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;
240 #ifdef HEAVY_STATISTICS
241 guint64 stat_objects_alloced_degraded = 0;
242 guint64 stat_bytes_alloced_degraded = 0;
244 guint64 stat_copy_object_called_nursery = 0;
245 guint64 stat_objects_copied_nursery = 0;
246 guint64 stat_copy_object_called_major = 0;
247 guint64 stat_objects_copied_major = 0;
249 guint64 stat_scan_object_called_nursery = 0;
250 guint64 stat_scan_object_called_major = 0;
252 guint64 stat_slots_allocated_in_vain;
254 guint64 stat_nursery_copy_object_failed_from_space = 0;
255 guint64 stat_nursery_copy_object_failed_forwarded = 0;
256 guint64 stat_nursery_copy_object_failed_pinned = 0;
257 guint64 stat_nursery_copy_object_failed_to_space = 0;
259 static guint64 stat_wbarrier_add_to_global_remset = 0;
260 static guint64 stat_wbarrier_arrayref_copy = 0;
261 static guint64 stat_wbarrier_generic_store = 0;
262 static guint64 stat_wbarrier_generic_store_atomic = 0;
263 static guint64 stat_wbarrier_set_root = 0;
266 static guint64 stat_pinned_objects = 0;
268 static guint64 time_minor_pre_collection_fragment_clear = 0;
269 static guint64 time_minor_pinning = 0;
270 static guint64 time_minor_scan_remsets = 0;
271 static guint64 time_minor_scan_major_blocks = 0;
272 static guint64 time_minor_scan_los = 0;
273 static guint64 time_minor_scan_pinned = 0;
274 static guint64 time_minor_scan_roots = 0;
275 static guint64 time_minor_finish_gray_stack = 0;
276 static guint64 time_minor_fragment_creation = 0;
278 static guint64 time_major_pre_collection_fragment_clear = 0;
279 static guint64 time_major_pinning = 0;
280 static guint64 time_major_scan_pinned = 0;
281 static guint64 time_major_scan_roots = 0;
282 static guint64 time_major_scan_mod_union = 0;
283 static guint64 time_major_finish_gray_stack = 0;
284 static guint64 time_major_free_bigobjs = 0;
285 static guint64 time_major_los_sweep = 0;
286 static guint64 time_major_sweep = 0;
287 static guint64 time_major_fragment_creation = 0;
289 static guint64 time_max = 0;
291 static SGEN_TV_DECLARE (time_major_conc_collection_start);
292 static SGEN_TV_DECLARE (time_major_conc_collection_end);
294 int gc_debug_level = 0;
296 static char* gc_params_options;
297 static char* gc_debug_options;
301 mono_gc_flush_info (void)
303 fflush (gc_debug_file);
307 #define TV_DECLARE SGEN_TV_DECLARE
308 #define TV_GETTIME SGEN_TV_GETTIME
309 #define TV_ELAPSED SGEN_TV_ELAPSED
311 static SGEN_TV_DECLARE (sgen_init_timestamp);
313 NurseryClearPolicy nursery_clear_policy = CLEAR_AT_TLAB_CREATION;
315 #define object_is_forwarded SGEN_OBJECT_IS_FORWARDED
316 #define object_is_pinned SGEN_OBJECT_IS_PINNED
317 #define pin_object SGEN_PIN_OBJECT
319 #define ptr_in_nursery sgen_ptr_in_nursery
321 #define LOAD_VTABLE SGEN_LOAD_VTABLE
324 nursery_canaries_enabled (void)
326 return enable_nursery_canaries;
329 #define safe_object_get_size sgen_safe_object_get_size
331 #if defined(HAVE_CONC_GC_AS_DEFAULT)
332 /* Use concurrent major on deskstop platforms */
333 #define DEFAULT_MAJOR_INIT sgen_marksweep_conc_init
334 #define DEFAULT_MAJOR_NAME "marksweep-conc"
336 #define DEFAULT_MAJOR_INIT sgen_marksweep_init
337 #define DEFAULT_MAJOR_NAME "marksweep"
341 * ######################################################################
342 * ######## Global data.
343 * ######################################################################
345 MonoCoopMutex gc_mutex;
347 #define SCAN_START_SIZE SGEN_SCAN_START_SIZE
349 size_t degraded_mode = 0;
351 static mword bytes_pinned_from_failed_allocation = 0;
353 GCMemSection *nursery_section = NULL;
354 static volatile mword lowest_heap_address = ~(mword)0;
355 static volatile mword highest_heap_address = 0;
357 MonoCoopMutex sgen_interruption_mutex;
359 int current_collection_generation = -1;
360 volatile gboolean concurrent_collection_in_progress = FALSE;
362 /* objects that are ready to be finalized */
363 static SgenPointerQueue fin_ready_queue = SGEN_POINTER_QUEUE_INIT (INTERNAL_MEM_FINALIZE_READY);
364 static SgenPointerQueue critical_fin_queue = SGEN_POINTER_QUEUE_INIT (INTERNAL_MEM_FINALIZE_READY);
366 /* registered roots: the key to the hash is the root start address */
368 * Different kinds of roots are kept separate to speed up pin_from_roots () for example.
370 SgenHashTable roots_hash [ROOT_TYPE_NUM] = {
371 SGEN_HASH_TABLE_INIT (INTERNAL_MEM_ROOTS_TABLE, INTERNAL_MEM_ROOT_RECORD, sizeof (RootRecord), sgen_aligned_addr_hash, NULL),
372 SGEN_HASH_TABLE_INIT (INTERNAL_MEM_ROOTS_TABLE, INTERNAL_MEM_ROOT_RECORD, sizeof (RootRecord), sgen_aligned_addr_hash, NULL),
373 SGEN_HASH_TABLE_INIT (INTERNAL_MEM_ROOTS_TABLE, INTERNAL_MEM_ROOT_RECORD, sizeof (RootRecord), sgen_aligned_addr_hash, NULL)
375 static mword roots_size = 0; /* amount of memory in the root set */
377 /* The size of a TLAB */
378 /* The bigger the value, the less often we have to go to the slow path to allocate a new
379 * one, but the more space is wasted by threads not allocating much memory.
381 * FIXME: Make this self-tuning for each thread.
383 guint32 tlab_size = (1024 * 4);
385 #define MAX_SMALL_OBJ_SIZE SGEN_MAX_SMALL_OBJ_SIZE
387 #define ALLOC_ALIGN SGEN_ALLOC_ALIGN
389 #define ALIGN_UP SGEN_ALIGN_UP
391 #ifdef SGEN_DEBUG_INTERNAL_ALLOC
392 MonoNativeThreadId main_gc_thread = NULL;
395 /*Object was pinned during the current collection*/
396 static mword objects_pinned;
399 * ######################################################################
400 * ######## Macros and function declarations.
401 * ######################################################################
404 /* forward declarations */
405 static void scan_from_registered_roots (char *addr_start, char *addr_end, int root_type, ScanCopyContext ctx);
407 static void pin_from_roots (void *start_nursery, void *end_nursery, ScanCopyContext ctx);
408 static void finish_gray_stack (int generation, ScanCopyContext ctx);
411 SgenMajorCollector major_collector;
412 SgenMinorCollector sgen_minor_collector;
414 static SgenRememberedSet remset;
417 * The gray queue a worker job must use. If we're not parallel or
418 * concurrent, we use the main gray queue.
420 static SgenGrayQueue*
421 sgen_workers_get_job_gray_queue (WorkerData *worker_data, SgenGrayQueue *default_gray_queue)
424 return &worker_data->private_gray_queue;
425 SGEN_ASSERT (0, default_gray_queue, "Why don't we have a default gray queue when we're not running in a worker thread?");
426 return default_gray_queue;
430 gray_queue_redirect (SgenGrayQueue *queue)
432 sgen_workers_take_from_queue (queue);
436 sgen_scan_area_with_callback (char *start, char *end, IterateObjectCallbackFunc callback, void *data, gboolean allow_flags, gboolean fail_on_canaries)
438 while (start < end) {
442 if (!*(void**)start) {
443 start += sizeof (void*); /* should be ALLOC_ALIGN, really */
448 if (!(obj = (char *)SGEN_OBJECT_IS_FORWARDED (start)))
454 if (!sgen_client_object_is_array_fill ((GCObject*)obj)) {
455 CHECK_CANARY_FOR_OBJECT ((GCObject*)obj, fail_on_canaries);
456 size = ALIGN_UP (safe_object_get_size ((GCObject*)obj));
457 callback ((GCObject*)obj, size, data);
458 CANARIFY_SIZE (size);
460 size = ALIGN_UP (safe_object_get_size ((GCObject*)obj));
468 * sgen_add_to_global_remset:
470 * The global remset contains locations which point into newspace after
471 * a minor collection. This can happen if the objects they point to are pinned.
473 * LOCKING: If called from a parallel collector, the global remset
474 * lock must be held. For serial collectors that is not necessary.
477 sgen_add_to_global_remset (gpointer ptr, GCObject *obj)
479 SGEN_ASSERT (5, sgen_ptr_in_nursery (obj), "Target pointer of global remset must be in the nursery");
481 HEAVY_STAT (++stat_wbarrier_add_to_global_remset);
483 if (!major_collector.is_concurrent) {
484 SGEN_ASSERT (5, current_collection_generation != -1, "Global remsets can only be added during collections");
486 if (current_collection_generation == -1)
487 SGEN_ASSERT (5, sgen_concurrent_collection_in_progress (), "Global remsets outside of collection pauses can only be added by the concurrent collector");
490 if (!object_is_pinned (obj))
491 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");
492 else if (sgen_cement_lookup_or_register (obj))
495 remset.record_pointer (ptr);
497 sgen_pin_stats_register_global_remset (obj);
499 SGEN_LOG (8, "Adding global remset for %p", ptr);
500 binary_protocol_global_remset (ptr, obj, (gpointer)SGEN_LOAD_VTABLE (obj));
504 * sgen_drain_gray_stack:
506 * Scan objects in the gray stack until the stack is empty. This should be called
507 * frequently after each object is copied, to achieve better locality and cache
512 sgen_drain_gray_stack (ScanCopyContext ctx)
514 SGEN_ASSERT (0, ctx.ops->drain_gray_stack, "Why do we have a scan/copy context with a missing drain gray stack function?");
516 return ctx.ops->drain_gray_stack (ctx.queue);
520 * Addresses in the pin queue are already sorted. This function finds
521 * the object header for each address and pins the object. The
522 * addresses must be inside the nursery section. The (start of the)
523 * address array is overwritten with the addresses of the actually
524 * pinned objects. Return the number of pinned objects.
527 pin_objects_from_nursery_pin_queue (gboolean do_scan_objects, ScanCopyContext ctx)
529 GCMemSection *section = nursery_section;
530 void **start = sgen_pinning_get_entry (section->pin_queue_first_entry);
531 void **end = sgen_pinning_get_entry (section->pin_queue_last_entry);
532 void *start_nursery = section->data;
533 void *end_nursery = section->end_data;
538 void *pinning_front = start_nursery;
540 void **definitely_pinned = start;
541 ScanObjectFunc scan_func = ctx.ops->scan_object;
542 SgenGrayQueue *queue = ctx.queue;
544 sgen_nursery_allocator_prepare_for_pinning ();
546 while (start < end) {
547 GCObject *obj_to_pin = NULL;
548 size_t obj_to_pin_size = 0;
553 SGEN_ASSERT (0, addr >= start_nursery && addr < end_nursery, "Potential pinning address out of range");
554 SGEN_ASSERT (0, addr >= last, "Pin queue not sorted");
561 SGEN_LOG (5, "Considering pinning addr %p", addr);
562 /* We've already processed everything up to pinning_front. */
563 if (addr < pinning_front) {
569 * Find the closest scan start <= addr. We might search backward in the
570 * scan_starts array because entries might be NULL. In the worst case we
571 * start at start_nursery.
573 idx = ((char*)addr - (char*)section->data) / SCAN_START_SIZE;
574 SGEN_ASSERT (0, idx < section->num_scan_start, "Scan start index out of range");
575 search_start = (void*)section->scan_starts [idx];
576 if (!search_start || search_start > addr) {
579 search_start = section->scan_starts [idx];
580 if (search_start && search_start <= addr)
583 if (!search_start || search_start > addr)
584 search_start = start_nursery;
588 * If the pinning front is closer than the scan start we found, start
589 * searching at the front.
591 if (search_start < pinning_front)
592 search_start = pinning_front;
595 * Now addr should be in an object a short distance from search_start.
597 * search_start must point to zeroed mem or point to an object.
600 size_t obj_size, canarified_obj_size;
603 if (!*(void**)search_start) {
604 search_start = (void*)ALIGN_UP ((mword)search_start + sizeof (gpointer));
605 /* The loop condition makes sure we don't overrun addr. */
609 canarified_obj_size = obj_size = ALIGN_UP (safe_object_get_size ((GCObject*)search_start));
612 * Filler arrays are marked by an invalid sync word. We don't
613 * consider them for pinning. They are not delimited by canaries,
616 if (!sgen_client_object_is_array_fill ((GCObject*)search_start)) {
617 CHECK_CANARY_FOR_OBJECT (search_start, TRUE);
618 CANARIFY_SIZE (canarified_obj_size);
620 if (addr >= search_start && (char*)addr < (char*)search_start + obj_size) {
621 /* This is the object we're looking for. */
622 obj_to_pin = (GCObject*)search_start;
623 obj_to_pin_size = canarified_obj_size;
628 /* Skip to the next object */
629 search_start = (void*)((char*)search_start + canarified_obj_size);
630 } while (search_start <= addr);
632 /* We've searched past the address we were looking for. */
634 pinning_front = search_start;
635 goto next_pin_queue_entry;
639 * We've found an object to pin. It might still be a dummy array, but we
640 * can advance the pinning front in any case.
642 pinning_front = (char*)obj_to_pin + obj_to_pin_size;
645 * If this is a dummy array marking the beginning of a nursery
646 * fragment, we don't pin it.
648 if (sgen_client_object_is_array_fill (obj_to_pin))
649 goto next_pin_queue_entry;
652 * Finally - pin the object!
654 desc = sgen_obj_get_descriptor_safe (obj_to_pin);
655 if (do_scan_objects) {
656 scan_func (obj_to_pin, desc, queue);
658 SGEN_LOG (4, "Pinned object %p, vtable %p (%s), count %d\n",
659 obj_to_pin, *(void**)obj_to_pin, sgen_client_vtable_get_name (SGEN_LOAD_VTABLE (obj_to_pin)), count);
660 binary_protocol_pin (obj_to_pin,
661 (gpointer)LOAD_VTABLE (obj_to_pin),
662 safe_object_get_size (obj_to_pin));
664 pin_object (obj_to_pin);
665 GRAY_OBJECT_ENQUEUE_SERIAL (queue, obj_to_pin, desc);
666 sgen_pin_stats_register_object (obj_to_pin, GENERATION_NURSERY);
667 definitely_pinned [count] = obj_to_pin;
670 if (concurrent_collection_in_progress)
671 sgen_pinning_register_pinned_in_nursery (obj_to_pin);
673 next_pin_queue_entry:
677 sgen_client_nursery_objects_pinned (definitely_pinned, count);
678 stat_pinned_objects += count;
683 pin_objects_in_nursery (gboolean do_scan_objects, ScanCopyContext ctx)
687 if (nursery_section->pin_queue_first_entry == nursery_section->pin_queue_last_entry)
690 reduced_to = pin_objects_from_nursery_pin_queue (do_scan_objects, ctx);
691 nursery_section->pin_queue_last_entry = nursery_section->pin_queue_first_entry + reduced_to;
695 * This function is only ever called (via `collector_pin_object()` in `sgen-copy-object.h`)
696 * when we can't promote an object because we're out of memory.
699 sgen_pin_object (GCObject *object, SgenGrayQueue *queue)
701 SGEN_ASSERT (0, sgen_ptr_in_nursery (object), "We're only supposed to use this for pinning nursery objects when out of memory.");
704 * All pinned objects are assumed to have been staged, so we need to stage as well.
705 * Also, the count of staged objects shows that "late pinning" happened.
707 sgen_pin_stage_ptr (object);
709 SGEN_PIN_OBJECT (object);
710 binary_protocol_pin (object, (gpointer)LOAD_VTABLE (object), safe_object_get_size (object));
713 sgen_pin_stats_register_object (object, GENERATION_NURSERY);
715 GRAY_OBJECT_ENQUEUE_SERIAL (queue, object, sgen_obj_get_descriptor_safe (object));
718 /* Sort the addresses in array in increasing order.
719 * Done using a by-the book heap sort. Which has decent and stable performance, is pretty cache efficient.
722 sgen_sort_addresses (void **array, size_t size)
727 for (i = 1; i < size; ++i) {
730 size_t parent = (child - 1) / 2;
732 if (array [parent] >= array [child])
735 tmp = array [parent];
736 array [parent] = array [child];
743 for (i = size - 1; i > 0; --i) {
746 array [i] = array [0];
752 while (root * 2 + 1 <= end) {
753 size_t child = root * 2 + 1;
755 if (child < end && array [child] < array [child + 1])
757 if (array [root] >= array [child])
761 array [root] = array [child];
770 * Scan the memory between start and end and queue values which could be pointers
771 * to the area between start_nursery and end_nursery for later consideration.
772 * Typically used for thread stacks.
775 sgen_conservatively_pin_objects_from (void **start, void **end, void *start_nursery, void *end_nursery, int pin_type)
779 SGEN_ASSERT (0, ((mword)start & (SIZEOF_VOID_P - 1)) == 0, "Why are we scanning for references in unaligned memory ?");
781 #if defined(VALGRIND_MAKE_MEM_DEFINED_IF_ADDRESSABLE) && !defined(_WIN64)
782 VALGRIND_MAKE_MEM_DEFINED_IF_ADDRESSABLE (start, (char*)end - (char*)start);
785 while (start < end) {
787 * *start can point to the middle of an object
788 * note: should we handle pointing at the end of an object?
789 * pinning in C# code disallows pointing at the end of an object
790 * but there is some small chance that an optimizing C compiler
791 * may keep the only reference to an object by pointing
792 * at the end of it. We ignore this small chance for now.
793 * Pointers to the end of an object are indistinguishable
794 * from pointers to the start of the next object in memory
795 * so if we allow that we'd need to pin two objects...
796 * We queue the pointer in an array, the
797 * array will then be sorted and uniqued. This way
798 * we can coalesce several pinning pointers and it should
799 * be faster since we'd do a memory scan with increasing
800 * addresses. Note: we can align the address to the allocation
801 * alignment, so the unique process is more effective.
803 mword addr = (mword)*start;
804 addr &= ~(ALLOC_ALIGN - 1);
805 if (addr >= (mword)start_nursery && addr < (mword)end_nursery) {
806 SGEN_LOG (6, "Pinning address %p from %p", (void*)addr, start);
807 sgen_pin_stage_ptr ((void*)addr);
808 binary_protocol_pin_stage (start, (void*)addr);
809 sgen_pin_stats_register_address ((char*)addr, pin_type);
815 SGEN_LOG (7, "found %d potential pinned heap pointers", count);
819 * The first thing we do in a collection is to identify pinned objects.
820 * This function considers all the areas of memory that need to be
821 * conservatively scanned.
824 pin_from_roots (void *start_nursery, void *end_nursery, ScanCopyContext ctx)
828 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);
829 /* objects pinned from the API are inside these roots */
830 SGEN_HASH_TABLE_FOREACH (&roots_hash [ROOT_TYPE_PINNED], void **, start_root, RootRecord *, root) {
831 SGEN_LOG (6, "Pinned roots %p-%p", start_root, root->end_root);
832 sgen_conservatively_pin_objects_from (start_root, (void**)root->end_root, start_nursery, end_nursery, PIN_TYPE_OTHER);
833 } SGEN_HASH_TABLE_FOREACH_END;
834 /* now deal with the thread stacks
835 * in the future we should be able to conservatively scan only:
836 * *) the cpu registers
837 * *) the unmanaged stack frames
838 * *) the _last_ managed stack frame
839 * *) pointers slots in managed frames
841 sgen_client_scan_thread_data (start_nursery, end_nursery, FALSE, ctx);
845 single_arg_user_copy_or_mark (GCObject **obj, void *gc_data)
847 ScanCopyContext *ctx = (ScanCopyContext *)gc_data;
848 ctx->ops->copy_or_mark_object (obj, ctx->queue);
852 * The memory area from start_root to end_root contains pointers to objects.
853 * Their position is precisely described by @desc (this means that the pointer
854 * can be either NULL or the pointer to the start of an object).
855 * This functions copies them to to_space updates them.
857 * This function is not thread-safe!
860 precisely_scan_objects_from (void** start_root, void** end_root, char* n_start, char *n_end, SgenDescriptor desc, ScanCopyContext ctx)
862 CopyOrMarkObjectFunc copy_func = ctx.ops->copy_or_mark_object;
863 ScanPtrFieldFunc scan_field_func = ctx.ops->scan_ptr_field;
864 SgenGrayQueue *queue = ctx.queue;
866 switch (desc & ROOT_DESC_TYPE_MASK) {
867 case ROOT_DESC_BITMAP:
868 desc >>= ROOT_DESC_TYPE_SHIFT;
870 if ((desc & 1) && *start_root) {
871 copy_func ((GCObject**)start_root, queue);
872 SGEN_LOG (9, "Overwrote root at %p with %p", start_root, *start_root);
878 case ROOT_DESC_COMPLEX: {
879 gsize *bitmap_data = (gsize *)sgen_get_complex_descriptor_bitmap (desc);
880 gsize bwords = (*bitmap_data) - 1;
881 void **start_run = start_root;
883 while (bwords-- > 0) {
884 gsize bmap = *bitmap_data++;
885 void **objptr = start_run;
887 if ((bmap & 1) && *objptr) {
888 copy_func ((GCObject**)objptr, queue);
889 SGEN_LOG (9, "Overwrote root at %p with %p", objptr, *objptr);
894 start_run += GC_BITS_PER_WORD;
898 case ROOT_DESC_VECTOR: {
901 for (p = start_root; p < end_root; p++) {
903 scan_field_func (NULL, (GCObject**)p, queue);
907 case ROOT_DESC_USER: {
908 SgenUserRootMarkFunc marker = sgen_get_user_descriptor_func (desc);
909 marker (start_root, single_arg_user_copy_or_mark, &ctx);
912 case ROOT_DESC_RUN_LEN:
913 g_assert_not_reached ();
915 g_assert_not_reached ();
920 reset_heap_boundaries (void)
922 lowest_heap_address = ~(mword)0;
923 highest_heap_address = 0;
927 sgen_update_heap_boundaries (mword low, mword high)
932 old = lowest_heap_address;
935 } while (SGEN_CAS_PTR ((gpointer*)&lowest_heap_address, (gpointer)low, (gpointer)old) != (gpointer)old);
938 old = highest_heap_address;
941 } while (SGEN_CAS_PTR ((gpointer*)&highest_heap_address, (gpointer)high, (gpointer)old) != (gpointer)old);
945 * Allocate and setup the data structures needed to be able to allocate objects
946 * in the nursery. The nursery is stored in nursery_section.
949 alloc_nursery (gboolean dynamic, size_t min_size, size_t max_size)
956 min_size = SGEN_DEFAULT_NURSERY_MIN_SIZE;
958 max_size = SGEN_DEFAULT_NURSERY_MAX_SIZE;
960 SGEN_ASSERT (0, min_size == max_size, "We can't have nursery ranges for static configuration.");
962 min_size = max_size = SGEN_DEFAULT_NURSERY_SIZE;
965 SGEN_ASSERT (0, !nursery_section, "Why are we allocating the nursery twice?");
966 SGEN_LOG (2, "Allocating nursery size: %zu, initial %zu", max_size, min_size);
968 /* FIXME: handle OOM */
969 nursery_section = (GCMemSection *)sgen_alloc_internal (INTERNAL_MEM_SECTION);
971 /* If there isn't enough space even for the nursery we should simply abort. */
972 g_assert (sgen_memgov_try_alloc_space (max_size, SPACE_NURSERY));
975 * The nursery section range represents the memory section where objects
976 * can be found. This is used when iterating for objects in the nursery,
977 * pinning etc. sgen_nursery_max_size represents the total allocated space
978 * for the nursery. sgen_nursery_size represents the current size of the
979 * nursery and it is used for allocation limits, heuristics etc. The
980 * nursery section is not always identical to the current nursery size
981 * because it can contain pinned objects from when the nursery was larger.
983 * sgen_nursery_size <= nursery_section size <= sgen_nursery_max_size
985 data = (char *)major_collector.alloc_heap (max_size, max_size);
986 sgen_update_heap_boundaries ((mword)data, (mword)(data + max_size));
987 nursery_section->data = data;
988 nursery_section->end_data = data + min_size;
989 scan_starts = (max_size + SCAN_START_SIZE - 1) / SCAN_START_SIZE;
990 nursery_section->scan_starts = (char **)sgen_alloc_internal_dynamic (sizeof (char*) * scan_starts, INTERNAL_MEM_SCAN_STARTS, TRUE);
991 nursery_section->num_scan_start = scan_starts;
993 sgen_nursery_allocator_set_nursery_bounds (data, min_size, max_size);
997 mono_gc_get_logfile (void)
999 return gc_debug_file;
1003 mono_gc_params_set (const char* options)
1005 if (gc_params_options)
1006 g_free (gc_params_options);
1008 gc_params_options = g_strdup (options);
1012 mono_gc_debug_set (const char* options)
1014 if (gc_debug_options)
1015 g_free (gc_debug_options);
1017 gc_debug_options = g_strdup (options);
1021 scan_finalizer_entries (SgenPointerQueue *fin_queue, ScanCopyContext ctx)
1023 CopyOrMarkObjectFunc copy_func = ctx.ops->copy_or_mark_object;
1024 SgenGrayQueue *queue = ctx.queue;
1027 for (i = 0; i < fin_queue->next_slot; ++i) {
1028 GCObject *obj = (GCObject *)fin_queue->data [i];
1031 SGEN_LOG (5, "Scan of fin ready object: %p (%s)\n", obj, sgen_client_vtable_get_name (SGEN_LOAD_VTABLE (obj)));
1032 copy_func ((GCObject**)&fin_queue->data [i], queue);
1037 generation_name (int generation)
1039 switch (generation) {
1040 case GENERATION_NURSERY: return "nursery";
1041 case GENERATION_OLD: return "old";
1042 default: g_assert_not_reached ();
1047 sgen_generation_name (int generation)
1049 return generation_name (generation);
1053 finish_gray_stack (int generation, ScanCopyContext ctx)
1057 int done_with_ephemerons, ephemeron_rounds = 0;
1058 char *start_addr = generation == GENERATION_NURSERY ? sgen_get_nursery_start () : NULL;
1059 char *end_addr = generation == GENERATION_NURSERY ? sgen_get_nursery_end () : (char*)-1;
1060 SgenGrayQueue *queue = ctx.queue;
1062 binary_protocol_finish_gray_stack_start (sgen_timestamp (), generation);
1064 * We copied all the reachable objects. Now it's the time to copy
1065 * the objects that were not referenced by the roots, but by the copied objects.
1066 * we built a stack of objects pointed to by gray_start: they are
1067 * additional roots and we may add more items as we go.
1068 * We loop until gray_start == gray_objects which means no more objects have
1069 * been added. Note this is iterative: no recursion is involved.
1070 * We need to walk the LO list as well in search of marked big objects
1071 * (use a flag since this is needed only on major collections). We need to loop
1072 * here as well, so keep a counter of marked LO (increasing it in copy_object).
1073 * To achieve better cache locality and cache usage, we drain the gray stack
1074 * frequently, after each object is copied, and just finish the work here.
1076 sgen_drain_gray_stack (ctx);
1078 SGEN_LOG (2, "%s generation done", generation_name (generation));
1081 Reset bridge data, we might have lingering data from a previous collection if this is a major
1082 collection trigged by minor overflow.
1084 We must reset the gathered bridges since their original block might be evacuated due to major
1085 fragmentation in the meanwhile and the bridge code should not have to deal with that.
1087 if (sgen_client_bridge_need_processing ())
1088 sgen_client_bridge_reset_data ();
1091 * Mark all strong toggleref objects. This must be done before we walk ephemerons or finalizers
1092 * to ensure they see the full set of live objects.
1094 sgen_client_mark_togglerefs (start_addr, end_addr, ctx);
1097 * Walk the ephemeron tables marking all values with reachable keys. This must be completely done
1098 * before processing finalizable objects and non-tracking weak links to avoid finalizing/clearing
1099 * objects that are in fact reachable.
1101 done_with_ephemerons = 0;
1103 done_with_ephemerons = sgen_client_mark_ephemerons (ctx);
1104 sgen_drain_gray_stack (ctx);
1106 } while (!done_with_ephemerons);
1108 if (sgen_client_bridge_need_processing ()) {
1109 /*Make sure the gray stack is empty before we process bridge objects so we get liveness right*/
1110 sgen_drain_gray_stack (ctx);
1111 sgen_collect_bridge_objects (generation, ctx);
1112 if (generation == GENERATION_OLD)
1113 sgen_collect_bridge_objects (GENERATION_NURSERY, ctx);
1116 Do the first bridge step here, as the collector liveness state will become useless after that.
1118 An important optimization is to only proccess the possibly dead part of the object graph and skip
1119 over all live objects as we transitively know everything they point must be alive too.
1121 The above invariant is completely wrong if we let the gray queue be drained and mark/copy everything.
1123 This has the unfortunate side effect of making overflow collections perform the first step twice, but
1124 given we now have heuristics that perform major GC in anticipation of minor overflows this should not
1127 sgen_client_bridge_processing_stw_step ();
1131 Make sure we drain the gray stack before processing disappearing links and finalizers.
1132 If we don't make sure it is empty we might wrongly see a live object as dead.
1134 sgen_drain_gray_stack (ctx);
1137 We must clear weak links that don't track resurrection before processing object ready for
1138 finalization so they can be cleared before that.
1140 sgen_null_link_in_range (generation, ctx, FALSE);
1141 if (generation == GENERATION_OLD)
1142 sgen_null_link_in_range (GENERATION_NURSERY, ctx, FALSE);
1145 /* walk the finalization queue and move also the objects that need to be
1146 * finalized: use the finalized objects as new roots so the objects they depend
1147 * on are also not reclaimed. As with the roots above, only objects in the nursery
1148 * are marked/copied.
1150 sgen_finalize_in_range (generation, ctx);
1151 if (generation == GENERATION_OLD)
1152 sgen_finalize_in_range (GENERATION_NURSERY, ctx);
1153 /* drain the new stack that might have been created */
1154 SGEN_LOG (6, "Precise scan of gray area post fin");
1155 sgen_drain_gray_stack (ctx);
1158 * This must be done again after processing finalizable objects since CWL slots are cleared only after the key is finalized.
1160 done_with_ephemerons = 0;
1162 done_with_ephemerons = sgen_client_mark_ephemerons (ctx);
1163 sgen_drain_gray_stack (ctx);
1165 } while (!done_with_ephemerons);
1167 sgen_client_clear_unreachable_ephemerons (ctx);
1170 * We clear togglerefs only after all possible chances of revival are done.
1171 * This is semantically more inline with what users expect and it allows for
1172 * user finalizers to correctly interact with TR objects.
1174 sgen_client_clear_togglerefs (start_addr, end_addr, ctx);
1177 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);
1180 * handle disappearing links
1181 * Note we do this after checking the finalization queue because if an object
1182 * survives (at least long enough to be finalized) we don't clear the link.
1183 * This also deals with a possible issue with the monitor reclamation: with the Boehm
1184 * GC a finalized object my lose the monitor because it is cleared before the finalizer is
1187 g_assert (sgen_gray_object_queue_is_empty (queue));
1189 sgen_null_link_in_range (generation, ctx, TRUE);
1190 if (generation == GENERATION_OLD)
1191 sgen_null_link_in_range (GENERATION_NURSERY, ctx, TRUE);
1192 if (sgen_gray_object_queue_is_empty (queue))
1194 sgen_drain_gray_stack (ctx);
1197 g_assert (sgen_gray_object_queue_is_empty (queue));
1199 binary_protocol_finish_gray_stack_end (sgen_timestamp (), generation);
1203 sgen_check_section_scan_starts (GCMemSection *section)
1206 for (i = 0; i < section->num_scan_start; ++i) {
1207 if (section->scan_starts [i]) {
1208 mword size = safe_object_get_size ((GCObject*) section->scan_starts [i]);
1209 SGEN_ASSERT (0, size >= SGEN_CLIENT_MINIMUM_OBJECT_SIZE && size <= MAX_SMALL_OBJ_SIZE, "Weird object size at scan starts.");
1215 check_scan_starts (void)
1217 if (!do_scan_starts_check)
1219 sgen_check_section_scan_starts (nursery_section);
1220 major_collector.check_scan_starts ();
1224 scan_from_registered_roots (char *addr_start, char *addr_end, int root_type, ScanCopyContext ctx)
1228 SGEN_HASH_TABLE_FOREACH (&roots_hash [root_type], void **, start_root, RootRecord *, root) {
1229 SGEN_LOG (6, "Precise root scan %p-%p (desc: %p)", start_root, root->end_root, (void*)root->root_desc);
1230 precisely_scan_objects_from (start_root, (void**)root->end_root, addr_start, addr_end, root->root_desc, ctx);
1231 } SGEN_HASH_TABLE_FOREACH_END;
1237 static gboolean inited = FALSE;
1242 mono_counters_register ("Collection max time", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME | MONO_COUNTER_MONOTONIC, &time_max);
1244 mono_counters_register ("Minor fragment clear", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_minor_pre_collection_fragment_clear);
1245 mono_counters_register ("Minor pinning", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_minor_pinning);
1246 mono_counters_register ("Minor scan remembered set", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_minor_scan_remsets);
1247 mono_counters_register ("Minor scan major blocks", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_minor_scan_major_blocks);
1248 mono_counters_register ("Minor scan los", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_minor_scan_los);
1249 mono_counters_register ("Minor scan pinned", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_minor_scan_pinned);
1250 mono_counters_register ("Minor scan roots", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_minor_scan_roots);
1251 mono_counters_register ("Minor fragment creation", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_minor_fragment_creation);
1253 mono_counters_register ("Major fragment clear", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_major_pre_collection_fragment_clear);
1254 mono_counters_register ("Major pinning", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_major_pinning);
1255 mono_counters_register ("Major scan pinned", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_major_scan_pinned);
1256 mono_counters_register ("Major scan roots", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_major_scan_roots);
1257 mono_counters_register ("Major scan mod union", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_major_scan_mod_union);
1258 mono_counters_register ("Major finish gray stack", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_major_finish_gray_stack);
1259 mono_counters_register ("Major free big objects", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_major_free_bigobjs);
1260 mono_counters_register ("Major LOS sweep", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_major_los_sweep);
1261 mono_counters_register ("Major sweep", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_major_sweep);
1262 mono_counters_register ("Major fragment creation", MONO_COUNTER_GC | MONO_COUNTER_ULONG | MONO_COUNTER_TIME, &time_major_fragment_creation);
1264 mono_counters_register ("Number of pinned objects", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_pinned_objects);
1266 #ifdef HEAVY_STATISTICS
1267 mono_counters_register ("WBarrier remember pointer", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_wbarrier_add_to_global_remset);
1268 mono_counters_register ("WBarrier arrayref copy", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_wbarrier_arrayref_copy);
1269 mono_counters_register ("WBarrier generic store called", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_wbarrier_generic_store);
1270 mono_counters_register ("WBarrier generic atomic store called", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_wbarrier_generic_store_atomic);
1271 mono_counters_register ("WBarrier set root", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_wbarrier_set_root);
1273 mono_counters_register ("# objects allocated degraded", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_objects_alloced_degraded);
1274 mono_counters_register ("bytes allocated degraded", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_bytes_alloced_degraded);
1276 mono_counters_register ("# copy_object() called (nursery)", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_copy_object_called_nursery);
1277 mono_counters_register ("# objects copied (nursery)", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_objects_copied_nursery);
1278 mono_counters_register ("# copy_object() called (major)", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_copy_object_called_major);
1279 mono_counters_register ("# objects copied (major)", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_objects_copied_major);
1281 mono_counters_register ("# scan_object() called (nursery)", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_scan_object_called_nursery);
1282 mono_counters_register ("# scan_object() called (major)", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_scan_object_called_major);
1284 mono_counters_register ("Slots allocated in vain", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_slots_allocated_in_vain);
1286 mono_counters_register ("# nursery copy_object() failed from space", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_nursery_copy_object_failed_from_space);
1287 mono_counters_register ("# nursery copy_object() failed forwarded", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_nursery_copy_object_failed_forwarded);
1288 mono_counters_register ("# nursery copy_object() failed pinned", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_nursery_copy_object_failed_pinned);
1289 mono_counters_register ("# nursery copy_object() failed to space", MONO_COUNTER_GC | MONO_COUNTER_ULONG, &stat_nursery_copy_object_failed_to_space);
1291 sgen_nursery_allocator_init_heavy_stats ();
1299 reset_pinned_from_failed_allocation (void)
1301 bytes_pinned_from_failed_allocation = 0;
1305 sgen_set_pinned_from_failed_allocation (mword objsize)
1307 bytes_pinned_from_failed_allocation += objsize;
1311 sgen_collection_is_concurrent (void)
1313 switch (current_collection_generation) {
1314 case GENERATION_NURSERY:
1316 case GENERATION_OLD:
1317 return concurrent_collection_in_progress;
1319 g_error ("Invalid current generation %d", current_collection_generation);
1325 sgen_concurrent_collection_in_progress (void)
1327 return concurrent_collection_in_progress;
1331 SgenThreadPoolJob job;
1332 SgenObjectOperations *ops;
1333 SgenGrayQueue *gc_thread_gray_queue;
1341 static ScanCopyContext
1342 scan_copy_context_for_scan_job (void *worker_data_untyped, ScanJob *job)
1344 WorkerData *worker_data = (WorkerData *)worker_data_untyped;
1346 return CONTEXT_FROM_OBJECT_OPERATIONS (job->ops, sgen_workers_get_job_gray_queue (worker_data, job->gc_thread_gray_queue));
1354 } ScanFromRegisteredRootsJob;
1357 job_scan_from_registered_roots (void *worker_data_untyped, SgenThreadPoolJob *job)
1359 ScanFromRegisteredRootsJob *job_data = (ScanFromRegisteredRootsJob*)job;
1360 ScanCopyContext ctx = scan_copy_context_for_scan_job (worker_data_untyped, &job_data->scan_job);
1362 scan_from_registered_roots (job_data->heap_start, job_data->heap_end, job_data->root_type, ctx);
1369 } ScanThreadDataJob;
1372 job_scan_thread_data (void *worker_data_untyped, SgenThreadPoolJob *job)
1374 ScanThreadDataJob *job_data = (ScanThreadDataJob*)job;
1375 ScanCopyContext ctx = scan_copy_context_for_scan_job (worker_data_untyped, &job_data->scan_job);
1377 sgen_client_scan_thread_data (job_data->heap_start, job_data->heap_end, TRUE, ctx);
1382 SgenPointerQueue *queue;
1383 } ScanFinalizerEntriesJob;
1386 job_scan_finalizer_entries (void *worker_data_untyped, SgenThreadPoolJob *job)
1388 ScanFinalizerEntriesJob *job_data = (ScanFinalizerEntriesJob*)job;
1389 ScanCopyContext ctx = scan_copy_context_for_scan_job (worker_data_untyped, &job_data->scan_job);
1391 scan_finalizer_entries (job_data->queue, ctx);
1395 job_scan_wbroots (void *worker_data_untyped, SgenThreadPoolJob *job)
1397 ScanJob *job_data = (ScanJob*)job;
1398 ScanCopyContext ctx = scan_copy_context_for_scan_job (worker_data_untyped, job_data);
1400 sgen_wbroots_scan_card_table (ctx);
1404 job_scan_major_card_table (void *worker_data_untyped, SgenThreadPoolJob *job)
1406 SGEN_TV_DECLARE (atv);
1407 SGEN_TV_DECLARE (btv);
1408 ParallelScanJob *job_data = (ParallelScanJob*)job;
1409 ScanCopyContext ctx = scan_copy_context_for_scan_job (worker_data_untyped, (ScanJob*)job_data);
1411 SGEN_TV_GETTIME (atv);
1412 major_collector.scan_card_table (CARDTABLE_SCAN_GLOBAL, ctx, job_data->job_index, sgen_workers_get_job_split_count ());
1413 SGEN_TV_GETTIME (btv);
1414 time_minor_scan_major_blocks += SGEN_TV_ELAPSED (atv, btv);
1418 job_scan_los_card_table (void *worker_data_untyped, SgenThreadPoolJob *job)
1420 SGEN_TV_DECLARE (atv);
1421 SGEN_TV_DECLARE (btv);
1422 ParallelScanJob *job_data = (ParallelScanJob*)job;
1423 ScanCopyContext ctx = scan_copy_context_for_scan_job (worker_data_untyped, (ScanJob*)job_data);
1425 SGEN_TV_GETTIME (atv);
1426 sgen_los_scan_card_table (CARDTABLE_SCAN_GLOBAL, ctx, job_data->job_index, sgen_workers_get_job_split_count ());
1427 SGEN_TV_GETTIME (btv);
1428 time_minor_scan_los += SGEN_TV_ELAPSED (atv, btv);
1432 job_scan_major_mod_union_card_table (void *worker_data_untyped, SgenThreadPoolJob *job)
1434 ParallelScanJob *job_data = (ParallelScanJob*)job;
1435 ScanCopyContext ctx = scan_copy_context_for_scan_job (worker_data_untyped, (ScanJob*)job_data);
1437 g_assert (concurrent_collection_in_progress);
1438 major_collector.scan_card_table (CARDTABLE_SCAN_MOD_UNION, ctx, job_data->job_index, sgen_workers_get_job_split_count ());
1442 job_scan_los_mod_union_card_table (void *worker_data_untyped, SgenThreadPoolJob *job)
1444 ParallelScanJob *job_data = (ParallelScanJob*)job;
1445 ScanCopyContext ctx = scan_copy_context_for_scan_job (worker_data_untyped, (ScanJob*)job_data);
1447 g_assert (concurrent_collection_in_progress);
1448 sgen_los_scan_card_table (CARDTABLE_SCAN_MOD_UNION, ctx, job_data->job_index, sgen_workers_get_job_split_count ());
1452 job_major_mod_union_preclean (void *worker_data_untyped, SgenThreadPoolJob *job)
1454 ParallelScanJob *job_data = (ParallelScanJob*)job;
1455 ScanCopyContext ctx = scan_copy_context_for_scan_job (worker_data_untyped, (ScanJob*)job_data);
1457 g_assert (concurrent_collection_in_progress);
1459 major_collector.scan_card_table (CARDTABLE_SCAN_MOD_UNION_PRECLEAN, ctx, job_data->job_index, sgen_workers_get_job_split_count ());
1463 job_los_mod_union_preclean (void *worker_data_untyped, SgenThreadPoolJob *job)
1465 ParallelScanJob *job_data = (ParallelScanJob*)job;
1466 ScanCopyContext ctx = scan_copy_context_for_scan_job (worker_data_untyped, (ScanJob*)job_data);
1468 g_assert (concurrent_collection_in_progress);
1470 sgen_los_scan_card_table (CARDTABLE_SCAN_MOD_UNION_PRECLEAN, ctx, job_data->job_index, sgen_workers_get_job_split_count ());
1474 job_scan_last_pinned (void *worker_data_untyped, SgenThreadPoolJob *job)
1476 ScanJob *job_data = (ScanJob*)job;
1477 ScanCopyContext ctx = scan_copy_context_for_scan_job (worker_data_untyped, job_data);
1479 g_assert (concurrent_collection_in_progress);
1481 sgen_scan_pin_queue_objects (ctx);
1485 workers_finish_callback (void)
1487 ParallelScanJob *psj;
1489 int split_count = sgen_workers_get_job_split_count ();
1491 /* Mod union preclean jobs */
1492 for (i = 0; i < split_count; i++) {
1493 psj = (ParallelScanJob*)sgen_thread_pool_job_alloc ("preclean major mod union cardtable", job_major_mod_union_preclean, sizeof (ParallelScanJob));
1494 psj->scan_job.ops = sgen_workers_get_idle_func_object_ops ();
1495 psj->scan_job.gc_thread_gray_queue = NULL;
1497 sgen_workers_enqueue_job (&psj->scan_job.job, TRUE);
1500 for (i = 0; i < split_count; i++) {
1501 psj = (ParallelScanJob*)sgen_thread_pool_job_alloc ("preclean los mod union cardtable", job_los_mod_union_preclean, sizeof (ParallelScanJob));
1502 psj->scan_job.ops = sgen_workers_get_idle_func_object_ops ();
1503 psj->scan_job.gc_thread_gray_queue = NULL;
1505 sgen_workers_enqueue_job (&psj->scan_job.job, TRUE);
1508 sj = (ScanJob*)sgen_thread_pool_job_alloc ("scan last pinned", job_scan_last_pinned, sizeof (ScanJob));
1509 sj->ops = sgen_workers_get_idle_func_object_ops ();
1510 sj->gc_thread_gray_queue = NULL;
1511 sgen_workers_enqueue_job (&sj->job, TRUE);
1515 init_gray_queue (SgenGrayQueue *gc_thread_gray_queue, gboolean use_workers)
1518 sgen_workers_init_distribute_gray_queue ();
1519 sgen_gray_object_queue_init (gc_thread_gray_queue, NULL, TRUE);
1523 enqueue_scan_remembered_set_jobs (SgenGrayQueue *gc_thread_gray_queue, SgenObjectOperations *ops, gboolean enqueue)
1525 int i, split_count = sgen_workers_get_job_split_count ();
1528 sj = (ScanJob*)sgen_thread_pool_job_alloc ("scan wbroots", job_scan_wbroots, sizeof (ScanJob));
1530 sj->gc_thread_gray_queue = gc_thread_gray_queue;
1531 sgen_workers_enqueue_job (&sj->job, enqueue);
1533 for (i = 0; i < split_count; i++) {
1534 ParallelScanJob *psj;
1536 psj = (ParallelScanJob*)sgen_thread_pool_job_alloc ("scan major remsets", job_scan_major_card_table, sizeof (ParallelScanJob));
1537 psj->scan_job.ops = ops;
1538 psj->scan_job.gc_thread_gray_queue = gc_thread_gray_queue;
1540 sgen_workers_enqueue_job (&psj->scan_job.job, enqueue);
1542 psj = (ParallelScanJob*)sgen_thread_pool_job_alloc ("scan LOS remsets", job_scan_los_card_table, sizeof (ParallelScanJob));
1543 psj->scan_job.ops = ops;
1544 psj->scan_job.gc_thread_gray_queue = gc_thread_gray_queue;
1546 sgen_workers_enqueue_job (&psj->scan_job.job, enqueue);
1551 enqueue_scan_from_roots_jobs (SgenGrayQueue *gc_thread_gray_queue, char *heap_start, char *heap_end, SgenObjectOperations *ops, gboolean enqueue)
1553 ScanFromRegisteredRootsJob *scrrj;
1554 ScanThreadDataJob *stdj;
1555 ScanFinalizerEntriesJob *sfej;
1557 /* registered roots, this includes static fields */
1559 scrrj = (ScanFromRegisteredRootsJob*)sgen_thread_pool_job_alloc ("scan from registered roots normal", job_scan_from_registered_roots, sizeof (ScanFromRegisteredRootsJob));
1560 scrrj->scan_job.ops = ops;
1561 scrrj->scan_job.gc_thread_gray_queue = gc_thread_gray_queue;
1562 scrrj->heap_start = heap_start;
1563 scrrj->heap_end = heap_end;
1564 scrrj->root_type = ROOT_TYPE_NORMAL;
1565 sgen_workers_enqueue_job (&scrrj->scan_job.job, enqueue);
1567 if (current_collection_generation == GENERATION_OLD) {
1568 /* During minors we scan the cardtable for these roots instead */
1569 scrrj = (ScanFromRegisteredRootsJob*)sgen_thread_pool_job_alloc ("scan from registered roots wbarrier", job_scan_from_registered_roots, sizeof (ScanFromRegisteredRootsJob));
1570 scrrj->scan_job.ops = ops;
1571 scrrj->scan_job.gc_thread_gray_queue = gc_thread_gray_queue;
1572 scrrj->heap_start = heap_start;
1573 scrrj->heap_end = heap_end;
1574 scrrj->root_type = ROOT_TYPE_WBARRIER;
1575 sgen_workers_enqueue_job (&scrrj->scan_job.job, enqueue);
1580 stdj = (ScanThreadDataJob*)sgen_thread_pool_job_alloc ("scan thread data", job_scan_thread_data, sizeof (ScanThreadDataJob));
1581 stdj->scan_job.ops = ops;
1582 stdj->scan_job.gc_thread_gray_queue = gc_thread_gray_queue;
1583 stdj->heap_start = heap_start;
1584 stdj->heap_end = heap_end;
1585 sgen_workers_enqueue_job (&stdj->scan_job.job, enqueue);
1587 /* Scan the list of objects ready for finalization. */
1589 sfej = (ScanFinalizerEntriesJob*)sgen_thread_pool_job_alloc ("scan finalizer entries", job_scan_finalizer_entries, sizeof (ScanFinalizerEntriesJob));
1590 sfej->scan_job.ops = ops;
1591 sfej->scan_job.gc_thread_gray_queue = gc_thread_gray_queue;
1592 sfej->queue = &fin_ready_queue;
1593 sgen_workers_enqueue_job (&sfej->scan_job.job, enqueue);
1595 sfej = (ScanFinalizerEntriesJob*)sgen_thread_pool_job_alloc ("scan critical finalizer entries", job_scan_finalizer_entries, sizeof (ScanFinalizerEntriesJob));
1596 sfej->scan_job.ops = ops;
1597 sfej->scan_job.gc_thread_gray_queue = gc_thread_gray_queue;
1598 sfej->queue = &critical_fin_queue;
1599 sgen_workers_enqueue_job (&sfej->scan_job.job, enqueue);
1603 * Perform a nursery collection.
1605 * Return whether any objects were late-pinned due to being out of memory.
1608 collect_nursery (const char *reason, gboolean is_overflow, SgenGrayQueue *unpin_queue)
1610 gboolean needs_major, is_parallel = FALSE;
1611 mword fragment_total;
1612 SgenGrayQueue gc_thread_gray_queue;
1613 SgenObjectOperations *object_ops_nopar, *object_ops_par = NULL;
1614 ScanCopyContext ctx;
1617 SGEN_TV_DECLARE (last_minor_collection_start_tv);
1618 SGEN_TV_DECLARE (last_minor_collection_end_tv);
1620 if (disable_minor_collections)
1623 TV_GETTIME (last_minor_collection_start_tv);
1624 atv = last_minor_collection_start_tv;
1626 binary_protocol_collection_begin (gc_stats.minor_gc_count, GENERATION_NURSERY);
1628 if (sgen_concurrent_collection_in_progress ()) {
1629 /* FIXME Support parallel nursery collections with concurrent major */
1630 object_ops_nopar = &sgen_minor_collector.serial_ops_with_concurrent_major;
1632 object_ops_nopar = &sgen_minor_collector.serial_ops;
1633 if (sgen_minor_collector.is_parallel) {
1634 object_ops_par = &sgen_minor_collector.parallel_ops;
1639 if (do_verify_nursery || do_dump_nursery_content)
1640 sgen_debug_verify_nursery (do_dump_nursery_content);
1642 current_collection_generation = GENERATION_NURSERY;
1644 SGEN_ASSERT (0, !sgen_collection_is_concurrent (), "Why is the nursery collection concurrent?");
1646 reset_pinned_from_failed_allocation ();
1648 check_scan_starts ();
1650 sgen_nursery_alloc_prepare_for_minor ();
1655 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));
1657 /* world must be stopped already */
1659 time_minor_pre_collection_fragment_clear += TV_ELAPSED (atv, btv);
1661 sgen_client_pre_collection_checks ();
1663 major_collector.start_nursery_collection ();
1665 sgen_memgov_minor_collection_start ();
1667 init_gray_queue (&gc_thread_gray_queue, is_parallel);
1668 ctx = CONTEXT_FROM_OBJECT_OPERATIONS (object_ops_nopar, &gc_thread_gray_queue);
1670 gc_stats.minor_gc_count ++;
1672 sgen_process_fin_stage_entries ();
1674 /* pin from pinned handles */
1675 sgen_init_pinning ();
1676 sgen_client_binary_protocol_mark_start (GENERATION_NURSERY);
1677 pin_from_roots (nursery_section->data, nursery_section->end_data, ctx);
1678 /* pin cemented objects */
1679 sgen_pin_cemented_objects ();
1680 /* identify pinned objects */
1681 sgen_optimize_pin_queue ();
1682 sgen_pinning_setup_section (nursery_section);
1684 pin_objects_in_nursery (FALSE, ctx);
1685 sgen_pinning_trim_queue_to_section (nursery_section);
1687 if (remset_consistency_checks)
1688 sgen_check_remset_consistency ();
1690 if (whole_heap_check_before_collection) {
1691 sgen_clear_nursery_fragments ();
1692 sgen_check_whole_heap (FALSE);
1696 time_minor_pinning += TV_ELAPSED (btv, atv);
1697 SGEN_LOG (2, "Finding pinned pointers: %zd in %lld usecs", sgen_get_pinned_count (), (long long)TV_ELAPSED (btv, atv));
1698 SGEN_LOG (4, "Start scan with %zd pinned objects", sgen_get_pinned_count ());
1700 remset.start_scan_remsets ();
1702 enqueue_scan_remembered_set_jobs (&gc_thread_gray_queue, is_parallel ? object_ops_par : object_ops_nopar, is_parallel);
1704 /* we don't have complete write barrier yet, so we scan all the old generation sections */
1706 time_minor_scan_remsets += TV_ELAPSED (atv, btv);
1707 SGEN_LOG (2, "Old generation scan: %lld usecs", (long long)TV_ELAPSED (atv, btv));
1709 sgen_pin_stats_report ();
1711 /* FIXME: Why do we do this at this specific, seemingly random, point? */
1712 sgen_client_collecting_minor (&fin_ready_queue, &critical_fin_queue);
1715 time_minor_scan_pinned += TV_ELAPSED (btv, atv);
1717 enqueue_scan_from_roots_jobs (&gc_thread_gray_queue, nursery_section->data, nursery_section->end_data, is_parallel ? object_ops_par : object_ops_nopar, is_parallel);
1720 gray_queue_redirect (&gc_thread_gray_queue);
1721 sgen_workers_start_all_workers (object_ops_nopar, object_ops_par, NULL);
1722 sgen_workers_join ();
1726 time_minor_scan_roots += TV_ELAPSED (atv, btv);
1728 finish_gray_stack (GENERATION_NURSERY, ctx);
1731 time_minor_finish_gray_stack += TV_ELAPSED (btv, atv);
1732 sgen_client_binary_protocol_mark_end (GENERATION_NURSERY);
1734 if (objects_pinned) {
1735 sgen_optimize_pin_queue ();
1736 sgen_pinning_setup_section (nursery_section);
1740 * This is the latest point at which we can do this check, because
1741 * sgen_build_nursery_fragments() unpins nursery objects again.
1743 if (remset_consistency_checks)
1744 sgen_check_remset_consistency ();
1746 sgen_resize_nursery ();
1748 /* walk the pin_queue, build up the fragment list of free memory, unmark
1749 * pinned objects as we go, memzero() the empty fragments so they are ready for the
1752 sgen_client_binary_protocol_reclaim_start (GENERATION_NURSERY);
1753 fragment_total = sgen_build_nursery_fragments (nursery_section, unpin_queue);
1754 if (!fragment_total)
1757 /* Clear TLABs for all threads */
1758 sgen_clear_tlabs ();
1760 sgen_client_binary_protocol_reclaim_end (GENERATION_NURSERY);
1762 time_minor_fragment_creation += TV_ELAPSED (atv, btv);
1763 SGEN_LOG (2, "Fragment creation: %lld usecs, %lu bytes available", (long long)TV_ELAPSED (atv, btv), (unsigned long)fragment_total);
1765 if (remset_consistency_checks)
1766 sgen_check_major_refs ();
1768 major_collector.finish_nursery_collection ();
1770 TV_GETTIME (last_minor_collection_end_tv);
1771 gc_stats.minor_gc_time += TV_ELAPSED (last_minor_collection_start_tv, last_minor_collection_end_tv);
1773 sgen_debug_dump_heap ("minor", gc_stats.minor_gc_count - 1, NULL);
1775 /* prepare the pin queue for the next collection */
1776 sgen_finish_pinning ();
1777 if (sgen_have_pending_finalizers ()) {
1778 SGEN_LOG (4, "Finalizer-thread wakeup");
1779 sgen_client_finalize_notify ();
1781 sgen_pin_stats_reset ();
1782 /* clear cemented hash */
1783 sgen_cement_clear_below_threshold ();
1785 sgen_gray_object_queue_dispose (&gc_thread_gray_queue);
1787 check_scan_starts ();
1789 binary_protocol_flush_buffers (FALSE);
1791 sgen_memgov_minor_collection_end (reason, is_overflow);
1793 /*objects are late pinned because of lack of memory, so a major is a good call*/
1794 needs_major = objects_pinned > 0;
1795 current_collection_generation = -1;
1798 binary_protocol_collection_end (gc_stats.minor_gc_count - 1, GENERATION_NURSERY, 0, 0);
1800 if (check_nursery_objects_pinned && !sgen_minor_collector.is_split)
1801 sgen_check_nursery_objects_pinned (unpin_queue != NULL);
1807 COPY_OR_MARK_FROM_ROOTS_SERIAL,
1808 COPY_OR_MARK_FROM_ROOTS_START_CONCURRENT,
1809 COPY_OR_MARK_FROM_ROOTS_FINISH_CONCURRENT
1810 } CopyOrMarkFromRootsMode;
1813 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)
1818 /* FIXME: only use these values for the precise scan
1819 * note that to_space pointers should be excluded anyway...
1821 char *heap_start = NULL;
1822 char *heap_end = (char*)-1;
1823 ScanCopyContext ctx = CONTEXT_FROM_OBJECT_OPERATIONS (object_ops_nopar, gc_thread_gray_queue);
1824 gboolean concurrent = mode != COPY_OR_MARK_FROM_ROOTS_SERIAL;
1826 SGEN_ASSERT (0, !!concurrent == !!concurrent_collection_in_progress, "We've been called with the wrong mode.");
1828 if (mode == COPY_OR_MARK_FROM_ROOTS_START_CONCURRENT) {
1829 /*This cleans up unused fragments */
1830 sgen_nursery_allocator_prepare_for_pinning ();
1832 if (do_concurrent_checks)
1833 sgen_debug_check_nursery_is_clean ();
1835 /* The concurrent collector doesn't touch the nursery. */
1836 sgen_nursery_alloc_prepare_for_major ();
1841 /* Pinning depends on this */
1842 sgen_clear_nursery_fragments ();
1844 if (whole_heap_check_before_collection)
1845 sgen_check_whole_heap (TRUE);
1848 time_major_pre_collection_fragment_clear += TV_ELAPSED (atv, btv);
1852 sgen_client_pre_collection_checks ();
1854 if (mode != COPY_OR_MARK_FROM_ROOTS_START_CONCURRENT) {
1855 /* Remsets are not useful for a major collection */
1856 remset.clear_cards ();
1859 sgen_process_fin_stage_entries ();
1862 sgen_init_pinning ();
1863 SGEN_LOG (6, "Collecting pinned addresses");
1864 pin_from_roots ((void*)lowest_heap_address, (void*)highest_heap_address, ctx);
1865 if (mode == COPY_OR_MARK_FROM_ROOTS_FINISH_CONCURRENT) {
1866 /* Pin cemented objects that were forced */
1867 sgen_pin_cemented_objects ();
1869 sgen_optimize_pin_queue ();
1870 if (mode == COPY_OR_MARK_FROM_ROOTS_START_CONCURRENT) {
1872 * Cemented objects that are in the pinned list will be marked. When
1873 * marking concurrently we won't mark mod-union cards for these objects.
1874 * Instead they will remain cemented until the next major collection,
1875 * when we will recheck if they are still pinned in the roots.
1877 sgen_cement_force_pinned ();
1880 sgen_client_collecting_major_1 ();
1883 * pin_queue now contains all candidate pointers, sorted and
1884 * uniqued. We must do two passes now to figure out which
1885 * objects are pinned.
1887 * The first is to find within the pin_queue the area for each
1888 * section. This requires that the pin_queue be sorted. We
1889 * also process the LOS objects and pinned chunks here.
1891 * The second, destructive, pass is to reduce the section
1892 * areas to pointers to the actually pinned objects.
1894 SGEN_LOG (6, "Pinning from sections");
1895 /* first pass for the sections */
1896 sgen_find_section_pin_queue_start_end (nursery_section);
1897 /* identify possible pointers to the insize of large objects */
1898 SGEN_LOG (6, "Pinning from large objects");
1899 for (bigobj = los_object_list; bigobj; bigobj = bigobj->next) {
1901 if (sgen_find_optimized_pin_queue_area ((char*)bigobj->data, (char*)bigobj->data + sgen_los_object_size (bigobj), &dummy, &dummy)) {
1902 binary_protocol_pin (bigobj->data, (gpointer)LOAD_VTABLE (bigobj->data), safe_object_get_size (bigobj->data));
1904 if (sgen_los_object_is_pinned (bigobj->data)) {
1905 SGEN_ASSERT (0, mode == COPY_OR_MARK_FROM_ROOTS_FINISH_CONCURRENT, "LOS objects can only be pinned here after concurrent marking.");
1908 sgen_los_pin_object (bigobj->data);
1909 if (SGEN_OBJECT_HAS_REFERENCES (bigobj->data))
1910 GRAY_OBJECT_ENQUEUE_SERIAL (gc_thread_gray_queue, bigobj->data, sgen_obj_get_descriptor ((GCObject*)bigobj->data));
1911 sgen_pin_stats_register_object (bigobj->data, GENERATION_OLD);
1912 SGEN_LOG (6, "Marked large object %p (%s) size: %lu from roots", bigobj->data,
1913 sgen_client_vtable_get_name (SGEN_LOAD_VTABLE (bigobj->data)),
1914 (unsigned long)sgen_los_object_size (bigobj));
1916 sgen_client_pinned_los_object (bigobj->data);
1920 pin_objects_in_nursery (mode == COPY_OR_MARK_FROM_ROOTS_START_CONCURRENT, ctx);
1921 if (check_nursery_objects_pinned && !sgen_minor_collector.is_split)
1922 sgen_check_nursery_objects_pinned (mode != COPY_OR_MARK_FROM_ROOTS_START_CONCURRENT);
1924 major_collector.pin_objects (gc_thread_gray_queue);
1925 if (old_next_pin_slot)
1926 *old_next_pin_slot = sgen_get_pinned_count ();
1929 time_major_pinning += TV_ELAPSED (atv, btv);
1930 SGEN_LOG (2, "Finding pinned pointers: %zd in %lld usecs", sgen_get_pinned_count (), (long long)TV_ELAPSED (atv, btv));
1931 SGEN_LOG (4, "Start scan with %zd pinned objects", sgen_get_pinned_count ());
1933 major_collector.init_to_space ();
1935 SGEN_ASSERT (0, sgen_workers_all_done (), "Why are the workers not done when we start or finish a major collection?");
1936 if (mode == COPY_OR_MARK_FROM_ROOTS_FINISH_CONCURRENT) {
1937 if (object_ops_par != NULL)
1938 sgen_workers_set_num_active_workers (0);
1939 if (sgen_workers_have_idle_work ()) {
1941 * We force the finish of the worker with the new object ops context
1942 * which can also do copying. We need to have finished pinning.
1944 sgen_workers_start_all_workers (object_ops_nopar, object_ops_par, NULL);
1946 sgen_workers_join ();
1950 #ifdef SGEN_DEBUG_INTERNAL_ALLOC
1951 main_gc_thread = mono_native_thread_self ();
1954 sgen_client_collecting_major_2 ();
1957 time_major_scan_pinned += TV_ELAPSED (btv, atv);
1959 sgen_client_collecting_major_3 (&fin_ready_queue, &critical_fin_queue);
1961 enqueue_scan_from_roots_jobs (gc_thread_gray_queue, heap_start, heap_end, object_ops_nopar, FALSE);
1964 time_major_scan_roots += TV_ELAPSED (atv, btv);
1967 * We start the concurrent worker after pinning and after we scanned the roots
1968 * in order to make sure that the worker does not finish before handling all
1971 if (mode == COPY_OR_MARK_FROM_ROOTS_START_CONCURRENT) {
1972 sgen_workers_set_num_active_workers (1);
1973 gray_queue_redirect (gc_thread_gray_queue);
1974 if (precleaning_enabled) {
1975 sgen_workers_start_all_workers (object_ops_nopar, object_ops_par, workers_finish_callback);
1977 sgen_workers_start_all_workers (object_ops_nopar, object_ops_par, NULL);
1981 if (mode == COPY_OR_MARK_FROM_ROOTS_FINISH_CONCURRENT) {
1982 int i, split_count = sgen_workers_get_job_split_count ();
1983 gboolean parallel = object_ops_par != NULL;
1985 /* If we're not parallel we finish the collection on the gc thread */
1987 gray_queue_redirect (gc_thread_gray_queue);
1989 /* Mod union card table */
1990 for (i = 0; i < split_count; i++) {
1991 ParallelScanJob *psj;
1993 psj = (ParallelScanJob*)sgen_thread_pool_job_alloc ("scan mod union cardtable", job_scan_major_mod_union_card_table, sizeof (ParallelScanJob));
1994 psj->scan_job.ops = object_ops_par ? object_ops_par : object_ops_nopar;
1995 psj->scan_job.gc_thread_gray_queue = gc_thread_gray_queue;
1997 sgen_workers_enqueue_job (&psj->scan_job.job, parallel);
1999 psj = (ParallelScanJob*)sgen_thread_pool_job_alloc ("scan LOS mod union cardtable", job_scan_los_mod_union_card_table, sizeof (ParallelScanJob));
2000 psj->scan_job.ops = object_ops_par ? object_ops_par : object_ops_nopar;
2001 psj->scan_job.gc_thread_gray_queue = gc_thread_gray_queue;
2003 sgen_workers_enqueue_job (&psj->scan_job.job, parallel);
2008 * If we enqueue a job while workers are running we need to sgen_workers_ensure_awake
2009 * in order to make sure that we are running the idle func and draining all worker
2010 * gray queues. The operation of starting workers implies this, so we start them after
2011 * in order to avoid doing this operation twice. The workers will drain the main gray
2012 * stack that contained roots and pinned objects and also scan the mod union card
2015 sgen_workers_start_all_workers (object_ops_nopar, object_ops_par, NULL);
2016 sgen_workers_join ();
2020 sgen_pin_stats_report ();
2022 if (mode == COPY_OR_MARK_FROM_ROOTS_START_CONCURRENT) {
2023 sgen_finish_pinning ();
2025 sgen_pin_stats_reset ();
2027 if (do_concurrent_checks)
2028 sgen_debug_check_nursery_is_clean ();
2033 major_start_collection (SgenGrayQueue *gc_thread_gray_queue, const char *reason, gboolean concurrent, size_t *old_next_pin_slot)
2035 SgenObjectOperations *object_ops_nopar, *object_ops_par = NULL;
2037 binary_protocol_collection_begin (gc_stats.major_gc_count, GENERATION_OLD);
2039 current_collection_generation = GENERATION_OLD;
2041 sgen_workers_assert_gray_queue_is_empty ();
2044 sgen_cement_reset ();
2047 g_assert (major_collector.is_concurrent);
2048 concurrent_collection_in_progress = TRUE;
2050 object_ops_nopar = &major_collector.major_ops_concurrent_start;
2051 if (major_collector.is_parallel)
2052 object_ops_par = &major_collector.major_ops_conc_par_start;
2055 object_ops_nopar = &major_collector.major_ops_serial;
2058 reset_pinned_from_failed_allocation ();
2060 sgen_memgov_major_collection_start (concurrent, reason);
2062 //count_ref_nonref_objs ();
2063 //consistency_check ();
2065 check_scan_starts ();
2068 SGEN_LOG (1, "Start major collection %d", gc_stats.major_gc_count);
2069 gc_stats.major_gc_count ++;
2071 if (major_collector.start_major_collection)
2072 major_collector.start_major_collection ();
2074 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);
2078 major_finish_collection (SgenGrayQueue *gc_thread_gray_queue, const char *reason, gboolean is_overflow, size_t old_next_pin_slot, gboolean forced)
2080 ScannedObjectCounts counts;
2081 SgenObjectOperations *object_ops_nopar;
2082 mword fragment_total;
2088 if (concurrent_collection_in_progress) {
2089 SgenObjectOperations *object_ops_par = NULL;
2091 object_ops_nopar = &major_collector.major_ops_concurrent_finish;
2092 if (major_collector.is_parallel)
2093 object_ops_par = &major_collector.major_ops_conc_par_finish;
2095 major_copy_or_mark_from_roots (gc_thread_gray_queue, NULL, COPY_OR_MARK_FROM_ROOTS_FINISH_CONCURRENT, object_ops_nopar, object_ops_par);
2097 #ifdef SGEN_DEBUG_INTERNAL_ALLOC
2098 main_gc_thread = NULL;
2101 object_ops_nopar = &major_collector.major_ops_serial;
2104 sgen_workers_assert_gray_queue_is_empty ();
2106 finish_gray_stack (GENERATION_OLD, CONTEXT_FROM_OBJECT_OPERATIONS (object_ops_nopar, gc_thread_gray_queue));
2108 time_major_finish_gray_stack += TV_ELAPSED (btv, atv);
2110 SGEN_ASSERT (0, sgen_workers_all_done (), "Can't have workers working after joining");
2112 if (objects_pinned) {
2113 g_assert (!concurrent_collection_in_progress);
2116 * This is slow, but we just OOM'd.
2118 * See comment at `sgen_pin_queue_clear_discarded_entries` for how the pin
2119 * queue is laid out at this point.
2121 sgen_pin_queue_clear_discarded_entries (nursery_section, old_next_pin_slot);
2123 * We need to reestablish all pinned nursery objects in the pin queue
2124 * because they're needed for fragment creation. Unpinning happens by
2125 * walking the whole queue, so it's not necessary to reestablish where major
2126 * heap block pins are - all we care is that they're still in there
2129 sgen_optimize_pin_queue ();
2130 sgen_find_section_pin_queue_start_end (nursery_section);
2134 reset_heap_boundaries ();
2135 sgen_update_heap_boundaries ((mword)sgen_get_nursery_start (), (mword)sgen_get_nursery_end ());
2137 /* walk the pin_queue, build up the fragment list of free memory, unmark
2138 * pinned objects as we go, memzero() the empty fragments so they are ready for the
2141 fragment_total = sgen_build_nursery_fragments (nursery_section, NULL);
2142 if (!fragment_total)
2144 SGEN_LOG (4, "Free space in nursery after major %ld", (long)fragment_total);
2146 if (do_concurrent_checks && concurrent_collection_in_progress)
2147 sgen_debug_check_nursery_is_clean ();
2149 /* prepare the pin queue for the next collection */
2150 sgen_finish_pinning ();
2152 /* Clear TLABs for all threads */
2153 sgen_clear_tlabs ();
2155 sgen_pin_stats_reset ();
2157 sgen_cement_clear_below_threshold ();
2159 if (check_mark_bits_after_major_collection)
2160 sgen_check_heap_marked (concurrent_collection_in_progress);
2163 time_major_fragment_creation += TV_ELAPSED (atv, btv);
2165 binary_protocol_sweep_begin (GENERATION_OLD, !major_collector.sweeps_lazily);
2166 sgen_memgov_major_pre_sweep ();
2169 time_major_free_bigobjs += TV_ELAPSED (btv, atv);
2174 time_major_los_sweep += TV_ELAPSED (atv, btv);
2176 major_collector.sweep ();
2178 binary_protocol_sweep_end (GENERATION_OLD, !major_collector.sweeps_lazily);
2181 time_major_sweep += TV_ELAPSED (btv, atv);
2183 sgen_debug_dump_heap ("major", gc_stats.major_gc_count - 1, reason);
2185 if (sgen_have_pending_finalizers ()) {
2186 SGEN_LOG (4, "Finalizer-thread wakeup");
2187 sgen_client_finalize_notify ();
2190 sgen_memgov_major_collection_end (forced, concurrent_collection_in_progress, reason, is_overflow);
2191 current_collection_generation = -1;
2193 memset (&counts, 0, sizeof (ScannedObjectCounts));
2194 major_collector.finish_major_collection (&counts);
2196 sgen_workers_assert_gray_queue_is_empty ();
2198 SGEN_ASSERT (0, sgen_workers_all_done (), "Can't have workers working after major collection has finished");
2199 if (concurrent_collection_in_progress)
2200 concurrent_collection_in_progress = FALSE;
2202 check_scan_starts ();
2204 binary_protocol_flush_buffers (FALSE);
2206 //consistency_check ();
2208 binary_protocol_collection_end (gc_stats.major_gc_count - 1, GENERATION_OLD, counts.num_scanned_objects, counts.num_unique_scanned_objects);
2212 major_do_collection (const char *reason, gboolean is_overflow, gboolean forced)
2214 TV_DECLARE (time_start);
2215 TV_DECLARE (time_end);
2216 size_t old_next_pin_slot;
2217 SgenGrayQueue gc_thread_gray_queue;
2219 if (disable_major_collections)
2222 if (major_collector.get_and_reset_num_major_objects_marked) {
2223 long long num_marked = major_collector.get_and_reset_num_major_objects_marked ();
2224 g_assert (!num_marked);
2227 /* world must be stopped already */
2228 TV_GETTIME (time_start);
2230 init_gray_queue (&gc_thread_gray_queue, FALSE);
2231 major_start_collection (&gc_thread_gray_queue, reason, FALSE, &old_next_pin_slot);
2232 major_finish_collection (&gc_thread_gray_queue, reason, is_overflow, old_next_pin_slot, forced);
2233 sgen_gray_object_queue_dispose (&gc_thread_gray_queue);
2235 TV_GETTIME (time_end);
2236 gc_stats.major_gc_time += TV_ELAPSED (time_start, time_end);
2238 /* FIXME: also report this to the user, preferably in gc-end. */
2239 if (major_collector.get_and_reset_num_major_objects_marked)
2240 major_collector.get_and_reset_num_major_objects_marked ();
2242 return bytes_pinned_from_failed_allocation > 0;
2246 major_start_concurrent_collection (const char *reason)
2248 TV_DECLARE (time_start);
2249 TV_DECLARE (time_end);
2250 long long num_objects_marked;
2251 SgenGrayQueue gc_thread_gray_queue;
2253 if (disable_major_collections)
2256 TV_GETTIME (time_start);
2257 SGEN_TV_GETTIME (time_major_conc_collection_start);
2259 num_objects_marked = major_collector.get_and_reset_num_major_objects_marked ();
2260 g_assert (num_objects_marked == 0);
2262 binary_protocol_concurrent_start ();
2264 init_gray_queue (&gc_thread_gray_queue, TRUE);
2265 // FIXME: store reason and pass it when finishing
2266 major_start_collection (&gc_thread_gray_queue, reason, TRUE, NULL);
2267 sgen_gray_object_queue_dispose (&gc_thread_gray_queue);
2269 num_objects_marked = major_collector.get_and_reset_num_major_objects_marked ();
2271 TV_GETTIME (time_end);
2272 gc_stats.major_gc_time += TV_ELAPSED (time_start, time_end);
2274 current_collection_generation = -1;
2278 * Returns whether the major collection has finished.
2281 major_should_finish_concurrent_collection (void)
2283 return sgen_workers_all_done ();
2287 major_update_concurrent_collection (void)
2289 TV_DECLARE (total_start);
2290 TV_DECLARE (total_end);
2292 TV_GETTIME (total_start);
2294 binary_protocol_concurrent_update ();
2296 major_collector.update_cardtable_mod_union ();
2297 sgen_los_update_cardtable_mod_union ();
2299 TV_GETTIME (total_end);
2300 gc_stats.major_gc_time += TV_ELAPSED (total_start, total_end);
2304 major_finish_concurrent_collection (gboolean forced)
2306 SgenGrayQueue gc_thread_gray_queue;
2307 TV_DECLARE (total_start);
2308 TV_DECLARE (total_end);
2310 TV_GETTIME (total_start);
2312 binary_protocol_concurrent_finish ();
2315 * We need to stop all workers since we're updating the cardtable below.
2316 * The workers will be resumed with a finishing pause context to avoid
2317 * additional cardtable and object scanning.
2319 sgen_workers_stop_all_workers ();
2321 SGEN_TV_GETTIME (time_major_conc_collection_end);
2322 gc_stats.major_gc_time_concurrent += SGEN_TV_ELAPSED (time_major_conc_collection_start, time_major_conc_collection_end);
2324 major_collector.update_cardtable_mod_union ();
2325 sgen_los_update_cardtable_mod_union ();
2327 if (mod_union_consistency_check)
2328 sgen_check_mod_union_consistency ();
2330 current_collection_generation = GENERATION_OLD;
2331 sgen_cement_reset ();
2332 init_gray_queue (&gc_thread_gray_queue, FALSE);
2333 major_finish_collection (&gc_thread_gray_queue, "finishing", FALSE, -1, forced);
2334 sgen_gray_object_queue_dispose (&gc_thread_gray_queue);
2336 TV_GETTIME (total_end);
2337 gc_stats.major_gc_time += TV_ELAPSED (total_start, total_end);
2339 current_collection_generation = -1;
2343 * Ensure an allocation request for @size will succeed by freeing enough memory.
2345 * LOCKING: The GC lock MUST be held.
2348 sgen_ensure_free_space (size_t size, int generation)
2350 int generation_to_collect = -1;
2351 const char *reason = NULL;
2353 if (generation == GENERATION_OLD) {
2354 if (sgen_need_major_collection (size)) {
2355 reason = "LOS overflow";
2356 generation_to_collect = GENERATION_OLD;
2359 if (degraded_mode) {
2360 if (sgen_need_major_collection (size)) {
2361 reason = "Degraded mode overflow";
2362 generation_to_collect = GENERATION_OLD;
2364 } else if (sgen_need_major_collection (size)) {
2365 reason = concurrent_collection_in_progress ? "Forced finish concurrent collection" : "Minor allowance";
2366 generation_to_collect = GENERATION_OLD;
2368 generation_to_collect = GENERATION_NURSERY;
2369 reason = "Nursery full";
2373 if (generation_to_collect == -1) {
2374 if (concurrent_collection_in_progress && sgen_workers_all_done ()) {
2375 generation_to_collect = GENERATION_OLD;
2376 reason = "Finish concurrent collection";
2380 if (generation_to_collect == -1)
2382 sgen_perform_collection (size, generation_to_collect, reason, FALSE, TRUE);
2386 * LOCKING: Assumes the GC lock is held.
2389 sgen_perform_collection (size_t requested_size, int generation_to_collect, const char *reason, gboolean wait_to_finish, gboolean stw)
2391 TV_DECLARE (gc_total_start);
2392 TV_DECLARE (gc_total_end);
2393 int overflow_generation_to_collect = -1;
2394 int oldest_generation_collected = generation_to_collect;
2395 const char *overflow_reason = NULL;
2396 gboolean finish_concurrent = concurrent_collection_in_progress && (major_should_finish_concurrent_collection () || generation_to_collect == GENERATION_OLD);
2398 binary_protocol_collection_requested (generation_to_collect, requested_size, wait_to_finish ? 1 : 0);
2400 SGEN_ASSERT (0, generation_to_collect == GENERATION_NURSERY || generation_to_collect == GENERATION_OLD, "What generation is this?");
2403 sgen_stop_world (generation_to_collect);
2405 SGEN_ASSERT (0, sgen_is_world_stopped (), "We can only collect if the world is stopped");
2408 TV_GETTIME (gc_total_start);
2410 // FIXME: extract overflow reason
2411 // FIXME: minor overflow for concurrent case
2412 if (generation_to_collect == GENERATION_NURSERY && !finish_concurrent) {
2413 if (concurrent_collection_in_progress)
2414 major_update_concurrent_collection ();
2416 if (collect_nursery (reason, FALSE, NULL) && !concurrent_collection_in_progress) {
2417 overflow_generation_to_collect = GENERATION_OLD;
2418 overflow_reason = "Minor overflow";
2420 } else if (finish_concurrent) {
2421 major_finish_concurrent_collection (wait_to_finish);
2422 oldest_generation_collected = GENERATION_OLD;
2424 SGEN_ASSERT (0, generation_to_collect == GENERATION_OLD, "We should have handled nursery collections above");
2425 if (major_collector.is_concurrent && !wait_to_finish) {
2426 collect_nursery ("Concurrent start", FALSE, NULL);
2427 major_start_concurrent_collection (reason);
2428 oldest_generation_collected = GENERATION_NURSERY;
2429 } else if (major_do_collection (reason, FALSE, wait_to_finish)) {
2430 overflow_generation_to_collect = GENERATION_NURSERY;
2431 overflow_reason = "Excessive pinning";
2435 if (overflow_generation_to_collect != -1) {
2436 SGEN_ASSERT (0, !concurrent_collection_in_progress, "We don't yet support overflow collections with the concurrent collector");
2439 * We need to do an overflow collection, either because we ran out of memory
2440 * or the nursery is fully pinned.
2443 if (overflow_generation_to_collect == GENERATION_NURSERY)
2444 collect_nursery (overflow_reason, TRUE, NULL);
2446 major_do_collection (overflow_reason, TRUE, wait_to_finish);
2448 oldest_generation_collected = MAX (oldest_generation_collected, overflow_generation_to_collect);
2451 SGEN_LOG (2, "Heap size: %lu, LOS size: %lu", (unsigned long)sgen_gc_get_total_heap_allocation (), (unsigned long)los_memory_usage);
2453 /* this also sets the proper pointers for the next allocation */
2454 if (generation_to_collect == GENERATION_NURSERY && !sgen_can_alloc_size (requested_size)) {
2455 /* TypeBuilder and MonoMethod are killing mcs with fragmentation */
2456 SGEN_LOG (1, "nursery collection didn't find enough room for %zd alloc (%zd pinned)", requested_size, sgen_get_pinned_count ());
2457 sgen_dump_pin_queue ();
2461 TV_GETTIME (gc_total_end);
2462 time_max = MAX (time_max, TV_ELAPSED (gc_total_start, gc_total_end));
2465 sgen_restart_world (oldest_generation_collected);
2469 * ######################################################################
2470 * ######## Memory allocation from the OS
2471 * ######################################################################
2472 * This section of code deals with getting memory from the OS and
2473 * allocating memory for GC-internal data structures.
2474 * Internal memory can be handled with a freelist for small objects.
2480 G_GNUC_UNUSED static void
2481 report_internal_mem_usage (void)
2483 printf ("Internal memory usage:\n");
2484 sgen_report_internal_mem_usage ();
2485 printf ("Pinned memory usage:\n");
2486 major_collector.report_pinned_memory_usage ();
2490 * ######################################################################
2491 * ######## Finalization support
2492 * ######################################################################
2496 * If the object has been forwarded it means it's still referenced from a root.
2497 * If it is pinned it's still alive as well.
2498 * A LOS object is only alive if we have pinned it.
2499 * Return TRUE if @obj is ready to be finalized.
2501 static inline gboolean
2502 sgen_is_object_alive (GCObject *object)
2504 if (ptr_in_nursery (object))
2505 return sgen_nursery_is_object_alive (object);
2507 return sgen_major_is_object_alive (object);
2511 * This function returns true if @object is either alive and belongs to the
2512 * current collection - major collections are full heap, so old gen objects
2513 * are never alive during a minor collection.
2516 sgen_is_object_alive_and_on_current_collection (GCObject *object)
2518 if (ptr_in_nursery (object))
2519 return sgen_nursery_is_object_alive (object);
2521 if (current_collection_generation == GENERATION_NURSERY)
2524 return sgen_major_is_object_alive (object);
2529 sgen_gc_is_object_ready_for_finalization (GCObject *object)
2531 return !sgen_is_object_alive (object);
2535 sgen_queue_finalization_entry (GCObject *obj)
2537 gboolean critical = sgen_client_object_has_critical_finalizer (obj);
2539 sgen_pointer_queue_add (critical ? &critical_fin_queue : &fin_ready_queue, obj);
2541 sgen_client_object_queued_for_finalization (obj);
2545 sgen_object_is_live (GCObject *obj)
2547 return sgen_is_object_alive_and_on_current_collection (obj);
2551 * `System.GC.WaitForPendingFinalizers` first checks `sgen_have_pending_finalizers()` to
2552 * determine whether it can exit quickly. The latter must therefore only return FALSE if
2553 * all finalizers have really finished running.
2555 * `sgen_gc_invoke_finalizers()` first dequeues a finalizable object, and then finalizes it.
2556 * This means that just checking whether the queues are empty leaves the possibility that an
2557 * object might have been dequeued but not yet finalized. That's why we need the additional
2558 * flag `pending_unqueued_finalizer`.
2561 static volatile gboolean pending_unqueued_finalizer = FALSE;
2562 volatile gboolean sgen_suspend_finalizers = FALSE;
2565 sgen_set_suspend_finalizers (void)
2567 sgen_suspend_finalizers = TRUE;
2571 sgen_gc_invoke_finalizers (void)
2575 g_assert (!pending_unqueued_finalizer);
2577 /* FIXME: batch to reduce lock contention */
2578 while (sgen_have_pending_finalizers ()) {
2584 * We need to set `pending_unqueued_finalizer` before dequeing the
2585 * finalizable object.
2587 if (!sgen_pointer_queue_is_empty (&fin_ready_queue)) {
2588 pending_unqueued_finalizer = TRUE;
2589 mono_memory_write_barrier ();
2590 obj = (GCObject *)sgen_pointer_queue_pop (&fin_ready_queue);
2591 } else if (!sgen_pointer_queue_is_empty (&critical_fin_queue)) {
2592 pending_unqueued_finalizer = TRUE;
2593 mono_memory_write_barrier ();
2594 obj = (GCObject *)sgen_pointer_queue_pop (&critical_fin_queue);
2600 SGEN_LOG (7, "Finalizing object %p (%s)", obj, sgen_client_vtable_get_name (SGEN_LOAD_VTABLE (obj)));
2608 /* the object is on the stack so it is pinned */
2609 /*g_print ("Calling finalizer for object: %p (%s)\n", obj, sgen_client_object_safe_name (obj));*/
2610 sgen_client_run_finalize (obj);
2613 if (pending_unqueued_finalizer) {
2614 mono_memory_write_barrier ();
2615 pending_unqueued_finalizer = FALSE;
2622 sgen_have_pending_finalizers (void)
2624 if (sgen_suspend_finalizers)
2626 return pending_unqueued_finalizer || !sgen_pointer_queue_is_empty (&fin_ready_queue) || !sgen_pointer_queue_is_empty (&critical_fin_queue);
2630 * ######################################################################
2631 * ######## registered roots support
2632 * ######################################################################
2636 * We do not coalesce roots.
2639 sgen_register_root (char *start, size_t size, SgenDescriptor descr, int root_type, int source, const char *msg)
2641 RootRecord new_root;
2644 for (i = 0; i < ROOT_TYPE_NUM; ++i) {
2645 RootRecord *root = (RootRecord *)sgen_hash_table_lookup (&roots_hash [i], start);
2646 /* we allow changing the size and the descriptor (for thread statics etc) */
2648 size_t old_size = root->end_root - start;
2649 root->end_root = start + size;
2650 SGEN_ASSERT (0, !!root->root_desc == !!descr, "Can't change whether a root is precise or conservative.");
2651 SGEN_ASSERT (0, root->source == source, "Can't change a root's source identifier.");
2652 SGEN_ASSERT (0, !!root->msg == !!msg, "Can't change a root's message.");
2653 root->root_desc = descr;
2655 roots_size -= old_size;
2661 new_root.end_root = start + size;
2662 new_root.root_desc = descr;
2663 new_root.source = source;
2666 sgen_hash_table_replace (&roots_hash [root_type], start, &new_root, NULL);
2669 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);
2676 sgen_deregister_root (char* addr)
2682 for (root_type = 0; root_type < ROOT_TYPE_NUM; ++root_type) {
2683 if (sgen_hash_table_remove (&roots_hash [root_type], addr, &root))
2684 roots_size -= (root.end_root - addr);
2690 sgen_wbroots_iterate_live_block_ranges (sgen_cardtable_block_callback cb)
2694 SGEN_HASH_TABLE_FOREACH (&roots_hash [ROOT_TYPE_WBARRIER], void **, start_root, RootRecord *, root) {
2695 cb ((mword)start_root, (mword)root->end_root - (mword)start_root);
2696 } SGEN_HASH_TABLE_FOREACH_END;
2699 /* Root equivalent of sgen_client_cardtable_scan_object */
2701 sgen_wbroot_scan_card_table (void** start_root, mword size, ScanCopyContext ctx)
2703 ScanPtrFieldFunc scan_field_func = ctx.ops->scan_ptr_field;
2704 guint8 *card_data = sgen_card_table_get_card_scan_address ((mword)start_root);
2705 guint8 *card_base = card_data;
2706 mword card_count = sgen_card_table_number_of_cards_in_range ((mword)start_root, size);
2707 guint8 *card_data_end = card_data + card_count;
2708 mword extra_idx = 0;
2709 char *obj_start = sgen_card_table_align_pointer (start_root);
2710 char *obj_end = (char*)start_root + size;
2711 #ifdef SGEN_HAVE_OVERLAPPING_CARDS
2712 guint8 *overflow_scan_end = NULL;
2715 #ifdef SGEN_HAVE_OVERLAPPING_CARDS
2716 /*Check for overflow and if so, setup to scan in two steps*/
2717 if (card_data_end >= SGEN_SHADOW_CARDTABLE_END) {
2718 overflow_scan_end = sgen_shadow_cardtable + (card_data_end - SGEN_SHADOW_CARDTABLE_END);
2719 card_data_end = SGEN_SHADOW_CARDTABLE_END;
2725 card_data = sgen_find_next_card (card_data, card_data_end);
2727 for (; card_data < card_data_end; card_data = sgen_find_next_card (card_data + 1, card_data_end)) {
2728 size_t idx = (card_data - card_base) + extra_idx;
2729 char *start = (char*)(obj_start + idx * CARD_SIZE_IN_BYTES);
2730 char *card_end = start + CARD_SIZE_IN_BYTES;
2731 char *elem = start, *first_elem = start;
2734 * Don't clean first and last card on 32bit systems since they
2735 * may also be part from other roots.
2737 if (card_data != card_base && card_data != (card_data_end - 1))
2738 sgen_card_table_prepare_card_for_scanning (card_data);
2740 card_end = MIN (card_end, obj_end);
2742 if (elem < (char*)start_root)
2743 first_elem = elem = (char*)start_root;
2745 for (; elem < card_end; elem += SIZEOF_VOID_P) {
2746 if (*(GCObject**)elem)
2747 scan_field_func (NULL, (GCObject**)elem, ctx.queue);
2750 binary_protocol_card_scan (first_elem, elem - first_elem);
2753 #ifdef SGEN_HAVE_OVERLAPPING_CARDS
2754 if (overflow_scan_end) {
2755 extra_idx = card_data - card_base;
2756 card_base = card_data = sgen_shadow_cardtable;
2757 card_data_end = overflow_scan_end;
2758 overflow_scan_end = NULL;
2765 sgen_wbroots_scan_card_table (ScanCopyContext ctx)
2770 SGEN_HASH_TABLE_FOREACH (&roots_hash [ROOT_TYPE_WBARRIER], void **, start_root, RootRecord *, root) {
2771 SGEN_ASSERT (0, (root->root_desc & ROOT_DESC_TYPE_MASK) == ROOT_DESC_VECTOR, "Unsupported root type");
2773 sgen_wbroot_scan_card_table (start_root, (mword)root->end_root - (mword)start_root, ctx);
2774 } SGEN_HASH_TABLE_FOREACH_END;
2778 * ######################################################################
2779 * ######## Thread handling (stop/start code)
2780 * ######################################################################
2784 sgen_get_current_collection_generation (void)
2786 return current_collection_generation;
2790 sgen_thread_register (SgenThreadInfo* info, void *stack_bottom_fallback)
2792 info->tlab_start = info->tlab_next = info->tlab_temp_end = info->tlab_real_end = NULL;
2794 sgen_client_thread_register (info, stack_bottom_fallback);
2800 sgen_thread_unregister (SgenThreadInfo *p)
2802 sgen_client_thread_unregister (p);
2806 * ######################################################################
2807 * ######## Write barriers
2808 * ######################################################################
2812 * Note: the write barriers first do the needed GC work and then do the actual store:
2813 * this way the value is visible to the conservative GC scan after the write barrier
2814 * itself. If a GC interrupts the barrier in the middle, value will be kept alive by
2815 * the conservative scan, otherwise by the remembered set scan.
2819 * mono_gc_wbarrier_arrayref_copy:
2822 mono_gc_wbarrier_arrayref_copy (gpointer dest_ptr, gpointer src_ptr, int count)
2824 HEAVY_STAT (++stat_wbarrier_arrayref_copy);
2825 /*This check can be done without taking a lock since dest_ptr array is pinned*/
2826 if (ptr_in_nursery (dest_ptr) || count <= 0) {
2827 mono_gc_memmove_aligned (dest_ptr, src_ptr, count * sizeof (gpointer));
2831 #ifdef SGEN_HEAVY_BINARY_PROTOCOL
2832 if (binary_protocol_is_heavy_enabled ()) {
2834 for (i = 0; i < count; ++i) {
2835 gpointer dest = (gpointer*)dest_ptr + i;
2836 gpointer obj = *((gpointer*)src_ptr + i);
2838 binary_protocol_wbarrier (dest, obj, (gpointer)LOAD_VTABLE (obj));
2843 remset.wbarrier_arrayref_copy (dest_ptr, src_ptr, count);
2847 * mono_gc_wbarrier_generic_nostore:
2850 mono_gc_wbarrier_generic_nostore (gpointer ptr)
2854 HEAVY_STAT (++stat_wbarrier_generic_store);
2856 sgen_client_wbarrier_generic_nostore_check (ptr);
2858 obj = *(gpointer*)ptr;
2860 binary_protocol_wbarrier (ptr, obj, (gpointer)LOAD_VTABLE (obj));
2863 * We need to record old->old pointer locations for the
2864 * concurrent collector.
2866 if (!ptr_in_nursery (obj) && !concurrent_collection_in_progress) {
2867 SGEN_LOG (8, "Skipping remset at %p", ptr);
2871 SGEN_LOG (8, "Adding remset at %p", ptr);
2873 remset.wbarrier_generic_nostore (ptr);
2877 * mono_gc_wbarrier_generic_store:
2880 mono_gc_wbarrier_generic_store (gpointer ptr, GCObject* value)
2882 SGEN_LOG (8, "Wbarrier store at %p to %p (%s)", ptr, value, value ? sgen_client_vtable_get_name (SGEN_LOAD_VTABLE (value)) : "null");
2883 SGEN_UPDATE_REFERENCE_ALLOW_NULL (ptr, value);
2884 if (ptr_in_nursery (value) || concurrent_collection_in_progress)
2885 mono_gc_wbarrier_generic_nostore (ptr);
2886 sgen_dummy_use (value);
2890 * mono_gc_wbarrier_generic_store_atomic:
2891 * Same as \c mono_gc_wbarrier_generic_store but performs the store
2892 * as an atomic operation with release semantics.
2895 mono_gc_wbarrier_generic_store_atomic (gpointer ptr, GCObject *value)
2897 HEAVY_STAT (++stat_wbarrier_generic_store_atomic);
2899 SGEN_LOG (8, "Wbarrier atomic store at %p to %p (%s)", ptr, value, value ? sgen_client_vtable_get_name (SGEN_LOAD_VTABLE (value)) : "null");
2901 InterlockedWritePointer ((volatile gpointer *)ptr, value);
2903 if (ptr_in_nursery (value) || concurrent_collection_in_progress)
2904 mono_gc_wbarrier_generic_nostore (ptr);
2906 sgen_dummy_use (value);
2910 sgen_wbarrier_range_copy (gpointer _dest, gpointer _src, int size)
2912 remset.wbarrier_range_copy (_dest,_src, size);
2916 * ######################################################################
2917 * ######## Other mono public interface functions.
2918 * ######################################################################
2922 sgen_gc_collect (int generation)
2927 sgen_perform_collection (0, generation, "user request", TRUE, TRUE);
2932 sgen_gc_collection_count (int generation)
2934 if (generation == 0)
2935 return gc_stats.minor_gc_count;
2936 return gc_stats.major_gc_count;
2940 sgen_gc_get_used_size (void)
2944 tot = los_memory_usage;
2945 tot += nursery_section->end_data - nursery_section->data;
2946 tot += major_collector.get_used_size ();
2947 /* FIXME: account for pinned objects */
2953 sgen_env_var_error (const char *env_var, const char *fallback, const char *description_format, ...)
2957 va_start (ap, description_format);
2959 fprintf (stderr, "Warning: In environment variable `%s': ", env_var);
2960 vfprintf (stderr, description_format, ap);
2962 fprintf (stderr, " - %s", fallback);
2963 fprintf (stderr, "\n");
2969 parse_double_in_interval (const char *env_var, const char *opt_name, const char *opt, double min, double max, double *result)
2972 double val = strtod (opt, &endptr);
2973 if (endptr == opt) {
2974 sgen_env_var_error (env_var, "Using default value.", "`%s` must be a number.", opt_name);
2977 else if (val < min || val > max) {
2978 sgen_env_var_error (env_var, "Using default value.", "`%s` must be between %.2f - %.2f.", opt_name, min, max);
2990 char *major_collector_opt = NULL;
2991 char *minor_collector_opt = NULL;
2992 char *params_opts = NULL;
2993 char *debug_opts = NULL;
2994 size_t max_heap = 0;
2995 size_t soft_limit = 0;
2997 gboolean debug_print_allowance = FALSE;
2998 double allowance_ratio = 0, save_target = 0;
2999 gboolean cement_enabled = TRUE;
3000 gboolean dynamic_nursery = FALSE;
3001 size_t min_nursery_size = 0, max_nursery_size = 0;
3004 result = InterlockedCompareExchange (&gc_initialized, -1, 0);
3007 /* already inited */
3010 /* being inited by another thread */
3011 mono_thread_info_usleep (1000);
3014 /* we will init it */
3017 g_assert_not_reached ();
3019 } while (result != 0);
3021 SGEN_TV_GETTIME (sgen_init_timestamp);
3023 #ifdef SGEN_WITHOUT_MONO
3024 mono_thread_smr_init ();
3027 mono_coop_mutex_init (&gc_mutex);
3029 gc_debug_file = stderr;
3031 mono_coop_mutex_init (&sgen_interruption_mutex);
3033 if ((env = g_getenv (MONO_GC_PARAMS_NAME)) || gc_params_options) {
3034 params_opts = g_strdup_printf ("%s,%s", gc_params_options ? gc_params_options : "", env ? env : "");
3039 opts = g_strsplit (params_opts, ",", -1);
3040 for (ptr = opts; *ptr; ++ptr) {
3042 if (g_str_has_prefix (opt, "major=")) {
3043 opt = strchr (opt, '=') + 1;
3044 major_collector_opt = g_strdup (opt);
3045 } else if (g_str_has_prefix (opt, "minor=")) {
3046 opt = strchr (opt, '=') + 1;
3047 minor_collector_opt = g_strdup (opt);
3055 sgen_init_internal_allocator ();
3056 sgen_init_nursery_allocator ();
3057 sgen_init_fin_weak_hash ();
3058 sgen_init_hash_table ();
3059 sgen_init_descriptors ();
3060 sgen_init_gray_queues ();
3061 sgen_init_allocator ();
3062 sgen_init_gchandles ();
3064 sgen_register_fixed_internal_mem_type (INTERNAL_MEM_SECTION, SGEN_SIZEOF_GC_MEM_SECTION);
3065 sgen_register_fixed_internal_mem_type (INTERNAL_MEM_GRAY_QUEUE, sizeof (GrayQueueSection));
3067 sgen_client_init ();
3069 if (!minor_collector_opt) {
3070 sgen_simple_nursery_init (&sgen_minor_collector, FALSE);
3072 if (!strcmp (minor_collector_opt, "simple")) {
3074 sgen_simple_nursery_init (&sgen_minor_collector, FALSE);
3075 } else if (!strcmp (minor_collector_opt, "simple-par")) {
3076 sgen_simple_nursery_init (&sgen_minor_collector, TRUE);
3077 } else if (!strcmp (minor_collector_opt, "split")) {
3078 sgen_split_nursery_init (&sgen_minor_collector);
3080 sgen_env_var_error (MONO_GC_PARAMS_NAME, "Using `simple` instead.", "Unknown minor collector `%s'.", minor_collector_opt);
3081 goto use_simple_nursery;
3085 if (!major_collector_opt) {
3087 DEFAULT_MAJOR_INIT (&major_collector);
3088 } else if (!strcmp (major_collector_opt, "marksweep")) {
3089 sgen_marksweep_init (&major_collector);
3090 } else if (!strcmp (major_collector_opt, "marksweep-conc")) {
3091 sgen_marksweep_conc_init (&major_collector);
3092 } else if (!strcmp (major_collector_opt, "marksweep-conc-par")) {
3093 sgen_marksweep_conc_par_init (&major_collector);
3095 sgen_env_var_error (MONO_GC_PARAMS_NAME, "Using `" DEFAULT_MAJOR_NAME "` instead.", "Unknown major collector `%s'.", major_collector_opt);
3096 goto use_default_major;
3100 gboolean usage_printed = FALSE;
3102 for (ptr = opts; *ptr; ++ptr) {
3104 if (!strcmp (opt, ""))
3106 if (g_str_has_prefix (opt, "major="))
3108 if (g_str_has_prefix (opt, "minor="))
3110 if (g_str_has_prefix (opt, "max-heap-size=")) {
3111 size_t page_size = mono_pagesize ();
3112 size_t max_heap_candidate = 0;
3113 opt = strchr (opt, '=') + 1;
3114 if (*opt && mono_gc_parse_environment_string_extract_number (opt, &max_heap_candidate)) {
3115 max_heap = (max_heap_candidate + page_size - 1) & ~(size_t)(page_size - 1);
3116 if (max_heap != max_heap_candidate)
3117 sgen_env_var_error (MONO_GC_PARAMS_NAME, "Rounding up.", "`max-heap-size` size must be a multiple of %d.", page_size);
3119 sgen_env_var_error (MONO_GC_PARAMS_NAME, NULL, "`max-heap-size` must be an integer.");
3123 if (g_str_has_prefix (opt, "soft-heap-limit=")) {
3124 opt = strchr (opt, '=') + 1;
3125 if (*opt && mono_gc_parse_environment_string_extract_number (opt, &soft_limit)) {
3126 if (soft_limit <= 0) {
3127 sgen_env_var_error (MONO_GC_PARAMS_NAME, NULL, "`soft-heap-limit` must be positive.");
3131 sgen_env_var_error (MONO_GC_PARAMS_NAME, NULL, "`soft-heap-limit` must be an integer.");
3135 if (g_str_has_prefix (opt, "nursery-size=")) {
3137 opt = strchr (opt, '=') + 1;
3138 if (*opt && mono_gc_parse_environment_string_extract_number (opt, &val)) {
3139 if ((val & (val - 1))) {
3140 sgen_env_var_error (MONO_GC_PARAMS_NAME, "Using default value.", "`nursery-size` must be a power of two.");
3144 if (val < SGEN_MAX_NURSERY_WASTE) {
3145 sgen_env_var_error (MONO_GC_PARAMS_NAME, "Using default value.",
3146 "`nursery-size` must be at least %d bytes.", SGEN_MAX_NURSERY_WASTE);
3150 min_nursery_size = max_nursery_size = val;
3151 dynamic_nursery = FALSE;
3153 sgen_env_var_error (MONO_GC_PARAMS_NAME, "Using default value.", "`nursery-size` must be an integer.");
3158 if (g_str_has_prefix (opt, "save-target-ratio=")) {
3160 opt = strchr (opt, '=') + 1;
3161 if (parse_double_in_interval (MONO_GC_PARAMS_NAME, "save-target-ratio", opt,
3162 SGEN_MIN_SAVE_TARGET_RATIO, SGEN_MAX_SAVE_TARGET_RATIO, &val)) {
3167 if (g_str_has_prefix (opt, "default-allowance-ratio=")) {
3169 opt = strchr (opt, '=') + 1;
3170 if (parse_double_in_interval (MONO_GC_PARAMS_NAME, "default-allowance-ratio", opt,
3171 SGEN_MIN_ALLOWANCE_NURSERY_SIZE_RATIO, SGEN_MAX_ALLOWANCE_NURSERY_SIZE_RATIO, &val)) {
3172 allowance_ratio = val;
3177 if (!strcmp (opt, "cementing")) {
3178 cement_enabled = TRUE;
3181 if (!strcmp (opt, "no-cementing")) {
3182 cement_enabled = FALSE;
3186 if (!strcmp (opt, "precleaning")) {
3187 precleaning_enabled = TRUE;
3190 if (!strcmp (opt, "no-precleaning")) {
3191 precleaning_enabled = FALSE;
3195 if (!strcmp (opt, "dynamic-nursery")) {
3196 if (sgen_minor_collector.is_split)
3197 sgen_env_var_error (MONO_GC_PARAMS_NAME, "Using default value.",
3198 "dynamic-nursery not supported with split-nursery.");
3200 dynamic_nursery = TRUE;
3203 if (!strcmp (opt, "no-dynamic-nursery")) {
3204 dynamic_nursery = FALSE;
3208 if (major_collector.handle_gc_param && major_collector.handle_gc_param (opt))
3211 if (sgen_minor_collector.handle_gc_param && sgen_minor_collector.handle_gc_param (opt))
3214 if (sgen_client_handle_gc_param (opt))
3217 sgen_env_var_error (MONO_GC_PARAMS_NAME, "Ignoring.", "Unknown option `%s`.", opt);
3222 fprintf (stderr, "\n%s must be a comma-delimited list of one or more of the following:\n", MONO_GC_PARAMS_NAME);
3223 fprintf (stderr, " max-heap-size=N (where N is an integer, possibly with a k, m or a g suffix)\n");
3224 fprintf (stderr, " soft-heap-limit=n (where N is an integer, possibly with a k, m or a g suffix)\n");
3225 fprintf (stderr, " nursery-size=N (where N is an integer, possibly with a k, m or a g suffix)\n");
3226 fprintf (stderr, " major=COLLECTOR (where COLLECTOR is `marksweep', `marksweep-conc', `marksweep-par')\n");
3227 fprintf (stderr, " minor=COLLECTOR (where COLLECTOR is `simple' or `split')\n");
3228 fprintf (stderr, " wbarrier=WBARRIER (where WBARRIER is `remset' or `cardtable')\n");
3229 fprintf (stderr, " [no-]cementing\n");
3230 fprintf (stderr, " [no-]dynamic-nursery\n");
3231 if (major_collector.print_gc_param_usage)
3232 major_collector.print_gc_param_usage ();
3233 if (sgen_minor_collector.print_gc_param_usage)
3234 sgen_minor_collector.print_gc_param_usage ();
3235 sgen_client_print_gc_params_usage ();
3236 fprintf (stderr, " Experimental options:\n");
3237 fprintf (stderr, " save-target-ratio=R (where R must be between %.2f - %.2f).\n", SGEN_MIN_SAVE_TARGET_RATIO, SGEN_MAX_SAVE_TARGET_RATIO);
3238 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);
3239 fprintf (stderr, "\n");
3241 usage_printed = TRUE;
3246 if (major_collector_opt)
3247 g_free (major_collector_opt);
3249 if (minor_collector_opt)
3250 g_free (minor_collector_opt);
3253 g_free (params_opts);
3255 alloc_nursery (dynamic_nursery, min_nursery_size, max_nursery_size);
3257 sgen_pinning_init ();
3258 sgen_cement_init (cement_enabled);
3260 if ((env = g_getenv (MONO_GC_DEBUG_NAME)) || gc_debug_options) {
3261 debug_opts = g_strdup_printf ("%s,%s", gc_debug_options ? gc_debug_options : "", env ? env : "");
3266 gboolean usage_printed = FALSE;
3268 opts = g_strsplit (debug_opts, ",", -1);
3269 for (ptr = opts; ptr && *ptr; ptr ++) {
3271 if (!strcmp (opt, ""))
3273 if (opt [0] >= '0' && opt [0] <= '9') {
3274 gc_debug_level = atoi (opt);
3279 char *rf = g_strdup_printf ("%s.%d", opt, mono_process_current_pid ());
3280 gc_debug_file = fopen (rf, "wb");
3282 gc_debug_file = stderr;
3285 } else if (!strcmp (opt, "print-allowance")) {
3286 debug_print_allowance = TRUE;
3287 } else if (!strcmp (opt, "print-pinning")) {
3288 sgen_pin_stats_enable ();
3289 } else if (!strcmp (opt, "verify-before-allocs")) {
3290 verify_before_allocs = 1;
3291 has_per_allocation_action = TRUE;
3292 } else if (g_str_has_prefix (opt, "max-valloc-size=")) {
3293 size_t max_valloc_size;
3294 char *arg = strchr (opt, '=') + 1;
3295 if (*opt && mono_gc_parse_environment_string_extract_number (arg, &max_valloc_size)) {
3296 mono_valloc_set_limit (max_valloc_size);
3298 sgen_env_var_error (MONO_GC_DEBUG_NAME, NULL, "`max-valloc-size` must be an integer.");
3301 } else if (g_str_has_prefix (opt, "verify-before-allocs=")) {
3302 char *arg = strchr (opt, '=') + 1;
3303 verify_before_allocs = atoi (arg);
3304 has_per_allocation_action = TRUE;
3305 } else if (!strcmp (opt, "collect-before-allocs")) {
3306 collect_before_allocs = 1;
3307 has_per_allocation_action = TRUE;
3308 } else if (g_str_has_prefix (opt, "collect-before-allocs=")) {
3309 char *arg = strchr (opt, '=') + 1;
3310 has_per_allocation_action = TRUE;
3311 collect_before_allocs = atoi (arg);
3312 } else if (!strcmp (opt, "verify-before-collections")) {
3313 whole_heap_check_before_collection = TRUE;
3314 } else if (!strcmp (opt, "check-remset-consistency")) {
3315 remset_consistency_checks = TRUE;
3316 nursery_clear_policy = CLEAR_AT_GC;
3317 } else if (!strcmp (opt, "mod-union-consistency-check")) {
3318 if (!major_collector.is_concurrent) {
3319 sgen_env_var_error (MONO_GC_DEBUG_NAME, "Ignoring.", "`mod-union-consistency-check` only works with concurrent major collector.");
3322 mod_union_consistency_check = TRUE;
3323 } else if (!strcmp (opt, "check-mark-bits")) {
3324 check_mark_bits_after_major_collection = TRUE;
3325 } else if (!strcmp (opt, "check-nursery-pinned")) {
3326 check_nursery_objects_pinned = TRUE;
3327 } else if (!strcmp (opt, "clear-at-gc")) {
3328 nursery_clear_policy = CLEAR_AT_GC;
3329 } else if (!strcmp (opt, "clear-nursery-at-gc")) {
3330 nursery_clear_policy = CLEAR_AT_GC;
3331 } else if (!strcmp (opt, "clear-at-tlab-creation")) {
3332 nursery_clear_policy = CLEAR_AT_TLAB_CREATION;
3333 } else if (!strcmp (opt, "debug-clear-at-tlab-creation")) {
3334 nursery_clear_policy = CLEAR_AT_TLAB_CREATION_DEBUG;
3335 } else if (!strcmp (opt, "check-scan-starts")) {
3336 do_scan_starts_check = TRUE;
3337 } else if (!strcmp (opt, "verify-nursery-at-minor-gc")) {
3338 do_verify_nursery = TRUE;
3339 } else if (!strcmp (opt, "check-concurrent")) {
3340 if (!major_collector.is_concurrent) {
3341 sgen_env_var_error (MONO_GC_DEBUG_NAME, "Ignoring.", "`check-concurrent` only works with concurrent major collectors.");
3344 nursery_clear_policy = CLEAR_AT_GC;
3345 do_concurrent_checks = TRUE;
3346 } else if (!strcmp (opt, "dump-nursery-at-minor-gc")) {
3347 do_dump_nursery_content = TRUE;
3348 } else if (!strcmp (opt, "disable-minor")) {
3349 disable_minor_collections = TRUE;
3350 } else if (!strcmp (opt, "disable-major")) {
3351 disable_major_collections = TRUE;
3352 } else if (g_str_has_prefix (opt, "heap-dump=")) {
3353 char *filename = strchr (opt, '=') + 1;
3354 nursery_clear_policy = CLEAR_AT_GC;
3355 sgen_debug_enable_heap_dump (filename);
3356 } else if (g_str_has_prefix (opt, "binary-protocol=")) {
3357 char *filename = strchr (opt, '=') + 1;
3358 char *colon = strrchr (filename, ':');
3361 if (!mono_gc_parse_environment_string_extract_number (colon + 1, &limit)) {
3362 sgen_env_var_error (MONO_GC_DEBUG_NAME, "Ignoring limit.", "Binary protocol file size limit must be an integer.");
3367 binary_protocol_init (filename, (long long)limit);
3368 } else if (!strcmp (opt, "nursery-canaries")) {
3369 do_verify_nursery = TRUE;
3370 enable_nursery_canaries = TRUE;
3371 } else if (!sgen_client_handle_gc_debug (opt)) {
3372 sgen_env_var_error (MONO_GC_DEBUG_NAME, "Ignoring.", "Unknown option `%s`.", opt);
3377 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);
3378 fprintf (stderr, "Valid <option>s are:\n");
3379 fprintf (stderr, " collect-before-allocs[=<n>]\n");
3380 fprintf (stderr, " verify-before-allocs[=<n>]\n");
3381 fprintf (stderr, " max-valloc-size=N (where N is an integer, possibly with a k, m or a g suffix)\n");
3382 fprintf (stderr, " check-remset-consistency\n");
3383 fprintf (stderr, " check-mark-bits\n");
3384 fprintf (stderr, " check-nursery-pinned\n");
3385 fprintf (stderr, " verify-before-collections\n");
3386 fprintf (stderr, " verify-nursery-at-minor-gc\n");
3387 fprintf (stderr, " dump-nursery-at-minor-gc\n");
3388 fprintf (stderr, " disable-minor\n");
3389 fprintf (stderr, " disable-major\n");
3390 fprintf (stderr, " check-concurrent\n");
3391 fprintf (stderr, " clear-[nursery-]at-gc\n");
3392 fprintf (stderr, " clear-at-tlab-creation\n");
3393 fprintf (stderr, " debug-clear-at-tlab-creation\n");
3394 fprintf (stderr, " check-scan-starts\n");
3395 fprintf (stderr, " print-allowance\n");
3396 fprintf (stderr, " print-pinning\n");
3397 fprintf (stderr, " heap-dump=<filename>\n");
3398 fprintf (stderr, " binary-protocol=<filename>[:<file-size-limit>]\n");
3399 fprintf (stderr, " nursery-canaries\n");
3400 sgen_client_print_gc_debug_usage ();
3401 fprintf (stderr, "\n");
3403 usage_printed = TRUE;
3410 g_free (debug_opts);
3412 if (check_mark_bits_after_major_collection)
3413 nursery_clear_policy = CLEAR_AT_GC;
3415 if (major_collector.post_param_init)
3416 major_collector.post_param_init (&major_collector);
3418 if (major_collector.is_concurrent || sgen_minor_collector.is_parallel) {
3419 int num_workers = 1;
3420 if (major_collector.is_parallel || sgen_minor_collector.is_parallel) {
3421 /* FIXME Detect the number of physical cores, instead of logical */
3422 num_workers = mono_cpu_count () / 2;
3423 if (num_workers < 1)
3426 sgen_workers_init (num_workers, (SgenWorkerCallback) major_collector.worker_init_cb);
3429 sgen_memgov_init (max_heap, soft_limit, debug_print_allowance, allowance_ratio, save_target);
3431 memset (&remset, 0, sizeof (remset));
3433 sgen_card_table_init (&remset);
3435 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");
3439 sgen_init_bridge ();
3443 sgen_gc_initialized ()
3445 return gc_initialized > 0;
3449 sgen_get_nursery_clear_policy (void)
3451 return nursery_clear_policy;
3457 mono_coop_mutex_lock (&gc_mutex);
3461 sgen_gc_unlock (void)
3463 mono_coop_mutex_unlock (&gc_mutex);
3467 sgen_major_collector_iterate_live_block_ranges (sgen_cardtable_block_callback callback)
3469 major_collector.iterate_live_block_ranges (callback);
3473 sgen_major_collector_iterate_block_ranges (sgen_cardtable_block_callback callback)
3475 major_collector.iterate_block_ranges (callback);
3479 sgen_get_major_collector (void)
3481 return &major_collector;
3485 sgen_get_minor_collector (void)
3487 return &sgen_minor_collector;
3491 sgen_get_remset (void)
3497 count_cards (long long *major_total, long long *major_marked, long long *los_total, long long *los_marked)
3499 sgen_get_major_collector ()->count_cards (major_total, major_marked);
3500 sgen_los_count_cards (los_total, los_marked);
3503 static gboolean world_is_stopped = FALSE;
3505 /* LOCKING: assumes the GC lock is held */
3507 sgen_stop_world (int generation)
3509 long long major_total = -1, major_marked = -1, los_total = -1, los_marked = -1;
3511 SGEN_ASSERT (0, !world_is_stopped, "Why are we stopping a stopped world?");
3513 binary_protocol_world_stopping (generation, sgen_timestamp (), (gpointer) (gsize) mono_native_thread_id_get ());
3515 sgen_client_stop_world (generation);
3517 world_is_stopped = TRUE;
3519 if (binary_protocol_is_heavy_enabled ())
3520 count_cards (&major_total, &major_marked, &los_total, &los_marked);
3521 binary_protocol_world_stopped (generation, sgen_timestamp (), major_total, major_marked, los_total, los_marked);
3524 /* LOCKING: assumes the GC lock is held */
3526 sgen_restart_world (int generation)
3528 long long major_total = -1, major_marked = -1, los_total = -1, los_marked = -1;
3531 SGEN_ASSERT (0, world_is_stopped, "Why are we restarting a running world?");
3533 if (binary_protocol_is_heavy_enabled ())
3534 count_cards (&major_total, &major_marked, &los_total, &los_marked);
3535 binary_protocol_world_restarting (generation, sgen_timestamp (), major_total, major_marked, los_total, los_marked);
3537 world_is_stopped = FALSE;
3539 sgen_client_restart_world (generation, &stw_time);
3541 binary_protocol_world_restarted (generation, sgen_timestamp ());
3543 if (sgen_client_bridge_need_processing ())
3544 sgen_client_bridge_processing_finish (generation);
3546 sgen_memgov_collection_end (generation, stw_time);
3550 sgen_is_world_stopped (void)
3552 return world_is_stopped;
3556 sgen_check_whole_heap_stw (void)
3558 sgen_stop_world (0);
3559 sgen_clear_nursery_fragments ();
3560 sgen_check_whole_heap (TRUE);
3561 sgen_restart_world (0);
3565 sgen_timestamp (void)
3567 SGEN_TV_DECLARE (timestamp);
3568 SGEN_TV_GETTIME (timestamp);
3569 return SGEN_TV_ELAPSED (sgen_init_timestamp, timestamp);
3572 #endif /* HAVE_SGEN_GC */