[Ping] Re: RFR [1/3]: 8057038 Speculative traps not robust when compilation and class unloading are concurrent

Hohensee, Paul hohensee at amazon.com
Wed Apr 14 17:39:53 UTC 2021


Lgtm.

The other two are 8139247 and 8231501. The latter I reviewed for 8u, but isn't in 11u: there's a comment that says that 11u isn’t affected by it because 8215889 isn't in 11u. But, 8057038 also references 8215889. The latter hasn't been backported to either 8u or 11u, presumably because it fixes a concurrent class unloading issue in ZGC, and ZGC issues are being ignored in 11u. But, Shenandoah has/will have concurrent class unloading, which could very well end up in 11u, so imo both 8215889 and 8231501 should be backported to 11u as part of this series.

Thanks,
Paul

-----Original Message-----
From: jdk8u-dev <jdk8u-dev-retn at openjdk.java.net> on behalf of Peter Zhelezniakov <peter.zhelezniakov at bell-sw.com>
Date: Tuesday, January 12, 2021 at 9:54 AM
To: "jdk8u-dev at openjdk.java.net" <jdk8u-dev at openjdk.java.net>
Subject: [Ping] Re: RFR [1/3]: 8057038 Speculative traps not robust when compilation and class unloading are concurrent

Any thoughts?

On Dec 22 2020, at 3:14 pm, Peter Zhelezniakov <peter.zhelezniakov at bell-sw.com> wrote:
> Hi,
>
> This is the first backport request in a series of 3: [1], [2], [3]. These fixes introduce atomicity and mutexes to Hotspot code. They help address crashes that occur in Apache Geode when running on a server with tens of CPU cores.
> The patch from jdk9 did not apply as is, some contextual changes were needed.
> This fix had a flaw that was later identified and fixed under [2]. Request for [2] is coming next in the series.
> Testing: Hotspot JTreg tests.
> JBS issue: https://bugs.openjdk.java.net/browse/JDK-8057038
> Webrev: http://cr.openjdk.java.net/~avoitylov/peterz/8057038/8057038/
>
>
> Thanks!
> Peter
>
> [1] https://bugs.openjdk.java.net/browse/JDK-8057038
> [2] https://bugs.openjdk.java.net/browse/JDK-8139247
> [3] https://bugs.openjdk.java.net/browse/JDK-8231501




More information about the jdk8u-dev mailing list