X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/7407f41105f8000bb3908d41a31daaf3a30d9440..9acc8cb9cd9ca4429712b0d31b647e9a6ecf2d96:/services/keep-web/handler.go diff --git a/services/keep-web/handler.go b/services/keep-web/handler.go index 643ca4f587..54b8c02165 100644 --- a/services/keep-web/handler.go +++ b/services/keep-web/handler.go @@ -2,10 +2,11 @@ // // SPDX-License-Identifier: AGPL-3.0 -package main +package keepweb import ( "encoding/json" + "fmt" "html" "html/template" "io" @@ -22,7 +23,6 @@ import ( "git.arvados.org/arvados.git/sdk/go/arvadosclient" "git.arvados.org/arvados.git/sdk/go/auth" "git.arvados.org/arvados.git/sdk/go/ctxlog" - "git.arvados.org/arvados.git/sdk/go/health" "git.arvados.org/arvados.git/sdk/go/httpserver" "git.arvados.org/arvados.git/sdk/go/keepclient" "github.com/sirupsen/logrus" @@ -30,38 +30,18 @@ import ( ) type handler struct { - Config *Config - MetricsAPI http.Handler - clientPool *arvadosclient.ClientPool - setupOnce sync.Once - healthHandler http.Handler - webdavLS webdav.LockSystem -} - -// parseCollectionIDFromDNSName returns a UUID or PDH if s begins with -// a UUID or URL-encoded PDH; otherwise "". -func parseCollectionIDFromDNSName(s string) string { - // Strip domain. - if i := strings.IndexRune(s, '.'); i >= 0 { - s = s[:i] - } - // Names like {uuid}--collections.example.com serve the same - // purpose as {uuid}.collections.example.com but can reduce - // cost/effort of using [additional] wildcard certificates. - if i := strings.Index(s, "--"); i >= 0 { - s = s[:i] - } - if arvadosclient.UUIDMatch(s) { - return s - } - if pdh := strings.Replace(s, "-", "+", 1); arvadosclient.PDHMatch(pdh) { - return pdh - } - return "" + Cache cache + Cluster *arvados.Cluster + clientPool *arvadosclient.ClientPool + setupOnce sync.Once + webdavLS webdav.LockSystem } var urlPDHDecoder = strings.NewReplacer(" ", "+", "-", "+") +var notFoundMessage = "404 Not found\r\n\r\nThe requested path was not found, or you do not have permission to access it.\r" +var unauthorizedMessage = "401 Unauthorized\r\n\r\nA valid Arvados token must be provided to access this resource.\r" + // parseCollectionIDFromURL returns a UUID or PDH if s is a UUID or a // PDH (even if it is a PDH with "+" replaced by " " or "-"); // otherwise "". @@ -77,16 +57,10 @@ func parseCollectionIDFromURL(s string) string { func (h *handler) setup() { // Errors will be handled at the client pool. - arv, _ := arvados.NewClientFromConfig(h.Config.cluster) + arv, _ := arvados.NewClientFromConfig(h.Cluster) h.clientPool = arvadosclient.MakeClientPoolWith(arv) - keepclient.RefreshServiceDiscoveryOnSIGHUP() - keepclient.DefaultBlockCache.MaxBlocks = h.Config.cluster.Collections.WebDAVCache.MaxBlockEntries - - h.healthHandler = &health.Handler{ - Token: h.Config.cluster.ManagementToken, - Prefix: "/_health/", - } + keepclient.DefaultBlockCache.MaxBlocks = h.Cluster.Collections.WebDAVCache.MaxBlockEntries // Even though we don't accept LOCK requests, every webdav // handler must have a non-nil LockSystem. @@ -181,25 +155,36 @@ var ( } ) +func stripDefaultPort(host string) string { + // Will consider port 80 and port 443 to be the same vhost. I think that's fine. + u := &url.URL{Host: host} + if p := u.Port(); p == "80" || p == "443" { + return strings.ToLower(u.Hostname()) + } else { + return strings.ToLower(host) + } +} + +// CheckHealth implements service.Handler. +func (h *handler) CheckHealth() error { + return nil +} + +// Done implements service.Handler. +func (h *handler) Done() <-chan struct{} { + return nil +} + // ServeHTTP implements http.Handler. func (h *handler) ServeHTTP(wOrig http.ResponseWriter, r *http.Request) { h.setupOnce.Do(h.setup) - remoteAddr := r.RemoteAddr - if xff := r.Header.Get("X-Forwarded-For"); xff != "" { - remoteAddr = xff + "," + remoteAddr - } if xfp := r.Header.Get("X-Forwarded-Proto"); xfp != "" && xfp != "http" { r.URL.Scheme = xfp } w := httpserver.WrapResponseWriter(wOrig) - if strings.HasPrefix(r.URL.Path, "/_health/") && r.Method == "GET" { - h.healthHandler.ServeHTTP(w, r) - return - } - if method := r.Header.Get("Access-Control-Request-Method"); method != "" && r.Method == "OPTIONS" { if !browserMethod[method] && !webdavMethod[method] { w.WriteHeader(http.StatusMethodNotAllowed) @@ -227,6 +212,10 @@ func (h *handler) ServeHTTP(wOrig http.ResponseWriter, r *http.Request) { w.Header().Set("Access-Control-Expose-Headers", "Content-Range") } + if h.serveS3(w, r) { + return + } + pathParts := strings.Split(r.URL.Path[1:], "/") var stripParts int @@ -236,24 +225,27 @@ func (h *handler) ServeHTTP(wOrig http.ResponseWriter, r *http.Request) { var pathToken bool var attachment bool var useSiteFS bool - credentialsOK := h.Config.cluster.Collections.TrustAllContent + credentialsOK := h.Cluster.Collections.TrustAllContent + reasonNotAcceptingCredentials := "" - if r.Host != "" && r.Host == h.Config.cluster.Services.WebDAVDownload.ExternalURL.Host { + if r.Host != "" && stripDefaultPort(r.Host) == stripDefaultPort(h.Cluster.Services.WebDAVDownload.ExternalURL.Host) { credentialsOK = true attachment = true } else if r.FormValue("disposition") == "attachment" { attachment = true } - if collectionID = parseCollectionIDFromDNSName(r.Host); collectionID != "" { + if !credentialsOK { + reasonNotAcceptingCredentials = fmt.Sprintf("vhost %q does not specify a single collection ID or match Services.WebDAVDownload.ExternalURL %q, and Collections.TrustAllContent is false", + r.Host, h.Cluster.Services.WebDAVDownload.ExternalURL) + } + + if collectionID = arvados.CollectionIDFromDNSName(r.Host); collectionID != "" { // http://ID.collections.example/PATH... credentialsOK = true } else if r.URL.Path == "/status.json" { h.serveStatus(w, r) return - } else if strings.HasPrefix(r.URL.Path, "/metrics") { - h.MetricsAPI.ServeHTTP(w, r) - return } else if siteFSDir[pathParts[0]] { useSiteFS = true } else if len(pathParts) >= 1 && strings.HasPrefix(pathParts[0], "c=") { @@ -275,11 +267,12 @@ func (h *handler) ServeHTTP(wOrig http.ResponseWriter, r *http.Request) { // data. Tokens provided with the request are // ignored. credentialsOK = false + reasonNotAcceptingCredentials = "the '/collections/UUID/PATH' form only works for public data" } } if collectionID == "" && !useSiteFS { - w.WriteHeader(http.StatusNotFound) + http.Error(w, notFoundMessage, http.StatusNotFound) return } @@ -293,27 +286,32 @@ func (h *handler) ServeHTTP(wOrig http.ResponseWriter, r *http.Request) { } formToken := r.FormValue("api_token") - if formToken != "" && r.Header.Get("Origin") != "" && attachment && r.URL.Query().Get("api_token") == "" { - // The client provided an explicit token in the POST - // body. The Origin header indicates this *might* be - // an AJAX request, in which case redirect-with-cookie - // won't work: we should just serve the content in the - // POST response. This is safe because: + origin := r.Header.Get("Origin") + cors := origin != "" && !strings.HasSuffix(origin, "://"+r.Host) + safeAjax := cors && (r.Method == http.MethodGet || r.Method == http.MethodHead) + safeAttachment := attachment && r.URL.Query().Get("api_token") == "" + if formToken == "" { + // No token to use or redact. + } else if safeAjax || safeAttachment { + // If this is a cross-origin request, the URL won't + // appear in the browser's address bar, so + // substituting a clipboard-safe URL is pointless. + // Redirect-with-cookie wouldn't work anyway, because + // it's not safe to allow third-party use of our + // cookie. // - // * We're supplying an attachment, not inline - // content, so we don't need to convert the POST to - // a GET and avoid the "really resubmit form?" - // problem. - // - // * The token isn't embedded in the URL, so we don't - // need to worry about bookmarks and copy/paste. + // If we're supplying an attachment, we don't need to + // convert POST to GET to avoid the "really resubmit + // form?" problem, so provided the token isn't + // embedded in the URL, there's no reason to do + // redirect-with-cookie in this case either. reqTokens = append(reqTokens, formToken) - } else if formToken != "" && browserMethod[r.Method] { - // The client provided an explicit token in the query - // string, or a form in POST body. We must put the - // token in an HttpOnly cookie, and redirect to the - // same URL with the query param redacted and method = - // GET. + } else if browserMethod[r.Method] { + // If this is a page view, and the client provided a + // token via query string or POST body, we must put + // the token in an HttpOnly cookie, and redirect to an + // equivalent URL with the query param redacted and + // method = GET. h.seeOtherWithCookie(w, r, "", credentialsOK) return } @@ -338,7 +336,19 @@ func (h *handler) ServeHTTP(wOrig http.ResponseWriter, r *http.Request) { } if tokens == nil { - tokens = append(reqTokens, h.Config.cluster.Users.AnonymousUserToken) + tokens = reqTokens + if h.Cluster.Users.AnonymousUserToken != "" { + tokens = append(tokens, h.Cluster.Users.AnonymousUserToken) + } + } + + if tokens == nil { + if !credentialsOK { + http.Error(w, fmt.Sprintf("Authorization tokens are not accepted here: %v, and no anonymous user token is configured.", reasonNotAcceptingCredentials), http.StatusUnauthorized) + } else { + http.Error(w, fmt.Sprintf("No authorization token in request, and no anonymous user token is configured."), http.StatusUnauthorized) + } + return } if len(targetPath) > 0 && targetPath[0] == "_" { @@ -360,10 +370,11 @@ func (h *handler) ServeHTTP(wOrig http.ResponseWriter, r *http.Request) { defer h.clientPool.Put(arv) var collection *arvados.Collection + var tokenUser *arvados.User tokenResult := make(map[string]int) for _, arv.ApiToken = range tokens { var err error - collection, err = h.Config.Cache.Get(arv, collectionID, forceReload) + collection, err = h.Cache.Get(arv, collectionID, forceReload) if err == nil { // Success break @@ -388,14 +399,14 @@ func (h *handler) ServeHTTP(wOrig http.ResponseWriter, r *http.Request) { // for additional credentials would just be // confusing), or we don't even accept // credentials at this path. - w.WriteHeader(http.StatusNotFound) + http.Error(w, notFoundMessage, http.StatusNotFound) return } for _, t := range reqTokens { if tokenResult[t] == 404 { // The client provided valid token(s), but the // collection was not found. - w.WriteHeader(http.StatusNotFound) + http.Error(w, notFoundMessage, http.StatusNotFound) return } } @@ -409,7 +420,7 @@ func (h *handler) ServeHTTP(wOrig http.ResponseWriter, r *http.Request) { // data that has been deleted. Allow a referrer to // provide this context somehow? w.Header().Add("WWW-Authenticate", "Basic realm=\"collections\"") - w.WriteHeader(http.StatusUnauthorized) + http.Error(w, unauthorizedMessage, http.StatusUnauthorized) return } @@ -445,7 +456,17 @@ func (h *handler) ServeHTTP(wOrig http.ResponseWriter, r *http.Request) { return } + // Check configured permission + _, sess, err := h.Cache.GetSession(arv.ApiToken) + tokenUser, err = h.Cache.GetTokenUser(arv.ApiToken) + if webdavMethod[r.Method] { + if !h.userPermittedToUploadOrDownload(r.Method, tokenUser) { + http.Error(w, "Not permitted", http.StatusForbidden) + return + } + h.logUploadOrDownload(r, sess.arvadosclient, nil, strings.Join(targetPath, "/"), collection, tokenUser) + if writeMethod[r.Method] { // Save the collection only if/when all // webdav->filesystem operations succeed -- @@ -455,7 +476,7 @@ func (h *handler) ServeHTTP(wOrig http.ResponseWriter, r *http.Request) { ResponseWriter: w, logger: ctxlog.FromContext(r.Context()), update: func() error { - return h.Config.Cache.Update(client, *collection, writefs) + return h.Cache.Update(client, *collection, writefs) }} } h := webdav.Handler{ @@ -477,13 +498,18 @@ func (h *handler) ServeHTTP(wOrig http.ResponseWriter, r *http.Request) { } openPath := "/" + strings.Join(targetPath, "/") - if f, err := fs.Open(openPath); os.IsNotExist(err) { + f, err := fs.Open(openPath) + if os.IsNotExist(err) { // Requested non-existent path - w.WriteHeader(http.StatusNotFound) + http.Error(w, notFoundMessage, http.StatusNotFound) + return } else if err != nil { // Some other (unexpected) error http.Error(w, "open: "+err.Error(), http.StatusInternalServerError) - } else if stat, err := f.Stat(); err != nil { + return + } + defer f.Close() + if stat, err := f.Stat(); err != nil { // Can't get Size/IsDir (shouldn't happen with a collectionFS!) http.Error(w, "stat: "+err.Error(), http.StatusInternalServerError) } else if stat.IsDir() && !strings.HasSuffix(r.URL.Path, "/") { @@ -495,51 +521,64 @@ func (h *handler) ServeHTTP(wOrig http.ResponseWriter, r *http.Request) { } else if stat.IsDir() { h.serveDirectory(w, r, collection.Name, fs, openPath, true) } else { + if !h.userPermittedToUploadOrDownload(r.Method, tokenUser) { + http.Error(w, "Not permitted", http.StatusForbidden) + return + } + h.logUploadOrDownload(r, sess.arvadosclient, nil, strings.Join(targetPath, "/"), collection, tokenUser) + http.ServeContent(w, r, basename, stat.ModTime(), f) - if wrote := int64(w.WroteBodyBytes()); wrote != stat.Size() && r.Header.Get("Range") == "" { + if wrote := int64(w.WroteBodyBytes()); wrote != stat.Size() && w.WroteStatus() == http.StatusOK { // If we wrote fewer bytes than expected, it's // too late to change the real response code // or send an error message to the client, but // at least we can try to put some useful // debugging info in the logs. n, err := f.Read(make([]byte, 1024)) - ctxlog.FromContext(r.Context()).Errorf("stat.Size()==%d but only wrote %d bytes; read(1024) returns %d, %s", stat.Size(), wrote, n, err) - + ctxlog.FromContext(r.Context()).Errorf("stat.Size()==%d but only wrote %d bytes; read(1024) returns %d, %v", stat.Size(), wrote, n, err) } } } +func (h *handler) getClients(reqID, token string) (arv *arvadosclient.ArvadosClient, kc *keepclient.KeepClient, client *arvados.Client, release func(), err error) { + arv = h.clientPool.Get() + if arv == nil { + err = h.clientPool.Err() + return + } + release = func() { h.clientPool.Put(arv) } + arv.ApiToken = token + kc, err = keepclient.MakeKeepClient(arv) + if err != nil { + release() + return + } + kc.RequestID = reqID + client = (&arvados.Client{ + APIHost: arv.ApiServer, + AuthToken: arv.ApiToken, + Insecure: arv.ApiInsecure, + }).WithRequestID(reqID) + return +} + func (h *handler) serveSiteFS(w http.ResponseWriter, r *http.Request, tokens []string, credentialsOK, attachment bool) { if len(tokens) == 0 { w.Header().Add("WWW-Authenticate", "Basic realm=\"collections\"") - http.Error(w, http.StatusText(http.StatusUnauthorized), http.StatusUnauthorized) + http.Error(w, unauthorizedMessage, http.StatusUnauthorized) return } if writeMethod[r.Method] { http.Error(w, errReadOnly.Error(), http.StatusMethodNotAllowed) return } - arv := h.clientPool.Get() - if arv == nil { - http.Error(w, "Pool failed: "+h.clientPool.Err().Error(), http.StatusInternalServerError) - return - } - defer h.clientPool.Put(arv) - arv.ApiToken = tokens[0] - kc, err := keepclient.MakeKeepClient(arv) + fs, sess, err := h.Cache.GetSession(tokens[0]) if err != nil { - http.Error(w, "error setting up keep client: "+err.Error(), http.StatusInternalServerError) + http.Error(w, err.Error(), http.StatusInternalServerError) return } - kc.RequestID = r.Header.Get("X-Request-Id") - client := (&arvados.Client{ - APIHost: arv.ApiServer, - AuthToken: arv.ApiToken, - Insecure: arv.ApiInsecure, - }).WithRequestID(r.Header.Get("X-Request-Id")) - fs := client.SiteFileSystem(kc) - fs.ForwardSlashNameSubstitution(h.Config.cluster.Collections.ForwardSlashNameSubstitution) + fs.ForwardSlashNameSubstitution(h.Cluster.Collections.ForwardSlashNameSubstitution) f, err := fs.Open(r.URL.Path) if os.IsNotExist(err) { http.Error(w, err.Error(), http.StatusNotFound) @@ -557,6 +596,14 @@ func (h *handler) serveSiteFS(w http.ResponseWriter, r *http.Request, tokens []s } return } + + tokenUser, err := h.Cache.GetTokenUser(tokens[0]) + if !h.userPermittedToUploadOrDownload(r.Method, tokenUser) { + http.Error(w, "Not permitted", http.StatusForbidden) + return + } + h.logUploadOrDownload(r, sess.arvadosclient, fs, r.URL.Path, nil, tokenUser) + if r.Method == "GET" { _, basename := filepath.Split(r.URL.Path) applyContentDispositionHdr(w, r, basename, attachment) @@ -756,6 +803,7 @@ func (h *handler) seeOtherWithCookie(w http.ResponseWriter, r *http.Request, loc Value: auth.EncodeTokenCookie([]byte(formToken)), Path: "/", HttpOnly: true, + SameSite: http.SameSiteLaxMode, }) } @@ -786,3 +834,125 @@ func (h *handler) seeOtherWithCookie(w http.ResponseWriter, r *http.Request, loc io.WriteString(w, html.EscapeString(redir)) io.WriteString(w, `">Continue`) } + +func (h *handler) userPermittedToUploadOrDownload(method string, tokenUser *arvados.User) bool { + var permitDownload bool + var permitUpload bool + if tokenUser != nil && tokenUser.IsAdmin { + permitUpload = h.Cluster.Collections.WebDAVPermission.Admin.Upload + permitDownload = h.Cluster.Collections.WebDAVPermission.Admin.Download + } else { + permitUpload = h.Cluster.Collections.WebDAVPermission.User.Upload + permitDownload = h.Cluster.Collections.WebDAVPermission.User.Download + } + if (method == "PUT" || method == "POST") && !permitUpload { + // Disallow operations that upload new files. + // Permit webdav operations that move existing files around. + return false + } else if method == "GET" && !permitDownload { + // Disallow downloading file contents. + // Permit webdav operations like PROPFIND that retrieve metadata + // but not file contents. + return false + } + return true +} + +func (h *handler) logUploadOrDownload( + r *http.Request, + client *arvadosclient.ArvadosClient, + fs arvados.CustomFileSystem, + filepath string, + collection *arvados.Collection, + user *arvados.User) { + + log := ctxlog.FromContext(r.Context()) + props := make(map[string]string) + props["reqPath"] = r.URL.Path + var useruuid string + if user != nil { + log = log.WithField("user_uuid", user.UUID). + WithField("user_full_name", user.FullName) + useruuid = user.UUID + } else { + useruuid = fmt.Sprintf("%s-tpzed-anonymouspublic", h.Cluster.ClusterID) + } + if collection == nil && fs != nil { + collection, filepath = h.determineCollection(fs, filepath) + } + if collection != nil { + log = log.WithField("collection_uuid", collection.UUID). + WithField("collection_file_path", filepath) + props["collection_uuid"] = collection.UUID + props["collection_file_path"] = filepath + // h.determineCollection populates the collection_uuid prop with the PDH, if + // this collection is being accessed via PDH. In that case, blank the + // collection_uuid field so that consumers of the log entries can rely on it + // being a UUID, or blank. The PDH remains available via the + // portable_data_hash property. + if props["collection_uuid"] == collection.PortableDataHash { + props["collection_uuid"] = "" + } + } + if r.Method == "PUT" || r.Method == "POST" { + log.Info("File upload") + if h.Cluster.Collections.WebDAVLogEvents { + go func() { + lr := arvadosclient.Dict{"log": arvadosclient.Dict{ + "object_uuid": useruuid, + "event_type": "file_upload", + "properties": props}} + err := client.Create("logs", lr, nil) + if err != nil { + log.WithError(err).Error("Failed to create upload log event on API server") + } + }() + } + } else if r.Method == "GET" { + if collection != nil && collection.PortableDataHash != "" { + log = log.WithField("portable_data_hash", collection.PortableDataHash) + props["portable_data_hash"] = collection.PortableDataHash + } + log.Info("File download") + if h.Cluster.Collections.WebDAVLogEvents { + go func() { + lr := arvadosclient.Dict{"log": arvadosclient.Dict{ + "object_uuid": useruuid, + "event_type": "file_download", + "properties": props}} + err := client.Create("logs", lr, nil) + if err != nil { + log.WithError(err).Error("Failed to create download log event on API server") + } + }() + } + } +} + +func (h *handler) determineCollection(fs arvados.CustomFileSystem, path string) (*arvados.Collection, string) { + segments := strings.Split(path, "/") + var i int + for i = 0; i < len(segments); i++ { + dir := append([]string{}, segments[0:i]...) + dir = append(dir, ".arvados#collection") + f, err := fs.OpenFile(strings.Join(dir, "/"), os.O_RDONLY, 0) + if f != nil { + defer f.Close() + } + if err != nil { + if !os.IsNotExist(err) { + return nil, "" + } + continue + } + // err is nil so we found it. + decoder := json.NewDecoder(f) + var collection arvados.Collection + err = decoder.Decode(&collection) + if err != nil { + return nil, "" + } + return &collection, strings.Join(segments[i:], "/") + } + return nil, "" +}