RFR (S) 8219459: oopDesc::is_valid() is broken

[Thomas Stüfe]

Hi Coleen, Martin,

I'm also fine with this patch. I agree this function is a bit overkillish.

Also, in Klass::is_valid(), we do:

if (!MetaspaceUtils::is_range_in_committed(k, k + 1)) return false;

which is strict but then again not really correct, since we only do test
range of the base Klass structure, excluding v/itable. We should probably
check with k->size() range instead.

Or, just do a Metaspace::contains() with the Klass pointer. I think the
cases where it points to valid Metaspace but its end does not are probably
rare.

Cheers, Thomas


On Wed, May 8, 2019 at 11:45 PM <coleen.phillimore at oracle.com> wrote:

> Summary: Use Metaspace::contains() to test address ranges.
>
> Tested with hs-tier1-3, and all tier1 on linux-x64-debug.  Tested
> manually in gdb.
>
> open webrev at http://cr.openjdk.java.net/~coleenp/2019/8219459.01/webrev
> bug link https://bugs.openjdk.java.net/browse/JDK-8219459
>
> Thanks,
> Coleen
>