A bug in C2 that causes a large amount of physical memory to be allocated

[Aleksey Shipilev]

On 3/15/19 11:42 AM, Tobias Hartmann wrote:
> On 15.03.19 11:08, 寒泉子 wrote:
>>     Thank you, will this patch consider backport to jdk8?
> 
> Probably a point fix is more suitable for a backport but I leave this to Coleen / the runtime team
> to decide.

What a coincidence! I think my build server OOMs on tests with similar symptoms.

If there are no objections and/or subtle problems with it, I am willing to pick this up for 11u and
8u backports:

$ hg qdiff
diff -r 3086207c8650 src/hotspot/share/code/nmethod.cpp
--- a/src/hotspot/share/code/nmethod.cpp        Tue Mar 05 08:24:58 2019 -0500
+++ b/src/hotspot/share/code/nmethod.cpp        Fri Mar 15 11:39:27 2019 +0100
@@ -1538,10 +1538,11 @@
           Metadata* md = r->metadata_value();
           if (md != _method) f(md);
         }
       } else if (iter.type() == relocInfo::virtual_call_type) {
         // Check compiledIC holders associated with this nmethod
+        ResourceMark rm;
         CompiledIC *ic = CompiledIC_at(&iter);
         if (ic->is_icholder_call()) {
           CompiledICHolder* cichk = ic->cached_icholder();
           f(cichk->holder_metadata());
           f(cichk->holder_klass());


-Aleksey