adding rsockets support into JDK

[Lu, Yingqi]

Hi Brian,

Thank you very much for the help here! I will apply your patch and try it here.

I am currently working on consolidating tests. I will add one specifically for non-Linux platforms to make sure they throw UOE on RDMA methods.

Thanks,
Lucy

From: Brian Burkhalter [mailto:brian.burkhalter at oracle.com]
Sent: Tuesday, September 18, 2018 10:28 AM
To: Lu, Yingqi <yingqi.lu at intel.com>
Cc: Alan Bateman <Alan.Bateman at oracle.com>; nio-dev at openjdk.java.net; Viswanathan, Sandhya <sandhya.viswanathan at intel.com>; Aundhe, Shirish <shirish.aundhe at intel.com>; Kaczmarek, Eric <eric.kaczmarek at intel.com>
Subject: Re: adding rsockets support into JDK

Hello,

I updated the prototype changes I previously posted. The webrev versus version 9 of the patch is here:

http://cr.openjdk.java.net/~bpb/8195160/webrev.09-03/

This should be able to be applied immediately on top of version 9. It builds on Linux and macOS and all the new tests pass on Linux.

The main changes are:

1. rdma.ch<http://rdma.ch> package refactored to jdk.internal.net.rdma.
2. jdk.net.internal.rdma refactored to jdk.internal.net.rdma.
3. Reflection replaced with shadow classes.

It might not hurt to add a test which verifies that the new jdk.net.Sockets methods all throw UOE on non-Linux platforms.

Thanks,

Brian

On Sep 17, 2018, at 11:01 AM, Brian Burkhalter <brian.burkhalter at oracle.com<mailto:brian.burkhalter at oracle.com>> wrote:


On Sep 15, 2018, at 7:38 AM, Alan Bateman <Alan.Bateman at oracle.com<mailto:Alan.Bateman at oracle.com>> wrote:


No objection to introducing jdk.net.internal.rdma.{SocketFactory,RdmaSocketFactoryRdmaProviderFactory} but it still means some ugly reflection in rdmaSocketProvider() to decide whether to return a provider or throw UOE. If you want something simpler just create a RdmaSocketImpl in shared code where the constructor throws UOE. The build will compile in this version on non-Linux platforms. We have a couple of other examples where platform specific classes shadow same named classes in share/classes.

I'll look into updating my most recent prototype.


jdk.internal.net.rdma seem okay to me for the package name (the webrev had jdk.net.internal.rmda which I assume wasn't intended).

Will change this also.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/nio-dev/attachments/20180918/197acf68/attachment.html>