ApacheDSRule has been moved to `ExternalResource` which requires less code and implements `TestRule` instead of `MethodRule`. `TestRule` supports `ClassRule`s as well as ordinary `Rule`s. A class rule is exactly what we need for the ldap tests since we want to start the ldap server once before class and shut it down after all tests (after class). Also made sure that the static fields are cleaned up, otherwise `StaticFieldsInvariantRule` barfs.
Added `extends ElasticsearchTestsCase` where missing also.
Closeselastic/elasticsearch#80
Original commit: elastic/x-pack-elasticsearch@2143a2dcc6
Added `ShieldRestTests` that extends `ElasticsearchRestTests` allowing to run REST tests against es+shield. Tests won't be run by default as they require additional configuration (e.g. rest tests and spec location on file system). They can be activated via `-Dtests.rest=true`. Rest tests and spec location can be provided as follows:
```
-Dtests.rest.spec=/path/to/elasticsearch/rest-api-spec/api -Dtests.rest.suite=/path/to/elasticsearch/master/rest-api-spec/test
```
Some tests need to be blacklisted at this moment as follows:
```
-Dtests.rest.blacklist=scroll/*/*,mpercolate/*/*,msearch/*/*
```
Closeselastic/elasticsearch#79
Original commit: elastic/x-pack-elasticsearch@6f3e72dd87
`randomInt` includes 0, thus the tampered id could stay the same as the signed scroll ids in some cases which would make everything work and the test fail.
Also cleared the scroll from a finally block, otherwise when the test fails the scroll stays around which might make after test checks fail.
Original commit: elastic/x-pack-elasticsearch@6f6b0d844d
- Key service provides un/signing functionality
- will initially be used to un/sign scroll ids (for the scroll api)
Original commit: elastic/x-pack-elasticsearch@256e0e3c5d
The current IP filter kicks in after the SSL handler, which only
makes sense, if you check things like the SSL certificate. For
now it makes most sense to really put this at the first place.
Original commit: elastic/x-pack-elasticsearch@bbaed67a3c
- Formalized the notion of a client vs. node mode. Introduced an `AbstractShieldModule` that takes care of that
- For now, standarized on the `Shield` name across the board (e.g. change `SecurityModule` to `ShieldModule`)
- Introduces static methods to `ShieldPlugin` to resolve shield specific config files (on the way fixed the file resolving of the ldap group mapper)
- The n2n ip filtering is now resolved at the module level. If not enabled, null is injected and the netty handler is then not injected to the pipeline
- updated code base with the latest changes in es-core around how relevant http headers are registered and copied over to the transport request
- Added new known action in es-core `indices:admin/get`
Original commit: elastic/x-pack-elasticsearch@ca8d85dc81
The wrong path was used as default path. Also added logging information
for all files, so one can at least check the paths.
Original commit: elastic/x-pack-elasticsearch@893493fd17
- Also made sure that we fallback on system token only if the system has permission to the action.
- While at it, change the binding of the different services to run as a singletons
Closeselastic/elasticsearch#64
Original commit: elastic/x-pack-elasticsearch@3705b7365a
- Also fixed a bug where if ldap/esuers module is disabled, injection failed (now injecting the appropriate `null` values to `Realms`
- Also updated `SecurityFilter` with latest changes in es core (action filters API changed)
Original commit: elastic/x-pack-elasticsearch@71de64e6ad
The automaton support changed quite a bit in 4.10 which required determinizing all the automatons used in the Privilege
Original commit: elastic/x-pack-elasticsearch@96a82f0f5d
Removed the caching of the token on the request context. Caching is now handled by the InternalAuthenticationService
Original commit: elastic/x-pack-elasticsearch@d60bc7af67
- ShieldModule loads the SecuredTransportModule which is in turn responsible for loading the required transport related sub-modules
- Removed n2n module and incorporated n2n into the transport module
- Updated the docs to reflect the new settings for n2n ip filtering
Original commit: elastic/x-pack-elasticsearch@dd82306a04
When a http request arrives, we first verify that it carries an authentication token (if it doesn't we throw an authentication exception). Beyond that, any action request that arrives, if it doesn't have an authentication token we assume system user identity. The rationale behind it is that if a request comes in via the transport, then the sending peer authenticated with a client auth cert (the cert acts as the guarantee here that the actor can be assumed as System)... otherwise, the request can come from the local node and triggered by the system (e.g. gateway recovery)
The System user only has permissions to internal apis (it doesn't have full access/permission to all the apis). when a System identity is assumed, the authorization service will grant/deny the request based on whether the request is an internal api or not.
Aso fixed the known actions (to be insync with 1.x branch)
Closeselastic/elasticsearch#45
Original commit: elastic/x-pack-elasticsearch@be27cb0e1b
- throw an error on initialization if the audit trail is configured with an unknown output
- removed unnecessary null checks
Original commit: elastic/x-pack-elasticsearch@8d26723553
Added `shield.n2n.enabled` property to disable this feature, so
it does not get added to the netty pipeline.
Original commit: elastic/x-pack-elasticsearch@d73d27619c
Removing all roles from a user result in an invalid line left in the
roles file. This commit simply removes the user from the roles file
in that case.
Original commit: elastic/x-pack-elasticsearch@c1f4a961fd
The roles file had been checked instead of the users file when checking
if the user already exists. This lead to wrong "user not found" error
messages.
Original commit: elastic/x-pack-elasticsearch@6fc5646ce7
The LDAP realm is a bind-per-user strategy and the group lookup strategy is configurable.
If a role mapping file is not defined, groups names are taken as role names.
Special configuration for active directory simplifies the configuration.
Integration Tests are using an embedded apache DS LDAP Server.
Original commit: elastic/x-pack-elasticsearch@ce20e1b3be
A strategy for extracting the related indices from a request. The default strategy relies on the request to either be `IndicesRequest` or `CompositeIndicesRequest`. But depending on the request type, there might be other strategies as well. For example, `GetIndexTemplatesRequest` only holds the template name and therefore requires a different strategy where the template is looked up in the cluster metadata and the indices are resolved from there.
This also cleans up the Permission class and makes sure that the indices are resolved once per request (and not once per configured group permission)
Original commit: elastic/x-pack-elasticsearch@95192ccdff
- It is now possible to assign index & cluster permission on an action level (not just the fixed privileges we defined).
- also added a test to check that all the actions elasticsearch has are indeed known to shield. So whenever a new action is introduced in elasticsearch, and shield is not aware of it, the build will fail. This will help us ensure that all actions in elasticsearch are 1) well formatted/categorized, 2) secured and "permissible"
Closeselastic/elasticsearch#19
Original commit: elastic/x-pack-elasticsearch@170c3b9185
Load transports in plugin. No need to specify any transport modules anymore.
Removed the transport modules.
Original commit: elastic/x-pack-elasticsearch@45f3245361
- Fixed a bug in the logic of applying the indices group permission check
- For the rest, mainly cosmetics refactoring (renaming & redundant code removal)
Original commit: elastic/x-pack-elasticsearch@748575dc2e
Created an initial set of documentation which can also be built
using `build_docs.pl` without errors.
Original commit: elastic/x-pack-elasticsearch@1a3ccf4eb9