RFR: 8292318: Memory corruption in remove_dumptime_info [v2]
Coleen Phillimore
coleenp at openjdk.org
Tue Aug 16 20:54:17 UTC 2022
On Tue, 16 Aug 2022 20:48:05 GMT, Ioi Lam <iklam at openjdk.org> wrote:
>> In [JDK-8290833](https://bugs.openjdk.org/browse/JDK-8290833) (#9759), I added a table (`SystemDictionaryShared::_saved_cpcache_entries_table`) that remembers the initial state of a `ConstantPoolCache` during CDS dumping. This table is indexed with a `ConstantPoolCache*`
>>
>> However, `ConstantPoolCache` has a complex lifecycle, especially with class redefinition. This makes it difficult to clean up the table. The crash reported in [the current bug](https://bugs.openjdk.org/browse/JDK-8292318) happened during clean up, probably because an `InstanceKlass` was still valid but its `ConstantPool` or `ConstantPoolCache` were not.
>>
>> For simplification, I am now storing the information inside the `ConstantPoolCache`. To compensate for the extra space used, I moved two 32-bit integers next to each other, so the net change in size is zero.
>>
>> instanceKlass.cpp was reverted to the state before #9759.
>
> Ioi Lam has updated the pull request incrementally with one additional commit since the last revision:
>
> added comments about what _initial_entries points to
This looks good. This makes sense to me with the comments.
-------------
Marked as reviewed by coleenp (Reviewer).
PR: https://git.openjdk.org/jdk/pull/9887
More information about the hotspot-dev
mailing list