diff options
author | GitLab Bot <gitlab-bot@gitlab.com> | 2020-12-17 11:59:07 +0000 |
---|---|---|
committer | GitLab Bot <gitlab-bot@gitlab.com> | 2020-12-17 11:59:07 +0000 |
commit | 8b573c94895dc0ac0e1d9d59cf3e8745e8b539ca (patch) | |
tree | 544930fb309b30317ae9797a9683768705d664c4 /workhorse/internal/artifacts | |
parent | 4b1de649d0168371549608993deac953eb692019 (diff) | |
download | gitlab-ce-8b573c94895dc0ac0e1d9d59cf3e8745e8b539ca.tar.gz |
Add latest changes from gitlab-org/gitlab@13-7-stable-eev13.7.0-rc42
Diffstat (limited to 'workhorse/internal/artifacts')
-rw-r--r-- | workhorse/internal/artifacts/artifacts_store_test.go | 338 | ||||
-rw-r--r-- | workhorse/internal/artifacts/artifacts_test.go | 19 | ||||
-rw-r--r-- | workhorse/internal/artifacts/artifacts_upload.go | 167 | ||||
-rw-r--r-- | workhorse/internal/artifacts/artifacts_upload_test.go | 322 | ||||
-rw-r--r-- | workhorse/internal/artifacts/entry.go | 123 | ||||
-rw-r--r-- | workhorse/internal/artifacts/entry_test.go | 134 | ||||
-rw-r--r-- | workhorse/internal/artifacts/escape_quotes.go | 10 |
7 files changed, 1113 insertions, 0 deletions
diff --git a/workhorse/internal/artifacts/artifacts_store_test.go b/workhorse/internal/artifacts/artifacts_store_test.go new file mode 100644 index 00000000000..bd56d9ea725 --- /dev/null +++ b/workhorse/internal/artifacts/artifacts_store_test.go @@ -0,0 +1,338 @@ +package artifacts + +import ( + "archive/zip" + "bytes" + "crypto/md5" + "encoding/hex" + "fmt" + "io/ioutil" + "mime/multipart" + "net/http" + "net/http/httptest" + "os" + "testing" + "time" + + "github.com/stretchr/testify/require" + + "gitlab.com/gitlab-org/gitlab-workhorse/internal/api" + "gitlab.com/gitlab-org/gitlab-workhorse/internal/objectstore/test" + "gitlab.com/gitlab-org/gitlab-workhorse/internal/testhelper" +) + +func createTestZipArchive(t *testing.T) (data []byte, md5Hash string) { + var buffer bytes.Buffer + archive := zip.NewWriter(&buffer) + fileInArchive, err := archive.Create("test.file") + require.NoError(t, err) + fmt.Fprint(fileInArchive, "test") + archive.Close() + data = buffer.Bytes() + + hasher := md5.New() + hasher.Write(data) + hexHash := hasher.Sum(nil) + md5Hash = hex.EncodeToString(hexHash) + + return data, md5Hash +} + +func createTestMultipartForm(t *testing.T, data []byte) (bytes.Buffer, string) { + var buffer bytes.Buffer + writer := multipart.NewWriter(&buffer) + file, err := writer.CreateFormFile("file", "my.file") + require.NoError(t, err) + file.Write(data) + writer.Close() + return buffer, writer.FormDataContentType() +} + +func testUploadArtifactsFromTestZip(t *testing.T, ts *httptest.Server) *httptest.ResponseRecorder { + archiveData, _ := createTestZipArchive(t) + contentBuffer, contentType := createTestMultipartForm(t, archiveData) + + return testUploadArtifacts(t, contentType, ts.URL+Path, &contentBuffer) +} + +func TestUploadHandlerSendingToExternalStorage(t *testing.T) { + tempPath, err := ioutil.TempDir("", "uploads") + if err != nil { + t.Fatal(err) + } + defer os.RemoveAll(tempPath) + + archiveData, md5 := createTestZipArchive(t) + archiveFile, err := ioutil.TempFile("", "artifact.zip") + require.NoError(t, err) + defer os.Remove(archiveFile.Name()) + _, err = archiveFile.Write(archiveData) + require.NoError(t, err) + archiveFile.Close() + + storeServerCalled := 0 + storeServerMux := http.NewServeMux() + storeServerMux.HandleFunc("/url/put", func(w http.ResponseWriter, r *http.Request) { + require.Equal(t, "PUT", r.Method) + + receivedData, err := ioutil.ReadAll(r.Body) + require.NoError(t, err) + require.Equal(t, archiveData, receivedData) + + storeServerCalled++ + w.Header().Set("ETag", md5) + w.WriteHeader(200) + }) + storeServerMux.HandleFunc("/store-id", func(w http.ResponseWriter, r *http.Request) { + http.ServeFile(w, r, archiveFile.Name()) + }) + + responseProcessorCalled := 0 + responseProcessor := func(w http.ResponseWriter, r *http.Request) { + require.Equal(t, "store-id", r.FormValue("file.remote_id")) + require.NotEmpty(t, r.FormValue("file.remote_url")) + w.WriteHeader(200) + responseProcessorCalled++ + } + + storeServer := httptest.NewServer(storeServerMux) + defer storeServer.Close() + + qs := fmt.Sprintf("?%s=%s", ArtifactFormatKey, ArtifactFormatZip) + + tests := []struct { + name string + preauth api.Response + }{ + { + name: "ObjectStore Upload", + preauth: api.Response{ + RemoteObject: api.RemoteObject{ + StoreURL: storeServer.URL + "/url/put" + qs, + ID: "store-id", + GetURL: storeServer.URL + "/store-id", + }, + }, + }, + } + + for _, test := range tests { + t.Run(test.name, func(t *testing.T) { + storeServerCalled = 0 + responseProcessorCalled = 0 + + ts := testArtifactsUploadServer(t, test.preauth, responseProcessor) + defer ts.Close() + + contentBuffer, contentType := createTestMultipartForm(t, archiveData) + response := testUploadArtifacts(t, contentType, ts.URL+Path+qs, &contentBuffer) + require.Equal(t, http.StatusOK, response.Code) + testhelper.RequireResponseHeader(t, response, MetadataHeaderKey, MetadataHeaderPresent) + require.Equal(t, 1, storeServerCalled, "store should be called only once") + require.Equal(t, 1, responseProcessorCalled, "response processor should be called only once") + }) + } +} + +func TestUploadHandlerSendingToExternalStorageAndStorageServerUnreachable(t *testing.T) { + tempPath, err := ioutil.TempDir("", "uploads") + if err != nil { + t.Fatal(err) + } + defer os.RemoveAll(tempPath) + + responseProcessor := func(w http.ResponseWriter, r *http.Request) { + t.Fatal("it should not be called") + } + + authResponse := api.Response{ + TempPath: tempPath, + RemoteObject: api.RemoteObject{ + StoreURL: "http://localhost:12323/invalid/url", + ID: "store-id", + }, + } + + ts := testArtifactsUploadServer(t, authResponse, responseProcessor) + defer ts.Close() + + response := testUploadArtifactsFromTestZip(t, ts) + require.Equal(t, http.StatusInternalServerError, response.Code) +} + +func TestUploadHandlerSendingToExternalStorageAndInvalidURLIsUsed(t *testing.T) { + tempPath, err := ioutil.TempDir("", "uploads") + if err != nil { + t.Fatal(err) + } + defer os.RemoveAll(tempPath) + + responseProcessor := func(w http.ResponseWriter, r *http.Request) { + t.Fatal("it should not be called") + } + + authResponse := api.Response{ + TempPath: tempPath, + RemoteObject: api.RemoteObject{ + StoreURL: "htt:////invalid-url", + ID: "store-id", + }, + } + + ts := testArtifactsUploadServer(t, authResponse, responseProcessor) + defer ts.Close() + + response := testUploadArtifactsFromTestZip(t, ts) + require.Equal(t, http.StatusInternalServerError, response.Code) +} + +func TestUploadHandlerSendingToExternalStorageAndItReturnsAnError(t *testing.T) { + putCalledTimes := 0 + + storeServerMux := http.NewServeMux() + storeServerMux.HandleFunc("/url/put", func(w http.ResponseWriter, r *http.Request) { + putCalledTimes++ + require.Equal(t, "PUT", r.Method) + w.WriteHeader(510) + }) + + responseProcessor := func(w http.ResponseWriter, r *http.Request) { + t.Fatal("it should not be called") + } + + storeServer := httptest.NewServer(storeServerMux) + defer storeServer.Close() + + authResponse := api.Response{ + RemoteObject: api.RemoteObject{ + StoreURL: storeServer.URL + "/url/put", + ID: "store-id", + }, + } + + ts := testArtifactsUploadServer(t, authResponse, responseProcessor) + defer ts.Close() + + response := testUploadArtifactsFromTestZip(t, ts) + require.Equal(t, http.StatusInternalServerError, response.Code) + require.Equal(t, 1, putCalledTimes, "upload should be called only once") +} + +func TestUploadHandlerSendingToExternalStorageAndSupportRequestTimeout(t *testing.T) { + putCalledTimes := 0 + + storeServerMux := http.NewServeMux() + storeServerMux.HandleFunc("/url/put", func(w http.ResponseWriter, r *http.Request) { + putCalledTimes++ + require.Equal(t, "PUT", r.Method) + time.Sleep(10 * time.Second) + w.WriteHeader(510) + }) + + responseProcessor := func(w http.ResponseWriter, r *http.Request) { + t.Fatal("it should not be called") + } + + storeServer := httptest.NewServer(storeServerMux) + defer storeServer.Close() + + authResponse := api.Response{ + RemoteObject: api.RemoteObject{ + StoreURL: storeServer.URL + "/url/put", + ID: "store-id", + Timeout: 1, + }, + } + + ts := testArtifactsUploadServer(t, authResponse, responseProcessor) + defer ts.Close() + + response := testUploadArtifactsFromTestZip(t, ts) + require.Equal(t, http.StatusInternalServerError, response.Code) + require.Equal(t, 1, putCalledTimes, "upload should be called only once") +} + +func TestUploadHandlerMultipartUploadSizeLimit(t *testing.T) { + os, server := test.StartObjectStore() + defer server.Close() + + err := os.InitiateMultipartUpload(test.ObjectPath) + require.NoError(t, err) + + objectURL := server.URL + test.ObjectPath + + uploadSize := 10 + preauth := api.Response{ + RemoteObject: api.RemoteObject{ + ID: "store-id", + MultipartUpload: &api.MultipartUploadParams{ + PartSize: 1, + PartURLs: []string{objectURL + "?partNumber=1"}, + AbortURL: objectURL, // DELETE + CompleteURL: objectURL, // POST + }, + }, + } + + responseProcessor := func(w http.ResponseWriter, r *http.Request) { + t.Fatal("it should not be called") + } + + ts := testArtifactsUploadServer(t, preauth, responseProcessor) + defer ts.Close() + + contentBuffer, contentType := createTestMultipartForm(t, make([]byte, uploadSize)) + response := testUploadArtifacts(t, contentType, ts.URL+Path, &contentBuffer) + require.Equal(t, http.StatusRequestEntityTooLarge, response.Code) + + // Poll because AbortMultipartUpload is async + for i := 0; os.IsMultipartUpload(test.ObjectPath) && i < 100; i++ { + time.Sleep(10 * time.Millisecond) + } + require.False(t, os.IsMultipartUpload(test.ObjectPath), "MultipartUpload should not be in progress anymore") + require.Empty(t, os.GetObjectMD5(test.ObjectPath), "upload should have failed, so the object should not exists") +} + +func TestUploadHandlerMultipartUploadMaximumSizeFromApi(t *testing.T) { + os, server := test.StartObjectStore() + defer server.Close() + + err := os.InitiateMultipartUpload(test.ObjectPath) + require.NoError(t, err) + + objectURL := server.URL + test.ObjectPath + + uploadSize := int64(10) + maxSize := uploadSize - 1 + preauth := api.Response{ + MaximumSize: maxSize, + RemoteObject: api.RemoteObject{ + ID: "store-id", + MultipartUpload: &api.MultipartUploadParams{ + PartSize: uploadSize, + PartURLs: []string{objectURL + "?partNumber=1"}, + AbortURL: objectURL, // DELETE + CompleteURL: objectURL, // POST + }, + }, + } + + responseProcessor := func(w http.ResponseWriter, r *http.Request) { + t.Fatal("it should not be called") + } + + ts := testArtifactsUploadServer(t, preauth, responseProcessor) + defer ts.Close() + + contentBuffer, contentType := createTestMultipartForm(t, make([]byte, uploadSize)) + response := testUploadArtifacts(t, contentType, ts.URL+Path, &contentBuffer) + require.Equal(t, http.StatusRequestEntityTooLarge, response.Code) + + testhelper.Retry(t, 5*time.Second, func() error { + if os.GetObjectMD5(test.ObjectPath) == "" { + return nil + } + + return fmt.Errorf("file is still present") + }) +} diff --git a/workhorse/internal/artifacts/artifacts_test.go b/workhorse/internal/artifacts/artifacts_test.go new file mode 100644 index 00000000000..b9a42cc60c1 --- /dev/null +++ b/workhorse/internal/artifacts/artifacts_test.go @@ -0,0 +1,19 @@ +package artifacts + +import ( + "os" + "testing" + + "gitlab.com/gitlab-org/labkit/log" + + "gitlab.com/gitlab-org/gitlab-workhorse/internal/testhelper" +) + +func TestMain(m *testing.M) { + if err := testhelper.BuildExecutables(); err != nil { + log.WithError(err).Fatal() + } + + os.Exit(m.Run()) + +} diff --git a/workhorse/internal/artifacts/artifacts_upload.go b/workhorse/internal/artifacts/artifacts_upload.go new file mode 100644 index 00000000000..3d4b8bf0931 --- /dev/null +++ b/workhorse/internal/artifacts/artifacts_upload.go @@ -0,0 +1,167 @@ +package artifacts + +import ( + "context" + "fmt" + "io" + "mime/multipart" + "net/http" + "os" + "os/exec" + "strings" + "syscall" + + "github.com/prometheus/client_golang/prometheus" + "github.com/prometheus/client_golang/prometheus/promauto" + "gitlab.com/gitlab-org/labkit/log" + + "gitlab.com/gitlab-org/gitlab-workhorse/internal/api" + "gitlab.com/gitlab-org/gitlab-workhorse/internal/filestore" + "gitlab.com/gitlab-org/gitlab-workhorse/internal/helper" + "gitlab.com/gitlab-org/gitlab-workhorse/internal/upload" + "gitlab.com/gitlab-org/gitlab-workhorse/internal/zipartifacts" +) + +// Sent by the runner: https://gitlab.com/gitlab-org/gitlab-runner/blob/c24da19ecce8808d9d2950896f70c94f5ea1cc2e/network/gitlab.go#L580 +const ( + ArtifactFormatKey = "artifact_format" + ArtifactFormatZip = "zip" + ArtifactFormatDefault = "" +) + +var zipSubcommandsErrorsCounter = promauto.NewCounterVec( + prometheus.CounterOpts{ + Name: "gitlab_workhorse_zip_subcommand_errors_total", + Help: "Errors comming from subcommands used for processing ZIP archives", + }, []string{"error"}) + +type artifactsUploadProcessor struct { + opts *filestore.SaveFileOpts + format string + + upload.SavedFileTracker +} + +func (a *artifactsUploadProcessor) generateMetadataFromZip(ctx context.Context, file *filestore.FileHandler) (*filestore.FileHandler, error) { + metaReader, metaWriter := io.Pipe() + defer metaWriter.Close() + + metaOpts := &filestore.SaveFileOpts{ + LocalTempPath: a.opts.LocalTempPath, + TempFilePrefix: "metadata.gz", + } + if metaOpts.LocalTempPath == "" { + metaOpts.LocalTempPath = os.TempDir() + } + + fileName := file.LocalPath + if fileName == "" { + fileName = file.RemoteURL + } + + zipMd := exec.CommandContext(ctx, "gitlab-zip-metadata", fileName) + zipMd.Stderr = log.ContextLogger(ctx).Writer() + zipMd.SysProcAttr = &syscall.SysProcAttr{Setpgid: true} + zipMd.Stdout = metaWriter + + if err := zipMd.Start(); err != nil { + return nil, err + } + defer helper.CleanUpProcessGroup(zipMd) + + type saveResult struct { + error + *filestore.FileHandler + } + done := make(chan saveResult) + go func() { + var result saveResult + result.FileHandler, result.error = filestore.SaveFileFromReader(ctx, metaReader, -1, metaOpts) + + done <- result + }() + + if err := zipMd.Wait(); err != nil { + st, ok := helper.ExitStatus(err) + + if !ok { + return nil, err + } + + zipSubcommandsErrorsCounter.WithLabelValues(zipartifacts.ErrorLabelByCode(st)).Inc() + + if st == zipartifacts.CodeNotZip { + return nil, nil + } + + if st == zipartifacts.CodeLimitsReached { + return nil, zipartifacts.ErrBadMetadata + } + } + + metaWriter.Close() + result := <-done + return result.FileHandler, result.error +} + +func (a *artifactsUploadProcessor) ProcessFile(ctx context.Context, formName string, file *filestore.FileHandler, writer *multipart.Writer) error { + // ProcessFile for artifacts requires file form-data field name to eq `file` + + if formName != "file" { + return fmt.Errorf("invalid form field: %q", formName) + } + if a.Count() > 0 { + return fmt.Errorf("artifacts request contains more than one file") + } + a.Track(formName, file.LocalPath) + + select { + case <-ctx.Done(): + return fmt.Errorf("ProcessFile: context done") + default: + } + + if !strings.EqualFold(a.format, ArtifactFormatZip) && a.format != ArtifactFormatDefault { + return nil + } + + // TODO: can we rely on disk for shipping metadata? Not if we split workhorse and rails in 2 different PODs + metadata, err := a.generateMetadataFromZip(ctx, file) + if err != nil { + return err + } + + if metadata != nil { + fields, err := metadata.GitLabFinalizeFields("metadata") + if err != nil { + return fmt.Errorf("finalize metadata field error: %v", err) + } + + for k, v := range fields { + writer.WriteField(k, v) + } + + a.Track("metadata", metadata.LocalPath) + } + + return nil +} + +func (a *artifactsUploadProcessor) Name() string { + return "artifacts" +} + +func UploadArtifacts(myAPI *api.API, h http.Handler, p upload.Preparer) http.Handler { + return myAPI.PreAuthorizeHandler(func(w http.ResponseWriter, r *http.Request, a *api.Response) { + opts, _, err := p.Prepare(a) + if err != nil { + helper.Fail500(w, r, fmt.Errorf("UploadArtifacts: error preparing file storage options")) + return + } + + format := r.URL.Query().Get(ArtifactFormatKey) + + mg := &artifactsUploadProcessor{opts: opts, format: format, SavedFileTracker: upload.SavedFileTracker{Request: r}} + upload.HandleFileUploads(w, r, h, a, mg, opts) + }, "/authorize") +} diff --git a/workhorse/internal/artifacts/artifacts_upload_test.go b/workhorse/internal/artifacts/artifacts_upload_test.go new file mode 100644 index 00000000000..c82ae791239 --- /dev/null +++ b/workhorse/internal/artifacts/artifacts_upload_test.go @@ -0,0 +1,322 @@ +package artifacts + +import ( + "archive/zip" + "bytes" + "compress/gzip" + "encoding/json" + "fmt" + "io" + "io/ioutil" + "mime/multipart" + "net/http" + "net/http/httptest" + "os" + "testing" + + "github.com/dgrijalva/jwt-go" + + "gitlab.com/gitlab-org/gitlab-workhorse/internal/api" + "gitlab.com/gitlab-org/gitlab-workhorse/internal/filestore" + "gitlab.com/gitlab-org/gitlab-workhorse/internal/helper" + "gitlab.com/gitlab-org/gitlab-workhorse/internal/proxy" + "gitlab.com/gitlab-org/gitlab-workhorse/internal/testhelper" + "gitlab.com/gitlab-org/gitlab-workhorse/internal/upload" + "gitlab.com/gitlab-org/gitlab-workhorse/internal/upstream/roundtripper" + "gitlab.com/gitlab-org/gitlab-workhorse/internal/zipartifacts" + + "github.com/stretchr/testify/require" +) + +const ( + MetadataHeaderKey = "Metadata-Status" + MetadataHeaderPresent = "present" + MetadataHeaderMissing = "missing" + Path = "/url/path" +) + +func testArtifactsUploadServer(t *testing.T, authResponse api.Response, bodyProcessor func(w http.ResponseWriter, r *http.Request)) *httptest.Server { + mux := http.NewServeMux() + mux.HandleFunc(Path+"/authorize", func(w http.ResponseWriter, r *http.Request) { + if r.Method != "POST" { + t.Fatal("Expected POST request") + } + + w.Header().Set("Content-Type", api.ResponseContentType) + + data, err := json.Marshal(&authResponse) + if err != nil { + t.Fatal("Expected to marshal") + } + w.Write(data) + }) + mux.HandleFunc(Path, func(w http.ResponseWriter, r *http.Request) { + opts, err := filestore.GetOpts(&authResponse) + require.NoError(t, err) + + if r.Method != "POST" { + t.Fatal("Expected POST request") + } + if opts.IsLocal() { + if r.FormValue("file.path") == "" { + t.Fatal("Expected file to be present") + return + } + + _, err := ioutil.ReadFile(r.FormValue("file.path")) + if err != nil { + t.Fatal("Expected file to be readable") + return + } + } else { + if r.FormValue("file.remote_url") == "" { + t.Fatal("Expected file to be remote accessible") + return + } + } + + if r.FormValue("metadata.path") != "" { + metadata, err := ioutil.ReadFile(r.FormValue("metadata.path")) + if err != nil { + t.Fatal("Expected metadata to be readable") + return + } + gz, err := gzip.NewReader(bytes.NewReader(metadata)) + if err != nil { + t.Fatal("Expected metadata to be valid gzip") + return + } + defer gz.Close() + metadata, err = ioutil.ReadAll(gz) + if err != nil { + t.Fatal("Expected metadata to be valid") + return + } + if !bytes.HasPrefix(metadata, []byte(zipartifacts.MetadataHeaderPrefix+zipartifacts.MetadataHeader)) { + t.Fatal("Expected metadata to be of valid format") + return + } + + w.Header().Set(MetadataHeaderKey, MetadataHeaderPresent) + + } else { + w.Header().Set(MetadataHeaderKey, MetadataHeaderMissing) + } + + w.WriteHeader(http.StatusOK) + + if bodyProcessor != nil { + bodyProcessor(w, r) + } + }) + return testhelper.TestServerWithHandler(nil, mux.ServeHTTP) +} + +type testServer struct { + url string + writer *multipart.Writer + buffer *bytes.Buffer + fileWriter io.Writer + cleanup func() +} + +func setupWithTmpPath(t *testing.T, filename string, includeFormat bool, format string, authResponse *api.Response, bodyProcessor func(w http.ResponseWriter, r *http.Request)) *testServer { + tempPath, err := ioutil.TempDir("", "uploads") + require.NoError(t, err) + + if authResponse == nil { + authResponse = &api.Response{TempPath: tempPath} + } + + ts := testArtifactsUploadServer(t, *authResponse, bodyProcessor) + + var buffer bytes.Buffer + writer := multipart.NewWriter(&buffer) + fileWriter, err := writer.CreateFormFile(filename, "my.file") + require.NotNil(t, fileWriter) + require.NoError(t, err) + + cleanup := func() { + ts.Close() + require.NoError(t, os.RemoveAll(tempPath)) + require.NoError(t, writer.Close()) + } + + qs := "" + + if includeFormat { + qs = fmt.Sprintf("?%s=%s", ArtifactFormatKey, format) + } + + return &testServer{url: ts.URL + Path + qs, writer: writer, buffer: &buffer, fileWriter: fileWriter, cleanup: cleanup} +} + +func testUploadArtifacts(t *testing.T, contentType, url string, body io.Reader) *httptest.ResponseRecorder { + httpRequest, err := http.NewRequest("POST", url, body) + require.NoError(t, err) + + httpRequest.Header.Set("Content-Type", contentType) + response := httptest.NewRecorder() + parsedURL := helper.URLMustParse(url) + roundTripper := roundtripper.NewTestBackendRoundTripper(parsedURL) + testhelper.ConfigureSecret() + apiClient := api.NewAPI(parsedURL, "123", roundTripper) + proxyClient := proxy.NewProxy(parsedURL, "123", roundTripper) + UploadArtifacts(apiClient, proxyClient, &upload.DefaultPreparer{}).ServeHTTP(response, httpRequest) + return response +} + +func TestUploadHandlerAddingMetadata(t *testing.T) { + testCases := []struct { + desc string + format string + includeFormat bool + }{ + { + desc: "ZIP format", + format: ArtifactFormatZip, + includeFormat: true, + }, + { + desc: "default format", + format: ArtifactFormatDefault, + includeFormat: true, + }, + { + desc: "default format without artifact_format", + format: ArtifactFormatDefault, + includeFormat: false, + }, + } + + for _, tc := range testCases { + t.Run(tc.desc, func(t *testing.T) { + s := setupWithTmpPath(t, "file", tc.includeFormat, tc.format, nil, + func(w http.ResponseWriter, r *http.Request) { + token, err := jwt.ParseWithClaims(r.Header.Get(upload.RewrittenFieldsHeader), &upload.MultipartClaims{}, testhelper.ParseJWT) + require.NoError(t, err) + + rewrittenFields := token.Claims.(*upload.MultipartClaims).RewrittenFields + require.Equal(t, 2, len(rewrittenFields)) + + require.Contains(t, rewrittenFields, "file") + require.Contains(t, rewrittenFields, "metadata") + require.Contains(t, r.PostForm, "file.gitlab-workhorse-upload") + require.Contains(t, r.PostForm, "metadata.gitlab-workhorse-upload") + }, + ) + defer s.cleanup() + + archive := zip.NewWriter(s.fileWriter) + file, err := archive.Create("test.file") + require.NotNil(t, file) + require.NoError(t, err) + + require.NoError(t, archive.Close()) + require.NoError(t, s.writer.Close()) + + response := testUploadArtifacts(t, s.writer.FormDataContentType(), s.url, s.buffer) + require.Equal(t, http.StatusOK, response.Code) + testhelper.RequireResponseHeader(t, response, MetadataHeaderKey, MetadataHeaderPresent) + }) + } +} + +func TestUploadHandlerTarArtifact(t *testing.T) { + s := setupWithTmpPath(t, "file", true, "tar", nil, + func(w http.ResponseWriter, r *http.Request) { + token, err := jwt.ParseWithClaims(r.Header.Get(upload.RewrittenFieldsHeader), &upload.MultipartClaims{}, testhelper.ParseJWT) + require.NoError(t, err) + + rewrittenFields := token.Claims.(*upload.MultipartClaims).RewrittenFields + require.Equal(t, 1, len(rewrittenFields)) + + require.Contains(t, rewrittenFields, "file") + require.Contains(t, r.PostForm, "file.gitlab-workhorse-upload") + }, + ) + defer s.cleanup() + + file, err := os.Open("../../testdata/tarfile.tar") + require.NoError(t, err) + + _, err = io.Copy(s.fileWriter, file) + require.NoError(t, err) + require.NoError(t, file.Close()) + require.NoError(t, s.writer.Close()) + + response := testUploadArtifacts(t, s.writer.FormDataContentType(), s.url, s.buffer) + require.Equal(t, http.StatusOK, response.Code) + testhelper.RequireResponseHeader(t, response, MetadataHeaderKey, MetadataHeaderMissing) +} + +func TestUploadHandlerForUnsupportedArchive(t *testing.T) { + s := setupWithTmpPath(t, "file", true, "other", nil, nil) + defer s.cleanup() + require.NoError(t, s.writer.Close()) + + response := testUploadArtifacts(t, s.writer.FormDataContentType(), s.url, s.buffer) + require.Equal(t, http.StatusOK, response.Code) + testhelper.RequireResponseHeader(t, response, MetadataHeaderKey, MetadataHeaderMissing) +} + +func TestUploadHandlerForMultipleFiles(t *testing.T) { + s := setupWithTmpPath(t, "file", true, "", nil, nil) + defer s.cleanup() + + file, err := s.writer.CreateFormFile("file", "my.file") + require.NotNil(t, file) + require.NoError(t, err) + require.NoError(t, s.writer.Close()) + + response := testUploadArtifacts(t, s.writer.FormDataContentType(), s.url, s.buffer) + require.Equal(t, http.StatusInternalServerError, response.Code) +} + +func TestUploadFormProcessing(t *testing.T) { + s := setupWithTmpPath(t, "metadata", true, "", nil, nil) + defer s.cleanup() + require.NoError(t, s.writer.Close()) + + response := testUploadArtifacts(t, s.writer.FormDataContentType(), s.url, s.buffer) + require.Equal(t, http.StatusInternalServerError, response.Code) +} + +func TestLsifFileProcessing(t *testing.T) { + tempPath, err := ioutil.TempDir("", "uploads") + require.NoError(t, err) + + s := setupWithTmpPath(t, "file", true, "zip", &api.Response{TempPath: tempPath, ProcessLsif: true}, nil) + defer s.cleanup() + + file, err := os.Open("../../testdata/lsif/valid.lsif.zip") + require.NoError(t, err) + + _, err = io.Copy(s.fileWriter, file) + require.NoError(t, err) + require.NoError(t, file.Close()) + require.NoError(t, s.writer.Close()) + + response := testUploadArtifacts(t, s.writer.FormDataContentType(), s.url, s.buffer) + require.Equal(t, http.StatusOK, response.Code) + testhelper.RequireResponseHeader(t, response, MetadataHeaderKey, MetadataHeaderPresent) +} + +func TestInvalidLsifFileProcessing(t *testing.T) { + tempPath, err := ioutil.TempDir("", "uploads") + require.NoError(t, err) + + s := setupWithTmpPath(t, "file", true, "zip", &api.Response{TempPath: tempPath, ProcessLsif: true}, nil) + defer s.cleanup() + + file, err := os.Open("../../testdata/lsif/invalid.lsif.zip") + require.NoError(t, err) + + _, err = io.Copy(s.fileWriter, file) + require.NoError(t, err) + require.NoError(t, file.Close()) + require.NoError(t, s.writer.Close()) + + response := testUploadArtifacts(t, s.writer.FormDataContentType(), s.url, s.buffer) + require.Equal(t, http.StatusInternalServerError, response.Code) +} diff --git a/workhorse/internal/artifacts/entry.go b/workhorse/internal/artifacts/entry.go new file mode 100644 index 00000000000..0c697d40020 --- /dev/null +++ b/workhorse/internal/artifacts/entry.go @@ -0,0 +1,123 @@ +package artifacts + +import ( + "bufio" + "context" + "fmt" + "io" + "mime" + "net/http" + "os" + "os/exec" + "path/filepath" + "strings" + "syscall" + + "gitlab.com/gitlab-org/labkit/log" + + "gitlab.com/gitlab-org/gitlab-workhorse/internal/helper" + "gitlab.com/gitlab-org/gitlab-workhorse/internal/senddata" + "gitlab.com/gitlab-org/gitlab-workhorse/internal/zipartifacts" +) + +type entry struct{ senddata.Prefix } +type entryParams struct{ Archive, Entry string } + +var SendEntry = &entry{"artifacts-entry:"} + +// Artifacts downloader doesn't support ranges when downloading a single file +func (e *entry) Inject(w http.ResponseWriter, r *http.Request, sendData string) { + var params entryParams + if err := e.Unpack(¶ms, sendData); err != nil { + helper.Fail500(w, r, fmt.Errorf("SendEntry: unpack sendData: %v", err)) + return + } + + log.WithContextFields(r.Context(), log.Fields{ + "entry": params.Entry, + "archive": params.Archive, + "path": r.URL.Path, + }).Print("SendEntry: sending") + + if params.Archive == "" || params.Entry == "" { + helper.Fail500(w, r, fmt.Errorf("SendEntry: Archive or Entry is empty")) + return + } + + err := unpackFileFromZip(r.Context(), params.Archive, params.Entry, w.Header(), w) + + if os.IsNotExist(err) { + http.NotFound(w, r) + } else if err != nil { + helper.Fail500(w, r, fmt.Errorf("SendEntry: %v", err)) + } +} + +func detectFileContentType(fileName string) string { + contentType := mime.TypeByExtension(filepath.Ext(fileName)) + if contentType == "" { + contentType = "application/octet-stream" + } + return contentType +} + +func unpackFileFromZip(ctx context.Context, archivePath, encodedFilename string, headers http.Header, output io.Writer) error { + fileName, err := zipartifacts.DecodeFileEntry(encodedFilename) + if err != nil { + return err + } + + catFile := exec.Command("gitlab-zip-cat") + catFile.Env = append(os.Environ(), + "ARCHIVE_PATH="+archivePath, + "ENCODED_FILE_NAME="+encodedFilename, + ) + catFile.Stderr = log.ContextLogger(ctx).Writer() + catFile.SysProcAttr = &syscall.SysProcAttr{Setpgid: true} + stdout, err := catFile.StdoutPipe() + if err != nil { + return fmt.Errorf("create gitlab-zip-cat stdout pipe: %v", err) + } + + if err := catFile.Start(); err != nil { + return fmt.Errorf("start %v: %v", catFile.Args, err) + } + defer helper.CleanUpProcessGroup(catFile) + + basename := filepath.Base(fileName) + reader := bufio.NewReader(stdout) + contentLength, err := reader.ReadString('\n') + if err != nil { + if catFileErr := waitCatFile(catFile); catFileErr != nil { + return catFileErr + } + return fmt.Errorf("read content-length: %v", err) + } + contentLength = strings.TrimSuffix(contentLength, "\n") + + // Write http headers about the file + headers.Set("Content-Length", contentLength) + headers.Set("Content-Type", detectFileContentType(fileName)) + headers.Set("Content-Disposition", "attachment; filename=\""+escapeQuotes(basename)+"\"") + // Copy file body to client + if _, err := io.Copy(output, reader); err != nil { + return fmt.Errorf("copy stdout of %v: %v", catFile.Args, err) + } + + return waitCatFile(catFile) +} + +func waitCatFile(cmd *exec.Cmd) error { + err := cmd.Wait() + if err == nil { + return nil + } + + st, ok := helper.ExitStatus(err) + + if ok && (st == zipartifacts.CodeArchiveNotFound || st == zipartifacts.CodeEntryNotFound) { + return os.ErrNotExist + } + return fmt.Errorf("wait for %v to finish: %v", cmd.Args, err) + +} diff --git a/workhorse/internal/artifacts/entry_test.go b/workhorse/internal/artifacts/entry_test.go new file mode 100644 index 00000000000..6f1e9d360aa --- /dev/null +++ b/workhorse/internal/artifacts/entry_test.go @@ -0,0 +1,134 @@ +package artifacts + +import ( + "archive/zip" + "encoding/base64" + "fmt" + "io/ioutil" + "net/http" + "net/http/httptest" + "os" + "path/filepath" + "testing" + + "github.com/stretchr/testify/require" + + "gitlab.com/gitlab-org/gitlab-workhorse/internal/testhelper" +) + +func testEntryServer(t *testing.T, archive string, entry string) *httptest.ResponseRecorder { + mux := http.NewServeMux() + mux.HandleFunc("/url/path", func(w http.ResponseWriter, r *http.Request) { + require.Equal(t, "GET", r.Method) + + encodedEntry := base64.StdEncoding.EncodeToString([]byte(entry)) + jsonParams := fmt.Sprintf(`{"Archive":"%s","Entry":"%s"}`, archive, encodedEntry) + data := base64.URLEncoding.EncodeToString([]byte(jsonParams)) + + SendEntry.Inject(w, r, data) + }) + + httpRequest, err := http.NewRequest("GET", "/url/path", nil) + require.NoError(t, err) + response := httptest.NewRecorder() + mux.ServeHTTP(response, httpRequest) + return response +} + +func TestDownloadingFromValidArchive(t *testing.T) { + tempFile, err := ioutil.TempFile("", "uploads") + require.NoError(t, err) + defer tempFile.Close() + defer os.Remove(tempFile.Name()) + + archive := zip.NewWriter(tempFile) + defer archive.Close() + fileInArchive, err := archive.Create("test.txt") + require.NoError(t, err) + fmt.Fprint(fileInArchive, "testtest") + archive.Close() + + response := testEntryServer(t, tempFile.Name(), "test.txt") + + require.Equal(t, 200, response.Code) + + testhelper.RequireResponseHeader(t, response, + "Content-Type", + "text/plain; charset=utf-8") + testhelper.RequireResponseHeader(t, response, + "Content-Disposition", + "attachment; filename=\"test.txt\"") + + testhelper.RequireResponseBody(t, response, "testtest") +} + +func TestDownloadingFromValidHTTPArchive(t *testing.T) { + tempDir, err := ioutil.TempDir("", "uploads") + require.NoError(t, err) + defer os.RemoveAll(tempDir) + + f, err := os.Create(filepath.Join(tempDir, "archive.zip")) + require.NoError(t, err) + defer f.Close() + + archive := zip.NewWriter(f) + defer archive.Close() + fileInArchive, err := archive.Create("test.txt") + require.NoError(t, err) + fmt.Fprint(fileInArchive, "testtest") + archive.Close() + f.Close() + + fileServer := httptest.NewServer(http.FileServer(http.Dir(tempDir))) + defer fileServer.Close() + + response := testEntryServer(t, fileServer.URL+"/archive.zip", "test.txt") + + require.Equal(t, 200, response.Code) + + testhelper.RequireResponseHeader(t, response, + "Content-Type", + "text/plain; charset=utf-8") + testhelper.RequireResponseHeader(t, response, + "Content-Disposition", + "attachment; filename=\"test.txt\"") + + testhelper.RequireResponseBody(t, response, "testtest") +} + +func TestDownloadingNonExistingFile(t *testing.T) { + tempFile, err := ioutil.TempFile("", "uploads") + require.NoError(t, err) + defer tempFile.Close() + defer os.Remove(tempFile.Name()) + + archive := zip.NewWriter(tempFile) + defer archive.Close() + archive.Close() + + response := testEntryServer(t, tempFile.Name(), "test") + require.Equal(t, 404, response.Code) +} + +func TestDownloadingFromInvalidArchive(t *testing.T) { + response := testEntryServer(t, "path/to/non/existing/file", "test") + require.Equal(t, 404, response.Code) +} + +func TestIncompleteApiResponse(t *testing.T) { + response := testEntryServer(t, "", "") + require.Equal(t, 500, response.Code) +} + +func TestDownloadingFromNonExistingHTTPArchive(t *testing.T) { + tempDir, err := ioutil.TempDir("", "uploads") + require.NoError(t, err) + defer os.RemoveAll(tempDir) + + fileServer := httptest.NewServer(http.FileServer(http.Dir(tempDir))) + defer fileServer.Close() + + response := testEntryServer(t, fileServer.URL+"/not-existing-archive-file.zip", "test.txt") + + require.Equal(t, 404, response.Code) +} diff --git a/workhorse/internal/artifacts/escape_quotes.go b/workhorse/internal/artifacts/escape_quotes.go new file mode 100644 index 00000000000..94db2be39b7 --- /dev/null +++ b/workhorse/internal/artifacts/escape_quotes.go @@ -0,0 +1,10 @@ +package artifacts + +import "strings" + +// taken from mime/multipart/writer.go +var quoteEscaper = strings.NewReplacer("\\", "\\\\", `"`, "\\\"") + +func escapeQuotes(s string) string { + return quoteEscaper.Replace(s) +} |