RFR:[jdk11]:JDK-8213574: Deadlock in string table expansion when dumping lots of CDS classes

[Jiangli Zhou]

Thank you, Robbin!

Best,
Jiangli

On Mon, Jun 8, 2020 at 1:42 AM Robbin Ehn <robbin.ehn at oracle.com> wrote:
>
> Looks good, thanks!
>
> /Robbin
>
> On 2020-06-05 19:03, Jiangli Zhou wrote:
> > [+christoph.langer at sap.com, +martinrb at google.com]
> >
> > Gentle ping
> >
> > Martin or Christoph, could you please review the backport?
> >
> > Thanks!
> > Jiangli
> >
> >
> >
> > On Mon, May 18, 2020 at 2:31 PM Jiangli Zhou <jianglizhou at google.com> wrote:
> >>
> >> Please review the backport for JDK-8213574 to JDK 11:
> >>
> >> backport webrev:
> >> http://cr.openjdk.java.net/~jiangli/8213574-jdk11-backport/webrev.00/
> >> https://bugs.openjdk.java.net/browse/JDK-8213574
> >>
> >> Original webrev: http://hg.openjdk.java.net/jdk/jdk/rev/d5eebe1e03fe
> >>
> >> CDS dumping may run into a deadlock situation during copying the
> >> interned j.l.String objects in the VM thread, while the service thread
> >> is concurrently expanding the string table. I ran into the deadlock
> >> recently when creating a CDS archive with an input classlist of ~60K
> >> classes on JDK 11. The backport addresses the deadlock issue.
> >>
> >> The diffs in symbolTable.cpp are not included in the backport as they
> >> only apply to ConcurrentHashTable. In JDK 11 the SymbolTable is not
> >> yet a ConcurrentHashTable. All other diffs in the patch applied
> >> cleanly.
> >>
> >> Tested with tier1, jtreg/runtime/SharedArchiveFile
> >>
> >> Best,
> >> Jiangli