ARTEMIS-5972: Replace JNI with Panama Foreign Function & Memory (FFM)… by mayankkunwar · Pull Request #6444 · apache/artemis

mayankkunwar · 2026-05-13T15:44:43Z

… API for Journal Native Layer

clebertsuconic · 2026-05-13T18:59:22Z

clebertsuconic · 2026-06-08T14:40:21Z

   </build>

   <profiles>
+      <profile>


Now that you made the optional compilation in the ffm module, this could always be present, right?

I just need to link artemis-ffm changes to artemis-journal, and after that we can always have it.

clebertsuconic · 2026-06-08T14:40:53Z

+         </activation>
+         <dependencies>
+            <dependency>
+               <groupId>org.apache.artemis</groupId>


this could always be present?

yes, let me push the linking part to artemis-journal, and after this we can.

clebertsuconic · 2026-06-08T14:41:01Z

+                  <artifactId>maven-compiler-plugin</artifactId>
+                  <executions>
+                     <execution>
+                        <id>java24-compile</id>


do we still need this?

yes, we will need this as latest changes has FFM related component like MemorySegment

what if we changed the signature to use ByteBuffer (at least what's exposed to this level?)

actually.. never mind... the implementation of the SequentialFactory itself is dependent on other internal parts. I think we need it.

clebertsuconic · 2026-06-08T14:41:23Z

+            <build>
+                <plugins>
+                    <plugin>
+                        <artifactId>maven-compiler-plugin</artifactId>


do we still need this?

I guess for this case we do, but just double checking with you.

yes, we will need this part as well

clebertsuconic · 2026-06-08T14:42:27Z

having the optional compilation in FFM simplifies thing a lot... nice one

gemmellr · 2026-06-10T11:44:49Z

   private boolean aio;

+   @Option(names = "--aio2", description = "Set the journal as asyncio 2 (Panama FFM).")
+   private boolean aio2;


I'd wonder about either something like --aio-ffm, (or maybe just reusing the existing aio option and having an additional toggle that says whether it is ffm or not (i.e jni).

'aio2' seems more like a distinct version of aio, not 'use different way to access same aio version'...and if there ever were such an aio version, it would clash.

aio could have more variants than this, like currently we are using libaio for kernel level I/O operations and later we can replace libaio with io_uring or something else.
So the naming conventions would again change in future.
What if we mention underlying I/O libs something like: --libaio-jni / --libaio-ffm / --iouring-ffm ? or --aio-libaio-jni / --aio-libaio-ffm / --aio-iouring-ffm?
Any suggestions from this POV?

gemmellr · 2026-06-10T11:45:10Z


-      if (aio) {
+      if (aio2) {
+         journalType = JournalType.ASYNCIO_2;


Let's decide on an agreeable naming convention.

gemmellr · 2026-06-10T11:48:59Z

+         <version>${project.version}</version>
+         <scope>compile</scope>


Shouldnt need the version here. Implies missing entry in artemis-bom. Compile scope is default so it can be omitted.

Good catch! Entry is missing in artemis-bom, will add an entry there and update this code

gemmellr · 2026-06-10T11:50:41Z

+      <dependency>
+         <groupId>org.apache.artemis</groupId>
+         <artifactId>artemis-ffm</artifactId>
+         <version>${project.version}</version>


Shouldnt need this at all if the journal depends on it, since the journal will bring it in (and note the journal isnt specified here either)

The artemis-server depends on artemis-journal, I guess we can remove this.

gemmellr · 2026-06-10T11:52:39Z

+   }
+
+   public static void setForceSyscall(boolean value) {
+      throw new UnsupportedOperationException("Not supported for JDK < 22.");


Since its building to 24+, I'd make this message match?

If all the messages are the same, perhaps extract this to a constant and then update the constant.

Will remove the unused methods.

gemmellr · 2026-06-10T12:55:38Z

+      <dependency>
+         <groupId>org.apache.artemis</groupId>
+         <artifactId>artemis-ffm</artifactId>
+         <version>${project.version}</version>


As earlier comment, version shouldnt be needed.

gemmellr · 2026-06-10T12:55:58Z

         </activation>
         <properties>
-            <its-surefire-extra-args>--add-exports java.security.jgss/sun.security.krb5=ALL-UNNAMED</its-surefire-extra-args>
+            <its-surefire-extra-args>--add-exports java.security.jgss/sun.security.krb5=ALL-UNNAMED


This seems a non-change and could be unwound.

gemmellr · 2026-06-10T12:56:41Z

+                <dependency>
+                    <groupId>org.apache.artemis</groupId>
+                    <artifactId>artemis-ffm</artifactId>
+                    <version>${project.version}</version>


As earlier comments.

gemmellr · 2026-06-10T12:57:35Z

+      <dependency>
+         <groupId>org.apache.artemis</groupId>
+         <artifactId>artemis-ffm</artifactId>
+         <version>${project.version}</version>


As earlier comments

gemmellr · 2026-06-10T13:00:09Z

      <module>artemis-jakarta-client-osgi</module>
      <module>artemis-jms-server</module>
      <module>artemis-jakarta-server</module>
+      <module>artemis-ffm</module>


I do wonder if this is the best module name. If its basically specific to the journal I'd wonder if it should have that in the name.

Or if the whole lot could go in the existing journal module, which is already depending on this and having its own multi-release build etc.

Can we rename it as "artemis-aio" or something similar, or any other suggestions on this?
As we have a totally different repo for artemis-native, I thought it would be better to have these changes in separate module.

…, which is now introducing libaio. In the configuration we are naming it as ASYNCIO_2. Co-Authored-By: Clebert Suconic <clebertsuconic@apache.org>

mayankkunwar force-pushed the ARTEMIS-5972 branch 2 times, most recently from 17ca830 to 758dfd6 Compare May 14, 2026 00:05

tabish121 reviewed May 14, 2026

View reviewed changes

Comment thread pom.xml Outdated

mayankkunwar force-pushed the ARTEMIS-5972 branch 8 times, most recently from a91e37f to f177963 Compare May 20, 2026 18:01

clebertsuconic reviewed May 21, 2026

View reviewed changes

Comment thread artemis-ffm/etc/checkstyle-suppressions.xml Outdated

mayankkunwar force-pushed the ARTEMIS-5972 branch from f177963 to 68f5845 Compare May 21, 2026 11:23

clebertsuconic reviewed May 21, 2026

View reviewed changes

Comment thread artemis-ffm/pom.xml Outdated

clebertsuconic reviewed May 21, 2026

View reviewed changes

Comment thread artemis-ffm/src/test/java/org/apache/artemis/nativo/jlibaio/test/jmh/AioCompareBenchmark.java Outdated

clebertsuconic reviewed May 21, 2026

View reviewed changes

Comment thread artemis-ffm/src/test/java/org/apache/artemis/nativo/jlibaio/test/jmh/AioCompareBenchmark.java Outdated

mayankkunwar force-pushed the ARTEMIS-5972 branch from 68f5845 to fb9503f Compare May 21, 2026 14:29

mayankkunwar force-pushed the ARTEMIS-5972 branch 7 times, most recently from 8ddbb55 to 8bcceee Compare June 2, 2026 17:25

clebertsuconic reviewed Jun 2, 2026

View reviewed changes

Comment thread ...a/org/apache/activemq/artemis/tests/unit/core/ffmaio/FfmMultiThreadAsynchronousFileTest.java Outdated

mayankkunwar force-pushed the ARTEMIS-5972 branch 4 times, most recently from bda5293 to fbd575c Compare June 2, 2026 20:28

mayankkunwar force-pushed the ARTEMIS-5972 branch 7 times, most recently from b602e56 to 9fc249c Compare June 5, 2026 17:07

clebertsuconic force-pushed the ARTEMIS-5972 branch 2 times, most recently from beabca5 to a89db8c Compare June 5, 2026 20:24

mayankkunwar force-pushed the ARTEMIS-5972 branch 6 times, most recently from 017c72c to 571ebec Compare June 8, 2026 08:28

clebertsuconic reviewed Jun 8, 2026

View reviewed changes

mayankkunwar force-pushed the ARTEMIS-5972 branch 6 times, most recently from 5c113de to dd41139 Compare June 10, 2026 08:42

gemmellr reviewed Jun 10, 2026

View reviewed changes

mayankkunwar force-pushed the ARTEMIS-5972 branch from dd41139 to 7b4612e Compare June 10, 2026 15:09

ARTEMIS-5972: Create a Panama / FFM module for native implementations…

8324ccb

…, which is now introducing libaio. In the configuration we are naming it as ASYNCIO_2. Co-Authored-By: Clebert Suconic <clebertsuconic@apache.org>

mayankkunwar force-pushed the ARTEMIS-5972 branch from 7b4612e to 8324ccb Compare June 10, 2026 20:34

Conversation

mayankkunwar commented May 13, 2026

Uh oh!

clebertsuconic commented May 13, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

clebertsuconic commented Jun 8, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone