xref: /haiku/src/system/kernel/slab/Slab.cpp (revision 3d4afef9cba2f328e238089d4609d00d4b1524f3)
1 /*
2  * Copyright 2010, Ingo Weinhold <ingo_weinhold@gmx.de>.
3  * Copyright 2008-2010, Axel Dörfler. All Rights Reserved.
4  * Copyright 2007, Hugo Santos. All Rights Reserved.
5  *
6  * Distributed under the terms of the MIT License.
7  */
8 
9 
10 #include <slab/Slab.h>
11 
12 #include <algorithm>
13 #include <new>
14 #include <stdlib.h>
15 #include <string.h>
16 
17 #include <KernelExport.h>
18 
19 #include <condition_variable.h>
20 #include <elf.h>
21 #include <kernel.h>
22 #include <low_resource_manager.h>
23 #include <slab/ObjectDepot.h>
24 #include <smp.h>
25 #include <tracing.h>
26 #include <util/AutoLock.h>
27 #include <util/DoublyLinkedList.h>
28 #include <vm/vm.h>
29 #include <vm/VMAddressSpace.h>
30 
31 #include "HashedObjectCache.h"
32 #include "MemoryManager.h"
33 #include "slab_debug.h"
34 #include "slab_private.h"
35 #include "SmallObjectCache.h"
36 
37 
38 #if !USE_GUARDED_HEAP_FOR_OBJECT_CACHE
39 
40 
41 typedef DoublyLinkedList<ObjectCache> ObjectCacheList;
42 
43 typedef DoublyLinkedList<ObjectCache,
44 	DoublyLinkedListMemberGetLink<ObjectCache, &ObjectCache::maintenance_link> >
45 		MaintenanceQueue;
46 
47 static ObjectCacheList sObjectCaches;
48 static mutex sObjectCacheListLock = MUTEX_INITIALIZER("object cache list");
49 
50 static mutex sMaintenanceLock
51 	= MUTEX_INITIALIZER("object cache resize requests");
52 static MaintenanceQueue sMaintenanceQueue;
53 static ConditionVariable sMaintenanceCondition;
54 
55 
56 #if SLAB_ALLOCATION_TRACKING_AVAILABLE
57 
58 struct caller_info {
59 	addr_t		caller;
60 	size_t		count;
61 	size_t		size;
62 };
63 
64 static const int32 kCallerInfoTableSize = 1024;
65 static caller_info sCallerInfoTable[kCallerInfoTableSize];
66 static int32 sCallerInfoCount = 0;
67 
68 static caller_info* get_caller_info(addr_t caller);
69 
70 
71 RANGE_MARKER_FUNCTION_PROTOTYPES(slab_allocator)
72 RANGE_MARKER_FUNCTION_PROTOTYPES(SlabHashedObjectCache)
73 RANGE_MARKER_FUNCTION_PROTOTYPES(SlabMemoryManager)
74 RANGE_MARKER_FUNCTION_PROTOTYPES(SlabObjectCache)
75 RANGE_MARKER_FUNCTION_PROTOTYPES(SlabObjectDepot)
76 RANGE_MARKER_FUNCTION_PROTOTYPES(Slab)
77 RANGE_MARKER_FUNCTION_PROTOTYPES(SlabSmallObjectCache)
78 
79 
80 static const addr_t kSlabCodeAddressRanges[] = {
81 	RANGE_MARKER_FUNCTION_ADDRESS_RANGE(slab_allocator),
82 	RANGE_MARKER_FUNCTION_ADDRESS_RANGE(SlabHashedObjectCache),
83 	RANGE_MARKER_FUNCTION_ADDRESS_RANGE(SlabMemoryManager),
84 	RANGE_MARKER_FUNCTION_ADDRESS_RANGE(SlabObjectCache),
85 	RANGE_MARKER_FUNCTION_ADDRESS_RANGE(SlabObjectDepot),
86 	RANGE_MARKER_FUNCTION_ADDRESS_RANGE(Slab),
87 	RANGE_MARKER_FUNCTION_ADDRESS_RANGE(SlabSmallObjectCache)
88 };
89 
90 static const uint32 kSlabCodeAddressRangeCount
91 	= B_COUNT_OF(kSlabCodeAddressRanges) / 2;
92 
93 #endif	// SLAB_ALLOCATION_TRACKING_AVAILABLE
94 
95 
96 RANGE_MARKER_FUNCTION_BEGIN(Slab)
97 
98 
99 #if SLAB_OBJECT_CACHE_TRACING
100 
101 
102 namespace SlabObjectCacheTracing {
103 
104 class ObjectCacheTraceEntry
105 	: public TRACE_ENTRY_SELECTOR(SLAB_OBJECT_CACHE_TRACING_STACK_TRACE) {
106 	public:
107 		ObjectCacheTraceEntry(ObjectCache* cache)
108 			:
109 			TraceEntryBase(SLAB_OBJECT_CACHE_TRACING_STACK_TRACE, 0, true),
110 			fCache(cache)
111 		{
112 		}
113 
114 	protected:
115 		ObjectCache*	fCache;
116 };
117 
118 
119 class Create : public ObjectCacheTraceEntry {
120 	public:
121 		Create(const char* name, size_t objectSize, size_t alignment,
122 				size_t maxByteUsage, uint32 flags, void* cookie,
123 				ObjectCache* cache)
124 			:
125 			ObjectCacheTraceEntry(cache),
126 			fObjectSize(objectSize),
127 			fAlignment(alignment),
128 			fMaxByteUsage(maxByteUsage),
129 			fFlags(flags),
130 			fCookie(cookie)
131 		{
132 			fName = alloc_tracing_buffer_strcpy(name, 64, false);
133 			Initialized();
134 		}
135 
136 		virtual void AddDump(TraceOutput& out)
137 		{
138 			out.Print("object cache create: name: \"%s\", object size: "
139 				"%" B_PRIuSIZE ", alignment: %" B_PRIuSIZE ", max usage: "
140 				"%" B_PRIuSIZE ", flags: 0x%" B_PRIx32 ", cookie: %p -> cache: %p",
141 					fName, fObjectSize, fAlignment, fMaxByteUsage, fFlags,
142 					fCookie, fCache);
143 		}
144 
145 	private:
146 		const char*	fName;
147 		size_t		fObjectSize;
148 		size_t		fAlignment;
149 		size_t		fMaxByteUsage;
150 		uint32		fFlags;
151 		void*		fCookie;
152 };
153 
154 
155 class Delete : public ObjectCacheTraceEntry {
156 	public:
157 		Delete(ObjectCache* cache)
158 			:
159 			ObjectCacheTraceEntry(cache)
160 		{
161 			Initialized();
162 		}
163 
164 		virtual void AddDump(TraceOutput& out)
165 		{
166 			out.Print("object cache delete: %p", fCache);
167 		}
168 };
169 
170 
171 class Alloc : public ObjectCacheTraceEntry {
172 	public:
173 		Alloc(ObjectCache* cache, uint32 flags, void* object)
174 			:
175 			ObjectCacheTraceEntry(cache),
176 			fFlags(flags),
177 			fObject(object)
178 		{
179 			Initialized();
180 		}
181 
182 		virtual void AddDump(TraceOutput& out)
183 		{
184 			out.Print("object cache alloc: cache: %p, flags: 0x%" B_PRIx32
185 				" -> object: %p", fCache, fFlags, fObject);
186 		}
187 
188 	private:
189 		uint32		fFlags;
190 		void*		fObject;
191 };
192 
193 
194 class Free : public ObjectCacheTraceEntry {
195 	public:
196 		Free(ObjectCache* cache, void* object)
197 			:
198 			ObjectCacheTraceEntry(cache),
199 			fObject(object)
200 		{
201 			Initialized();
202 		}
203 
204 		virtual void AddDump(TraceOutput& out)
205 		{
206 			out.Print("object cache free: cache: %p, object: %p", fCache,
207 				fObject);
208 		}
209 
210 	private:
211 		void*		fObject;
212 };
213 
214 
215 class Reserve : public ObjectCacheTraceEntry {
216 	public:
217 		Reserve(ObjectCache* cache, size_t count, uint32 flags)
218 			:
219 			ObjectCacheTraceEntry(cache),
220 			fCount(count),
221 			fFlags(flags)
222 		{
223 			Initialized();
224 		}
225 
226 		virtual void AddDump(TraceOutput& out)
227 		{
228 			out.Print("object cache reserve: cache: %p, count: %" B_PRIu32 ", "
229 				"flags: 0x%" B_PRIx32, fCache, fCount, fFlags);
230 		}
231 
232 	private:
233 		uint32		fCount;
234 		uint32		fFlags;
235 };
236 
237 
238 }	// namespace SlabObjectCacheTracing
239 
240 #	define T(x)	new(std::nothrow) SlabObjectCacheTracing::x
241 
242 #else
243 #	define T(x)
244 #endif	// SLAB_OBJECT_CACHE_TRACING
245 
246 
247 // #pragma mark -
248 
249 
250 static void
251 dump_slab(::slab* slab)
252 {
253 	kprintf("  %p  %p  %6" B_PRIuSIZE " %6" B_PRIuSIZE " %6" B_PRIuSIZE "  %p\n",
254 		slab, slab->pages, slab->size, slab->count, slab->offset, slab->free);
255 }
256 
257 
258 static int
259 dump_slabs(int argc, char* argv[])
260 {
261 	kprintf("%*s %22s %8s %8s %8s %6s %8s %8s %8s\n",
262 		B_PRINTF_POINTER_WIDTH + 2, "address", "name", "objsize", "align",
263 		"usage", "empty", "usedobj", "total", "flags");
264 
265 	ObjectCacheList::Iterator it = sObjectCaches.GetIterator();
266 
267 	while (it.HasNext()) {
268 		ObjectCache* cache = it.Next();
269 
270 		kprintf("%p %22s %8lu %8" B_PRIuSIZE " %8lu %6lu %8lu %8lu %8" B_PRIx32
271 			"\n", cache, cache->name, cache->object_size, cache->alignment,
272 			cache->usage, cache->empty_count, cache->used_count,
273 			cache->total_objects, cache->flags);
274 	}
275 
276 	return 0;
277 }
278 
279 
280 static int
281 dump_cache_info(int argc, char* argv[])
282 {
283 	if (argc < 2) {
284 		kprintf("usage: slab_cache [address]\n");
285 		return 0;
286 	}
287 
288 	ObjectCache* cache = (ObjectCache*)parse_expression(argv[1]);
289 
290 	kprintf("name:              %s\n", cache->name);
291 	kprintf("lock:              %p\n", &cache->lock);
292 	kprintf("object_size:       %lu\n", cache->object_size);
293 	kprintf("alignment:         %" B_PRIuSIZE "\n", cache->alignment);
294 	kprintf("cache_color_cycle: %lu\n", cache->cache_color_cycle);
295 	kprintf("total_objects:     %lu\n", cache->total_objects);
296 	kprintf("used_count:        %lu\n", cache->used_count);
297 	kprintf("empty_count:       %lu\n", cache->empty_count);
298 	kprintf("pressure:          %lu\n", cache->pressure);
299 	kprintf("slab_size:         %lu\n", cache->slab_size);
300 	kprintf("usage:             %lu\n", cache->usage);
301 	kprintf("maximum:           %lu\n", cache->maximum);
302 	kprintf("flags:             0x%" B_PRIx32 "\n", cache->flags);
303 	kprintf("cookie:            %p\n", cache->cookie);
304 	kprintf("resize entry don't wait: %p\n", cache->resize_entry_dont_wait);
305 	kprintf("resize entry can wait:   %p\n", cache->resize_entry_can_wait);
306 
307 	kprintf("  %-*s    %-*s      size   used offset  free\n",
308 		B_PRINTF_POINTER_WIDTH, "slab", B_PRINTF_POINTER_WIDTH, "chunk");
309 
310 	SlabList::Iterator iterator = cache->empty.GetIterator();
311 	if (iterator.HasNext())
312 		kprintf("empty:\n");
313 	while (::slab* slab = iterator.Next())
314 		dump_slab(slab);
315 
316 	iterator = cache->partial.GetIterator();
317 	if (iterator.HasNext())
318 		kprintf("partial:\n");
319 	while (::slab* slab = iterator.Next())
320 		dump_slab(slab);
321 
322 	iterator = cache->full.GetIterator();
323 	if (iterator.HasNext())
324 		kprintf("full:\n");
325 	while (::slab* slab = iterator.Next())
326 		dump_slab(slab);
327 
328 	if ((cache->flags & CACHE_NO_DEPOT) == 0) {
329 		kprintf("depot:\n");
330 		dump_object_depot(&cache->depot);
331 	}
332 
333 	return 0;
334 }
335 
336 
337 // #pragma mark - AllocationTrackingCallback
338 
339 
340 #if SLAB_ALLOCATION_TRACKING_AVAILABLE
341 
342 AllocationTrackingCallback::~AllocationTrackingCallback()
343 {
344 }
345 
346 #endif	// SLAB_ALLOCATION_TRACKING_AVAILABLE
347 
348 
349 // #pragma mark -
350 
351 
352 #if SLAB_ALLOCATION_TRACKING_AVAILABLE
353 
354 namespace {
355 
356 class AllocationCollectorCallback : public AllocationTrackingCallback {
357 public:
358 	AllocationCollectorCallback(bool resetInfos)
359 		:
360 		fResetInfos(resetInfos)
361 	{
362 	}
363 
364 	virtual bool ProcessTrackingInfo(AllocationTrackingInfo* info,
365 		void* allocation, size_t allocationSize)
366 	{
367 		if (!info->IsInitialized())
368 			return true;
369 
370 		addr_t caller = 0;
371 		AbstractTraceEntryWithStackTrace* traceEntry = info->TraceEntry();
372 
373 		if (traceEntry != NULL && info->IsTraceEntryValid()) {
374 			caller = tracing_find_caller_in_stack_trace(
375 				traceEntry->StackTrace(), kSlabCodeAddressRanges,
376 				kSlabCodeAddressRangeCount);
377 		}
378 
379 		caller_info* callerInfo = get_caller_info(caller);
380 		if (callerInfo == NULL) {
381 			kprintf("out of space for caller infos\n");
382 			return false;
383 		}
384 
385 		callerInfo->count++;
386 		callerInfo->size += allocationSize;
387 
388 		if (fResetInfos)
389 			info->Clear();
390 
391 		return true;
392 	}
393 
394 private:
395 	bool	fResetInfos;
396 };
397 
398 
399 class AllocationInfoPrinterCallback : public AllocationTrackingCallback {
400 public:
401 	AllocationInfoPrinterCallback(bool printStackTrace, addr_t addressFilter,
402 		team_id teamFilter, thread_id threadFilter)
403 		:
404 		fPrintStackTrace(printStackTrace),
405 		fAddressFilter(addressFilter),
406 		fTeamFilter(teamFilter),
407 		fThreadFilter(threadFilter)
408 	{
409 	}
410 
411 	virtual bool ProcessTrackingInfo(AllocationTrackingInfo* info,
412 		void* allocation, size_t allocationSize)
413 	{
414 		if (!info->IsInitialized())
415 			return true;
416 
417 		if (fAddressFilter != 0 && (addr_t)allocation != fAddressFilter)
418 			return true;
419 
420 		AbstractTraceEntryWithStackTrace* traceEntry = info->TraceEntry();
421 		if (traceEntry != NULL && !info->IsTraceEntryValid())
422 			traceEntry = NULL;
423 
424 		if (traceEntry != NULL) {
425 			if (fTeamFilter != -1 && traceEntry->TeamID() != fTeamFilter)
426 				return true;
427 			if (fThreadFilter != -1 && traceEntry->ThreadID() != fThreadFilter)
428 				return true;
429 		} else {
430 			// we need the info if we have filters set
431 			if (fTeamFilter != -1 || fThreadFilter != -1)
432 				return true;
433 		}
434 
435 		kprintf("allocation %p, size: %" B_PRIuSIZE, allocation,
436 			allocationSize);
437 
438 		if (traceEntry != NULL) {
439 			kprintf(", team: %" B_PRId32 ", thread %" B_PRId32
440 				", time %" B_PRId64 "\n", traceEntry->TeamID(),
441 				traceEntry->ThreadID(), traceEntry->Time());
442 
443 			if (fPrintStackTrace)
444 				tracing_print_stack_trace(traceEntry->StackTrace());
445 		} else
446 			kprintf("\n");
447 
448 		return true;
449 	}
450 
451 private:
452 	bool		fPrintStackTrace;
453 	addr_t		fAddressFilter;
454 	team_id		fTeamFilter;
455 	thread_id	fThreadFilter;
456 };
457 
458 
459 class AllocationDetailPrinterCallback : public AllocationTrackingCallback {
460 public:
461 	AllocationDetailPrinterCallback(addr_t caller)
462 		:
463 		fCaller(caller)
464 	{
465 	}
466 
467 	virtual bool ProcessTrackingInfo(AllocationTrackingInfo* info,
468 		void* allocation, size_t allocationSize)
469 	{
470 		if (!info->IsInitialized())
471 			return true;
472 
473 		addr_t caller = 0;
474 		AbstractTraceEntryWithStackTrace* traceEntry = info->TraceEntry();
475 		if (traceEntry != NULL && !info->IsTraceEntryValid())
476 			traceEntry = NULL;
477 
478 		if (traceEntry != NULL) {
479 			caller = tracing_find_caller_in_stack_trace(
480 				traceEntry->StackTrace(), kSlabCodeAddressRanges,
481 				kSlabCodeAddressRangeCount);
482 		}
483 
484 		if (caller != fCaller)
485 			return true;
486 
487 		kprintf("allocation %p, size: %" B_PRIuSIZE "\n", allocation,
488 			allocationSize);
489 		if (traceEntry != NULL)
490 			tracing_print_stack_trace(traceEntry->StackTrace());
491 
492 		return true;
493 	}
494 
495 private:
496 	addr_t	fCaller;
497 };
498 
499 }	// unnamed namespace
500 
501 static caller_info*
502 get_caller_info(addr_t caller)
503 {
504 	// find the caller info
505 	for (int32 i = 0; i < sCallerInfoCount; i++) {
506 		if (caller == sCallerInfoTable[i].caller)
507 			return &sCallerInfoTable[i];
508 	}
509 
510 	// not found, add a new entry, if there are free slots
511 	if (sCallerInfoCount >= kCallerInfoTableSize)
512 		return NULL;
513 
514 	caller_info* info = &sCallerInfoTable[sCallerInfoCount++];
515 	info->caller = caller;
516 	info->count = 0;
517 	info->size = 0;
518 
519 	return info;
520 }
521 
522 
523 static int
524 caller_info_compare_size(const void* _a, const void* _b)
525 {
526 	const caller_info* a = (const caller_info*)_a;
527 	const caller_info* b = (const caller_info*)_b;
528 	return (int)(b->size - a->size);
529 }
530 
531 
532 static int
533 caller_info_compare_count(const void* _a, const void* _b)
534 {
535 	const caller_info* a = (const caller_info*)_a;
536 	const caller_info* b = (const caller_info*)_b;
537 	return (int)(b->count - a->count);
538 }
539 
540 
541 #if SLAB_OBJECT_CACHE_ALLOCATION_TRACKING
542 
543 static bool
544 analyze_allocation_callers(ObjectCache* cache, slab* slab,
545 	AllocationTrackingCallback& callback)
546 {
547 	for (uint32 i = 0; i < slab->size; i++) {
548 		if (!callback.ProcessTrackingInfo(&slab->tracking[i],
549 				cache->ObjectAtIndex(slab, i), cache->object_size)) {
550 			return false;
551 		}
552 	}
553 
554 	return true;
555 }
556 
557 
558 static bool
559 analyze_allocation_callers(ObjectCache* cache, const SlabList& slabList,
560 	AllocationTrackingCallback& callback)
561 {
562 	for (SlabList::ConstIterator it = slabList.GetIterator();
563 			slab* slab = it.Next();) {
564 		if (!analyze_allocation_callers(cache, slab, callback))
565 			return false;
566 	}
567 
568 	return true;
569 }
570 
571 
572 static bool
573 analyze_allocation_callers(ObjectCache* cache,
574 	AllocationTrackingCallback& callback)
575 {
576 	return analyze_allocation_callers(cache, cache->full, callback)
577 		&& analyze_allocation_callers(cache, cache->partial, callback);
578 }
579 
580 #endif	// SLAB_OBJECT_CACHE_ALLOCATION_TRACKING
581 
582 
583 static int
584 dump_allocation_infos(int argc, char **argv)
585 {
586 	ObjectCache* cache = NULL;
587 	slab* slab = NULL;
588 	addr_t addressFilter = 0;
589 	team_id teamFilter = -1;
590 	thread_id threadFilter = -1;
591 	bool printStackTraces = false;
592 
593 	for (int32 i = 1; i < argc; i++) {
594 		if (strcmp(argv[i], "--stacktrace") == 0)
595 			printStackTraces = true;
596 		else if (strcmp(argv[i], "-a") == 0) {
597 			uint64 address;
598 			if (++i >= argc
599 				|| !evaluate_debug_expression(argv[i], &address, true)) {
600 				print_debugger_command_usage(argv[0]);
601 				return 0;
602 			}
603 
604 			addressFilter = address;
605 		} else if (strcmp(argv[i], "-o") == 0) {
606 			uint64 cacheAddress;
607 			if (++i >= argc
608 				|| !evaluate_debug_expression(argv[i], &cacheAddress, true)) {
609 				print_debugger_command_usage(argv[0]);
610 				return 0;
611 			}
612 
613 			cache = (ObjectCache*)(addr_t)cacheAddress;
614 		} else if (strcasecmp(argv[i], "-s") == 0) {
615 			uint64 slabAddress;
616 			if (++i >= argc
617 				|| !evaluate_debug_expression(argv[i], &slabAddress, true)) {
618 				print_debugger_command_usage(argv[0]);
619 				return 0;
620 			}
621 
622 			void* slabPages = (void*)slabAddress;
623 			if (strcmp(argv[i], "-s") == 0) {
624 				slab = (struct slab*)(addr_t)slabAddress;
625 				slabPages = slab->pages;
626 			}
627 
628 			cache = MemoryManager::DebugObjectCacheForAddress(slabPages);
629 			if (cache == NULL) {
630 				kprintf("Couldn't find object cache for address %p.\n",
631 					slabPages);
632 				return 0;
633 			}
634 
635 			if (slab == NULL) {
636 				slab = cache->ObjectSlab(slabPages);
637 
638 				if (slab == NULL) {
639 					kprintf("Couldn't find slab for address %p.\n", slabPages);
640 					return 0;
641 				}
642 			}
643 		} else if (strcmp(argv[i], "--team") == 0) {
644 			uint64 team;
645 			if (++i >= argc
646 				|| !evaluate_debug_expression(argv[i], &team, true)) {
647 				print_debugger_command_usage(argv[0]);
648 				return 0;
649 			}
650 
651 			teamFilter = team;
652 		} else if (strcmp(argv[i], "--thread") == 0) {
653 			uint64 thread;
654 			if (++i >= argc
655 				|| !evaluate_debug_expression(argv[i], &thread, true)) {
656 				print_debugger_command_usage(argv[0]);
657 				return 0;
658 			}
659 
660 			threadFilter = thread;
661 		} else {
662 			print_debugger_command_usage(argv[0]);
663 			return 0;
664 		}
665 	}
666 
667 	AllocationInfoPrinterCallback callback(printStackTraces, addressFilter,
668 		teamFilter, threadFilter);
669 
670 	if (slab != NULL || cache != NULL) {
671 #if SLAB_OBJECT_CACHE_ALLOCATION_TRACKING
672 		if (slab != NULL) {
673 			if (!analyze_allocation_callers(cache, slab, callback))
674 				return 0;
675 		} else if (cache != NULL) {
676 			if (!analyze_allocation_callers(cache, callback))
677 				return 0;
678 		}
679 #else
680 		kprintf("Object cache allocation tracking not available. "
681 			"SLAB_OBJECT_CACHE_TRACING (%d) and "
682 			"SLAB_OBJECT_CACHE_TRACING_STACK_TRACE (%d) must be enabled.\n",
683 			SLAB_OBJECT_CACHE_TRACING, SLAB_OBJECT_CACHE_TRACING_STACK_TRACE);
684 		return 0;
685 #endif
686 	} else {
687 #if SLAB_OBJECT_CACHE_ALLOCATION_TRACKING
688 
689 		for (ObjectCacheList::Iterator it = sObjectCaches.GetIterator();
690 				it.HasNext();) {
691 			if (!analyze_allocation_callers(it.Next(), callback))
692 				return 0;
693 		}
694 #endif
695 
696 #if SLAB_MEMORY_MANAGER_ALLOCATION_TRACKING
697 		if (!MemoryManager::AnalyzeAllocationCallers(callback))
698 			return 0;
699 #endif
700 	}
701 
702 	return 0;
703 }
704 
705 
706 static int
707 dump_allocations_per_caller(int argc, char **argv)
708 {
709 	bool sortBySize = true;
710 	bool resetAllocationInfos = false;
711 	bool printDetails = false;
712 	ObjectCache* cache = NULL;
713 	addr_t caller = 0;
714 
715 	for (int32 i = 1; i < argc; i++) {
716 		if (strcmp(argv[i], "-c") == 0) {
717 			sortBySize = false;
718 		} else if (strcmp(argv[i], "-d") == 0) {
719 			uint64 callerAddress;
720 			if (++i >= argc
721 				|| !evaluate_debug_expression(argv[i], &callerAddress, true)) {
722 				print_debugger_command_usage(argv[0]);
723 				return 0;
724 			}
725 
726 			caller = callerAddress;
727 			printDetails = true;
728 		} else if (strcmp(argv[i], "-o") == 0) {
729 			uint64 cacheAddress;
730 			if (++i >= argc
731 				|| !evaluate_debug_expression(argv[i], &cacheAddress, true)) {
732 				print_debugger_command_usage(argv[0]);
733 				return 0;
734 			}
735 
736 			cache = (ObjectCache*)(addr_t)cacheAddress;
737 		} else if (strcmp(argv[i], "-r") == 0) {
738 			resetAllocationInfos = true;
739 		} else {
740 			print_debugger_command_usage(argv[0]);
741 			return 0;
742 		}
743 	}
744 
745 	sCallerInfoCount = 0;
746 
747 	AllocationCollectorCallback collectorCallback(resetAllocationInfos);
748 	AllocationDetailPrinterCallback detailsCallback(caller);
749 	AllocationTrackingCallback& callback = printDetails
750 		? (AllocationTrackingCallback&)detailsCallback
751 		: (AllocationTrackingCallback&)collectorCallback;
752 
753 	if (cache != NULL) {
754 #if SLAB_OBJECT_CACHE_ALLOCATION_TRACKING
755 		if (!analyze_allocation_callers(cache, callback))
756 			return 0;
757 #else
758 		kprintf("Object cache allocation tracking not available. "
759 			"SLAB_OBJECT_CACHE_TRACING (%d) and "
760 			"SLAB_OBJECT_CACHE_TRACING_STACK_TRACE (%d) must be enabled.\n",
761 			SLAB_OBJECT_CACHE_TRACING, SLAB_OBJECT_CACHE_TRACING_STACK_TRACE);
762 		return 0;
763 #endif
764 	} else {
765 #if SLAB_OBJECT_CACHE_ALLOCATION_TRACKING
766 
767 		for (ObjectCacheList::Iterator it = sObjectCaches.GetIterator();
768 				it.HasNext();) {
769 			if (!analyze_allocation_callers(it.Next(), callback))
770 				return 0;
771 		}
772 #endif
773 
774 #if SLAB_MEMORY_MANAGER_ALLOCATION_TRACKING
775 		if (!MemoryManager::AnalyzeAllocationCallers(callback))
776 			return 0;
777 #endif
778 	}
779 
780 	if (printDetails)
781 		return 0;
782 
783 	// sort the array
784 	qsort(sCallerInfoTable, sCallerInfoCount, sizeof(caller_info),
785 		sortBySize ? &caller_info_compare_size : &caller_info_compare_count);
786 
787 	kprintf("%" B_PRId32 " different callers, sorted by %s...\n\n",
788 		sCallerInfoCount, sortBySize ? "size" : "count");
789 
790 	size_t totalAllocationSize = 0;
791 	size_t totalAllocationCount = 0;
792 
793 	kprintf("     count        size      caller\n");
794 	kprintf("----------------------------------\n");
795 	for (int32 i = 0; i < sCallerInfoCount; i++) {
796 		caller_info& info = sCallerInfoTable[i];
797 		kprintf("%10" B_PRIuSIZE "  %10" B_PRIuSIZE "  %p", info.count,
798 			info.size, (void*)info.caller);
799 
800 		const char* symbol;
801 		const char* imageName;
802 		bool exactMatch;
803 		addr_t baseAddress;
804 
805 		if (elf_debug_lookup_symbol_address(info.caller, &baseAddress, &symbol,
806 				&imageName, &exactMatch) == B_OK) {
807 			kprintf("  %s + %#" B_PRIxADDR " (%s)%s\n", symbol,
808 				info.caller - baseAddress, imageName,
809 				exactMatch ? "" : " (nearest)");
810 		} else
811 			kprintf("\n");
812 
813 		totalAllocationCount += info.count;
814 		totalAllocationSize += info.size;
815 	}
816 
817 	kprintf("\ntotal allocations: %" B_PRIuSIZE ", %" B_PRIuSIZE " bytes\n",
818 		totalAllocationCount, totalAllocationSize);
819 
820 	return 0;
821 }
822 
823 #endif	// SLAB_ALLOCATION_TRACKING_AVAILABLE
824 
825 
826 void
827 add_alloc_tracing_entry(ObjectCache* cache, uint32 flags, void* object)
828 {
829 #if SLAB_OBJECT_CACHE_TRACING
830 #if SLAB_OBJECT_CACHE_ALLOCATION_TRACKING
831 	MutexLocker _(cache->lock);
832 	cache->TrackingInfoFor(object)->Init(T(Alloc(cache, flags, object)));
833 #else
834 	T(Alloc(cache, flags, object));
835 #endif
836 #endif
837 }
838 
839 
840 // #pragma mark -
841 
842 
843 void
844 request_memory_manager_maintenance()
845 {
846 	MutexLocker locker(sMaintenanceLock);
847 	sMaintenanceCondition.NotifyAll();
848 }
849 
850 
851 // #pragma mark -
852 
853 
854 static void
855 delete_object_cache_internal(object_cache* cache)
856 {
857 	if (!(cache->flags & CACHE_NO_DEPOT))
858 		object_depot_destroy(&cache->depot, 0);
859 
860 	mutex_lock(&cache->lock);
861 
862 	if (!cache->full.IsEmpty())
863 		panic("cache destroy: still has full slabs");
864 
865 	if (!cache->partial.IsEmpty())
866 		panic("cache destroy: still has partial slabs");
867 
868 	while (!cache->empty.IsEmpty())
869 		cache->ReturnSlab(cache->empty.RemoveHead(), 0);
870 
871 	mutex_destroy(&cache->lock);
872 	cache->Delete();
873 }
874 
875 
876 static void
877 increase_object_reserve(ObjectCache* cache)
878 {
879 	MutexLocker locker(sMaintenanceLock);
880 
881 	cache->maintenance_resize = true;
882 
883 	if (!cache->maintenance_pending) {
884 		cache->maintenance_pending = true;
885 		sMaintenanceQueue.Add(cache);
886 		sMaintenanceCondition.NotifyAll();
887 	}
888 }
889 
890 
891 /*!	Makes sure that \a objectCount objects can be allocated.
892 */
893 static status_t
894 object_cache_reserve_internal(ObjectCache* cache, size_t objectCount,
895 	uint32 flags)
896 {
897 	// If someone else is already adding slabs, we wait for that to be finished
898 	// first.
899 	thread_id thread = find_thread(NULL);
900 	while (true) {
901 		if (objectCount <= cache->total_objects - cache->used_count)
902 			return B_OK;
903 
904 		ObjectCacheResizeEntry* resizeEntry = NULL;
905 		if (cache->resize_entry_dont_wait != NULL) {
906 			resizeEntry = cache->resize_entry_dont_wait;
907 			if (thread == resizeEntry->thread)
908 				return B_WOULD_BLOCK;
909 			// Note: We could still have reentered the function, i.e.
910 			// resize_entry_can_wait would be ours. That doesn't matter much,
911 			// though, since after the don't-wait thread has done its job
912 			// everyone will be happy.
913 		} else if (cache->resize_entry_can_wait != NULL) {
914 			resizeEntry = cache->resize_entry_can_wait;
915 			if (thread == resizeEntry->thread)
916 				return B_WOULD_BLOCK;
917 
918 			if ((flags & CACHE_DONT_WAIT_FOR_MEMORY) != 0)
919 				break;
920 		} else
921 			break;
922 
923 		ConditionVariableEntry entry;
924 		resizeEntry->condition.Add(&entry);
925 
926 		cache->Unlock();
927 		entry.Wait();
928 		cache->Lock();
929 	}
930 
931 	// prepare the resize entry others can wait on
932 	ObjectCacheResizeEntry*& resizeEntry
933 		= (flags & CACHE_DONT_WAIT_FOR_MEMORY) != 0
934 			? cache->resize_entry_dont_wait : cache->resize_entry_can_wait;
935 
936 	ObjectCacheResizeEntry myResizeEntry;
937 	resizeEntry = &myResizeEntry;
938 	resizeEntry->condition.Init(cache, "wait for slabs");
939 	resizeEntry->thread = thread;
940 
941 	// add new slabs until there are as many free ones as requested
942 	while (objectCount > cache->total_objects - cache->used_count) {
943 		slab* newSlab = cache->CreateSlab(flags);
944 		if (newSlab == NULL) {
945 			resizeEntry->condition.NotifyAll();
946 			resizeEntry = NULL;
947 			return B_NO_MEMORY;
948 		}
949 
950 		cache->usage += cache->slab_size;
951 		cache->total_objects += newSlab->size;
952 
953 		cache->empty.Add(newSlab);
954 		cache->empty_count++;
955 	}
956 
957 	resizeEntry->condition.NotifyAll();
958 	resizeEntry = NULL;
959 
960 	return B_OK;
961 }
962 
963 
964 static void
965 object_cache_low_memory(void* dummy, uint32 resources, int32 level)
966 {
967 	if (level == B_NO_LOW_RESOURCE)
968 		return;
969 
970 	MutexLocker cacheListLocker(sObjectCacheListLock);
971 
972 	// Append the first cache to the end of the queue. We assume that it is
973 	// one of the caches that will never be deleted and thus we use it as a
974 	// marker.
975 	ObjectCache* firstCache = sObjectCaches.RemoveHead();
976 	sObjectCaches.Add(firstCache);
977 	cacheListLocker.Unlock();
978 
979 	ObjectCache* cache;
980 	do {
981 		cacheListLocker.Lock();
982 
983 		cache = sObjectCaches.RemoveHead();
984 		sObjectCaches.Add(cache);
985 
986 		MutexLocker maintenanceLocker(sMaintenanceLock);
987 		if (cache->maintenance_pending || cache->maintenance_in_progress) {
988 			// We don't want to mess with caches in maintenance.
989 			continue;
990 		}
991 
992 		cache->maintenance_pending = true;
993 		cache->maintenance_in_progress = true;
994 
995 		maintenanceLocker.Unlock();
996 		cacheListLocker.Unlock();
997 
998 		// We are calling the reclaimer without the object cache lock
999 		// to give the owner a chance to return objects to the slabs.
1000 
1001 		if (cache->reclaimer)
1002 			cache->reclaimer(cache->cookie, level);
1003 
1004 		if ((cache->flags & CACHE_NO_DEPOT) == 0)
1005 			object_depot_make_empty(&cache->depot, 0);
1006 
1007 		MutexLocker cacheLocker(cache->lock);
1008 		size_t minimumAllowed;
1009 
1010 		switch (level) {
1011 			case B_LOW_RESOURCE_NOTE:
1012 				minimumAllowed = cache->pressure / 2 + 1;
1013 				cache->pressure -= cache->pressure / 8;
1014 				break;
1015 
1016 			case B_LOW_RESOURCE_WARNING:
1017 				cache->pressure /= 2;
1018 				minimumAllowed = 0;
1019 				break;
1020 
1021 			default:
1022 				cache->pressure = 0;
1023 				minimumAllowed = 0;
1024 				break;
1025 		}
1026 
1027 		while (cache->empty_count > minimumAllowed) {
1028 			// make sure we respect the cache's minimum object reserve
1029 			size_t objectsPerSlab = cache->empty.Head()->size;
1030 			size_t freeObjects = cache->total_objects - cache->used_count;
1031 			if (freeObjects < cache->min_object_reserve + objectsPerSlab)
1032 				break;
1033 
1034 			cache->ReturnSlab(cache->empty.RemoveHead(), 0);
1035 			cache->empty_count--;
1036 		}
1037 
1038 		cacheLocker.Unlock();
1039 
1040 		// Check whether in the meantime someone has really requested
1041 		// maintenance for the cache.
1042 		maintenanceLocker.Lock();
1043 
1044 		if (cache->maintenance_delete) {
1045 			delete_object_cache_internal(cache);
1046 			continue;
1047 		}
1048 
1049 		cache->maintenance_in_progress = false;
1050 
1051 		if (cache->maintenance_resize)
1052 			sMaintenanceQueue.Add(cache);
1053 		else
1054 			cache->maintenance_pending = false;
1055 	} while (cache != firstCache);
1056 }
1057 
1058 
1059 static status_t
1060 object_cache_maintainer(void*)
1061 {
1062 	while (true) {
1063 		MutexLocker locker(sMaintenanceLock);
1064 
1065 		// wait for the next request
1066 		while (sMaintenanceQueue.IsEmpty()) {
1067 			// perform memory manager maintenance, if needed
1068 			if (MemoryManager::MaintenanceNeeded()) {
1069 				locker.Unlock();
1070 				MemoryManager::PerformMaintenance();
1071 				locker.Lock();
1072 				continue;
1073 			}
1074 
1075 			ConditionVariableEntry entry;
1076 			sMaintenanceCondition.Add(&entry);
1077 			locker.Unlock();
1078 			entry.Wait();
1079 			locker.Lock();
1080 		}
1081 
1082 		ObjectCache* cache = sMaintenanceQueue.RemoveHead();
1083 
1084 		while (true) {
1085 			bool resizeRequested = cache->maintenance_resize;
1086 			bool deleteRequested = cache->maintenance_delete;
1087 
1088 			if (!resizeRequested && !deleteRequested) {
1089 				cache->maintenance_pending = false;
1090 				cache->maintenance_in_progress = false;
1091 				break;
1092 			}
1093 
1094 			cache->maintenance_resize = false;
1095 			cache->maintenance_in_progress = true;
1096 
1097 			locker.Unlock();
1098 
1099 			if (deleteRequested) {
1100 				delete_object_cache_internal(cache);
1101 				break;
1102 			}
1103 
1104 			// resize the cache, if necessary
1105 
1106 			MutexLocker cacheLocker(cache->lock);
1107 
1108 			if (resizeRequested) {
1109 				status_t error = object_cache_reserve_internal(cache,
1110 					cache->min_object_reserve, 0);
1111 				if (error != B_OK) {
1112 					dprintf("object cache resizer: Failed to resize object "
1113 						"cache %p!\n", cache);
1114 					break;
1115 				}
1116 			}
1117 
1118 			locker.Lock();
1119 		}
1120 	}
1121 
1122 	// never can get here
1123 	return B_OK;
1124 }
1125 
1126 
1127 // #pragma mark - public API
1128 
1129 
1130 object_cache*
1131 create_object_cache(const char* name, size_t object_size, size_t alignment,
1132 	void* cookie, object_cache_constructor constructor,
1133 	object_cache_destructor destructor)
1134 {
1135 	return create_object_cache_etc(name, object_size, alignment, 0, 0, 0, 0,
1136 		cookie, constructor, destructor, NULL);
1137 }
1138 
1139 
1140 object_cache*
1141 create_object_cache_etc(const char* name, size_t objectSize, size_t alignment,
1142 	size_t maximum, size_t magazineCapacity, size_t maxMagazineCount,
1143 	uint32 flags, void* cookie, object_cache_constructor constructor,
1144 	object_cache_destructor destructor, object_cache_reclaimer reclaimer)
1145 {
1146 	ObjectCache* cache;
1147 
1148 	if (objectSize == 0) {
1149 		cache = NULL;
1150 	} else if (objectSize <= 256) {
1151 		cache = SmallObjectCache::Create(name, objectSize, alignment, maximum,
1152 			magazineCapacity, maxMagazineCount, flags, cookie, constructor,
1153 			destructor, reclaimer);
1154 	} else {
1155 		cache = HashedObjectCache::Create(name, objectSize, alignment, maximum,
1156 			magazineCapacity, maxMagazineCount, flags, cookie, constructor,
1157 			destructor, reclaimer);
1158 	}
1159 
1160 	if (cache != NULL) {
1161 		MutexLocker _(sObjectCacheListLock);
1162 		sObjectCaches.Add(cache);
1163 	}
1164 
1165 	T(Create(name, objectSize, alignment, maximum, flags, cookie, cache));
1166 	return cache;
1167 }
1168 
1169 
1170 void
1171 delete_object_cache(object_cache* cache)
1172 {
1173 	T(Delete(cache));
1174 
1175 	{
1176 		MutexLocker _(sObjectCacheListLock);
1177 		sObjectCaches.Remove(cache);
1178 	}
1179 
1180 	MutexLocker cacheLocker(cache->lock);
1181 
1182 	{
1183 		MutexLocker maintenanceLocker(sMaintenanceLock);
1184 		if (cache->maintenance_in_progress) {
1185 			// The maintainer thread is working with the cache. Just mark it
1186 			// to be deleted.
1187 			cache->maintenance_delete = true;
1188 			return;
1189 		}
1190 
1191 		// unschedule maintenance
1192 		if (cache->maintenance_pending)
1193 			sMaintenanceQueue.Remove(cache);
1194 	}
1195 
1196 	// at this point no-one should have a reference to the cache anymore
1197 	cacheLocker.Unlock();
1198 
1199 	delete_object_cache_internal(cache);
1200 }
1201 
1202 
1203 status_t
1204 object_cache_set_minimum_reserve(object_cache* cache, size_t objectCount)
1205 {
1206 	MutexLocker _(cache->lock);
1207 
1208 	if (cache->min_object_reserve == objectCount)
1209 		return B_OK;
1210 
1211 	cache->min_object_reserve = objectCount;
1212 
1213 	increase_object_reserve(cache);
1214 
1215 	return B_OK;
1216 }
1217 
1218 
1219 void*
1220 object_cache_alloc(object_cache* cache, uint32 flags)
1221 {
1222 	if (!(cache->flags & CACHE_NO_DEPOT)) {
1223 		void* object = object_depot_obtain(&cache->depot);
1224 		if (object) {
1225 			add_alloc_tracing_entry(cache, flags, object);
1226 			return fill_allocated_block(object, cache->object_size);
1227 		}
1228 	}
1229 
1230 	MutexLocker locker(cache->lock);
1231 	slab* source = NULL;
1232 
1233 	while (true) {
1234 		source = cache->partial.Head();
1235 		if (source != NULL)
1236 			break;
1237 
1238 		source = cache->empty.RemoveHead();
1239 		if (source != NULL) {
1240 			cache->empty_count--;
1241 			cache->partial.Add(source);
1242 			break;
1243 		}
1244 
1245 		if (object_cache_reserve_internal(cache, 1, flags) != B_OK) {
1246 			T(Alloc(cache, flags, NULL));
1247 			return NULL;
1248 		}
1249 
1250 		cache->pressure++;
1251 	}
1252 
1253 	ParanoiaChecker _2(source);
1254 
1255 	object_link* link = _pop(source->free);
1256 	source->count--;
1257 	cache->used_count++;
1258 
1259 	if (cache->total_objects - cache->used_count < cache->min_object_reserve)
1260 		increase_object_reserve(cache);
1261 
1262 	REMOVE_PARANOIA_CHECK(PARANOIA_SUSPICIOUS, source, &link->next,
1263 		sizeof(void*));
1264 
1265 	TRACE_CACHE(cache, "allocate %p (%p) from %p, %lu remaining.",
1266 		link_to_object(link, cache->object_size), link, source, source->count);
1267 
1268 	if (source->count == 0) {
1269 		cache->partial.Remove(source);
1270 		cache->full.Add(source);
1271 	}
1272 
1273 	void* object = link_to_object(link, cache->object_size);
1274 	locker.Unlock();
1275 
1276 	add_alloc_tracing_entry(cache, flags, object);
1277 	return fill_allocated_block(object, cache->object_size);
1278 }
1279 
1280 
1281 void
1282 object_cache_free(object_cache* cache, void* object, uint32 flags)
1283 {
1284 	if (object == NULL)
1285 		return;
1286 
1287 	T(Free(cache, object));
1288 
1289 #if PARANOID_KERNEL_FREE
1290 	// TODO: allow forcing the check even if we don't find deadbeef
1291 	if (*(uint32*)object == 0xdeadbeef) {
1292 		if (!cache->AssertObjectNotFreed(object))
1293 			return;
1294 
1295 		if ((cache->flags & CACHE_NO_DEPOT) == 0) {
1296 			if (object_depot_contains_object(&cache->depot, object)) {
1297 				panic("object_cache: object %p is already freed", object);
1298 				return;
1299 			}
1300 		}
1301 	}
1302 
1303 	fill_freed_block(object, cache->object_size);
1304 #endif
1305 
1306 #if SLAB_OBJECT_CACHE_ALLOCATION_TRACKING
1307 	mutex_lock(&cache->lock);
1308 	cache->TrackingInfoFor(object)->Clear();
1309 	mutex_unlock(&cache->lock);
1310 #endif
1311 
1312 	if ((cache->flags & CACHE_NO_DEPOT) == 0) {
1313 		object_depot_store(&cache->depot, object, flags);
1314 		return;
1315 	}
1316 
1317 	MutexLocker _(cache->lock);
1318 	cache->ReturnObjectToSlab(cache->ObjectSlab(object), object, flags);
1319 }
1320 
1321 
1322 status_t
1323 object_cache_reserve(object_cache* cache, size_t objectCount, uint32 flags)
1324 {
1325 	if (objectCount == 0)
1326 		return B_OK;
1327 
1328 	T(Reserve(cache, objectCount, flags));
1329 
1330 	MutexLocker _(cache->lock);
1331 	return object_cache_reserve_internal(cache, objectCount, flags);
1332 }
1333 
1334 
1335 void
1336 object_cache_get_usage(object_cache* cache, size_t* _allocatedMemory)
1337 {
1338 	MutexLocker _(cache->lock);
1339 	*_allocatedMemory = cache->usage;
1340 }
1341 
1342 
1343 void
1344 slab_init(kernel_args* args)
1345 {
1346 	MemoryManager::Init(args);
1347 
1348 	new (&sObjectCaches) ObjectCacheList();
1349 
1350 	block_allocator_init_boot();
1351 }
1352 
1353 
1354 void
1355 slab_init_post_area()
1356 {
1357 	MemoryManager::InitPostArea();
1358 
1359 	add_debugger_command("slabs", dump_slabs, "list all object caches");
1360 	add_debugger_command("slab_cache", dump_cache_info,
1361 		"dump information about a specific object cache");
1362 	add_debugger_command("slab_depot", dump_object_depot,
1363 		"dump contents of an object depot");
1364 	add_debugger_command("slab_magazine", dump_depot_magazine,
1365 		"dump contents of a depot magazine");
1366 #if SLAB_ALLOCATION_TRACKING_AVAILABLE
1367 	add_debugger_command_etc("allocations_per_caller",
1368 		&dump_allocations_per_caller,
1369 		"Dump current slab allocations summed up per caller",
1370 		"[ -c ] [ -d <caller> ] [ -o <object cache> ] [ -r ]\n"
1371 		"The current allocations will by summed up by caller (their count and\n"
1372 		"size) printed in decreasing order by size or, if \"-c\" is\n"
1373 		"specified, by allocation count. If given <object cache> specifies\n"
1374 		"the address of the object cache for which to print the allocations.\n"
1375 		"If \"-d\" is given, each allocation for caller <caller> is printed\n"
1376 		"including the respective stack trace.\n"
1377 		"If \"-r\" is given, the allocation infos are reset after gathering\n"
1378 		"the information, so the next command invocation will only show the\n"
1379 		"allocations made after the reset.\n", 0);
1380 	add_debugger_command_etc("allocation_infos",
1381 		&dump_allocation_infos,
1382 		"Dump current slab allocations",
1383 		"[ --stacktrace ] [ -o <object cache> | -s <slab> | -S <address> ] "
1384 		"[ -a <allocation> ] [ --team <team ID> ] [ --thread <thread ID> ]\n"
1385 		"The current allocations filtered by optional values will be printed.\n"
1386 		"If given, <object cache> specifies the address of the object cache\n"
1387 		"or <slab> specifies the address of a slab, for which to print the\n"
1388 		"allocations. Alternatively <address> specifies any address within\n"
1389 		"a slab allocation range.\n"
1390 		"The optional \"-a\" address filters for a specific allocation,\n"
1391 		"with \"--team\" and \"--thread\" allocations by specific teams\n"
1392 		"and/or threads can be filtered (these only work if a corresponding\n"
1393 		"tracing entry is still available).\n"
1394 		"If \"--stacktrace\" is given, then stack traces of the allocation\n"
1395 		"callers are printed, where available\n", 0);
1396 #endif	// SLAB_ALLOCATION_TRACKING_AVAILABLE
1397 }
1398 
1399 
1400 void
1401 slab_init_post_sem()
1402 {
1403 	register_low_resource_handler(object_cache_low_memory, NULL,
1404 		B_KERNEL_RESOURCE_PAGES | B_KERNEL_RESOURCE_MEMORY
1405 			| B_KERNEL_RESOURCE_ADDRESS_SPACE, 5);
1406 
1407 	block_allocator_init_rest();
1408 }
1409 
1410 
1411 void
1412 slab_init_post_thread()
1413 {
1414 	new(&sMaintenanceQueue) MaintenanceQueue;
1415 	sMaintenanceCondition.Init(&sMaintenanceQueue, "object cache maintainer");
1416 
1417 	thread_id objectCacheResizer = spawn_kernel_thread(object_cache_maintainer,
1418 		"object cache resizer", B_URGENT_PRIORITY, NULL);
1419 	if (objectCacheResizer < 0) {
1420 		panic("slab_init_post_thread(): failed to spawn object cache resizer "
1421 			"thread\n");
1422 		return;
1423 	}
1424 
1425 	resume_thread(objectCacheResizer);
1426 }
1427 
1428 
1429 RANGE_MARKER_FUNCTION_END(Slab)
1430 
1431 
1432 #endif	// !USE_GUARDED_HEAP_FOR_OBJECT_CACHE
1433