New macro ALWAYS_SMALL_CLEAR_STACK to avoid clearing large stack sections
(a cherry-pick of commit 95f5767 from Unity-Technologies/bdwgc)
Issue #173 (bdwgc).
Avoid clearing large sections of the stack. It was clearing
up to 16k which may not be available in cases of small stacks or stack
overflow situations (for some targets).
* misc.c [ALWAYS_SMALL_CLEAR_STACK] (GC_clear_stack): Do only BZERO for
dummy where dummy is a local array of length SMALL_CLEAR_SIZE.
Ivan Maidanski [Wed, 31 Jan 2018 08:11:31 +0000 (11:11 +0300)]
Do not define clear_stack_inner and next_random_no if not used
(code refactoring of commits 4eaea0afa, b850e1c9b)
* misc.c (SMALL_CLEAR_SIZE): Do not define if already defined; define
even if THREADS is not defined.
* misc.c [STACK_NOT_SCANNED] (GC_clear_stack): Standalone definition
(just return arg).
* misc.c (BIG_CLEAR_SIZE, GC_stack_last_cleared, GC_min_sp,
GC_high_water, GC_bytes_allocd_at_reset, DEGRADE_RATE,
GC_clear_stack_inner, next_random_no): Do not define if
STACK_NOT_SCANNED.
* misc.c [!STACK_NOT_SCANNED] (CLEAR_SIZE): Do not define if
ASM_CLEAR_CODE.
* misc.c (GC_stack_last_cleared, GC_min_sp, GC_high_water,
GC_bytes_allocd_at_reset, GC_clear_stack_inner, next_random_no,
GC_clear_stack): Adjust indentation.
Gustavo Giraldez [Thu, 18 Jan 2018 21:52:25 +0000 (00:52 +0300)]
Fix GC allocation mutex in child after a fork
Even though after a fork the child only inherits the single thread
that called the fork(), if another thread in the parent was attempting
to lock the mutex while being held in fork_child_prepare(), the mutex
will be left in an inconsistent state in the child after the UNLOCK().
This is the case, at least, in Mac OS X and leads to an unusable GC in
the child which will block when attempting to perform any GC operation
that acquires the mutex.
* pthread_support.c [CAN_HANDLE_FORK && USE_PTHREAD_LOCKS]
(fork_child_proc): Add assertion (after UNLOCK) that the lock is not
held actually, and, then, re-initialize GC_allocate_ml; add comments.
Ivan Maidanski [Wed, 17 Jan 2018 07:42:04 +0000 (10:42 +0300)]
Add note of set_free_space_divisor, set_warn_proc ABI change after gc-7.1
Issue #197 (bdwgc).
GC_set_free_space_divisor() and GC_set_warn_proc() used to return
some value. But starting from version 7.1alpha2, these API functions
do not return any value.
* include/gc.h (GC_set_free_space_divisor, GC_set_warn_proc): Add
comment that the setter used to returned the old value in gc-7.1 (and
previous versions).
Ivan Maidanski [Mon, 15 Jan 2018 05:20:02 +0000 (08:20 +0300)]
Fix global operator delete definition for C++14 in gc_cpp
Issue #195 (bdwgc).
Sized variants of global operator delete should be defined along with
the non-sized ones. Otherwise compiler might invoke the one from the
standard library (as observed in e.g. Cygwin) leading to a crash.
* gc_cpp.cc [!_MSC_VER && __cplusplus>201103L] (operator delete):
Define sized variant (the size argument is ignored for now).
* gc_cpp.cc [!_MSC_VER && __cplusplus>201103L && GC_OPERATOR_NEW_ARRAY
&& !CPPCHECK] (operator delete[]): Likewise.
Ivan Maidanski [Sun, 14 Jan 2018 07:51:06 +0000 (10:51 +0300)]
Fix last_reclaimed..gc_no interval comparison to threshold in unmap_old
(fix commit 14c324f8b)
* allchblk.c [USE_MUNMAP] (GC_unmap_old): Replace
(unsigned short)GC_gc_no-hb_last_reclaimed to
(unsigned short)(GC_gc_no-hb_last_reclaimed) to handle
value wrapping in hb_last_reclaimed (and the truncated GC_gc_no)
correctly.
Peter Wang [Fri, 12 Jan 2018 16:50:47 +0000 (19:50 +0300)]
Remove obsolete advice about linking with _DYNAMIC=0 (Linux)
Issue #196 (bdwgc).
The mentioned GCC option is not needed since gc-5.0alpha3 (when
_DYNAMIC symbol was declared as weak in dyn_load.c).
Even more, the advice is problematic for Alpine Linux (and probably
other systems using musl) as building a program with gcc -static
produces a binary containing a _DYNAMIC symbol with a non-zero address,
and forcing the address to 0 causes a crash during GC initialization.
* doc/README.linux: Remove note that static executable should be
linked with the gcc option "-Wl,-defsym,_DYNAMIC=0".
Ivan Maidanski [Thu, 11 Jan 2018 23:57:32 +0000 (02:57 +0300)]
Add check that gc_cpp operator delete is called (test_cpp)
Issue #195 (bdwgc).
* tests/test_cpp.cc (GC_CHECKED_DELETE): New macro.
* tests/test_cpp.cc (main): Use GC_CHECKED_DELETE() instead of operator
delete (where it is expected that the operator defined in gc_cpp.c/h
is called).
Ivan Maidanski [Thu, 11 Jan 2018 22:08:37 +0000 (01:08 +0300)]
New field (expl_freed_bytes_since_gc) in public prof_stats_s
* include/gc.h (GC_prof_stats_s): Add expl_freed_bytes_since_gc field
(to the end of the structure).
* misc.c [!GC_GET_HEAP_USAGE_NOT_NEEDED] (fill_prof_stats): Store
GC_bytes_freed value to pstats->expl_freed_bytes_since_gc.
Ivan Maidanski [Thu, 11 Jan 2018 21:56:36 +0000 (00:56 +0300)]
New API function (get_expl_freed_bytes_since_gc)
Note: this function could be used in test_cpp to check that the proper
operator delete is called.
* include/gc.h (GC_get_expl_freed_bytes_since_gc): New function
prototype.
* mallocx.c (GC_get_expl_freed_bytes_since_gc): New function definition
(which returns GC_bytes_freed).
Ivan Maidanski [Thu, 28 Dec 2017 08:18:25 +0000 (11:18 +0300)]
Do not call sem_getvalue in stop_world if one thread exists
* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL] (GC_stop_world):
Do not call sem_getvalue() if n_live_threads is zero (matters only if
GC_retry_signals).
Ivan Maidanski [Wed, 27 Dec 2017 08:09:48 +0000 (11:09 +0300)]
Fix error code in abort message if sem_wait failed in start_world (NetBSD)
* pthread_stop_world.c [GC_NETBSD_THREADS_WORKAROUND] (GC_start_world):
Print errno instead of the value returned by sem_wait if the latter has
failed; remove "code" local variable.
Ivan Maidanski [Fri, 15 Dec 2017 07:15:05 +0000 (10:15 +0300)]
Workaround TSan hang in free_inner when called from at-fork child handler
* pthread_support.c [CAN_HANDLE_FORK] (GC_remove_all_threads_but_me):
Do not call GC_INTERNAL_FREE(p) if THREAD_SANITIZER and CAN_CALL_ATFORK;
add comment.
Ivan Maidanski [Thu, 14 Dec 2017 07:24:34 +0000 (10:24 +0300)]
Workaround TSan false positive in remove_all_threads_but_me
* pthread_support.c [CAN_HANDLE_FORK] (store_to_threads_table): New
static function (defined with GC_ATTR_NO_SANITIZE_THREAD attribute
if CAN_CALL_ATFORK); add comment.
* pthread_support.c [CAN_CALL_ATFORK] (GC_remove_all_threads_but_me):
Call store_to_threads_table instead of GC_threads[hv]=me.
Ivan Maidanski [Tue, 12 Dec 2017 16:24:02 +0000 (19:24 +0300)]
Fix gctest failure if compiled with TSan and parallel marker
As of clang-4.0, Thread Sanitizer does not support creation of threads
in the forked process (before exec). So, GC_start_mark_threads()
is a no-op now if TSan is enabled.
* misc.c [THREADS && PARALLEL_MARK && CAN_HANDLE_FORK]
(GC_start_mark_threads): Do not call GC_start_mark_threads_inner()
if THREAD_SANITIZER; add comment.
Fix marking of disclaim-reachable objects in the incremental mode
Issue #192 (bdwgc).
Unconditional marking is done analogously to marking from uncollectible
blocks, since they are meant to secure access to data for disclaim
notifiers after the data is no longer marked.
* mark.c [ENABLE_DISCLAIM] (GC_push_next_marked_dirty): Honor
MARK_UNCONDITIONALLY when rescanning dirty blocks during incremental
marking.
Ivan Maidanski [Fri, 1 Dec 2017 16:38:02 +0000 (19:38 +0300)]
Fix data race in do_local_mark when comparing active_count to helper_count
* mark.c [PARALLEL_MARK] (has_inactive_helpers): New static function
(which compares GC_active_count to GC_helper_count with the mark lock
held).
* mark.c [PARALLEL_MARK && GC_ASSERTIONS] (GC_do_local_mark): Remove
GC_acquire_mark_lock() with immediate GC_acquire_mark_unlock(); add
the appropriate comment to the function description instead.
* mark.c [PARALLEL_MARK] (GC_do_local_mark): Call has_inactive_helpers()
instead of GC_active_count<GC_helper_count; do not call
has_inactive_helpers() unless local_top>local_mark_stack+1.
Ivan Maidanski [Fri, 1 Dec 2017 08:48:08 +0000 (11:48 +0300)]
Workaround TSan warning about data race in generic_malloc_many
It is acceptable to update GC_bytes_found counter without
synchronization (so, it is OK to have inaccurate value in
GC_bytes_found in favor of performance).
* mallocx.c [PARALLEL_MARK && THREAD_SANITIZER]
(GC_generic_malloc_many): Acquire the allocation lock to update
GC_bytes_found (after successful GC_reclaim_generic).
* mallocx.c [PARALLEL_MARK && !THREAD_SANITIZER]
(GC_generic_malloc_many): Move GC_bytes_found update into the block
executed with the mark lock held.
Ivan Maidanski [Wed, 29 Nov 2017 22:18:19 +0000 (01:18 +0300)]
Remove done_init static variable from fnlz_mlc.c
(code refactoring)
GC_init_finalized_malloc!=0 is used instead of boolean done_init.
* fnlz_mlc.c [ENABLE_DISCLAIM] (done_init): Remove static variable.
* fnlz_mlc.c [ENABLE_DISCLAIM] (GC_init_finalized_malloc,
GC_finalized_malloc): Use GC_finalized_kind instead of done_init.
* fnlz_mlc.c [ENABLE_DISCLAIM] (GC_init_finalized_malloc): Add
assertion that the result of GC_new_kind_inner() is non-zero.
Ivan Maidanski [Wed, 29 Nov 2017 18:26:28 +0000 (21:26 +0300)]
Add a sanity check that load_acquire and store_release are available
* pthread_stop_world.c [(!AO_HAVE_load_acquire
|| !AO_HAVE_store_release) && !CPPCHECK]: Issue error that
libatomic_ops does not define AO_load_acquire and/or AO_store_release
so the client should manually define AO_REQUIRE_CAS macro.
Ivan Maidanski [Wed, 29 Nov 2017 18:09:52 +0000 (21:09 +0300)]
Remove explicit case of TRUE/FALSE to AO_t in suspend/resume_thread
(fix commit ce09fd5)
* pthread_stop_world.c [GC_ENABLE_SUSPEND_THREAD
&& !GC_OPENBSD_UTHREADS && !NACL] (GC_suspend_thread,
GC_resume_thread): Do not case TRUE/FALSE to AO_t explicitly.
Ivan Maidanski [Wed, 29 Nov 2017 17:15:49 +0000 (20:15 +0300)]
Fix data race in make_descriptor when setting explicit_typing_initialized
(fix commit 5f350a0)
* typd_mlc.c [AO_HAVE_load_acquire] (GC_explicit_typing_initialized):
Add volatile qualifier.
* typd_mlc.c [AO_HAVE_load_acquire] (GC_make_descriptor): Remove
cast of &GC_explicit_typing_initialized; do not check THREADS macro;
fallback to locked checking of GC_explicit_typing_initialized if
AO_HAVE_store_release is not defined; use AO_store_release to set
GC_explicit_typing_initialized (to true).
* typd_mlc.c (GC_make_descriptor): Reformat code dealing with
GC_explicit_typing_initialized (check whether AO_HAVE_load_acquire is
defined only once).
Ivan Maidanski [Wed, 29 Nov 2017 08:46:05 +0000 (11:46 +0300)]
Eliminate TSan false positive for stop_info.stack_ptr (v2)
Without this patch, Thread Sanitizer reports a data race between
GC_has_other_debug_info() and the code which sets stop_info.stack_ptr.
* include/private/pthread_support.h [THREAD_SANITIZER]: Include
dbg_mlc.h file.
* include/private/pthread_support.h [THREAD_SANITIZER] (GC_Thread_Rep):
Add dummy field (as the first field of the structure).
* pthread_support.c [THREAD_SANITIZER && CPPCHECK] (GC_new_thread):
Call GC_noop1(first_thread.dummy[0]) (to suppress cppcheck warning
about unused field).
projects / gc / log