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 * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
23 * Use is subject to license terms.
24 * Copyright (c) 2015, Joyent, Inc.
25 */
26
27 /* Copyright (c) 1984, 1986, 1987, 1988, 1989 AT&T */
28 /* All Rights Reserved */
29
30 /*
31 * University Copyright- Copyright (c) 1982, 1986, 1988
32 * The Regents of the University of California
33 * All Rights Reserved
34 *
35 * University Acknowledgment- Portions of this document are derived from
36 * software developed by the University of California, Berkeley, and its
37 * contributors.
38 */
39
40 #ifndef _VM_SEG_H
41 #define _VM_SEG_H
42
43 #include <sys/vnode.h>
44 #include <sys/avl.h>
45 #include <vm/seg_enum.h>
46 #include <vm/faultcode.h>
47 #include <vm/hat.h>
48
49 #ifdef __cplusplus
50 extern "C" {
51 #endif
52
53 /*
54 * VM - Segments.
55 */
56
57 struct anon_map;
58
59 /*
60 * kstat statistics for segment advise
61 */
62 typedef struct {
63 kstat_named_t MADV_FREE_hit;
64 kstat_named_t MADV_FREE_miss;
65 } segadvstat_t;
66
67 /*
68 * memory object ids
69 */
70 typedef struct memid { u_longlong_t val[2]; } memid_t;
71
72 /*
73 * An address space contains a set of segments, managed by drivers.
74 * Drivers support mapped devices, sharing, copy-on-write, etc.
75 *
76 * The seg structure contains a lock to prevent races, the base virtual
77 * address and size of the segment, a back pointer to the containing
78 * address space, pointers to maintain an AVL tree of segments in the
79 * same address space, and procedure and data hooks for the driver.
80 * The AVL tree of segments for the address space is sorted by
81 * ascending base addresses and overlapping segments are not allowed.
82 *
83 * After a segment is created, faults may occur on pages of the segment.
84 * When a fault occurs, the fault handling code must get the desired
85 * object and set up the hardware translation to the object. For some
86 * objects, the fault handling code also implements copy-on-write.
87 *
88 * When the hat wants to unload a translation, it can call the unload
89 * routine which is responsible for processing reference and modify bits.
90 *
91 * Each segment is protected by it's containing address space lock. To
92 * access any field in the segment structure, the "as" must be locked.
93 * If a segment field is to be modified, the address space lock must be
94 * write locked.
95 */
96
97 typedef struct pcache_link {
98 struct pcache_link *p_lnext;
99 struct pcache_link *p_lprev;
100 } pcache_link_t;
101
102 typedef struct seg {
103 caddr_t s_base; /* base virtual address */
104 size_t s_size; /* size in bytes */
105 uint_t s_szc; /* max page size code */
106 uint_t s_flags; /* flags for segment, see below */
107 struct as *s_as; /* containing address space */
108 avl_node_t s_tree; /* AVL tree links to segs in this as */
109 struct seg_ops *s_ops; /* ops vector: see below */
110 void *s_data; /* private data for instance */
111 kmutex_t s_pmtx; /* protects seg's pcache list */
112 pcache_link_t s_phead; /* head of seg's pcache list */
113 } seg_t;
114
115 #define S_PURGE (0x01) /* seg should be purged in as_gap() */
116
117 struct seg_ops {
118 int (*dup)(struct seg *, struct seg *);
119 int (*unmap)(struct seg *, caddr_t, size_t);
120 void (*free)(struct seg *);
121 faultcode_t (*fault)(struct hat *, struct seg *, caddr_t, size_t,
122 enum fault_type, enum seg_rw);
123 faultcode_t (*faulta)(struct seg *, caddr_t);
124 int (*setprot)(struct seg *, caddr_t, size_t, uint_t);
125 int (*checkprot)(struct seg *, caddr_t, size_t, uint_t);
126 int (*kluster)(struct seg *, caddr_t, ssize_t);
127 int (*sync)(struct seg *, caddr_t, size_t, int, uint_t);
128 size_t (*incore)(struct seg *, caddr_t, size_t, char *);
129 int (*lockop)(struct seg *, caddr_t, size_t, int, int, ulong_t *,
130 size_t);
131 int (*getprot)(struct seg *, caddr_t, size_t, uint_t *);
132 u_offset_t (*getoffset)(struct seg *, caddr_t);
133 int (*gettype)(struct seg *, caddr_t);
134 int (*getvp)(struct seg *, caddr_t, struct vnode **);
135 int (*advise)(struct seg *, caddr_t, size_t, uint_t);
136 void (*dump)(struct seg *);
137 int (*pagelock)(struct seg *, caddr_t, size_t, struct page ***,
138 enum lock_type, enum seg_rw);
139 int (*setpagesize)(struct seg *, caddr_t, size_t, uint_t);
140 int (*getmemid)(struct seg *, caddr_t, memid_t *);
141 struct lgrp_mem_policy_info *(*getpolicy)(struct seg *, caddr_t);
142 int (*capable)(struct seg *, segcapability_t);
143 int (*inherit)(struct seg *, caddr_t, size_t, uint_t);
144 };
145
146 #ifdef _KERNEL
147
148 /*
149 * Generic segment operations
150 */
151 extern void seg_init(void);
152 extern struct seg *seg_alloc(struct as *as, caddr_t base, size_t size);
153 extern int seg_attach(struct as *as, caddr_t base, size_t size,
154 struct seg *seg);
155 extern void seg_unmap(struct seg *seg);
156 extern void seg_free(struct seg *seg);
157
158 /*
159 * functions for pagelock cache support
160 */
161 typedef int (*seg_preclaim_cbfunc_t)(void *, caddr_t, size_t,
162 struct page **, enum seg_rw, int);
163
164 extern struct page **seg_plookup(struct seg *seg, struct anon_map *amp,
165 caddr_t addr, size_t len, enum seg_rw rw, uint_t flags);
166 extern void seg_pinactive(struct seg *seg, struct anon_map *amp,
167 caddr_t addr, size_t len, struct page **pp, enum seg_rw rw,
168 uint_t flags, seg_preclaim_cbfunc_t callback);
169
170 extern void seg_ppurge(struct seg *seg, struct anon_map *amp,
171 uint_t flags);
172 extern void seg_ppurge_wiredpp(struct page **pp);
173
174 extern int seg_pinsert_check(struct seg *seg, struct anon_map *amp,
175 caddr_t addr, size_t len, uint_t flags);
176 extern int seg_pinsert(struct seg *seg, struct anon_map *amp,
177 caddr_t addr, size_t len, size_t wlen, struct page **pp, enum seg_rw rw,
178 uint_t flags, seg_preclaim_cbfunc_t callback);
179
180 extern void seg_pasync_thread(void);
181 extern void seg_preap(void);
182 extern int seg_p_disable(void);
183 extern void seg_p_enable(void);
184
185 extern segadvstat_t segadvstat;
186
187 /*
188 * Flags for pagelock cache support.
189 * Flags argument is passed as uint_t to pcache routines. upper 16 bits of
190 * the flags argument are reserved for alignment page shift when SEGP_PSHIFT
191 * is set.
192 */
193 #define SEGP_FORCE_WIRED 0x1 /* skip check against seg_pwindow */
194 #define SEGP_AMP 0x2 /* anon map's pcache entry */
195 #define SEGP_PSHIFT 0x4 /* addr pgsz shift for hash function */
196
197 /*
198 * Return values for seg_pinsert and seg_pinsert_check functions.
199 */
200 #define SEGP_SUCCESS 0 /* seg_pinsert() succeeded */
201 #define SEGP_FAIL 1 /* seg_pinsert() failed */
202
203 /* Page status bits for segop_incore */
204 #define SEG_PAGE_INCORE 0x01 /* VA has a page backing it */
205 #define SEG_PAGE_LOCKED 0x02 /* VA has a page that is locked */
206 #define SEG_PAGE_HASCOW 0x04 /* VA has a page with a copy-on-write */
207 #define SEG_PAGE_SOFTLOCK 0x08 /* VA has a page with softlock held */
208 #define SEG_PAGE_VNODEBACKED 0x10 /* Segment is backed by a vnode */
209 #define SEG_PAGE_ANON 0x20 /* VA has an anonymous page */
210 #define SEG_PAGE_VNODE 0x40 /* VA has a vnode page backing it */
211
212 #define SEGOP_DUP(s, n) (*(s)->s_ops->dup)((s), (n))
213 #define SEGOP_UNMAP(s, a, l) (*(s)->s_ops->unmap)((s), (a), (l))
214 #define SEGOP_FREE(s) (*(s)->s_ops->free)((s))
215 #define SEGOP_FAULT(h, s, a, l, t, rw) \
216 (*(s)->s_ops->fault)((h), (s), (a), (l), (t), (rw))
217 #define SEGOP_FAULTA(s, a) (*(s)->s_ops->faulta)((s), (a))
218 #define SEGOP_SETPROT(s, a, l, p) (*(s)->s_ops->setprot)((s), (a), (l), (p))
219 #define SEGOP_CHECKPROT(s, a, l, p) (*(s)->s_ops->checkprot)((s), (a), (l), (p))
220 #define SEGOP_KLUSTER(s, a, d) (*(s)->s_ops->kluster)((s), (a), (d))
221 #define SEGOP_SYNC(s, a, l, atr, f) \
222 (*(s)->s_ops->sync)((s), (a), (l), (atr), (f))
223 #define SEGOP_INCORE(s, a, l, v) (*(s)->s_ops->incore)((s), (a), (l), (v))
224 #define SEGOP_LOCKOP(s, a, l, atr, op, b, p) \
225 (*(s)->s_ops->lockop)((s), (a), (l), (atr), (op), (b), (p))
226 #define SEGOP_GETPROT(s, a, l, p) (*(s)->s_ops->getprot)((s), (a), (l), (p))
227 #define SEGOP_GETOFFSET(s, a) (*(s)->s_ops->getoffset)((s), (a))
228 #define SEGOP_GETTYPE(s, a) (*(s)->s_ops->gettype)((s), (a))
229 #define SEGOP_GETVP(s, a, vpp) (*(s)->s_ops->getvp)((s), (a), (vpp))
230 #define SEGOP_ADVISE(s, a, l, b) (*(s)->s_ops->advise)((s), (a), (l), (b))
231 #define SEGOP_DUMP(s) (*(s)->s_ops->dump)((s))
232 #define SEGOP_PAGELOCK(s, a, l, p, t, rw) \
233 (*(s)->s_ops->pagelock)((s), (a), (l), (p), (t), (rw))
234 #define SEGOP_SETPAGESIZE(s, a, l, szc) \
235 (*(s)->s_ops->setpagesize)((s), (a), (l), (szc))
236 #define SEGOP_GETMEMID(s, a, mp) (*(s)->s_ops->getmemid)((s), (a), (mp))
237 #define SEGOP_GETPOLICY(s, a) (*(s)->s_ops->getpolicy)((s), (a))
238 #define SEGOP_CAPABLE(s, c) (*(s)->s_ops->capable)((s), (c))
239 #define SEGOP_INHERIT(s, a, l, b) (*(s)->s_ops->inherit)((s), (a), (l), (b))
240
241 #define seg_page(seg, addr) \
242 (((uintptr_t)((addr) - (seg)->s_base)) >> PAGESHIFT)
243
244 #define seg_pages(seg) \
245 (((uintptr_t)((seg)->s_size + PAGEOFFSET)) >> PAGESHIFT)
246
247 #define IE_NOMEM -1 /* internal to seg layer */
248 #define IE_RETRY -2 /* internal to seg layer */
249 #define IE_REATTACH -3 /* internal to seg layer */
250
251 /* Values for SEGOP_INHERIT */
252 #define SEGP_INH_ZERO 0x01
253
254 int seg_inherit_notsup(struct seg *, caddr_t, size_t, uint_t);
255
256 /* Delay/retry factors for seg_p_mem_config_pre_del */
257 #define SEGP_PREDEL_DELAY_FACTOR 4
258 /*
259 * As a workaround to being unable to purge the pagelock
260 * cache during a DR delete memory operation, we use
261 * a stall threshold that is twice the maximum seen
262 * during testing. This workaround will be removed
263 * when a suitable fix is found.
264 */
265 #define SEGP_STALL_SECONDS 25
266 #define SEGP_STALL_THRESHOLD \
267 (SEGP_STALL_SECONDS * SEGP_PREDEL_DELAY_FACTOR)
268
269 #ifdef VMDEBUG
270
271 uint_t seg_page(struct seg *, caddr_t);
272 uint_t seg_pages(struct seg *);
273
274 #endif /* VMDEBUG */
275
276 boolean_t seg_can_change_zones(struct seg *);
277 size_t seg_swresv(struct seg *);
278
279 #endif /* _KERNEL */
280
281 #ifdef __cplusplus
282 }
283 #endif
284
285 #endif /* _VM_SEG_H */