X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/60a60a176c77488cfacc7fd02e03b7b8c69de587..1123c446a167b531b0683004ff3283d50aa037d0:/services/keep-web/handler.go diff --git a/services/keep-web/handler.go b/services/keep-web/handler.go index 161864a8b9..563a59df01 100644 --- a/services/keep-web/handler.go +++ b/services/keep-web/handler.go @@ -6,30 +6,32 @@ package main import ( "encoding/json" - "fmt" "html" "html/template" "io" "net/http" "net/url" "os" + "path/filepath" "sort" "strconv" "strings" "sync" - "git.curoverse.com/arvados.git/sdk/go/arvados" - "git.curoverse.com/arvados.git/sdk/go/arvadosclient" - "git.curoverse.com/arvados.git/sdk/go/auth" - "git.curoverse.com/arvados.git/sdk/go/health" - "git.curoverse.com/arvados.git/sdk/go/httpserver" - "git.curoverse.com/arvados.git/sdk/go/keepclient" - arvadosVersion "git.curoverse.com/arvados.git/sdk/go/version" + "git.arvados.org/arvados.git/sdk/go/arvados" + "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" "golang.org/x/net/webdav" ) type handler struct { Config *Config + MetricsAPI http.Handler clientPool *arvadosclient.ClientPool setupOnce sync.Once healthHandler http.Handler @@ -77,9 +79,10 @@ func (h *handler) setup() { h.clientPool = arvadosclient.MakeClientPool() keepclient.RefreshServiceDiscoveryOnSIGHUP() + keepclient.DefaultBlockCache.MaxBlocks = h.Config.cluster.Collections.WebDAVCache.MaxBlockEntries h.healthHandler = &health.Handler{ - Token: h.Config.ManagementToken, + Token: h.Config.cluster.ManagementToken, Prefix: "/_health/", } @@ -89,55 +92,106 @@ func (h *handler) setup() { } func (h *handler) serveStatus(w http.ResponseWriter, r *http.Request) { - status := struct { - cacheStats - Version string - }{ - cacheStats: h.Config.Cache.Stats(), - Version: arvadosVersion.GetVersion(), - } - json.NewEncoder(w).Encode(status) + json.NewEncoder(w).Encode(struct{ Version string }{version}) +} + +// updateOnSuccess wraps httpserver.ResponseWriter. If the handler +// sends an HTTP header indicating success, updateOnSuccess first +// calls the provided update func. If the update func fails, a 500 +// response is sent, and the status code and body sent by the handler +// are ignored (all response writes return the update error). +type updateOnSuccess struct { + httpserver.ResponseWriter + logger logrus.FieldLogger + update func() error + sentHeader bool + err error +} + +func (uos *updateOnSuccess) Write(p []byte) (int, error) { + if !uos.sentHeader { + uos.WriteHeader(http.StatusOK) + } + if uos.err != nil { + return 0, uos.err + } + return uos.ResponseWriter.Write(p) +} + +func (uos *updateOnSuccess) WriteHeader(code int) { + if !uos.sentHeader { + uos.sentHeader = true + if code >= 200 && code < 400 { + if uos.err = uos.update(); uos.err != nil { + code := http.StatusInternalServerError + if err, ok := uos.err.(*arvados.TransactionError); ok { + code = err.StatusCode + } + uos.logger.WithError(uos.err).Errorf("update() returned error type %T, changing response to HTTP %d", uos.err, code) + http.Error(uos.ResponseWriter, uos.err.Error(), code) + return + } + } + } + uos.ResponseWriter.WriteHeader(code) } var ( + corsAllowHeadersHeader = strings.Join([]string{ + "Authorization", "Content-Type", "Range", + // WebDAV request headers: + "Depth", "Destination", "If", "Lock-Token", "Overwrite", "Timeout", + }, ", ") + writeMethod = map[string]bool{ + "COPY": true, + "DELETE": true, + "LOCK": true, + "MKCOL": true, + "MOVE": true, + "PROPPATCH": true, + "PUT": true, + "RMCOL": true, + "UNLOCK": true, + } webdavMethod = map[string]bool{ - "OPTIONS": true, - "PROPFIND": true, + "COPY": true, + "DELETE": true, + "LOCK": true, + "MKCOL": true, + "MOVE": true, + "OPTIONS": true, + "PROPFIND": true, + "PROPPATCH": true, + "PUT": true, + "RMCOL": true, + "UNLOCK": true, } browserMethod = map[string]bool{ "GET": true, "HEAD": true, "POST": true, } + // top-level dirs to serve with siteFS + siteFSDir = map[string]bool{ + "": true, // root directory + "by_id": true, + "users": true, + } ) // ServeHTTP implements http.Handler. func (h *handler) ServeHTTP(wOrig http.ResponseWriter, r *http.Request) { h.setupOnce.Do(h.setup) - var statusCode = 0 - var statusText string - 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) - defer func() { - if statusCode == 0 { - statusCode = w.WroteStatus() - } else if w.WroteStatus() == 0 { - w.WriteHeader(statusCode) - } else if w.WroteStatus() != statusCode { - httpserver.Log(r.RemoteAddr, "WARNING", - fmt.Sprintf("Our status changed from %d to %d after we sent headers", w.WroteStatus(), statusCode)) - } - if statusText == "" { - statusText = http.StatusText(statusCode) - } - httpserver.Log(remoteAddr, statusCode, statusText, w.WroteBodyBytes(), r.Method, r.Host, r.URL.Path, r.URL.RawQuery) - }() if strings.HasPrefix(r.URL.Path, "/_health/") && r.Method == "GET" { h.healthHandler.ServeHTTP(w, r) @@ -146,19 +200,18 @@ func (h *handler) ServeHTTP(wOrig http.ResponseWriter, r *http.Request) { if method := r.Header.Get("Access-Control-Request-Method"); method != "" && r.Method == "OPTIONS" { if !browserMethod[method] && !webdavMethod[method] { - statusCode = http.StatusMethodNotAllowed + w.WriteHeader(http.StatusMethodNotAllowed) return } - w.Header().Set("Access-Control-Allow-Headers", "Authorization, Content-Type, Range") - w.Header().Set("Access-Control-Allow-Methods", "GET, POST, OPTIONS, PROPFIND") + w.Header().Set("Access-Control-Allow-Headers", corsAllowHeadersHeader) + w.Header().Set("Access-Control-Allow-Methods", "COPY, DELETE, GET, LOCK, MKCOL, MOVE, OPTIONS, POST, PROPFIND, PROPPATCH, PUT, RMCOL, UNLOCK") w.Header().Set("Access-Control-Allow-Origin", "*") w.Header().Set("Access-Control-Max-Age", "86400") - statusCode = http.StatusOK return } if !browserMethod[r.Method] && !webdavMethod[r.Method] { - statusCode, statusText = http.StatusMethodNotAllowed, r.Method + w.WriteHeader(http.StatusMethodNotAllowed) return } @@ -172,60 +225,71 @@ func (h *handler) ServeHTTP(wOrig http.ResponseWriter, r *http.Request) { w.Header().Set("Access-Control-Expose-Headers", "Content-Range") } - arv := h.clientPool.Get() - if arv == nil { - statusCode, statusText = http.StatusInternalServerError, "Pool failed: "+h.clientPool.Err().Error() - return - } - defer h.clientPool.Put(arv) - pathParts := strings.Split(r.URL.Path[1:], "/") var stripParts int - var targetID string + var collectionID string var tokens []string var reqTokens []string var pathToken bool var attachment bool - credentialsOK := h.Config.TrustAllContent + var useSiteFS bool + credentialsOK := h.Config.cluster.Collections.TrustAllContent - if r.Host != "" && r.Host == h.Config.AttachmentOnlyHost { + if r.Host != "" && r.Host == h.Config.cluster.Services.WebDAVDownload.ExternalURL.Host { credentialsOK = true attachment = true } else if r.FormValue("disposition") == "attachment" { attachment = true } - if targetID = parseCollectionIDFromDNSName(r.Host); targetID != "" { + if collectionID = parseCollectionIDFromDNSName(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=") { // /c=ID[/PATH...] - targetID = parseCollectionIDFromURL(pathParts[0][2:]) + collectionID = parseCollectionIDFromURL(pathParts[0][2:]) stripParts = 1 } else if len(pathParts) >= 2 && pathParts[0] == "collections" { if len(pathParts) >= 4 && pathParts[1] == "download" { // /collections/download/ID/TOKEN/PATH... - targetID = parseCollectionIDFromURL(pathParts[2]) + collectionID = parseCollectionIDFromURL(pathParts[2]) tokens = []string{pathParts[3]} stripParts = 4 pathToken = true } else { // /collections/ID/PATH... - targetID = parseCollectionIDFromURL(pathParts[1]) - tokens = h.Config.AnonymousTokens + collectionID = parseCollectionIDFromURL(pathParts[1]) stripParts = 2 + // This path is only meant to work for public + // data. Tokens provided with the request are + // ignored. + credentialsOK = false } } - if targetID == "" { - statusCode = http.StatusNotFound + if collectionID == "" && !useSiteFS { + w.WriteHeader(http.StatusNotFound) return } + forceReload := false + if cc := r.Header.Get("Cache-Control"); strings.Contains(cc, "no-cache") || strings.Contains(cc, "must-revalidate") { + forceReload = true + } + + if credentialsOK { + reqTokens = auth.CredentialsFromRequest(r).Tokens + } + 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 @@ -241,7 +305,7 @@ func (h *handler) ServeHTTP(wOrig http.ResponseWriter, r *http.Request) { // // * The token isn't embedded in the URL, so we don't // need to worry about bookmarks and copy/paste. - tokens = append(tokens, formToken) + 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 @@ -252,6 +316,11 @@ func (h *handler) ServeHTTP(wOrig http.ResponseWriter, r *http.Request) { return } + if useSiteFS { + h.serveSiteFS(w, r, reqTokens, credentialsOK, attachment) + return + } + targetPath := pathParts[stripParts:] if tokens == nil && len(targetPath) > 0 && strings.HasPrefix(targetPath[0], "t=") { // http://ID.example/t=TOKEN/PATH... @@ -267,10 +336,7 @@ func (h *handler) ServeHTTP(wOrig http.ResponseWriter, r *http.Request) { } if tokens == nil { - if credentialsOK { - reqTokens = auth.NewCredentialsFromHTTPRequest(r).Tokens - } - tokens = append(reqTokens, h.Config.AnonymousTokens...) + tokens = append(reqTokens, h.Config.cluster.Users.AnonymousUserToken) } if len(targetPath) > 0 && targetPath[0] == "_" { @@ -284,16 +350,18 @@ func (h *handler) ServeHTTP(wOrig http.ResponseWriter, r *http.Request) { stripParts++ } - forceReload := false - if cc := r.Header.Get("Cache-Control"); strings.Contains(cc, "no-cache") || strings.Contains(cc, "must-revalidate") { - forceReload = true + arv := h.clientPool.Get() + if arv == nil { + http.Error(w, "client pool error: "+h.clientPool.Err().Error(), http.StatusInternalServerError) + return } + defer h.clientPool.Put(arv) var collection *arvados.Collection tokenResult := make(map[string]int) for _, arv.ApiToken = range tokens { var err error - collection, err = h.Config.Cache.Get(arv, targetID, forceReload) + collection, err = h.Config.Cache.Get(arv, collectionID, forceReload) if err == nil { // Success break @@ -308,7 +376,7 @@ func (h *handler) ServeHTTP(wOrig http.ResponseWriter, r *http.Request) { } } // Something more serious is wrong - statusCode, statusText = http.StatusInternalServerError, err.Error() + http.Error(w, "cache error: "+err.Error(), http.StatusInternalServerError) return } if collection == nil { @@ -318,14 +386,14 @@ 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. - statusCode = http.StatusNotFound + w.WriteHeader(http.StatusNotFound) return } for _, t := range reqTokens { if tokenResult[t] == 404 { // The client provided valid token(s), but the // collection was not found. - statusCode = http.StatusNotFound + w.WriteHeader(http.StatusNotFound) return } } @@ -339,15 +407,16 @@ 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\"") - statusCode = http.StatusUnauthorized + w.WriteHeader(http.StatusUnauthorized) return } kc, err := keepclient.MakeKeepClient(arv) if err != nil { - statusCode, statusText = http.StatusInternalServerError, err.Error() + http.Error(w, "error setting up keep client: "+err.Error(), http.StatusInternalServerError) return } + kc.RequestID = r.Header.Get("X-Request-Id") var basename string if len(targetPath) > 0 { @@ -355,21 +424,49 @@ func (h *handler) ServeHTTP(wOrig http.ResponseWriter, r *http.Request) { } applyContentDispositionHdr(w, r, basename, attachment) - fs := collection.FileSystem(&arvados.Client{ + client := (&arvados.Client{ APIHost: arv.ApiServer, AuthToken: arv.ApiToken, Insecure: arv.ApiInsecure, - }, kc) + }).WithRequestID(r.Header.Get("X-Request-Id")) + + fs, err := collection.FileSystem(client, kc) + if err != nil { + http.Error(w, "error creating collection filesystem: "+err.Error(), http.StatusInternalServerError) + return + } + + writefs, writeOK := fs.(arvados.CollectionFileSystem) + targetIsPDH := arvadosclient.PDHMatch(collectionID) + if (targetIsPDH || !writeOK) && writeMethod[r.Method] { + http.Error(w, errReadOnly.Error(), http.StatusMethodNotAllowed) + return + } + if webdavMethod[r.Method] { + if writeMethod[r.Method] { + // Save the collection only if/when all + // webdav->filesystem operations succeed -- + // and send a 500 error if the modified + // collection can't be saved. + w = &updateOnSuccess{ + ResponseWriter: w, + logger: ctxlog.FromContext(r.Context()), + update: func() error { + return h.Config.Cache.Update(client, *collection, writefs) + }} + } h := webdav.Handler{ - Prefix: "/" + strings.Join(pathParts[:stripParts], "/"), - FileSystem: &webdavFS{collfs: fs}, + Prefix: "/" + strings.Join(pathParts[:stripParts], "/"), + FileSystem: &webdavFS{ + collfs: fs, + writing: writeMethod[r.Method], + alwaysReadEOF: r.Method == "PROPFIND", + }, LockSystem: h.webdavLS, Logger: func(_ *http.Request, err error) { - if os.IsNotExist(err) { - statusCode, statusText = http.StatusNotFound, err.Error() - } else if err != nil { - statusCode, statusText = http.StatusInternalServerError, err.Error() + if err != nil { + ctxlog.FromContext(r.Context()).WithError(err).Error("error reported by webdav handler") } }, } @@ -380,13 +477,13 @@ func (h *handler) ServeHTTP(wOrig http.ResponseWriter, r *http.Request) { openPath := "/" + strings.Join(targetPath, "/") if f, err := fs.Open(openPath); os.IsNotExist(err) { // Requested non-existent path - statusCode = http.StatusNotFound + w.WriteHeader(http.StatusNotFound) } else if err != nil { // Some other (unexpected) error - statusCode, statusText = http.StatusInternalServerError, err.Error() + http.Error(w, "open: "+err.Error(), http.StatusInternalServerError) } else if stat, err := f.Stat(); err != nil { // Can't get Size/IsDir (shouldn't happen with a collectionFS!) - statusCode, statusText = http.StatusInternalServerError, err.Error() + http.Error(w, "stat: "+err.Error(), http.StatusInternalServerError) } else if stat.IsDir() && !strings.HasSuffix(r.URL.Path, "/") { // If client requests ".../dirname", redirect to // ".../dirname/". This way, relative links in the @@ -394,26 +491,95 @@ func (h *handler) ServeHTTP(wOrig http.ResponseWriter, r *http.Request) { // "dirname/fnm". h.seeOtherWithCookie(w, r, r.URL.Path+"/", credentialsOK) } else if stat.IsDir() { - h.serveDirectory(w, r, collection.Name, fs, openPath, stripParts) + h.serveDirectory(w, r, collection.Name, fs, openPath, true) } else { http.ServeContent(w, r, basename, stat.ModTime(), f) - if r.Header.Get("Range") == "" && int64(w.WroteBodyBytes()) != stat.Size() { + if wrote := int64(w.WroteBodyBytes()); wrote != stat.Size() && r.Header.Get("Range") == "" { // 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)) - statusCode, statusText = http.StatusInternalServerError, fmt.Sprintf("f.Size()==%d but only wrote %d bytes; read(1024) returns %d, %s", stat.Size(), w.WroteBodyBytes(), n, err) + ctxlog.FromContext(r.Context()).Errorf("stat.Size()==%d but only wrote %d bytes; read(1024) returns %d, %s", stat.Size(), wrote, n, err) } } } +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) + 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) + if err != nil { + http.Error(w, "error setting up keep client: "+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) + f, err := fs.Open(r.URL.Path) + if os.IsNotExist(err) { + http.Error(w, err.Error(), http.StatusNotFound) + return + } else if err != nil { + http.Error(w, err.Error(), http.StatusInternalServerError) + return + } + defer f.Close() + if fi, err := f.Stat(); err == nil && fi.IsDir() && r.Method == "GET" { + if !strings.HasSuffix(r.URL.Path, "/") { + h.seeOtherWithCookie(w, r, r.URL.Path+"/", credentialsOK) + } else { + h.serveDirectory(w, r, fi.Name(), fs, r.URL.Path, false) + } + return + } + if r.Method == "GET" { + _, basename := filepath.Split(r.URL.Path) + applyContentDispositionHdr(w, r, basename, attachment) + } + wh := webdav.Handler{ + Prefix: "/", + FileSystem: &webdavFS{ + collfs: fs, + writing: writeMethod[r.Method], + alwaysReadEOF: r.Method == "PROPFIND", + }, + LockSystem: h.webdavLS, + Logger: func(_ *http.Request, err error) { + if err != nil { + ctxlog.FromContext(r.Context()).WithError(err).Error("error reported by webdav handler") + } + }, + } + wh.ServeHTTP(w, r) +} + var dirListingTemplate = ` - {{ .Collection.Name }} + {{ .CollectionName }} +

{{ .CollectionName }}

This collection of data files is being shared with you through Arvados. You can download individual files listed below. To download -the entire collection with wget, try:

+the entire directory tree with wget, try:

-
$ wget --mirror --no-parent --no-host --cut-dirs={{ .StripParts }} https://{{ .Request.Host }}{{ .Request.URL }}
+
$ wget --mirror --no-parent --no-host --cut-dirs={{ .StripParts }} https://{{ .Request.Host }}{{ .Request.URL.Path }}

File Listing

{{if .Files}} {{else}}

(No files; this collection is empty.)

@@ -469,11 +642,12 @@ the entire collection with wget, try:

` type fileListEnt struct { - Name string - Size int64 + Name string + Size int64 + IsDir bool } -func (h *handler) serveDirectory(w http.ResponseWriter, r *http.Request, collectionName string, fs http.FileSystem, base string, stripParts int) { +func (h *handler) serveDirectory(w http.ResponseWriter, r *http.Request, collectionName string, fs http.FileSystem, base string, recurse bool) { var files []fileListEnt var walk func(string) error if !strings.HasSuffix(base, "/") { @@ -493,22 +667,23 @@ func (h *handler) serveDirectory(w http.ResponseWriter, r *http.Request, collect return err } for _, ent := range ents { - if ent.IsDir() { + if recurse && ent.IsDir() { err = walk(path + ent.Name() + "/") if err != nil { return err } } else { files = append(files, fileListEnt{ - Name: path + ent.Name(), - Size: ent.Size(), + Name: path + ent.Name(), + Size: ent.Size(), + IsDir: ent.IsDir(), }) } } return nil } if err := walk(""); err != nil { - http.Error(w, err.Error(), http.StatusInternalServerError) + http.Error(w, "error getting directory listing: "+err.Error(), http.StatusInternalServerError) return } @@ -519,7 +694,7 @@ func (h *handler) serveDirectory(w http.ResponseWriter, r *http.Request, collect } tmpl, err := template.New("dir").Funcs(funcs).Parse(dirListingTemplate) if err != nil { - http.Error(w, err.Error(), http.StatusInternalServerError) + http.Error(w, "error parsing template: "+err.Error(), http.StatusInternalServerError) return } sort.Slice(files, func(i, j int) bool { @@ -530,7 +705,7 @@ func (h *handler) serveDirectory(w http.ResponseWriter, r *http.Request, collect "CollectionName": collectionName, "Files": files, "Request": r, - "StripParts": stripParts, + "StripParts": strings.Count(strings.TrimRight(r.URL.Path, "/"), "/"), }) } @@ -559,7 +734,7 @@ func (h *handler) seeOtherWithCookie(w http.ResponseWriter, r *http.Request, loc // into a cookie unless the current vhost // (origin) serves only a single collection or // we are in TrustAllContent mode. - w.WriteHeader(http.StatusBadRequest) + http.Error(w, "cannot serve inline content at this URL (possible configuration error; see https://doc.arvados.org/install/install-keep-web.html#dns)", http.StatusBadRequest) return } @@ -591,12 +766,13 @@ func (h *handler) seeOtherWithCookie(w http.ResponseWriter, r *http.Request, loc if location != "" { newu, err := u.Parse(location) if err != nil { - w.WriteHeader(http.StatusInternalServerError) + http.Error(w, "error resolving redirect target: "+err.Error(), http.StatusInternalServerError) return } u = newu } redir := (&url.URL{ + Scheme: r.URL.Scheme, Host: r.Host, Path: u.Path, RawQuery: redirQuery.Encode(),