xref: /titanic_50/usr/src/lib/libumem/common/vmem.c (revision 2eeaed14a5e2ed9bd811643ad5bffc3510ca0310)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 
22 /*
23  * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
24  * Use is subject to license terms.
25  */
26 
27 #pragma ident	"%Z%%M%	%I%	%E% SMI"
28 
29 /*
30  * For a more complete description of the main ideas, see:
31  *
32  *	Jeff Bonwick and Jonathan Adams,
33  *
34  *	Magazines and vmem: Extending the Slab Allocator to Many CPUs and
35  *	Arbitrary Resources.
36  *
37  *	Proceedings of the 2001 Usenix Conference.
38  *	Available as /shared/sac/PSARC/2000/550/materials/vmem.pdf.
39  *
40  * For the "Big Theory Statement", see usr/src/uts/common/os/vmem.c
41  *
42  * 1. Overview of changes
43  * ------------------------------
44  * There have been a few changes to vmem in order to support umem.  The
45  * main areas are:
46  *
47  *	* VM_SLEEP unsupported
48  *
49  *	* Reaping changes
50  *
51  *	* initialization changes
52  *
53  *	* _vmem_extend_alloc
54  *
55  *
56  * 2. VM_SLEEP Removed
57  * -------------------
58  * Since VM_SLEEP allocations can hold locks (in vmem_populate()) for
59  * possibly infinite amounts of time, they are not supported in this
60  * version of vmem.  Sleep-like behavior can be achieved through
61  * UMEM_NOFAIL umem allocations.
62  *
63  *
64  * 3. Reaping changes
65  * ------------------
66  * Unlike kmem_reap(), which just asynchronously schedules work, umem_reap()
67  * can do allocations and frees synchronously.  This is a problem if it
68  * occurs during a vmem_populate() allocation.
69  *
70  * Instead, we delay reaps while populates are active.
71  *
72  *
73  * 4. Initialization changes
74  * -------------------------
75  * In the kernel, vmem_init() allows you to create a single, top-level arena,
76  * which has vmem_internal_arena as a child.  For umem, we want to be able
77  * to extend arenas dynamically.  It is much easier to support this if we
78  * allow a two-level "heap" arena:
79  *
80  *	+----------+
81  *	|  "fake"  |
82  *	+----------+
83  *	      |
84  *	+----------+
85  *	|  "heap"  |
86  *	+----------+
87  *	  |    \ \
88  *	  |     +-+-- ... <other children>
89  *	  |
90  *	+---------------+
91  *	| vmem_internal |
92  *	+---------------+
93  *	    | | | |
94  *	   <children>
95  *
96  * The new vmem_init() allows you to specify a "parent" of the heap, along
97  * with allocation functions.
98  *
99  *
100  * 5. _vmem_extend_alloc
101  * ---------------------
102  * The other part of extending is _vmem_extend_alloc.  This function allows
103  * you to extend (expand current spans, if possible) an arena and allocate
104  * a chunk of the newly extened span atomically.  This is needed to support
105  * extending the heap while vmem_populate()ing it.
106  *
107  * In order to increase the usefulness of extending, non-imported spans are
108  * sorted in address order.
109  */
110 
111 #include "c_synonyms.h"
112 #include <sys/vmem_impl_user.h>
113 #include <alloca.h>
114 #include <sys/sysmacros.h>
115 #include <stdio.h>
116 #include <strings.h>
117 #include <atomic.h>
118 
119 #include "vmem_base.h"
120 #include "umem_base.h"
121 
122 #define	VMEM_INITIAL		6	/* early vmem arenas */
123 #define	VMEM_SEG_INITIAL	100	/* early segments */
124 
125 /*
126  * Adding a new span to an arena requires two segment structures: one to
127  * represent the span, and one to represent the free segment it contains.
128  */
129 #define	VMEM_SEGS_PER_SPAN_CREATE	2
130 
131 /*
132  * Allocating a piece of an existing segment requires 0-2 segment structures
133  * depending on how much of the segment we're allocating.
134  *
135  * To allocate the entire segment, no new segment structures are needed; we
136  * simply move the existing segment structure from the freelist to the
137  * allocation hash table.
138  *
139  * To allocate a piece from the left or right end of the segment, we must
140  * split the segment into two pieces (allocated part and remainder), so we
141  * need one new segment structure to represent the remainder.
142  *
143  * To allocate from the middle of a segment, we need two new segment strucures
144  * to represent the remainders on either side of the allocated part.
145  */
146 #define	VMEM_SEGS_PER_EXACT_ALLOC	0
147 #define	VMEM_SEGS_PER_LEFT_ALLOC	1
148 #define	VMEM_SEGS_PER_RIGHT_ALLOC	1
149 #define	VMEM_SEGS_PER_MIDDLE_ALLOC	2
150 
151 /*
152  * vmem_populate() preallocates segment structures for vmem to do its work.
153  * It must preallocate enough for the worst case, which is when we must import
154  * a new span and then allocate from the middle of it.
155  */
156 #define	VMEM_SEGS_PER_ALLOC_MAX		\
157 	(VMEM_SEGS_PER_SPAN_CREATE + VMEM_SEGS_PER_MIDDLE_ALLOC)
158 
159 /*
160  * The segment structures themselves are allocated from vmem_seg_arena, so
161  * we have a recursion problem when vmem_seg_arena needs to populate itself.
162  * We address this by working out the maximum number of segment structures
163  * this act will require, and multiplying by the maximum number of threads
164  * that we'll allow to do it simultaneously.
165  *
166  * The worst-case segment consumption to populate vmem_seg_arena is as
167  * follows (depicted as a stack trace to indicate why events are occurring):
168  *
169  * vmem_alloc(vmem_seg_arena)		-> 2 segs (span create + exact alloc)
170  *  vmem_alloc(vmem_internal_arena)	-> 2 segs (span create + exact alloc)
171  *   heap_alloc(heap_arena)
172  *    vmem_alloc(heap_arena)		-> 4 seg (span create + alloc)
173  *     parent_alloc(parent_arena)
174  *	_vmem_extend_alloc(parent_arena) -> 3 seg (span create + left alloc)
175  *
176  * Note:  The reservation for heap_arena must be 4, since vmem_xalloc()
177  * is overly pessimistic on allocations where parent_arena has a stricter
178  * alignment than heap_arena.
179  *
180  * The worst-case consumption for any arena is 4 segment structures.
181  * For now, we only support VM_NOSLEEP allocations, so as long as we
182  * serialize all vmem_populates, a 4-seg reserve is sufficient.
183  */
184 #define	VMEM_POPULATE_SEGS_PER_ARENA	4
185 #define	VMEM_POPULATE_LOCKS		1
186 
187 #define	VMEM_POPULATE_RESERVE		\
188 	(VMEM_POPULATE_SEGS_PER_ARENA * VMEM_POPULATE_LOCKS)
189 
190 /*
191  * vmem_populate() ensures that each arena has VMEM_MINFREE seg structures
192  * so that it can satisfy the worst-case allocation *and* participate in
193  * worst-case allocation from vmem_seg_arena.
194  */
195 #define	VMEM_MINFREE	(VMEM_POPULATE_RESERVE + VMEM_SEGS_PER_ALLOC_MAX)
196 
197 /* Don't assume new statics are zeroed - see vmem_startup() */
198 static vmem_t vmem0[VMEM_INITIAL];
199 static vmem_t *vmem_populator[VMEM_INITIAL];
200 static uint32_t vmem_id;
201 static uint32_t vmem_populators;
202 static vmem_seg_t vmem_seg0[VMEM_SEG_INITIAL];
203 static vmem_seg_t *vmem_segfree;
204 static mutex_t vmem_list_lock;
205 static mutex_t vmem_segfree_lock;
206 static vmem_populate_lock_t vmem_nosleep_lock;
207 #define	IN_POPULATE()	(vmem_nosleep_lock.vmpl_thr == thr_self())
208 static vmem_t *vmem_list;
209 static vmem_t *vmem_internal_arena;
210 static vmem_t *vmem_seg_arena;
211 static vmem_t *vmem_hash_arena;
212 static vmem_t *vmem_vmem_arena;
213 
214 vmem_t *vmem_heap;
215 vmem_alloc_t *vmem_heap_alloc;
216 vmem_free_t *vmem_heap_free;
217 
218 uint32_t vmem_mtbf;		/* mean time between failures [default: off] */
219 size_t vmem_seg_size = sizeof (vmem_seg_t);
220 
221 /*
222  * Insert/delete from arena list (type 'a') or next-of-kin list (type 'k').
223  */
224 #define	VMEM_INSERT(vprev, vsp, type)					\
225 {									\
226 	vmem_seg_t *vnext = (vprev)->vs_##type##next;			\
227 	(vsp)->vs_##type##next = (vnext);				\
228 	(vsp)->vs_##type##prev = (vprev);				\
229 	(vprev)->vs_##type##next = (vsp);				\
230 	(vnext)->vs_##type##prev = (vsp);				\
231 }
232 
233 #define	VMEM_DELETE(vsp, type)						\
234 {									\
235 	vmem_seg_t *vprev = (vsp)->vs_##type##prev;			\
236 	vmem_seg_t *vnext = (vsp)->vs_##type##next;			\
237 	(vprev)->vs_##type##next = (vnext);				\
238 	(vnext)->vs_##type##prev = (vprev);				\
239 }
240 
241 /*
242  * Get a vmem_seg_t from the global segfree list.
243  */
244 static vmem_seg_t *
245 vmem_getseg_global(void)
246 {
247 	vmem_seg_t *vsp;
248 
249 	(void) mutex_lock(&vmem_segfree_lock);
250 	if ((vsp = vmem_segfree) != NULL)
251 		vmem_segfree = vsp->vs_knext;
252 	(void) mutex_unlock(&vmem_segfree_lock);
253 
254 	return (vsp);
255 }
256 
257 /*
258  * Put a vmem_seg_t on the global segfree list.
259  */
260 static void
261 vmem_putseg_global(vmem_seg_t *vsp)
262 {
263 	(void) mutex_lock(&vmem_segfree_lock);
264 	vsp->vs_knext = vmem_segfree;
265 	vmem_segfree = vsp;
266 	(void) mutex_unlock(&vmem_segfree_lock);
267 }
268 
269 /*
270  * Get a vmem_seg_t from vmp's segfree list.
271  */
272 static vmem_seg_t *
273 vmem_getseg(vmem_t *vmp)
274 {
275 	vmem_seg_t *vsp;
276 
277 	ASSERT(vmp->vm_nsegfree > 0);
278 
279 	vsp = vmp->vm_segfree;
280 	vmp->vm_segfree = vsp->vs_knext;
281 	vmp->vm_nsegfree--;
282 
283 	return (vsp);
284 }
285 
286 /*
287  * Put a vmem_seg_t on vmp's segfree list.
288  */
289 static void
290 vmem_putseg(vmem_t *vmp, vmem_seg_t *vsp)
291 {
292 	vsp->vs_knext = vmp->vm_segfree;
293 	vmp->vm_segfree = vsp;
294 	vmp->vm_nsegfree++;
295 }
296 
297 /*
298  * Add vsp to the appropriate freelist.
299  */
300 static void
301 vmem_freelist_insert(vmem_t *vmp, vmem_seg_t *vsp)
302 {
303 	vmem_seg_t *vprev;
304 
305 	ASSERT(*VMEM_HASH(vmp, vsp->vs_start) != vsp);
306 
307 	vprev = (vmem_seg_t *)&vmp->vm_freelist[highbit(VS_SIZE(vsp)) - 1];
308 	vsp->vs_type = VMEM_FREE;
309 	vmp->vm_freemap |= VS_SIZE(vprev);
310 	VMEM_INSERT(vprev, vsp, k);
311 
312 	(void) cond_broadcast(&vmp->vm_cv);
313 }
314 
315 /*
316  * Take vsp from the freelist.
317  */
318 static void
319 vmem_freelist_delete(vmem_t *vmp, vmem_seg_t *vsp)
320 {
321 	ASSERT(*VMEM_HASH(vmp, vsp->vs_start) != vsp);
322 	ASSERT(vsp->vs_type == VMEM_FREE);
323 
324 	if (vsp->vs_knext->vs_start == 0 && vsp->vs_kprev->vs_start == 0) {
325 		/*
326 		 * The segments on both sides of 'vsp' are freelist heads,
327 		 * so taking vsp leaves the freelist at vsp->vs_kprev empty.
328 		 */
329 		ASSERT(vmp->vm_freemap & VS_SIZE(vsp->vs_kprev));
330 		vmp->vm_freemap ^= VS_SIZE(vsp->vs_kprev);
331 	}
332 	VMEM_DELETE(vsp, k);
333 }
334 
335 /*
336  * Add vsp to the allocated-segment hash table and update kstats.
337  */
338 static void
339 vmem_hash_insert(vmem_t *vmp, vmem_seg_t *vsp)
340 {
341 	vmem_seg_t **bucket;
342 
343 	vsp->vs_type = VMEM_ALLOC;
344 	bucket = VMEM_HASH(vmp, vsp->vs_start);
345 	vsp->vs_knext = *bucket;
346 	*bucket = vsp;
347 
348 	if (vmem_seg_size == sizeof (vmem_seg_t)) {
349 		vsp->vs_depth = (uint8_t)getpcstack(vsp->vs_stack,
350 		    VMEM_STACK_DEPTH, 0);
351 		vsp->vs_thread = thr_self();
352 		vsp->vs_timestamp = gethrtime();
353 	} else {
354 		vsp->vs_depth = 0;
355 	}
356 
357 	vmp->vm_kstat.vk_alloc++;
358 	vmp->vm_kstat.vk_mem_inuse += VS_SIZE(vsp);
359 }
360 
361 /*
362  * Remove vsp from the allocated-segment hash table and update kstats.
363  */
364 static vmem_seg_t *
365 vmem_hash_delete(vmem_t *vmp, uintptr_t addr, size_t size)
366 {
367 	vmem_seg_t *vsp, **prev_vspp;
368 
369 	prev_vspp = VMEM_HASH(vmp, addr);
370 	while ((vsp = *prev_vspp) != NULL) {
371 		if (vsp->vs_start == addr) {
372 			*prev_vspp = vsp->vs_knext;
373 			break;
374 		}
375 		vmp->vm_kstat.vk_lookup++;
376 		prev_vspp = &vsp->vs_knext;
377 	}
378 
379 	if (vsp == NULL) {
380 		umem_panic("vmem_hash_delete(%p, %lx, %lu): bad free",
381 		    vmp, addr, size);
382 	}
383 	if (VS_SIZE(vsp) != size) {
384 		umem_panic("vmem_hash_delete(%p, %lx, %lu): wrong size "
385 		    "(expect %lu)", vmp, addr, size, VS_SIZE(vsp));
386 	}
387 
388 	vmp->vm_kstat.vk_free++;
389 	vmp->vm_kstat.vk_mem_inuse -= size;
390 
391 	return (vsp);
392 }
393 
394 /*
395  * Create a segment spanning the range [start, end) and add it to the arena.
396  */
397 static vmem_seg_t *
398 vmem_seg_create(vmem_t *vmp, vmem_seg_t *vprev, uintptr_t start, uintptr_t end)
399 {
400 	vmem_seg_t *newseg = vmem_getseg(vmp);
401 
402 	newseg->vs_start = start;
403 	newseg->vs_end = end;
404 	newseg->vs_type = 0;
405 	newseg->vs_import = 0;
406 
407 	VMEM_INSERT(vprev, newseg, a);
408 
409 	return (newseg);
410 }
411 
412 /*
413  * Remove segment vsp from the arena.
414  */
415 static void
416 vmem_seg_destroy(vmem_t *vmp, vmem_seg_t *vsp)
417 {
418 	ASSERT(vsp->vs_type != VMEM_ROTOR);
419 	VMEM_DELETE(vsp, a);
420 
421 	vmem_putseg(vmp, vsp);
422 }
423 
424 /*
425  * Add the span [vaddr, vaddr + size) to vmp and update kstats.
426  */
427 static vmem_seg_t *
428 vmem_span_create(vmem_t *vmp, void *vaddr, size_t size, uint8_t import)
429 {
430 	vmem_seg_t *knext;
431 	vmem_seg_t *newseg, *span;
432 	uintptr_t start = (uintptr_t)vaddr;
433 	uintptr_t end = start + size;
434 
435 	knext = &vmp->vm_seg0;
436 	if (!import && vmp->vm_source_alloc == NULL) {
437 		vmem_seg_t *kend, *kprev;
438 		/*
439 		 * non-imported spans are sorted in address order.  This
440 		 * makes vmem_extend_unlocked() much more effective.
441 		 *
442 		 * We search in reverse order, since new spans are
443 		 * generally at higher addresses.
444 		 */
445 		kend = &vmp->vm_seg0;
446 		for (kprev = kend->vs_kprev; kprev != kend;
447 		    kprev = kprev->vs_kprev) {
448 			if (!kprev->vs_import && (kprev->vs_end - 1) < start)
449 				break;
450 		}
451 		knext = kprev->vs_knext;
452 	}
453 
454 	ASSERT(MUTEX_HELD(&vmp->vm_lock));
455 
456 	if ((start | end) & (vmp->vm_quantum - 1)) {
457 		umem_panic("vmem_span_create(%p, %p, %lu): misaligned",
458 		    vmp, vaddr, size);
459 	}
460 
461 	span = vmem_seg_create(vmp, knext->vs_aprev, start, end);
462 	span->vs_type = VMEM_SPAN;
463 	VMEM_INSERT(knext->vs_kprev, span, k);
464 
465 	newseg = vmem_seg_create(vmp, span, start, end);
466 	vmem_freelist_insert(vmp, newseg);
467 
468 	newseg->vs_import = import;
469 	if (import)
470 		vmp->vm_kstat.vk_mem_import += size;
471 	vmp->vm_kstat.vk_mem_total += size;
472 
473 	return (newseg);
474 }
475 
476 /*
477  * Remove span vsp from vmp and update kstats.
478  */
479 static void
480 vmem_span_destroy(vmem_t *vmp, vmem_seg_t *vsp)
481 {
482 	vmem_seg_t *span = vsp->vs_aprev;
483 	size_t size = VS_SIZE(vsp);
484 
485 	ASSERT(MUTEX_HELD(&vmp->vm_lock));
486 	ASSERT(span->vs_type == VMEM_SPAN);
487 
488 	if (vsp->vs_import)
489 		vmp->vm_kstat.vk_mem_import -= size;
490 	vmp->vm_kstat.vk_mem_total -= size;
491 
492 	VMEM_DELETE(span, k);
493 
494 	vmem_seg_destroy(vmp, vsp);
495 	vmem_seg_destroy(vmp, span);
496 }
497 
498 /*
499  * Allocate the subrange [addr, addr + size) from segment vsp.
500  * If there are leftovers on either side, place them on the freelist.
501  * Returns a pointer to the segment representing [addr, addr + size).
502  */
503 static vmem_seg_t *
504 vmem_seg_alloc(vmem_t *vmp, vmem_seg_t *vsp, uintptr_t addr, size_t size)
505 {
506 	uintptr_t vs_start = vsp->vs_start;
507 	uintptr_t vs_end = vsp->vs_end;
508 	size_t vs_size = vs_end - vs_start;
509 	size_t realsize = P2ROUNDUP(size, vmp->vm_quantum);
510 	uintptr_t addr_end = addr + realsize;
511 
512 	ASSERT(P2PHASE(vs_start, vmp->vm_quantum) == 0);
513 	ASSERT(P2PHASE(addr, vmp->vm_quantum) == 0);
514 	ASSERT(vsp->vs_type == VMEM_FREE);
515 	ASSERT(addr >= vs_start && addr_end - 1 <= vs_end - 1);
516 	ASSERT(addr - 1 <= addr_end - 1);
517 
518 	/*
519 	 * If we're allocating from the start of the segment, and the
520 	 * remainder will be on the same freelist, we can save quite
521 	 * a bit of work.
522 	 */
523 	if (P2SAMEHIGHBIT(vs_size, vs_size - realsize) && addr == vs_start) {
524 		ASSERT(highbit(vs_size) == highbit(vs_size - realsize));
525 		vsp->vs_start = addr_end;
526 		vsp = vmem_seg_create(vmp, vsp->vs_aprev, addr, addr + size);
527 		vmem_hash_insert(vmp, vsp);
528 		return (vsp);
529 	}
530 
531 	vmem_freelist_delete(vmp, vsp);
532 
533 	if (vs_end != addr_end)
534 		vmem_freelist_insert(vmp,
535 		    vmem_seg_create(vmp, vsp, addr_end, vs_end));
536 
537 	if (vs_start != addr)
538 		vmem_freelist_insert(vmp,
539 		    vmem_seg_create(vmp, vsp->vs_aprev, vs_start, addr));
540 
541 	vsp->vs_start = addr;
542 	vsp->vs_end = addr + size;
543 
544 	vmem_hash_insert(vmp, vsp);
545 	return (vsp);
546 }
547 
548 /*
549  * We cannot reap if we are in the middle of a vmem_populate().
550  */
551 void
552 vmem_reap(void)
553 {
554 	if (!IN_POPULATE())
555 		umem_reap();
556 }
557 
558 /*
559  * Populate vmp's segfree list with VMEM_MINFREE vmem_seg_t structures.
560  */
561 static int
562 vmem_populate(vmem_t *vmp, int vmflag)
563 {
564 	char *p;
565 	vmem_seg_t *vsp;
566 	ssize_t nseg;
567 	size_t size;
568 	vmem_populate_lock_t *lp;
569 	int i;
570 
571 	while (vmp->vm_nsegfree < VMEM_MINFREE &&
572 	    (vsp = vmem_getseg_global()) != NULL)
573 		vmem_putseg(vmp, vsp);
574 
575 	if (vmp->vm_nsegfree >= VMEM_MINFREE)
576 		return (1);
577 
578 	/*
579 	 * If we're already populating, tap the reserve.
580 	 */
581 	if (vmem_nosleep_lock.vmpl_thr == thr_self()) {
582 		ASSERT(vmp->vm_cflags & VMC_POPULATOR);
583 		return (1);
584 	}
585 
586 	(void) mutex_unlock(&vmp->vm_lock);
587 
588 	ASSERT(vmflag & VM_NOSLEEP);	/* we do not allow sleep allocations */
589 	lp = &vmem_nosleep_lock;
590 
591 	/*
592 	 * Cannot be just a mutex_lock(), since that has no effect if
593 	 * libthread is not linked.
594 	 */
595 	(void) mutex_lock(&lp->vmpl_mutex);
596 	ASSERT(lp->vmpl_thr == 0);
597 	lp->vmpl_thr = thr_self();
598 
599 	nseg = VMEM_MINFREE + vmem_populators * VMEM_POPULATE_RESERVE;
600 	size = P2ROUNDUP(nseg * vmem_seg_size, vmem_seg_arena->vm_quantum);
601 	nseg = size / vmem_seg_size;
602 
603 	/*
604 	 * The following vmem_alloc() may need to populate vmem_seg_arena
605 	 * and all the things it imports from.  When doing so, it will tap
606 	 * each arena's reserve to prevent recursion (see the block comment
607 	 * above the definition of VMEM_POPULATE_RESERVE).
608 	 *
609 	 * During this allocation, vmem_reap() is a no-op.  If the allocation
610 	 * fails, we call vmem_reap() after dropping the population lock.
611 	 */
612 	p = vmem_alloc(vmem_seg_arena, size, vmflag & VM_UMFLAGS);
613 	if (p == NULL) {
614 		lp->vmpl_thr = 0;
615 		(void) mutex_unlock(&lp->vmpl_mutex);
616 		vmem_reap();
617 
618 		(void) mutex_lock(&vmp->vm_lock);
619 		vmp->vm_kstat.vk_populate_fail++;
620 		return (0);
621 	}
622 	/*
623 	 * Restock the arenas that may have been depleted during population.
624 	 */
625 	for (i = 0; i < vmem_populators; i++) {
626 		(void) mutex_lock(&vmem_populator[i]->vm_lock);
627 		while (vmem_populator[i]->vm_nsegfree < VMEM_POPULATE_RESERVE)
628 			vmem_putseg(vmem_populator[i],
629 			    (vmem_seg_t *)(p + --nseg * vmem_seg_size));
630 		(void) mutex_unlock(&vmem_populator[i]->vm_lock);
631 	}
632 
633 	lp->vmpl_thr = 0;
634 	(void) mutex_unlock(&lp->vmpl_mutex);
635 	(void) mutex_lock(&vmp->vm_lock);
636 
637 	/*
638 	 * Now take our own segments.
639 	 */
640 	ASSERT(nseg >= VMEM_MINFREE);
641 	while (vmp->vm_nsegfree < VMEM_MINFREE)
642 		vmem_putseg(vmp, (vmem_seg_t *)(p + --nseg * vmem_seg_size));
643 
644 	/*
645 	 * Give the remainder to charity.
646 	 */
647 	while (nseg > 0)
648 		vmem_putseg_global((vmem_seg_t *)(p + --nseg * vmem_seg_size));
649 
650 	return (1);
651 }
652 
653 /*
654  * Advance a walker from its previous position to 'afterme'.
655  * Note: may drop and reacquire vmp->vm_lock.
656  */
657 static void
658 vmem_advance(vmem_t *vmp, vmem_seg_t *walker, vmem_seg_t *afterme)
659 {
660 	vmem_seg_t *vprev = walker->vs_aprev;
661 	vmem_seg_t *vnext = walker->vs_anext;
662 	vmem_seg_t *vsp = NULL;
663 
664 	VMEM_DELETE(walker, a);
665 
666 	if (afterme != NULL)
667 		VMEM_INSERT(afterme, walker, a);
668 
669 	/*
670 	 * The walker segment's presence may have prevented its neighbors
671 	 * from coalescing.  If so, coalesce them now.
672 	 */
673 	if (vprev->vs_type == VMEM_FREE) {
674 		if (vnext->vs_type == VMEM_FREE) {
675 			ASSERT(vprev->vs_end == vnext->vs_start);
676 			vmem_freelist_delete(vmp, vnext);
677 			vmem_freelist_delete(vmp, vprev);
678 			vprev->vs_end = vnext->vs_end;
679 			vmem_freelist_insert(vmp, vprev);
680 			vmem_seg_destroy(vmp, vnext);
681 		}
682 		vsp = vprev;
683 	} else if (vnext->vs_type == VMEM_FREE) {
684 		vsp = vnext;
685 	}
686 
687 	/*
688 	 * vsp could represent a complete imported span,
689 	 * in which case we must return it to the source.
690 	 */
691 	if (vsp != NULL && vsp->vs_import && vmp->vm_source_free != NULL &&
692 	    vsp->vs_aprev->vs_type == VMEM_SPAN &&
693 	    vsp->vs_anext->vs_type == VMEM_SPAN) {
694 		void *vaddr = (void *)vsp->vs_start;
695 		size_t size = VS_SIZE(vsp);
696 		ASSERT(size == VS_SIZE(vsp->vs_aprev));
697 		vmem_freelist_delete(vmp, vsp);
698 		vmem_span_destroy(vmp, vsp);
699 		(void) mutex_unlock(&vmp->vm_lock);
700 		vmp->vm_source_free(vmp->vm_source, vaddr, size);
701 		(void) mutex_lock(&vmp->vm_lock);
702 	}
703 }
704 
705 /*
706  * VM_NEXTFIT allocations deliberately cycle through all virtual addresses
707  * in an arena, so that we avoid reusing addresses for as long as possible.
708  * This helps to catch used-after-freed bugs.  It's also the perfect policy
709  * for allocating things like process IDs, where we want to cycle through
710  * all values in order.
711  */
712 static void *
713 vmem_nextfit_alloc(vmem_t *vmp, size_t size, int vmflag)
714 {
715 	vmem_seg_t *vsp, *rotor;
716 	uintptr_t addr;
717 	size_t realsize = P2ROUNDUP(size, vmp->vm_quantum);
718 	size_t vs_size;
719 
720 	(void) mutex_lock(&vmp->vm_lock);
721 
722 	if (vmp->vm_nsegfree < VMEM_MINFREE && !vmem_populate(vmp, vmflag)) {
723 		(void) mutex_unlock(&vmp->vm_lock);
724 		return (NULL);
725 	}
726 
727 	/*
728 	 * The common case is that the segment right after the rotor is free,
729 	 * and large enough that extracting 'size' bytes won't change which
730 	 * freelist it's on.  In this case we can avoid a *lot* of work.
731 	 * Instead of the normal vmem_seg_alloc(), we just advance the start
732 	 * address of the victim segment.  Instead of moving the rotor, we
733 	 * create the new segment structure *behind the rotor*, which has
734 	 * the same effect.  And finally, we know we don't have to coalesce
735 	 * the rotor's neighbors because the new segment lies between them.
736 	 */
737 	rotor = &vmp->vm_rotor;
738 	vsp = rotor->vs_anext;
739 	if (vsp->vs_type == VMEM_FREE && (vs_size = VS_SIZE(vsp)) > realsize &&
740 	    P2SAMEHIGHBIT(vs_size, vs_size - realsize)) {
741 		ASSERT(highbit(vs_size) == highbit(vs_size - realsize));
742 		addr = vsp->vs_start;
743 		vsp->vs_start = addr + realsize;
744 		vmem_hash_insert(vmp,
745 		    vmem_seg_create(vmp, rotor->vs_aprev, addr, addr + size));
746 		(void) mutex_unlock(&vmp->vm_lock);
747 		return ((void *)addr);
748 	}
749 
750 	/*
751 	 * Starting at the rotor, look for a segment large enough to
752 	 * satisfy the allocation.
753 	 */
754 	for (;;) {
755 		vmp->vm_kstat.vk_search++;
756 		if (vsp->vs_type == VMEM_FREE && VS_SIZE(vsp) >= size)
757 			break;
758 		vsp = vsp->vs_anext;
759 		if (vsp == rotor) {
760 			int cancel_state;
761 
762 			/*
763 			 * We've come full circle.  One possibility is that the
764 			 * there's actually enough space, but the rotor itself
765 			 * is preventing the allocation from succeeding because
766 			 * it's sitting between two free segments.  Therefore,
767 			 * we advance the rotor and see if that liberates a
768 			 * suitable segment.
769 			 */
770 			vmem_advance(vmp, rotor, rotor->vs_anext);
771 			vsp = rotor->vs_aprev;
772 			if (vsp->vs_type == VMEM_FREE && VS_SIZE(vsp) >= size)
773 				break;
774 			/*
775 			 * If there's a lower arena we can import from, or it's
776 			 * a VM_NOSLEEP allocation, let vmem_xalloc() handle it.
777 			 * Otherwise, wait until another thread frees something.
778 			 */
779 			if (vmp->vm_source_alloc != NULL ||
780 			    (vmflag & VM_NOSLEEP)) {
781 				(void) mutex_unlock(&vmp->vm_lock);
782 				return (vmem_xalloc(vmp, size, vmp->vm_quantum,
783 				    0, 0, NULL, NULL, vmflag & VM_UMFLAGS));
784 			}
785 			vmp->vm_kstat.vk_wait++;
786 			(void) pthread_setcancelstate(PTHREAD_CANCEL_DISABLE,
787 			    &cancel_state);
788 			(void) cond_wait(&vmp->vm_cv, &vmp->vm_lock);
789 			(void) pthread_setcancelstate(cancel_state, NULL);
790 			vsp = rotor->vs_anext;
791 		}
792 	}
793 
794 	/*
795 	 * We found a segment.  Extract enough space to satisfy the allocation.
796 	 */
797 	addr = vsp->vs_start;
798 	vsp = vmem_seg_alloc(vmp, vsp, addr, size);
799 	ASSERT(vsp->vs_type == VMEM_ALLOC &&
800 	    vsp->vs_start == addr && vsp->vs_end == addr + size);
801 
802 	/*
803 	 * Advance the rotor to right after the newly-allocated segment.
804 	 * That's where the next VM_NEXTFIT allocation will begin searching.
805 	 */
806 	vmem_advance(vmp, rotor, vsp);
807 	(void) mutex_unlock(&vmp->vm_lock);
808 	return ((void *)addr);
809 }
810 
811 /*
812  * Allocate size bytes at offset phase from an align boundary such that the
813  * resulting segment [addr, addr + size) is a subset of [minaddr, maxaddr)
814  * that does not straddle a nocross-aligned boundary.
815  */
816 void *
817 vmem_xalloc(vmem_t *vmp, size_t size, size_t align, size_t phase,
818 	size_t nocross, void *minaddr, void *maxaddr, int vmflag)
819 {
820 	vmem_seg_t *vsp;
821 	vmem_seg_t *vbest = NULL;
822 	uintptr_t addr, taddr, start, end;
823 	void *vaddr;
824 	int hb, flist, resv;
825 	uint32_t mtbf;
826 
827 	if (phase > 0 && phase >= align)
828 		umem_panic("vmem_xalloc(%p, %lu, %lu, %lu, %lu, %p, %p, %x): "
829 		    "invalid phase",
830 		    (void *)vmp, size, align, phase, nocross,
831 		    minaddr, maxaddr, vmflag);
832 
833 	if (align == 0)
834 		align = vmp->vm_quantum;
835 
836 	if ((align | phase | nocross) & (vmp->vm_quantum - 1)) {
837 		umem_panic("vmem_xalloc(%p, %lu, %lu, %lu, %lu, %p, %p, %x): "
838 		    "parameters not vm_quantum aligned",
839 		    (void *)vmp, size, align, phase, nocross,
840 		    minaddr, maxaddr, vmflag);
841 	}
842 
843 	if (nocross != 0 &&
844 	    (align > nocross || P2ROUNDUP(phase + size, align) > nocross)) {
845 		umem_panic("vmem_xalloc(%p, %lu, %lu, %lu, %lu, %p, %p, %x): "
846 		    "overconstrained allocation",
847 		    (void *)vmp, size, align, phase, nocross,
848 		    minaddr, maxaddr, vmflag);
849 	}
850 
851 	if ((mtbf = vmem_mtbf | vmp->vm_mtbf) != 0 && gethrtime() % mtbf == 0 &&
852 	    (vmflag & (VM_NOSLEEP | VM_PANIC)) == VM_NOSLEEP)
853 		return (NULL);
854 
855 	(void) mutex_lock(&vmp->vm_lock);
856 	for (;;) {
857 		int cancel_state;
858 
859 		if (vmp->vm_nsegfree < VMEM_MINFREE &&
860 		    !vmem_populate(vmp, vmflag))
861 			break;
862 
863 		/*
864 		 * highbit() returns the highest bit + 1, which is exactly
865 		 * what we want: we want to search the first freelist whose
866 		 * members are *definitely* large enough to satisfy our
867 		 * allocation.  However, there are certain cases in which we
868 		 * want to look at the next-smallest freelist (which *might*
869 		 * be able to satisfy the allocation):
870 		 *
871 		 * (1)	The size is exactly a power of 2, in which case
872 		 *	the smaller freelist is always big enough;
873 		 *
874 		 * (2)	All other freelists are empty;
875 		 *
876 		 * (3)	We're in the highest possible freelist, which is
877 		 *	always empty (e.g. the 4GB freelist on 32-bit systems);
878 		 *
879 		 * (4)	We're doing a best-fit or first-fit allocation.
880 		 */
881 		if ((size & (size - 1)) == 0) {
882 			flist = lowbit(P2ALIGN(vmp->vm_freemap, size));
883 		} else {
884 			hb = highbit(size);
885 			if ((vmp->vm_freemap >> hb) == 0 ||
886 			    hb == VMEM_FREELISTS ||
887 			    (vmflag & (VM_BESTFIT | VM_FIRSTFIT)))
888 				hb--;
889 			flist = lowbit(P2ALIGN(vmp->vm_freemap, 1UL << hb));
890 		}
891 
892 		for (vbest = NULL, vsp = (flist == 0) ? NULL :
893 		    vmp->vm_freelist[flist - 1].vs_knext;
894 		    vsp != NULL; vsp = vsp->vs_knext) {
895 			vmp->vm_kstat.vk_search++;
896 			if (vsp->vs_start == 0) {
897 				/*
898 				 * We're moving up to a larger freelist,
899 				 * so if we've already found a candidate,
900 				 * the fit can't possibly get any better.
901 				 */
902 				if (vbest != NULL)
903 					break;
904 				/*
905 				 * Find the next non-empty freelist.
906 				 */
907 				flist = lowbit(P2ALIGN(vmp->vm_freemap,
908 				    VS_SIZE(vsp)));
909 				if (flist-- == 0)
910 					break;
911 				vsp = (vmem_seg_t *)&vmp->vm_freelist[flist];
912 				ASSERT(vsp->vs_knext->vs_type == VMEM_FREE);
913 				continue;
914 			}
915 			if (vsp->vs_end - 1 < (uintptr_t)minaddr)
916 				continue;
917 			if (vsp->vs_start > (uintptr_t)maxaddr - 1)
918 				continue;
919 			start = MAX(vsp->vs_start, (uintptr_t)minaddr);
920 			end = MIN(vsp->vs_end - 1, (uintptr_t)maxaddr - 1) + 1;
921 			taddr = P2PHASEUP(start, align, phase);
922 			if (P2CROSS(taddr, taddr + size - 1, nocross))
923 				taddr +=
924 				    P2ROUNDUP(P2NPHASE(taddr, nocross), align);
925 			if ((taddr - start) + size > end - start ||
926 			    (vbest != NULL && VS_SIZE(vsp) >= VS_SIZE(vbest)))
927 				continue;
928 			vbest = vsp;
929 			addr = taddr;
930 			if (!(vmflag & VM_BESTFIT) || VS_SIZE(vbest) == size)
931 				break;
932 		}
933 		if (vbest != NULL)
934 			break;
935 		if (size == 0)
936 			umem_panic("vmem_xalloc(): size == 0");
937 		if (vmp->vm_source_alloc != NULL && nocross == 0 &&
938 		    minaddr == NULL && maxaddr == NULL) {
939 			size_t asize = P2ROUNDUP(size + phase,
940 			    MAX(align, vmp->vm_source->vm_quantum));
941 			if (asize < size) {		/* overflow */
942 				(void) mutex_unlock(&vmp->vm_lock);
943 				if (vmflag & VM_NOSLEEP)
944 					return (NULL);
945 
946 				umem_panic("vmem_xalloc(): "
947 				    "overflow on VM_SLEEP allocation");
948 			}
949 			/*
950 			 * Determine how many segment structures we'll consume.
951 			 * The calculation must be presise because if we're
952 			 * here on behalf of vmem_populate(), we are taking
953 			 * segments from a very limited reserve.
954 			 */
955 			resv = (size == asize) ?
956 			    VMEM_SEGS_PER_SPAN_CREATE +
957 			    VMEM_SEGS_PER_EXACT_ALLOC :
958 			    VMEM_SEGS_PER_ALLOC_MAX;
959 			ASSERT(vmp->vm_nsegfree >= resv);
960 			vmp->vm_nsegfree -= resv;	/* reserve our segs */
961 			(void) mutex_unlock(&vmp->vm_lock);
962 			vaddr = vmp->vm_source_alloc(vmp->vm_source, asize,
963 			    vmflag & VM_UMFLAGS);
964 			(void) mutex_lock(&vmp->vm_lock);
965 			vmp->vm_nsegfree += resv;	/* claim reservation */
966 			if (vaddr != NULL) {
967 				vbest = vmem_span_create(vmp, vaddr, asize, 1);
968 				addr = P2PHASEUP(vbest->vs_start, align, phase);
969 				break;
970 			}
971 		}
972 		(void) mutex_unlock(&vmp->vm_lock);
973 		vmem_reap();
974 		(void) mutex_lock(&vmp->vm_lock);
975 		if (vmflag & VM_NOSLEEP)
976 			break;
977 		vmp->vm_kstat.vk_wait++;
978 		(void) pthread_setcancelstate(PTHREAD_CANCEL_DISABLE,
979 		    &cancel_state);
980 		(void) cond_wait(&vmp->vm_cv, &vmp->vm_lock);
981 		(void) pthread_setcancelstate(cancel_state, NULL);
982 	}
983 	if (vbest != NULL) {
984 		ASSERT(vbest->vs_type == VMEM_FREE);
985 		ASSERT(vbest->vs_knext != vbest);
986 		(void) vmem_seg_alloc(vmp, vbest, addr, size);
987 		(void) mutex_unlock(&vmp->vm_lock);
988 		ASSERT(P2PHASE(addr, align) == phase);
989 		ASSERT(!P2CROSS(addr, addr + size - 1, nocross));
990 		ASSERT(addr >= (uintptr_t)minaddr);
991 		ASSERT(addr + size - 1 <= (uintptr_t)maxaddr - 1);
992 		return ((void *)addr);
993 	}
994 	vmp->vm_kstat.vk_fail++;
995 	(void) mutex_unlock(&vmp->vm_lock);
996 	if (vmflag & VM_PANIC)
997 		umem_panic("vmem_xalloc(%p, %lu, %lu, %lu, %lu, %p, %p, %x): "
998 		    "cannot satisfy mandatory allocation",
999 		    (void *)vmp, size, align, phase, nocross,
1000 		    minaddr, maxaddr, vmflag);
1001 	return (NULL);
1002 }
1003 
1004 /*
1005  * Free the segment [vaddr, vaddr + size), where vaddr was a constrained
1006  * allocation.  vmem_xalloc() and vmem_xfree() must always be paired because
1007  * both routines bypass the quantum caches.
1008  */
1009 void
1010 vmem_xfree(vmem_t *vmp, void *vaddr, size_t size)
1011 {
1012 	vmem_seg_t *vsp, *vnext, *vprev;
1013 
1014 	(void) mutex_lock(&vmp->vm_lock);
1015 
1016 	vsp = vmem_hash_delete(vmp, (uintptr_t)vaddr, size);
1017 	vsp->vs_end = P2ROUNDUP(vsp->vs_end, vmp->vm_quantum);
1018 
1019 	/*
1020 	 * Attempt to coalesce with the next segment.
1021 	 */
1022 	vnext = vsp->vs_anext;
1023 	if (vnext->vs_type == VMEM_FREE) {
1024 		ASSERT(vsp->vs_end == vnext->vs_start);
1025 		vmem_freelist_delete(vmp, vnext);
1026 		vsp->vs_end = vnext->vs_end;
1027 		vmem_seg_destroy(vmp, vnext);
1028 	}
1029 
1030 	/*
1031 	 * Attempt to coalesce with the previous segment.
1032 	 */
1033 	vprev = vsp->vs_aprev;
1034 	if (vprev->vs_type == VMEM_FREE) {
1035 		ASSERT(vprev->vs_end == vsp->vs_start);
1036 		vmem_freelist_delete(vmp, vprev);
1037 		vprev->vs_end = vsp->vs_end;
1038 		vmem_seg_destroy(vmp, vsp);
1039 		vsp = vprev;
1040 	}
1041 
1042 	/*
1043 	 * If the entire span is free, return it to the source.
1044 	 */
1045 	if (vsp->vs_import && vmp->vm_source_free != NULL &&
1046 	    vsp->vs_aprev->vs_type == VMEM_SPAN &&
1047 	    vsp->vs_anext->vs_type == VMEM_SPAN) {
1048 		vaddr = (void *)vsp->vs_start;
1049 		size = VS_SIZE(vsp);
1050 		ASSERT(size == VS_SIZE(vsp->vs_aprev));
1051 		vmem_span_destroy(vmp, vsp);
1052 		(void) mutex_unlock(&vmp->vm_lock);
1053 		vmp->vm_source_free(vmp->vm_source, vaddr, size);
1054 	} else {
1055 		vmem_freelist_insert(vmp, vsp);
1056 		(void) mutex_unlock(&vmp->vm_lock);
1057 	}
1058 }
1059 
1060 /*
1061  * Allocate size bytes from arena vmp.  Returns the allocated address
1062  * on success, NULL on failure.  vmflag specifies VM_SLEEP or VM_NOSLEEP,
1063  * and may also specify best-fit, first-fit, or next-fit allocation policy
1064  * instead of the default instant-fit policy.  VM_SLEEP allocations are
1065  * guaranteed to succeed.
1066  */
1067 void *
1068 vmem_alloc(vmem_t *vmp, size_t size, int vmflag)
1069 {
1070 	vmem_seg_t *vsp;
1071 	uintptr_t addr;
1072 	int hb;
1073 	int flist = 0;
1074 	uint32_t mtbf;
1075 
1076 	if (size - 1 < vmp->vm_qcache_max) {
1077 		ASSERT(vmflag & VM_NOSLEEP);
1078 		return (_umem_cache_alloc(vmp->vm_qcache[(size - 1) >>
1079 		    vmp->vm_qshift], UMEM_DEFAULT));
1080 	}
1081 
1082 	if ((mtbf = vmem_mtbf | vmp->vm_mtbf) != 0 && gethrtime() % mtbf == 0 &&
1083 	    (vmflag & (VM_NOSLEEP | VM_PANIC)) == VM_NOSLEEP)
1084 		return (NULL);
1085 
1086 	if (vmflag & VM_NEXTFIT)
1087 		return (vmem_nextfit_alloc(vmp, size, vmflag));
1088 
1089 	if (vmflag & (VM_BESTFIT | VM_FIRSTFIT))
1090 		return (vmem_xalloc(vmp, size, vmp->vm_quantum, 0, 0,
1091 		    NULL, NULL, vmflag));
1092 
1093 	/*
1094 	 * Unconstrained instant-fit allocation from the segment list.
1095 	 */
1096 	(void) mutex_lock(&vmp->vm_lock);
1097 
1098 	if (vmp->vm_nsegfree >= VMEM_MINFREE || vmem_populate(vmp, vmflag)) {
1099 		if ((size & (size - 1)) == 0)
1100 			flist = lowbit(P2ALIGN(vmp->vm_freemap, size));
1101 		else if ((hb = highbit(size)) < VMEM_FREELISTS)
1102 			flist = lowbit(P2ALIGN(vmp->vm_freemap, 1UL << hb));
1103 	}
1104 
1105 	if (flist-- == 0) {
1106 		(void) mutex_unlock(&vmp->vm_lock);
1107 		return (vmem_xalloc(vmp, size, vmp->vm_quantum,
1108 		    0, 0, NULL, NULL, vmflag));
1109 	}
1110 
1111 	ASSERT(size <= (1UL << flist));
1112 	vsp = vmp->vm_freelist[flist].vs_knext;
1113 	addr = vsp->vs_start;
1114 	(void) vmem_seg_alloc(vmp, vsp, addr, size);
1115 	(void) mutex_unlock(&vmp->vm_lock);
1116 	return ((void *)addr);
1117 }
1118 
1119 /*
1120  * Free the segment [vaddr, vaddr + size).
1121  */
1122 void
1123 vmem_free(vmem_t *vmp, void *vaddr, size_t size)
1124 {
1125 	if (size - 1 < vmp->vm_qcache_max)
1126 		_umem_cache_free(vmp->vm_qcache[(size - 1) >> vmp->vm_qshift],
1127 		    vaddr);
1128 	else
1129 		vmem_xfree(vmp, vaddr, size);
1130 }
1131 
1132 /*
1133  * Determine whether arena vmp contains the segment [vaddr, vaddr + size).
1134  */
1135 int
1136 vmem_contains(vmem_t *vmp, void *vaddr, size_t size)
1137 {
1138 	uintptr_t start = (uintptr_t)vaddr;
1139 	uintptr_t end = start + size;
1140 	vmem_seg_t *vsp;
1141 	vmem_seg_t *seg0 = &vmp->vm_seg0;
1142 
1143 	(void) mutex_lock(&vmp->vm_lock);
1144 	vmp->vm_kstat.vk_contains++;
1145 	for (vsp = seg0->vs_knext; vsp != seg0; vsp = vsp->vs_knext) {
1146 		vmp->vm_kstat.vk_contains_search++;
1147 		ASSERT(vsp->vs_type == VMEM_SPAN);
1148 		if (start >= vsp->vs_start && end - 1 <= vsp->vs_end - 1)
1149 			break;
1150 	}
1151 	(void) mutex_unlock(&vmp->vm_lock);
1152 	return (vsp != seg0);
1153 }
1154 
1155 /*
1156  * Add the span [vaddr, vaddr + size) to arena vmp.
1157  */
1158 void *
1159 vmem_add(vmem_t *vmp, void *vaddr, size_t size, int vmflag)
1160 {
1161 	if (vaddr == NULL || size == 0) {
1162 		umem_panic("vmem_add(%p, %p, %lu): bad arguments",
1163 		    vmp, vaddr, size);
1164 	}
1165 
1166 	ASSERT(!vmem_contains(vmp, vaddr, size));
1167 
1168 	(void) mutex_lock(&vmp->vm_lock);
1169 	if (vmem_populate(vmp, vmflag))
1170 		(void) vmem_span_create(vmp, vaddr, size, 0);
1171 	else
1172 		vaddr = NULL;
1173 	(void) cond_broadcast(&vmp->vm_cv);
1174 	(void) mutex_unlock(&vmp->vm_lock);
1175 	return (vaddr);
1176 }
1177 
1178 /*
1179  * Adds the address range [addr, endaddr) to arena vmp, by either:
1180  *   1. joining two existing spans, [x, addr), and [endaddr, y) (which
1181  *      are in that order) into a single [x, y) span,
1182  *   2. expanding an existing [x, addr) span to [x, endaddr),
1183  *   3. expanding an existing [endaddr, x) span to [addr, x), or
1184  *   4. creating a new [addr, endaddr) span.
1185  *
1186  * Called with vmp->vm_lock held, and a successful vmem_populate() completed.
1187  * Cannot fail.  Returns the new segment.
1188  *
1189  * NOTE:  this algorithm is linear-time in the number of spans, but is
1190  *      constant-time when you are extending the last (highest-addressed)
1191  *      span.
1192  */
1193 static vmem_seg_t *
1194 vmem_extend_unlocked(vmem_t *vmp, uintptr_t addr, uintptr_t endaddr)
1195 {
1196 	vmem_seg_t *span;
1197 	vmem_seg_t *vsp;
1198 
1199 	vmem_seg_t *end = &vmp->vm_seg0;
1200 
1201 	ASSERT(MUTEX_HELD(&vmp->vm_lock));
1202 
1203 	/*
1204 	 * the second "if" clause below relies on the direction of this search
1205 	 */
1206 	for (span = end->vs_kprev; span != end; span = span->vs_kprev) {
1207 		if (span->vs_end == addr || span->vs_start == endaddr)
1208 			break;
1209 	}
1210 
1211 	if (span == end)
1212 		return (vmem_span_create(vmp, (void *)addr, endaddr - addr, 0));
1213 	if (span->vs_kprev->vs_end == addr && span->vs_start == endaddr) {
1214 		vmem_seg_t *prevspan = span->vs_kprev;
1215 		vmem_seg_t *nextseg = span->vs_anext;
1216 		vmem_seg_t *prevseg = span->vs_aprev;
1217 
1218 		/*
1219 		 * prevspan becomes the span marker for the full range
1220 		 */
1221 		prevspan->vs_end = span->vs_end;
1222 
1223 		/*
1224 		 * Notionally, span becomes a free segment representing
1225 		 * [addr, endaddr).
1226 		 *
1227 		 * However, if either of its neighbors are free, we coalesce
1228 		 * by destroying span and changing the free segment.
1229 		 */
1230 		if (prevseg->vs_type == VMEM_FREE &&
1231 		    nextseg->vs_type == VMEM_FREE) {
1232 			/*
1233 			 * coalesce both ways
1234 			 */
1235 			ASSERT(prevseg->vs_end == addr &&
1236 			    nextseg->vs_start == endaddr);
1237 
1238 			vmem_freelist_delete(vmp, prevseg);
1239 			prevseg->vs_end = nextseg->vs_end;
1240 
1241 			vmem_freelist_delete(vmp, nextseg);
1242 			VMEM_DELETE(span, k);
1243 			vmem_seg_destroy(vmp, nextseg);
1244 			vmem_seg_destroy(vmp, span);
1245 
1246 			vsp = prevseg;
1247 		} else if (prevseg->vs_type == VMEM_FREE) {
1248 			/*
1249 			 * coalesce left
1250 			 */
1251 			ASSERT(prevseg->vs_end == addr);
1252 
1253 			VMEM_DELETE(span, k);
1254 			vmem_seg_destroy(vmp, span);
1255 
1256 			vmem_freelist_delete(vmp, prevseg);
1257 			prevseg->vs_end = endaddr;
1258 
1259 			vsp = prevseg;
1260 		} else if (nextseg->vs_type == VMEM_FREE) {
1261 			/*
1262 			 * coalesce right
1263 			 */
1264 			ASSERT(nextseg->vs_start == endaddr);
1265 
1266 			VMEM_DELETE(span, k);
1267 			vmem_seg_destroy(vmp, span);
1268 
1269 			vmem_freelist_delete(vmp, nextseg);
1270 			nextseg->vs_start = addr;
1271 
1272 			vsp = nextseg;
1273 		} else {
1274 			/*
1275 			 * cannnot coalesce
1276 			 */
1277 			VMEM_DELETE(span, k);
1278 			span->vs_start = addr;
1279 			span->vs_end = endaddr;
1280 
1281 			vsp = span;
1282 		}
1283 	} else if (span->vs_end == addr) {
1284 		vmem_seg_t *oldseg = span->vs_knext->vs_aprev;
1285 		span->vs_end = endaddr;
1286 
1287 		ASSERT(oldseg->vs_type != VMEM_SPAN);
1288 		if (oldseg->vs_type == VMEM_FREE) {
1289 			ASSERT(oldseg->vs_end == addr);
1290 			vmem_freelist_delete(vmp, oldseg);
1291 			oldseg->vs_end = endaddr;
1292 			vsp = oldseg;
1293 		} else
1294 			vsp = vmem_seg_create(vmp, oldseg, addr, endaddr);
1295 	} else {
1296 		vmem_seg_t *oldseg = span->vs_anext;
1297 		ASSERT(span->vs_start == endaddr);
1298 		span->vs_start = addr;
1299 
1300 		ASSERT(oldseg->vs_type != VMEM_SPAN);
1301 		if (oldseg->vs_type == VMEM_FREE) {
1302 			ASSERT(oldseg->vs_start == endaddr);
1303 			vmem_freelist_delete(vmp, oldseg);
1304 			oldseg->vs_start = addr;
1305 			vsp = oldseg;
1306 		} else
1307 			vsp = vmem_seg_create(vmp, span, addr, endaddr);
1308 	}
1309 	vmem_freelist_insert(vmp, vsp);
1310 	vmp->vm_kstat.vk_mem_total += (endaddr - addr);
1311 	return (vsp);
1312 }
1313 
1314 /*
1315  * Does some error checking, calls vmem_extend_unlocked to add
1316  * [vaddr, vaddr+size) to vmp, then allocates alloc bytes from the
1317  * newly merged segment.
1318  */
1319 void *
1320 _vmem_extend_alloc(vmem_t *vmp, void *vaddr, size_t size, size_t alloc,
1321     int vmflag)
1322 {
1323 	uintptr_t addr = (uintptr_t)vaddr;
1324 	uintptr_t endaddr = addr + size;
1325 	vmem_seg_t *vsp;
1326 
1327 	ASSERT(vaddr != NULL && size != 0 && endaddr > addr);
1328 	ASSERT(alloc <= size && alloc != 0);
1329 	ASSERT(((addr | size | alloc) & (vmp->vm_quantum - 1)) == 0);
1330 
1331 	ASSERT(!vmem_contains(vmp, vaddr, size));
1332 
1333 	(void) mutex_lock(&vmp->vm_lock);
1334 	if (!vmem_populate(vmp, vmflag)) {
1335 		(void) mutex_unlock(&vmp->vm_lock);
1336 		return (NULL);
1337 	}
1338 	/*
1339 	 * if there is a source, we can't mess with the spans
1340 	 */
1341 	if (vmp->vm_source_alloc != NULL)
1342 		vsp = vmem_span_create(vmp, vaddr, size, 0);
1343 	else
1344 		vsp = vmem_extend_unlocked(vmp, addr, endaddr);
1345 
1346 	ASSERT(VS_SIZE(vsp) >= alloc);
1347 
1348 	addr = vsp->vs_start;
1349 	(void) vmem_seg_alloc(vmp, vsp, addr, alloc);
1350 	vaddr = (void *)addr;
1351 
1352 	(void) cond_broadcast(&vmp->vm_cv);
1353 	(void) mutex_unlock(&vmp->vm_lock);
1354 
1355 	return (vaddr);
1356 }
1357 
1358 /*
1359  * Walk the vmp arena, applying func to each segment matching typemask.
1360  * If VMEM_REENTRANT is specified, the arena lock is dropped across each
1361  * call to func(); otherwise, it is held for the duration of vmem_walk()
1362  * to ensure a consistent snapshot.  Note that VMEM_REENTRANT callbacks
1363  * are *not* necessarily consistent, so they may only be used when a hint
1364  * is adequate.
1365  */
1366 void
1367 vmem_walk(vmem_t *vmp, int typemask,
1368 	void (*func)(void *, void *, size_t), void *arg)
1369 {
1370 	vmem_seg_t *vsp;
1371 	vmem_seg_t *seg0 = &vmp->vm_seg0;
1372 	vmem_seg_t walker;
1373 
1374 	if (typemask & VMEM_WALKER)
1375 		return;
1376 
1377 	bzero(&walker, sizeof (walker));
1378 	walker.vs_type = VMEM_WALKER;
1379 
1380 	(void) mutex_lock(&vmp->vm_lock);
1381 	VMEM_INSERT(seg0, &walker, a);
1382 	for (vsp = seg0->vs_anext; vsp != seg0; vsp = vsp->vs_anext) {
1383 		if (vsp->vs_type & typemask) {
1384 			void *start = (void *)vsp->vs_start;
1385 			size_t size = VS_SIZE(vsp);
1386 			if (typemask & VMEM_REENTRANT) {
1387 				vmem_advance(vmp, &walker, vsp);
1388 				(void) mutex_unlock(&vmp->vm_lock);
1389 				func(arg, start, size);
1390 				(void) mutex_lock(&vmp->vm_lock);
1391 				vsp = &walker;
1392 			} else {
1393 				func(arg, start, size);
1394 			}
1395 		}
1396 	}
1397 	vmem_advance(vmp, &walker, NULL);
1398 	(void) mutex_unlock(&vmp->vm_lock);
1399 }
1400 
1401 /*
1402  * Return the total amount of memory whose type matches typemask.  Thus:
1403  *
1404  *	typemask VMEM_ALLOC yields total memory allocated (in use).
1405  *	typemask VMEM_FREE yields total memory free (available).
1406  *	typemask (VMEM_ALLOC | VMEM_FREE) yields total arena size.
1407  */
1408 size_t
1409 vmem_size(vmem_t *vmp, int typemask)
1410 {
1411 	uint64_t size = 0;
1412 
1413 	if (typemask & VMEM_ALLOC)
1414 		size += vmp->vm_kstat.vk_mem_inuse;
1415 	if (typemask & VMEM_FREE)
1416 		size += vmp->vm_kstat.vk_mem_total -
1417 		    vmp->vm_kstat.vk_mem_inuse;
1418 	return ((size_t)size);
1419 }
1420 
1421 /*
1422  * Create an arena called name whose initial span is [base, base + size).
1423  * The arena's natural unit of currency is quantum, so vmem_alloc()
1424  * guarantees quantum-aligned results.  The arena may import new spans
1425  * by invoking afunc() on source, and may return those spans by invoking
1426  * ffunc() on source.  To make small allocations fast and scalable,
1427  * the arena offers high-performance caching for each integer multiple
1428  * of quantum up to qcache_max.
1429  */
1430 vmem_t *
1431 vmem_create(const char *name, void *base, size_t size, size_t quantum,
1432 	vmem_alloc_t *afunc, vmem_free_t *ffunc, vmem_t *source,
1433 	size_t qcache_max, int vmflag)
1434 {
1435 	int i;
1436 	size_t nqcache;
1437 	vmem_t *vmp, *cur, **vmpp;
1438 	vmem_seg_t *vsp;
1439 	vmem_freelist_t *vfp;
1440 	uint32_t id = atomic_add_32_nv(&vmem_id, 1);
1441 
1442 	if (vmem_vmem_arena != NULL) {
1443 		vmp = vmem_alloc(vmem_vmem_arena, sizeof (vmem_t),
1444 		    vmflag & VM_UMFLAGS);
1445 	} else {
1446 		ASSERT(id <= VMEM_INITIAL);
1447 		vmp = &vmem0[id - 1];
1448 	}
1449 
1450 	if (vmp == NULL)
1451 		return (NULL);
1452 	bzero(vmp, sizeof (vmem_t));
1453 
1454 	(void) snprintf(vmp->vm_name, VMEM_NAMELEN, "%s", name);
1455 	(void) mutex_init(&vmp->vm_lock, USYNC_THREAD, NULL);
1456 	(void) cond_init(&vmp->vm_cv, USYNC_THREAD, NULL);
1457 	vmp->vm_cflags = vmflag;
1458 	vmflag &= VM_UMFLAGS;
1459 
1460 	vmp->vm_quantum = quantum;
1461 	vmp->vm_qshift = highbit(quantum) - 1;
1462 	nqcache = MIN(qcache_max >> vmp->vm_qshift, VMEM_NQCACHE_MAX);
1463 
1464 	for (i = 0; i <= VMEM_FREELISTS; i++) {
1465 		vfp = &vmp->vm_freelist[i];
1466 		vfp->vs_end = 1UL << i;
1467 		vfp->vs_knext = (vmem_seg_t *)(vfp + 1);
1468 		vfp->vs_kprev = (vmem_seg_t *)(vfp - 1);
1469 	}
1470 
1471 	vmp->vm_freelist[0].vs_kprev = NULL;
1472 	vmp->vm_freelist[VMEM_FREELISTS].vs_knext = NULL;
1473 	vmp->vm_freelist[VMEM_FREELISTS].vs_end = 0;
1474 	vmp->vm_hash_table = vmp->vm_hash0;
1475 	vmp->vm_hash_mask = VMEM_HASH_INITIAL - 1;
1476 	vmp->vm_hash_shift = highbit(vmp->vm_hash_mask);
1477 
1478 	vsp = &vmp->vm_seg0;
1479 	vsp->vs_anext = vsp;
1480 	vsp->vs_aprev = vsp;
1481 	vsp->vs_knext = vsp;
1482 	vsp->vs_kprev = vsp;
1483 	vsp->vs_type = VMEM_SPAN;
1484 
1485 	vsp = &vmp->vm_rotor;
1486 	vsp->vs_type = VMEM_ROTOR;
1487 	VMEM_INSERT(&vmp->vm_seg0, vsp, a);
1488 
1489 	vmp->vm_id = id;
1490 	if (source != NULL)
1491 		vmp->vm_kstat.vk_source_id = source->vm_id;
1492 	vmp->vm_source = source;
1493 	vmp->vm_source_alloc = afunc;
1494 	vmp->vm_source_free = ffunc;
1495 
1496 	if (nqcache != 0) {
1497 		vmp->vm_qcache_max = nqcache << vmp->vm_qshift;
1498 		for (i = 0; i < nqcache; i++) {
1499 			char buf[VMEM_NAMELEN + 21];
1500 			(void) snprintf(buf, sizeof (buf), "%s_%lu",
1501 			    vmp->vm_name, (long)((i + 1) * quantum));
1502 			vmp->vm_qcache[i] = umem_cache_create(buf,
1503 			    (i + 1) * quantum, quantum, NULL, NULL, NULL,
1504 			    NULL, vmp, UMC_QCACHE | UMC_NOTOUCH);
1505 			if (vmp->vm_qcache[i] == NULL) {
1506 				vmp->vm_qcache_max = i * quantum;
1507 				break;
1508 			}
1509 		}
1510 	}
1511 
1512 	(void) mutex_lock(&vmem_list_lock);
1513 	vmpp = &vmem_list;
1514 	while ((cur = *vmpp) != NULL)
1515 		vmpp = &cur->vm_next;
1516 	*vmpp = vmp;
1517 	(void) mutex_unlock(&vmem_list_lock);
1518 
1519 	if (vmp->vm_cflags & VMC_POPULATOR) {
1520 		uint_t pop_id = atomic_add_32_nv(&vmem_populators, 1);
1521 		ASSERT(pop_id <= VMEM_INITIAL);
1522 		vmem_populator[pop_id - 1] = vmp;
1523 		(void) mutex_lock(&vmp->vm_lock);
1524 		(void) vmem_populate(vmp, vmflag | VM_PANIC);
1525 		(void) mutex_unlock(&vmp->vm_lock);
1526 	}
1527 
1528 	if ((base || size) && vmem_add(vmp, base, size, vmflag) == NULL) {
1529 		vmem_destroy(vmp);
1530 		return (NULL);
1531 	}
1532 
1533 	return (vmp);
1534 }
1535 
1536 /*
1537  * Destroy arena vmp.
1538  */
1539 void
1540 vmem_destroy(vmem_t *vmp)
1541 {
1542 	vmem_t *cur, **vmpp;
1543 	vmem_seg_t *seg0 = &vmp->vm_seg0;
1544 	vmem_seg_t *vsp;
1545 	size_t leaked;
1546 	int i;
1547 
1548 	(void) mutex_lock(&vmem_list_lock);
1549 	vmpp = &vmem_list;
1550 	while ((cur = *vmpp) != vmp)
1551 		vmpp = &cur->vm_next;
1552 	*vmpp = vmp->vm_next;
1553 	(void) mutex_unlock(&vmem_list_lock);
1554 
1555 	for (i = 0; i < VMEM_NQCACHE_MAX; i++)
1556 		if (vmp->vm_qcache[i])
1557 			umem_cache_destroy(vmp->vm_qcache[i]);
1558 
1559 	leaked = vmem_size(vmp, VMEM_ALLOC);
1560 	if (leaked != 0)
1561 		umem_printf("vmem_destroy('%s'): leaked %lu bytes",
1562 		    vmp->vm_name, leaked);
1563 
1564 	if (vmp->vm_hash_table != vmp->vm_hash0)
1565 		vmem_free(vmem_hash_arena, vmp->vm_hash_table,
1566 		    (vmp->vm_hash_mask + 1) * sizeof (void *));
1567 
1568 	/*
1569 	 * Give back the segment structures for anything that's left in the
1570 	 * arena, e.g. the primary spans and their free segments.
1571 	 */
1572 	VMEM_DELETE(&vmp->vm_rotor, a);
1573 	for (vsp = seg0->vs_anext; vsp != seg0; vsp = vsp->vs_anext)
1574 		vmem_putseg_global(vsp);
1575 
1576 	while (vmp->vm_nsegfree > 0)
1577 		vmem_putseg_global(vmem_getseg(vmp));
1578 
1579 	(void) mutex_destroy(&vmp->vm_lock);
1580 	(void) cond_destroy(&vmp->vm_cv);
1581 	vmem_free(vmem_vmem_arena, vmp, sizeof (vmem_t));
1582 }
1583 
1584 /*
1585  * Resize vmp's hash table to keep the average lookup depth near 1.0.
1586  */
1587 static void
1588 vmem_hash_rescale(vmem_t *vmp)
1589 {
1590 	vmem_seg_t **old_table, **new_table, *vsp;
1591 	size_t old_size, new_size, h, nseg;
1592 
1593 	nseg = (size_t)(vmp->vm_kstat.vk_alloc - vmp->vm_kstat.vk_free);
1594 
1595 	new_size = MAX(VMEM_HASH_INITIAL, 1 << (highbit(3 * nseg + 4) - 2));
1596 	old_size = vmp->vm_hash_mask + 1;
1597 
1598 	if ((old_size >> 1) <= new_size && new_size <= (old_size << 1))
1599 		return;
1600 
1601 	new_table = vmem_alloc(vmem_hash_arena, new_size * sizeof (void *),
1602 	    VM_NOSLEEP);
1603 	if (new_table == NULL)
1604 		return;
1605 	bzero(new_table, new_size * sizeof (void *));
1606 
1607 	(void) mutex_lock(&vmp->vm_lock);
1608 
1609 	old_size = vmp->vm_hash_mask + 1;
1610 	old_table = vmp->vm_hash_table;
1611 
1612 	vmp->vm_hash_mask = new_size - 1;
1613 	vmp->vm_hash_table = new_table;
1614 	vmp->vm_hash_shift = highbit(vmp->vm_hash_mask);
1615 
1616 	for (h = 0; h < old_size; h++) {
1617 		vsp = old_table[h];
1618 		while (vsp != NULL) {
1619 			uintptr_t addr = vsp->vs_start;
1620 			vmem_seg_t *next_vsp = vsp->vs_knext;
1621 			vmem_seg_t **hash_bucket = VMEM_HASH(vmp, addr);
1622 			vsp->vs_knext = *hash_bucket;
1623 			*hash_bucket = vsp;
1624 			vsp = next_vsp;
1625 		}
1626 	}
1627 
1628 	(void) mutex_unlock(&vmp->vm_lock);
1629 
1630 	if (old_table != vmp->vm_hash0)
1631 		vmem_free(vmem_hash_arena, old_table,
1632 		    old_size * sizeof (void *));
1633 }
1634 
1635 /*
1636  * Perform periodic maintenance on all vmem arenas.
1637  */
1638 /*ARGSUSED*/
1639 void
1640 vmem_update(void *dummy)
1641 {
1642 	vmem_t *vmp;
1643 
1644 	(void) mutex_lock(&vmem_list_lock);
1645 	for (vmp = vmem_list; vmp != NULL; vmp = vmp->vm_next) {
1646 		/*
1647 		 * If threads are waiting for resources, wake them up
1648 		 * periodically so they can issue another vmem_reap()
1649 		 * to reclaim resources cached by the slab allocator.
1650 		 */
1651 		(void) cond_broadcast(&vmp->vm_cv);
1652 
1653 		/*
1654 		 * Rescale the hash table to keep the hash chains short.
1655 		 */
1656 		vmem_hash_rescale(vmp);
1657 	}
1658 	(void) mutex_unlock(&vmem_list_lock);
1659 }
1660 
1661 /*
1662  * If vmem_init is called again, we need to be able to reset the world.
1663  * That includes resetting the statics back to their original values.
1664  */
1665 void
1666 vmem_startup(void)
1667 {
1668 #ifdef UMEM_STANDALONE
1669 	vmem_id = 0;
1670 	vmem_populators = 0;
1671 	vmem_segfree = NULL;
1672 	vmem_list = NULL;
1673 	vmem_internal_arena = NULL;
1674 	vmem_seg_arena = NULL;
1675 	vmem_hash_arena = NULL;
1676 	vmem_vmem_arena = NULL;
1677 	vmem_heap = NULL;
1678 	vmem_heap_alloc = NULL;
1679 	vmem_heap_free = NULL;
1680 
1681 	bzero(vmem0, sizeof (vmem0));
1682 	bzero(vmem_populator, sizeof (vmem_populator));
1683 	bzero(vmem_seg0, sizeof (vmem_seg0));
1684 #endif
1685 }
1686 
1687 /*
1688  * Prepare vmem for use.
1689  */
1690 vmem_t *
1691 vmem_init(const char *parent_name, size_t parent_quantum,
1692     vmem_alloc_t *parent_alloc, vmem_free_t *parent_free,
1693     const char *heap_name, void *heap_start, size_t heap_size,
1694     size_t heap_quantum, vmem_alloc_t *heap_alloc, vmem_free_t *heap_free)
1695 {
1696 	uint32_t id;
1697 	int nseg = VMEM_SEG_INITIAL;
1698 	vmem_t *parent, *heap;
1699 
1700 	ASSERT(vmem_internal_arena == NULL);
1701 
1702 	while (--nseg >= 0)
1703 		vmem_putseg_global(&vmem_seg0[nseg]);
1704 
1705 	if (parent_name != NULL) {
1706 		parent = vmem_create(parent_name,
1707 		    heap_start, heap_size, parent_quantum,
1708 		    NULL, NULL, NULL, 0,
1709 		    VM_SLEEP | VMC_POPULATOR);
1710 		heap_start = NULL;
1711 		heap_size = 0;
1712 	} else {
1713 		ASSERT(parent_alloc == NULL && parent_free == NULL);
1714 		parent = NULL;
1715 	}
1716 
1717 	heap = vmem_create(heap_name,
1718 	    heap_start, heap_size, heap_quantum,
1719 	    parent_alloc, parent_free, parent, 0,
1720 	    VM_SLEEP | VMC_POPULATOR);
1721 
1722 	vmem_heap = heap;
1723 	vmem_heap_alloc = heap_alloc;
1724 	vmem_heap_free = heap_free;
1725 
1726 	vmem_internal_arena = vmem_create("vmem_internal",
1727 	    NULL, 0, heap_quantum,
1728 	    heap_alloc, heap_free, heap, 0,
1729 	    VM_SLEEP | VMC_POPULATOR);
1730 
1731 	vmem_seg_arena = vmem_create("vmem_seg",
1732 	    NULL, 0, heap_quantum,
1733 	    vmem_alloc, vmem_free, vmem_internal_arena, 0,
1734 	    VM_SLEEP | VMC_POPULATOR);
1735 
1736 	vmem_hash_arena = vmem_create("vmem_hash",
1737 	    NULL, 0, 8,
1738 	    vmem_alloc, vmem_free, vmem_internal_arena, 0,
1739 	    VM_SLEEP);
1740 
1741 	vmem_vmem_arena = vmem_create("vmem_vmem",
1742 	    vmem0, sizeof (vmem0), 1,
1743 	    vmem_alloc, vmem_free, vmem_internal_arena, 0,
1744 	    VM_SLEEP);
1745 
1746 	for (id = 0; id < vmem_id; id++)
1747 		(void) vmem_xalloc(vmem_vmem_arena, sizeof (vmem_t),
1748 		    1, 0, 0, &vmem0[id], &vmem0[id + 1],
1749 		    VM_NOSLEEP | VM_BESTFIT | VM_PANIC);
1750 
1751 	return (heap);
1752 }
1753 
1754 void
1755 vmem_no_debug(void)
1756 {
1757 	/*
1758 	 * This size must be a multiple of the minimum required alignment,
1759 	 * since vmem_populate allocates them compactly.
1760 	 */
1761 	vmem_seg_size = P2ROUNDUP(offsetof(vmem_seg_t, vs_thread),
1762 	    sizeof (hrtime_t));
1763 }
1764 
1765 /*
1766  * Lockup and release, for fork1(2) handling.
1767  */
1768 void
1769 vmem_lockup(void)
1770 {
1771 	vmem_t *cur;
1772 
1773 	(void) mutex_lock(&vmem_list_lock);
1774 	(void) mutex_lock(&vmem_nosleep_lock.vmpl_mutex);
1775 
1776 	/*
1777 	 * Lock up and broadcast all arenas.
1778 	 */
1779 	for (cur = vmem_list; cur != NULL; cur = cur->vm_next) {
1780 		(void) mutex_lock(&cur->vm_lock);
1781 		(void) cond_broadcast(&cur->vm_cv);
1782 	}
1783 
1784 	(void) mutex_lock(&vmem_segfree_lock);
1785 }
1786 
1787 void
1788 vmem_release(void)
1789 {
1790 	vmem_t *cur;
1791 
1792 	(void) mutex_unlock(&vmem_nosleep_lock.vmpl_mutex);
1793 
1794 	for (cur = vmem_list; cur != NULL; cur = cur->vm_next)
1795 		(void) mutex_unlock(&cur->vm_lock);
1796 
1797 	(void) mutex_unlock(&vmem_segfree_lock);
1798 	(void) mutex_unlock(&vmem_list_lock);
1799 }
1800