001/* 002 * Licensed to DuraSpace under one or more contributor license agreements. 003 * See the NOTICE file distributed with this work for additional information 004 * regarding copyright ownership. 005 * 006 * DuraSpace licenses this file to you under the Apache License, 007 * Version 2.0 (the "License"); you may not use this file except in 008 * compliance with the License. You may obtain a copy of the License at 009 * 010 * http://www.apache.org/licenses/LICENSE-2.0 011 * 012 * Unless required by applicable law or agreed to in writing, software 013 * distributed under the License is distributed on an "AS IS" BASIS, 014 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 015 * See the License for the specific language governing permissions and 016 * limitations under the License. 017 */ 018package org.fcrepo.auth.webac; 019 020import org.apache.jena.graph.Triple; 021import org.apache.jena.rdf.model.Statement; 022 023import org.fcrepo.config.AuthPropsConfig; 024import org.fcrepo.kernel.api.Transaction; 025import org.fcrepo.kernel.api.exception.PathNotFoundException; 026import org.fcrepo.kernel.api.exception.PathNotFoundRuntimeException; 027import org.fcrepo.kernel.api.exception.RepositoryException; 028import org.fcrepo.kernel.api.identifiers.FedoraId; 029import org.fcrepo.kernel.api.models.FedoraResource; 030import org.fcrepo.kernel.api.models.NonRdfSourceDescription; 031import org.fcrepo.kernel.api.models.ResourceFactory; 032import org.fcrepo.kernel.api.models.TimeMap; 033import org.fcrepo.kernel.api.models.WebacAcl; 034import org.slf4j.Logger; 035import org.springframework.stereotype.Component; 036 037import javax.annotation.PostConstruct; 038import javax.inject.Inject; 039import java.net.URI; 040import java.util.ArrayList; 041import java.util.Collection; 042import java.util.HashMap; 043import java.util.HashSet; 044import java.util.List; 045import java.util.Map; 046import java.util.Objects; 047import java.util.Optional; 048import java.util.Set; 049import java.util.function.Function; 050import java.util.function.Predicate; 051import java.util.stream.Collectors; 052import java.util.stream.Stream; 053 054import static java.util.Arrays.asList; 055import static java.util.Collections.emptyList; 056import static java.util.stream.Collectors.toList; 057import static java.util.stream.Collectors.toSet; 058import static java.util.stream.IntStream.range; 059import static java.util.stream.Stream.concat; 060import static java.util.stream.Stream.empty; 061import static java.util.stream.Stream.of; 062import static org.apache.jena.graph.NodeFactory.createURI; 063import static org.fcrepo.auth.webac.URIConstants.FOAF_AGENT_VALUE; 064import static org.fcrepo.auth.webac.URIConstants.VCARD_GROUP_VALUE; 065import static org.fcrepo.auth.webac.URIConstants.VCARD_MEMBER_VALUE; 066import static org.fcrepo.auth.webac.URIConstants.WEBAC_ACCESSTO_CLASS_VALUE; 067import static org.fcrepo.auth.webac.URIConstants.WEBAC_ACCESSTO_VALUE; 068import static org.fcrepo.auth.webac.URIConstants.WEBAC_AGENT_CLASS_VALUE; 069import static org.fcrepo.auth.webac.URIConstants.WEBAC_AGENT_GROUP_VALUE; 070import static org.fcrepo.auth.webac.URIConstants.WEBAC_AGENT_VALUE; 071import static org.fcrepo.auth.webac.URIConstants.WEBAC_AUTHENTICATED_AGENT_VALUE; 072import static org.fcrepo.auth.webac.URIConstants.WEBAC_AUTHORIZATION_VALUE; 073import static org.fcrepo.auth.webac.URIConstants.WEBAC_DEFAULT_VALUE; 074import static org.fcrepo.auth.webac.URIConstants.WEBAC_MODE_VALUE; 075import static org.fcrepo.auth.webac.URIConstants.WEBAC_NAMESPACE_VALUE; 076import static org.fcrepo.http.api.FedoraAcl.getDefaultAcl; 077import static org.fcrepo.kernel.api.FedoraTypes.FEDORA_ID_PREFIX; 078import static org.fcrepo.kernel.api.RdfLexicon.RDF_NAMESPACE; 079import static org.slf4j.LoggerFactory.getLogger; 080 081 082/** 083 * @author acoburn 084 * @since 9/3/15 085 */ 086@Component 087public class WebACRolesProvider { 088 089 private static final Logger LOGGER = getLogger(WebACRolesProvider.class); 090 091 private static final org.apache.jena.graph.Node RDF_TYPE_NODE = createURI(RDF_NAMESPACE + "type"); 092 private static final org.apache.jena.graph.Node VCARD_GROUP_NODE = createURI(VCARD_GROUP_VALUE); 093 private static final org.apache.jena.graph.Node VCARD_MEMBER_NODE = createURI(VCARD_MEMBER_VALUE); 094 095 @Inject 096 private AuthPropsConfig authPropsConfig; 097 098 @Inject 099 private ResourceFactory resourceFactory; 100 101 private String userBaseUri; 102 private String groupBaseUri; 103 104 @PostConstruct 105 public void setup() { 106 this.userBaseUri = authPropsConfig.getUserAgentBaseUri(); 107 this.groupBaseUri = authPropsConfig.getGroupAgentBaseUri(); 108 } 109 110 /** 111 * Get the roles assigned to this Node. 112 * 113 * @param resource the subject resource 114 * @param transaction the transaction being acted upon 115 * @return a set of roles for each principal 116 */ 117 public Map<String, Collection<String>> getRoles(final FedoraResource resource, final Transaction transaction) { 118 LOGGER.debug("Getting agent roles for: {}", resource.getId()); 119 120 // Get the effective ACL by searching the target node and any ancestors. 121 final Optional<ACLHandle> effectiveAcl = getEffectiveAcl(resource, false); 122 123 // Construct a list of acceptable acl:accessTo values for the target resource. 124 final List<String> resourcePaths = new ArrayList<>(); 125 if (resource instanceof WebacAcl) { 126 // ACLs don't describe their resource, but we still want the container which is the resource. 127 resourcePaths.add(resource.getContainer().getId()); 128 } else { 129 resourcePaths.add(resource.getDescribedResource().getId()); 130 } 131 132 // Construct a list of acceptable acl:accessToClass values for the target resource. 133 final List<URI> rdfTypes = resource.getDescription().getTypes(); 134 135 // Add the resource location and types of the ACL-bearing parent, 136 // if present and if different than the target resource. 137 effectiveAcl 138 .map(aclHandle -> aclHandle.resource) 139 .filter(effectiveResource -> !effectiveResource.getId().equals(resource.getId())) 140 .ifPresent(effectiveResource -> { 141 resourcePaths.add(effectiveResource.getId()); 142 rdfTypes.addAll(effectiveResource.getTypes()); 143 }); 144 145 // If we fall through to the system/classpath-based Authorization and it 146 // contains any acl:accessTo properties, it is necessary to add each ancestor 147 // path up the node hierarchy, starting at the resource location up to the 148 // root location. This way, the checkAccessTo predicate (below) can be properly 149 // created to match any acl:accessTo values that are part of the getDefaultAuthorization. 150 // This is not relevant if an effectiveAcl is present. 151 if (!effectiveAcl.isPresent()) { 152 resourcePaths.addAll(getAllPathAncestors(resource.getId())); 153 } 154 155 // Create a function to check acl:accessTo, scoped to the given resourcePaths 156 final Predicate<WebACAuthorization> checkAccessTo = accessTo.apply(resourcePaths); 157 158 // Create a function to check acl:accessToClass, scoped to the given rdf:type values, 159 // but transform the URIs to Strings first. 160 final Predicate<WebACAuthorization> checkAccessToClass = 161 accessToClass.apply(rdfTypes.stream().map(URI::toString).collect(toList())); 162 163 // Read the effective Acl and return a list of acl:Authorization statements 164 final List<WebACAuthorization> authorizations = effectiveAcl 165 .map(auth -> auth.authorizations) 166 .orElseGet(() -> getDefaultAuthorizations()); 167 168 // Filter the acl:Authorization statements so that they correspond only to statements that apply to 169 // the target (or acl-bearing ancestor) resource path or rdf:type. 170 // Then, assign all acceptable acl:mode values to the relevant acl:agent values: this creates a UNION 171 // of acl:modes for each particular acl:agent. 172 final Map<String, Collection<String>> effectiveRoles = new HashMap<>(); 173 authorizations.stream() 174 .filter(checkAccessTo.or(checkAccessToClass)) 175 .forEach(auth -> { 176 concat(auth.getAgents().stream(), 177 dereferenceAgentGroups(transaction, auth.getAgentGroups()).stream()) 178 .filter(agent -> !agent.equals(FOAF_AGENT_VALUE) && 179 !agent.equals(WEBAC_AUTHENTICATED_AGENT_VALUE)) 180 .forEach(agent -> { 181 effectiveRoles.computeIfAbsent(agent, key -> new HashSet<>()) 182 .addAll(auth.getModes().stream().map(URI::toString).collect(toSet())); 183 }); 184 auth.getAgentClasses().stream().filter(agentClass -> agentClass.equals(FOAF_AGENT_VALUE) || 185 agentClass.equals( 186 WEBAC_AUTHENTICATED_AGENT_VALUE)) 187 .forEach(agentClass -> { 188 effectiveRoles.computeIfAbsent(agentClass, key -> new HashSet<>()) 189 .addAll(auth.getModes().stream().map(URI::toString).collect(toSet())); 190 }); 191 }); 192 193 LOGGER.debug("Unfiltered ACL: {}", effectiveRoles); 194 195 return effectiveRoles; 196 } 197 198 /** 199 * Given a path (e.g. /a/b/c/d) retrieve a list of all ancestor paths. 200 * In this case, that would be a list of "/a/b/c", "/a/b", "/a" and "/". 201 */ 202 private static List<String> getAllPathAncestors(final String path) { 203 final List<String> segments = asList(path.replace(FEDORA_ID_PREFIX, "").split("/")); 204 return range(1, segments.size()) 205 .mapToObj(frameSize -> { 206 final var subpath = String.join("/", segments.subList(1, frameSize)); 207 return FEDORA_ID_PREFIX + (!subpath.isBlank() ? "/" : "") + subpath; 208 }) 209 .collect(toList()); 210 } 211 212 /** 213 * This is a function for generating a Predicate that filters WebACAuthorizations according 214 * to whether the given acl:accessToClass values contain any of the rdf:type values provided 215 * when creating the predicate. 216 */ 217 private static final Function<List<String>, Predicate<WebACAuthorization>> accessToClass = uris -> auth -> 218 uris.stream().anyMatch(uri -> auth.getAccessToClassURIs().contains(uri)); 219 220 /** 221 * This is a function for generating a Predicate that filters WebACAuthorizations according 222 * to whether the given acl:accessTo values contain any of the target resource values provided 223 * when creating the predicate. 224 */ 225 private static final Function<List<String>, Predicate<WebACAuthorization>> accessTo = uris -> auth -> 226 uris.stream().anyMatch(uri -> auth.getAccessToURIs().contains(uri)); 227 228 /** 229 * This maps a Collection of acl:agentGroup values to a List of agents. 230 * Any out-of-domain URIs are silently ignored. 231 */ 232 private List<String> dereferenceAgentGroups(final Transaction transaction, final Collection<String> agentGroups) { 233 final List<String> members = agentGroups.stream().flatMap(agentGroup -> { 234 if (agentGroup.startsWith(FEDORA_ID_PREFIX)) { 235 //strip off trailing hash. 236 final int hashIndex = agentGroup.indexOf("#"); 237 final String agentGroupNoHash = hashIndex > 0 ? 238 agentGroup.substring(0, hashIndex) : 239 agentGroup; 240 final String hashedSuffix = hashIndex > 0 ? agentGroup.substring(hashIndex) : null; 241 try { 242 final FedoraId fedoraId = FedoraId.create(agentGroupNoHash); 243 final FedoraResource resource = resourceFactory.getResource(transaction, fedoraId); 244 return getAgentMembers(resource, hashedSuffix); 245 } catch (final PathNotFoundException e) { 246 throw new PathNotFoundRuntimeException(e.getMessage(), e); 247 } 248 } else if (agentGroup.equals(FOAF_AGENT_VALUE)) { 249 return of(agentGroup); 250 } else { 251 LOGGER.info("Ignoring agentGroup: {}", agentGroup); 252 return empty(); 253 } 254 }).collect(toList()); 255 256 if (LOGGER.isDebugEnabled() && !agentGroups.isEmpty()) { 257 LOGGER.debug("Found {} members in {} agentGroups resources", members.size(), agentGroups.size()); 258 } 259 260 return members; 261 } 262 263 /** 264 * Given a FedoraResource, return a list of agents. 265 */ 266 private Stream<String> getAgentMembers(final FedoraResource resource, final String hashPortion) { 267 //resolve list of triples, accounting for hash-uris. 268 final List<Triple> triples = resource.getTriples().filter( 269 triple -> hashPortion == null || triple.getSubject().getURI().endsWith(hashPortion)).collect(toList()); 270 //determine if there is a rdf:type vcard:Group 271 final boolean hasVcardGroup = triples.stream().anyMatch( 272 triple -> triple.matches(triple.getSubject(), RDF_TYPE_NODE, VCARD_GROUP_NODE)); 273 //return members only if there is an associated vcard:Group 274 if (hasVcardGroup) { 275 return triples.stream() 276 .filter(triple -> triple.predicateMatches(VCARD_MEMBER_NODE)) 277 .map(Triple::getObject).flatMap(WebACRolesProvider::nodeToStringStream) 278 .map(this::stripUserAgentBaseURI); 279 } else { 280 return empty(); 281 } 282 } 283 284 private String stripUserAgentBaseURI(final String object) { 285 if (userBaseUri != null && object.startsWith(userBaseUri)) { 286 return object.substring(userBaseUri.length()); 287 } 288 return object; 289 } 290 291 /** 292 * Map a Jena Node to a Stream of Strings. Any non-URI, non-Literals map to an empty Stream, 293 * making this suitable to use with flatMap. 294 */ 295 private static Stream<String> nodeToStringStream(final org.apache.jena.graph.Node object) { 296 if (object.isURI()) { 297 return of(object.getURI()); 298 } else if (object.isLiteral()) { 299 return of(object.getLiteralValue().toString()); 300 } else { 301 return empty(); 302 } 303 } 304 305 306 /** 307 * A simple predicate for filtering out any non-acl triples. 308 */ 309 private static final Predicate<Triple> hasAclPredicate = triple -> 310 triple.getPredicate().getNameSpace().equals(WEBAC_NAMESPACE_VALUE); 311 312 /** 313 * This function reads a Fedora ACL resource and all of its acl:Authorization children. 314 * The RDF from each child resource is put into a WebACAuthorization object, and the 315 * full list is returned. 316 * 317 * @param aclResource the ACL resource 318 * @param ancestorAcl flag indicating whether or not the ACL resource associated with an ancestor of the target 319 * resource 320 * @return a list of acl:Authorization objects 321 */ 322 private List<WebACAuthorization> getAuthorizations(final FedoraResource aclResource, 323 final boolean ancestorAcl) { 324 325 final List<WebACAuthorization> authorizations = new ArrayList<>(); 326 327 if (LOGGER.isDebugEnabled()) { 328 LOGGER.debug("ACL: {}", aclResource.getId()); 329 } 330 331 if (aclResource.isAcl()) { 332 //resolve set of subjects that are of type acl:authorization 333 final List<Triple> triples = aclResource.getTriples().collect(toList()); 334 335 final Set<org.apache.jena.graph.Node> authSubjects = triples.stream().filter(t -> { 336 return t.getPredicate().getURI().equals(RDF_NAMESPACE + "type") && 337 t.getObject().getURI().equals(WEBAC_AUTHORIZATION_VALUE); 338 }).map(t -> t.getSubject()).collect(Collectors.toSet()); 339 340 // Read resource, keeping only acl-prefixed triples. 341 final Map<String, Map<String, List<String>>> authMap = new HashMap<>(); 342 triples.stream().filter(hasAclPredicate) 343 .forEach(triple -> { 344 if (authSubjects.contains(triple.getSubject())) { 345 final Map<String, List<String>> aclTriples = 346 authMap.computeIfAbsent(triple.getSubject().getURI(), key -> new HashMap<>()); 347 348 final String predicate = triple.getPredicate().getURI(); 349 final List<String> values = aclTriples.computeIfAbsent(predicate, 350 key -> new ArrayList<>()); 351 nodeToStringStream(triple.getObject()).forEach(values::add); 352 if (predicate.equals(WEBAC_AGENT_VALUE)) { 353 additionalAgentValues(triple.getObject()).forEach(values::add); 354 } 355 } 356 }); 357 // Create a WebACAuthorization object from the provided triples. 358 if (LOGGER.isDebugEnabled()) { 359 LOGGER.debug("Adding acl:Authorization from {}", aclResource.getId()); 360 } 361 authMap.values().forEach(aclTriples -> { 362 final WebACAuthorization authorization = createAuthorizationFromMap(aclTriples); 363 //only include authorizations if the acl resource is not an ancestor acl 364 //or the authorization has at least one acl:default 365 if (!ancestorAcl || authorization.getDefaults().size() > 0) { 366 authorizations.add(authorization); 367 } 368 }); 369 } 370 371 return authorizations; 372 } 373 374 private static WebACAuthorization createAuthorizationFromMap(final Map<String, List<String>> data) { 375 return new WebACAuthorization( 376 data.getOrDefault(WEBAC_AGENT_VALUE, emptyList()), 377 data.getOrDefault(WEBAC_AGENT_CLASS_VALUE, emptyList()), 378 data.getOrDefault(WEBAC_MODE_VALUE, emptyList()).stream() 379 .map(URI::create).collect(toList()), 380 data.getOrDefault(WEBAC_ACCESSTO_VALUE, emptyList()), 381 data.getOrDefault(WEBAC_ACCESSTO_CLASS_VALUE, emptyList()), 382 data.getOrDefault(WEBAC_AGENT_GROUP_VALUE, emptyList()), 383 data.getOrDefault(WEBAC_DEFAULT_VALUE, emptyList())); 384 } 385 386 /** 387 * Recursively find the effective ACL as a URI along with the FedoraResource that points to it. 388 * This way, if the effective ACL is pointed to from a parent resource, the child will inherit 389 * any permissions that correspond to access to that parent. This ACL resource may or may not exist, 390 * and it may be external to the fedora repository. 391 * @param resource the Fedora resource 392 * @param ancestorAcl the flag for looking up ACL from ancestor hierarchy resources 393 */ 394 Optional<ACLHandle> getEffectiveAcl(final FedoraResource resource, final boolean ancestorAcl) { 395 try { 396 397 final FedoraResource aclResource = resource.getAcl(); 398 399 if (aclResource != null) { 400 final List<WebACAuthorization> authorizations = 401 getAuthorizations(aclResource, ancestorAcl); 402 if (authorizations.size() > 0) { 403 return Optional.of( 404 new ACLHandle(resource, authorizations)); 405 } 406 } 407 408 FedoraResource container = resource.getContainer(); 409 // The resource is not ldp:contained by anything, so checked its described resource. 410 if (container == null && (resource instanceof NonRdfSourceDescription || resource instanceof TimeMap)) { 411 final var described = resource.getDescribedResource(); 412 if (!Objects.equals(resource, described)) { 413 container = described; 414 } 415 } 416 if (container == null) { 417 LOGGER.debug("No ACLs defined on this node or in parent hierarchy"); 418 return Optional.empty(); 419 } else { 420 LOGGER.trace("Checking parent resource for ACL. No ACL found at {}", resource.getId()); 421 return getEffectiveAcl(container, true); 422 } 423 } catch (final RepositoryException ex) { 424 LOGGER.debug("Exception finding effective ACL: {}", ex.getMessage()); 425 return Optional.empty(); 426 } 427 } 428 429 private List<WebACAuthorization> getDefaultAuthorizations() { 430 final Map<String, List<String>> aclTriples = new HashMap<>(); 431 final List<WebACAuthorization> authorizations = new ArrayList<>(); 432 433 getDefaultAcl(null, authPropsConfig.getRootAuthAclPath()) 434 .listStatements().mapWith(Statement::asTriple).forEachRemaining(triple -> { 435 if (hasAclPredicate.test(triple)) { 436 final String predicate = triple.getPredicate().getURI(); 437 final List<String> values = aclTriples.computeIfAbsent(predicate, 438 key -> new ArrayList<>()); 439 nodeToStringStream(triple.getObject()).forEach(values::add); 440 if (predicate.equals(WEBAC_AGENT_VALUE)) { 441 additionalAgentValues(triple.getObject()).forEach(values::add); 442 } 443 } 444 }); 445 446 authorizations.add(createAuthorizationFromMap(aclTriples)); 447 return authorizations; 448 } 449 450 private Stream<String> additionalAgentValues(final org.apache.jena.graph.Node object) { 451 if (object.isURI()) { 452 final String uri = object.getURI(); 453 if (userBaseUri != null && uri.startsWith(userBaseUri)) { 454 return of(uri.substring(userBaseUri.length())); 455 } else if (groupBaseUri != null && uri.startsWith(groupBaseUri)) { 456 return of(uri.substring(groupBaseUri.length())); 457 } 458 } 459 return empty(); 460 } 461 462 /** 463 * @param userBaseUri the user base uri 464 */ 465 public void setUserBaseUri(final String userBaseUri) { 466 this.userBaseUri = userBaseUri; 467 } 468 469 /** 470 * @param groupBaseUri the group base uri 471 */ 472 public void setGroupBaseUri(final String groupBaseUri) { 473 this.groupBaseUri = groupBaseUri; 474 } 475}