3 * Simple generational GC.
6 * Paolo Molaro (lupus@ximian.com)
7 * Rodrigo Kumpera (kumpera@gmail.com)
9 * Copyright 2005-2011 Novell, Inc (http://www.novell.com)
10 * Copyright 2011 Xamarin Inc (http://www.xamarin.com)
12 * Thread start/stop adapted from Boehm's GC:
13 * Copyright (c) 1994 by Xerox Corporation. All rights reserved.
14 * Copyright (c) 1996 by Silicon Graphics. All rights reserved.
15 * Copyright (c) 1998 by Fergus Henderson. All rights reserved.
16 * Copyright (c) 2000-2004 by Hewlett-Packard Company. All rights reserved.
17 * Copyright 2001-2003 Ximian, Inc
18 * Copyright 2003-2010 Novell, Inc.
19 * Copyright 2011 Xamarin, Inc.
20 * Copyright (C) 2012 Xamarin Inc
22 * Licensed under the MIT license. See LICENSE file in the project root for full license information.
24 * Important: allocation provides always zeroed memory, having to do
25 * a memset after allocation is deadly for performance.
26 * Memory usage at startup is currently as follows:
28 * 64 KB internal space
30 * We should provide a small memory config with half the sizes
32 * We currently try to make as few mono assumptions as possible:
33 * 1) 2-word header with no GC pointers in it (first vtable, second to store the
35 * 2) gc descriptor is the second word in the vtable (first word in the class)
36 * 3) 8 byte alignment is the minimum and enough (not true for special structures (SIMD), FIXME)
37 * 4) there is a function to get an object's size and the number of
38 * elements in an array.
39 * 5) we know the special way bounds are allocated for complex arrays
40 * 6) we know about proxies and how to treat them when domains are unloaded
42 * Always try to keep stack usage to a minimum: no recursive behaviour
43 * and no large stack allocs.
45 * General description.
46 * Objects are initially allocated in a nursery using a fast bump-pointer technique.
47 * When the nursery is full we start a nursery collection: this is performed with a
49 * When the old generation is full we start a copying GC of the old generation as well:
50 * this will be changed to mark&sweep with copying when fragmentation becomes to severe
51 * in the future. Maybe we'll even do both during the same collection like IMMIX.
53 * The things that complicate this description are:
54 * *) pinned objects: we can't move them so we need to keep track of them
55 * *) no precise info of the thread stacks and registers: we need to be able to
56 * quickly find the objects that may be referenced conservatively and pin them
57 * (this makes the first issues more important)
58 * *) large objects are too expensive to be dealt with using copying GC: we handle them
59 * with mark/sweep during major collections
60 * *) some objects need to not move even if they are small (interned strings, Type handles):
61 * we use mark/sweep for them, too: they are not allocated in the nursery, but inside
62 * PinnedChunks regions
68 *) we could have a function pointer in MonoClass to implement
69 customized write barriers for value types
71 *) investigate the stuff needed to advance a thread to a GC-safe
72 point (single-stepping, read from unmapped memory etc) and implement it.
73 This would enable us to inline allocations and write barriers, for example,
74 or at least parts of them, like the write barrier checks.
75 We may need this also for handling precise info on stacks, even simple things
76 as having uninitialized data on the stack and having to wait for the prolog
77 to zero it. Not an issue for the last frame that we scan conservatively.
78 We could always not trust the value in the slots anyway.
80 *) modify the jit to save info about references in stack locations:
81 this can be done just for locals as a start, so that at least
82 part of the stack is handled precisely.
84 *) test/fix endianess issues
86 *) Implement a card table as the write barrier instead of remembered
87 sets? Card tables are not easy to implement with our current
88 memory layout. We have several different kinds of major heap
89 objects: Small objects in regular blocks, small objects in pinned
90 chunks and LOS objects. If we just have a pointer we have no way
91 to tell which kind of object it points into, therefore we cannot
92 know where its card table is. The least we have to do to make
93 this happen is to get rid of write barriers for indirect stores.
96 *) Get rid of write barriers for indirect stores. We can do this by
97 telling the GC to wbarrier-register an object once we do an ldloca
98 or ldelema on it, and to unregister it once it's not used anymore
99 (it can only travel downwards on the stack). The problem with
100 unregistering is that it needs to happen eventually no matter
101 what, even if exceptions are thrown, the thread aborts, etc.
102 Rodrigo suggested that we could do only the registering part and
103 let the collector find out (pessimistically) when it's safe to
104 unregister, namely when the stack pointer of the thread that
105 registered the object is higher than it was when the registering
106 happened. This might make for a good first implementation to get
107 some data on performance.
109 *) Some sort of blacklist support? Blacklists is a concept from the
110 Boehm GC: if during a conservative scan we find pointers to an
111 area which we might use as heap, we mark that area as unusable, so
112 pointer retention by random pinning pointers is reduced.
114 *) experiment with max small object size (very small right now - 2kb,
115 because it's tied to the max freelist size)
117 *) add an option to mmap the whole heap in one chunk: it makes for many
118 simplifications in the checks (put the nursery at the top and just use a single
119 check for inclusion/exclusion): the issue this has is that on 32 bit systems it's
120 not flexible (too much of the address space may be used by default or we can't
121 increase the heap as needed) and we'd need a race-free mechanism to return memory
122 back to the system (mprotect(PROT_NONE) will still keep the memory allocated if it
123 was written to, munmap is needed, but the following mmap may not find the same segment
126 *) memzero the major fragments after restarting the world and optionally a smaller
129 *) investigate having fragment zeroing threads
131 *) separate locks for finalization and other minor stuff to reduce
134 *) try a different copying order to improve memory locality
136 *) a thread abort after a store but before the write barrier will
137 prevent the write barrier from executing
139 *) specialized dynamically generated markers/copiers
141 *) Dynamically adjust TLAB size to the number of threads. If we have
142 too many threads that do allocation, we might need smaller TLABs,
143 and we might get better performance with larger TLABs if we only
144 have a handful of threads. We could sum up the space left in all
145 assigned TLABs and if that's more than some percentage of the
146 nursery size, reduce the TLAB size.
148 *) Explore placing unreachable objects on unused nursery memory.
149 Instead of memset'ng a region to zero, place an int[] covering it.
150 A good place to start is add_nursery_frag. The tricky thing here is
151 placing those objects atomically outside of a collection.
153 *) Allocation should use asymmetric Dekker synchronization:
154 http://blogs.oracle.com/dave/resource/Asymmetric-Dekker-Synchronization.txt
155 This should help weak consistency archs.
162 #define _XOPEN_SOURCE
163 #define _DARWIN_C_SOURCE
169 #ifdef HAVE_PTHREAD_H
172 #ifdef HAVE_PTHREAD_NP_H
173 #include <pthread_np.h>
181 #include "mono/sgen/sgen-gc.h"
182 #include "mono/sgen/sgen-cardtable.h"
183 #include "mono/sgen/sgen-protocol.h"
184 #include "mono/sgen/sgen-memory-governor.h"
185 #include "mono/sgen/sgen-hash-table.h"
186 #include "mono/sgen/sgen-pinning.h"
187 #include "mono/sgen/sgen-workers.h"
188 #include "mono/sgen/sgen-client.h"
189 #include "mono/sgen/sgen-pointer-queue.h"
190 #include "mono/sgen/gc-internal-agnostic.h"
191 #include "mono/utils/mono-proclib.h"
192 #include "mono/utils/mono-memory-model.h"
193 #include "mono/utils/hazard-pointer.h"
195 #include <mono/utils/memcheck.h>
196 #include <mono/utils/mono-mmap-internals.h>
198 #undef pthread_create
200 #undef pthread_detach
203 * ######################################################################
204 * ######## Types and constants used by the GC.
205 * ######################################################################
208 /* 0 means not initialized, 1 is initialized, -1 means in progress */
209 static int gc_initialized = 0;
210 /* If set, check if we need to do something every X allocations */
211 gboolean has_per_allocation_action;
212 /* If set, do a heap check every X allocation */
213 guint32 verify_before_allocs = 0;
214 /* If set, do a minor collection before every X allocation */
215 guint32 collect_before_allocs = 0;
216 /* If set, do a whole heap check before each collection */
217 static gboolean whole_heap_check_before_collection = FALSE;
218 /* If set, do a remset consistency check at various opportunities */
219 static gboolean remset_consistency_checks = FALSE;
220 /* If set, do a mod union consistency check before each finishing collection pause */
221 static gboolean mod_union_consistency_check = FALSE;
222 /* If set, check whether mark bits are consistent after major collections */
223 static gboolean check_mark_bits_after_major_collection = FALSE;
224 /* If set, check that all nursery objects are pinned/not pinned, depending on context */
225 static gboolean check_nursery_objects_pinned = FALSE;
226 /* If set, do a few checks when the concurrent collector is used */
227 static gboolean do_concurrent_checks = FALSE;
228 /* If set, do a plausibility check on the scan_starts before and after
230 static gboolean do_scan_starts_check = FALSE;
232 static gboolean disable_minor_collections = FALSE;
233 static gboolean disable_major_collections = FALSE;
234 static gboolean do_verify_nursery = FALSE;
235 static gboolean do_dump_nursery_content = FALSE;
236 static gboolean enable_nursery_canaries = FALSE;
238 static gboolean precleaning_enabled = TRUE;
239 static gboolean dynamic_nursery = FALSE;
240 static size_t min_nursery_size = 0;
241 static size_t max_nursery_size = 0;
243 #ifdef HEAVY_STATISTICS
244 guint64 stat_objects_alloced_degraded = 0;
245 guint64 stat_bytes_alloced_degraded = 0;
247 guint64 stat_copy_object_called_nursery = 0;
248 guint64 stat_objects_copied_nursery = 0;
249 guint64 stat_copy_object_called_major = 0;
250 guint64 stat_objects_copied_major = 0;
252 guint64 stat_scan_object_called_nursery = 0;
253 guint64 stat_scan_object_called_major = 0;
255 guint64 stat_slots_allocated_in_vain;
257 guint64 stat_nursery_copy_object_failed_from_space = 0;
258 guint64 stat_nursery_copy_object_failed_forwarded = 0;
259 guint64 stat_nursery_copy_object_failed_pinned = 0;
260 guint64 stat_nursery_copy_object_failed_to_space = 0;
262 static guint64 stat_wbarrier_add_to_global_remset = 0;
263 static guint64 stat_wbarrier_arrayref_copy = 0;
264 static guint64 stat_wbarrier_generic_store = 0;
265 static guint64 stat_wbarrier_generic_store_atomic = 0;
266 static guint64 stat_wbarrier_set_root = 0;
269 static guint64 stat_pinned_objects = 0;
271 static guint64 time_minor_pre_collection_fragment_clear = 0;
272 static guint64 time_minor_pinning = 0;
273 static guint64 time_minor_scan_remsets = 0;
274 static guint64 time_minor_scan_major_blocks = 0;
275 static guint64 time_minor_scan_los = 0;
276 static guint64 time_minor_scan_pinned = 0;
277 static guint64 time_minor_scan_roots = 0;
278 static guint64 time_minor_finish_gray_stack = 0;
279 static guint64 time_minor_fragment_creation = 0;
281 static guint64 time_major_pre_collection_fragment_clear = 0;
282 static guint64 time_major_pinning = 0;
283 static guint64 time_major_scan_pinned = 0;
284 static guint64 time_major_scan_roots = 0;
285 static guint64 time_major_scan_mod_union_blocks = 0;
286 static guint64 time_major_scan_mod_union_los = 0;
287 static guint64 time_major_finish_gray_stack = 0;
288 static guint64 time_major_free_bigobjs = 0;
289 static guint64 time_major_los_sweep = 0;
290 static guint64 time_major_sweep = 0;
291 static guint64 time_major_fragment_creation = 0;
293 static guint64 time_max = 0;
295 static int sgen_max_pause_time = SGEN_DEFAULT_MAX_PAUSE_TIME;
296 static float sgen_max_pause_margin = SGEN_DEFAULT_MAX_PAUSE_MARGIN;
298 static SGEN_TV_DECLARE (time_major_conc_collection_start);
299 static SGEN_TV_DECLARE (time_major_conc_collection_end);
301 int gc_debug_level = 0;
303 static char* gc_params_options;
304 static char* gc_debug_options;
308 mono_gc_flush_info (void)
310 fflush (gc_debug_file);
314 #define TV_DECLARE SGEN_TV_DECLARE
315 #define TV_GETTIME SGEN_TV_GETTIME
316 #define TV_ELAPSED SGEN_TV_ELAPSED
318 static SGEN_TV_DECLARE (sgen_init_timestamp);
320 NurseryClearPolicy nursery_clear_policy = CLEAR_AT_TLAB_CREATION;
322 #define object_is_forwarded SGEN_OBJECT_IS_FORWARDED
323 #define object_is_pinned SGEN_OBJECT_IS_PINNED
324 #define pin_object SGEN_PIN_OBJECT
326 #define ptr_in_nursery sgen_ptr_in_nursery
328 #define LOAD_VTABLE SGEN_LOAD_VTABLE
331 nursery_canaries_enabled (void)
333 return enable_nursery_canaries;
336 #define safe_object_get_size sgen_safe_object_get_size
338 #if defined(HAVE_CONC_GC_AS_DEFAULT)
339 /* Use concurrent major on deskstop platforms */
340 #define DEFAULT_MAJOR SGEN_MAJOR_CONCURRENT
342 #define DEFAULT_MAJOR SGEN_MAJOR_SERIAL
348 SGEN_MAJOR_CONCURRENT,
349 SGEN_MAJOR_CONCURRENT_PARALLEL
355 SGEN_MINOR_SIMPLE_PARALLEL,
362 SGEN_MODE_THROUGHPUT,
367 * ######################################################################
368 * ######## Global data.
369 * ######################################################################
371 MonoCoopMutex gc_mutex;
373 #define SCAN_START_SIZE SGEN_SCAN_START_SIZE
375 size_t degraded_mode = 0;
377 static mword bytes_pinned_from_failed_allocation = 0;
379 GCMemSection *nursery_section = NULL;
380 static volatile mword lowest_heap_address = ~(mword)0;
381 static volatile mword highest_heap_address = 0;
383 MonoCoopMutex sgen_interruption_mutex;
385 int current_collection_generation = -1;
386 volatile gboolean concurrent_collection_in_progress = FALSE;
388 /* objects that are ready to be finalized */
389 static SgenPointerQueue fin_ready_queue = SGEN_POINTER_QUEUE_INIT (INTERNAL_MEM_FINALIZE_READY);
390 static SgenPointerQueue critical_fin_queue = SGEN_POINTER_QUEUE_INIT (INTERNAL_MEM_FINALIZE_READY);
392 /* registered roots: the key to the hash is the root start address */
394 * Different kinds of roots are kept separate to speed up pin_from_roots () for example.
396 SgenHashTable roots_hash [ROOT_TYPE_NUM] = {
397 SGEN_HASH_TABLE_INIT (INTERNAL_MEM_ROOTS_TABLE, INTERNAL_MEM_ROOT_RECORD, sizeof (RootRecord), sgen_aligned_addr_hash, NULL),
398 SGEN_HASH_TABLE_INIT (INTERNAL_MEM_ROOTS_TABLE, INTERNAL_MEM_ROOT_RECORD, sizeof (RootRecord), sgen_aligned_addr_hash, NULL),
399 SGEN_HASH_TABLE_INIT (INTERNAL_MEM_ROOTS_TABLE, INTERNAL_MEM_ROOT_RECORD, sizeof (RootRecord), sgen_aligned_addr_hash, NULL)
401 static mword roots_size = 0; /* amount of memory in the root set */
403 /* The size of a TLAB */
404 /* The bigger the value, the less often we have to go to the slow path to allocate a new
405 * one, but the more space is wasted by threads not allocating much memory.
407 * FIXME: Make this self-tuning for each thread.
409 guint32 tlab_size = (1024 * 4);
411 #define MAX_SMALL_OBJ_SIZE SGEN_MAX_SMALL_OBJ_SIZE
413 #define ALLOC_ALIGN SGEN_ALLOC_ALIGN
415 #define ALIGN_UP SGEN_ALIGN_UP
417 #ifdef SGEN_DEBUG_INTERNAL_ALLOC
418 MonoNativeThreadId main_gc_thread = NULL;
421 /*Object was pinned during the current collection*/
422 static mword objects_pinned;
425 * ######################################################################
426 * ######## Macros and function declarations.
427 * ######################################################################
430 /* forward declarations */
431 static void scan_from_registered_roots (char *addr_start, char *addr_end, int root_type, ScanCopyContext ctx);
433 static void pin_from_roots (void *start_nursery, void *end_nursery, ScanCopyContext ctx);
434 static void finish_gray_stack (int generation, ScanCopyContext ctx);
437 SgenMajorCollector major_collector;
438 SgenMinorCollector sgen_minor_collector;
440 static SgenRememberedSet remset;
443 * The gray queue a worker job must use. If we're not parallel or
444 * concurrent, we use the main gray queue.
446 static SgenGrayQueue*
447 sgen_workers_get_job_gray_queue (WorkerData *worker_data, SgenGrayQueue *default_gray_queue)
450 return &worker_data->private_gray_queue;
451 SGEN_ASSERT (0, default_gray_queue, "Why don't we have a default gray queue when we're not running in a worker thread?");
452 return default_gray_queue;
456 gray_queue_redirect (SgenGrayQueue *queue)
458 sgen_workers_take_from_queue (current_collection_generation, queue);
462 sgen_scan_area_with_callback (char *start, char *end, IterateObjectCallbackFunc callback, void *data, gboolean allow_flags, gboolean fail_on_canaries)
464 while (start < end) {
468 if (!*(void**)start) {
469 start += sizeof (void*); /* should be ALLOC_ALIGN, really */
474 if (!(obj = (char *)SGEN_OBJECT_IS_FORWARDED (start)))
480 if (!sgen_client_object_is_array_fill ((GCObject*)obj)) {
481 CHECK_CANARY_FOR_OBJECT ((GCObject*)obj, fail_on_canaries);
482 size = ALIGN_UP (safe_object_get_size ((GCObject*)obj));
483 callback ((GCObject*)obj, size, data);
484 CANARIFY_SIZE (size);
486 size = ALIGN_UP (safe_object_get_size ((GCObject*)obj));
494 * sgen_add_to_global_remset:
496 * The global remset contains locations which point into newspace after
497 * a minor collection. This can happen if the objects they point to are pinned.
499 * LOCKING: If called from a parallel collector, the global remset
500 * lock must be held. For serial collectors that is not necessary.
503 sgen_add_to_global_remset (gpointer ptr, GCObject *obj)
505 SGEN_ASSERT (5, sgen_ptr_in_nursery (obj), "Target pointer of global remset must be in the nursery");
507 HEAVY_STAT (++stat_wbarrier_add_to_global_remset);
509 if (!major_collector.is_concurrent) {
510 SGEN_ASSERT (5, current_collection_generation != -1, "Global remsets can only be added during collections");
512 if (current_collection_generation == -1)
513 SGEN_ASSERT (5, sgen_concurrent_collection_in_progress (), "Global remsets outside of collection pauses can only be added by the concurrent collector");
516 if (!object_is_pinned (obj))
517 SGEN_ASSERT (5, sgen_minor_collector.is_split || sgen_concurrent_collection_in_progress (), "Non-pinned objects can only remain in nursery if it is a split nursery");
518 else if (sgen_cement_lookup_or_register (obj))
521 remset.record_pointer (ptr);
523 sgen_pin_stats_register_global_remset (obj);
525 SGEN_LOG (8, "Adding global remset for %p", ptr);
526 binary_protocol_global_remset (ptr, obj, (gpointer)SGEN_LOAD_VTABLE (obj));
530 * sgen_drain_gray_stack:
532 * Scan objects in the gray stack until the stack is empty. This should be called
533 * frequently after each object is copied, to achieve better locality and cache
538 sgen_drain_gray_stack (ScanCopyContext ctx)
540 SGEN_ASSERT (0, ctx.ops->drain_gray_stack, "Why do we have a scan/copy context with a missing drain gray stack function?");
542 return ctx.ops->drain_gray_stack (ctx.queue);
546 * Addresses in the pin queue are already sorted. This function finds
547 * the object header for each address and pins the object. The
548 * addresses must be inside the nursery section. The (start of the)
549 * address array is overwritten with the addresses of the actually
550 * pinned objects. Return the number of pinned objects.
553 pin_objects_from_nursery_pin_queue (gboolean do_scan_objects, ScanCopyContext ctx)
555 GCMemSection *section = nursery_section;
556 void **start = sgen_pinning_get_entry (section->pin_queue_first_entry);
557 void **end = sgen_pinning_get_entry (section->pin_queue_last_entry);
558 void *start_nursery = section->data;
559 void *end_nursery = section->end_data;
564 void *pinning_front = start_nursery;
566 void **definitely_pinned = start;
567 ScanObjectFunc scan_func = ctx.ops->scan_object;
568 SgenGrayQueue *queue = ctx.queue;
570 sgen_nursery_allocator_prepare_for_pinning ();
572 while (start < end) {
573 GCObject *obj_to_pin = NULL;
574 size_t obj_to_pin_size = 0;
579 SGEN_ASSERT (0, addr >= start_nursery && addr < end_nursery, "Potential pinning address out of range");
580 SGEN_ASSERT (0, addr >= last, "Pin queue not sorted");
587 SGEN_LOG (5, "Considering pinning addr %p", addr);
588 /* We've already processed everything up to pinning_front. */
589 if (addr < pinning_front) {
595 * Find the closest scan start <= addr. We might search backward in the
596 * scan_starts array because entries might be NULL. In the worst case we
597 * start at start_nursery.
599 idx = ((char*)addr - (char*)section->data) / SCAN_START_SIZE;
600 SGEN_ASSERT (0, idx < section->num_scan_start, "Scan start index out of range");
601 search_start = (void*)section->scan_starts [idx];
602 if (!search_start || search_start > addr) {
605 search_start = section->scan_starts [idx];
606 if (search_start && search_start <= addr)
609 if (!search_start || search_start > addr)
610 search_start = start_nursery;
614 * If the pinning front is closer than the scan start we found, start
615 * searching at the front.
617 if (search_start < pinning_front)
618 search_start = pinning_front;
621 * Now addr should be in an object a short distance from search_start.
623 * search_start must point to zeroed mem or point to an object.
626 size_t obj_size, canarified_obj_size;
629 if (!*(void**)search_start) {
630 search_start = (void*)ALIGN_UP ((mword)search_start + sizeof (gpointer));
631 /* The loop condition makes sure we don't overrun addr. */
635 canarified_obj_size = obj_size = ALIGN_UP (safe_object_get_size ((GCObject*)search_start));
638 * Filler arrays are marked by an invalid sync word. We don't
639 * consider them for pinning. They are not delimited by canaries,
642 if (!sgen_client_object_is_array_fill ((GCObject*)search_start)) {
643 CHECK_CANARY_FOR_OBJECT (search_start, TRUE);
644 CANARIFY_SIZE (canarified_obj_size);
646 if (addr >= search_start && (char*)addr < (char*)search_start + obj_size) {
647 /* This is the object we're looking for. */
648 obj_to_pin = (GCObject*)search_start;
649 obj_to_pin_size = canarified_obj_size;
654 /* Skip to the next object */
655 search_start = (void*)((char*)search_start + canarified_obj_size);
656 } while (search_start <= addr);
658 /* We've searched past the address we were looking for. */
660 pinning_front = search_start;
661 goto next_pin_queue_entry;
665 * We've found an object to pin. It might still be a dummy array, but we
666 * can advance the pinning front in any case.
668 pinning_front = (char*)obj_to_pin + obj_to_pin_size;
671 * If this is a dummy array marking the beginning of a nursery
672 * fragment, we don't pin it.
674 if (sgen_client_object_is_array_fill (obj_to_pin))
675 goto next_pin_queue_entry;
678 * Finally - pin the object!
680 desc = sgen_obj_get_descriptor_safe (obj_to_pin);
681 if (do_scan_objects) {
682 scan_func (obj_to_pin, desc, queue);
684 SGEN_LOG (4, "Pinned object %p, vtable %p (%s), count %d\n",
685 obj_to_pin, *(void**)obj_to_pin, sgen_client_vtable_get_name (SGEN_LOAD_VTABLE (obj_to_pin)), count);
686 binary_protocol_pin (obj_to_pin,
687 (gpointer)LOAD_VTABLE (obj_to_pin),
688 safe_object_get_size (obj_to_pin));
690 pin_object (obj_to_pin);
691 GRAY_OBJECT_ENQUEUE_SERIAL (queue, obj_to_pin, desc);
692 sgen_pin_stats_register_object (obj_to_pin, GENERATION_NURSERY);
693 definitely_pinned [count] = obj_to_pin;
696 if (concurrent_collection_in_progress)
697 sgen_pinning_register_pinned_in_nursery (obj_to_pin);
699 next_pin_queue_entry:
703 sgen_client_nursery_objects_pinned (definitely_pinned, count);
704 stat_pinned_objects += count;
709 pin_objects_in_nursery (gboolean do_scan_objects, ScanCopyContext ctx)
713 if (nursery_section->pin_queue_first_entry == nursery_section->pin_queue_last_entry)
716 reduced_to = pin_objects_from_nursery_pin_queue (do_scan_objects, ctx);
717 nursery_section->pin_queue_last_entry = nursery_section->pin_queue_first_entry + reduced_to;
721 * This function is only ever called (via `collector_pin_object()` in `sgen-copy-object.h`)
722 * when we can't promote an object because we're out of memory.
725 sgen_pin_object (GCObject *object, SgenGrayQueue *queue)
727 SGEN_ASSERT (0, sgen_ptr_in_nursery (object), "We're only supposed to use this for pinning nursery objects when out of memory.");
730 * All pinned objects are assumed to have been staged, so we need to stage as well.
731 * Also, the count of staged objects shows that "late pinning" happened.
733 sgen_pin_stage_ptr (object);
735 SGEN_PIN_OBJECT (object);
736 binary_protocol_pin (object, (gpointer)LOAD_VTABLE (object), safe_object_get_size (object));
739 sgen_pin_stats_register_object (object, GENERATION_NURSERY);
741 GRAY_OBJECT_ENQUEUE_SERIAL (queue, object, sgen_obj_get_descriptor_safe (object));
744 /* Sort the addresses in array in increasing order.
745 * Done using a by-the book heap sort. Which has decent and stable performance, is pretty cache efficient.
748 sgen_sort_addresses (void **array, size_t size)
753 for (i = 1; i < size; ++i) {
756 size_t parent = (child - 1) / 2;
758 if (array [parent] >= array [child])
761 tmp = array [parent];
762 array [parent] = array [child];
769 for (i = size - 1; i > 0; --i) {
772 array [i] = array [0];
778 while (root * 2 + 1 <= end) {
779 size_t child = root * 2 + 1;
781 if (child < end && array [child] < array [child + 1])
783 if (array [root] >= array [child])
787 array [root] = array [child];
796 * Scan the memory between start and end and queue values which could be pointers
797 * to the area between start_nursery and end_nursery for later consideration.
798 * Typically used for thread stacks.
801 sgen_conservatively_pin_objects_from (void **start, void **end, void *start_nursery, void *end_nursery, int pin_type)
805 SGEN_ASSERT (0, ((mword)start & (SIZEOF_VOID_P - 1)) == 0, "Why are we scanning for references in unaligned memory ?");
807 #if defined(VALGRIND_MAKE_MEM_DEFINED_IF_ADDRESSABLE) && !defined(_WIN64)
808 VALGRIND_MAKE_MEM_DEFINED_IF_ADDRESSABLE (start, (char*)end - (char*)start);
811 while (start < end) {
813 * *start can point to the middle of an object
814 * note: should we handle pointing at the end of an object?
815 * pinning in C# code disallows pointing at the end of an object
816 * but there is some small chance that an optimizing C compiler
817 * may keep the only reference to an object by pointing
818 * at the end of it. We ignore this small chance for now.
819 * Pointers to the end of an object are indistinguishable
820 * from pointers to the start of the next object in memory
821 * so if we allow that we'd need to pin two objects...
822 * We queue the pointer in an array, the
823 * array will then be sorted and uniqued. This way
824 * we can coalesce several pinning pointers and it should
825 * be faster since we'd do a memory scan with increasing
826 * addresses. Note: we can align the address to the allocation
827 * alignment, so the unique process is more effective.
829 mword addr = (mword)*start;
830 addr &= ~(ALLOC_ALIGN - 1);
831 if (addr >= (mword)start_nursery && addr < (mword)end_nursery) {
832 SGEN_LOG (6, "Pinning address %p from %p", (void*)addr, start);
833 sgen_pin_stage_ptr ((void*)addr);
834 binary_protocol_pin_stage (start, (void*)addr);
835 sgen_pin_stats_register_address ((char*)addr, pin_type);
841 SGEN_LOG (7, "found %d potential pinned heap pointers", count);
845 * The first thing we do in a collection is to identify pinned objects.
846 * This function considers all the areas of memory that need to be
847 * conservatively scanned.
850 pin_from_roots (void *start_nursery, void *end_nursery, ScanCopyContext ctx)
854 SGEN_LOG (2, "Scanning pinned roots (%d bytes, %d/%d entries)", (int)roots_size, roots_hash [ROOT_TYPE_NORMAL].num_entries, roots_hash [ROOT_TYPE_PINNED].num_entries);
855 /* objects pinned from the API are inside these roots */
856 SGEN_HASH_TABLE_FOREACH (&roots_hash [ROOT_TYPE_PINNED], void **, start_root, RootRecord *, root) {
857 SGEN_LOG (6, "Pinned roots %p-%p", start_root, root->end_root);
858 sgen_conservatively_pin_objects_from (start_root, (void**)root->end_root, start_nursery, end_nursery, PIN_TYPE_OTHER);
859 } SGEN_HASH_TABLE_FOREACH_END;
860 /* now deal with the thread stacks
861 * in the future we should be able to conservatively scan only:
862 * *) the cpu registers
863 * *) the unmanaged stack frames
864 * *) the _last_ managed stack frame
865 * *) pointers slots in managed frames
867 sgen_client_scan_thread_data (start_nursery, end_nursery, FALSE, ctx);
871 single_arg_user_copy_or_mark (GCObject **obj, void *gc_data)
873 ScanCopyContext *ctx = (ScanCopyContext *)gc_data;
874 ctx->ops->copy_or_mark_object (obj, ctx->queue);
878 * The memory area from start_root to end_root contains pointers to objects.
879 * Their position is precisely described by @desc (this means that the pointer
880 * can be either NULL or the pointer to the start of an object).
881 * This functions copies them to to_space updates them.
883 * This function is not thread-safe!
886 precisely_scan_objects_from (void** start_root, void** end_root, char* n_start, char *n_end, SgenDescriptor desc, ScanCopyContext ctx)
888 CopyOrMarkObjectFunc copy_func = ctx.ops->copy_or_mark_object;
889 ScanPtrFieldFunc scan_field_func = ctx.ops->scan_ptr_field;
890 SgenGrayQueue *queue = ctx.queue;
892 switch (desc & ROOT_DESC_TYPE_MASK) {
893 case ROOT_DESC_BITMAP:
894 desc >>= ROOT_DESC_TYPE_SHIFT;
896 if ((desc & 1) && *start_root) {
897 copy_func ((GCObject**)start_root, queue);
898 SGEN_LOG (9, "Overwrote root at %p with %p", start_root, *start_root);
904 case ROOT_DESC_COMPLEX: {
905 gsize *bitmap_data = (gsize *)sgen_get_complex_descriptor_bitmap (desc);
906 gsize bwords = (*bitmap_data) - 1;
907 void **start_run = start_root;
909 while (bwords-- > 0) {
910 gsize bmap = *bitmap_data++;
911 void **objptr = start_run;
913 if ((bmap & 1) && *objptr) {
914 copy_func ((GCObject**)objptr, queue);
915 SGEN_LOG (9, "Overwrote root at %p with %p", objptr, *objptr);
920 start_run += GC_BITS_PER_WORD;
924 case ROOT_DESC_VECTOR: {
927 for (p = start_root; p < end_root; p++) {
929 scan_field_func (NULL, (GCObject**)p, queue);
933 case ROOT_DESC_USER: {
934 SgenUserRootMarkFunc marker = sgen_get_user_descriptor_func (desc);
935 marker (start_root, single_arg_user_copy_or_mark, &ctx);
938 case ROOT_DESC_RUN_LEN:
939 g_assert_not_reached ();
941 g_assert_not_reached ();
946 reset_heap_boundaries (void)
948 lowest_heap_address = ~(mword)0;
949 highest_heap_address = 0;
953 sgen_update_heap_boundaries (mword low, mword high)
958 old = lowest_heap_address;
961 } while (SGEN_CAS_PTR ((gpointer*)&lowest_heap_address, (gpointer)low, (gpointer)old) != (gpointer)old);
964 old = highest_heap_address;
967 } while (SGEN_CAS_PTR ((gpointer*)&highest_heap_address, (gpointer)high, (gpointer)old) != (gpointer)old);
971 * Allocate and setup the data structures needed to be able to allocate objects
972 * in the nursery. The nursery is stored in nursery_section.
975 alloc_nursery (gboolean dynamic, size_t min_size, size_t max_size)
982 min_size = SGEN_DEFAULT_NURSERY_MIN_SIZE;
984 max_size = SGEN_DEFAULT_NURSERY_MAX_SIZE;
986 SGEN_ASSERT (0, min_size == max_size, "We can't have nursery ranges for static configuration.");
988 min_size = max_size = SGEN_DEFAULT_NURSERY_SIZE;
991 SGEN_ASSERT (0, !nursery_section, "Why are we allocating the nursery twice?");
992 SGEN_LOG (2, "Allocating nursery size: %zu, initial %zu", max_size, min_size);
994 /* FIXME: handle OOM */
995 nursery_section = (GCMemSection *)sgen_alloc_internal (INTERNAL_MEM_SECTION);
997 /* If there isn't enough space even for the nursery we should simply abort. */
998 g_assert (sgen_memgov_try_alloc_space (max_size, SPACE_NURSERY));
1001 * The nursery section range represents the memory section where objects
1002 * can be found. This is used when iterating for objects in the nursery,
1003 * pinning etc. sgen_nursery_max_size represents the total allocated space
1004 * for the nursery. sgen_nursery_size represents the current size of the
1005 * nursery and it is used for allocation limits, heuristics etc. The
1006 * nursery section is not always identical to the current nursery size
1007 * because it can contain pinned objects from when the nursery was larger.
1009 * sgen_nursery_size <= nursery_section size <= sgen_nursery_max_size
1011 data = (char *)major_collector.alloc_heap (max_size, max_size);
1012 sgen_update_heap_boundaries ((mword)data, (mword)(data + max_size));
1013 nursery_section->data = data;
1014 nursery_section->end_data = data + min_size;
1015 scan_starts = (max_size + SCAN_START_SIZE - 1) / SCAN_START_SIZE;
1016 nursery_section->scan_starts = (char **)sgen_alloc_internal_dynamic (sizeof (char*) * scan_starts, INTERNAL_MEM_SCAN_STARTS, TRUE);
1017 nursery_section->num_scan_start = scan_starts;
1019 sgen_nursery_allocator_set_nursery_bounds (data, min_size, max_size);
1023 mono_gc_get_logfile (void)
1025 return gc_debug_file;
1029 mono_gc_params_set (const char* options)
1031 if (gc_params_options)
1032 g_free (gc_params_options);
1034 gc_params_options = g_strdup (options);
1038 mono_gc_debug_set (const char* options)
1040 if (gc_debug_options)
1041 g_free (gc_debug_options);
1043 gc_debug_options = g_strdup (options);
1047 scan_finalizer_entries (SgenPointerQueue *fin_queue, ScanCopyContext ctx)
1049 CopyOrMarkObjectFunc copy_func = ctx.ops->copy_or_mark_object;
1050 SgenGrayQueue *queue = ctx.queue;
1053 for (i = 0; i < fin_queue->next_slot; ++i) {
1054 GCObject *obj = (GCObject *)fin_queue->data [i];
1057 SGEN_LOG (5, "Scan of fin ready object: %p (%s)\n", obj, sgen_client_vtable_get_name (SGEN_LOAD_VTABLE (obj)));
1058 copy_func ((GCObject**)&fin_queue->data [i], queue);
1063 generation_name (int generation)
1065 switch (generation) {
1066 case GENERATION_NURSERY: return "nursery";
1067 case GENERATION_OLD: return "old";
1068 default: g_assert_not_reached ();
1073 sgen_generation_name (int generation)
1075 return generation_name (generation);
1079 finish_gray_stack (int generation, ScanCopyContext ctx)
1083 int done_with_ephemerons, ephemeron_rounds = 0;
1084 char *start_addr = generation == GENERATION_NURSERY ? sgen_get_nursery_start () : NULL;
1085 char *end_addr = generation == GENERATION_NURSERY ? sgen_get_nursery_end () : (char*)-1;
1086 SgenGrayQueue *queue = ctx.queue;
1088 binary_protocol_finish_gray_stack_start (sgen_timestamp (), generation);
1090 * We copied all the reachable objects. Now it's the time to copy
1091 * the objects that were not referenced by the roots, but by the copied objects.
1092 * we built a stack of objects pointed to by gray_start: they are
1093 * additional roots and we may add more items as we go.
1094 * We loop until gray_start == gray_objects which means no more objects have
1095 * been added. Note this is iterative: no recursion is involved.
1096 * We need to walk the LO list as well in search of marked big objects
1097 * (use a flag since this is needed only on major collections). We need to loop
1098 * here as well, so keep a counter of marked LO (increasing it in copy_object).
1099 * To achieve better cache locality and cache usage, we drain the gray stack
1100 * frequently, after each object is copied, and just finish the work here.
1102 sgen_drain_gray_stack (ctx);
1104 SGEN_LOG (2, "%s generation done", generation_name (generation));
1107 Reset bridge data, we might have lingering data from a previous collection if this is a major
1108 collection trigged by minor overflow.
1110 We must reset the gathered bridges since their original block might be evacuated due to major
1111 fragmentation in the meanwhile and the bridge code should not have to deal with that.
1113 if (sgen_client_bridge_need_processing ())
1114 sgen_client_bridge_reset_data ();
1117 * Mark all strong toggleref objects. This must be done before we walk ephemerons or finalizers
1118 * to ensure they see the full set of live objects.
1120 sgen_client_mark_togglerefs (start_addr, end_addr, ctx);
1123 * Walk the ephemeron tables marking all values with reachable keys. This must be completely done
1124 * before processing finalizable objects and non-tracking weak links to avoid finalizing/clearing
1125 * objects that are in fact reachable.
1127 done_with_ephemerons = 0;
1129 done_with_ephemerons = sgen_client_mark_ephemerons (ctx);
1130 sgen_drain_gray_stack (ctx);
1132 } while (!done_with_ephemerons);
1134 if (sgen_client_bridge_need_processing ()) {
1135 /*Make sure the gray stack is empty before we process bridge objects so we get liveness right*/
1136 sgen_drain_gray_stack (ctx);
1137 sgen_collect_bridge_objects (generation, ctx);
1138 if (generation == GENERATION_OLD)
1139 sgen_collect_bridge_objects (GENERATION_NURSERY, ctx);
1142 Do the first bridge step here, as the collector liveness state will become useless after that.
1144 An important optimization is to only proccess the possibly dead part of the object graph and skip
1145 over all live objects as we transitively know everything they point must be alive too.
1147 The above invariant is completely wrong if we let the gray queue be drained and mark/copy everything.
1149 This has the unfortunate side effect of making overflow collections perform the first step twice, but
1150 given we now have heuristics that perform major GC in anticipation of minor overflows this should not
1153 sgen_client_bridge_processing_stw_step ();
1157 Make sure we drain the gray stack before processing disappearing links and finalizers.
1158 If we don't make sure it is empty we might wrongly see a live object as dead.
1160 sgen_drain_gray_stack (ctx);
1163 We must clear weak links that don't track resurrection before processing object ready for
1164 finalization so they can be cleared before that.
1166 sgen_null_link_in_range (generation, ctx, FALSE);
1167 if (generation == GENERATION_OLD)
1168 sgen_null_link_in_range (GENERATION_NURSERY, ctx, FALSE);
1171 /* walk the finalization queue and move also the objects that need to be
1172 * finalized: use the finalized objects as new roots so the objects they depend
1173 * on are also not reclaimed. As with the roots above, only objects in the nursery
1174 * are marked/copied.
1176 sgen_finalize_in_range (generation, ctx);
1177 if (generation == GENERATION_OLD)
1178 sgen_finalize_in_range (GENERATION_NURSERY, ctx);
1179 /* drain the new stack that might have been created */
1180 SGEN_LOG (6, "Precise scan of gray area post fin");
1181 sgen_drain_gray_stack (ctx);
1184 * This must be done again after processing finalizable objects since CWL slots are cleared only after the key is finalized.
1186 done_with_ephemerons = 0;
1188 done_with_ephemerons = sgen_client_mark_ephemerons (ctx);
1189 sgen_drain_gray_stack (ctx);
1191 } while (!done_with_ephemerons);
1193 sgen_client_clear_unreachable_ephemerons (ctx);
1196 * We clear togglerefs only after all possible chances of revival are done.
1197 * This is semantically more inline with what users expect and it allows for
1198 * user finalizers to correctly interact with TR objects.
1200 sgen_client_clear_togglerefs (start_addr, end_addr, ctx);
1203 SGEN_LOG (2, "Finalize queue handling scan for %s generation: %lld usecs %d ephemeron rounds", generation_name (generation), (long long)TV_ELAPSED (atv, btv), ephemeron_rounds);
1206 * handle disappearing links
1207 * Note we do this after checking the finalization queue because if an object
1208 * survives (at least long enough to be finalized) we don't clear the link.
1209 * This also deals with a possible issue with the monitor reclamation: with the Boehm
1210 * GC a finalized object my lose the monitor because it is cleared before the finalizer is
1213 g_assert (sgen_gray_object_queue_is_empty (queue));
1215 sgen_null_link_in_range (generation, ctx, TRUE);
1216 if (generation == GENERATION_OLD)
1217 sgen_null_link_in_range (GENERATION_NURSERY, ctx, TRUE);
1218 if (sgen_gray_object_queue_is_empty (queue))
1220 sgen_drain_gray_stack (ctx);
1223 g_assert (sgen_gray_object_queue_is_empty (queue));
1225 binary_protocol_finish_gray_stack_end (sgen_timestamp (), generation);
1229 sgen_check_section_scan_starts (GCMemSection *section)
1232 for (i = 0; i < section->num_scan_start; ++i) {
1233 if (section->scan_starts [i]) {
1234 mword size = safe_object_get_size ((GCObject*) section->scan_starts [i]);
1235 SGEN_ASSERT (0, size >= SGEN_CLIENT_MINIMUM_OBJECT_SIZE && size <= MAX_SMALL_OBJ_SIZE, "Weird object size at scan starts.");
1241 check_scan_starts (void)
1243 if (!do_scan_starts_check)
1245 sgen_check_section_scan_starts (nursery_section);
1246 major_collector.check_scan_starts ();
1250 scan_from_registered_roots (char *addr_start, char *addr_end, int root_type, ScanCopyContext ctx)
1254 SGEN_HASH_TABLE_FOREACH (&roots_hash [root_type], void **, start_root, RootRecord *, root) {
1255 SGEN_LOG (6, "Precise root scan %p-%p (desc: %p)", start_root, root->end_root, (void*)root->root_desc);
1256 precisely_scan_objects_from (start_root, (void**)root->end_root, addr_start, addr_end, root->root_desc, ctx);
1257 } SGEN_HASH_TABLE_FOREACH_END;
1263 static gboolean inited = FALSE;
1268 mono_counters_register ("Collection max time", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME | MONO_COUNTER_MONOTONIC, &time_max);
1270 mono_counters_register ("Minor fragment clear", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_minor_pre_collection_fragment_clear);
1271 mono_counters_register ("Minor pinning", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_minor_pinning);
1272 mono_counters_register ("Minor scan remembered set", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_minor_scan_remsets);
1273 mono_counters_register ("Minor scan major blocks", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_minor_scan_major_blocks);
1274 mono_counters_register ("Minor scan los", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_minor_scan_los);
1275 mono_counters_register ("Minor scan pinned", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_minor_scan_pinned);
1276 mono_counters_register ("Minor scan roots", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_minor_scan_roots);
1277 mono_counters_register ("Minor fragment creation", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_minor_fragment_creation);
1279 mono_counters_register ("Major fragment clear", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_major_pre_collection_fragment_clear);
1280 mono_counters_register ("Major pinning", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_major_pinning);
1281 mono_counters_register ("Major scan pinned", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_major_scan_pinned);
1282 mono_counters_register ("Major scan roots", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_major_scan_roots);
1283 mono_counters_register ("Major scan mod union blocks", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_major_scan_mod_union_blocks);
1284 mono_counters_register ("Major scan mod union los", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_major_scan_mod_union_los);
1285 mono_counters_register ("Major finish gray stack", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_major_finish_gray_stack);
1286 mono_counters_register ("Major free big objects", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_major_free_bigobjs);
1287 mono_counters_register ("Major LOS sweep", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_major_los_sweep);
1288 mono_counters_register ("Major sweep", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_major_sweep);
1289 mono_counters_register ("Major fragment creation", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_major_fragment_creation);
1291 mono_counters_register ("Number of pinned objects", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_pinned_objects);
1293 #ifdef HEAVY_STATISTICS
1294 mono_counters_register ("WBarrier remember pointer", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_wbarrier_add_to_global_remset);
1295 mono_counters_register ("WBarrier arrayref copy", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_wbarrier_arrayref_copy);
1296 mono_counters_register ("WBarrier generic store called", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_wbarrier_generic_store);
1297 mono_counters_register ("WBarrier generic atomic store called", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_wbarrier_generic_store_atomic);
1298 mono_counters_register ("WBarrier set root", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_wbarrier_set_root);
1300 mono_counters_register ("# objects allocated degraded", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_objects_alloced_degraded);
1301 mono_counters_register ("bytes allocated degraded", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_bytes_alloced_degraded);
1303 mono_counters_register ("# copy_object() called (nursery)", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_copy_object_called_nursery);
1304 mono_counters_register ("# objects copied (nursery)", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_objects_copied_nursery);
1305 mono_counters_register ("# copy_object() called (major)", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_copy_object_called_major);
1306 mono_counters_register ("# objects copied (major)", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_objects_copied_major);
1308 mono_counters_register ("# scan_object() called (nursery)", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_scan_object_called_nursery);
1309 mono_counters_register ("# scan_object() called (major)", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_scan_object_called_major);
1311 mono_counters_register ("Slots allocated in vain", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_slots_allocated_in_vain);
1313 mono_counters_register ("# nursery copy_object() failed from space", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_nursery_copy_object_failed_from_space);
1314 mono_counters_register ("# nursery copy_object() failed forwarded", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_nursery_copy_object_failed_forwarded);
1315 mono_counters_register ("# nursery copy_object() failed pinned", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_nursery_copy_object_failed_pinned);
1316 mono_counters_register ("# nursery copy_object() failed to space", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_nursery_copy_object_failed_to_space);
1318 sgen_nursery_allocator_init_heavy_stats ();
1326 reset_pinned_from_failed_allocation (void)
1328 bytes_pinned_from_failed_allocation = 0;
1332 sgen_set_pinned_from_failed_allocation (mword objsize)
1334 bytes_pinned_from_failed_allocation += objsize;
1338 sgen_collection_is_concurrent (void)
1340 switch (current_collection_generation) {
1341 case GENERATION_NURSERY:
1343 case GENERATION_OLD:
1344 return concurrent_collection_in_progress;
1346 g_error ("Invalid current generation %d", current_collection_generation);
1352 sgen_concurrent_collection_in_progress (void)
1354 return concurrent_collection_in_progress;
1358 SgenThreadPoolJob job;
1359 SgenObjectOperations *ops;
1360 SgenGrayQueue *gc_thread_gray_queue;
1365 int job_index, job_split_count;
1369 static ScanCopyContext
1370 scan_copy_context_for_scan_job (void *worker_data_untyped, ScanJob *job)
1372 WorkerData *worker_data = (WorkerData *)worker_data_untyped;
1376 * For jobs enqueued on workers we set the ops at job runtime in order
1377 * to be able to profit from on the fly optimized object ops or other
1378 * object ops changes, like forced concurrent finish.
1380 SGEN_ASSERT (0, sgen_workers_is_worker_thread (mono_native_thread_id_get ()), "We need a context for the scan job");
1381 job->ops = sgen_workers_get_idle_func_object_ops (worker_data);
1384 return CONTEXT_FROM_OBJECT_OPERATIONS (job->ops, sgen_workers_get_job_gray_queue (worker_data, job->gc_thread_gray_queue));
1392 } ScanFromRegisteredRootsJob;
1395 job_scan_from_registered_roots (void *worker_data_untyped, SgenThreadPoolJob *job)
1397 ScanFromRegisteredRootsJob *job_data = (ScanFromRegisteredRootsJob*)job;
1398 ScanCopyContext ctx = scan_copy_context_for_scan_job (worker_data_untyped, &job_data->scan_job);
1400 scan_from_registered_roots (job_data->heap_start, job_data->heap_end, job_data->root_type, ctx);
1407 } ScanThreadDataJob;
1410 job_scan_thread_data (void *worker_data_untyped, SgenThreadPoolJob *job)
1412 ScanThreadDataJob *job_data = (ScanThreadDataJob*)job;
1413 ScanCopyContext ctx = scan_copy_context_for_scan_job (worker_data_untyped, &job_data->scan_job);
1415 sgen_client_scan_thread_data (job_data->heap_start, job_data->heap_end, TRUE, ctx);
1420 SgenPointerQueue *queue;
1421 } ScanFinalizerEntriesJob;
1424 job_scan_finalizer_entries (void *worker_data_untyped, SgenThreadPoolJob *job)
1426 ScanFinalizerEntriesJob *job_data = (ScanFinalizerEntriesJob*)job;
1427 ScanCopyContext ctx = scan_copy_context_for_scan_job (worker_data_untyped, &job_data->scan_job);
1429 scan_finalizer_entries (job_data->queue, ctx);
1433 job_scan_wbroots (void *worker_data_untyped, SgenThreadPoolJob *job)
1435 ScanJob *job_data = (ScanJob*)job;
1436 ScanCopyContext ctx = scan_copy_context_for_scan_job (worker_data_untyped, job_data);
1438 sgen_wbroots_scan_card_table (ctx);
1442 job_scan_major_card_table (void *worker_data_untyped, SgenThreadPoolJob *job)
1444 SGEN_TV_DECLARE (atv);
1445 SGEN_TV_DECLARE (btv);
1446 ParallelScanJob *job_data = (ParallelScanJob*)job;
1447 ScanCopyContext ctx = scan_copy_context_for_scan_job (worker_data_untyped, (ScanJob*)job_data);
1449 SGEN_TV_GETTIME (atv);
1450 major_collector.scan_card_table (CARDTABLE_SCAN_GLOBAL, ctx, job_data->job_index, job_data->job_split_count, job_data->data);
1451 SGEN_TV_GETTIME (btv);
1452 time_minor_scan_major_blocks += SGEN_TV_ELAPSED (atv, btv);
1454 if (worker_data_untyped)
1455 ((WorkerData*)worker_data_untyped)->major_scan_time += SGEN_TV_ELAPSED (atv, btv);
1459 job_scan_los_card_table (void *worker_data_untyped, SgenThreadPoolJob *job)
1461 SGEN_TV_DECLARE (atv);
1462 SGEN_TV_DECLARE (btv);
1463 ParallelScanJob *job_data = (ParallelScanJob*)job;
1464 ScanCopyContext ctx = scan_copy_context_for_scan_job (worker_data_untyped, (ScanJob*)job_data);
1466 SGEN_TV_GETTIME (atv);
1467 sgen_los_scan_card_table (CARDTABLE_SCAN_GLOBAL, ctx, job_data->job_index, job_data->job_split_count);
1468 SGEN_TV_GETTIME (btv);
1469 time_minor_scan_los += SGEN_TV_ELAPSED (atv, btv);
1471 if (worker_data_untyped)
1472 ((WorkerData*)worker_data_untyped)->los_scan_time += SGEN_TV_ELAPSED (atv, btv);
1476 job_scan_major_mod_union_card_table (void *worker_data_untyped, SgenThreadPoolJob *job)
1478 SGEN_TV_DECLARE (atv);
1479 SGEN_TV_DECLARE (btv);
1480 ParallelScanJob *job_data = (ParallelScanJob*)job;
1481 ScanCopyContext ctx = scan_copy_context_for_scan_job (worker_data_untyped, (ScanJob*)job_data);
1483 g_assert (concurrent_collection_in_progress);
1484 SGEN_TV_GETTIME (atv);
1485 major_collector.scan_card_table (CARDTABLE_SCAN_MOD_UNION, ctx, job_data->job_index, job_data->job_split_count, job_data->data);
1486 SGEN_TV_GETTIME (btv);
1487 time_major_scan_mod_union_blocks += SGEN_TV_ELAPSED (atv, btv);
1489 if (worker_data_untyped)
1490 ((WorkerData*)worker_data_untyped)->major_scan_time += SGEN_TV_ELAPSED (atv, btv);
1494 job_scan_los_mod_union_card_table (void *worker_data_untyped, SgenThreadPoolJob *job)
1496 SGEN_TV_DECLARE (atv);
1497 SGEN_TV_DECLARE (btv);
1498 ParallelScanJob *job_data = (ParallelScanJob*)job;
1499 ScanCopyContext ctx = scan_copy_context_for_scan_job (worker_data_untyped, (ScanJob*)job_data);
1501 g_assert (concurrent_collection_in_progress);
1502 SGEN_TV_GETTIME (atv);
1503 sgen_los_scan_card_table (CARDTABLE_SCAN_MOD_UNION, ctx, job_data->job_index, job_data->job_split_count);
1504 SGEN_TV_GETTIME (btv);
1505 time_major_scan_mod_union_los += SGEN_TV_ELAPSED (atv, btv);
1507 if (worker_data_untyped)
1508 ((WorkerData*)worker_data_untyped)->los_scan_time += SGEN_TV_ELAPSED (atv, btv);
1512 job_major_mod_union_preclean (void *worker_data_untyped, SgenThreadPoolJob *job)
1514 SGEN_TV_DECLARE (atv);
1515 SGEN_TV_DECLARE (btv);
1516 ParallelScanJob *job_data = (ParallelScanJob*)job;
1517 ScanCopyContext ctx = scan_copy_context_for_scan_job (worker_data_untyped, (ScanJob*)job_data);
1519 g_assert (concurrent_collection_in_progress);
1520 SGEN_TV_GETTIME (atv);
1521 major_collector.scan_card_table (CARDTABLE_SCAN_MOD_UNION_PRECLEAN, ctx, job_data->job_index, job_data->job_split_count, job_data->data);
1522 SGEN_TV_GETTIME (btv);
1524 g_assert (worker_data_untyped);
1525 ((WorkerData*)worker_data_untyped)->major_scan_time += SGEN_TV_ELAPSED (atv, btv);
1529 job_los_mod_union_preclean (void *worker_data_untyped, SgenThreadPoolJob *job)
1531 SGEN_TV_DECLARE (atv);
1532 SGEN_TV_DECLARE (btv);
1533 ParallelScanJob *job_data = (ParallelScanJob*)job;
1534 ScanCopyContext ctx = scan_copy_context_for_scan_job (worker_data_untyped, (ScanJob*)job_data);
1536 g_assert (concurrent_collection_in_progress);
1537 SGEN_TV_GETTIME (atv);
1538 sgen_los_scan_card_table (CARDTABLE_SCAN_MOD_UNION_PRECLEAN, ctx, job_data->job_index, job_data->job_split_count);
1539 SGEN_TV_GETTIME (btv);
1541 g_assert (worker_data_untyped);
1542 ((WorkerData*)worker_data_untyped)->los_scan_time += SGEN_TV_ELAPSED (atv, btv);
1546 job_scan_last_pinned (void *worker_data_untyped, SgenThreadPoolJob *job)
1548 ScanJob *job_data = (ScanJob*)job;
1549 ScanCopyContext ctx = scan_copy_context_for_scan_job (worker_data_untyped, job_data);
1551 g_assert (concurrent_collection_in_progress);
1553 sgen_scan_pin_queue_objects (ctx);
1557 workers_finish_callback (void)
1559 ParallelScanJob *psj;
1561 size_t num_major_sections = major_collector.get_num_major_sections ();
1562 int split_count = sgen_workers_get_job_split_count (GENERATION_OLD);
1564 /* Mod union preclean jobs */
1565 for (i = 0; i < split_count; i++) {
1566 psj = (ParallelScanJob*)sgen_thread_pool_job_alloc ("preclean major mod union cardtable", job_major_mod_union_preclean, sizeof (ParallelScanJob));
1567 psj->scan_job.gc_thread_gray_queue = NULL;
1569 psj->job_split_count = split_count;
1570 psj->data = num_major_sections / split_count;
1571 sgen_workers_enqueue_job (GENERATION_OLD, &psj->scan_job.job, TRUE);
1574 for (i = 0; i < split_count; i++) {
1575 psj = (ParallelScanJob*)sgen_thread_pool_job_alloc ("preclean los mod union cardtable", job_los_mod_union_preclean, sizeof (ParallelScanJob));
1576 psj->scan_job.gc_thread_gray_queue = NULL;
1578 psj->job_split_count = split_count;
1579 sgen_workers_enqueue_job (GENERATION_OLD, &psj->scan_job.job, TRUE);
1582 sj = (ScanJob*)sgen_thread_pool_job_alloc ("scan last pinned", job_scan_last_pinned, sizeof (ScanJob));
1583 sj->gc_thread_gray_queue = NULL;
1584 sgen_workers_enqueue_job (GENERATION_OLD, &sj->job, TRUE);
1588 init_gray_queue (SgenGrayQueue *gc_thread_gray_queue)
1590 sgen_gray_object_queue_init (gc_thread_gray_queue, NULL, TRUE);
1594 enqueue_scan_remembered_set_jobs (SgenGrayQueue *gc_thread_gray_queue, SgenObjectOperations *ops, gboolean enqueue)
1596 int i, split_count = sgen_workers_get_job_split_count (GENERATION_NURSERY);
1597 size_t num_major_sections = major_collector.get_num_major_sections ();
1600 sj = (ScanJob*)sgen_thread_pool_job_alloc ("scan wbroots", job_scan_wbroots, sizeof (ScanJob));
1602 sj->gc_thread_gray_queue = gc_thread_gray_queue;
1603 sgen_workers_enqueue_job (GENERATION_NURSERY, &sj->job, enqueue);
1605 for (i = 0; i < split_count; i++) {
1606 ParallelScanJob *psj;
1608 psj = (ParallelScanJob*)sgen_thread_pool_job_alloc ("scan major remsets", job_scan_major_card_table, sizeof (ParallelScanJob));
1609 psj->scan_job.ops = ops;
1610 psj->scan_job.gc_thread_gray_queue = gc_thread_gray_queue;
1612 psj->job_split_count = split_count;
1613 psj->data = num_major_sections / split_count;
1614 sgen_workers_enqueue_job (GENERATION_NURSERY, &psj->scan_job.job, enqueue);
1616 psj = (ParallelScanJob*)sgen_thread_pool_job_alloc ("scan LOS remsets", job_scan_los_card_table, sizeof (ParallelScanJob));
1617 psj->scan_job.ops = ops;
1618 psj->scan_job.gc_thread_gray_queue = gc_thread_gray_queue;
1620 psj->job_split_count = split_count;
1621 sgen_workers_enqueue_job (GENERATION_NURSERY, &psj->scan_job.job, enqueue);
1626 enqueue_scan_from_roots_jobs (SgenGrayQueue *gc_thread_gray_queue, char *heap_start, char *heap_end, SgenObjectOperations *ops, gboolean enqueue)
1628 ScanFromRegisteredRootsJob *scrrj;
1629 ScanThreadDataJob *stdj;
1630 ScanFinalizerEntriesJob *sfej;
1632 /* registered roots, this includes static fields */
1634 scrrj = (ScanFromRegisteredRootsJob*)sgen_thread_pool_job_alloc ("scan from registered roots normal", job_scan_from_registered_roots, sizeof (ScanFromRegisteredRootsJob));
1635 scrrj->scan_job.ops = ops;
1636 scrrj->scan_job.gc_thread_gray_queue = gc_thread_gray_queue;
1637 scrrj->heap_start = heap_start;
1638 scrrj->heap_end = heap_end;
1639 scrrj->root_type = ROOT_TYPE_NORMAL;
1640 sgen_workers_enqueue_job (current_collection_generation, &scrrj->scan_job.job, enqueue);
1642 if (current_collection_generation == GENERATION_OLD) {
1643 /* During minors we scan the cardtable for these roots instead */
1644 scrrj = (ScanFromRegisteredRootsJob*)sgen_thread_pool_job_alloc ("scan from registered roots wbarrier", job_scan_from_registered_roots, sizeof (ScanFromRegisteredRootsJob));
1645 scrrj->scan_job.ops = ops;
1646 scrrj->scan_job.gc_thread_gray_queue = gc_thread_gray_queue;
1647 scrrj->heap_start = heap_start;
1648 scrrj->heap_end = heap_end;
1649 scrrj->root_type = ROOT_TYPE_WBARRIER;
1650 sgen_workers_enqueue_job (current_collection_generation, &scrrj->scan_job.job, enqueue);
1655 stdj = (ScanThreadDataJob*)sgen_thread_pool_job_alloc ("scan thread data", job_scan_thread_data, sizeof (ScanThreadDataJob));
1656 stdj->scan_job.ops = ops;
1657 stdj->scan_job.gc_thread_gray_queue = gc_thread_gray_queue;
1658 stdj->heap_start = heap_start;
1659 stdj->heap_end = heap_end;
1660 sgen_workers_enqueue_job (current_collection_generation, &stdj->scan_job.job, enqueue);
1662 /* Scan the list of objects ready for finalization. */
1664 sfej = (ScanFinalizerEntriesJob*)sgen_thread_pool_job_alloc ("scan finalizer entries", job_scan_finalizer_entries, sizeof (ScanFinalizerEntriesJob));
1665 sfej->scan_job.ops = ops;
1666 sfej->scan_job.gc_thread_gray_queue = gc_thread_gray_queue;
1667 sfej->queue = &fin_ready_queue;
1668 sgen_workers_enqueue_job (current_collection_generation, &sfej->scan_job.job, enqueue);
1670 sfej = (ScanFinalizerEntriesJob*)sgen_thread_pool_job_alloc ("scan critical finalizer entries", job_scan_finalizer_entries, sizeof (ScanFinalizerEntriesJob));
1671 sfej->scan_job.ops = ops;
1672 sfej->scan_job.gc_thread_gray_queue = gc_thread_gray_queue;
1673 sfej->queue = &critical_fin_queue;
1674 sgen_workers_enqueue_job (current_collection_generation, &sfej->scan_job.job, enqueue);
1678 * Perform a nursery collection.
1680 * Return whether any objects were late-pinned due to being out of memory.
1683 collect_nursery (const char *reason, gboolean is_overflow, SgenGrayQueue *unpin_queue)
1685 gboolean needs_major, is_parallel = FALSE;
1686 mword fragment_total;
1687 SgenGrayQueue gc_thread_gray_queue;
1688 SgenObjectOperations *object_ops_nopar, *object_ops_par = NULL;
1689 ScanCopyContext ctx;
1692 SGEN_TV_DECLARE (last_minor_collection_start_tv);
1693 SGEN_TV_DECLARE (last_minor_collection_end_tv);
1694 guint64 major_scan_start = time_minor_scan_major_blocks;
1695 guint64 los_scan_start = time_minor_scan_los;
1696 guint64 finish_gray_start = time_minor_finish_gray_stack;
1698 if (disable_minor_collections)
1701 TV_GETTIME (last_minor_collection_start_tv);
1702 atv = last_minor_collection_start_tv;
1704 binary_protocol_collection_begin (gc_stats.minor_gc_count, GENERATION_NURSERY);
1706 object_ops_nopar = sgen_concurrent_collection_in_progress ()
1707 ? &sgen_minor_collector.serial_ops_with_concurrent_major
1708 : &sgen_minor_collector.serial_ops;
1709 if (sgen_minor_collector.is_parallel && sgen_nursery_size >= SGEN_PARALLEL_MINOR_MIN_NURSERY_SIZE) {
1710 object_ops_par = sgen_concurrent_collection_in_progress ()
1711 ? &sgen_minor_collector.parallel_ops_with_concurrent_major
1712 : &sgen_minor_collector.parallel_ops;
1716 if (do_verify_nursery || do_dump_nursery_content)
1717 sgen_debug_verify_nursery (do_dump_nursery_content);
1719 current_collection_generation = GENERATION_NURSERY;
1721 SGEN_ASSERT (0, !sgen_collection_is_concurrent (), "Why is the nursery collection concurrent?");
1723 reset_pinned_from_failed_allocation ();
1725 check_scan_starts ();
1727 sgen_nursery_alloc_prepare_for_minor ();
1732 SGEN_LOG (1, "Start nursery collection %d %p-%p, size: %d", gc_stats.minor_gc_count, nursery_section->data, nursery_section->end_data, (int)(nursery_section->end_data - nursery_section->data));
1734 /* world must be stopped already */
1736 time_minor_pre_collection_fragment_clear += TV_ELAPSED (atv, btv);
1738 sgen_client_pre_collection_checks ();
1740 major_collector.start_nursery_collection ();
1742 sgen_memgov_minor_collection_start ();
1744 init_gray_queue (&gc_thread_gray_queue);
1745 ctx = CONTEXT_FROM_OBJECT_OPERATIONS (object_ops_nopar, &gc_thread_gray_queue);
1747 gc_stats.minor_gc_count ++;
1749 sgen_process_fin_stage_entries ();
1751 /* pin from pinned handles */
1752 sgen_init_pinning ();
1753 if (concurrent_collection_in_progress)
1754 sgen_init_pinning_for_conc ();
1755 sgen_client_binary_protocol_mark_start (GENERATION_NURSERY);
1756 pin_from_roots (nursery_section->data, nursery_section->end_data, ctx);
1757 /* pin cemented objects */
1758 sgen_pin_cemented_objects ();
1759 /* identify pinned objects */
1760 sgen_optimize_pin_queue ();
1761 sgen_pinning_setup_section (nursery_section);
1763 pin_objects_in_nursery (FALSE, ctx);
1764 sgen_pinning_trim_queue_to_section (nursery_section);
1765 if (concurrent_collection_in_progress)
1766 sgen_finish_pinning_for_conc ();
1768 if (remset_consistency_checks)
1769 sgen_check_remset_consistency ();
1771 if (whole_heap_check_before_collection) {
1772 sgen_clear_nursery_fragments ();
1773 sgen_check_whole_heap (FALSE);
1777 time_minor_pinning += TV_ELAPSED (btv, atv);
1778 SGEN_LOG (2, "Finding pinned pointers: %zd in %lld usecs", sgen_get_pinned_count (), (long long)TV_ELAPSED (btv, atv));
1779 SGEN_LOG (4, "Start scan with %zd pinned objects", sgen_get_pinned_count ());
1781 remset.start_scan_remsets ();
1783 enqueue_scan_remembered_set_jobs (&gc_thread_gray_queue, is_parallel ? NULL : object_ops_nopar, is_parallel);
1785 /* we don't have complete write barrier yet, so we scan all the old generation sections */
1787 time_minor_scan_remsets += TV_ELAPSED (atv, btv);
1788 SGEN_LOG (2, "Old generation scan: %lld usecs", (long long)TV_ELAPSED (atv, btv));
1790 sgen_pin_stats_report ();
1792 /* FIXME: Why do we do this at this specific, seemingly random, point? */
1793 sgen_client_collecting_minor (&fin_ready_queue, &critical_fin_queue);
1796 time_minor_scan_pinned += TV_ELAPSED (btv, atv);
1798 enqueue_scan_from_roots_jobs (&gc_thread_gray_queue, nursery_section->data, nursery_section->end_data, is_parallel ? NULL : object_ops_nopar, is_parallel);
1801 gray_queue_redirect (&gc_thread_gray_queue);
1802 sgen_workers_start_all_workers (GENERATION_NURSERY, object_ops_nopar, object_ops_par, NULL);
1803 sgen_workers_join (GENERATION_NURSERY);
1807 time_minor_scan_roots += TV_ELAPSED (atv, btv);
1809 finish_gray_stack (GENERATION_NURSERY, ctx);
1812 time_minor_finish_gray_stack += TV_ELAPSED (btv, atv);
1813 sgen_client_binary_protocol_mark_end (GENERATION_NURSERY);
1815 if (objects_pinned) {
1816 sgen_optimize_pin_queue ();
1817 sgen_pinning_setup_section (nursery_section);
1821 * This is the latest point at which we can do this check, because
1822 * sgen_build_nursery_fragments() unpins nursery objects again.
1824 if (remset_consistency_checks)
1825 sgen_check_remset_consistency ();
1828 if (sgen_max_pause_time) {
1832 duration = (int)(TV_ELAPSED (last_minor_collection_start_tv, btv) / 10000);
1833 if (duration > (sgen_max_pause_time * sgen_max_pause_margin))
1834 sgen_resize_nursery (TRUE);
1836 sgen_resize_nursery (FALSE);
1838 sgen_resize_nursery (FALSE);
1841 /* walk the pin_queue, build up the fragment list of free memory, unmark
1842 * pinned objects as we go, memzero() the empty fragments so they are ready for the
1845 sgen_client_binary_protocol_reclaim_start (GENERATION_NURSERY);
1846 fragment_total = sgen_build_nursery_fragments (nursery_section, unpin_queue);
1847 if (!fragment_total)
1850 /* Clear TLABs for all threads */
1851 sgen_clear_tlabs ();
1853 sgen_client_binary_protocol_reclaim_end (GENERATION_NURSERY);
1855 time_minor_fragment_creation += TV_ELAPSED (atv, btv);
1856 SGEN_LOG (2, "Fragment creation: %lld usecs, %lu bytes available", (long long)TV_ELAPSED (atv, btv), (unsigned long)fragment_total);
1858 if (remset_consistency_checks)
1859 sgen_check_major_refs ();
1861 major_collector.finish_nursery_collection ();
1863 TV_GETTIME (last_minor_collection_end_tv);
1864 gc_stats.minor_gc_time += TV_ELAPSED (last_minor_collection_start_tv, last_minor_collection_end_tv);
1866 sgen_debug_dump_heap ("minor", gc_stats.minor_gc_count - 1, NULL);
1868 /* prepare the pin queue for the next collection */
1869 sgen_finish_pinning ();
1870 if (sgen_have_pending_finalizers ()) {
1871 SGEN_LOG (4, "Finalizer-thread wakeup");
1872 sgen_client_finalize_notify ();
1874 sgen_pin_stats_reset ();
1875 /* clear cemented hash */
1876 sgen_cement_clear_below_threshold ();
1878 sgen_gray_object_queue_dispose (&gc_thread_gray_queue);
1880 check_scan_starts ();
1882 binary_protocol_flush_buffers (FALSE);
1884 sgen_memgov_minor_collection_end (reason, is_overflow);
1886 /*objects are late pinned because of lack of memory, so a major is a good call*/
1887 needs_major = objects_pinned > 0;
1888 current_collection_generation = -1;
1892 binary_protocol_collection_end_stats (0, 0, time_minor_finish_gray_stack - finish_gray_start);
1894 binary_protocol_collection_end_stats (
1895 time_minor_scan_major_blocks - major_scan_start,
1896 time_minor_scan_los - los_scan_start,
1897 time_minor_finish_gray_stack - finish_gray_start);
1899 binary_protocol_collection_end (gc_stats.minor_gc_count - 1, GENERATION_NURSERY, 0, 0);
1901 if (check_nursery_objects_pinned && !sgen_minor_collector.is_split)
1902 sgen_check_nursery_objects_pinned (unpin_queue != NULL);
1908 COPY_OR_MARK_FROM_ROOTS_SERIAL,
1909 COPY_OR_MARK_FROM_ROOTS_START_CONCURRENT,
1910 COPY_OR_MARK_FROM_ROOTS_FINISH_CONCURRENT
1911 } CopyOrMarkFromRootsMode;
1914 major_copy_or_mark_from_roots (SgenGrayQueue *gc_thread_gray_queue, size_t *old_next_pin_slot, CopyOrMarkFromRootsMode mode, SgenObjectOperations *object_ops_nopar, SgenObjectOperations *object_ops_par)
1919 /* FIXME: only use these values for the precise scan
1920 * note that to_space pointers should be excluded anyway...
1922 char *heap_start = NULL;
1923 char *heap_end = (char*)-1;
1924 ScanCopyContext ctx = CONTEXT_FROM_OBJECT_OPERATIONS (object_ops_nopar, gc_thread_gray_queue);
1925 gboolean concurrent = mode != COPY_OR_MARK_FROM_ROOTS_SERIAL;
1927 SGEN_ASSERT (0, !!concurrent == !!concurrent_collection_in_progress, "We've been called with the wrong mode.");
1929 if (mode == COPY_OR_MARK_FROM_ROOTS_START_CONCURRENT) {
1930 /*This cleans up unused fragments */
1931 sgen_nursery_allocator_prepare_for_pinning ();
1933 if (do_concurrent_checks)
1934 sgen_debug_check_nursery_is_clean ();
1936 /* The concurrent collector doesn't touch the nursery. */
1937 sgen_nursery_alloc_prepare_for_major ();
1942 /* Pinning depends on this */
1943 sgen_clear_nursery_fragments ();
1945 if (whole_heap_check_before_collection)
1946 sgen_check_whole_heap (TRUE);
1949 time_major_pre_collection_fragment_clear += TV_ELAPSED (atv, btv);
1953 sgen_client_pre_collection_checks ();
1955 if (mode != COPY_OR_MARK_FROM_ROOTS_START_CONCURRENT) {
1956 /* Remsets are not useful for a major collection */
1957 remset.clear_cards ();
1960 sgen_process_fin_stage_entries ();
1963 sgen_init_pinning ();
1964 if (mode == COPY_OR_MARK_FROM_ROOTS_START_CONCURRENT)
1965 sgen_init_pinning_for_conc ();
1966 SGEN_LOG (6, "Collecting pinned addresses");
1967 pin_from_roots ((void*)lowest_heap_address, (void*)highest_heap_address, ctx);
1968 if (mode == COPY_OR_MARK_FROM_ROOTS_FINISH_CONCURRENT) {
1969 /* Pin cemented objects that were forced */
1970 sgen_pin_cemented_objects ();
1972 sgen_optimize_pin_queue ();
1973 if (mode == COPY_OR_MARK_FROM_ROOTS_START_CONCURRENT) {
1975 * Cemented objects that are in the pinned list will be marked. When
1976 * marking concurrently we won't mark mod-union cards for these objects.
1977 * Instead they will remain cemented until the next major collection,
1978 * when we will recheck if they are still pinned in the roots.
1980 sgen_cement_force_pinned ();
1983 sgen_client_collecting_major_1 ();
1986 * pin_queue now contains all candidate pointers, sorted and
1987 * uniqued. We must do two passes now to figure out which
1988 * objects are pinned.
1990 * The first is to find within the pin_queue the area for each
1991 * section. This requires that the pin_queue be sorted. We
1992 * also process the LOS objects and pinned chunks here.
1994 * The second, destructive, pass is to reduce the section
1995 * areas to pointers to the actually pinned objects.
1997 SGEN_LOG (6, "Pinning from sections");
1998 /* first pass for the sections */
1999 sgen_find_section_pin_queue_start_end (nursery_section);
2000 /* identify possible pointers to the insize of large objects */
2001 SGEN_LOG (6, "Pinning from large objects");
2002 for (bigobj = los_object_list; bigobj; bigobj = bigobj->next) {
2004 if (sgen_find_optimized_pin_queue_area ((char*)bigobj->data, (char*)bigobj->data + sgen_los_object_size (bigobj), &dummy, &dummy)) {
2005 binary_protocol_pin (bigobj->data, (gpointer)LOAD_VTABLE (bigobj->data), safe_object_get_size (bigobj->data));
2007 if (sgen_los_object_is_pinned (bigobj->data)) {
2008 SGEN_ASSERT (0, mode == COPY_OR_MARK_FROM_ROOTS_FINISH_CONCURRENT, "LOS objects can only be pinned here after concurrent marking.");
2011 sgen_los_pin_object (bigobj->data);
2012 if (SGEN_OBJECT_HAS_REFERENCES (bigobj->data))
2013 GRAY_OBJECT_ENQUEUE_SERIAL (gc_thread_gray_queue, bigobj->data, sgen_obj_get_descriptor ((GCObject*)bigobj->data));
2014 sgen_pin_stats_register_object (bigobj->data, GENERATION_OLD);
2015 SGEN_LOG (6, "Marked large object %p (%s) size: %lu from roots", bigobj->data,
2016 sgen_client_vtable_get_name (SGEN_LOAD_VTABLE (bigobj->data)),
2017 (unsigned long)sgen_los_object_size (bigobj));
2019 sgen_client_pinned_los_object (bigobj->data);
2023 pin_objects_in_nursery (mode == COPY_OR_MARK_FROM_ROOTS_START_CONCURRENT, ctx);
2024 if (check_nursery_objects_pinned && !sgen_minor_collector.is_split)
2025 sgen_check_nursery_objects_pinned (mode != COPY_OR_MARK_FROM_ROOTS_START_CONCURRENT);
2027 major_collector.pin_objects (gc_thread_gray_queue);
2028 if (old_next_pin_slot)
2029 *old_next_pin_slot = sgen_get_pinned_count ();
2032 time_major_pinning += TV_ELAPSED (atv, btv);
2033 SGEN_LOG (2, "Finding pinned pointers: %zd in %lld usecs", sgen_get_pinned_count (), (long long)TV_ELAPSED (atv, btv));
2034 SGEN_LOG (4, "Start scan with %zd pinned objects", sgen_get_pinned_count ());
2036 if (mode == COPY_OR_MARK_FROM_ROOTS_START_CONCURRENT)
2037 sgen_finish_pinning_for_conc ();
2039 major_collector.init_to_space ();
2041 SGEN_ASSERT (0, sgen_workers_all_done (), "Why are the workers not done when we start or finish a major collection?");
2042 if (mode == COPY_OR_MARK_FROM_ROOTS_FINISH_CONCURRENT) {
2043 if (object_ops_par != NULL)
2044 sgen_workers_set_num_active_workers (GENERATION_OLD, 0);
2045 if (object_ops_par == NULL && sgen_workers_have_idle_work (GENERATION_OLD)) {
2047 * We force the finish of the worker with the new object ops context
2048 * which can also do copying. We need to have finished pinning. On the
2049 * parallel collector, there is no need to drain the private queues
2050 * here, since we can do it as part of the finishing work, achieving
2051 * better work distribution.
2053 sgen_workers_start_all_workers (GENERATION_OLD, object_ops_nopar, object_ops_par, NULL);
2055 sgen_workers_join (GENERATION_OLD);
2059 #ifdef SGEN_DEBUG_INTERNAL_ALLOC
2060 main_gc_thread = mono_native_thread_self ();
2063 sgen_client_collecting_major_2 ();
2066 time_major_scan_pinned += TV_ELAPSED (btv, atv);
2068 sgen_client_collecting_major_3 (&fin_ready_queue, &critical_fin_queue);
2070 enqueue_scan_from_roots_jobs (gc_thread_gray_queue, heap_start, heap_end, object_ops_nopar, FALSE);
2073 time_major_scan_roots += TV_ELAPSED (atv, btv);
2076 * We start the concurrent worker after pinning and after we scanned the roots
2077 * in order to make sure that the worker does not finish before handling all
2080 if (mode == COPY_OR_MARK_FROM_ROOTS_START_CONCURRENT) {
2081 sgen_workers_set_num_active_workers (GENERATION_OLD, 1);
2082 gray_queue_redirect (gc_thread_gray_queue);
2083 if (precleaning_enabled) {
2084 sgen_workers_start_all_workers (GENERATION_OLD, object_ops_nopar, object_ops_par, workers_finish_callback);
2086 sgen_workers_start_all_workers (GENERATION_OLD, object_ops_nopar, object_ops_par, NULL);
2090 if (mode == COPY_OR_MARK_FROM_ROOTS_FINISH_CONCURRENT) {
2091 int i, split_count = sgen_workers_get_job_split_count (GENERATION_OLD);
2092 size_t num_major_sections = major_collector.get_num_major_sections ();
2093 gboolean parallel = object_ops_par != NULL;
2095 /* If we're not parallel we finish the collection on the gc thread */
2097 gray_queue_redirect (gc_thread_gray_queue);
2099 /* Mod union card table */
2100 for (i = 0; i < split_count; i++) {
2101 ParallelScanJob *psj;
2103 psj = (ParallelScanJob*)sgen_thread_pool_job_alloc ("scan mod union cardtable", job_scan_major_mod_union_card_table, sizeof (ParallelScanJob));
2104 psj->scan_job.ops = parallel ? NULL : object_ops_nopar;
2105 psj->scan_job.gc_thread_gray_queue = gc_thread_gray_queue;
2107 psj->job_split_count = split_count;
2108 psj->data = num_major_sections / split_count;
2109 sgen_workers_enqueue_job (GENERATION_OLD, &psj->scan_job.job, parallel);
2111 psj = (ParallelScanJob*)sgen_thread_pool_job_alloc ("scan LOS mod union cardtable", job_scan_los_mod_union_card_table, sizeof (ParallelScanJob));
2112 psj->scan_job.ops = parallel ? NULL : object_ops_nopar;
2113 psj->scan_job.gc_thread_gray_queue = gc_thread_gray_queue;
2115 psj->job_split_count = split_count;
2116 sgen_workers_enqueue_job (GENERATION_OLD, &psj->scan_job.job, parallel);
2121 * If we enqueue a job while workers are running we need to sgen_workers_ensure_awake
2122 * in order to make sure that we are running the idle func and draining all worker
2123 * gray queues. The operation of starting workers implies this, so we start them after
2124 * in order to avoid doing this operation twice. The workers will drain the main gray
2125 * stack that contained roots and pinned objects and also scan the mod union card
2128 sgen_workers_start_all_workers (GENERATION_OLD, object_ops_nopar, object_ops_par, NULL);
2129 sgen_workers_join (GENERATION_OLD);
2133 sgen_pin_stats_report ();
2135 if (mode == COPY_OR_MARK_FROM_ROOTS_START_CONCURRENT) {
2136 sgen_finish_pinning ();
2138 sgen_pin_stats_reset ();
2140 if (do_concurrent_checks)
2141 sgen_debug_check_nursery_is_clean ();
2146 major_start_collection (SgenGrayQueue *gc_thread_gray_queue, const char *reason, gboolean concurrent, size_t *old_next_pin_slot)
2148 SgenObjectOperations *object_ops_nopar, *object_ops_par = NULL;
2150 binary_protocol_collection_begin (gc_stats.major_gc_count, GENERATION_OLD);
2152 current_collection_generation = GENERATION_OLD;
2154 sgen_workers_assert_gray_queue_is_empty (GENERATION_OLD);
2157 sgen_cement_reset ();
2160 g_assert (major_collector.is_concurrent);
2161 concurrent_collection_in_progress = TRUE;
2163 object_ops_nopar = &major_collector.major_ops_concurrent_start;
2164 if (major_collector.is_parallel)
2165 object_ops_par = &major_collector.major_ops_conc_par_start;
2168 object_ops_nopar = &major_collector.major_ops_serial;
2171 reset_pinned_from_failed_allocation ();
2173 sgen_memgov_major_collection_start (concurrent, reason);
2175 //count_ref_nonref_objs ();
2176 //consistency_check ();
2178 check_scan_starts ();
2181 SGEN_LOG (1, "Start major collection %d", gc_stats.major_gc_count);
2182 gc_stats.major_gc_count ++;
2184 if (major_collector.start_major_collection)
2185 major_collector.start_major_collection ();
2187 major_copy_or_mark_from_roots (gc_thread_gray_queue, old_next_pin_slot, concurrent ? COPY_OR_MARK_FROM_ROOTS_START_CONCURRENT : COPY_OR_MARK_FROM_ROOTS_SERIAL, object_ops_nopar, object_ops_par);
2191 major_finish_collection (SgenGrayQueue *gc_thread_gray_queue, const char *reason, gboolean is_overflow, size_t old_next_pin_slot, gboolean forced)
2193 ScannedObjectCounts counts;
2194 SgenObjectOperations *object_ops_nopar;
2195 mword fragment_total;
2198 guint64 major_scan_start = time_major_scan_mod_union_blocks;
2199 guint64 los_scan_start = time_major_scan_mod_union_los;
2200 guint64 finish_gray_start = time_major_finish_gray_stack;
2202 if (concurrent_collection_in_progress) {
2203 SgenObjectOperations *object_ops_par = NULL;
2205 object_ops_nopar = &major_collector.major_ops_concurrent_finish;
2206 if (major_collector.is_parallel)
2207 object_ops_par = &major_collector.major_ops_conc_par_finish;
2209 major_copy_or_mark_from_roots (gc_thread_gray_queue, NULL, COPY_OR_MARK_FROM_ROOTS_FINISH_CONCURRENT, object_ops_nopar, object_ops_par);
2211 #ifdef SGEN_DEBUG_INTERNAL_ALLOC
2212 main_gc_thread = NULL;
2215 object_ops_nopar = &major_collector.major_ops_serial;
2218 sgen_workers_assert_gray_queue_is_empty (GENERATION_OLD);
2221 finish_gray_stack (GENERATION_OLD, CONTEXT_FROM_OBJECT_OPERATIONS (object_ops_nopar, gc_thread_gray_queue));
2223 time_major_finish_gray_stack += TV_ELAPSED (btv, atv);
2225 SGEN_ASSERT (0, sgen_workers_all_done (), "Can't have workers working after joining");
2227 if (objects_pinned) {
2228 g_assert (!concurrent_collection_in_progress);
2231 * This is slow, but we just OOM'd.
2233 * See comment at `sgen_pin_queue_clear_discarded_entries` for how the pin
2234 * queue is laid out at this point.
2236 sgen_pin_queue_clear_discarded_entries (nursery_section, old_next_pin_slot);
2238 * We need to reestablish all pinned nursery objects in the pin queue
2239 * because they're needed for fragment creation. Unpinning happens by
2240 * walking the whole queue, so it's not necessary to reestablish where major
2241 * heap block pins are - all we care is that they're still in there
2244 sgen_optimize_pin_queue ();
2245 sgen_find_section_pin_queue_start_end (nursery_section);
2249 reset_heap_boundaries ();
2250 sgen_update_heap_boundaries ((mword)sgen_get_nursery_start (), (mword)sgen_get_nursery_end ());
2252 /* walk the pin_queue, build up the fragment list of free memory, unmark
2253 * pinned objects as we go, memzero() the empty fragments so they are ready for the
2256 fragment_total = sgen_build_nursery_fragments (nursery_section, NULL);
2257 if (!fragment_total)
2259 SGEN_LOG (4, "Free space in nursery after major %ld", (long)fragment_total);
2261 if (do_concurrent_checks && concurrent_collection_in_progress)
2262 sgen_debug_check_nursery_is_clean ();
2264 /* prepare the pin queue for the next collection */
2265 sgen_finish_pinning ();
2267 /* Clear TLABs for all threads */
2268 sgen_clear_tlabs ();
2270 sgen_pin_stats_reset ();
2272 sgen_cement_clear_below_threshold ();
2274 if (check_mark_bits_after_major_collection)
2275 sgen_check_heap_marked (concurrent_collection_in_progress);
2278 time_major_fragment_creation += TV_ELAPSED (atv, btv);
2280 binary_protocol_sweep_begin (GENERATION_OLD, !major_collector.sweeps_lazily);
2281 sgen_memgov_major_pre_sweep ();
2284 time_major_free_bigobjs += TV_ELAPSED (btv, atv);
2289 time_major_los_sweep += TV_ELAPSED (atv, btv);
2291 major_collector.sweep ();
2293 binary_protocol_sweep_end (GENERATION_OLD, !major_collector.sweeps_lazily);
2296 time_major_sweep += TV_ELAPSED (btv, atv);
2298 sgen_debug_dump_heap ("major", gc_stats.major_gc_count - 1, reason);
2300 if (sgen_have_pending_finalizers ()) {
2301 SGEN_LOG (4, "Finalizer-thread wakeup");
2302 sgen_client_finalize_notify ();
2305 sgen_memgov_major_collection_end (forced, concurrent_collection_in_progress, reason, is_overflow);
2306 current_collection_generation = -1;
2308 memset (&counts, 0, sizeof (ScannedObjectCounts));
2309 major_collector.finish_major_collection (&counts);
2311 sgen_workers_assert_gray_queue_is_empty (GENERATION_OLD);
2313 SGEN_ASSERT (0, sgen_workers_all_done (), "Can't have workers working after major collection has finished");
2314 if (concurrent_collection_in_progress)
2315 concurrent_collection_in_progress = FALSE;
2317 check_scan_starts ();
2319 binary_protocol_flush_buffers (FALSE);
2321 //consistency_check ();
2322 if (major_collector.is_parallel)
2323 binary_protocol_collection_end_stats (0, 0, time_major_finish_gray_stack - finish_gray_start);
2325 binary_protocol_collection_end_stats (
2326 time_major_scan_mod_union_blocks - major_scan_start,
2327 time_major_scan_mod_union_los - los_scan_start,
2328 time_major_finish_gray_stack - finish_gray_start);
2330 binary_protocol_collection_end (gc_stats.major_gc_count - 1, GENERATION_OLD, counts.num_scanned_objects, counts.num_unique_scanned_objects);
2334 major_do_collection (const char *reason, gboolean is_overflow, gboolean forced)
2336 TV_DECLARE (time_start);
2337 TV_DECLARE (time_end);
2338 size_t old_next_pin_slot;
2339 SgenGrayQueue gc_thread_gray_queue;
2341 if (disable_major_collections)
2344 if (major_collector.get_and_reset_num_major_objects_marked) {
2345 long long num_marked = major_collector.get_and_reset_num_major_objects_marked ();
2346 g_assert (!num_marked);
2349 /* world must be stopped already */
2350 TV_GETTIME (time_start);
2352 init_gray_queue (&gc_thread_gray_queue);
2353 major_start_collection (&gc_thread_gray_queue, reason, FALSE, &old_next_pin_slot);
2354 major_finish_collection (&gc_thread_gray_queue, reason, is_overflow, old_next_pin_slot, forced);
2355 sgen_gray_object_queue_dispose (&gc_thread_gray_queue);
2357 TV_GETTIME (time_end);
2358 gc_stats.major_gc_time += TV_ELAPSED (time_start, time_end);
2360 /* FIXME: also report this to the user, preferably in gc-end. */
2361 if (major_collector.get_and_reset_num_major_objects_marked)
2362 major_collector.get_and_reset_num_major_objects_marked ();
2364 return bytes_pinned_from_failed_allocation > 0;
2368 major_start_concurrent_collection (const char *reason)
2370 TV_DECLARE (time_start);
2371 TV_DECLARE (time_end);
2372 long long num_objects_marked;
2373 SgenGrayQueue gc_thread_gray_queue;
2375 if (disable_major_collections)
2378 TV_GETTIME (time_start);
2379 SGEN_TV_GETTIME (time_major_conc_collection_start);
2381 num_objects_marked = major_collector.get_and_reset_num_major_objects_marked ();
2382 g_assert (num_objects_marked == 0);
2384 binary_protocol_concurrent_start ();
2386 init_gray_queue (&gc_thread_gray_queue);
2387 // FIXME: store reason and pass it when finishing
2388 major_start_collection (&gc_thread_gray_queue, reason, TRUE, NULL);
2389 sgen_gray_object_queue_dispose (&gc_thread_gray_queue);
2391 num_objects_marked = major_collector.get_and_reset_num_major_objects_marked ();
2393 TV_GETTIME (time_end);
2394 gc_stats.major_gc_time += TV_ELAPSED (time_start, time_end);
2396 current_collection_generation = -1;
2400 * Returns whether the major collection has finished.
2403 major_should_finish_concurrent_collection (void)
2405 return sgen_workers_all_done ();
2409 major_update_concurrent_collection (void)
2411 TV_DECLARE (total_start);
2412 TV_DECLARE (total_end);
2414 TV_GETTIME (total_start);
2416 binary_protocol_concurrent_update ();
2418 major_collector.update_cardtable_mod_union ();
2419 sgen_los_update_cardtable_mod_union ();
2421 TV_GETTIME (total_end);
2422 gc_stats.major_gc_time += TV_ELAPSED (total_start, total_end);
2426 major_finish_concurrent_collection (gboolean forced)
2428 SgenGrayQueue gc_thread_gray_queue;
2429 TV_DECLARE (total_start);
2430 TV_DECLARE (total_end);
2432 TV_GETTIME (total_start);
2434 binary_protocol_concurrent_finish ();
2437 * We need to stop all workers since we're updating the cardtable below.
2438 * The workers will be resumed with a finishing pause context to avoid
2439 * additional cardtable and object scanning.
2441 sgen_workers_stop_all_workers (GENERATION_OLD);
2443 SGEN_TV_GETTIME (time_major_conc_collection_end);
2444 gc_stats.major_gc_time_concurrent += SGEN_TV_ELAPSED (time_major_conc_collection_start, time_major_conc_collection_end);
2446 major_collector.update_cardtable_mod_union ();
2447 sgen_los_update_cardtable_mod_union ();
2449 if (mod_union_consistency_check)
2450 sgen_check_mod_union_consistency ();
2452 current_collection_generation = GENERATION_OLD;
2453 sgen_cement_reset ();
2454 init_gray_queue (&gc_thread_gray_queue);
2455 major_finish_collection (&gc_thread_gray_queue, "finishing", FALSE, -1, forced);
2456 sgen_gray_object_queue_dispose (&gc_thread_gray_queue);
2458 TV_GETTIME (total_end);
2459 gc_stats.major_gc_time += TV_ELAPSED (total_start, total_end);
2461 current_collection_generation = -1;
2465 * Ensure an allocation request for @size will succeed by freeing enough memory.
2467 * LOCKING: The GC lock MUST be held.
2470 sgen_ensure_free_space (size_t size, int generation)
2472 int generation_to_collect = -1;
2473 const char *reason = NULL;
2475 if (generation == GENERATION_OLD) {
2476 if (sgen_need_major_collection (size)) {
2477 reason = "LOS overflow";
2478 generation_to_collect = GENERATION_OLD;
2481 if (degraded_mode) {
2482 if (sgen_need_major_collection (size)) {
2483 reason = "Degraded mode overflow";
2484 generation_to_collect = GENERATION_OLD;
2486 } else if (sgen_need_major_collection (size)) {
2487 reason = concurrent_collection_in_progress ? "Forced finish concurrent collection" : "Minor allowance";
2488 generation_to_collect = GENERATION_OLD;
2490 generation_to_collect = GENERATION_NURSERY;
2491 reason = "Nursery full";
2495 if (generation_to_collect == -1) {
2496 if (concurrent_collection_in_progress && sgen_workers_all_done ()) {
2497 generation_to_collect = GENERATION_OLD;
2498 reason = "Finish concurrent collection";
2502 if (generation_to_collect == -1)
2504 sgen_perform_collection (size, generation_to_collect, reason, FALSE, TRUE);
2508 * LOCKING: Assumes the GC lock is held.
2511 sgen_perform_collection (size_t requested_size, int generation_to_collect, const char *reason, gboolean wait_to_finish, gboolean stw)
2513 TV_DECLARE (gc_total_start);
2514 TV_DECLARE (gc_total_end);
2515 int overflow_generation_to_collect = -1;
2516 int oldest_generation_collected = generation_to_collect;
2517 const char *overflow_reason = NULL;
2518 gboolean finish_concurrent = concurrent_collection_in_progress && (major_should_finish_concurrent_collection () || generation_to_collect == GENERATION_OLD);
2520 binary_protocol_collection_requested (generation_to_collect, requested_size, wait_to_finish ? 1 : 0);
2522 SGEN_ASSERT (0, generation_to_collect == GENERATION_NURSERY || generation_to_collect == GENERATION_OLD, "What generation is this?");
2525 sgen_stop_world (generation_to_collect);
2527 SGEN_ASSERT (0, sgen_is_world_stopped (), "We can only collect if the world is stopped");
2530 TV_GETTIME (gc_total_start);
2532 // FIXME: extract overflow reason
2533 // FIXME: minor overflow for concurrent case
2534 if (generation_to_collect == GENERATION_NURSERY && !finish_concurrent) {
2535 if (concurrent_collection_in_progress)
2536 major_update_concurrent_collection ();
2538 if (collect_nursery (reason, FALSE, NULL) && !concurrent_collection_in_progress) {
2539 overflow_generation_to_collect = GENERATION_OLD;
2540 overflow_reason = "Minor overflow";
2542 } else if (finish_concurrent) {
2543 major_finish_concurrent_collection (wait_to_finish);
2544 oldest_generation_collected = GENERATION_OLD;
2546 SGEN_ASSERT (0, generation_to_collect == GENERATION_OLD, "We should have handled nursery collections above");
2547 if (major_collector.is_concurrent && !wait_to_finish) {
2548 collect_nursery ("Concurrent start", FALSE, NULL);
2549 major_start_concurrent_collection (reason);
2550 oldest_generation_collected = GENERATION_NURSERY;
2551 } else if (major_do_collection (reason, FALSE, wait_to_finish)) {
2552 overflow_generation_to_collect = GENERATION_NURSERY;
2553 overflow_reason = "Excessive pinning";
2557 if (overflow_generation_to_collect != -1) {
2558 SGEN_ASSERT (0, !concurrent_collection_in_progress, "We don't yet support overflow collections with the concurrent collector");
2561 * We need to do an overflow collection, either because we ran out of memory
2562 * or the nursery is fully pinned.
2565 if (overflow_generation_to_collect == GENERATION_NURSERY)
2566 collect_nursery (overflow_reason, TRUE, NULL);
2568 major_do_collection (overflow_reason, TRUE, wait_to_finish);
2570 oldest_generation_collected = MAX (oldest_generation_collected, overflow_generation_to_collect);
2573 SGEN_LOG (2, "Heap size: %lu, LOS size: %lu", (unsigned long)sgen_gc_get_total_heap_allocation (), (unsigned long)los_memory_usage);
2575 /* this also sets the proper pointers for the next allocation */
2576 if (generation_to_collect == GENERATION_NURSERY && !sgen_can_alloc_size (requested_size)) {
2577 /* TypeBuilder and MonoMethod are killing mcs with fragmentation */
2578 SGEN_LOG (1, "nursery collection didn't find enough room for %zd alloc (%zd pinned)", requested_size, sgen_get_pinned_count ());
2579 sgen_dump_pin_queue ();
2583 TV_GETTIME (gc_total_end);
2584 time_max = MAX (time_max, TV_ELAPSED (gc_total_start, gc_total_end));
2587 sgen_restart_world (oldest_generation_collected);
2591 * ######################################################################
2592 * ######## Memory allocation from the OS
2593 * ######################################################################
2594 * This section of code deals with getting memory from the OS and
2595 * allocating memory for GC-internal data structures.
2596 * Internal memory can be handled with a freelist for small objects.
2602 G_GNUC_UNUSED static void
2603 report_internal_mem_usage (void)
2605 printf ("Internal memory usage:\n");
2606 sgen_report_internal_mem_usage ();
2607 printf ("Pinned memory usage:\n");
2608 major_collector.report_pinned_memory_usage ();
2612 * ######################################################################
2613 * ######## Finalization support
2614 * ######################################################################
2618 * If the object has been forwarded it means it's still referenced from a root.
2619 * If it is pinned it's still alive as well.
2620 * A LOS object is only alive if we have pinned it.
2621 * Return TRUE if @obj is ready to be finalized.
2623 static inline gboolean
2624 sgen_is_object_alive (GCObject *object)
2626 if (ptr_in_nursery (object))
2627 return sgen_nursery_is_object_alive (object);
2629 return sgen_major_is_object_alive (object);
2633 * This function returns true if @object is either alive and belongs to the
2634 * current collection - major collections are full heap, so old gen objects
2635 * are never alive during a minor collection.
2638 sgen_is_object_alive_and_on_current_collection (GCObject *object)
2640 if (ptr_in_nursery (object))
2641 return sgen_nursery_is_object_alive (object);
2643 if (current_collection_generation == GENERATION_NURSERY)
2646 return sgen_major_is_object_alive (object);
2651 sgen_gc_is_object_ready_for_finalization (GCObject *object)
2653 return !sgen_is_object_alive (object);
2657 sgen_queue_finalization_entry (GCObject *obj)
2659 gboolean critical = sgen_client_object_has_critical_finalizer (obj);
2661 sgen_pointer_queue_add (critical ? &critical_fin_queue : &fin_ready_queue, obj);
2663 sgen_client_object_queued_for_finalization (obj);
2667 sgen_object_is_live (GCObject *obj)
2669 return sgen_is_object_alive_and_on_current_collection (obj);
2673 * `System.GC.WaitForPendingFinalizers` first checks `sgen_have_pending_finalizers()` to
2674 * determine whether it can exit quickly. The latter must therefore only return FALSE if
2675 * all finalizers have really finished running.
2677 * `sgen_gc_invoke_finalizers()` first dequeues a finalizable object, and then finalizes it.
2678 * This means that just checking whether the queues are empty leaves the possibility that an
2679 * object might have been dequeued but not yet finalized. That's why we need the additional
2680 * flag `pending_unqueued_finalizer`.
2683 static volatile gboolean pending_unqueued_finalizer = FALSE;
2684 volatile gboolean sgen_suspend_finalizers = FALSE;
2687 sgen_set_suspend_finalizers (void)
2689 sgen_suspend_finalizers = TRUE;
2693 sgen_gc_invoke_finalizers (void)
2697 g_assert (!pending_unqueued_finalizer);
2699 /* FIXME: batch to reduce lock contention */
2700 while (sgen_have_pending_finalizers ()) {
2706 * We need to set `pending_unqueued_finalizer` before dequeing the
2707 * finalizable object.
2709 if (!sgen_pointer_queue_is_empty (&fin_ready_queue)) {
2710 pending_unqueued_finalizer = TRUE;
2711 mono_memory_write_barrier ();
2712 obj = (GCObject *)sgen_pointer_queue_pop (&fin_ready_queue);
2713 } else if (!sgen_pointer_queue_is_empty (&critical_fin_queue)) {
2714 pending_unqueued_finalizer = TRUE;
2715 mono_memory_write_barrier ();
2716 obj = (GCObject *)sgen_pointer_queue_pop (&critical_fin_queue);
2722 SGEN_LOG (7, "Finalizing object %p (%s)", obj, sgen_client_vtable_get_name (SGEN_LOAD_VTABLE (obj)));
2730 /* the object is on the stack so it is pinned */
2731 /*g_print ("Calling finalizer for object: %p (%s)\n", obj, sgen_client_object_safe_name (obj));*/
2732 sgen_client_run_finalize (obj);
2735 if (pending_unqueued_finalizer) {
2736 mono_memory_write_barrier ();
2737 pending_unqueued_finalizer = FALSE;
2744 sgen_have_pending_finalizers (void)
2746 if (sgen_suspend_finalizers)
2748 return pending_unqueued_finalizer || !sgen_pointer_queue_is_empty (&fin_ready_queue) || !sgen_pointer_queue_is_empty (&critical_fin_queue);
2752 * ######################################################################
2753 * ######## registered roots support
2754 * ######################################################################
2758 * We do not coalesce roots.
2761 sgen_register_root (char *start, size_t size, SgenDescriptor descr, int root_type, int source, const char *msg)
2763 RootRecord new_root;
2766 for (i = 0; i < ROOT_TYPE_NUM; ++i) {
2767 RootRecord *root = (RootRecord *)sgen_hash_table_lookup (&roots_hash [i], start);
2768 /* we allow changing the size and the descriptor (for thread statics etc) */
2770 size_t old_size = root->end_root - start;
2771 root->end_root = start + size;
2772 SGEN_ASSERT (0, !!root->root_desc == !!descr, "Can't change whether a root is precise or conservative.");
2773 SGEN_ASSERT (0, root->source == source, "Can't change a root's source identifier.");
2774 SGEN_ASSERT (0, !!root->msg == !!msg, "Can't change a root's message.");
2775 root->root_desc = descr;
2777 roots_size -= old_size;
2783 new_root.end_root = start + size;
2784 new_root.root_desc = descr;
2785 new_root.source = source;
2788 sgen_hash_table_replace (&roots_hash [root_type], start, &new_root, NULL);
2791 SGEN_LOG (3, "Added root for range: %p-%p, descr: %llx (%d/%d bytes)", start, new_root.end_root, (long long)descr, (int)size, (int)roots_size);
2798 sgen_deregister_root (char* addr)
2804 for (root_type = 0; root_type < ROOT_TYPE_NUM; ++root_type) {
2805 if (sgen_hash_table_remove (&roots_hash [root_type], addr, &root))
2806 roots_size -= (root.end_root - addr);
2812 sgen_wbroots_iterate_live_block_ranges (sgen_cardtable_block_callback cb)
2816 SGEN_HASH_TABLE_FOREACH (&roots_hash [ROOT_TYPE_WBARRIER], void **, start_root, RootRecord *, root) {
2817 cb ((mword)start_root, (mword)root->end_root - (mword)start_root);
2818 } SGEN_HASH_TABLE_FOREACH_END;
2821 /* Root equivalent of sgen_client_cardtable_scan_object */
2823 sgen_wbroot_scan_card_table (void** start_root, mword size, ScanCopyContext ctx)
2825 ScanPtrFieldFunc scan_field_func = ctx.ops->scan_ptr_field;
2826 guint8 *card_data = sgen_card_table_get_card_scan_address ((mword)start_root);
2827 guint8 *card_base = card_data;
2828 mword card_count = sgen_card_table_number_of_cards_in_range ((mword)start_root, size);
2829 guint8 *card_data_end = card_data + card_count;
2830 mword extra_idx = 0;
2831 char *obj_start = sgen_card_table_align_pointer (start_root);
2832 char *obj_end = (char*)start_root + size;
2833 #ifdef SGEN_HAVE_OVERLAPPING_CARDS
2834 guint8 *overflow_scan_end = NULL;
2837 #ifdef SGEN_HAVE_OVERLAPPING_CARDS
2838 /*Check for overflow and if so, setup to scan in two steps*/
2839 if (card_data_end >= SGEN_SHADOW_CARDTABLE_END) {
2840 overflow_scan_end = sgen_shadow_cardtable + (card_data_end - SGEN_SHADOW_CARDTABLE_END);
2841 card_data_end = SGEN_SHADOW_CARDTABLE_END;
2847 card_data = sgen_find_next_card (card_data, card_data_end);
2849 for (; card_data < card_data_end; card_data = sgen_find_next_card (card_data + 1, card_data_end)) {
2850 size_t idx = (card_data - card_base) + extra_idx;
2851 char *start = (char*)(obj_start + idx * CARD_SIZE_IN_BYTES);
2852 char *card_end = start + CARD_SIZE_IN_BYTES;
2853 char *elem = start, *first_elem = start;
2856 * Don't clean first and last card on 32bit systems since they
2857 * may also be part from other roots.
2859 if (card_data != card_base && card_data != (card_data_end - 1))
2860 sgen_card_table_prepare_card_for_scanning (card_data);
2862 card_end = MIN (card_end, obj_end);
2864 if (elem < (char*)start_root)
2865 first_elem = elem = (char*)start_root;
2867 for (; elem < card_end; elem += SIZEOF_VOID_P) {
2868 if (*(GCObject**)elem)
2869 scan_field_func (NULL, (GCObject**)elem, ctx.queue);
2872 binary_protocol_card_scan (first_elem, elem - first_elem);
2875 #ifdef SGEN_HAVE_OVERLAPPING_CARDS
2876 if (overflow_scan_end) {
2877 extra_idx = card_data - card_base;
2878 card_base = card_data = sgen_shadow_cardtable;
2879 card_data_end = overflow_scan_end;
2880 overflow_scan_end = NULL;
2887 sgen_wbroots_scan_card_table (ScanCopyContext ctx)
2892 SGEN_HASH_TABLE_FOREACH (&roots_hash [ROOT_TYPE_WBARRIER], void **, start_root, RootRecord *, root) {
2893 SGEN_ASSERT (0, (root->root_desc & ROOT_DESC_TYPE_MASK) == ROOT_DESC_VECTOR, "Unsupported root type");
2895 sgen_wbroot_scan_card_table (start_root, (mword)root->end_root - (mword)start_root, ctx);
2896 } SGEN_HASH_TABLE_FOREACH_END;
2900 * ######################################################################
2901 * ######## Thread handling (stop/start code)
2902 * ######################################################################
2906 sgen_get_current_collection_generation (void)
2908 return current_collection_generation;
2912 sgen_thread_attach (SgenThreadInfo* info)
2914 info->tlab_start = info->tlab_next = info->tlab_temp_end = info->tlab_real_end = NULL;
2916 sgen_client_thread_attach (info);
2922 sgen_thread_detach_with_lock (SgenThreadInfo *p)
2924 sgen_client_thread_detach_with_lock (p);
2928 * ######################################################################
2929 * ######## Write barriers
2930 * ######################################################################
2934 * Note: the write barriers first do the needed GC work and then do the actual store:
2935 * this way the value is visible to the conservative GC scan after the write barrier
2936 * itself. If a GC interrupts the barrier in the middle, value will be kept alive by
2937 * the conservative scan, otherwise by the remembered set scan.
2941 * mono_gc_wbarrier_arrayref_copy:
2944 mono_gc_wbarrier_arrayref_copy (gpointer dest_ptr, gpointer src_ptr, int count)
2946 HEAVY_STAT (++stat_wbarrier_arrayref_copy);
2947 /*This check can be done without taking a lock since dest_ptr array is pinned*/
2948 if (ptr_in_nursery (dest_ptr) || count <= 0) {
2949 mono_gc_memmove_aligned (dest_ptr, src_ptr, count * sizeof (gpointer));
2953 #ifdef SGEN_HEAVY_BINARY_PROTOCOL
2954 if (binary_protocol_is_heavy_enabled ()) {
2956 for (i = 0; i < count; ++i) {
2957 gpointer dest = (gpointer*)dest_ptr + i;
2958 gpointer obj = *((gpointer*)src_ptr + i);
2960 binary_protocol_wbarrier (dest, obj, (gpointer)LOAD_VTABLE (obj));
2965 remset.wbarrier_arrayref_copy (dest_ptr, src_ptr, count);
2969 * mono_gc_wbarrier_generic_nostore:
2972 mono_gc_wbarrier_generic_nostore (gpointer ptr)
2976 HEAVY_STAT (++stat_wbarrier_generic_store);
2978 sgen_client_wbarrier_generic_nostore_check (ptr);
2980 obj = *(gpointer*)ptr;
2982 binary_protocol_wbarrier (ptr, obj, (gpointer)LOAD_VTABLE (obj));
2985 * We need to record old->old pointer locations for the
2986 * concurrent collector.
2988 if (!ptr_in_nursery (obj) && !concurrent_collection_in_progress) {
2989 SGEN_LOG (8, "Skipping remset at %p", ptr);
2993 SGEN_LOG (8, "Adding remset at %p", ptr);
2995 remset.wbarrier_generic_nostore (ptr);
2999 * mono_gc_wbarrier_generic_store:
3002 mono_gc_wbarrier_generic_store (gpointer ptr, GCObject* value)
3004 SGEN_LOG (8, "Wbarrier store at %p to %p (%s)", ptr, value, value ? sgen_client_vtable_get_name (SGEN_LOAD_VTABLE (value)) : "null");
3005 SGEN_UPDATE_REFERENCE_ALLOW_NULL (ptr, value);
3006 if (ptr_in_nursery (value) || concurrent_collection_in_progress)
3007 mono_gc_wbarrier_generic_nostore (ptr);
3008 sgen_dummy_use (value);
3012 * mono_gc_wbarrier_generic_store_atomic:
3013 * Same as \c mono_gc_wbarrier_generic_store but performs the store
3014 * as an atomic operation with release semantics.
3017 mono_gc_wbarrier_generic_store_atomic (gpointer ptr, GCObject *value)
3019 HEAVY_STAT (++stat_wbarrier_generic_store_atomic);
3021 SGEN_LOG (8, "Wbarrier atomic store at %p to %p (%s)", ptr, value, value ? sgen_client_vtable_get_name (SGEN_LOAD_VTABLE (value)) : "null");
3023 InterlockedWritePointer ((volatile gpointer *)ptr, value);
3025 if (ptr_in_nursery (value) || concurrent_collection_in_progress)
3026 mono_gc_wbarrier_generic_nostore (ptr);
3028 sgen_dummy_use (value);
3032 sgen_wbarrier_range_copy (gpointer _dest, gpointer _src, int size)
3034 remset.wbarrier_range_copy (_dest,_src, size);
3038 * ######################################################################
3039 * ######## Other mono public interface functions.
3040 * ######################################################################
3044 sgen_gc_collect (int generation)
3049 sgen_perform_collection (0, generation, "user request", TRUE, TRUE);
3054 sgen_gc_collection_count (int generation)
3056 if (generation == 0)
3057 return gc_stats.minor_gc_count;
3058 return gc_stats.major_gc_count;
3062 sgen_gc_get_used_size (void)
3066 tot = los_memory_usage;
3067 tot += nursery_section->end_data - nursery_section->data;
3068 tot += major_collector.get_used_size ();
3069 /* FIXME: account for pinned objects */
3075 sgen_env_var_error (const char *env_var, const char *fallback, const char *description_format, ...)
3079 va_start (ap, description_format);
3081 fprintf (stderr, "Warning: In environment variable `%s': ", env_var);
3082 vfprintf (stderr, description_format, ap);
3084 fprintf (stderr, " - %s", fallback);
3085 fprintf (stderr, "\n");
3091 parse_double_in_interval (const char *env_var, const char *opt_name, const char *opt, double min, double max, double *result)
3094 double val = strtod (opt, &endptr);
3095 if (endptr == opt) {
3096 sgen_env_var_error (env_var, "Using default value.", "`%s` must be a number.", opt_name);
3099 else if (val < min || val > max) {
3100 sgen_env_var_error (env_var, "Using default value.", "`%s` must be between %.2f - %.2f.", opt_name, min, max);
3108 parse_sgen_minor (const char *opt)
3111 return SGEN_MINOR_DEFAULT;
3113 if (!strcmp (opt, "simple")) {
3114 return SGEN_MINOR_SIMPLE;
3115 } else if (!strcmp (opt, "simple-par")) {
3116 return SGEN_MINOR_SIMPLE_PARALLEL;
3117 } else if (!strcmp (opt, "split")) {
3118 return SGEN_MINOR_SPLIT;
3120 sgen_env_var_error (MONO_GC_PARAMS_NAME, "Using default instead.", "Unknown minor collector `%s'.", opt);
3121 return SGEN_MINOR_DEFAULT;
3126 parse_sgen_major (const char *opt)
3129 return SGEN_MAJOR_DEFAULT;
3131 if (!strcmp (opt, "marksweep")) {
3132 return SGEN_MAJOR_SERIAL;
3133 } else if (!strcmp (opt, "marksweep-conc")) {
3134 return SGEN_MAJOR_CONCURRENT;
3135 } else if (!strcmp (opt, "marksweep-conc-par")) {
3136 return SGEN_MAJOR_CONCURRENT_PARALLEL;
3138 sgen_env_var_error (MONO_GC_PARAMS_NAME, "Using default instead.", "Unknown major collector `%s'.", opt);
3139 return SGEN_MAJOR_DEFAULT;
3145 parse_sgen_mode (const char *opt)
3148 return SGEN_MODE_NONE;
3150 if (!strcmp (opt, "balanced")) {
3151 return SGEN_MODE_BALANCED;
3152 } else if (!strcmp (opt, "throughput")) {
3153 return SGEN_MODE_THROUGHPUT;
3154 } else if (!strcmp (opt, "pause") || g_str_has_prefix (opt, "pause:")) {
3155 return SGEN_MODE_PAUSE;
3157 sgen_env_var_error (MONO_GC_PARAMS_NAME, "Using default configurations.", "Unknown mode `%s'.", opt);
3158 return SGEN_MODE_NONE;
3163 init_sgen_minor (SgenMinor minor)
3166 case SGEN_MINOR_DEFAULT:
3167 case SGEN_MINOR_SIMPLE:
3168 sgen_simple_nursery_init (&sgen_minor_collector, FALSE);
3170 case SGEN_MINOR_SIMPLE_PARALLEL:
3171 sgen_simple_nursery_init (&sgen_minor_collector, TRUE);
3173 case SGEN_MINOR_SPLIT:
3174 sgen_split_nursery_init (&sgen_minor_collector);
3177 g_assert_not_reached ();
3182 init_sgen_major (SgenMajor major)
3184 if (major == SGEN_MAJOR_DEFAULT)
3185 major = DEFAULT_MAJOR;
3188 case SGEN_MAJOR_SERIAL:
3189 sgen_marksweep_init (&major_collector);
3191 case SGEN_MAJOR_CONCURRENT:
3192 sgen_marksweep_conc_init (&major_collector);
3194 case SGEN_MAJOR_CONCURRENT_PARALLEL:
3195 sgen_marksweep_conc_par_init (&major_collector);
3198 g_assert_not_reached ();
3203 * If sgen mode is set, major/minor configuration is fixed. The other gc_params
3204 * are parsed and processed after major/minor initialization, so it can potentially
3205 * override some knobs set by the sgen mode. We can consider locking out additional
3206 * configurations when gc_modes are used.
3209 init_sgen_mode (SgenMode mode)
3211 SgenMinor minor = SGEN_MINOR_DEFAULT;
3212 SgenMajor major = SGEN_MAJOR_DEFAULT;
3215 case SGEN_MODE_BALANCED:
3217 * Use a dynamic parallel nursery with a major concurrent collector.
3218 * This uses the default values for max pause time and nursery size.
3220 minor = SGEN_MINOR_SIMPLE;
3221 major = SGEN_MAJOR_CONCURRENT;
3222 dynamic_nursery = TRUE;
3224 case SGEN_MODE_THROUGHPUT:
3226 * Use concurrent major to let the mutator do more work. Use a larger
3227 * nursery, without pause time constraints, in order to collect more
3228 * objects in parallel and avoid repetitive collection tasks (pinning,
3229 * root scanning etc)
3231 minor = SGEN_MINOR_SIMPLE_PARALLEL;
3232 major = SGEN_MAJOR_CONCURRENT;
3233 dynamic_nursery = TRUE;
3234 sgen_max_pause_time = 0;
3236 case SGEN_MODE_PAUSE:
3238 * Use concurrent major and dynamic nursery with a more
3239 * aggressive shrinking relative to pause times.
3241 minor = SGEN_MINOR_SIMPLE_PARALLEL;
3242 major = SGEN_MAJOR_CONCURRENT;
3243 dynamic_nursery = TRUE;
3244 sgen_max_pause_margin = SGEN_PAUSE_MODE_MAX_PAUSE_MARGIN;
3247 g_assert_not_reached ();
3250 init_sgen_minor (minor);
3251 init_sgen_major (major);
3259 SgenMajor sgen_major = SGEN_MAJOR_DEFAULT;
3260 SgenMinor sgen_minor = SGEN_MINOR_DEFAULT;
3261 SgenMode sgen_mode = SGEN_MODE_NONE;
3262 char *params_opts = NULL;
3263 char *debug_opts = NULL;
3264 size_t max_heap = 0;
3265 size_t soft_limit = 0;
3267 gboolean debug_print_allowance = FALSE;
3268 double allowance_ratio = 0, save_target = 0;
3269 gboolean cement_enabled = TRUE;
3272 result = InterlockedCompareExchange (&gc_initialized, -1, 0);
3275 /* already inited */
3278 /* being inited by another thread */
3279 mono_thread_info_usleep (1000);
3282 /* we will init it */
3285 g_assert_not_reached ();
3287 } while (result != 0);
3289 SGEN_TV_GETTIME (sgen_init_timestamp);
3291 #ifdef SGEN_WITHOUT_MONO
3292 mono_thread_smr_init ();
3295 mono_coop_mutex_init (&gc_mutex);
3297 gc_debug_file = stderr;
3299 mono_coop_mutex_init (&sgen_interruption_mutex);
3301 if ((env = g_getenv (MONO_GC_PARAMS_NAME)) || gc_params_options) {
3302 params_opts = g_strdup_printf ("%s,%s", gc_params_options ? gc_params_options : "", env ? env : "");
3307 opts = g_strsplit (params_opts, ",", -1);
3308 for (ptr = opts; *ptr; ++ptr) {
3310 if (g_str_has_prefix (opt, "major=")) {
3311 opt = strchr (opt, '=') + 1;
3312 sgen_major = parse_sgen_major (opt);
3313 } else if (g_str_has_prefix (opt, "minor=")) {
3314 opt = strchr (opt, '=') + 1;
3315 sgen_minor = parse_sgen_minor (opt);
3316 } else if (g_str_has_prefix (opt, "mode=")) {
3317 opt = strchr (opt, '=') + 1;
3318 sgen_mode = parse_sgen_mode (opt);
3326 sgen_init_internal_allocator ();
3327 sgen_init_nursery_allocator ();
3328 sgen_init_fin_weak_hash ();
3329 sgen_init_hash_table ();
3330 sgen_init_descriptors ();
3331 sgen_init_gray_queues ();
3332 sgen_init_allocator ();
3333 sgen_init_gchandles ();
3335 sgen_register_fixed_internal_mem_type (INTERNAL_MEM_SECTION, SGEN_SIZEOF_GC_MEM_SECTION);
3336 sgen_register_fixed_internal_mem_type (INTERNAL_MEM_GRAY_QUEUE, sizeof (GrayQueueSection));
3338 sgen_client_init ();
3340 if (sgen_mode != SGEN_MODE_NONE) {
3341 if (sgen_minor != SGEN_MINOR_DEFAULT || sgen_major != SGEN_MAJOR_DEFAULT)
3342 sgen_env_var_error (MONO_GC_PARAMS_NAME, "Ignoring major/minor configuration", "Major/minor configurations cannot be used with sgen modes");
3343 init_sgen_mode (sgen_mode);
3345 init_sgen_minor (sgen_minor);
3346 init_sgen_major (sgen_major);
3350 gboolean usage_printed = FALSE;
3352 for (ptr = opts; *ptr; ++ptr) {
3354 if (!strcmp (opt, ""))
3356 if (g_str_has_prefix (opt, "major="))
3358 if (g_str_has_prefix (opt, "minor="))
3360 if (g_str_has_prefix (opt, "mode=")) {
3361 if (g_str_has_prefix (opt, "mode=pause:")) {
3362 char *str_pause = strchr (opt, ':') + 1;
3363 int pause = atoi (str_pause);
3365 sgen_max_pause_time = pause;
3367 sgen_env_var_error (MONO_GC_PARAMS_NAME, "Using default", "Invalid maximum pause time for `pause` sgen mode");
3371 if (g_str_has_prefix (opt, "max-heap-size=")) {
3372 size_t page_size = mono_pagesize ();
3373 size_t max_heap_candidate = 0;
3374 opt = strchr (opt, '=') + 1;
3375 if (*opt && mono_gc_parse_environment_string_extract_number (opt, &max_heap_candidate)) {
3376 max_heap = (max_heap_candidate + page_size - 1) & ~(size_t)(page_size - 1);
3377 if (max_heap != max_heap_candidate)
3378 sgen_env_var_error (MONO_GC_PARAMS_NAME, "Rounding up.", "`max-heap-size` size must be a multiple of %d.", page_size);
3380 sgen_env_var_error (MONO_GC_PARAMS_NAME, NULL, "`max-heap-size` must be an integer.");
3384 if (g_str_has_prefix (opt, "soft-heap-limit=")) {
3385 opt = strchr (opt, '=') + 1;
3386 if (*opt && mono_gc_parse_environment_string_extract_number (opt, &soft_limit)) {
3387 if (soft_limit <= 0) {
3388 sgen_env_var_error (MONO_GC_PARAMS_NAME, NULL, "`soft-heap-limit` must be positive.");
3392 sgen_env_var_error (MONO_GC_PARAMS_NAME, NULL, "`soft-heap-limit` must be an integer.");
3396 if (g_str_has_prefix (opt, "nursery-size=")) {
3398 opt = strchr (opt, '=') + 1;
3399 if (*opt && mono_gc_parse_environment_string_extract_number (opt, &val)) {
3400 if ((val & (val - 1))) {
3401 sgen_env_var_error (MONO_GC_PARAMS_NAME, "Using default value.", "`nursery-size` must be a power of two.");
3405 if (val < SGEN_MAX_NURSERY_WASTE) {
3406 sgen_env_var_error (MONO_GC_PARAMS_NAME, "Using default value.",
3407 "`nursery-size` must be at least %d bytes.", SGEN_MAX_NURSERY_WASTE);
3411 min_nursery_size = max_nursery_size = val;
3412 dynamic_nursery = FALSE;
3414 sgen_env_var_error (MONO_GC_PARAMS_NAME, "Using default value.", "`nursery-size` must be an integer.");
3419 if (g_str_has_prefix (opt, "save-target-ratio=")) {
3421 opt = strchr (opt, '=') + 1;
3422 if (parse_double_in_interval (MONO_GC_PARAMS_NAME, "save-target-ratio", opt,
3423 SGEN_MIN_SAVE_TARGET_RATIO, SGEN_MAX_SAVE_TARGET_RATIO, &val)) {
3428 if (g_str_has_prefix (opt, "default-allowance-ratio=")) {
3430 opt = strchr (opt, '=') + 1;
3431 if (parse_double_in_interval (MONO_GC_PARAMS_NAME, "default-allowance-ratio", opt,
3432 SGEN_MIN_ALLOWANCE_NURSERY_SIZE_RATIO, SGEN_MAX_ALLOWANCE_NURSERY_SIZE_RATIO, &val)) {
3433 allowance_ratio = val;
3438 if (!strcmp (opt, "cementing")) {
3439 cement_enabled = TRUE;
3442 if (!strcmp (opt, "no-cementing")) {
3443 cement_enabled = FALSE;
3447 if (!strcmp (opt, "precleaning")) {
3448 precleaning_enabled = TRUE;
3451 if (!strcmp (opt, "no-precleaning")) {
3452 precleaning_enabled = FALSE;
3456 if (!strcmp (opt, "dynamic-nursery")) {
3457 if (sgen_minor_collector.is_split)
3458 sgen_env_var_error (MONO_GC_PARAMS_NAME, "Using default value.",
3459 "dynamic-nursery not supported with split-nursery.");
3461 dynamic_nursery = TRUE;
3464 if (!strcmp (opt, "no-dynamic-nursery")) {
3465 dynamic_nursery = FALSE;
3469 if (major_collector.handle_gc_param && major_collector.handle_gc_param (opt))
3472 if (sgen_minor_collector.handle_gc_param && sgen_minor_collector.handle_gc_param (opt))
3475 if (sgen_client_handle_gc_param (opt))
3478 sgen_env_var_error (MONO_GC_PARAMS_NAME, "Ignoring.", "Unknown option `%s`.", opt);
3483 fprintf (stderr, "\n%s must be a comma-delimited list of one or more of the following:\n", MONO_GC_PARAMS_NAME);
3484 fprintf (stderr, " max-heap-size=N (where N is an integer, possibly with a k, m or a g suffix)\n");
3485 fprintf (stderr, " soft-heap-limit=n (where N is an integer, possibly with a k, m or a g suffix)\n");
3486 fprintf (stderr, " mode=MODE (where MODE is 'balanced', 'throughput' or 'pause[:N]' and N is maximum pause in milliseconds)\n");
3487 fprintf (stderr, " nursery-size=N (where N is an integer, possibly with a k, m or a g suffix)\n");
3488 fprintf (stderr, " major=COLLECTOR (where COLLECTOR is `marksweep', `marksweep-conc', `marksweep-par')\n");
3489 fprintf (stderr, " minor=COLLECTOR (where COLLECTOR is `simple' or `split')\n");
3490 fprintf (stderr, " wbarrier=WBARRIER (where WBARRIER is `remset' or `cardtable')\n");
3491 fprintf (stderr, " [no-]cementing\n");
3492 fprintf (stderr, " [no-]dynamic-nursery\n");
3493 if (major_collector.print_gc_param_usage)
3494 major_collector.print_gc_param_usage ();
3495 if (sgen_minor_collector.print_gc_param_usage)
3496 sgen_minor_collector.print_gc_param_usage ();
3497 sgen_client_print_gc_params_usage ();
3498 fprintf (stderr, " Experimental options:\n");
3499 fprintf (stderr, " save-target-ratio=R (where R must be between %.2f - %.2f).\n", SGEN_MIN_SAVE_TARGET_RATIO, SGEN_MAX_SAVE_TARGET_RATIO);
3500 fprintf (stderr, " default-allowance-ratio=R (where R must be between %.2f - %.2f).\n", SGEN_MIN_ALLOWANCE_NURSERY_SIZE_RATIO, SGEN_MAX_ALLOWANCE_NURSERY_SIZE_RATIO);
3501 fprintf (stderr, "\n");
3503 usage_printed = TRUE;
3509 g_free (params_opts);
3511 alloc_nursery (dynamic_nursery, min_nursery_size, max_nursery_size);
3513 sgen_pinning_init ();
3514 sgen_cement_init (cement_enabled);
3516 if ((env = g_getenv (MONO_GC_DEBUG_NAME)) || gc_debug_options) {
3517 debug_opts = g_strdup_printf ("%s,%s", gc_debug_options ? gc_debug_options : "", env ? env : "");
3522 gboolean usage_printed = FALSE;
3524 opts = g_strsplit (debug_opts, ",", -1);
3525 for (ptr = opts; ptr && *ptr; ptr ++) {
3527 if (!strcmp (opt, ""))
3529 if (opt [0] >= '0' && opt [0] <= '9') {
3530 gc_debug_level = atoi (opt);
3535 char *rf = g_strdup_printf ("%s.%d", opt, mono_process_current_pid ());
3536 gc_debug_file = fopen (rf, "wb");
3538 gc_debug_file = stderr;
3541 } else if (!strcmp (opt, "print-allowance")) {
3542 debug_print_allowance = TRUE;
3543 } else if (!strcmp (opt, "print-pinning")) {
3544 sgen_pin_stats_enable ();
3545 } else if (!strcmp (opt, "verify-before-allocs")) {
3546 verify_before_allocs = 1;
3547 has_per_allocation_action = TRUE;
3548 } else if (g_str_has_prefix (opt, "max-valloc-size=")) {
3549 size_t max_valloc_size;
3550 char *arg = strchr (opt, '=') + 1;
3551 if (*opt && mono_gc_parse_environment_string_extract_number (arg, &max_valloc_size)) {
3552 mono_valloc_set_limit (max_valloc_size);
3554 sgen_env_var_error (MONO_GC_DEBUG_NAME, NULL, "`max-valloc-size` must be an integer.");
3557 } else if (g_str_has_prefix (opt, "verify-before-allocs=")) {
3558 char *arg = strchr (opt, '=') + 1;
3559 verify_before_allocs = atoi (arg);
3560 has_per_allocation_action = TRUE;
3561 } else if (!strcmp (opt, "collect-before-allocs")) {
3562 collect_before_allocs = 1;
3563 has_per_allocation_action = TRUE;
3564 } else if (g_str_has_prefix (opt, "collect-before-allocs=")) {
3565 char *arg = strchr (opt, '=') + 1;
3566 has_per_allocation_action = TRUE;
3567 collect_before_allocs = atoi (arg);
3568 } else if (!strcmp (opt, "verify-before-collections")) {
3569 whole_heap_check_before_collection = TRUE;
3570 } else if (!strcmp (opt, "check-remset-consistency")) {
3571 remset_consistency_checks = TRUE;
3572 nursery_clear_policy = CLEAR_AT_GC;
3573 } else if (!strcmp (opt, "mod-union-consistency-check")) {
3574 if (!major_collector.is_concurrent) {
3575 sgen_env_var_error (MONO_GC_DEBUG_NAME, "Ignoring.", "`mod-union-consistency-check` only works with concurrent major collector.");
3578 mod_union_consistency_check = TRUE;
3579 } else if (!strcmp (opt, "check-mark-bits")) {
3580 check_mark_bits_after_major_collection = TRUE;
3581 } else if (!strcmp (opt, "check-nursery-pinned")) {
3582 check_nursery_objects_pinned = TRUE;
3583 } else if (!strcmp (opt, "clear-at-gc")) {
3584 nursery_clear_policy = CLEAR_AT_GC;
3585 } else if (!strcmp (opt, "clear-nursery-at-gc")) {
3586 nursery_clear_policy = CLEAR_AT_GC;
3587 } else if (!strcmp (opt, "clear-at-tlab-creation")) {
3588 nursery_clear_policy = CLEAR_AT_TLAB_CREATION;
3589 } else if (!strcmp (opt, "debug-clear-at-tlab-creation")) {
3590 nursery_clear_policy = CLEAR_AT_TLAB_CREATION_DEBUG;
3591 } else if (!strcmp (opt, "check-scan-starts")) {
3592 do_scan_starts_check = TRUE;
3593 } else if (!strcmp (opt, "verify-nursery-at-minor-gc")) {
3594 do_verify_nursery = TRUE;
3595 } else if (!strcmp (opt, "check-concurrent")) {
3596 if (!major_collector.is_concurrent) {
3597 sgen_env_var_error (MONO_GC_DEBUG_NAME, "Ignoring.", "`check-concurrent` only works with concurrent major collectors.");
3600 nursery_clear_policy = CLEAR_AT_GC;
3601 do_concurrent_checks = TRUE;
3602 } else if (!strcmp (opt, "dump-nursery-at-minor-gc")) {
3603 do_dump_nursery_content = TRUE;
3604 } else if (!strcmp (opt, "disable-minor")) {
3605 disable_minor_collections = TRUE;
3606 } else if (!strcmp (opt, "disable-major")) {
3607 disable_major_collections = TRUE;
3608 } else if (g_str_has_prefix (opt, "heap-dump=")) {
3609 char *filename = strchr (opt, '=') + 1;
3610 nursery_clear_policy = CLEAR_AT_GC;
3611 sgen_debug_enable_heap_dump (filename);
3612 } else if (g_str_has_prefix (opt, "binary-protocol=")) {
3613 char *filename = strchr (opt, '=') + 1;
3614 char *colon = strrchr (filename, ':');
3617 if (!mono_gc_parse_environment_string_extract_number (colon + 1, &limit)) {
3618 sgen_env_var_error (MONO_GC_DEBUG_NAME, "Ignoring limit.", "Binary protocol file size limit must be an integer.");
3623 binary_protocol_init (filename, (long long)limit);
3624 } else if (!strcmp (opt, "nursery-canaries")) {
3625 do_verify_nursery = TRUE;
3626 enable_nursery_canaries = TRUE;
3627 } else if (!sgen_client_handle_gc_debug (opt)) {
3628 sgen_env_var_error (MONO_GC_DEBUG_NAME, "Ignoring.", "Unknown option `%s`.", opt);
3633 fprintf (stderr, "\n%s must be of the format [<l>[:<filename>]|<option>]+ where <l> is a debug level 0-9.\n", MONO_GC_DEBUG_NAME);
3634 fprintf (stderr, "Valid <option>s are:\n");
3635 fprintf (stderr, " collect-before-allocs[=<n>]\n");
3636 fprintf (stderr, " verify-before-allocs[=<n>]\n");
3637 fprintf (stderr, " max-valloc-size=N (where N is an integer, possibly with a k, m or a g suffix)\n");
3638 fprintf (stderr, " check-remset-consistency\n");
3639 fprintf (stderr, " check-mark-bits\n");
3640 fprintf (stderr, " check-nursery-pinned\n");
3641 fprintf (stderr, " verify-before-collections\n");
3642 fprintf (stderr, " verify-nursery-at-minor-gc\n");
3643 fprintf (stderr, " dump-nursery-at-minor-gc\n");
3644 fprintf (stderr, " disable-minor\n");
3645 fprintf (stderr, " disable-major\n");
3646 fprintf (stderr, " check-concurrent\n");
3647 fprintf (stderr, " clear-[nursery-]at-gc\n");
3648 fprintf (stderr, " clear-at-tlab-creation\n");
3649 fprintf (stderr, " debug-clear-at-tlab-creation\n");
3650 fprintf (stderr, " check-scan-starts\n");
3651 fprintf (stderr, " print-allowance\n");
3652 fprintf (stderr, " print-pinning\n");
3653 fprintf (stderr, " heap-dump=<filename>\n");
3654 fprintf (stderr, " binary-protocol=<filename>[:<file-size-limit>]\n");
3655 fprintf (stderr, " nursery-canaries\n");
3656 sgen_client_print_gc_debug_usage ();
3657 fprintf (stderr, "\n");
3659 usage_printed = TRUE;
3666 g_free (debug_opts);
3668 if (check_mark_bits_after_major_collection)
3669 nursery_clear_policy = CLEAR_AT_GC;
3671 if (major_collector.post_param_init)
3672 major_collector.post_param_init (&major_collector);
3674 sgen_thread_pool_start ();
3676 sgen_memgov_init (max_heap, soft_limit, debug_print_allowance, allowance_ratio, save_target);
3678 memset (&remset, 0, sizeof (remset));
3680 sgen_card_table_init (&remset);
3682 sgen_register_root (NULL, 0, sgen_make_user_root_descriptor (sgen_mark_normal_gc_handles), ROOT_TYPE_NORMAL, MONO_ROOT_SOURCE_GC_HANDLE, "normal gc handles");
3686 sgen_init_bridge ();
3690 sgen_gc_initialized ()
3692 return gc_initialized > 0;
3696 sgen_get_nursery_clear_policy (void)
3698 return nursery_clear_policy;
3704 mono_coop_mutex_lock (&gc_mutex);
3708 sgen_gc_unlock (void)
3710 mono_coop_mutex_unlock (&gc_mutex);
3714 sgen_major_collector_iterate_live_block_ranges (sgen_cardtable_block_callback callback)
3716 major_collector.iterate_live_block_ranges (callback);
3720 sgen_major_collector_iterate_block_ranges (sgen_cardtable_block_callback callback)
3722 major_collector.iterate_block_ranges (callback);
3726 sgen_get_major_collector (void)
3728 return &major_collector;
3732 sgen_get_minor_collector (void)
3734 return &sgen_minor_collector;
3738 sgen_get_remset (void)
3744 count_cards (long long *major_total, long long *major_marked, long long *los_total, long long *los_marked)
3746 sgen_get_major_collector ()->count_cards (major_total, major_marked);
3747 sgen_los_count_cards (los_total, los_marked);
3750 static gboolean world_is_stopped = FALSE;
3752 /* LOCKING: assumes the GC lock is held */
3754 sgen_stop_world (int generation)
3756 long long major_total = -1, major_marked = -1, los_total = -1, los_marked = -1;
3758 SGEN_ASSERT (0, !world_is_stopped, "Why are we stopping a stopped world?");
3760 binary_protocol_world_stopping (generation, sgen_timestamp (), (gpointer) (gsize) mono_native_thread_id_get ());
3762 sgen_client_stop_world (generation);
3764 world_is_stopped = TRUE;
3766 if (binary_protocol_is_heavy_enabled ())
3767 count_cards (&major_total, &major_marked, &los_total, &los_marked);
3768 binary_protocol_world_stopped (generation, sgen_timestamp (), major_total, major_marked, los_total, los_marked);
3771 /* LOCKING: assumes the GC lock is held */
3773 sgen_restart_world (int generation)
3775 long long major_total = -1, major_marked = -1, los_total = -1, los_marked = -1;
3778 SGEN_ASSERT (0, world_is_stopped, "Why are we restarting a running world?");
3780 if (binary_protocol_is_heavy_enabled ())
3781 count_cards (&major_total, &major_marked, &los_total, &los_marked);
3782 binary_protocol_world_restarting (generation, sgen_timestamp (), major_total, major_marked, los_total, los_marked);
3784 world_is_stopped = FALSE;
3786 sgen_client_restart_world (generation, &stw_time);
3788 binary_protocol_world_restarted (generation, sgen_timestamp ());
3790 if (sgen_client_bridge_need_processing ())
3791 sgen_client_bridge_processing_finish (generation);
3793 sgen_memgov_collection_end (generation, stw_time);
3797 sgen_is_world_stopped (void)
3799 return world_is_stopped;
3803 sgen_check_whole_heap_stw (void)
3805 sgen_stop_world (0);
3806 sgen_clear_nursery_fragments ();
3807 sgen_check_whole_heap (TRUE);
3808 sgen_restart_world (0);
3812 sgen_timestamp (void)
3814 SGEN_TV_DECLARE (timestamp);
3815 SGEN_TV_GETTIME (timestamp);
3816 return SGEN_TV_ELAPSED (sgen_init_timestamp, timestamp);
3819 #endif /* HAVE_SGEN_GC */