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