kern_malloc.c revision 15543 
1/*
2 * Copyright (c) 1987, 1991, 1993
3 *	The Regents of the University of California.  All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 * 3. All advertising materials mentioning features or use of this software
14 *    must display the following acknowledgement:
15 *	This product includes software developed by the University of
16 *	California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 *    may be used to endorse or promote products derived from this software
19 *    without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 *	@(#)kern_malloc.c	8.3 (Berkeley) 1/4/94
34 * $Id: kern_malloc.c,v 1.20 1996/05/02 10:43:17 phk Exp $
35 */
36
37#include <sys/param.h>
38#include <sys/systm.h>
39#include <sys/proc.h>
40#include <sys/kernel.h>
41#include <sys/malloc.h>
42#include <sys/mbuf.h>
43#include <sys/vmmeter.h>
44
45#include <vm/vm.h>
46#include <vm/vm_param.h>
47#include <vm/vm_kern.h>
48#include <vm/vm_extern.h>
49
50static void kmeminit __P((void *));
51SYSINIT(kmem, SI_SUB_KMEM, SI_ORDER_FIRST, kmeminit, NULL)
52
53static struct kmembuckets bucket[MINBUCKET + 16];
54struct kmemstats kmemstats[M_LAST];
55struct kmemusage *kmemusage;
56char *kmembase, *kmemlimit;
57char *memname[] = INITKMEMNAMES;
58
59#ifdef DIAGNOSTIC
60/*
61 * This structure provides a set of masks to catch unaligned frees.
62 */
63static long addrmask[] = { 0,
64	0x00000001, 0x00000003, 0x00000007, 0x0000000f,
65	0x0000001f, 0x0000003f, 0x0000007f, 0x000000ff,
66	0x000001ff, 0x000003ff, 0x000007ff, 0x00000fff,
67	0x00001fff, 0x00003fff, 0x00007fff, 0x0000ffff,
68};
69
70/*
71 * The WEIRD_ADDR is used as known text to copy into free objects so
72 * that modifications after frees can be detected.
73 */
74#define WEIRD_ADDR	0xdeadc0de
75#define MAX_COPY	64
76
77/*
78 * Normally the first word of the structure is used to hold the list
79 * pointer for free objects. However, when running with diagnostics,
80 * we use the third and fourth fields, so as to catch modifications
81 * in the most commonly trashed first two words.
82 */
83struct freelist {
84	long	spare0;
85	short	type;
86	long	spare1;
87	caddr_t	next;
88};
89#else /* !DIAGNOSTIC */
90struct freelist {
91	caddr_t	next;
92};
93#endif /* DIAGNOSTIC */
94
95/*
96 * Allocate a block of memory
97 */
98void *
99malloc(size, type, flags)
100	unsigned long size;
101	int type, flags;
102{
103	register struct kmembuckets *kbp;
104	register struct kmemusage *kup;
105	register struct freelist *freep;
106	long indx, npg, allocsize;
107	int s;
108	caddr_t va, cp, savedlist;
109#ifdef DIAGNOSTIC
110	long *end, *lp;
111	int copysize;
112	char *savedtype;
113#endif
114#ifdef KMEMSTATS
115	register struct kmemstats *ksp = &kmemstats[type];
116
117	if (((unsigned long)type) > M_LAST)
118		panic("malloc - bogus type");
119#endif
120	indx = BUCKETINDX(size);
121	kbp = &bucket[indx];
122	s = splhigh();
123#ifdef KMEMSTATS
124	while (ksp->ks_memuse >= ksp->ks_limit) {
125		if (flags & M_NOWAIT) {
126			splx(s);
127			return ((void *) NULL);
128		}
129		if (ksp->ks_limblocks < 65535)
130			ksp->ks_limblocks++;
131		tsleep((caddr_t)ksp, PSWP+2, memname[type], 0);
132	}
133	ksp->ks_size |= 1 << indx;
134#endif
135#ifdef DIAGNOSTIC
136	copysize = 1 << indx < MAX_COPY ? 1 << indx : MAX_COPY;
137#endif
138	if (kbp->kb_next == NULL) {
139		kbp->kb_last = NULL;
140		if (size > MAXALLOCSAVE)
141			allocsize = roundup(size, PAGE_SIZE);
142		else
143			allocsize = 1 << indx;
144		npg = btoc(allocsize);
145		va = (caddr_t) kmem_malloc(kmem_map, (vm_size_t)ctob(npg), flags);
146		if (va == NULL) {
147			splx(s);
148			return ((void *) NULL);
149		}
150#ifdef KMEMSTATS
151		kbp->kb_total += kbp->kb_elmpercl;
152#endif
153		kup = btokup(va);
154		kup->ku_indx = indx;
155		if (allocsize > MAXALLOCSAVE) {
156			if (npg > 65535)
157				panic("malloc: allocation too large");
158			kup->ku_pagecnt = npg;
159#ifdef KMEMSTATS
160			ksp->ks_memuse += allocsize;
161#endif
162			goto out;
163		}
164#ifdef KMEMSTATS
165		kup->ku_freecnt = kbp->kb_elmpercl;
166		kbp->kb_totalfree += kbp->kb_elmpercl;
167#endif
168		/*
169		 * Just in case we blocked while allocating memory,
170		 * and someone else also allocated memory for this
171		 * bucket, don't assume the list is still empty.
172		 */
173		savedlist = kbp->kb_next;
174		kbp->kb_next = cp = va + (npg * PAGE_SIZE) - allocsize;
175		for (;;) {
176			freep = (struct freelist *)cp;
177#ifdef DIAGNOSTIC
178			/*
179			 * Copy in known text to detect modification
180			 * after freeing.
181			 */
182			end = (long *)&cp[copysize];
183			for (lp = (long *)cp; lp < end; lp++)
184				*lp = WEIRD_ADDR;
185			freep->type = M_FREE;
186#endif /* DIAGNOSTIC */
187			if (cp <= va)
188				break;
189			cp -= allocsize;
190			freep->next = cp;
191		}
192		freep->next = savedlist;
193		if (kbp->kb_last == NULL)
194			kbp->kb_last = (caddr_t)freep;
195	}
196	va = kbp->kb_next;
197	kbp->kb_next = ((struct freelist *)va)->next;
198#ifdef DIAGNOSTIC
199	freep = (struct freelist *)va;
200	savedtype = (unsigned)freep->type < M_LAST ?
201		memname[freep->type] : "???";
202	if (kbp->kb_next &&
203	    !kernacc(kbp->kb_next, sizeof(struct freelist), 0)) {
204		printf("%s of object %p size %ld %s %s (invalid addr %p)\n",
205			"Data modified on freelist: word 2.5", va, size,
206			"previous type", savedtype, kbp->kb_next);
207		kbp->kb_next = NULL;
208	}
209#if BYTE_ORDER == BIG_ENDIAN
210	freep->type = WEIRD_ADDR >> 16;
211#endif
212#if BYTE_ORDER == LITTLE_ENDIAN
213	freep->type = (short)WEIRD_ADDR;
214#endif
215	if (((long)(&freep->next)) & 0x2)
216		freep->next = (caddr_t)((WEIRD_ADDR >> 16)|(WEIRD_ADDR << 16));
217	else
218		freep->next = (caddr_t)WEIRD_ADDR;
219	end = (long *)&va[copysize];
220	for (lp = (long *)va; lp < end; lp++) {
221		if (*lp == WEIRD_ADDR)
222			continue;
223		printf("%s %d of object %p size %ld %s %s (0x%lx != 0x%x)\n",
224			"Data modified on freelist: word", lp - (long *)va,
225			va, size, "previous type", savedtype, *lp, WEIRD_ADDR);
226		break;
227	}
228	freep->spare0 = 0;
229#endif /* DIAGNOSTIC */
230#ifdef KMEMSTATS
231	kup = btokup(va);
232	if (kup->ku_indx != indx)
233		panic("malloc: wrong bucket");
234	if (kup->ku_freecnt == 0)
235		panic("malloc: lost data");
236	kup->ku_freecnt--;
237	kbp->kb_totalfree--;
238	ksp->ks_memuse += 1 << indx;
239out:
240	kbp->kb_calls++;
241	ksp->ks_inuse++;
242	ksp->ks_calls++;
243	if (ksp->ks_memuse > ksp->ks_maxused)
244		ksp->ks_maxused = ksp->ks_memuse;
245#else
246out:
247#endif
248	splx(s);
249	return ((void *) va);
250}
251
252/*
253 * Free a block of memory allocated by malloc.
254 */
255void
256free(addr, type)
257	void *addr;
258	int type;
259{
260	register struct kmembuckets *kbp;
261	register struct kmemusage *kup;
262	register struct freelist *freep;
263	long size;
264	int s;
265#ifdef DIAGNOSTIC
266	caddr_t cp;
267	long *end, *lp, alloc, copysize;
268#endif
269#ifdef KMEMSTATS
270	register struct kmemstats *ksp = &kmemstats[type];
271#endif
272
273#ifdef DIAGNOSTIC
274	if ((char *)addr < kmembase || (char *)addr >= kmemlimit) {
275		panic("free: address 0x%x out of range", addr);
276	}
277	if ((u_long)type > M_LAST) {
278		panic("free: type %d out of range", type);
279	}
280#endif
281	kup = btokup(addr);
282	size = 1 << kup->ku_indx;
283	kbp = &bucket[kup->ku_indx];
284	s = splhigh();
285#ifdef DIAGNOSTIC
286	/*
287	 * Check for returns of data that do not point to the
288	 * beginning of the allocation.
289	 */
290	if (size > PAGE_SIZE)
291		alloc = addrmask[BUCKETINDX(PAGE_SIZE)];
292	else
293		alloc = addrmask[kup->ku_indx];
294	if (((u_long)addr & alloc) != 0)
295		panic("free: unaligned addr 0x%x, size %d, type %s, mask %d",
296			addr, size, memname[type], alloc);
297#endif /* DIAGNOSTIC */
298	if (size > MAXALLOCSAVE) {
299		kmem_free(kmem_map, (vm_offset_t)addr, ctob(kup->ku_pagecnt));
300#ifdef KMEMSTATS
301		size = kup->ku_pagecnt << PAGE_SHIFT;
302		ksp->ks_memuse -= size;
303		kup->ku_indx = 0;
304		kup->ku_pagecnt = 0;
305		if (ksp->ks_memuse + size >= ksp->ks_limit &&
306		    ksp->ks_memuse < ksp->ks_limit)
307			wakeup((caddr_t)ksp);
308		ksp->ks_inuse--;
309		kbp->kb_total -= 1;
310#endif
311		splx(s);
312		return;
313	}
314	freep = (struct freelist *)addr;
315#ifdef DIAGNOSTIC
316	/*
317	 * Check for multiple frees. Use a quick check to see if
318	 * it looks free before laboriously searching the freelist.
319	 */
320	if (freep->spare0 == WEIRD_ADDR) {
321		for (cp = kbp->kb_next; cp; cp = *(caddr_t *)cp) {
322			if (addr != cp)
323				continue;
324			printf("multiply freed item %p\n", addr);
325			panic("free: duplicated free");
326		}
327	}
328	/*
329	 * Copy in known text to detect modification after freeing
330	 * and to make it look free. Also, save the type being freed
331	 * so we can list likely culprit if modification is detected
332	 * when the object is reallocated.
333	 */
334	copysize = size < MAX_COPY ? size : MAX_COPY;
335	end = (long *)&((caddr_t)addr)[copysize];
336	for (lp = (long *)addr; lp < end; lp++)
337		*lp = WEIRD_ADDR;
338	freep->type = type;
339#endif /* DIAGNOSTIC */
340#ifdef KMEMSTATS
341	kup->ku_freecnt++;
342	if (kup->ku_freecnt >= kbp->kb_elmpercl)
343		if (kup->ku_freecnt > kbp->kb_elmpercl)
344			panic("free: multiple frees");
345		else if (kbp->kb_totalfree > kbp->kb_highwat)
346			kbp->kb_couldfree++;
347	kbp->kb_totalfree++;
348	ksp->ks_memuse -= size;
349	if (ksp->ks_memuse + size >= ksp->ks_limit &&
350	    ksp->ks_memuse < ksp->ks_limit)
351		wakeup((caddr_t)ksp);
352	ksp->ks_inuse--;
353#endif
354	if (kbp->kb_next == NULL)
355		kbp->kb_next = addr;
356	else
357		((struct freelist *)kbp->kb_last)->next = addr;
358	freep->next = NULL;
359	kbp->kb_last = addr;
360	splx(s);
361}
362
363/*
364 * Initialize the kernel memory allocator
365 */
366/* ARGSUSED*/
367static void
368kmeminit(dummy)
369	void *dummy;
370{
371	register long indx;
372	int npg;
373
374#if	((MAXALLOCSAVE & (MAXALLOCSAVE - 1)) != 0)
375		ERROR!_kmeminit:_MAXALLOCSAVE_not_power_of_2
376#endif
377#if	(MAXALLOCSAVE > MINALLOCSIZE * 32768)
378		ERROR!_kmeminit:_MAXALLOCSAVE_too_big
379#endif
380#if	(MAXALLOCSAVE < PAGE_SIZE)
381		ERROR!_kmeminit:_MAXALLOCSAVE_too_small
382#endif
383	npg = (nmbclusters * MCLBYTES + VM_KMEM_SIZE) / PAGE_SIZE;
384
385	kmemusage = (struct kmemusage *) kmem_alloc(kernel_map,
386		(vm_size_t)(npg * sizeof(struct kmemusage)));
387	kmem_map = kmem_suballoc(kernel_map, (vm_offset_t *)&kmembase,
388		(vm_offset_t *)&kmemlimit, (vm_size_t)(npg * PAGE_SIZE),
389		FALSE);
390#ifdef KMEMSTATS
391	for (indx = 0; indx < MINBUCKET + 16; indx++) {
392		if (1 << indx >= PAGE_SIZE)
393			bucket[indx].kb_elmpercl = 1;
394		else
395			bucket[indx].kb_elmpercl = PAGE_SIZE / (1 << indx);
396		bucket[indx].kb_highwat = 5 * bucket[indx].kb_elmpercl;
397	}
398	/*
399	 * Limit maximum memory for each type to 60% of malloc area size or
400	 * 60% of physical memory, whichever is smaller.
401	 */
402	for (indx = 0; indx < M_LAST; indx++) {
403		kmemstats[indx].ks_limit = min(cnt.v_page_count * PAGE_SIZE,
404			(npg * PAGE_SIZE - nmbclusters * MCLBYTES)) * 6 / 10;
405	}
406#endif
407}
408