) - - cases := []struct { - input error - oldnew []string - expected string - }{ - // case 0 - { - errors.New("error while secret on test"), - []string{"secret", "replaced"}, - "error while replaced on test", - }, - // case 1 - { - errors.New("error while sec-ret on test"), - []string{"secret", "replaced"}, - "error while sec-ret on test", - }, - } - - for n, c := range cases { - err := NewSanitizedError(c.input, c.oldnew...) - - assert.Equal(t, c.expected, err.Error(), "case %d: error should match", n) - } + func TestSanitizeErrorCredentialURLs(t *testing.T) { + err := errors.New("error with https://a@b.com") + se := SanitizeErrorCredentialURLs(err) + assert.Equal(t, "error with https://"+userPlaceholder+"@b.com", se.Error()) } - func TestNewStringURLSanitizer(t *testing.T) { + func TestSanitizeCredentialURLs(t *testing.T) { cases := []struct { - input string - placeholder bool - expected string + input string + expected string }{ - // case 0 { "https://github.com/go-gitea/test_repo.git", - true, "https://github.com/go-gitea/test_repo.git", }, - // case 1 - { - "https://github.com/go-gitea/test_repo.git", - false, - "https://github.com/go-gitea/test_repo.git", - }, - // case 2 { "https://mytoken@github.com/go-gitea/test_repo.git", - true, "https://" + userPlaceholder + "@github.com/go-gitea/test_repo.git", }, - // case 3 - { - "https://mytoken@github.com/go-gitea/test_repo.git", - false, - "https://github.com/go-gitea/test_repo.git", - }, - // case 4 { "https://user:password@github.com/go-gitea/test_repo.git", - true, "https://" + userPlaceholder + "@github.com/go-gitea/test_repo.git", }, - // case 5 { - "https://user:password@github.com/go-gitea/test_repo.git", - false, - "https://github.com/go-gitea/test_repo.git", + "ftp://x@", + "ftp://" + userPlaceholder + "@", }, - // case 6 { - "https://gi\nthub.com/go-gitea/test_repo.git", - false, - unparsableURL, + "ftp://x/@", + "ftp://x/@", }, - } - - for n, c := range cases { - // uses NewURLSanitizer internally - result := NewStringURLSanitizer(c.input, c.placeholder).Replace(c.input) - - assert.Equal(t, c.expected, result, "case %d: error should match", n) - } - } - - func TestNewStringURLSanitizedError(t *testing.T) { - cases := []struct { - input string - placeholder bool - expected string - }{ - // case 0 { - "https://github.com/go-gitea/test_repo.git", - true, - "https://github.com/go-gitea/test_repo.git", - }, - // case 1 - { - "https://github.com/go-gitea/test_repo.git", - false, - "https://github.com/go-gitea/test_repo.git", + "ftp://u@x/@", // test multiple @ chars + "ftp://" + userPlaceholder + "@x/@", }, - // case 2 { - "https://mytoken@github.com/go-gitea/test_repo.git", - true, - "https://" + userPlaceholder + "@github.com/go-gitea/test_repo.git", + "😊ftp://u@x😊", // test unicode + "😊ftp://" + userPlaceholder + "@x😊", }, - // case 3 { - "https://mytoken@github.com/go-gitea/test_repo.git", - false, - "https://github.com/go-gitea/test_repo.git", + "://@", + "://@", }, - // case 4 { - "https://user:password@github.com/go-gitea/test_repo.git", - true, - "https://" + userPlaceholder + "@github.com/go-gitea/test_repo.git", + "//u:p@h", // do not process URLs without explicit scheme, they are not treated as "valid" URLs because there is no scheme context in string + "//u:p@h", }, - // case 5 { - "https://user:password@github.com/go-gitea/test_repo.git", - false, - "https://github.com/go-gitea/test_repo.git", + "s://u@h", // the minimal pattern to be sanitized + "s://" + userPlaceholder + "@h", }, - // case 6 { - "https://gi\nthub.com/go-gitea/test_repo.git", - false, - unparsableURL, + "URLs in log https://u:b@h and https://u:b@h:80/, with https://h.com and u@h.com", + "URLs in log https://" + userPlaceholder + "@h and https://" + userPlaceholder + "@h:80/, with https://h.com and u@h.com", }, } - encloseText := func(input string) string { - return "test " + input + " test" - } - for n, c := range cases { - err := errors.New(encloseText(c.input)) - - result := NewStringURLSanitizedError(err, c.input, c.placeholder) - - assert.Equal(t, encloseText(c.expected), result.Error(), "case %d: error should match", n) + result := SanitizeCredentialURLs(c.input) + assert.Equal(t, c.expected, result, "case %d: error should match", n) } }
• routers/api/v1/repo/migrate.go
  • @@ -236,7 +236,7 @@ func handleMigrateError(ctx *context.APIContext, repoOwner *user_model.User, rem
    case base.IsErrNotSupported(err): ctx.Error(http.StatusUnprocessableEntity, "", err) default: - err = util.NewStringURLSanitizedError(err, remoteAddr, true) + err = util.SanitizeErrorCredentialURLs(err) if strings.Contains(err.Error(), "Authentication failed") || strings.Contains(err.Error(), "Bad credentials") || strings.Contains(err.Error(), "could not read Username") {
• routers/web/repo/migrate.go
  • @@ -106,8 +106,7 @@ func handleMigrateError(ctx *context.Context, owner *user_model.User, err error,
    ctx.Data["Err_RepoName"] = true ctx.RenderWithErr(ctx.Tr("repo.form.name_pattern_not_allowed", err.(db.ErrNamePatternNotAllowed).Pattern), tpl, form) default: - remoteAddr, _ := forms.ParseRemoteAddr(form.CloneAddr, form.AuthUsername, form.AuthPassword) - err = util.NewStringURLSanitizedError(err, remoteAddr, true) + err = util.SanitizeErrorCredentialURLs(err) if strings.Contains(err.Error(), "Authentication failed") || strings.Contains(err.Error(), "Bad credentials") || strings.Contains(err.Error(), "could not read Username") {
• services/mirror/mirror_pull.go
  • @@ -40,7 +40,7 @@ func UpdateAddress(ctx context.Context, m *repo_model.Mirror, addr string) error
    cmd := git.NewCommand(ctx, "remote", "add", remoteName, "--mirror=fetch", addr) if strings.Contains(addr, "://") && strings.Contains(addr, "@") { - cmd.SetDescription(fmt.Sprintf("remote add %s --mirror=fetch %s [repo_path: %s]", remoteName, util.NewStringURLSanitizer(addr, true).Replace(addr), repoPath)) + cmd.SetDescription(fmt.Sprintf("remote add %s --mirror=fetch %s [repo_path: %s]", remoteName, util.SanitizeCredentialURLs(addr), repoPath)) } else { cmd.SetDescription(fmt.Sprintf("remote add %s --mirror=fetch %s [repo_path: %s]", remoteName, addr, repoPath)) }
  • @@ -60,7 +60,7 @@ func UpdateAddress(ctx context.Context, m *repo_model.Mirror, addr string) error
    cmd = git.NewCommand(ctx, "remote", "add", remoteName, "--mirror=fetch", wikiRemotePath) if strings.Contains(wikiRemotePath, "://") && strings.Contains(wikiRemotePath, "@") { - cmd.SetDescription(fmt.Sprintf("remote add %s --mirror=fetch %s [repo_path: %s]", remoteName, util.NewStringURLSanitizer(wikiRemotePath, true).Replace(wikiRemotePath), wikiPath)) + cmd.SetDescription(fmt.Sprintf("remote add %s --mirror=fetch %s [repo_path: %s]", remoteName, util.SanitizeCredentialURLs(wikiRemotePath), wikiPath)) } else { cmd.SetDescription(fmt.Sprintf("remote add %s --mirror=fetch %s [repo_path: %s]", remoteName, wikiRemotePath, wikiPath)) }
  • @@ -160,7 +160,6 @@ func pruneBrokenReferences(ctx context.Context,
    repoPath string, timeout time.Duration, stdoutBuilder, stderrBuilder *strings.Builder, - sanitizer *strings.Replacer, isWiki bool, ) error { wiki := ""
  • @@ -184,8 +183,8 @@ func pruneBrokenReferences(ctx context.Context,
    // sanitize the output, since it may contain the remote address, which may // contain a password - stderrMessage := sanitizer.Replace(stderr) - stdoutMessage := sanitizer.Replace(stdout) + stderrMessage := util.SanitizeCredentialURLs(stderr) + stdoutMessage := util.SanitizeCredentialURLs(stdout) log.Error("Failed to prune mirror repository %s%-v references:\nStdout: %s\nStderr: %s\nErr: %v", wiki, m.Repo, stdoutMessage, stderrMessage, pruneErr) desc := fmt.Sprintf("Failed to prune mirror repository %s'%s' references: %s", wiki, repoPath, stderrMessage)
  • @@ -229,11 +228,9 @@ func runSync(ctx context.Context, m *repo_model.Mirror) ([]*mirrorSyncResult, bo
    stdout := stdoutBuilder.String() stderr := stderrBuilder.String() - // sanitize the output, since it may contain the remote address, which may - // contain a password - sanitizer := util.NewURLSanitizer(remoteAddr, true) - stderrMessage := sanitizer.Replace(stderr) - stdoutMessage := sanitizer.Replace(stdout) + // sanitize the output, since it may contain the remote address, which may contain a password + stderrMessage := util.SanitizeCredentialURLs(stderr) + stdoutMessage := util.SanitizeCredentialURLs(stdout) // Now check if the error is a resolve reference due to broken reference if strings.Contains(stderr, "unable to resolve reference") && strings.Contains(stderr, "reference broken") {
  • @@ -241,7 +238,7 @@ func runSync(ctx context.Context, m *repo_model.Mirror) ([]*mirrorSyncResult, bo
    err = nil // Attempt prune - pruneErr := pruneBrokenReferences(ctx, m, repoPath, timeout, &stdoutBuilder, &stderrBuilder, sanitizer, false) + pruneErr := pruneBrokenReferences(ctx, m, repoPath, timeout, &stdoutBuilder, &stderrBuilder, false) if pruneErr == nil { // Successful prune - reattempt mirror stderrBuilder.Reset()
  • @@ -259,8 +256,8 @@ func runSync(ctx context.Context, m *repo_model.Mirror) ([]*mirrorSyncResult, bo
    // sanitize the output, since it may contain the remote address, which may // contain a password - stderrMessage = sanitizer.Replace(stderr) - stdoutMessage = sanitizer.Replace(stdout) + stderrMessage = util.SanitizeCredentialURLs(stderr) + stdoutMessage = util.SanitizeCredentialURLs(stdout) } } }
  • @@ -322,19 +319,9 @@ func runSync(ctx context.Context, m *repo_model.Mirror) ([]*mirrorSyncResult, bo
    stdout := stdoutBuilder.String() stderr := stderrBuilder.String() - // sanitize the output, since it may contain the remote address, which may - // contain a password - - remoteAddr, remoteErr := git.GetRemoteAddress(ctx, wikiPath, m.GetRemoteName()) - if remoteErr != nil { - log.Error("SyncMirrors [repo: %-v Wiki]: unable to get GetRemoteAddress Error %v", m.Repo, remoteErr) - } - - // sanitize the output, since it may contain the remote address, which may - // contain a password - sanitizer := util.NewURLSanitizer(remoteAddr, true) - stderrMessage := sanitizer.Replace(stderr) - stdoutMessage := sanitizer.Replace(stdout) + // sanitize the output, since it may contain the remote address, which may contain a password + stderrMessage := util.SanitizeCredentialURLs(stderr) + stdoutMessage := util.SanitizeCredentialURLs(stdout) // Now check if the error is a resolve reference due to broken reference if strings.Contains(stderrMessage, "unable to resolve reference") && strings.Contains(stderrMessage, "reference broken") {
  • @@ -342,7 +329,7 @@ func runSync(ctx context.Context, m *repo_model.Mirror) ([]*mirrorSyncResult, bo
    err = nil // Attempt prune - pruneErr := pruneBrokenReferences(ctx, m, repoPath, timeout, &stdoutBuilder, &stderrBuilder, sanitizer, true) + pruneErr := pruneBrokenReferences(ctx, m, repoPath, timeout, &stdoutBuilder, &stderrBuilder, true) if pruneErr == nil { // Successful prune - reattempt mirror stderrBuilder.Reset()
  • @@ -358,8 +345,8 @@ func runSync(ctx context.Context, m *repo_model.Mirror) ([]*mirrorSyncResult, bo
    }); err != nil { stdout := stdoutBuilder.String() stderr := stderrBuilder.String() - stderrMessage = sanitizer.Replace(stderr) - stdoutMessage = sanitizer.Replace(stdout) + stderrMessage = util.SanitizeCredentialURLs(stderr) + stdoutMessage = util.SanitizeCredentialURLs(stdout) } } }
• services/mirror/mirror_push.go
  • @@ -31,7 +31,7 @@ func AddPushMirrorRemote(ctx context.Context, m *repo_model.PushMirror, addr str
    addRemoteAndConfig := func(addr, path string) error { cmd := git.NewCommand(ctx, "remote", "add", "--mirror=push", m.RemoteName, addr) if strings.Contains(addr, "://") && strings.Contains(addr, "@") { - cmd.SetDescription(fmt.Sprintf("remote add %s --mirror=push %s [repo_path: %s]", m.RemoteName, util.NewStringURLSanitizer(addr, true).Replace(addr), path)) + cmd.SetDescription(fmt.Sprintf("remote add %s --mirror=push %s [repo_path: %s]", m.RemoteName, util.SanitizeCredentialURLs(addr), path)) } else { cmd.SetDescription(fmt.Sprintf("remote add %s --mirror=push %s [repo_path: %s]", m.RemoteName, addr, path)) }
  • @@ -147,7 +147,7 @@ func runPushSync(ctx context.Context, m *repo_model.PushMirror) error {
    endpoint := lfs.DetermineEndpoint(remoteAddr.String(), "") lfsClient := lfs.NewClient(endpoint, nil) if err := pushAllLFSObjects(ctx, gitRepo, lfsClient); err != nil { - return util.NewURLSanitizedError(err, remoteAddr, true) + return util.SanitizeErrorCredentialURLs(err) } }
  • @@ -161,7 +161,7 @@ func runPushSync(ctx context.Context, m *repo_model.PushMirror) error {
    }); err != nil { log.Error("Error pushing %s mirror[%d] remote %s: %v", path, m.ID, m.RemoteName, err) - return util.NewURLSanitizedError(err, remoteAddr, true) + return util.SanitizeErrorCredentialURLs(err) } return nil
• services/task/migrate.go
  • @@ -129,7 +129,7 @@ func runMigrateTask(t *models.Task) (err error) {
    } // remoteAddr may contain credentials, so we sanitize it - err = util.NewStringURLSanitizedError(err, opts.CloneAddr, true) + err = util.SanitizeErrorCredentialURLs(err) if strings.Contains(err.Error(), "Authentication failed") || strings.Contains(err.Error(), "could not read Username") { return fmt.Errorf("Authentication failed: %v", err.Error())
• services/task/task.go
  • @@ -77,7 +77,7 @@ func CreateMigrateTask(doer, u *user_model.User, opts base.MigrateOptions) (*mod
    if err != nil { return nil, err } - opts.CloneAddr = util.NewStringURLSanitizer(opts.CloneAddr, true).Replace(opts.CloneAddr) + opts.CloneAddr = util.SanitizeCredentialURLs(opts.CloneAddr) opts.AuthPasswordEncrypted, err = secret.EncryptSecret(setting.SecretKey, opts.AuthPassword) if err != nil { return nil, err