[8u] RFR (S) 8231949: [PPC64, s390]: Make async profiling more reliable

[Thomas Stüfe]

Looks fine to me. I am not sure whether the const void* cast was needed
though.

Cheers, Thomas

On Thu, Apr 15, 2021 at 2:31 PM Aleksey Shipilev <shade at redhat.com> wrote:

> Original bug:
>    https://bugs.openjdk.java.net/browse/JDK-8231949
>    https://hg.openjdk.java.net/jdk/jdk/rev/c6cbcc673cd3
>
> We suspect this causes some of the real-world failures on 8u ppc64le, but
> there is no clear
> reproducer to see if it actually helps. So, this is somewhat of a blind
> backport.
>
> The 8u patch follows the same footsteps Martin did for 11u backport:
>
> https://mail.openjdk.java.net/pipermail/jdk-updates-dev/2019-November/002137.html
>
> Namely, we are using the old style for these checks:
>
> -      if (!Method::is_valid_method(m)) return false;
> -      if (!Metaspace::contains(m->constMethod())) return false;
> +      if (m == NULL || !m->is_valid_method()) return false;
> +      if (!Metaspace::contains((const void*)m->constMethod())) return
> false;
>
> (I would suggest backporting JDK-8207779 that would introduce
> Method::is_valid_method in 8u, then we
> would rewrite this line back to upstream state).
>
> There are also no s390x parts, because that port is not available in 8u.
>
> 8u webrev:
>    https://cr.openjdk.java.net/~shade/8231949/webrev.8u.01/
>
> Testing: Linux ppc64le jdk_jfr (no new failures, no new passes)
>
> --
> Thanks,
> -Aleksey
>
>