RFR: JDK-8114823 - G1 doesn't honor request to disable class unloading (original) (raw)

Jesper Wilhelmsson jesper.wilhelmsson at oracle.com
Thu Mar 31 15:29:18 UTC 2016

Hi Mikael,

Den 31/3/16 kl. 17:04, skrev Mikael Gerdin:

Hi Jesper,

On 2016-03-31 16:50, Jesper Wilhelmsson wrote: Hi Mikael,

Thanks for looking at this! I agree that it would look better (and probably be more semantically correct) to disable CMSClassUnloadingEnabled as well. It will look weird to users who are used to look at this flag if it says true but class unloading still doesn't happen. I also updated the description of CMSClassUnloadingEnabled to say that it is equivalent to ClassUnloadingWithConcurrentMark. New webrev: http://cr.openjdk.java.net/~jwilhelm/8114823/webrev.01/ One final comment: now it feels a bit strange that if -ClassUnloading is set then CMSClassUnloading will not have its default value and we will end up inside the second if statement always. Perhaps it would be better to reorder the ifs so that we first check for CMSClassUnloadingEnabled and adjust ClassUnloadingWithConcurrentMark properly, then if class unloading is completely disabled then just make them all false. Does that make sense to you?

I thought about this as well, but the reason for having it in the order it is now is if someone would set -XX:-ClassUnloading -XX:+CMSClassUnloadingEnabled which is a valid combination. In this case, if we change the order, we would disable CMSClassUnloadingEnabled. Of course we could add checks to see if it is set on the command line before disabling it, but that would in my opinion add a lot of code just to avoid setting ClassUnloadingWithConcurrentMark to false twice.

If you have a strong opinion about this I can try to come up with something better, but I would prefer to keep it the way it is. /Jesper

/Mikael

Thanks, /Jesper

Den 31/3/16 kl. 16:21, skrev Mikael Gerdin: Hi Jesper, On 2016-03-31 15:51, Jesper Wilhelmsson wrote: Hi,

Please review this change to make the GCs disable class unloading if disabled on the command line. There are a few parts in this change: 1. The flag ClassUnloadingWithConcurrentMark was not set to false if ClassUnloading was disabled on the command line. This is now done. 2. CMS was using CMSClassUnloadingEnabled instead of the more generic ClassUnloadingWithConcurrentMark. I changed so that ClassUnloadingWithConcurrentMark is used all over and CMSClassUnloadingEnabled is only used as an alias when initializing the arguments. At some point it would be nice to deprecate CMSClassUnloadingEnabled but it is a fairly well known flag so that's not part of this change. 3. CMS, Serial, and PS did correctly disable class unloading when told so, but the outermost code witch included some logging was still enabled making class unloading output present in the log file. I changed so that this code is also disabled. 4. G1 did not disable class unloading during full GCs at all. This is now done. Testing: I used SecureDBBTest.java as suggested in the bug while fixing to verify that class unloading was happening before the fix and not after. I also ran it through JPRT. Bug: https://bugs.openjdk.java.net/browse/JDK-8114823 Webrev: http://cr.openjdk.java.net/~jwilhelm/8114823/webrev.00/ I think this change seems good overall. I have one small concern regarding CMSClassUnloadingEnabled, and that is that if I set -ClassUnloading then CMSClassUnloadingEnabled will still be true, would it make sense to still disabled it in the !ClassUnloading block even if nobody's supposed to look at CMSClassUnloadingEnabled? Another approach could be to just add CMSClassUnloadingEnabled as an alias for ClassUnloadingWithConcurrentMark but I haven't researched exactly how the aliasedjvmflags table works. /Mikael

Thanks, /Jesper

More information about the hotspot-gc-dev mailing list