Ivan Maidanski [Fri, 18 May 2018 18:36:37 +0000 (21:36 +0300)]
Fix comments style in configure.ac and Makefile.am
(a cherry-pick of commit 05f4d2c7 from 'release-7_6')
Do not allow the comments that are meaningless in the auto-generated
configure and Makefile.in to be put to these files.
* Makefile.am (EXTRA_DIST, SINGLE_GC_OBJ, EXTRA_DIST,
include doc/doc.am): Change comment style from "#" to "##".
* Makefile.am (CPLUSPLUS): Move "#" inside the conditional block.
* configure.ac: Update the copyright.
* configure.ac: Change comment style from "#" to "dnl" for the comments
that are meaningless in auto-generated configure file.
Naoyuki Sawa [Wed, 2 May 2018 04:10:23 +0000 (13:10 +0900)]
Fix register_dynamic_libraries on Windows 10
Issue #219 (bdwgc).
In the past (Windows XP, Windows 7, and older Windows 10), VirtualQuery
always returned PAGE_READWRITE for the data section.
In the April 2018 Update of Windows 10, it seems that PAGE_WRITECOPY
is returned sometimes (thus causing GC_register_dynamic_libraries not
to call GC_cond_add_roots for the section).
* dyn_load.c [MSWIN32 || MSWINCE || CYGWIN32]
(GC_register_dynamic_libraries): Call GC_cond_add_roots()
also for PAGE_EXECUTE_WRITECOPY and PAGE_WRITECOPY pages.
Ivan Maidanski [Tue, 10 Apr 2018 08:07:08 +0000 (11:07 +0300)]
Fix data race during apply_to_each_object(reset_back_edge)
(back-port of commit d7ea0b1 from 'master')
Issue #144 (bdwgc).
* backgraph.c [MAKE_BACK_GRAPH] (GC_print_back_graph_stats): Add
assertion that the allocation lock is held; store GC_deepest_obj to
obj local variable (while holding the lock); place UNLOCK/LOCK around
GC_print_heap_obj() call; pass obj to GC_print_heap_obj().
* finalize.c [MAKE_BACK_GRAPH] (GC_notify_or_invoke_finalizers): Remove
UNLOCK/LOCK around GC_print_back_graph_stats() call.
Ivan Maidanski [Tue, 17 Apr 2018 21:20:46 +0000 (00:20 +0300)]
Fix the collector hang when it is configured with --enable-gc-debug
(back-port of commits 7881f44e, 8f63e5b3, 9a46da7e from 'release-7_6')
Issue #205 (bdwgc).
* dbg_mlc.c (GC_store_debug_info_inner): Remove comment (as it exists
in the header); change from STATIC to GC_INNER; change p type from
ptr_t to void*.
* dbg_mlc.c [GC_ASSERTIONS] (GC_store_debug_info_inner): Cast p to
ptr_t when passed to CROSSES_HBLK().
* dbg_mlc.c (GC_store_debug_info): Change from GC_INNER to static;
remove GC_ prefix; replace ptr_t to void*; add fn argument;
replace "const char *string, int linenum" with GC_EXTRA_PARAMS;
change "word sz" argument to "size_t lb"; allow p to be null (print
error message in this case); call GC_start_debugging_inner unless
GC_debugging_started; call ADD_CALL_CHAIN.
* dbg_mlc.c (GC_start_debugging_inner): Define as GC_INNER (instead of
STATIC).
* dbg_mlc.c (GC_start_debugging): Remove.
* dbg_mlc.c (GC_debug_malloc, GC_debug_malloc_ignore_off_page,
GC_debug_malloc_atomic_ignore_off_page, GC_debug_generic_malloc,
GC_debug_malloc_atomic, GC_debug_malloc_uncollectable): Call
store_debug_info() instead of checking result for null and calling
GC_start_debugging, ADD_CALL_CHAIN, GC_store_debug_info; use OPT_RA.
* dbg_mlc.c [STUBBORN_ALLOC] (GC_debug_malloc_stubborn): Likewise.
* dbg_mlc.c [GC_ATOMIC_UNCOLLECTABLE]
(GC_debug_malloc_atomic_uncollectable): Likewise.
* gcj_mlc.c [GC_GCJ_SUPPORT] (GC_debug_gcj_malloc): Call ADD_CALL_CHAIN
while holding the lock; call GC_store_debug_info_inner (holding the
lock) instead of GC_store_debug_info.
* gcj_mlc.c (GC_debug_gcj_malloc): Call GC_start_debugging_inner
(holding the lock) instead of GC_start_debugging.
* include/private/dbg_mlc.h (ADD_CALL_CHAIN): Update comment.
* include/private/gc_priv.h (GC_store_debug_info): Replace with
GC_store_debug_info_inner; update comment; change ptr_t to void*.
* include/private/gc_priv.h (GC_start_debugging): Rename to
GC_start_debugging_inner; improve usage comment.
* os_dep.c [SAVE_CALL_CHAIN] (GC_save_callers): Add assertion that the
allocation lock is held; add comment.
Ivan Maidanski [Fri, 6 Apr 2018 16:07:50 +0000 (19:07 +0300)]
Fix null dereference in print_callers on backtrace_symbols failure
* os_dep.c [NEED_CALLINFO && GC_HAVE_BUILTIN_BACKTRACE
&& !GC_BACKTRACE_SYMBOLS_BROKEN] (GC_print_callers): If sym_name is
NULL then print info[i].ci_pc to buf and set name to buf (instead of
to sym_name[0]).
* os_dep.c [NEED_CALLINFO && GC_HAVE_BUILTIN_BACKTRACE
&& !GC_BACKTRACE_SYMBOLS_BROKEN] (GC_print_callers): Do not call
free(sym_name) if sym_name is NULL.
Ivan Maidanski [Fri, 6 Apr 2018 06:32:12 +0000 (09:32 +0300)]
Fix null pointer dereference in get_private_path_and_zero_file (Symbian)
(back-port of commit 9d019e2e from 'master')
* extra/symbian.cpp (GC_get_private_path_and_zero_file): Do not call
memcpy() if allocation of copyChar is failed.
* os_dep.c [MMAP_SUPPORTED && !USE_MMAP_ANON] (zero_fd): Initialize
global variable to -1 (instead of 0).
* os_dep.c [MMAP_SUPPORTED && !USE_MMAP_ANON && SYMBIAN]
(GC_unix_mmap_get_mem): Do not call open() and free() if path is NULL.
Ivan Maidanski [Fri, 23 Mar 2018 08:13:23 +0000 (11:13 +0300)]
Fix thread_suspend fail for threads registered from key destructor (OS X)
Issue #213 (bdwgc).
* pthread_support.c [GC_DARWIN_THREADS] (GC_register_my_thread):
Reinitialize stop_info.mach_thread if the thread is registered from
the client thread key destructor; add comment.
Natanael Copa [Tue, 6 Feb 2018 20:21:23 +0000 (21:21 +0100)]
Fix gctest with musl libc on s390x
Issue #202 (bdwgc).
DEFAULT_STACK_MAYBE_SMALL is only set if PARALLEL_MARK is set, but stack
is too small on musl libc even if PARALLEL_MARK is unset. We solve this
by detecting musl via NO_GETCONTEXT.
Ivan Maidanski [Thu, 22 Feb 2018 20:39:38 +0000 (23:39 +0300)]
Fix type of local variables receiving result of PHT_HASH
(back-port of commit 037991eb from 'master')
The variables should be of word type, not int or size_t.
* os_dep.c (GC_write_fault_handler, GC_remove_protection): Change type
of index local variable from size_t to word.
* os_dep.c [MPROTECT_VDB && DARWIN] (catch_exception_raise): Change
type of index local variable from int to word.
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, 27 Dec 2017 08:09:48 +0000 (11:09 +0300)]
Fix error code in abort message if sem_wait failed in start_world (NetBSD)
(backport of commit 8f42658f from 'release-7_6')
* 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 [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.
Ivan Maidanski [Wed, 29 Nov 2017 08:42:47 +0000 (11:42 +0300)]
Eliminate 'this statement may fall through' GCC warnings
* cord/cordprnt.c (extract_conv_spec): Eliminate fall through in
a switch statement.
* cord/tests/de.c (do_command): Replace "fall through:" comment with
"FALLTHRU" formal comment (on a new line).
* win32_threads.c [!GC_PTHREADS && !GC_NO_THREADS_DISCOVERY]
(GC_DllMain): Likewise.
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
(Cherry-pick commit 66ae55e3 from 'master' branch.)
* 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 [Thu, 23 Nov 2017 23:21:46 +0000 (02:21 +0300)]
Fix lack of barriers to synchronize memory for suspend_handler
pthread_kill is not on the list of functions which synchronize memory.
* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL]
(GC_world_is_stopped): Reformat comment.
* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL]
(GC_suspend_handler_inner): Use AO_load_acquire instead of AO_load
to fetch the value of GC_stop_count; add comment.
* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL]
(GC_start_world): Use AO_store_release instead of AO_store to reset
GC_world_is_stopped; add comment.
Ivan Maidanski [Thu, 23 Nov 2017 17:15:54 +0000 (20:15 +0300)]
Fix data race in last_stop_count access (suspend_handler_inner)
(Cherry-pick commit 3961ab98 from 'release-7_6' branch.)
* include/private/pthread_stop_world.h [!GC_OPENBSD_UTHREADS]
(thread_stop_info.last_stop_count): Do not define if NACL; change the
type from word to AO_t; add volatile qualifier; fix a typo in comment
("GC_stop_count").
* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL]
(GC_suspend_handler_inner): Replace
me->stop_info.last_stop_count=my_stop_count with
AO_store_release(&me->stop_info.last_stop_count,my_stop_count).
* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL]
(GC_suspend_all): Replace p->stop_info.last_stop_count with
AO_load(&p->stop_info.last_stop_count).
Ivan Maidanski [Fri, 10 Nov 2017 17:07:16 +0000 (20:07 +0300)]
Fix data race when getting object size in explicitly-typed allocators
(Cherry-pick commit eb33bda5 from 'release-7_6' branch.)
* typd_mlc.c (COMPLEX): Reformat comment.
* typd_mlc.c (GC_malloc_explicitly_typed,
GC_malloc_explicitly_typed_ignore_off_page, GC_calloc_explicitly_typed):
Always use BYTES_TO_GRANULES(GC_size(op)) instead of GC_size_map[lb] to
determine size of the allocated object in granules (because the value
of GC_size_map[lb] might be updated by another thread since the value
use in GC_malloc_kind); add comment.
Ivan Maidanski [Fri, 3 Nov 2017 17:08:09 +0000 (20:08 +0300)]
Fix data race in a list referenced by A.aa (gctest)
(Cherry-pick commit 2f1b8e8 from 'release-7_6' branch.)
* tests/test.c [!AO_CLEAR] (AO_t): Define.
* tests/test.c [!AO_HAVE_load_acquire] (AO_load_acquire): New static
function (which uses FINALIZER_LOCK/FINALIZER_UNLOCK).
* tests/test.c [!AO_HAVE_store_release] (AO_store_release): Likewise.
* tests/test.c (A): Add volatile qualifier; change the type of aa
field from sexpr to AO_t.
* tests/test.c (a): Do not define (and undefine).
* tests/test.c (a_set, a_get): New macro (based on AO_store_release
and AO_load_acquire, respectively).
* tests/test.c (reverse_test_inner): Use a_set() and a_get() to store
and fetch the value of a, respectively.
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 [Fri, 3 Nov 2017 08:03:04 +0000 (11:03 +0300)]
Avoid data race in finalized_count (gctest)
(Cherry-pick commit ffc89ea1 from 'release-7_6' branch.)
* tests/test.c (FINALIZER_LOCK, FINALIZER_UNLOCK): New macro.
* tests/test.c [GC_PTHREADS] (incr_lock): Move static variable out
of finalizer() and mktree() (use a single lock variable to access
finalized_count in finalizer, mktree and tree_test).
* tests/test.c (dropped_something): Change type back to int; remove
volatile.
* tests/test.c (finalizer, mktree): Use FINALIZER_[UN]LOCK() instead
of PCR_ThCrSec_Enter/ExitSys(), pthread_mutex_[un]lock(),
Enter/LeaveCriticalSection().
* tests/test.c (tree_test): Use FINALIZER_[UN]LOCK() to avoid data
race when accessing finalized_count and dropped_something; do not use
AO_load() and AO_store() to access dropped_something.
Ivan Maidanski [Mon, 16 Oct 2017 08:00:53 +0000 (11:00 +0300)]
Allow PKG_CHECK_MODULES in configure.ac to be commented out easily
(Cherry-pick commits 3d34335, fbb22e9 from 'master' branch.)
* configure.ac [$with_libatomic_ops!=no && $THREADS!=none]
(missing_libatomic_ops): Set to true (before PKG_CHECK_MODULES).
* configure.ac [$missing_libatomic_ops] (PKG_CHECK_MODULES): Set
missing_libatomic_ops back to false if atomic_ops module is not found;
add comment (about pkg-config and the possibility to comment out
PKG_CHECK_MODULES).
Ivan Maidanski [Thu, 24 Aug 2017 08:34:05 +0000 (11:34 +0300)]
Eliminate most of collisions in GC_threads on Linux/x64
(Cherry-pick commits 85fce54, e897b41 from 'master' branch.)
On some targets (e.g., Linux/x64) pthread_t values always have zero
in the lowest byte (e.g. if pthread_t is actually a page-aligned
pointer) thus causing all the created thread descriptors to have
a collision at GC_threads[0]. This commit mixes 3 lowest bytes
of thread id with a XOR operation (thus greatly reduces the
probability of collisions in GC_threads).
Windows thread Id rarely has non-zero highest half of DWORD, so
it is OK to use only the lowest one for hash value computation.
* include/private/pthread_support.h (THREAD_TABLE_SZ): Do not
define if already defined; refine comment.
* pthread_support.c (THREAD_TABLE_INDEX): New macro (similar to that
in win32_threads.c).
* pthread_support.c (GC_new_thread, GC_delete_thread,
GC_delete_gc_thread, GC_lookup_thread): Replace
NUMERIC_THREAD_ID(id)%THREAD_TABLE_SZ to THREAD_TABLE_INDEX(id).
* pthread_support.c (GC_lookup_thread): Remove hv local variable.
* win32_threads.c (THREAD_TABLE_INDEX): Replace (id>>2) to
((id>>8)^id).
* win32_threads.c [GC_PTHREADS] (GC_lookup_pthread): Define win32_id
local variable to pass it to THREAD_TABLE_INDEX() instead of passing
GET_PTHREAD_MAP_CACHE(id) directly.
Ivan Maidanski [Wed, 23 Aug 2017 08:46:41 +0000 (11:46 +0300)]
Change type of THREAD_TABLE_INDEX result to int in win32_threads.c
(code refactoring)
* win32_threads.c (THREAD_TABLE_INDEX): Cast result to int; remove cast
of argument to word type; add comment.
* win32_threads.c (GC_new_thread, GC_delete_gc_thread_no_free,
GC_delete_thread): Change type of hv local variable from word to int.
* win32_threads.c (GC_lookup_thread_inner): Remove hv local variable;
remove "register" keyword for p local variable.
* win32_threads.c (GC_lookup_pthread): Change type of hv_guess local
variable from word to int.
Ivan Maidanski [Tue, 26 Sep 2017 21:56:14 +0000 (00:56 +0300)]
Fix get_maps on proc maps file asynchronous growth
Issue #179 (bdwgc).
Asynchronous growth of /proc/self/maps could be caused by some thread
which is not currently registered (e.g. stack allocation on thread
creation or a non-GC heap allocation from an unregistered thread).
Ivan Maidanski [Tue, 19 Sep 2017 07:51:24 +0000 (10:51 +0300)]
Fix infinite mark_some calls after memory mapping disappeared (Glibc)
Issue #179 (bdwgc).
* mark.c [WRAP_MARK_SOME && REGISTER_LIBRARIES_EARLY] (GC_mark_some):
Call GC_cond_register_dynamic_libraries (with the world running)
before GC_invalidate_mark_state().
Ivan Maidanski [Mon, 18 Sep 2017 07:54:37 +0000 (10:54 +0300)]
Fix SIGSEGV in mark_from called from do_local_mark if WRAP_MARK_SOME
(Cherry-pick commits f4d690e, b874d02 from 'master' branch.)
Issue #179 (bdwgc).
Parallel marker is disabled (for now) if there is a chance of
scanning unmapped data roots (because GC_do_local_mark is not
prepared to deal with memory protection faults gracefully).
* include/private/gc_pmark.h [HAVE_CONFIG_H]: Do not include config.h
if already included from gc_priv.h (to avoid PARALLEL_MARK macro
redefinition in config.h after it was undefined in gcconfig.h).
* mark.c [MSWIN32 || USE_PROC_FOR_LIBRARIES] (WRAP_MARK_SOME): Move
definition to gcconfig.h.
* include/private/gcconfig.h [WRAP_MARK_SOME && PARALLEL_MARK]
(PARALLEL_MARK): Undefine (as not implemented yet).
Ivan Maidanski [Wed, 27 Sep 2017 07:52:09 +0000 (10:52 +0300)]
Fix pthread_detach for threads not yet registered (Cygwin, winpthreads)
* win32_threads.c [GC_PTHREADS] (GC_pthread_detach): Repeat
GC_lookup_pthread() (with a 10 msec delay) while t is null; add comment
and TODO; remove ABORT if t is null (as t is never null after
pthread_detach call).
Ivan Maidanski [Tue, 26 Sep 2017 08:44:44 +0000 (11:44 +0300)]
Fix pthread_join when thread is registered in thread key destructor
* pthread_support.c (WRAP_FUNC(pthread_join)): Remove assertion that
thread is finished; call GC_delete_gc_thread only if thread is
finished; add comment.
* win32_threads.c [GC_PTHREADS] (GC_pthread_join): Call
GC_delete_gc_thread_no_free and GC_INTERNAL_FREE only if thread is
finished.
Ivan Maidanski [Tue, 26 Sep 2017 08:01:29 +0000 (11:01 +0300)]
Fix removal of dead threads in a child process
GC_threads table may contain several elements with the same pthread id,
so GC_remove_all_threads_but_me should remove all elements except for
the first found element matching the id of the current thread.
* pthread_support.c [CAN_HANDLE_FORK] (GC_remove_all_threads_but_me):
Do not assign me if it is already non-null; add comment.
* win32_threads.c [CAN_HANDLE_FORK] (GC_remove_all_threads_but_me):
Likewise.
* win32_threads.c [CAN_HANDLE_FORK] (GC_remove_all_threads_but_me):
Remove GC_ASSERT(me==NULL).
Ivan Maidanski [Tue, 26 Sep 2017 08:29:54 +0000 (11:29 +0300)]
Fix pthread_join to avoid thread removal on failure (Cygwin, winpthreads)
(Cherry-pick commit 29269f1 from 'release-7_6' branch.)
* win32_threads.c [GC_WIN32_PTHREADS] (GC_pthread_join): Do not call
GC_lookup_pthread if pthread_join failed.
* win32_threads.c [GC_PTHREADS] (GC_pthread_join): Do not call
GC_delete_gc_thread_no_free, GC_INTERNAL_FREE if pthread_join failed.
* win32_threads.c [GC_PTHREADS && DEBUG_THREADS] (GC_pthread_join):
Log whether pthread_join succeeded or failed.
Ivan Maidanski [Mon, 28 Aug 2017 06:27:07 +0000 (09:27 +0300)]
Fix GC_remove_specific invocation from remove_all_threads_but_me
(Cherry-pick commit 3d85c61 from 'release-7_6 branch.)
* include/private/specific.h (GC_remove_specific): Define as macro
(calls GC_remove_specific_after_fork).
* include/private/specific.h (GC_remove_specific_after_fork): New
GC_INNER function declaration.
* include/private/thread_local_alloc.h [USE_PTHREAD_SPECIFIC
|| USE_COMPILER_TLS || USE_WIN32_COMPILER_TLS || USE_WIN32_SPECIFIC]
(GC_remove_specific_after_fork): New macro (defined to no-op).
* pthread_support.c [CAN_HANDLE_FORK && THREAD_LOCAL_ALLOC]
(GC_remove_all_threads_but_me): Call GC_remove_specific_after_fork
instead of GC_remove_specific (i.e. delete thread-specific control data
for the given thread instead of the current one).
* win32_threads.c [CAN_HANDLE_FORK && THREAD_LOCAL_ALLOC] Like
* specific.c [USE_CUSTOM_SPECIFIC] (GC_setspecific): Add assertion that
the allocation lock is held; remove the corresponding comment.
* specific.c [USE_CUSTOM_SPECIFIC] (GC_remove_specific): Replace to
GC_remove_specific_after_fork (add t argument); replace self local
variable with t argument; update comment;
* specific.c [USE_CUSTOM_SPECIFIC && CAN_HANDLE_FORK]
(GC_remove_specific_after_fork): Add assertion that the allocation lock
is held; add comment.
Ivan Maidanski [Tue, 19 Sep 2017 22:23:43 +0000 (01:23 +0300)]
Fix thread id leaks in subthread_create and threadkey_test
(Cherry-pick commit 7574e33 from 'release-7_6' branch.)
* tests/subthread_create.c [GC_PTHREADS] (entry): Call pthread_detach
for the created thread.
* tests/threadkey_test.c (main): Likewise.
* tests/threadkey_test.c (on_thread_exit_inner): New local variable
attr; call pthread_attr_init and pthread_attr_setdetachstate;
pass attr to GC_pthread_create; call pthread_attr_destroy.
Ivan Maidanski [Tue, 22 Aug 2017 08:23:27 +0000 (11:23 +0300)]
Fix deadlock in GC_help_marker caused by use of mark_cv of parent process
(Cherry-pick commit 9b7ef4d from 'release-7_6' branch.)
The marker threads of the parent process are blocked on mark_cv at
fork. So pthread_cond_wait() malfunction (hang) is possible in the
child process without mark_cv state cleanup (reinitialization).
* pthread_support.c [PARALLEL_MARK] (mark_cv): Move static variable
definition upper to be before start_mark_threads).
* win32_threads.c [GC_PTHREADS_PARAMARK] (mark_cv): Likewise.
* pthread_support.c [PARALLEL_MARK && CAN_HANDLE_FORK] (mark_cv):
Do not initialize statically; add comment.
* win32_threads.c [GC_PTHREADS_PARAMARK && CAN_HANDLE_FORK] (mark_cv):
Likewise.
* pthread_support.c [PARALLEL_MARK && CAN_HANDLE_FORK]
(start_mark_threads): Initialize mark_cv to
PTHREAD_COND_INITIALIZER (unless available_markers_m1 <= 0 or
GC_parallel); add comment.
* win32_threads.c [GC_PTHREADS_PARAMARK && CAN_HANDLE_FORK]
(start_mark_threads): Likewise.
* pthread_support.c [PARALLEL_MARK] (GC_wait_marker,
GC_notify_all_marker): Add assertion that GC_parallel is true (so
mark_cv is initialized).
* win32_threads.c [GC_PTHREADS_PARAMARK] (GC_wait_marker,
GC_notify_all_marker): Likewise.
Ivan Maidanski [Tue, 22 Aug 2017 05:56:31 +0000 (08:56 +0300)]
Avoid busy waiting in mark_thread while GC_parallel is false
(Cherry-pick commit 46d2a44 from 'release-7_6' branch.)
* mark.c [PARALLEL_MARK] (GC_help_marker): Add assertion that
GC_parallel is true (i.e. GC_markers_m1 is non-zero).
* pthread_support.c [PARALLEL_MARK && CAN_HANDLE_FORK]
(start_mark_threads): Set GC_markers_m1 value before starting
the first marker thread (it is already set if fork handling is off).
* win32_threads.c [GC_PTHREADS_PARAMARK && CAN_HANDLE_FORK]
(start_mark_threads): Likewise.
* pthread_support.c [PARALLEL_MARK] (start_mark_threads):
Adjust GC_markers_m1 value only if pthread_create failed.
* win32_threads.c [GC_PTHREADS_PARAMARK] (start_mark_threads):
Likewise.
projects / gc / log