1 /*- 2 * SPDX-License-Identifier: (BSD-3-Clause AND MIT-CMU) 3 * 4 * Copyright (c) 1991, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * This code is derived from software contributed to Berkeley by 8 * The Mach Operating System project at Carnegie-Mellon University. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 * from: @(#)vm_page.h 8.2 (Berkeley) 12/13/93 35 * 36 * 37 * Copyright (c) 1987, 1990 Carnegie-Mellon University. 38 * All rights reserved. 39 * 40 * Authors: Avadis Tevanian, Jr., Michael Wayne Young 41 * 42 * Permission to use, copy, modify and distribute this software and 43 * its documentation is hereby granted, provided that both the copyright 44 * notice and this permission notice appear in all copies of the 45 * software, derivative works or modified versions, and any portions 46 * thereof, and that both notices appear in supporting documentation. 47 * 48 * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS" 49 * CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND 50 * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE. 51 * 52 * Carnegie Mellon requests users of this software to return to 53 * 54 * Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU 55 * School of Computer Science 56 * Carnegie Mellon University 57 * Pittsburgh PA 15213-3890 58 * 59 * any improvements or extensions that they make and grant Carnegie the 60 * rights to redistribute these changes. 61 * 62 * $FreeBSD$ 63 */ 64 65 #ifndef _VM_PAGEQUEUE_ 66 #define _VM_PAGEQUEUE_ 67 68 #ifdef _KERNEL 69 struct vm_pagequeue { 70 struct mtx pq_mutex; 71 struct pglist pq_pl; 72 int pq_cnt; 73 const char * const pq_name; 74 uint64_t pq_pdpages; 75 } __aligned(CACHE_LINE_SIZE); 76 77 #ifndef VM_BATCHQUEUE_SIZE 78 #define VM_BATCHQUEUE_SIZE 7 79 #endif 80 81 struct vm_batchqueue { 82 vm_page_t bq_pa[VM_BATCHQUEUE_SIZE]; 83 int bq_cnt; 84 } __aligned(CACHE_LINE_SIZE); 85 86 #include <vm/uma.h> 87 #include <sys/pidctrl.h> 88 struct sysctl_oid; 89 90 /* 91 * One vm_domain per-numa domain. Contains pagequeues, free page structures, 92 * and accounting. 93 * 94 * Lock Key: 95 * f vmd_free_mtx 96 * p vmd_pageout_mtx 97 * d vm_domainset_lock 98 * a atomic 99 * c const after boot 100 * q page queue lock 101 */ 102 struct vm_domain { 103 struct vm_pagequeue vmd_pagequeues[PQ_COUNT]; 104 struct mtx_padalign vmd_free_mtx; 105 struct mtx_padalign vmd_pageout_mtx; 106 uma_zone_t vmd_pgcache; /* (c) page free cache. */ 107 struct vmem *vmd_kernel_arena; /* (c) per-domain kva R/W arena. */ 108 struct vmem *vmd_kernel_rwx_arena; /* (c) per-domain kva R/W/X arena. */ 109 u_int vmd_domain; /* (c) Domain number. */ 110 u_int vmd_page_count; /* (c) Total page count. */ 111 long vmd_segs; /* (c) bitmask of the segments */ 112 u_int __aligned(CACHE_LINE_SIZE) vmd_free_count; /* (a,f) free page count */ 113 u_int vmd_pageout_deficit; /* (a) Estimated number of pages deficit */ 114 uint8_t vmd_pad[CACHE_LINE_SIZE - (sizeof(u_int) * 2)]; 115 116 /* Paging control variables, used within single threaded page daemon. */ 117 struct pidctrl vmd_pid; /* Pageout controller. */ 118 boolean_t vmd_oom; 119 int vmd_oom_seq; 120 int vmd_last_active_scan; 121 struct vm_page vmd_markers[PQ_COUNT]; /* (q) markers for queue scans */ 122 struct vm_page vmd_inacthead; /* marker for LRU-defeating insertions */ 123 struct vm_page vmd_clock[2]; /* markers for active queue scan */ 124 125 int vmd_pageout_wanted; /* (a, p) pageout daemon wait channel */ 126 int vmd_pageout_pages_needed; /* (d) page daemon waiting for pages? */ 127 bool vmd_minset; /* (d) Are we in vm_min_domains? */ 128 bool vmd_severeset; /* (d) Are we in vm_severe_domains? */ 129 enum { 130 VM_LAUNDRY_IDLE = 0, 131 VM_LAUNDRY_BACKGROUND, 132 VM_LAUNDRY_SHORTFALL 133 } vmd_laundry_request; 134 135 /* Paging thresholds and targets. */ 136 u_int vmd_clean_pages_freed; /* (q) accumulator for laundry thread */ 137 u_int vmd_background_launder_target; /* (c) */ 138 u_int vmd_free_reserved; /* (c) pages reserved for deadlock */ 139 u_int vmd_free_target; /* (c) pages desired free */ 140 u_int vmd_free_min; /* (c) pages desired free */ 141 u_int vmd_inactive_target; /* (c) pages desired inactive */ 142 u_int vmd_pageout_free_min; /* (c) min pages reserved for kernel */ 143 u_int vmd_pageout_wakeup_thresh;/* (c) min pages to wake pagedaemon */ 144 u_int vmd_interrupt_free_min; /* (c) reserved pages for int code */ 145 u_int vmd_free_severe; /* (c) severe page depletion point */ 146 147 /* Name for sysctl etc. */ 148 struct sysctl_oid *vmd_oid; 149 char vmd_name[sizeof(__XSTRING(MAXMEMDOM))]; 150 } __aligned(CACHE_LINE_SIZE); 151 152 extern struct vm_domain vm_dom[MAXMEMDOM]; 153 154 #define VM_DOMAIN(n) (&vm_dom[(n)]) 155 #define VM_DOMAIN_EMPTY(n) (vm_dom[(n)].vmd_page_count == 0) 156 157 #define vm_pagequeue_assert_locked(pq) mtx_assert(&(pq)->pq_mutex, MA_OWNED) 158 #define vm_pagequeue_lock(pq) mtx_lock(&(pq)->pq_mutex) 159 #define vm_pagequeue_lockptr(pq) (&(pq)->pq_mutex) 160 #define vm_pagequeue_trylock(pq) mtx_trylock(&(pq)->pq_mutex) 161 #define vm_pagequeue_unlock(pq) mtx_unlock(&(pq)->pq_mutex) 162 163 #define vm_domain_free_assert_locked(n) \ 164 mtx_assert(vm_domain_free_lockptr((n)), MA_OWNED) 165 #define vm_domain_free_assert_unlocked(n) \ 166 mtx_assert(vm_domain_free_lockptr((n)), MA_NOTOWNED) 167 #define vm_domain_free_lock(d) \ 168 mtx_lock(vm_domain_free_lockptr((d))) 169 #define vm_domain_free_lockptr(d) \ 170 (&(d)->vmd_free_mtx) 171 #define vm_domain_free_trylock(d) \ 172 mtx_trylock(vm_domain_free_lockptr((d))) 173 #define vm_domain_free_unlock(d) \ 174 mtx_unlock(vm_domain_free_lockptr((d))) 175 176 #define vm_domain_pageout_lockptr(d) \ 177 (&(d)->vmd_pageout_mtx) 178 #define vm_domain_pageout_assert_locked(n) \ 179 mtx_assert(vm_domain_pageout_lockptr((n)), MA_OWNED) 180 #define vm_domain_pageout_assert_unlocked(n) \ 181 mtx_assert(vm_domain_pageout_lockptr((n)), MA_NOTOWNED) 182 #define vm_domain_pageout_lock(d) \ 183 mtx_lock(vm_domain_pageout_lockptr((d))) 184 #define vm_domain_pageout_unlock(d) \ 185 mtx_unlock(vm_domain_pageout_lockptr((d))) 186 187 static __inline void 188 vm_pagequeue_cnt_add(struct vm_pagequeue *pq, int addend) 189 { 190 191 vm_pagequeue_assert_locked(pq); 192 pq->pq_cnt += addend; 193 } 194 #define vm_pagequeue_cnt_inc(pq) vm_pagequeue_cnt_add((pq), 1) 195 #define vm_pagequeue_cnt_dec(pq) vm_pagequeue_cnt_add((pq), -1) 196 197 static inline void 198 vm_batchqueue_init(struct vm_batchqueue *bq) 199 { 200 201 bq->bq_cnt = 0; 202 } 203 204 static inline bool 205 vm_batchqueue_insert(struct vm_batchqueue *bq, vm_page_t m) 206 { 207 208 if (bq->bq_cnt < nitems(bq->bq_pa)) { 209 bq->bq_pa[bq->bq_cnt++] = m; 210 return (true); 211 } 212 return (false); 213 } 214 215 static inline vm_page_t 216 vm_batchqueue_pop(struct vm_batchqueue *bq) 217 { 218 219 if (bq->bq_cnt == 0) 220 return (NULL); 221 return (bq->bq_pa[--bq->bq_cnt]); 222 } 223 224 void vm_domain_set(struct vm_domain *vmd); 225 void vm_domain_clear(struct vm_domain *vmd); 226 int vm_domain_allocate(struct vm_domain *vmd, int req, int npages); 227 228 /* 229 * vm_pagequeue_domain: 230 * 231 * Return the memory domain the page belongs to. 232 */ 233 static inline struct vm_domain * 234 vm_pagequeue_domain(vm_page_t m) 235 { 236 237 return (VM_DOMAIN(vm_phys_domain(m))); 238 } 239 240 /* 241 * Return the number of pages we need to free-up or cache 242 * A positive number indicates that we do not have enough free pages. 243 */ 244 static inline int 245 vm_paging_target(struct vm_domain *vmd) 246 { 247 248 return (vmd->vmd_free_target - vmd->vmd_free_count); 249 } 250 251 /* 252 * Returns TRUE if the pagedaemon needs to be woken up. 253 */ 254 static inline int 255 vm_paging_needed(struct vm_domain *vmd, u_int free_count) 256 { 257 258 return (free_count < vmd->vmd_pageout_wakeup_thresh); 259 } 260 261 /* 262 * Returns TRUE if the domain is below the min paging target. 263 */ 264 static inline int 265 vm_paging_min(struct vm_domain *vmd) 266 { 267 268 return (vmd->vmd_free_min > vmd->vmd_free_count); 269 } 270 271 /* 272 * Returns TRUE if the domain is below the severe paging target. 273 */ 274 static inline int 275 vm_paging_severe(struct vm_domain *vmd) 276 { 277 278 return (vmd->vmd_free_severe > vmd->vmd_free_count); 279 } 280 281 /* 282 * Return the number of pages we need to launder. 283 * A positive number indicates that we have a shortfall of clean pages. 284 */ 285 static inline int 286 vm_laundry_target(struct vm_domain *vmd) 287 { 288 289 return (vm_paging_target(vmd)); 290 } 291 292 void pagedaemon_wakeup(int domain); 293 294 static inline void 295 vm_domain_freecnt_inc(struct vm_domain *vmd, int adj) 296 { 297 u_int old, new; 298 299 old = atomic_fetchadd_int(&vmd->vmd_free_count, adj); 300 new = old + adj; 301 /* 302 * Only update bitsets on transitions. Notice we short-circuit the 303 * rest of the checks if we're above min already. 304 */ 305 if (old < vmd->vmd_free_min && (new >= vmd->vmd_free_min || 306 (old < vmd->vmd_free_severe && new >= vmd->vmd_free_severe) || 307 (old < vmd->vmd_pageout_free_min && 308 new >= vmd->vmd_pageout_free_min))) 309 vm_domain_clear(vmd); 310 } 311 312 #endif /* _KERNEL */ 313 #endif /* !_VM_PAGEQUEUE_ */ 314