RawMemoryFreeList can't get more space from mmap

[wks]

I observed the following error when the CI was running for #1067

The GitHub Action link: https://github.com/mmtk/mmtk-core/actions/runs/8293161626/job/22695757005?pr=1067

---- util::raw_memory_freelist::tests::alloc_exhaust stdout ----
thread 'util::raw_memory_freelist::tests::alloc_exhaust' panicked at 'Can't get more space with mmap()', src/util/raw_memory_freelist.rs:202:9
stack backtrace:
   0: rust_begin_unwind
             at /rustc/eb26296b556cef10fb713a38f3d16b9886080f26/library/std/src/panicking.rs:593:5
   1: core::panicking::panic_fmt
             at /rustc/eb26296b556cef10fb713a38f3d16b9886080f26/library/core/src/panicking.rs:67:14
   2: mmtk::util::raw_memory_freelist::RawMemoryFreeList::mmap
             at ./src/util/raw_memory_freelist.rs:202:9
   3: mmtk::util::raw_memory_freelist::RawMemoryFreeList::raise_high_water
             at ./src/util/raw_memory_freelist.rs:196:9
   4: mmtk::util::raw_memory_freelist::RawMemoryFreeList::grow_list_by_blocks
             at ./src/util/raw_memory_freelist.rs:144:13
   5: mmtk::util::raw_memory_freelist::RawMemoryFreeList::grow_freelist
             at ./src/util/raw_memory_freelist.rs:134:9
   6: mmtk::util::raw_memory_freelist::tests::new_raw_memory_freelist
             at ./src/util/raw_memory_freelist.rs:279:9
   7: mmtk::util::raw_memory_freelist::tests::alloc_exhaust
             at ./src/util/raw_memory_freelist.rs:474:40
   8: mmtk::util::raw_memory_freelist::tests::alloc_exhaust::{{closure}}
             at ./src/util/raw_memory_freelist.rs:473:24
   9: core::ops::function::FnOnce::call_once
             at /rustc/eb26296b556cef10fb713a38f3d16b9886080f26/library/core/src/ops/function.rs:250:5
  10: core::ops::function::FnOnce::call_once
             at /rustc/eb26296b556cef10fb713a38f3d16b9886080f26/library/core/src/ops/function.rs:250:5
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.

I think it is unrelated to #1067. I suspect it is related to address space layout randomization (ASLR) and the tight address space of 32-bit architecture, and that particular execution may unfortunately allocated the free list at a place that cannot be grown.

I'll re-run the test and see if the problem persists.

[wks]

This error does not occur in the second run. So it may be related to some random factors.

[k-sareen]

We need to be more tolerant to mmap-failures anyway. Key for robustness

[qinsoon]

We need to be more tolerant to mmap-failures anyway. Key for robustness

I agree with what you said. But this issue is caused by using a fixed address range for testing.

[qinsoon]

The unit tests for raw_memory_freelist use a fixed address range. It happened that on our CI, that range is frequently occupied, possibly due to the runner image update. #1095 mitigated the issue by using a different fixed address range. If the issue occurs again, we may consider just using a random address we get from mmap for the tests.