"net/http"
"net/url"
"os"
+ "path/filepath"
"sort"
"strconv"
"strings"
"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"
+ log "github.com/Sirupsen/logrus"
+ "golang.org/x/net/webdav"
)
type handler struct {
- Config *Config
- clientPool *arvadosclient.ClientPool
- setupOnce sync.Once
+ Config *Config
+ clientPool *arvadosclient.ClientPool
+ setupOnce sync.Once
+ healthHandler http.Handler
+ webdavLS webdav.LockSystem
}
// parseCollectionIDFromDNSName returns a UUID or PDH if s begins with
func (h *handler) setup() {
h.clientPool = arvadosclient.MakeClientPool()
+
keepclient.RefreshServiceDiscoveryOnSIGHUP()
+
+ h.healthHandler = &health.Handler{
+ Token: h.Config.ManagementToken,
+ Prefix: "/_health/",
+ }
+
+ // Even though we don't accept LOCK requests, every webdav
+ // handler must have a non-nil LockSystem.
+ h.webdavLS = &noLockSystem{}
}
func (h *handler) serveStatus(w http.ResponseWriter, r *http.Request) {
status := struct {
cacheStats
+ Version string
}{
cacheStats: h.Config.Cache.Stats(),
+ Version: version,
}
json.NewEncoder(w).Encode(status)
}
+// 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
+ 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
+ }
+ log.Printf("update() changes response to HTTP %d: %T %q", code, uos.err, uos.err)
+ http.Error(uos.ResponseWriter, uos.err.Error(), code)
+ return
+ }
+ }
+ }
+ uos.ResponseWriter.WriteHeader(code)
+}
+
+var (
+ writeMethod = map[string]bool{
+ "COPY": true,
+ "DELETE": true,
+ "MKCOL": true,
+ "MOVE": true,
+ "PUT": true,
+ "RMCOL": true,
+ }
+ webdavMethod = map[string]bool{
+ "COPY": true,
+ "DELETE": true,
+ "MKCOL": true,
+ "MOVE": true,
+ "OPTIONS": true,
+ "PROPFIND": true,
+ "PUT": true,
+ "RMCOL": 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)
} else if w.WroteStatus() == 0 {
w.WriteHeader(statusCode)
} else if w.WroteStatus() != statusCode {
- httpserver.Log(r.RemoteAddr, "WARNING",
+ log.WithField("RequestID", r.Header.Get("X-Request-Id")).Warn(
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 r.Method == "OPTIONS" {
- method := r.Header.Get("Access-Control-Request-Method")
- if method != "GET" && method != "POST" {
+ 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] {
statusCode = http.StatusMethodNotAllowed
return
}
- w.Header().Set("Access-Control-Allow-Headers", "Range")
- w.Header().Set("Access-Control-Allow-Methods", "GET, POST")
+ w.Header().Set("Access-Control-Allow-Headers", "Authorization, Content-Type, Range")
+ w.Header().Set("Access-Control-Allow-Methods", "COPY, DELETE, GET, MKCOL, MOVE, OPTIONS, POST, PROPFIND, PUT, RMCOL")
w.Header().Set("Access-Control-Allow-Origin", "*")
w.Header().Set("Access-Control-Max-Age", "86400")
statusCode = http.StatusOK
return
}
- if r.Method != "GET" && r.Method != "POST" {
+ if !browserMethod[r.Method] && !webdavMethod[r.Method] {
statusCode, statusText = http.StatusMethodNotAllowed, r.Method
return
}
// SSL certificates. See
// http://www.w3.org/TR/cors/#user-credentials).
w.Header().Set("Access-Control-Allow-Origin", "*")
+ 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
+ var useSiteFS bool
credentialsOK := h.Config.TrustAllContent
if r.Host != "" && r.Host == h.Config.AttachmentOnlyHost {
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 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])
+ collectionID = parseCollectionIDFromURL(pathParts[1])
tokens = h.Config.AnonymousTokens
stripParts = 2
}
}
- if targetID == "" {
+ if collectionID == "" && !useSiteFS {
statusCode = http.StatusNotFound
return
}
+ forceReload := false
+ if cc := r.Header.Get("Cache-Control"); strings.Contains(cc, "no-cache") || strings.Contains(cc, "must-revalidate") {
+ forceReload = true
+ }
+
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
// * The token isn't embedded in the URL, so we don't
// need to worry about bookmarks and copy/paste.
tokens = append(tokens, formToken)
- } else if 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
return
}
+ if useSiteFS {
+ if tokens == nil {
+ tokens = auth.NewCredentialsFromHTTPRequest(r).Tokens
+ }
+ h.serveSiteFS(w, r, tokens, credentialsOK, attachment)
+ return
+ }
+
targetPath := pathParts[stripParts:]
if tokens == nil && len(targetPath) > 0 && strings.HasPrefix(targetPath[0], "t=") {
// http://ID.example/t=TOKEN/PATH...
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 {
+ statusCode, statusText = http.StatusInternalServerError, "Pool failed: "+h.clientPool.Err().Error()
+ 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
statusCode, statusText = http.StatusInternalServerError, err.Error()
return
}
+ kc.RequestID = r.Header.Get("X-Request-Id")
- basename := targetPath[len(targetPath)-1]
+ var basename string
+ if len(targetPath) > 0 {
+ basename = targetPath[len(targetPath)-1]
+ }
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 {
+ statusCode, statusText = http.StatusInternalServerError, err.Error()
+ return
+ }
+
+ writefs, writeOK := fs.(arvados.CollectionFileSystem)
+ targetIsPDH := arvadosclient.PDHMatch(collectionID)
+ if (targetIsPDH || !writeOK) && writeMethod[r.Method] {
+ statusCode, statusText = http.StatusMethodNotAllowed, errReadOnly.Error()
+ 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,
+ update: func() error {
+ return h.Config.Cache.Update(client, *collection, writefs)
+ }}
+ }
+ h := webdav.Handler{
+ 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 err != nil {
+ log.Printf("error from webdav handler: %q", err)
+ }
+ },
+ }
+ h.ServeHTTP(w, r)
+ return
+ }
+
openPath := "/" + strings.Join(targetPath, "/")
if f, err := fs.Open(openPath); os.IsNotExist(err) {
// Requested non-existent path
// ".../dirname/". This way, relative links in the
// listing for "dirname" can always be "fnm", never
// "dirname/fnm".
- h.seeOtherWithCookie(w, r, basename+"/", credentialsOK)
+ 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() {
}
}
+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, 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)
+ 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 {
+ log.Printf("error from webdav handler: %q", err)
+ }
+ },
+ }
+ wh.ServeHTTP(w, r)
+}
+
var dirListingTemplate = `<!DOCTYPE HTML>
<HTML><HEAD>
<META name="robots" content="NOINDEX">
- <TITLE>{{ .Collection.Name }}</TITLE>
+ <TITLE>{{ .CollectionName }}</TITLE>
<STYLE type="text/css">
body {
margin: 1.5em;
</STYLE>
</HEAD>
<BODY>
+
<H1>{{ .CollectionName }}</H1>
<P>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:</P>
+the entire directory tree with wget, try:</P>
-<PRE>$ wget --mirror --no-parent --no-host --cut-dirs={{ .StripParts }} https://{{ .Request.Host }}{{ .Request.URL }}</PRE>
+<PRE>$ wget --mirror --no-parent --no-host --cut-dirs={{ .StripParts }} https://{{ .Request.Host }}{{ .Request.URL.Path }}</PRE>
<H2>File Listing</H2>
{{if .Files}}
<UL>
-{{range .Files}} <LI>{{.Size | printf "%15d " | nbsp}}<A href="{{.Name}}">{{.Name}}</A></LI>{{end}}
+{{range .Files}}
+{{if .IsDir }}
+ <LI>{{" " | printf "%15s " | nbsp}}<A href="{{print "./" .Name}}/">{{.Name}}/</A></LI>
+{{else}}
+ <LI>{{.Size | printf "%15d " | nbsp}}<A href="{{print "./" .Name}}">{{.Name}}</A></LI>
+{{end}}
+{{end}}
</UL>
{{else}}
<P>(No files; this collection is empty.)</P>
`
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, "/") {
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(),
})
}
}
"CollectionName": collectionName,
"Files": files,
"Request": r,
- "StripParts": stripParts,
+ "StripParts": strings.Count(strings.TrimRight(r.URL.Path, "/"), "/"),
})
}
}
func (h *handler) seeOtherWithCookie(w http.ResponseWriter, r *http.Request, location string, credentialsOK bool) {
- if !credentialsOK {
- // It is not safe to copy the provided token
- // into a cookie unless the current vhost
- // (origin) serves only a single collection or
- // we are in TrustAllContent mode.
- w.WriteHeader(http.StatusBadRequest)
- return
- }
-
if formToken := r.FormValue("api_token"); formToken != "" {
+ if !credentialsOK {
+ // It is not safe to copy the provided token
+ // into a cookie unless the current vhost
+ // (origin) serves only a single collection or
+ // we are in TrustAllContent mode.
+ w.WriteHeader(http.StatusBadRequest)
+ return
+ }
+
// The HttpOnly flag is necessary to prevent
// JavaScript code (included in, or loaded by, a page
// in the collection being served) from employing the
// bar, and in the case of a POST request to avoid
// raising warnings when the user refreshes the
// resulting page.
-
http.SetCookie(w, &http.Cookie{
Name: "arvados_api_token",
Value: auth.EncodeTokenCookie([]byte(formToken)),