On Mon, 28 Oct 2024 15:50:00 GMT, Robert Toyonaga <d...@openjdk.org> wrote:
>> Yes I think so. `MemTracker::print_containing_region` first acquires the >> NMT lock then the `SharedDecoder_lock` like this: >> >> `MemTracker::print_containing_region` --> >> `PrintRegionWalker::do_allocation_site` --> >> `NativeCallStack::print_frame` --> >> `Decoder::get_source_info` --> >> `Decoder::shared_decoder_lock()` > > Otherwise `make test TEST=hotspot_nmt` throws a lock rank error when running > the test `location_printing_mmap_1`. That test uses > `MemTracker::print_containing_region`. Thanks. But eew. Okay, ugly but not easy to fix. Can you please add a trailing comment behind SharedDecoder lock that it needs to be lower than the NMT lock? ------------- PR Review Comment: https://git.openjdk.org/jdk/pull/20852#discussion_r1819338305
