[11u] RFR 8244340: Handshake processing thread lacks yielding

[Reingruber, Richard]

Hi Goetz,

> Thanks for downporting this bug. 
> It is good you target it for 11.0.11. I cannot see why Oracle 
> pushed it to 11.0.10 that late.
> Maye they just downported the tiny fix !is_mp()... discussed 
> in the mail thread.

Maybe they have done that. I (partially) reread the mail thread and learned that
the fix solves also issues on MP systems. But I think these issues relate to
direct handshakes and biased lock revocation (both introduced after
jdk11). Interesting also that the original repro case was for jdk14.

> The downport looks good.

Thanks.

> I can sponsor this for you. Please ping me once it is
> labeled jdk11u-fix-yes.

Thanks, the label is there now.

Cheers, Richard.

-----Original Message-----
From: Lindenmaier, Goetz <goetz.lindenmaier at sap.com> 
Sent: Dienstag, 15. Dezember 2020 08:24
To: Reingruber, Richard <richard.reingruber at sap.com>; jdk-updates-dev <jdk-updates-dev at openjdk.java.net>
Subject: RE: [11u] RFR 8244340: Handshake processing thread lacks yielding

Hi Richard, 

Thanks for downporting this bug. 
It is good you target it for 11.0.11. I cannot see why Oracle 
pushed it to 11.0.10 that late.
Maye they just downported the tiny fix !is_mp()... discussed 
in the mail thread.

The downport looks good. 

I can sponsor this for you. Please ping me once it is
labeled jdk11u-fix-yes.

Best regards,
 Goetz



> -----Original Message-----
> From: jdk-updates-dev <jdk-updates-dev-retn at openjdk.java.net> On Behalf
> Of Reingruber, Richard
> Sent: Friday, December 11, 2020 4:17 PM
> To: jdk-updates-dev <jdk-updates-dev at openjdk.java.net>
> Subject: [11u] RFR 8244340: Handshake processing thread lacks yielding
> 
> Hi,
> 
> Original bug:
>   https://bugs.openjdk.java.net/browse/JDK-8244340
>   https://hg.openjdk.java.net/jdk/jdk/rev/67a435ce2dc2
> 
> Prerequisite downport:
>   https://bugs.openjdk.java.net/browse/JDK-8214180
> 
> Original patch does not apply cleanly to 11u because it is based on
> handshakes
> in jdk15 which is quite different from handshakes in 11u, e.g. jdk15 has direct
> handshakes.
> 
> 11u webrev:
> 
> http://cr.openjdk.java.net/~rrich/webrevs/8244340_11u_Handshake_proces
> sing_thread_lacks_yielding/webrev.0/
> 
> Testing:
> 
> Tier1 with and without ZGC.
> 
> With ZGC I got a few failures. Most of them if not all were OOME because
> 11u ZGC
> cannot handle small heaps. The same tests fail without the patch.
> 
> CI testing @sap.
> 
> Thanks, Richard.