22184: Force sync in depth header test sequence.
[arvados.git] / services / keep-web / s3_test.go
1 // Copyright (C) The Arvados Authors. All rights reserved.
2 //
3 // SPDX-License-Identifier: AGPL-3.0
4
5 package keepweb
6
7 import (
8         "bytes"
9         "context"
10         "crypto/rand"
11         "crypto/sha256"
12         "fmt"
13         "io/ioutil"
14         "mime"
15         "net/http"
16         "net/http/httptest"
17         "net/url"
18         "os"
19         "os/exec"
20         "sort"
21         "strings"
22         "sync"
23         "time"
24
25         "git.arvados.org/arvados.git/sdk/go/arvados"
26         "git.arvados.org/arvados.git/sdk/go/arvadosclient"
27         "git.arvados.org/arvados.git/sdk/go/arvadostest"
28         "git.arvados.org/arvados.git/sdk/go/keepclient"
29         "github.com/AdRoll/goamz/aws"
30         "github.com/AdRoll/goamz/s3"
31         aws_aws "github.com/aws/aws-sdk-go/aws"
32         aws_credentials "github.com/aws/aws-sdk-go/aws/credentials"
33         aws_session "github.com/aws/aws-sdk-go/aws/session"
34         aws_s3 "github.com/aws/aws-sdk-go/service/s3"
35         check "gopkg.in/check.v1"
36 )
37
38 type CachedS3SecretSuite struct{}
39
40 var _ = check.Suite(&CachedS3SecretSuite{})
41
42 func (s *CachedS3SecretSuite) activeACA(expiresAt time.Time) *arvados.APIClientAuthorization {
43         return &arvados.APIClientAuthorization{
44                 UUID:      arvadostest.ActiveTokenUUID,
45                 APIToken:  arvadostest.ActiveToken,
46                 ExpiresAt: expiresAt,
47         }
48 }
49
50 func (s *CachedS3SecretSuite) TestNewCachedS3SecretExpiresBeforeTTL(c *check.C) {
51         expected := time.Unix(1<<29, 0)
52         aca := s.activeACA(expected)
53         actual := newCachedS3Secret(aca, time.Unix(1<<30, 0))
54         c.Check(actual.expiry, check.Equals, expected)
55 }
56
57 func (s *CachedS3SecretSuite) TestNewCachedS3SecretExpiresAfterTTL(c *check.C) {
58         expected := time.Unix(1<<29, 0)
59         aca := s.activeACA(time.Unix(1<<30, 0))
60         actual := newCachedS3Secret(aca, expected)
61         c.Check(actual.expiry, check.Equals, expected)
62 }
63
64 func (s *CachedS3SecretSuite) TestNewCachedS3SecretWithoutExpiry(c *check.C) {
65         expected := time.Unix(1<<29, 0)
66         aca := s.activeACA(time.Time{})
67         actual := newCachedS3Secret(aca, expected)
68         c.Check(actual.expiry, check.Equals, expected)
69 }
70
71 func (s *CachedS3SecretSuite) cachedSecretWithExpiry(expiry time.Time) *cachedS3Secret {
72         return &cachedS3Secret{
73                 auth:   s.activeACA(expiry),
74                 expiry: expiry,
75         }
76 }
77
78 func (s *CachedS3SecretSuite) TestIsValidAtEmpty(c *check.C) {
79         cache := &cachedS3Secret{}
80         c.Check(cache.isValidAt(time.Unix(0, 0)), check.Equals, false)
81         c.Check(cache.isValidAt(time.Unix(1<<31, 0)), check.Equals, false)
82 }
83
84 func (s *CachedS3SecretSuite) TestIsValidAtNoAuth(c *check.C) {
85         cache := &cachedS3Secret{expiry: time.Unix(3, 0)}
86         c.Check(cache.isValidAt(time.Unix(2, 0)), check.Equals, false)
87         c.Check(cache.isValidAt(time.Unix(4, 0)), check.Equals, false)
88 }
89
90 func (s *CachedS3SecretSuite) TestIsValidAtNoExpiry(c *check.C) {
91         cache := &cachedS3Secret{auth: s.activeACA(time.Unix(3, 0))}
92         c.Check(cache.isValidAt(time.Unix(2, 0)), check.Equals, false)
93         c.Check(cache.isValidAt(time.Unix(4, 0)), check.Equals, false)
94 }
95
96 func (s *CachedS3SecretSuite) TestIsValidAtTimeAfterExpiry(c *check.C) {
97         expiry := time.Unix(10, 0)
98         cache := s.cachedSecretWithExpiry(expiry)
99         c.Check(cache.isValidAt(expiry), check.Equals, false)
100         c.Check(cache.isValidAt(time.Unix(1<<25, 0)), check.Equals, false)
101         c.Check(cache.isValidAt(time.Unix(1<<30, 0)), check.Equals, false)
102 }
103
104 func (s *CachedS3SecretSuite) TestIsValidAtTimeBeforeExpiry(c *check.C) {
105         cache := s.cachedSecretWithExpiry(time.Unix(1<<30, 0))
106         c.Check(cache.isValidAt(time.Unix(1<<25, 0)), check.Equals, true)
107         c.Check(cache.isValidAt(time.Unix(1<<27, 0)), check.Equals, true)
108         c.Check(cache.isValidAt(time.Unix(1<<29, 0)), check.Equals, true)
109 }
110
111 func (s *CachedS3SecretSuite) TestIsValidAtZeroTime(c *check.C) {
112         cache := s.cachedSecretWithExpiry(time.Unix(10, 0))
113         c.Check(cache.isValidAt(time.Time{}), check.Equals, false)
114 }
115
116 type s3stage struct {
117         arv        *arvados.Client
118         ac         *arvadosclient.ArvadosClient
119         kc         *keepclient.KeepClient
120         proj       arvados.Group
121         projbucket *s3.Bucket
122         subproj    arvados.Group
123         coll       arvados.Collection
124         collbucket *s3.Bucket
125 }
126
127 func (s *IntegrationSuite) s3setup(c *check.C) s3stage {
128         var proj, subproj arvados.Group
129         var coll arvados.Collection
130         arv := arvados.NewClientFromEnv()
131         arv.AuthToken = arvadostest.ActiveToken
132         err := arv.RequestAndDecode(&proj, "POST", "arvados/v1/groups", nil, map[string]interface{}{
133                 "group": map[string]interface{}{
134                         "group_class": "project",
135                         "name":        "keep-web s3 test",
136                         "properties": map[string]interface{}{
137                                 "project-properties-key": "project properties value",
138                         },
139                 },
140                 "ensure_unique_name": true,
141         })
142         c.Assert(err, check.IsNil)
143         err = arv.RequestAndDecode(&subproj, "POST", "arvados/v1/groups", nil, map[string]interface{}{
144                 "group": map[string]interface{}{
145                         "owner_uuid":  proj.UUID,
146                         "group_class": "project",
147                         "name":        "keep-web s3 test subproject",
148                         "properties": map[string]interface{}{
149                                 "subproject_properties_key": "subproject properties value",
150                                 "invalid header key":        "this value will not be returned because key contains spaces",
151                         },
152                 },
153         })
154         c.Assert(err, check.IsNil)
155         err = arv.RequestAndDecode(&coll, "POST", "arvados/v1/collections", nil, map[string]interface{}{"collection": map[string]interface{}{
156                 "owner_uuid":    proj.UUID,
157                 "name":          "keep-web s3 test collection",
158                 "manifest_text": ". d41d8cd98f00b204e9800998ecf8427e+0 0:0:emptyfile\n./emptydir d41d8cd98f00b204e9800998ecf8427e+0 0:0:.\n",
159                 "properties": map[string]interface{}{
160                         "string":   "string value",
161                         "array":    []string{"element1", "element2"},
162                         "object":   map[string]interface{}{"key": map[string]interface{}{"key2": "value⛵"}},
163                         "nonascii": "⛵",
164                         "newline":  "foo\r\nX-Bad: header",
165                         // This key cannot be expressed as a MIME
166                         // header key, so it will be silently skipped
167                         // (see "Inject" in PropertiesAsMetadata test)
168                         "a: a\r\nInject": "bogus",
169                 },
170         }})
171         c.Assert(err, check.IsNil)
172         ac, err := arvadosclient.New(arv)
173         c.Assert(err, check.IsNil)
174         kc, err := keepclient.MakeKeepClient(ac)
175         c.Assert(err, check.IsNil)
176         fs, err := coll.FileSystem(arv, kc)
177         c.Assert(err, check.IsNil)
178         f, err := fs.OpenFile("sailboat.txt", os.O_CREATE|os.O_WRONLY, 0644)
179         c.Assert(err, check.IsNil)
180         _, err = f.Write([]byte("⛵\n"))
181         c.Assert(err, check.IsNil)
182         err = f.Close()
183         c.Assert(err, check.IsNil)
184         err = fs.Sync()
185         c.Assert(err, check.IsNil)
186         err = arv.RequestAndDecode(&coll, "GET", "arvados/v1/collections/"+coll.UUID, nil, nil)
187         c.Assert(err, check.IsNil)
188
189         auth := aws.NewAuth(arvadostest.ActiveTokenUUID, arvadostest.ActiveToken, "", time.Now().Add(time.Hour))
190         region := aws.Region{
191                 Name:       "zzzzz",
192                 S3Endpoint: s.testServer.URL,
193         }
194         client := s3.New(*auth, region)
195         client.Signature = aws.V4Signature
196         return s3stage{
197                 arv:  arv,
198                 ac:   ac,
199                 kc:   kc,
200                 proj: proj,
201                 projbucket: &s3.Bucket{
202                         S3:   client,
203                         Name: proj.UUID,
204                 },
205                 subproj: subproj,
206                 coll:    coll,
207                 collbucket: &s3.Bucket{
208                         S3:   client,
209                         Name: coll.UUID,
210                 },
211         }
212 }
213
214 func (stage s3stage) teardown(c *check.C) {
215         if stage.coll.UUID != "" {
216                 err := stage.arv.RequestAndDecode(&stage.coll, "DELETE", "arvados/v1/collections/"+stage.coll.UUID, nil, nil)
217                 c.Check(err, check.IsNil)
218         }
219         if stage.proj.UUID != "" {
220                 err := stage.arv.RequestAndDecode(&stage.proj, "DELETE", "arvados/v1/groups/"+stage.proj.UUID, nil, nil)
221                 c.Check(err, check.IsNil)
222         }
223 }
224
225 func (s *IntegrationSuite) TestS3Signatures(c *check.C) {
226         stage := s.s3setup(c)
227         defer stage.teardown(c)
228
229         bucket := stage.collbucket
230         for _, trial := range []struct {
231                 success   bool
232                 signature int
233                 accesskey string
234                 secretkey string
235         }{
236                 {true, aws.V2Signature, arvadostest.ActiveToken, "none"},
237                 {true, aws.V2Signature, url.QueryEscape(arvadostest.ActiveTokenV2), "none"},
238                 {true, aws.V2Signature, strings.Replace(arvadostest.ActiveTokenV2, "/", "_", -1), "none"},
239                 {false, aws.V2Signature, "none", "none"},
240                 {false, aws.V2Signature, "none", arvadostest.ActiveToken},
241
242                 {true, aws.V4Signature, arvadostest.ActiveTokenUUID, arvadostest.ActiveToken},
243                 {true, aws.V4Signature, arvadostest.ActiveToken, arvadostest.ActiveToken},
244                 {true, aws.V4Signature, url.QueryEscape(arvadostest.ActiveTokenV2), url.QueryEscape(arvadostest.ActiveTokenV2)},
245                 {true, aws.V4Signature, strings.Replace(arvadostest.ActiveTokenV2, "/", "_", -1), strings.Replace(arvadostest.ActiveTokenV2, "/", "_", -1)},
246                 {false, aws.V4Signature, arvadostest.ActiveToken, ""},
247                 {false, aws.V4Signature, arvadostest.ActiveToken, "none"},
248                 {false, aws.V4Signature, "none", arvadostest.ActiveToken},
249                 {false, aws.V4Signature, "none", "none"},
250         } {
251                 c.Logf("%#v", trial)
252                 bucket.S3.Auth = *(aws.NewAuth(trial.accesskey, trial.secretkey, "", time.Now().Add(time.Hour)))
253                 bucket.S3.Signature = trial.signature
254                 _, err := bucket.GetReader("emptyfile")
255                 if trial.success {
256                         c.Check(err, check.IsNil)
257                 } else {
258                         c.Check(err, check.NotNil)
259                 }
260         }
261 }
262
263 func (s *IntegrationSuite) TestS3SecretCacheUpdates(c *check.C) {
264         stage := s.s3setup(c)
265         defer stage.teardown(c)
266         reqUrl, err := url.Parse("https://" + stage.collbucket.Name + ".example.com/")
267         c.Assert(err, check.IsNil)
268
269         for trialName, trialAuth := range map[string]string{
270                 "v1 token":                    arvadostest.ActiveToken,
271                 "token UUID":                  arvadostest.ActiveTokenUUID,
272                 "v2 token query escaped":      url.QueryEscape(arvadostest.ActiveTokenV2),
273                 "v2 token underscore escaped": strings.Replace(arvadostest.ActiveTokenV2, "/", "_", -1),
274         } {
275                 s.handler.s3SecretCache = nil
276                 req, err := http.NewRequest("GET", reqUrl.String(), bytes.NewReader(nil))
277                 if !c.Check(err, check.IsNil) {
278                         continue
279                 }
280                 secret := trialAuth
281                 if secret[5:12] == "-gj3su-" {
282                         secret = arvadostest.ActiveToken
283                 }
284                 s.sign(c, req, trialAuth, secret)
285                 rec := httptest.NewRecorder()
286                 s.handler.ServeHTTP(rec, req)
287                 if !c.Check(rec.Result().StatusCode, check.Equals, http.StatusOK,
288                         check.Commentf("%s auth did not get 200 OK response: %v", trialName, req)) {
289                         continue
290                 }
291
292                 for name, key := range map[string]string{
293                         "v1 token":   arvadostest.ActiveToken,
294                         "token UUID": arvadostest.ActiveTokenUUID,
295                         "v2 token":   arvadostest.ActiveTokenV2,
296                 } {
297                         actual, ok := s.handler.s3SecretCache[key]
298                         if c.Check(ok, check.Equals, true, check.Commentf("%s not cached from %s", name, trialName)) {
299                                 c.Check(actual.auth.UUID, check.Equals, arvadostest.ActiveTokenUUID)
300                         }
301                 }
302         }
303 }
304
305 func (s *IntegrationSuite) TestS3SecretCacheUsed(c *check.C) {
306         stage := s.s3setup(c)
307         defer stage.teardown(c)
308
309         token := arvadostest.ActiveToken
310         // Step 1: Make a request to get the active token in the cache.
311         reqUrl, err := url.Parse("https://" + stage.collbucket.Name + ".example.com/")
312         c.Assert(err, check.IsNil)
313         req, err := http.NewRequest("GET", reqUrl.String(), bytes.NewReader(nil))
314         s.sign(c, req, token, token)
315         rec := httptest.NewRecorder()
316         s.handler.ServeHTTP(rec, req)
317         resp := rec.Result()
318         c.Assert(resp.StatusCode, check.Equals, http.StatusOK,
319                 check.Commentf("first request did not get 200 OK response"))
320
321         // Step 2: Remove some cache keys our request doesn't rely upon.
322         c.Assert(s.handler.s3SecretCache[arvadostest.ActiveTokenUUID], check.NotNil)
323         delete(s.handler.s3SecretCache, arvadostest.ActiveTokenUUID)
324         c.Assert(s.handler.s3SecretCache[arvadostest.ActiveTokenV2], check.NotNil)
325         delete(s.handler.s3SecretCache, arvadostest.ActiveTokenV2)
326
327         // Step 3: Repeat the original request.
328         rec = httptest.NewRecorder()
329         s.handler.ServeHTTP(rec, req)
330         resp = rec.Result()
331         c.Assert(resp.StatusCode, check.Equals, http.StatusOK,
332                 check.Commentf("cached auth request did not get 200 OK response"))
333
334         // Step 4: Confirm the deleted cache keys were not re-added
335         // (which would imply the authorization was re-requested and cached).
336         c.Check(s.handler.s3SecretCache[arvadostest.ActiveTokenUUID], check.IsNil,
337                 check.Commentf("token UUID re-added to cache after removal"))
338         c.Check(s.handler.s3SecretCache[arvadostest.ActiveTokenV2], check.IsNil,
339                 check.Commentf("v2 token re-added to cache after removal"))
340 }
341
342 func (s *IntegrationSuite) TestS3SecretCacheCleanup(c *check.C) {
343         stage := s.s3setup(c)
344         defer stage.teardown(c)
345         td := -2 * s3SecretCacheTidyInterval
346         startTidied := time.Now().Add(td)
347         s.handler.s3SecretCacheNextTidy = startTidied
348         s.handler.s3SecretCache = make(map[string]*cachedS3Secret)
349         s.handler.s3SecretCache["old"] = &cachedS3Secret{expiry: startTidied.Add(td)}
350
351         reqUrl, err := url.Parse("https://" + stage.collbucket.Name + ".example.com/")
352         c.Assert(err, check.IsNil)
353         req, err := http.NewRequest("GET", reqUrl.String(), bytes.NewReader(nil))
354         token := arvadostest.ActiveToken
355         s.sign(c, req, token, token)
356         rec := httptest.NewRecorder()
357         s.handler.ServeHTTP(rec, req)
358
359         c.Check(s.handler.s3SecretCache["old"], check.IsNil,
360                 check.Commentf("expired token not removed from cache"))
361         c.Check(s.handler.s3SecretCacheNextTidy.After(startTidied), check.Equals, true,
362                 check.Commentf("s3SecretCacheNextTidy not updated"))
363         c.Check(s.handler.s3SecretCache[token], check.NotNil,
364                 check.Commentf("just-used token not found in cache"))
365 }
366
367 func (s *IntegrationSuite) TestS3HeadBucket(c *check.C) {
368         stage := s.s3setup(c)
369         defer stage.teardown(c)
370
371         for _, bucket := range []*s3.Bucket{stage.collbucket, stage.projbucket} {
372                 c.Logf("bucket %s", bucket.Name)
373                 exists, err := bucket.Exists("")
374                 c.Check(err, check.IsNil)
375                 c.Check(exists, check.Equals, true)
376         }
377 }
378
379 func (s *IntegrationSuite) TestS3CollectionGetObject(c *check.C) {
380         stage := s.s3setup(c)
381         defer stage.teardown(c)
382         s.testS3GetObject(c, stage.collbucket, "")
383 }
384 func (s *IntegrationSuite) TestS3ProjectGetObject(c *check.C) {
385         stage := s.s3setup(c)
386         defer stage.teardown(c)
387         s.testS3GetObject(c, stage.projbucket, stage.coll.Name+"/")
388 }
389 func (s *IntegrationSuite) testS3GetObject(c *check.C, bucket *s3.Bucket, prefix string) {
390         rdr, err := bucket.GetReader(prefix + "emptyfile")
391         c.Assert(err, check.IsNil)
392         buf, err := ioutil.ReadAll(rdr)
393         c.Check(err, check.IsNil)
394         c.Check(len(buf), check.Equals, 0)
395         err = rdr.Close()
396         c.Check(err, check.IsNil)
397
398         // GetObject
399         rdr, err = bucket.GetReader(prefix + "missingfile")
400         c.Check(err.(*s3.Error).StatusCode, check.Equals, 404)
401         c.Check(err.(*s3.Error).Code, check.Equals, `NoSuchKey`)
402         c.Check(err, check.ErrorMatches, `The specified key does not exist.`)
403
404         // HeadObject
405         exists, err := bucket.Exists(prefix + "missingfile")
406         c.Check(err, check.IsNil)
407         c.Check(exists, check.Equals, false)
408
409         // GetObject
410         rdr, err = bucket.GetReader(prefix + "sailboat.txt")
411         c.Assert(err, check.IsNil)
412         buf, err = ioutil.ReadAll(rdr)
413         c.Check(err, check.IsNil)
414         c.Check(buf, check.DeepEquals, []byte("⛵\n"))
415         err = rdr.Close()
416         c.Check(err, check.IsNil)
417
418         // HeadObject
419         resp, err := bucket.Head(prefix+"sailboat.txt", nil)
420         c.Check(err, check.IsNil)
421         c.Check(resp.StatusCode, check.Equals, http.StatusOK)
422         c.Check(resp.ContentLength, check.Equals, int64(4))
423
424         // HeadObject with superfluous leading slashes
425         exists, err = bucket.Exists(prefix + "//sailboat.txt")
426         c.Check(err, check.IsNil)
427         c.Check(exists, check.Equals, false)
428 }
429
430 func (s *IntegrationSuite) checkMetaEquals(c *check.C, hdr http.Header, expect map[string]string) {
431         got := map[string]string{}
432         for hk, hv := range hdr {
433                 if k := strings.TrimPrefix(hk, "X-Amz-Meta-"); k != hk && len(hv) == 1 {
434                         got[k] = hv[0]
435                 }
436         }
437         c.Check(got, check.DeepEquals, expect)
438 }
439
440 func (s *IntegrationSuite) TestS3PropertiesAsMetadata(c *check.C) {
441         stage := s.s3setup(c)
442         defer stage.teardown(c)
443
444         expectCollectionTags := map[string]string{
445                 "String":   "string value",
446                 "Array":    `["element1","element2"]`,
447                 "Object":   mime.BEncoding.Encode("UTF-8", `{"key":{"key2":"value⛵"}}`),
448                 "Nonascii": "=?UTF-8?b?4pu1?=",
449                 "Newline":  mime.BEncoding.Encode("UTF-8", "foo\r\nX-Bad: header"),
450         }
451         expectSubprojectTags := map[string]string{
452                 "Subproject_properties_key": "subproject properties value",
453         }
454         expectProjectTags := map[string]string{
455                 "Project-Properties-Key": "project properties value",
456         }
457
458         c.Log("HEAD object with metadata from collection")
459         resp, err := stage.collbucket.Head("sailboat.txt", nil)
460         c.Assert(err, check.IsNil)
461         s.checkMetaEquals(c, resp.Header, expectCollectionTags)
462
463         c.Log("GET object with metadata from collection")
464         rdr, hdr, err := stage.collbucket.GetReaderWithHeaders("sailboat.txt")
465         c.Assert(err, check.IsNil)
466         content, err := ioutil.ReadAll(rdr)
467         c.Check(err, check.IsNil)
468         rdr.Close()
469         c.Check(content, check.HasLen, 4)
470         s.checkMetaEquals(c, hdr, expectCollectionTags)
471         c.Check(hdr["Inject"], check.IsNil)
472
473         c.Log("HEAD bucket with metadata from collection")
474         resp, err = stage.collbucket.Head("/", nil)
475         c.Assert(err, check.IsNil)
476         s.checkMetaEquals(c, resp.Header, expectCollectionTags)
477
478         c.Log("HEAD directory placeholder with metadata from collection")
479         resp, err = stage.projbucket.Head("keep-web s3 test collection/", nil)
480         c.Assert(err, check.IsNil)
481         s.checkMetaEquals(c, resp.Header, expectCollectionTags)
482
483         c.Log("HEAD file with metadata from collection")
484         resp, err = stage.projbucket.Head("keep-web s3 test collection/sailboat.txt", nil)
485         c.Assert(err, check.IsNil)
486         s.checkMetaEquals(c, resp.Header, expectCollectionTags)
487
488         c.Log("HEAD directory placeholder with metadata from subproject")
489         resp, err = stage.projbucket.Head("keep-web s3 test subproject/", nil)
490         c.Assert(err, check.IsNil)
491         s.checkMetaEquals(c, resp.Header, expectSubprojectTags)
492
493         c.Log("HEAD bucket with metadata from project")
494         resp, err = stage.projbucket.Head("/", nil)
495         c.Assert(err, check.IsNil)
496         s.checkMetaEquals(c, resp.Header, expectProjectTags)
497 }
498
499 func (s *IntegrationSuite) TestS3CollectionPutObjectSuccess(c *check.C) {
500         stage := s.s3setup(c)
501         defer stage.teardown(c)
502         s.testS3PutObjectSuccess(c, stage.collbucket, "", stage.coll.UUID)
503 }
504 func (s *IntegrationSuite) TestS3ProjectPutObjectSuccess(c *check.C) {
505         stage := s.s3setup(c)
506         defer stage.teardown(c)
507         s.testS3PutObjectSuccess(c, stage.projbucket, stage.coll.Name+"/", stage.coll.UUID)
508 }
509 func (s *IntegrationSuite) testS3PutObjectSuccess(c *check.C, bucket *s3.Bucket, prefix string, collUUID string) {
510         // We insert a delay between test cases to ensure we exercise
511         // rollover of expired sessions.
512         sleep := time.Second / 100
513         s.handler.Cluster.Collections.WebDAVCache.TTL = arvados.Duration(sleep * 3)
514
515         for _, trial := range []struct {
516                 path        string
517                 size        int
518                 contentType string
519         }{
520                 {
521                         path:        "newfile",
522                         size:        128000000,
523                         contentType: "application/octet-stream",
524                 }, {
525                         path:        "newdir/newfile",
526                         size:        1 << 26,
527                         contentType: "application/octet-stream",
528                 }, {
529                         path:        "ccc/",
530                         size:        0,
531                         contentType: "application/x-directory",
532                 }, {
533                         path:        "newdir1/newdir2/newfile",
534                         size:        0,
535                         contentType: "application/octet-stream",
536                 }, {
537                         path:        "newdir1/newdir2/newdir3/",
538                         size:        0,
539                         contentType: "application/x-directory",
540                 },
541         } {
542                 time.Sleep(sleep)
543                 c.Logf("=== %v", trial)
544
545                 objname := prefix + trial.path
546
547                 _, err := bucket.GetReader(objname)
548                 if !c.Check(err, check.NotNil) {
549                         continue
550                 }
551                 c.Check(err.(*s3.Error).StatusCode, check.Equals, http.StatusNotFound)
552                 c.Check(err.(*s3.Error).Code, check.Equals, `NoSuchKey`)
553                 if !c.Check(err, check.ErrorMatches, `The specified key does not exist.`) {
554                         continue
555                 }
556
557                 buf := make([]byte, trial.size)
558                 rand.Read(buf)
559
560                 err = bucket.PutReader(objname, bytes.NewReader(buf), int64(len(buf)), trial.contentType, s3.Private, s3.Options{})
561                 c.Check(err, check.IsNil)
562
563                 rdr, err := bucket.GetReader(objname)
564                 if strings.HasSuffix(trial.path, "/") && !s.handler.Cluster.Collections.S3FolderObjects {
565                         c.Check(err, check.NotNil)
566                         continue
567                 } else if !c.Check(err, check.IsNil) {
568                         continue
569                 }
570                 buf2, err := ioutil.ReadAll(rdr)
571                 c.Check(err, check.IsNil)
572                 c.Check(buf2, check.HasLen, len(buf))
573                 c.Check(bytes.Equal(buf, buf2), check.Equals, true)
574
575                 // Check that the change is immediately visible via
576                 // (non-S3) webdav request.
577                 _, resp := s.do("GET", "http://"+collUUID+".keep-web.example/"+trial.path, arvadostest.ActiveTokenV2, nil)
578                 c.Check(resp.Code, check.Equals, http.StatusOK)
579                 if !strings.HasSuffix(trial.path, "/") {
580                         c.Check(resp.Body.Len(), check.Equals, trial.size)
581                 }
582         }
583 }
584
585 func (s *IntegrationSuite) TestS3ProjectPutObjectNotSupported(c *check.C) {
586         stage := s.s3setup(c)
587         defer stage.teardown(c)
588         bucket := stage.projbucket
589
590         for _, trial := range []struct {
591                 path        string
592                 size        int
593                 contentType string
594         }{
595                 {
596                         path:        "newfile",
597                         size:        1234,
598                         contentType: "application/octet-stream",
599                 }, {
600                         path:        "newdir/newfile",
601                         size:        1234,
602                         contentType: "application/octet-stream",
603                 }, {
604                         path:        "newdir2/",
605                         size:        0,
606                         contentType: "application/x-directory",
607                 },
608         } {
609                 c.Logf("=== %v", trial)
610
611                 _, err := bucket.GetReader(trial.path)
612                 c.Check(err.(*s3.Error).StatusCode, check.Equals, 404)
613                 c.Check(err.(*s3.Error).Code, check.Equals, `NoSuchKey`)
614                 c.Assert(err, check.ErrorMatches, `The specified key does not exist.`)
615
616                 buf := make([]byte, trial.size)
617                 rand.Read(buf)
618
619                 err = bucket.PutReader(trial.path, bytes.NewReader(buf), int64(len(buf)), trial.contentType, s3.Private, s3.Options{})
620                 c.Check(err.(*s3.Error).StatusCode, check.Equals, 400)
621                 c.Check(err.(*s3.Error).Code, check.Equals, `InvalidArgument`)
622                 c.Check(err, check.ErrorMatches, `(mkdir "/by_id/zzzzz-j7d0g-[a-z0-9]{15}/newdir2?"|open "/zzzzz-j7d0g-[a-z0-9]{15}/newfile") failed: invalid (argument|operation)`)
623
624                 _, err = bucket.GetReader(trial.path)
625                 c.Check(err.(*s3.Error).StatusCode, check.Equals, 404)
626                 c.Check(err.(*s3.Error).Code, check.Equals, `NoSuchKey`)
627                 c.Assert(err, check.ErrorMatches, `The specified key does not exist.`)
628         }
629 }
630
631 func (s *IntegrationSuite) TestS3CollectionDeleteObject(c *check.C) {
632         stage := s.s3setup(c)
633         defer stage.teardown(c)
634         s.testS3DeleteObject(c, stage.collbucket, "")
635 }
636 func (s *IntegrationSuite) TestS3ProjectDeleteObject(c *check.C) {
637         stage := s.s3setup(c)
638         defer stage.teardown(c)
639         s.testS3DeleteObject(c, stage.projbucket, stage.coll.Name+"/")
640 }
641 func (s *IntegrationSuite) testS3DeleteObject(c *check.C, bucket *s3.Bucket, prefix string) {
642         s.handler.Cluster.Collections.S3FolderObjects = true
643         for _, trial := range []struct {
644                 path string
645         }{
646                 {"/"},
647                 {"nonexistentfile"},
648                 {"emptyfile"},
649                 {"sailboat.txt"},
650                 {"sailboat.txt/"},
651                 {"emptydir"},
652                 {"emptydir/"},
653         } {
654                 objname := prefix + trial.path
655                 comment := check.Commentf("objname %q", objname)
656
657                 err := bucket.Del(objname)
658                 if trial.path == "/" {
659                         c.Check(err, check.NotNil)
660                         continue
661                 }
662                 c.Check(err, check.IsNil, comment)
663                 _, err = bucket.GetReader(objname)
664                 c.Check(err, check.NotNil, comment)
665         }
666 }
667
668 func (s *IntegrationSuite) TestS3CollectionPutObjectFailure(c *check.C) {
669         stage := s.s3setup(c)
670         defer stage.teardown(c)
671         s.testS3PutObjectFailure(c, stage.collbucket, "")
672 }
673 func (s *IntegrationSuite) TestS3ProjectPutObjectFailure(c *check.C) {
674         stage := s.s3setup(c)
675         defer stage.teardown(c)
676         s.testS3PutObjectFailure(c, stage.projbucket, stage.coll.Name+"/")
677 }
678 func (s *IntegrationSuite) testS3PutObjectFailure(c *check.C, bucket *s3.Bucket, prefix string) {
679         s.handler.Cluster.Collections.S3FolderObjects = false
680
681         var wg sync.WaitGroup
682         for _, trial := range []struct {
683                 path string
684         }{
685                 {
686                         path: "emptyfile/newname", // emptyfile exists, see s3setup()
687                 }, {
688                         path: "emptyfile/", // emptyfile exists, see s3setup()
689                 }, {
690                         path: "emptydir", // dir already exists, see s3setup()
691                 }, {
692                         path: "emptydir/",
693                 }, {
694                         path: "newdir/",
695                 }, {
696                         path: "",
697                 },
698         } {
699                 trial := trial
700                 wg.Add(1)
701                 go func() {
702                         defer wg.Done()
703                         c.Logf("=== %v", trial)
704
705                         objname := prefix + trial.path
706
707                         buf := make([]byte, 1234)
708                         rand.Read(buf)
709
710                         err := bucket.PutReader(objname, bytes.NewReader(buf), int64(len(buf)), "application/octet-stream", s3.Private, s3.Options{})
711                         if !c.Check(err, check.ErrorMatches, `(invalid object name.*|open ".*" failed.*|object name conflicts with existing object|Missing object name in PUT request.)`, check.Commentf("PUT %q should fail", objname)) {
712                                 return
713                         }
714
715                         if objname != "" && objname != "/" {
716                                 _, err = bucket.GetReader(objname)
717                                 c.Check(err.(*s3.Error).StatusCode, check.Equals, 404)
718                                 c.Check(err.(*s3.Error).Code, check.Equals, `NoSuchKey`)
719                                 c.Check(err, check.ErrorMatches, `The specified key does not exist.`, check.Commentf("GET %q should return 404", objname))
720                         }
721                 }()
722         }
723         wg.Wait()
724 }
725
726 func (stage *s3stage) writeBigDirs(c *check.C, dirs int, filesPerDir int) {
727         fs, err := stage.coll.FileSystem(stage.arv, stage.kc)
728         c.Assert(err, check.IsNil)
729         for d := 0; d < dirs; d++ {
730                 dir := fmt.Sprintf("dir%d", d)
731                 c.Assert(fs.Mkdir(dir, 0755), check.IsNil)
732                 for i := 0; i < filesPerDir; i++ {
733                         f, err := fs.OpenFile(fmt.Sprintf("%s/file%d.txt", dir, i), os.O_CREATE|os.O_WRONLY, 0644)
734                         c.Assert(err, check.IsNil)
735                         c.Assert(f.Close(), check.IsNil)
736                 }
737         }
738         c.Assert(fs.Sync(), check.IsNil)
739 }
740
741 func (s *IntegrationSuite) sign(c *check.C, req *http.Request, key, secret string) {
742         scope := "20200202/zzzzz/service/aws4_request"
743         signedHeaders := "date"
744         req.Header.Set("Date", time.Now().UTC().Format(time.RFC1123))
745         stringToSign, err := s3stringToSign(s3SignAlgorithm, scope, signedHeaders, req)
746         c.Assert(err, check.IsNil)
747         sig, err := s3signature(secret, scope, signedHeaders, stringToSign)
748         c.Assert(err, check.IsNil)
749         req.Header.Set("Authorization", s3SignAlgorithm+" Credential="+key+"/"+scope+", SignedHeaders="+signedHeaders+", Signature="+sig)
750 }
751
752 func (s *IntegrationSuite) TestS3VirtualHostStyleRequests(c *check.C) {
753         stage := s.s3setup(c)
754         defer stage.teardown(c)
755         for _, trial := range []struct {
756                 url            string
757                 method         string
758                 body           string
759                 responseCode   int
760                 responseRegexp []string
761                 checkEtag      bool
762         }{
763                 {
764                         url:            "https://" + stage.collbucket.Name + ".example.com/",
765                         method:         "GET",
766                         responseCode:   http.StatusOK,
767                         responseRegexp: []string{`(?ms).*sailboat\.txt.*`},
768                 },
769                 {
770                         url:            "https://" + strings.Replace(stage.coll.PortableDataHash, "+", "-", -1) + ".example.com/",
771                         method:         "GET",
772                         responseCode:   http.StatusOK,
773                         responseRegexp: []string{`(?ms).*sailboat\.txt.*`},
774                 },
775                 {
776                         url:            "https://" + stage.projbucket.Name + ".example.com/?prefix=" + stage.coll.Name + "/&delimiter=/",
777                         method:         "GET",
778                         responseCode:   http.StatusOK,
779                         responseRegexp: []string{`(?ms).*sailboat\.txt.*`},
780                 },
781                 {
782                         url:            "https://" + stage.projbucket.Name + ".example.com/" + stage.coll.Name + "/sailboat.txt",
783                         method:         "GET",
784                         responseCode:   http.StatusOK,
785                         responseRegexp: []string{`⛵\n`},
786                         checkEtag:      true,
787                 },
788                 {
789                         url:          "https://" + stage.projbucket.Name + ".example.com/" + stage.coll.Name + "/beep",
790                         method:       "PUT",
791                         body:         "boop",
792                         responseCode: http.StatusOK,
793                 },
794                 {
795                         url:            "https://" + stage.projbucket.Name + ".example.com/" + stage.coll.Name + "/beep",
796                         method:         "GET",
797                         responseCode:   http.StatusOK,
798                         responseRegexp: []string{`boop`},
799                         checkEtag:      true,
800                 },
801                 {
802                         url:          "https://" + stage.projbucket.Name + ".example.com/" + stage.coll.Name + "//boop",
803                         method:       "GET",
804                         responseCode: http.StatusNotFound,
805                 },
806                 {
807                         url:          "https://" + stage.projbucket.Name + ".example.com/" + stage.coll.Name + "//boop",
808                         method:       "PUT",
809                         body:         "boop",
810                         responseCode: http.StatusOK,
811                 },
812                 {
813                         url:            "https://" + stage.projbucket.Name + ".example.com/" + stage.coll.Name + "//boop",
814                         method:         "GET",
815                         responseCode:   http.StatusOK,
816                         responseRegexp: []string{`boop`},
817                         checkEtag:      true,
818                 },
819         } {
820                 url, err := url.Parse(trial.url)
821                 c.Assert(err, check.IsNil)
822                 req, err := http.NewRequest(trial.method, url.String(), bytes.NewReader([]byte(trial.body)))
823                 c.Assert(err, check.IsNil)
824                 s.sign(c, req, arvadostest.ActiveTokenUUID, arvadostest.ActiveToken)
825                 rr := httptest.NewRecorder()
826                 s.handler.ServeHTTP(rr, req)
827                 resp := rr.Result()
828                 c.Check(resp.StatusCode, check.Equals, trial.responseCode)
829                 body, err := ioutil.ReadAll(resp.Body)
830                 c.Assert(err, check.IsNil)
831                 for _, re := range trial.responseRegexp {
832                         c.Check(string(body), check.Matches, re)
833                 }
834                 if trial.checkEtag {
835                         c.Check(resp.Header.Get("Etag"), check.Matches, `"[\da-f]{32}\+\d+"`)
836                 }
837         }
838 }
839
840 func (s *IntegrationSuite) TestS3NormalizeURIForSignature(c *check.C) {
841         stage := s.s3setup(c)
842         defer stage.teardown(c)
843         for _, trial := range []struct {
844                 rawPath        string
845                 normalizedPath string
846         }{
847                 {"/foo", "/foo"},                           // boring case
848                 {"/foo%5fbar", "/foo_bar"},                 // _ must not be escaped
849                 {"/foo%2fbar", "/foo/bar"},                 // / must not be escaped
850                 {"/(foo)/[];,", "/%28foo%29/%5B%5D%3B%2C"}, // ()[];, must be escaped
851                 {"/foo%5bbar", "/foo%5Bbar"},               // %XX must be uppercase
852                 // unicode chars must be UTF-8 encoded and escaped
853                 {"/\u26f5", "/%E2%9B%B5"},
854                 // "//" and "///" must not be squashed -- see example,
855                 // https://docs.aws.amazon.com/AmazonS3/latest/API/sig-v4-header-based-auth.html
856                 {"//foo///.bar", "//foo///.bar"},
857         } {
858                 c.Logf("trial %q", trial)
859
860                 date := time.Now().UTC().Format("20060102T150405Z")
861                 scope := "20200202/zzzzz/S3/aws4_request"
862                 canonicalRequest := fmt.Sprintf("%s\n%s\n%s\n%s\n%s\n%s", "GET", trial.normalizedPath, "", "host:host.example.com\n", "host", "")
863                 c.Logf("canonicalRequest %q", canonicalRequest)
864                 expect := fmt.Sprintf("%s\n%s\n%s\n%s", s3SignAlgorithm, date, scope, hashdigest(sha256.New(), canonicalRequest))
865                 c.Logf("expected stringToSign %q", expect)
866
867                 req, err := http.NewRequest("GET", "https://host.example.com"+trial.rawPath, nil)
868                 req.Header.Set("X-Amz-Date", date)
869                 req.Host = "host.example.com"
870                 c.Assert(err, check.IsNil)
871
872                 obtained, err := s3stringToSign(s3SignAlgorithm, scope, "host", req)
873                 if !c.Check(err, check.IsNil) {
874                         continue
875                 }
876                 c.Check(obtained, check.Equals, expect)
877         }
878 }
879
880 func (s *IntegrationSuite) TestS3GetBucketLocation(c *check.C) {
881         stage := s.s3setup(c)
882         defer stage.teardown(c)
883         for _, bucket := range []*s3.Bucket{stage.collbucket, stage.projbucket} {
884                 req, err := http.NewRequest("GET", bucket.URL("/"), nil)
885                 c.Check(err, check.IsNil)
886                 req.Header.Set("Authorization", "AWS "+arvadostest.ActiveTokenV2+":none")
887                 req.URL.RawQuery = "location"
888                 resp, err := http.DefaultClient.Do(req)
889                 c.Assert(err, check.IsNil)
890                 c.Check(resp.Header.Get("Content-Type"), check.Equals, "application/xml")
891                 buf, err := ioutil.ReadAll(resp.Body)
892                 c.Assert(err, check.IsNil)
893                 c.Check(string(buf), check.Equals, "<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<LocationConstraint><LocationConstraint xmlns=\"http://s3.amazonaws.com/doc/2006-03-01/\">zzzzz</LocationConstraint></LocationConstraint>\n")
894         }
895 }
896
897 func (s *IntegrationSuite) TestS3GetBucketVersioning(c *check.C) {
898         stage := s.s3setup(c)
899         defer stage.teardown(c)
900         for _, bucket := range []*s3.Bucket{stage.collbucket, stage.projbucket} {
901                 req, err := http.NewRequest("GET", bucket.URL("/"), nil)
902                 c.Check(err, check.IsNil)
903                 req.Header.Set("Authorization", "AWS "+arvadostest.ActiveTokenV2+":none")
904                 req.URL.RawQuery = "versioning"
905                 resp, err := http.DefaultClient.Do(req)
906                 c.Assert(err, check.IsNil)
907                 c.Check(resp.Header.Get("Content-Type"), check.Equals, "application/xml")
908                 buf, err := ioutil.ReadAll(resp.Body)
909                 c.Assert(err, check.IsNil)
910                 c.Check(string(buf), check.Equals, "<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<VersioningConfiguration xmlns=\"http://s3.amazonaws.com/doc/2006-03-01/\"/>\n")
911         }
912 }
913
914 func (s *IntegrationSuite) TestS3UnsupportedAPIs(c *check.C) {
915         stage := s.s3setup(c)
916         defer stage.teardown(c)
917         for _, trial := range []struct {
918                 method   string
919                 path     string
920                 rawquery string
921         }{
922                 {"GET", "/", "acl&versionId=1234"},    // GetBucketAcl
923                 {"GET", "/foo", "acl&versionId=1234"}, // GetObjectAcl
924                 {"PUT", "/", "acl"},                   // PutBucketAcl
925                 {"PUT", "/foo", "acl"},                // PutObjectAcl
926                 {"DELETE", "/", "tagging"},            // DeleteBucketTagging
927                 {"DELETE", "/foo", "tagging"},         // DeleteObjectTagging
928         } {
929                 for _, bucket := range []*s3.Bucket{stage.collbucket, stage.projbucket} {
930                         c.Logf("trial %v bucket %v", trial, bucket)
931                         req, err := http.NewRequest(trial.method, bucket.URL(trial.path), nil)
932                         c.Check(err, check.IsNil)
933                         req.Header.Set("Authorization", "AWS "+arvadostest.ActiveTokenV2+":none")
934                         req.URL.RawQuery = trial.rawquery
935                         resp, err := http.DefaultClient.Do(req)
936                         c.Assert(err, check.IsNil)
937                         c.Check(resp.Header.Get("Content-Type"), check.Equals, "application/xml")
938                         buf, err := ioutil.ReadAll(resp.Body)
939                         c.Assert(err, check.IsNil)
940                         c.Check(string(buf), check.Matches, "(?ms).*InvalidRequest.*API not supported.*")
941                 }
942         }
943 }
944
945 // If there are no CommonPrefixes entries, the CommonPrefixes XML tag
946 // should not appear at all.
947 func (s *IntegrationSuite) TestS3ListNoCommonPrefixes(c *check.C) {
948         stage := s.s3setup(c)
949         defer stage.teardown(c)
950
951         req, err := http.NewRequest("GET", stage.collbucket.URL("/"), nil)
952         c.Assert(err, check.IsNil)
953         req.Header.Set("Authorization", "AWS "+arvadostest.ActiveTokenV2+":none")
954         req.URL.RawQuery = "prefix=asdfasdfasdf&delimiter=/"
955         resp, err := http.DefaultClient.Do(req)
956         c.Assert(err, check.IsNil)
957         buf, err := ioutil.ReadAll(resp.Body)
958         c.Assert(err, check.IsNil)
959         c.Check(string(buf), check.Not(check.Matches), `(?ms).*CommonPrefixes.*`)
960 }
961
962 // If there is no delimiter in the request, or the results are not
963 // truncated, the NextMarker XML tag should not appear in the response
964 // body.
965 func (s *IntegrationSuite) TestS3ListNoNextMarker(c *check.C) {
966         stage := s.s3setup(c)
967         defer stage.teardown(c)
968
969         for _, query := range []string{"prefix=e&delimiter=/", ""} {
970                 req, err := http.NewRequest("GET", stage.collbucket.URL("/"), nil)
971                 c.Assert(err, check.IsNil)
972                 req.Header.Set("Authorization", "AWS "+arvadostest.ActiveTokenV2+":none")
973                 req.URL.RawQuery = query
974                 resp, err := http.DefaultClient.Do(req)
975                 c.Assert(err, check.IsNil)
976                 buf, err := ioutil.ReadAll(resp.Body)
977                 c.Assert(err, check.IsNil)
978                 c.Check(string(buf), check.Not(check.Matches), `(?ms).*NextMarker.*`)
979         }
980 }
981
982 // List response should include KeyCount field.
983 func (s *IntegrationSuite) TestS3ListKeyCount(c *check.C) {
984         stage := s.s3setup(c)
985         defer stage.teardown(c)
986
987         req, err := http.NewRequest("GET", stage.collbucket.URL("/"), nil)
988         c.Assert(err, check.IsNil)
989         req.Header.Set("Authorization", "AWS "+arvadostest.ActiveTokenV2+":none")
990         req.URL.RawQuery = "prefix=&delimiter=/"
991         resp, err := http.DefaultClient.Do(req)
992         c.Assert(err, check.IsNil)
993         buf, err := ioutil.ReadAll(resp.Body)
994         c.Assert(err, check.IsNil)
995         c.Check(string(buf), check.Matches, `(?ms).*<KeyCount>2</KeyCount>.*`)
996 }
997
998 func (s *IntegrationSuite) TestS3CollectionList(c *check.C) {
999         stage := s.s3setup(c)
1000         defer stage.teardown(c)
1001
1002         var markers int
1003         for markers, s.handler.Cluster.Collections.S3FolderObjects = range []bool{false, true} {
1004                 dirs := 2000
1005                 filesPerDir := 2
1006                 stage.writeBigDirs(c, dirs, filesPerDir)
1007                 // Total # objects is:
1008                 //                 2 file entries from s3setup (emptyfile and sailboat.txt)
1009                 //                +1 fake "directory" marker from s3setup (emptydir) (if enabled)
1010                 //             +dirs fake "directory" marker from writeBigDirs (dir0/, dir1/) (if enabled)
1011                 // +filesPerDir*dirs file entries from writeBigDirs (dir0/file0.txt, etc.)
1012                 s.testS3List(c, stage.collbucket, "", 4000, markers+2+(filesPerDir+markers)*dirs)
1013                 s.testS3List(c, stage.collbucket, "", 131, markers+2+(filesPerDir+markers)*dirs)
1014                 s.testS3List(c, stage.collbucket, "", 51, markers+2+(filesPerDir+markers)*dirs)
1015                 s.testS3List(c, stage.collbucket, "dir0/", 71, filesPerDir+markers)
1016         }
1017 }
1018 func (s *IntegrationSuite) testS3List(c *check.C, bucket *s3.Bucket, prefix string, pageSize, expectFiles int) {
1019         c.Logf("testS3List: prefix=%q pageSize=%d S3FolderObjects=%v", prefix, pageSize, s.handler.Cluster.Collections.S3FolderObjects)
1020         expectPageSize := pageSize
1021         if expectPageSize > 1000 {
1022                 expectPageSize = 1000
1023         }
1024         gotKeys := map[string]s3.Key{}
1025         nextMarker := ""
1026         pages := 0
1027         for {
1028                 resp, err := bucket.List(prefix, "", nextMarker, pageSize)
1029                 if !c.Check(err, check.IsNil) {
1030                         break
1031                 }
1032                 c.Check(len(resp.Contents) <= expectPageSize, check.Equals, true)
1033                 if pages++; !c.Check(pages <= (expectFiles/expectPageSize)+1, check.Equals, true) {
1034                         break
1035                 }
1036                 for _, key := range resp.Contents {
1037                         if _, dup := gotKeys[key.Key]; dup {
1038                                 c.Errorf("got duplicate key %q on page %d", key.Key, pages)
1039                         }
1040                         gotKeys[key.Key] = key
1041                         if strings.Contains(key.Key, "sailboat.txt") {
1042                                 c.Check(key.Size, check.Equals, int64(4))
1043                         }
1044                 }
1045                 if !resp.IsTruncated {
1046                         c.Check(resp.NextMarker, check.Equals, "")
1047                         break
1048                 }
1049                 if !c.Check(resp.NextMarker, check.Not(check.Equals), "") {
1050                         break
1051                 }
1052                 nextMarker = resp.NextMarker
1053         }
1054         if !c.Check(len(gotKeys), check.Equals, expectFiles) {
1055                 var sorted []string
1056                 for k := range gotKeys {
1057                         sorted = append(sorted, k)
1058                 }
1059                 sort.Strings(sorted)
1060                 for _, k := range sorted {
1061                         c.Logf("got %s", k)
1062                 }
1063         }
1064 }
1065
1066 func (s *IntegrationSuite) TestS3CollectionListRollup(c *check.C) {
1067         for _, s.handler.Cluster.Collections.S3FolderObjects = range []bool{false, true} {
1068                 s.testS3CollectionListRollup(c)
1069         }
1070 }
1071
1072 func (s *IntegrationSuite) testS3CollectionListRollup(c *check.C) {
1073         stage := s.s3setup(c)
1074         defer stage.teardown(c)
1075
1076         dirs := 2
1077         filesPerDir := 500
1078         stage.writeBigDirs(c, dirs, filesPerDir)
1079         err := stage.collbucket.PutReader("dingbats", &bytes.Buffer{}, 0, "application/octet-stream", s3.Private, s3.Options{})
1080         c.Assert(err, check.IsNil)
1081         var allfiles []string
1082         for marker := ""; ; {
1083                 resp, err := stage.collbucket.List("", "", marker, 20000)
1084                 c.Check(err, check.IsNil)
1085                 for _, key := range resp.Contents {
1086                         if len(allfiles) == 0 || allfiles[len(allfiles)-1] != key.Key {
1087                                 allfiles = append(allfiles, key.Key)
1088                         }
1089                 }
1090                 marker = resp.NextMarker
1091                 if marker == "" {
1092                         break
1093                 }
1094         }
1095         markers := 0
1096         if s.handler.Cluster.Collections.S3FolderObjects {
1097                 markers = 1
1098         }
1099         c.Check(allfiles, check.HasLen, dirs*(filesPerDir+markers)+3+markers)
1100
1101         gotDirMarker := map[string]bool{}
1102         for _, name := range allfiles {
1103                 isDirMarker := strings.HasSuffix(name, "/")
1104                 if markers == 0 {
1105                         c.Check(isDirMarker, check.Equals, false, check.Commentf("name %q", name))
1106                 } else if isDirMarker {
1107                         gotDirMarker[name] = true
1108                 } else if i := strings.LastIndex(name, "/"); i >= 0 {
1109                         c.Check(gotDirMarker[name[:i+1]], check.Equals, true, check.Commentf("name %q", name))
1110                         gotDirMarker[name[:i+1]] = true // skip redundant complaints about this dir marker
1111                 }
1112         }
1113
1114         for _, trial := range []struct {
1115                 prefix    string
1116                 delimiter string
1117                 marker    string
1118         }{
1119                 {"", "", ""},
1120                 {"di", "/", ""},
1121                 {"di", "r", ""},
1122                 {"di", "n", ""},
1123                 {"dir0", "/", ""},
1124                 {"dir0/", "/", ""},
1125                 {"dir0/f", "/", ""},
1126                 {"dir0", "", ""},
1127                 {"dir0/", "", ""},
1128                 {"dir0/f", "", ""},
1129                 {"dir0", "/", "dir0/file14.txt"},       // one commonprefix, "dir0/"
1130                 {"dir0", "/", "dir0/zzzzfile.txt"},     // no commonprefixes
1131                 {"", "", "dir0/file14.txt"},            // middle page, skip walking dir1
1132                 {"", "", "dir1/file14.txt"},            // middle page, skip walking dir0
1133                 {"", "", "dir1/file498.txt"},           // last page of results
1134                 {"dir1/file", "", "dir1/file498.txt"},  // last page of results, with prefix
1135                 {"dir1/file", "/", "dir1/file498.txt"}, // last page of results, with prefix + delimiter
1136                 {"dir1", "Z", "dir1/file498.txt"},      // delimiter "Z" never appears
1137                 {"dir2", "/", ""},                      // prefix "dir2" does not exist
1138                 {"", "/", ""},
1139         } {
1140                 c.Logf("\n\n=== trial %+v markers=%d", trial, markers)
1141
1142                 maxKeys := 20
1143                 resp, err := stage.collbucket.List(trial.prefix, trial.delimiter, trial.marker, maxKeys)
1144                 c.Check(err, check.IsNil)
1145                 if resp.IsTruncated && trial.delimiter == "" {
1146                         // goamz List method fills in the missing
1147                         // NextMarker field if resp.IsTruncated, so
1148                         // now we can't really tell whether it was
1149                         // sent by the server or by goamz. In cases
1150                         // where it should be empty but isn't, assume
1151                         // it's goamz's fault.
1152                         resp.NextMarker = ""
1153                 }
1154
1155                 var expectKeys []string
1156                 var expectPrefixes []string
1157                 var expectNextMarker string
1158                 var expectTruncated bool
1159                 for _, key := range allfiles {
1160                         full := len(expectKeys)+len(expectPrefixes) >= maxKeys
1161                         if !strings.HasPrefix(key, trial.prefix) || key <= trial.marker {
1162                                 continue
1163                         } else if idx := strings.Index(key[len(trial.prefix):], trial.delimiter); trial.delimiter != "" && idx >= 0 {
1164                                 prefix := key[:len(trial.prefix)+idx+1]
1165                                 if len(expectPrefixes) > 0 && expectPrefixes[len(expectPrefixes)-1] == prefix {
1166                                         // same prefix as previous key
1167                                 } else if full {
1168                                         expectTruncated = true
1169                                 } else {
1170                                         expectPrefixes = append(expectPrefixes, prefix)
1171                                         expectNextMarker = prefix
1172                                 }
1173                         } else if full {
1174                                 expectTruncated = true
1175                                 break
1176                         } else {
1177                                 expectKeys = append(expectKeys, key)
1178                                 if trial.delimiter != "" {
1179                                         expectNextMarker = key
1180                                 }
1181                         }
1182                 }
1183                 if !expectTruncated {
1184                         expectNextMarker = ""
1185                 }
1186
1187                 var gotKeys []string
1188                 for _, key := range resp.Contents {
1189                         gotKeys = append(gotKeys, key.Key)
1190                 }
1191                 var gotPrefixes []string
1192                 for _, prefix := range resp.CommonPrefixes {
1193                         gotPrefixes = append(gotPrefixes, prefix)
1194                 }
1195                 commentf := check.Commentf("trial %+v markers=%d", trial, markers)
1196                 c.Check(gotKeys, check.DeepEquals, expectKeys, commentf)
1197                 c.Check(gotPrefixes, check.DeepEquals, expectPrefixes, commentf)
1198                 c.Check(resp.NextMarker, check.Equals, expectNextMarker, commentf)
1199                 c.Check(resp.IsTruncated, check.Equals, expectTruncated, commentf)
1200                 c.Logf("=== trial %+v keys %q prefixes %q nextMarker %q", trial, gotKeys, gotPrefixes, resp.NextMarker)
1201         }
1202 }
1203
1204 func (s *IntegrationSuite) TestS3ListObjectsV2ManySubprojects(c *check.C) {
1205         stage := s.s3setup(c)
1206         defer stage.teardown(c)
1207         projects := 50
1208         collectionsPerProject := 2
1209         for i := 0; i < projects; i++ {
1210                 var subproj arvados.Group
1211                 err := stage.arv.RequestAndDecode(&subproj, "POST", "arvados/v1/groups", nil, map[string]interface{}{
1212                         "group": map[string]interface{}{
1213                                 "owner_uuid":  stage.subproj.UUID,
1214                                 "group_class": "project",
1215                                 "name":        fmt.Sprintf("keep-web s3 test subproject %d", i),
1216                         },
1217                 })
1218                 c.Assert(err, check.IsNil)
1219                 for j := 0; j < collectionsPerProject; j++ {
1220                         err = stage.arv.RequestAndDecode(nil, "POST", "arvados/v1/collections", nil, map[string]interface{}{"collection": map[string]interface{}{
1221                                 "owner_uuid":    subproj.UUID,
1222                                 "name":          fmt.Sprintf("keep-web s3 test collection %d", j),
1223                                 "manifest_text": ". d41d8cd98f00b204e9800998ecf8427e+0 0:0:emptyfile\n./emptydir d41d8cd98f00b204e9800998ecf8427e+0 0:0:.\n",
1224                         }})
1225                         c.Assert(err, check.IsNil)
1226                 }
1227         }
1228         c.Logf("setup complete")
1229
1230         sess := aws_session.Must(aws_session.NewSession(&aws_aws.Config{
1231                 Region:           aws_aws.String("auto"),
1232                 Endpoint:         aws_aws.String(s.testServer.URL),
1233                 Credentials:      aws_credentials.NewStaticCredentials(url.QueryEscape(arvadostest.ActiveTokenV2), url.QueryEscape(arvadostest.ActiveTokenV2), ""),
1234                 S3ForcePathStyle: aws_aws.Bool(true),
1235         }))
1236         client := aws_s3.New(sess)
1237         ctx := context.Background()
1238         params := aws_s3.ListObjectsV2Input{
1239                 Bucket:    aws_aws.String(stage.proj.UUID),
1240                 Delimiter: aws_aws.String("/"),
1241                 Prefix:    aws_aws.String("keep-web s3 test subproject/"),
1242                 MaxKeys:   aws_aws.Int64(int64(projects / 2)),
1243         }
1244         for page := 1; ; page++ {
1245                 t0 := time.Now()
1246                 result, err := client.ListObjectsV2WithContext(ctx, &params)
1247                 if !c.Check(err, check.IsNil) {
1248                         break
1249                 }
1250                 c.Logf("got page %d in %v with len(Contents) == %d, len(CommonPrefixes) == %d", page, time.Since(t0), len(result.Contents), len(result.CommonPrefixes))
1251                 if !*result.IsTruncated {
1252                         break
1253                 }
1254                 params.ContinuationToken = result.NextContinuationToken
1255                 *params.MaxKeys = *params.MaxKeys/2 + 1
1256         }
1257 }
1258
1259 func (s *IntegrationSuite) TestS3ListObjectsV2(c *check.C) {
1260         stage := s.s3setup(c)
1261         defer stage.teardown(c)
1262         dirs := 2
1263         filesPerDir := 40
1264         stage.writeBigDirs(c, dirs, filesPerDir)
1265
1266         sess := aws_session.Must(aws_session.NewSession(&aws_aws.Config{
1267                 Region:           aws_aws.String("auto"),
1268                 Endpoint:         aws_aws.String(s.testServer.URL),
1269                 Credentials:      aws_credentials.NewStaticCredentials(url.QueryEscape(arvadostest.ActiveTokenV2), url.QueryEscape(arvadostest.ActiveTokenV2), ""),
1270                 S3ForcePathStyle: aws_aws.Bool(true),
1271         }))
1272
1273         stringOrNil := func(s string) *string {
1274                 if s == "" {
1275                         return nil
1276                 } else {
1277                         return &s
1278                 }
1279         }
1280
1281         client := aws_s3.New(sess)
1282         ctx := context.Background()
1283
1284         for _, trial := range []struct {
1285                 prefix               string
1286                 delimiter            string
1287                 startAfter           string
1288                 maxKeys              int
1289                 expectKeys           int
1290                 expectCommonPrefixes map[string]bool
1291         }{
1292                 {
1293                         // Expect {filesPerDir plus the dir itself}
1294                         // for each dir, plus emptydir, emptyfile, and
1295                         // sailboat.txt.
1296                         expectKeys: (filesPerDir+1)*dirs + 3,
1297                 },
1298                 {
1299                         maxKeys:    15,
1300                         expectKeys: (filesPerDir+1)*dirs + 3,
1301                 },
1302                 {
1303                         startAfter: "dir0/z",
1304                         maxKeys:    15,
1305                         // Expect {filesPerDir plus the dir itself}
1306                         // for each dir except dir0, plus emptydir,
1307                         // emptyfile, and sailboat.txt.
1308                         expectKeys: (filesPerDir+1)*(dirs-1) + 3,
1309                 },
1310                 {
1311                         maxKeys:              1,
1312                         delimiter:            "/",
1313                         expectKeys:           2, // emptyfile, sailboat.txt
1314                         expectCommonPrefixes: map[string]bool{"dir0/": true, "dir1/": true, "emptydir/": true},
1315                 },
1316                 {
1317                         startAfter:           "dir0/z",
1318                         maxKeys:              15,
1319                         delimiter:            "/",
1320                         expectKeys:           2, // emptyfile, sailboat.txt
1321                         expectCommonPrefixes: map[string]bool{"dir1/": true, "emptydir/": true},
1322                 },
1323                 {
1324                         startAfter:           "dir0/file10.txt",
1325                         maxKeys:              15,
1326                         delimiter:            "/",
1327                         expectKeys:           2,
1328                         expectCommonPrefixes: map[string]bool{"dir0/": true, "dir1/": true, "emptydir/": true},
1329                 },
1330                 {
1331                         startAfter:           "dir0/file10.txt",
1332                         maxKeys:              15,
1333                         prefix:               "d",
1334                         delimiter:            "/",
1335                         expectKeys:           0,
1336                         expectCommonPrefixes: map[string]bool{"dir0/": true, "dir1/": true},
1337                 },
1338         } {
1339                 c.Logf("[trial %+v]", trial)
1340                 params := aws_s3.ListObjectsV2Input{
1341                         Bucket:     aws_aws.String(stage.collbucket.Name),
1342                         Prefix:     stringOrNil(trial.prefix),
1343                         Delimiter:  stringOrNil(trial.delimiter),
1344                         StartAfter: stringOrNil(trial.startAfter),
1345                         MaxKeys:    aws_aws.Int64(int64(trial.maxKeys)),
1346                 }
1347                 keySeen := map[string]bool{}
1348                 prefixSeen := map[string]bool{}
1349                 for {
1350                         result, err := client.ListObjectsV2WithContext(ctx, &params)
1351                         if !c.Check(err, check.IsNil) {
1352                                 break
1353                         }
1354                         c.Check(result.Name, check.DeepEquals, aws_aws.String(stage.collbucket.Name))
1355                         c.Check(result.Prefix, check.DeepEquals, aws_aws.String(trial.prefix))
1356                         c.Check(result.Delimiter, check.DeepEquals, aws_aws.String(trial.delimiter))
1357                         // The following two fields are expected to be
1358                         // nil (i.e., no tag in XML response) rather
1359                         // than "" when the corresponding request
1360                         // field was empty or nil.
1361                         c.Check(result.StartAfter, check.DeepEquals, stringOrNil(trial.startAfter))
1362                         c.Check(result.ContinuationToken, check.DeepEquals, params.ContinuationToken)
1363
1364                         if trial.maxKeys > 0 {
1365                                 c.Check(result.MaxKeys, check.DeepEquals, aws_aws.Int64(int64(trial.maxKeys)))
1366                                 c.Check(len(result.Contents)+len(result.CommonPrefixes) <= trial.maxKeys, check.Equals, true)
1367                         } else {
1368                                 c.Check(result.MaxKeys, check.DeepEquals, aws_aws.Int64(int64(s3MaxKeys)))
1369                         }
1370
1371                         for _, ent := range result.Contents {
1372                                 c.Assert(ent.Key, check.NotNil)
1373                                 c.Check(*ent.Key > trial.startAfter, check.Equals, true)
1374                                 c.Check(keySeen[*ent.Key], check.Equals, false, check.Commentf("dup key %q", *ent.Key))
1375                                 keySeen[*ent.Key] = true
1376                         }
1377                         for _, ent := range result.CommonPrefixes {
1378                                 c.Assert(ent.Prefix, check.NotNil)
1379                                 c.Check(strings.HasSuffix(*ent.Prefix, trial.delimiter), check.Equals, true, check.Commentf("bad CommonPrefix %q", *ent.Prefix))
1380                                 if strings.HasPrefix(trial.startAfter, *ent.Prefix) {
1381                                         // If we asked for
1382                                         // startAfter=dir0/file10.txt,
1383                                         // we expect dir0/ to be
1384                                         // returned as a common prefix
1385                                 } else {
1386                                         c.Check(*ent.Prefix > trial.startAfter, check.Equals, true)
1387                                 }
1388                                 c.Check(prefixSeen[*ent.Prefix], check.Equals, false, check.Commentf("dup common prefix %q", *ent.Prefix))
1389                                 prefixSeen[*ent.Prefix] = true
1390                         }
1391                         if *result.IsTruncated && c.Check(result.NextContinuationToken, check.Not(check.Equals), "") {
1392                                 params.ContinuationToken = aws_aws.String(*result.NextContinuationToken)
1393                         } else {
1394                                 break
1395                         }
1396                 }
1397                 c.Check(keySeen, check.HasLen, trial.expectKeys)
1398                 c.Check(prefixSeen, check.HasLen, len(trial.expectCommonPrefixes))
1399                 if len(trial.expectCommonPrefixes) > 0 {
1400                         c.Check(prefixSeen, check.DeepEquals, trial.expectCommonPrefixes)
1401                 }
1402         }
1403 }
1404
1405 func (s *IntegrationSuite) TestS3ListObjectsV2EncodingTypeURL(c *check.C) {
1406         stage := s.s3setup(c)
1407         defer stage.teardown(c)
1408         dirs := 2
1409         filesPerDir := 40
1410         stage.writeBigDirs(c, dirs, filesPerDir)
1411
1412         sess := aws_session.Must(aws_session.NewSession(&aws_aws.Config{
1413                 Region:           aws_aws.String("auto"),
1414                 Endpoint:         aws_aws.String(s.testServer.URL),
1415                 Credentials:      aws_credentials.NewStaticCredentials(url.QueryEscape(arvadostest.ActiveTokenV2), url.QueryEscape(arvadostest.ActiveTokenV2), ""),
1416                 S3ForcePathStyle: aws_aws.Bool(true),
1417         }))
1418
1419         client := aws_s3.New(sess)
1420         ctx := context.Background()
1421
1422         result, err := client.ListObjectsV2WithContext(ctx, &aws_s3.ListObjectsV2Input{
1423                 Bucket:       aws_aws.String(stage.collbucket.Name),
1424                 Prefix:       aws_aws.String("dir0/"),
1425                 Delimiter:    aws_aws.String("/"),
1426                 StartAfter:   aws_aws.String("dir0/"),
1427                 EncodingType: aws_aws.String("url"),
1428         })
1429         c.Assert(err, check.IsNil)
1430         c.Check(*result.Prefix, check.Equals, "dir0%2F")
1431         c.Check(*result.Delimiter, check.Equals, "%2F")
1432         c.Check(*result.StartAfter, check.Equals, "dir0%2F")
1433         for _, ent := range result.Contents {
1434                 c.Check(*ent.Key, check.Matches, "dir0%2F.*")
1435         }
1436         result, err = client.ListObjectsV2WithContext(ctx, &aws_s3.ListObjectsV2Input{
1437                 Bucket:       aws_aws.String(stage.collbucket.Name),
1438                 Delimiter:    aws_aws.String("/"),
1439                 EncodingType: aws_aws.String("url"),
1440         })
1441         c.Assert(err, check.IsNil)
1442         c.Check(*result.Delimiter, check.Equals, "%2F")
1443         c.Check(result.CommonPrefixes, check.HasLen, dirs+1)
1444         for _, ent := range result.CommonPrefixes {
1445                 c.Check(*ent.Prefix, check.Matches, ".*%2F")
1446         }
1447 }
1448
1449 // TestS3cmd checks compatibility with the s3cmd command line tool, if
1450 // it's installed. As of Debian buster, s3cmd is only in backports, so
1451 // `arvados-server install` don't install it, and this test skips if
1452 // it's not installed.
1453 func (s *IntegrationSuite) TestS3cmd(c *check.C) {
1454         if _, err := exec.LookPath("s3cmd"); err != nil {
1455                 c.Skip("s3cmd not found")
1456                 return
1457         }
1458
1459         stage := s.s3setup(c)
1460         defer stage.teardown(c)
1461
1462         cmd := exec.Command("s3cmd", "--no-ssl", "--host="+s.testServer.URL[7:], "--host-bucket="+s.testServer.URL[7:], "--access_key="+arvadostest.ActiveTokenUUID, "--secret_key="+arvadostest.ActiveToken, "ls", "s3://"+arvadostest.FooCollection)
1463         buf, err := cmd.CombinedOutput()
1464         c.Check(err, check.IsNil)
1465         c.Check(string(buf), check.Matches, `.* 3 +s3://`+arvadostest.FooCollection+`/foo\n`)
1466
1467         // This tests whether s3cmd's path normalization agrees with
1468         // keep-web's signature verification wrt chars like "|"
1469         // (neither reserved nor unreserved) and "," (not normally
1470         // percent-encoded in a path).
1471         tmpfile := c.MkDir() + "/dstfile"
1472         cmd = exec.Command("s3cmd", "--no-ssl", "--host="+s.testServer.URL[7:], "--host-bucket="+s.testServer.URL[7:], "--access_key="+arvadostest.ActiveTokenUUID, "--secret_key="+arvadostest.ActiveToken, "get", "s3://"+arvadostest.FooCollection+"/foo,;$[|]bar", tmpfile)
1473         buf, err = cmd.CombinedOutput()
1474         c.Check(err, check.NotNil)
1475         // As of commit b7520e5c25e1bf25c1a8bf5aa2eadb299be8f606
1476         // (between debian bullseye and bookworm versions), s3cmd
1477         // started catching the NoSuchKey error code and replacing it
1478         // with "Source object '%s' does not exist.".
1479         c.Check(string(buf), check.Matches, `(?ms).*(NoSuchKey|Source object.*does not exist).*\n`)
1480
1481         tmpfile = c.MkDir() + "/foo"
1482         cmd = exec.Command("s3cmd", "--no-ssl", "--host="+s.testServer.URL[7:], "--host-bucket="+s.testServer.URL[7:], "--access_key="+arvadostest.ActiveTokenUUID, "--secret_key="+arvadostest.ActiveToken, "get", "s3://"+arvadostest.FooCollection+"/foo", tmpfile)
1483         buf, err = cmd.CombinedOutput()
1484         c.Logf("%s", buf)
1485         if c.Check(err, check.IsNil) {
1486                 checkcontent, err := os.ReadFile(tmpfile)
1487                 c.Check(err, check.IsNil)
1488                 c.Check(string(checkcontent), check.Equals, "foo")
1489         }
1490 }
1491
1492 func (s *IntegrationSuite) TestS3BucketInHost(c *check.C) {
1493         stage := s.s3setup(c)
1494         defer stage.teardown(c)
1495
1496         hdr, body, _ := s.runCurl(c, "AWS "+arvadostest.ActiveTokenV2+":none", stage.coll.UUID+".collections.example.com", "/sailboat.txt")
1497         c.Check(hdr, check.Matches, `(?s)HTTP/1.1 200 OK\r\n.*`)
1498         c.Check(body, check.Equals, "⛵\n")
1499 }