Launch 5.4.0 · jemalloc/jemalloc · GitHub

This launch comprises over 160 commits, specializing in the technical money owed
cleansing together with refactorings, bug fixes, take a look at protection enchancment, and
choice cleanups. The launch additionally contains portability enhancements per
upstream points report.

New options:

• Add EXTENT_ALLOC_FLAG_PINNED so customized extent-allocation hooks can
  mark non-reclaimable mappings, akin to HugeTLB pages, for preferential
  reuse outdoors the decay and purge pipeline. Add the mallctl interfaces
  stats.pinned, stats.arenas..pinned,
  stats.arenas..extents..npinned,
  stats.arenas..extents..pinned_bytes, and
  stats.arenas..mutexes.extents_pinned.{counter} to report
  pinned-memory utilization and mutex statistics. (@binliu19: be2de8c)
• Allow resuming per-CPU area choice through thread.area.
  (@Algunenano: 68c35f6)
• Better align the contents of human-readable and JSON malloc statistics.
  (@spredolac: 68fe1be, 30b1a41, 04aad97)
• Replace the runtime experimental_infallible_new choice with the
  compile-time --enable-cxx-infallible-new choice, enabling
  compiler-level optimizations and optimization in transfer constructors,
  and repair the new(std::nothrow) contract (@spredolac: 160ab9d, fe33667)

Incompatible modifications:

• Adapt tcache fill and retention targets per bin to demand noticed
  between GC occasions, changing the fastened refill/flush coverage. Remove
  seven legacy non-experimental controls: lg_tcache_nslots_mul,
  tcache_nslots_small_min, tcache_nslots_small_max,
  tcache_nslots_large, tcache_gc_delay_bytes,
  lg_tcache_flush_small_div, and lg_tcache_flush_large_div.
  Corresponding malloc_conf settings are silently ignored, matching
  decide.* mallctls return ENOENT, and tcache_ncached_max stays
  supported. (@spredolac: d13fe91)

Bug fixes:

• Preserve errno throughout free, free_sized, and free_aligned_sized,
  and throughout process_madvise-based web page purging. (@spredolac: 3a77966,
  86f0582)
• Fix numeric overflow checks in measurement courses. (@spredolac: 6b24522)
• Accept NULL in free_sized() and free_aligned_sized() (C23
  correctness). (@bigbruno: 7ce8b91)
• Fix TSD lifecycle edge instances by 1) initializing thread-cache bins earlier than
  marking the cache enabled, stopping reentrant bootstrap allocations
  from utilizing uninitialized state, and a pair of) avoiding TSD recreation for late
  deallocations after thread teardown on generic-TSD platforms.
  (@fzakaria: 54f22c8, @spredolac: fb5499a)
• Use O_CLOEXEC when opening the THP sysfs file in init_thp_state.
  (@ibookstein: d070554)
• Fix duplicate decide.stats_print and decide.stats_print_opts fields in
  malloc_stats_print output. (@spredolac: dfe3a2e)
• Fix a possible impasse throughout arena_reset. (@guangli-dai: 6957341)
• Fix a prof-sampling / guard-page interplay bug within the SAN. (@gctony:
  e36a0fa)

Optimizations and refactors:

• Modularize jemalloc’s entrance finish by extracting area administration,
  initialization, fork orchestration, and allocation dispatch from
  jemalloc.c; untangle tcache/area possession; and consolidate the
  inside header graph to remove round dependencies. (@spredolac:
  ba1e2fe, 8874597, 9d75722, de9ad14, …)
• Simplify ctl dispatch by refactoring area helpers, internalizing an
  implementation-only management, changing control-flow macros with typed
  helpers, and organizing ctl.c by subsystem. (@spredolac: 9e14346,
  901365d, 4e903a0, 5bc8d6e)
• Cap the base-block progress heuristic to keep away from virtual-memory exhaustion
  below uncommon racy situations. (@guangli-dai: 2f4db8c)
• Move background-thread lifecycle and state operations into the
  background-thread module, clarifying possession independently of PAC/HPA
  callers. (@guangli-dai: f1f0792)
• Simplify the page-allocation boundary by changing PAI vtable dispatch
  with direct PAC/HPA calls, eradicating the out of date pai_t/pai.h
  abstraction, and shifting deferred-work and decay orchestration out of the
  area. (@guangli-dai: 1dfa6f7, 8edd101, d410f43)
• Refactor statistics assortment and rendering into separate
  collect/emission levels and descriptor-driven tables. (@spredolac:
  184f304, 80c8fcb)
• Introduce an OS abstraction layer and transfer platform-dependent
  file/course of I/O, time, synchronization, CPU, virtual-memory, atfork,
  error-handling, profiling, thread-yield, and configuration-access
  operations out of allocator core code. (@guangli-dai: c4158ac,
  c8e2e01, …)

Portability enhancements:

• Replace the std::__throw_bad_alloc name with customary C++ (#2900).
  (@lexprfuncall: 1a15fe3)
• Make arena_s use a versatile array member (bin_t all_bins[]) for C99
  or newer. (@grueninger: 300b58b)
• Fix rdtscp detection with --with-lg-vaddr. (@xinydev: e8a0d2b)
• Fix malloc_getcpu on macOS to learn the present CPU quantity appropriately.
  (@Algunenano: 5aabbc8)
• Use CLOCK_MONOTONIC for background-thread sleep to forestall
  clock-rollback stalls, and detect monotonic-condvar assist at configure
  time. (@antonio2368: ebacec3, 8361239)
• Fix compilation warnings on macOS. (@gctony: abb0a8a)
• Fix thread-exit TSD cleanup on MinGW builds. (@gctony: 1e92317)
• Fix GCC 16 construct warnings by eradicating -Wpedantic violations in macro
  and performance syntax and explicitly NUL-terminating profiling thread-name
  copies to resolve -Wstringop-truncation. (@grueninger: 5acdcee,
  c22d929, @jasonangelov: 6a245e0)
• Parse PID-namespace symlinks with out glibc-dependent strtok/atol,
  and return identifiers as uint64_t. (@guangli-dai: 278d90a)