-
Notifications
You must be signed in to change notification settings - Fork 336
Fix Namespace resolution on grant/revoke privilege operations #2170
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Fix Namespace resolution on grant/revoke privilege operations #2170
Conversation
dimas-b
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Hi @poojanilangekar , this fix makes sense to me, but I've got some comments/suggestions :)
| isNamespaceFullyResolvedMethod.invoke( | ||
| adminService, null, "test-catalog", Namespace.of("ns1")); | ||
|
|
||
| assertThat(result).isFalse(); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
nit: Why not inline result? The expression is already multi-line 🤔
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Done.
| (boolean) | ||
| isNamespaceFullyResolvedMethod.invoke( | ||
| adminService, resolvedPathWrapper, catalogName, namespace); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This pattern is repeated... Maybe add a help method in this test class?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I've moved the test, so that fixed it.
|
|
||
| List<PolarisEntity> fullPath = resolvedPathWrapper.getRawFullPath(); | ||
| int expectedPathLength = 1 + namespace.levels().length; | ||
| if (fullPath.size() < expectedPathLength) { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
why not ==?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Done, changed it to !=
| * @param namespace the namespace we're trying to resolve | ||
| * @return true if the namespace is considered fully resolved for the given catalog type | ||
| */ | ||
| private boolean isNamespaceFullyResolved( |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Why keep it as a private instance method? It does not use any state from this... It could be a package-private static method in a utility class.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
... or perhaps a public method in PolarisResolvedPathWrapper? 🤔
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Done.
poojanilangekar
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Done, I have addressed your comments.
|
|
||
| List<PolarisEntity> fullPath = resolvedPathWrapper.getRawFullPath(); | ||
| int expectedPathLength = 1 + namespace.levels().length; | ||
| if (fullPath.size() < expectedPathLength) { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Done, changed it to !=
| * @param namespace the namespace we're trying to resolve | ||
| * @return true if the namespace is considered fully resolved for the given catalog type | ||
| */ | ||
| private boolean isNamespaceFullyResolved( |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Done.
| isNamespaceFullyResolvedMethod.invoke( | ||
| adminService, null, "test-catalog", Namespace.of("ns1")); | ||
|
|
||
| assertThat(result).isFalse(); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Done.
| (boolean) | ||
| isNamespaceFullyResolvedMethod.invoke( | ||
| adminService, resolvedPathWrapper, catalogName, namespace); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I've moved the test, so that fixed it.
| } | ||
|
|
||
| for (int i = 0; i < namespace.levels().length; i++) { | ||
| if (!fullPath.get(i + 1).getName().equals(namespace.levels()[i]) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This may not be a big deal, but since we're always scanning this list from head to tail, using iterators might be more portable and easier to understand. Probably as efficient as .get(N) for array-based lists, but potentially more efficient for linked lists.
It might be preferable to move namespace.levels() to a local variable. Iterate it in for, but do simple iterator.next() for fullPath.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Done.
|
|
||
| boolean result = wrapper.isFullyResolvedNamespace("test-catalog", Namespace.of("ns1")); | ||
|
|
||
| assertThat(result).isFalse(); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
nit: inline?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Done.
| throws Exception { | ||
| String catalogName = "test-catalog"; | ||
| String catalogRoleName = "test-role"; | ||
| Namespace namespace = Namespace.of("incomplete-ns"); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is a single-level namespace... How is "incomplete" different from "non existent"?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Done.
| Namespace namespace = Namespace.of("ns1"); | ||
|
|
||
| PolarisEntity catalogEntity = createEntity(catalogName, PolarisEntityType.CATALOG); | ||
| PolarisEntity wrongTypeEntity = createEntity("ns1", PolarisEntityType.TABLE_LIKE); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
For completeness this test should probably also have some NS level with the right type and some with another type.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Done.
dimas-b
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks, @poojanilangekar ! I'll be merging late Friday if there are no more review (or Mon if I forget :) )
|
Hi @dimas-b, could you please merge this PR? |
* fix(deps): update dependency com.google.errorprone:error_prone_core to v2.41.0 (apache#2181) * Simplify bootstrapServiceAndCreatePolarisPrincipalForRealm (apache#2172) this is a small follow-up to 5faa371 because the same pattern existed for this method. note that we do some minor additional "formatting" changes to minimize the diff between the two files (as they were originally copy pasted). this could lead to having a common base class in the future. * fix(deps): update dependency boto3 to v1.39.13 (apache#2182) * Add podman support (apache#2143) * Add Polaris Community Meeting 2025-07-24 (apache#2184) * fix(deps): update dependency com.adobe.testing:s3mock-testcontainers to v4.7.0 (apache#2185) * Push AccessConfig creation to PolarisStorageIntegration (apache#2171) This refactoring does not change Polaris behaviour. * Move storage-specific access properties processing logic from core code to storage integration implementations. * Add `isExpirationTimestamp` flag to `StorageAccessProperty` to allow them to be processed uniformly. * Prepare for supporting access config properties that may have different values in Polaris Servers and Clients. This enables future enhancements to support different S3 endpoint DNS names in servers and clients for apache#1530 * Fix doc to remove privileges may take up to one hour to take effect and add Policy to securable object (apache#2009) * fix(deps): update dependency boto3 to v1.39.14 (apache#2186) * chore(deps): update plugin jetbrains-changelog to v2.3.0 (apache#2187) * fix(deps): update dependency software.amazon.awssdk:bom to v2.32.9 (apache#2191) * Add Principal lookup helpers to PolarisMetaStoreManager (apache#2174) `PolarisMetaStoreManager.readEntityByName` is quite a low-level api, so we can simplify a lot of callers with additional helpers: - add `PolarisMetaStoreManager.findRootPrincipal` - add `PolarisMetaStoreManager.findPrincipalByName` - add `PolarisMetaStoreManager.findPrincipalRoleByName` also we now prefer `PolarisEntityConstants` where applicable * Remove PolarisDiagnostics from json utils (apache#2176) With transitive cleanups of, PolarisStorageConfigurationInfo, ConnectionConfigInfoDpo, BaseMetaStoreManager, PolarisObjectMapperUtil, CurrentContext * Fix Namespace resolution on grant/revoke privilege operations (apache#2170) * Fix Namespace resolution on grant/revoke privilege operations * Move isFullyResolvedNamespace to PolarisResolvedPathWrapper * fix(deps): update dependency boto3 to v1.39.15 (apache#2199) * fix(deps): update dependency com.google.cloud:google-cloud-storage-bom to v2.54.0 (apache#2200) * fix(deps): update dependency boto3 to v1.39.16 (apache#2209) * chore(deps): update actions/stale digest to a92fd57 (apache#2208) * fix(deps): update quarkus platform and group to v3.25.0 (apache#2167) * NoSQL updates * Last merged commit 914be46 --------- Co-authored-by: Mend Renovate <bot@renovateapp.com> Co-authored-by: Christopher Lambert <xn137@gmx.de> Co-authored-by: Yong Zheng <yongzheng0809@gmail.com> Co-authored-by: JB Onofré <jbonofre@apache.org> Co-authored-by: Dmitri Bourlatchkov <dmitri.bourlatchkov@gmail.com> Co-authored-by: CG <cgpoh@users.noreply.github.com> Co-authored-by: Pooja Nilangekar <poojan@umd.edu>
I encountered the bug while implementing RBAC for federation.
For grant/revoke operations on tables, polaris checks the subtype of the leaf entity and make sure it matches. However, for namespace the only check is that the resolved path is not null.
Ideally we should check that the entire path matches the one sent in the request. Further, all entities except the root should be of the type NAMESPACE.
Testing:
Added unit tests.