1/*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 1987, 1993
5 *	The Regents of the University of California.
6 * Copyright (c) 2005, 2009 Robert N. M. Watson
7 * All rights reserved.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 * 1. Redistributions of source code must retain the above copyright
13 *    notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 *    notice, this list of conditions and the following disclaimer in the
16 *    documentation and/or other materials provided with the distribution.
17 * 3. 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 *	@(#)malloc.h	8.5 (Berkeley) 5/3/95
34 * $FreeBSD$
35 */
36
37#ifndef _SYS_MALLOC_H_
38#define	_SYS_MALLOC_H_
39
40#ifndef _STANDALONE
41#include <sys/param.h>
42#ifdef _KERNEL
43#include <sys/systm.h>
44#endif
45#include <sys/queue.h>
46#include <sys/_lock.h>
47#include <sys/_mutex.h>
48#include <machine/_limits.h>
49
50#define	MINALLOCSIZE	UMA_SMALLEST_UNIT
51
52/*
53 * Flags to memory allocation functions.
54 */
55#define	M_NOWAIT	0x0001		/* do not block */
56#define	M_WAITOK	0x0002		/* ok to block */
57#define	M_ZERO		0x0100		/* bzero the allocation */
58#define	M_NOVM		0x0200		/* don't ask VM for pages */
59#define	M_USE_RESERVE	0x0400		/* can alloc out of reserve memory */
60#define	M_NODUMP	0x0800		/* don't dump pages in this allocation */
61#define	M_FIRSTFIT	0x1000		/* only for vmem, fast fit */
62#define	M_BESTFIT	0x2000		/* only for vmem, low fragmentation */
63#define	M_EXEC		0x4000		/* allocate executable space */
64#define	M_NEXTFIT	0x8000		/* only for vmem, follow cursor */
65
66#define	M_MAGIC		877983977	/* time when first defined :-) */
67
68/*
69 * Two malloc type structures are present: malloc_type, which is used by a
70 * type owner to declare the type, and malloc_type_internal, which holds
71 * malloc-owned statistics and other ABI-sensitive fields, such as the set of
72 * malloc statistics indexed by the compile-time MAXCPU constant.
73 * Applications should avoid introducing dependence on the allocator private
74 * data layout and size.
75 *
76 * The malloc_type ks_next field is protected by malloc_mtx.  Other fields in
77 * malloc_type are static after initialization so unsynchronized.
78 *
79 * Statistics in malloc_type_stats are written only when holding a critical
80 * section and running on the CPU associated with the index into the stat
81 * array, but read lock-free resulting in possible (minor) races, which the
82 * monitoring app should take into account.
83 */
84struct malloc_type_stats {
85	uint64_t	mts_memalloced;	/* Bytes allocated on CPU. */
86	uint64_t	mts_memfreed;	/* Bytes freed on CPU. */
87	uint64_t	mts_numallocs;	/* Number of allocates on CPU. */
88	uint64_t	mts_numfrees;	/* number of frees on CPU. */
89	uint64_t	mts_size;	/* Bitmask of sizes allocated on CPU. */
90	uint64_t	_mts_reserved1;	/* Reserved field. */
91	uint64_t	_mts_reserved2;	/* Reserved field. */
92	uint64_t	_mts_reserved3;	/* Reserved field. */
93};
94
95/*
96 * Index definitions for the mti_probes[] array.
97 */
98#define DTMALLOC_PROBE_MALLOC		0
99#define DTMALLOC_PROBE_FREE		1
100#define DTMALLOC_PROBE_MAX		2
101
102struct malloc_type_internal {
103	uint32_t	mti_probes[DTMALLOC_PROBE_MAX];
104					/* DTrace probe ID array. */
105	u_char		mti_zone;
106	struct malloc_type_stats	*mti_stats;
107};
108
109/*
110 * Public data structure describing a malloc type.  Private data is hung off
111 * of ks_handle to avoid encoding internal malloc(9) data structures in
112 * modules, which will statically allocate struct malloc_type.
113 */
114struct malloc_type {
115	struct malloc_type *ks_next;	/* Next in global chain. */
116	u_long		 ks_magic;	/* Detect programmer error. */
117	const char	*ks_shortdesc;	/* Printable type name. */
118	void		*ks_handle;	/* Priv. data, was lo_class. */
119};
120
121/*
122 * Statistics structure headers for user space.  The kern.malloc sysctl
123 * exposes a structure stream consisting of a stream header, then a series of
124 * malloc type headers and statistics structures (quantity maxcpus).  For
125 * convenience, the kernel will provide the current value of maxcpus at the
126 * head of the stream.
127 */
128#define	MALLOC_TYPE_STREAM_VERSION	0x00000001
129struct malloc_type_stream_header {
130	uint32_t	mtsh_version;	/* Stream format version. */
131	uint32_t	mtsh_maxcpus;	/* Value of MAXCPU for stream. */
132	uint32_t	mtsh_count;	/* Number of records. */
133	uint32_t	_mtsh_pad;	/* Pad/reserved field. */
134};
135
136#define	MALLOC_MAX_NAME	32
137struct malloc_type_header {
138	char				mth_name[MALLOC_MAX_NAME];
139};
140
141#ifdef _KERNEL
142#define	MALLOC_DEFINE(type, shortdesc, longdesc)			\
143	struct malloc_type type[1] = {					\
144		{ NULL, M_MAGIC, shortdesc, NULL }			\
145	};								\
146	SYSINIT(type##_init, SI_SUB_KMEM, SI_ORDER_THIRD, malloc_init,	\
147	    type);							\
148	SYSUNINIT(type##_uninit, SI_SUB_KMEM, SI_ORDER_ANY,		\
149	    malloc_uninit, type)
150
151#define	MALLOC_DECLARE(type) \
152	extern struct malloc_type type[1]
153
154MALLOC_DECLARE(M_CACHE);
155MALLOC_DECLARE(M_DEVBUF);
156MALLOC_DECLARE(M_TEMP);
157
158/*
159 * XXX this should be declared in <sys/uio.h>, but that tends to fail
160 * because <sys/uio.h> is included in a header before the source file
161 * has a chance to include <sys/malloc.h> to get MALLOC_DECLARE() defined.
162 */
163MALLOC_DECLARE(M_IOV);
164
165struct domainset;
166extern struct mtx malloc_mtx;
167
168/*
169 * Function type used when iterating over the list of malloc types.
170 */
171typedef void malloc_type_list_func_t(struct malloc_type *, void *);
172
173void	contigfree(void *addr, unsigned long size, struct malloc_type *type);
174void	*contigmalloc(unsigned long size, struct malloc_type *type, int flags,
175	    vm_paddr_t low, vm_paddr_t high, unsigned long alignment,
176	    vm_paddr_t boundary) __malloc_like __result_use_check
177	    __alloc_size(1) __alloc_align(6);
178void	*contigmalloc_domainset(unsigned long size, struct malloc_type *type,
179	    struct domainset *ds, int flags, vm_paddr_t low, vm_paddr_t high,
180	    unsigned long alignment, vm_paddr_t boundary)
181	    __malloc_like __result_use_check __alloc_size(1) __alloc_align(7);
182void	free(void *addr, struct malloc_type *type);
183void	zfree(void *addr, struct malloc_type *type);
184void	*malloc(size_t size, struct malloc_type *type, int flags) __malloc_like
185	    __result_use_check __alloc_size(1);
186/*
187 * Try to optimize malloc(..., ..., M_ZERO) allocations by doing zeroing in
188 * place if the size is known at compilation time.
189 *
190 * Passing the flag down requires malloc to blindly zero the entire object.
191 * In practice a lot of the zeroing can be avoided if most of the object
192 * gets explicitly initialized after the allocation. Letting the compiler
193 * zero in place gives it the opportunity to take advantage of this state.
194 *
195 * Note that the operation is only applicable if both flags and size are
196 * known at compilation time. If M_ZERO is passed but M_WAITOK is not, the
197 * allocation can fail and a NULL check is needed. However, if M_WAITOK is
198 * passed we know the allocation must succeed and the check can be elided.
199 *
200 *	_malloc_item = malloc(_size, type, (flags) &~ M_ZERO);
201 *	if (((flags) & M_WAITOK) != 0 || _malloc_item != NULL)
202 *		bzero(_malloc_item, _size);
203 *
204 * If the flag is set, the compiler knows the left side is always true,
205 * therefore the entire statement is true and the callsite is:
206 *
207 *	_malloc_item = malloc(_size, type, (flags) &~ M_ZERO);
208 *	bzero(_malloc_item, _size);
209 *
210 * If the flag is not set, the compiler knows the left size is always false
211 * and the NULL check is needed, therefore the callsite is:
212 *
213 * 	_malloc_item = malloc(_size, type, (flags) &~ M_ZERO);
214 *	if (_malloc_item != NULL)
215 *		bzero(_malloc_item, _size);
216 *
217 * The implementation is a macro because of what appears to be a clang 6 bug:
218 * an inline function variant ended up being compiled to a mere malloc call
219 * regardless of argument. gcc generates expected code (like the above).
220 */
221#define	malloc(size, type, flags) ({					\
222	void *_malloc_item;						\
223	size_t _size = (size);						\
224	if (__builtin_constant_p(size) && __builtin_constant_p(flags) &&\
225	    ((flags) & M_ZERO) != 0) {					\
226		_malloc_item = malloc(_size, type, (flags) &~ M_ZERO);	\
227		if (((flags) & M_WAITOK) != 0 ||			\
228		    __predict_true(_malloc_item != NULL))		\
229			bzero(_malloc_item, _size);			\
230	} else {							\
231		_malloc_item = malloc(_size, type, flags);		\
232	}								\
233	_malloc_item;							\
234})
235
236void	*malloc_domainset(size_t size, struct malloc_type *type,
237	    struct domainset *ds, int flags) __malloc_like __result_use_check
238	    __alloc_size(1);
239void	*mallocarray(size_t nmemb, size_t size, struct malloc_type *type,
240	    int flags) __malloc_like __result_use_check
241	    __alloc_size2(1, 2);
242void	malloc_init(void *);
243int	malloc_last_fail(void);
244void	malloc_type_allocated(struct malloc_type *type, unsigned long size);
245void	malloc_type_freed(struct malloc_type *type, unsigned long size);
246void	malloc_type_list(malloc_type_list_func_t *, void *);
247void	malloc_uninit(void *);
248size_t	malloc_usable_size(const void *);
249void	*realloc(void *addr, size_t size, struct malloc_type *type, int flags)
250	    __result_use_check __alloc_size(2);
251void	*reallocf(void *addr, size_t size, struct malloc_type *type, int flags)
252	    __result_use_check __alloc_size(2);
253
254struct malloc_type *malloc_desc2type(const char *desc);
255
256/*
257 * This is sqrt(SIZE_MAX+1), as s1*s2 <= SIZE_MAX
258 * if both s1 < MUL_NO_OVERFLOW and s2 < MUL_NO_OVERFLOW
259 */
260#define MUL_NO_OVERFLOW		(1UL << (sizeof(size_t) * 8 / 2))
261static inline bool
262WOULD_OVERFLOW(size_t nmemb, size_t size)
263{
264
265	return ((nmemb >= MUL_NO_OVERFLOW || size >= MUL_NO_OVERFLOW) &&
266	    nmemb > 0 && __SIZE_T_MAX / nmemb < size);
267}
268#undef MUL_NO_OVERFLOW
269#endif /* _KERNEL */
270
271#else
272/*
273 * The native stand malloc / free interface we're mapping to
274 */
275extern void Free(void *p, const char *file, int line);
276extern void *Malloc(size_t bytes, const char *file, int line);
277
278/*
279 * Minimal standalone malloc implementation / environment. None of the
280 * flags mean anything and there's no need declare malloc types.
281 * Define the simple alloc / free routines in terms of Malloc and
282 * Free. None of the kernel features that this stuff disables are needed.
283 *
284 * XXX we are setting ourselves up for a potential crash if we can't allocate
285 * memory for a M_WAITOK call.
286 */
287#define M_WAITOK 0
288#define M_ZERO 0
289#define M_NOWAIT 0
290#define MALLOC_DECLARE(x)
291
292#define kmem_zalloc(size, flags) Malloc((size), __FILE__, __LINE__)
293#define kmem_free(p, size) Free(p, __FILE__, __LINE__)
294
295/*
296 * ZFS mem.h define that's the OpenZFS porting layer way of saying
297 * M_WAITOK. Given the above, it will also be a nop.
298 */
299#define KM_SLEEP M_WAITOK
300#endif /* _STANDALONE */
301#endif /* !_SYS_MALLOC_H_ */
302