RFR: 8351661: NMT: VMATree should support separate call-stacks for reserve and commit operations [v2]

[Afshin Zafari]

On Mon, 17 Mar 2025 09:59:42 GMT, Johan Sjölen <jsjolen at openjdk.org> wrote:

>> Afshin Zafari has updated the pull request incrementally with one additional commit since the last revision:
>> 
>>   fixed a bug.
>
> src/hotspot/share/nmt/nmtNativeCallStackStorage.hpp line 53:
> 
>> 51: 
>> 52:   static bool is_invalid(StackIndex a) {
>> 53:     return a == invalid || a < 0;
> 
> If this change is necessary, then there's a bug in the code.

fixed.

> src/hotspot/share/nmt/vmatree.hpp line 95:
> 
>> 93:     uint8_t type_tag[2];
>> 94:     NativeCallStackStorage::StackIndex sidx;       // call-stack of all operations
>> 95:     NativeCallStackStorage::StackIndex second_idx; // call-stack when committing/uncommitting the start-node of a reserved region
> 
> These can be called `primary_stack` and `secondary_stack`. Does an uncommitted region really have a stack?

renamed.

Yes uncommit also has stack although never reported. `VMATree::register_mapping()` will save the stack in the nodes, regardless of the operation. It is done when constructing the `stA` and `stB` interval-states:

IntervalChange stA{
      IntervalState{StateType::Released, empty_regiondata},
      IntervalState{              state,   metadata}
  };
  IntervalChange stB{
      IntervalState{              state,   metadata},
      IntervalState{StateType::Released, empty_regiondata}
  };

The stack may be empty or whatever passed in by `metadata` arg.

-------------

PR Review Comment: https://git.openjdk.org/jdk/pull/24028#discussion_r2000569649
PR Review Comment: https://git.openjdk.org/jdk/pull/24028#discussion_r2000569358