vm_meter.c revision 159475
1/*- 2 * Copyright (c) 1982, 1986, 1989, 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 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)vm_meter.c 8.4 (Berkeley) 1/4/94 30 */ 31 32#include <sys/cdefs.h> 33__FBSDID("$FreeBSD: head/sys/vm/vm_meter.c 159475 2006-06-10 05:29:37Z alc $"); 34 35#include <sys/param.h> 36#include <sys/systm.h> 37#include <sys/kernel.h> 38#include <sys/lock.h> 39#include <sys/mutex.h> 40#include <sys/proc.h> 41#include <sys/resource.h> 42#include <sys/sx.h> 43#include <sys/vmmeter.h> 44#include <sys/smp.h> 45 46#include <vm/vm.h> 47#include <vm/vm_page.h> 48#include <vm/vm_extern.h> 49#include <vm/vm_param.h> 50#include <vm/pmap.h> 51#include <vm/vm_map.h> 52#include <vm/vm_object.h> 53#include <sys/sysctl.h> 54 55/* 56 * Virtual memory MPSAFE temporary workarounds. 57 */ 58#if !defined(__powerpc__) 59int debug_mpsafevm = 1; 60#else 61int debug_mpsafevm; 62#endif 63TUNABLE_INT("debug.mpsafevm", &debug_mpsafevm); 64SYSCTL_INT(_debug, OID_AUTO, mpsafevm, CTLFLAG_RD, &debug_mpsafevm, 0, 65 "Enable/disable MPSAFE virtual memory support"); 66 67struct vmmeter cnt; 68 69int maxslp = MAXSLP; 70 71SYSCTL_UINT(_vm, VM_V_FREE_MIN, v_free_min, 72 CTLFLAG_RW, &cnt.v_free_min, 0, ""); 73SYSCTL_UINT(_vm, VM_V_FREE_TARGET, v_free_target, 74 CTLFLAG_RW, &cnt.v_free_target, 0, ""); 75SYSCTL_UINT(_vm, VM_V_FREE_RESERVED, v_free_reserved, 76 CTLFLAG_RW, &cnt.v_free_reserved, 0, ""); 77SYSCTL_UINT(_vm, VM_V_INACTIVE_TARGET, v_inactive_target, 78 CTLFLAG_RW, &cnt.v_inactive_target, 0, ""); 79SYSCTL_UINT(_vm, VM_V_CACHE_MIN, v_cache_min, 80 CTLFLAG_RW, &cnt.v_cache_min, 0, ""); 81SYSCTL_UINT(_vm, VM_V_CACHE_MAX, v_cache_max, 82 CTLFLAG_RW, &cnt.v_cache_max, 0, ""); 83SYSCTL_UINT(_vm, VM_V_PAGEOUT_FREE_MIN, v_pageout_free_min, 84 CTLFLAG_RW, &cnt.v_pageout_free_min, 0, ""); 85SYSCTL_UINT(_vm, OID_AUTO, v_free_severe, 86 CTLFLAG_RW, &cnt.v_free_severe, 0, ""); 87 88static int 89sysctl_vm_loadavg(SYSCTL_HANDLER_ARGS) 90{ 91#ifdef SCTL_MASK32 92 u_int32_t la[4]; 93 94 if (req->flags & SCTL_MASK32) { 95 la[0] = averunnable.ldavg[0]; 96 la[1] = averunnable.ldavg[1]; 97 la[2] = averunnable.ldavg[2]; 98 la[3] = averunnable.fscale; 99 return SYSCTL_OUT(req, la, sizeof(la)); 100 } else 101#endif 102 return SYSCTL_OUT(req, &averunnable, sizeof(averunnable)); 103} 104SYSCTL_PROC(_vm, VM_LOADAVG, loadavg, CTLTYPE_STRUCT|CTLFLAG_RD, 105 NULL, 0, sysctl_vm_loadavg, "S,loadavg", "Machine loadaverage history"); 106 107static int 108vmtotal(SYSCTL_HANDLER_ARGS) 109{ 110/* XXXKSE almost completely broken */ 111 struct proc *p; 112 struct vmtotal total, *totalp; 113 vm_map_entry_t entry; 114 vm_object_t object; 115 vm_map_t map; 116 int paging; 117 struct thread *td; 118 struct vmspace *vm; 119 120 totalp = &total; 121 bzero(totalp, sizeof *totalp); 122 /* 123 * Mark all objects as inactive. 124 */ 125 GIANT_REQUIRED; 126 mtx_lock(&vm_object_list_mtx); 127 TAILQ_FOREACH(object, &vm_object_list, object_list) { 128 if (!VM_OBJECT_TRYLOCK(object)) { 129 /* 130 * Avoid a lock-order reversal. Consequently, 131 * the reported number of active pages may be 132 * greater than the actual number. 133 */ 134 continue; 135 } 136 vm_object_clear_flag(object, OBJ_ACTIVE); 137 VM_OBJECT_UNLOCK(object); 138 } 139 mtx_unlock(&vm_object_list_mtx); 140 /* 141 * Calculate process statistics. 142 */ 143 sx_slock(&allproc_lock); 144 FOREACH_PROC_IN_SYSTEM(p) { 145 if (p->p_flag & P_SYSTEM) 146 continue; 147 mtx_lock_spin(&sched_lock); 148 switch (p->p_state) { 149 case PRS_NEW: 150 mtx_unlock_spin(&sched_lock); 151 continue; 152 break; 153 default: 154 FOREACH_THREAD_IN_PROC(p, td) { 155 /* Need new statistics XXX */ 156 switch (td->td_state) { 157 case TDS_INHIBITED: 158 if (TD_ON_LOCK(td) || 159 (td->td_inhibitors == 160 TDI_SWAPPED)) { 161 totalp->t_sw++; 162 } else if (TD_IS_SLEEPING(td) || 163 TD_AWAITING_INTR(td) || 164 TD_IS_SUSPENDED(td)) { 165 if (td->td_priority <= PZERO) 166 totalp->t_dw++; 167 else 168 totalp->t_sl++; 169 } 170 break; 171 172 case TDS_CAN_RUN: 173 totalp->t_sw++; 174 break; 175 case TDS_RUNQ: 176 case TDS_RUNNING: 177 totalp->t_rq++; 178 continue; 179 default: 180 break; 181 } 182 } 183 } 184 mtx_unlock_spin(&sched_lock); 185 /* 186 * Note active objects. 187 */ 188 paging = 0; 189 vm = vmspace_acquire_ref(p); 190 if (vm == NULL) 191 continue; 192 map = &vm->vm_map; 193 vm_map_lock_read(map); 194 for (entry = map->header.next; 195 entry != &map->header; entry = entry->next) { 196 if ((entry->eflags & MAP_ENTRY_IS_SUB_MAP) || 197 (object = entry->object.vm_object) == NULL) 198 continue; 199 VM_OBJECT_LOCK(object); 200 vm_object_set_flag(object, OBJ_ACTIVE); 201 paging |= object->paging_in_progress; 202 VM_OBJECT_UNLOCK(object); 203 } 204 vm_map_unlock_read(map); 205 vmspace_free(vm); 206 if (paging) 207 totalp->t_pw++; 208 } 209 sx_sunlock(&allproc_lock); 210 /* 211 * Calculate object memory usage statistics. 212 */ 213 mtx_lock(&vm_object_list_mtx); 214 TAILQ_FOREACH(object, &vm_object_list, object_list) { 215 /* 216 * Perform unsynchronized reads on the object to avoid 217 * a lock-order reversal. In this case, the lack of 218 * synchronization should not impair the accuracy of 219 * the reported statistics. 220 */ 221 if (object->type == OBJT_DEVICE) { 222 /* 223 * Devices, like /dev/mem, will badly skew our totals. 224 */ 225 continue; 226 } 227 totalp->t_vm += object->size; 228 totalp->t_rm += object->resident_page_count; 229 if (object->flags & OBJ_ACTIVE) { 230 totalp->t_avm += object->size; 231 totalp->t_arm += object->resident_page_count; 232 } 233 if (object->shadow_count > 1) { 234 /* shared object */ 235 totalp->t_vmshr += object->size; 236 totalp->t_rmshr += object->resident_page_count; 237 if (object->flags & OBJ_ACTIVE) { 238 totalp->t_avmshr += object->size; 239 totalp->t_armshr += object->resident_page_count; 240 } 241 } 242 } 243 mtx_unlock(&vm_object_list_mtx); 244 totalp->t_free = cnt.v_free_count + cnt.v_cache_count; 245 return (sysctl_handle_opaque(oidp, totalp, sizeof total, req)); 246} 247 248/* 249 * vcnt() - accumulate statistics from all cpus and the global cnt 250 * structure. 251 * 252 * The vmmeter structure is now per-cpu as well as global. Those 253 * statistics which can be kept on a per-cpu basis (to avoid cache 254 * stalls between cpus) can be moved to the per-cpu vmmeter. Remaining 255 * statistics, such as v_free_reserved, are left in the global 256 * structure. 257 * 258 * (sysctl_oid *oidp, void *arg1, int arg2, struct sysctl_req *req) 259 */ 260static int 261vcnt(SYSCTL_HANDLER_ARGS) 262{ 263 int count = *(int *)arg1; 264 int offset = (char *)arg1 - (char *)&cnt; 265#ifdef SMP 266 int i; 267 268 for (i = 0; i < mp_ncpus; ++i) { 269 struct pcpu *pcpu = pcpu_find(i); 270 count += *(int *)((char *)&pcpu->pc_cnt + offset); 271 } 272#else 273 count += *(int *)((char *)PCPU_PTR(cnt) + offset); 274#endif 275 return (SYSCTL_OUT(req, &count, sizeof(int))); 276} 277 278SYSCTL_PROC(_vm, VM_TOTAL, vmtotal, CTLTYPE_OPAQUE|CTLFLAG_RD, 279 0, sizeof(struct vmtotal), vmtotal, "S,vmtotal", 280 "System virtual memory statistics"); 281SYSCTL_NODE(_vm, OID_AUTO, stats, CTLFLAG_RW, 0, "VM meter stats"); 282static SYSCTL_NODE(_vm_stats, OID_AUTO, sys, CTLFLAG_RW, 0, 283 "VM meter sys stats"); 284static SYSCTL_NODE(_vm_stats, OID_AUTO, vm, CTLFLAG_RW, 0, 285 "VM meter vm stats"); 286SYSCTL_NODE(_vm_stats, OID_AUTO, misc, CTLFLAG_RW, 0, "VM meter misc stats"); 287 288SYSCTL_PROC(_vm_stats_sys, OID_AUTO, v_swtch, CTLTYPE_UINT|CTLFLAG_RD, 289 &cnt.v_swtch, 0, vcnt, "IU", "Context switches"); 290SYSCTL_PROC(_vm_stats_sys, OID_AUTO, v_trap, CTLTYPE_UINT|CTLFLAG_RD, 291 &cnt.v_trap, 0, vcnt, "IU", "Traps"); 292SYSCTL_PROC(_vm_stats_sys, OID_AUTO, v_syscall, CTLTYPE_UINT|CTLFLAG_RD, 293 &cnt.v_syscall, 0, vcnt, "IU", "Syscalls"); 294SYSCTL_PROC(_vm_stats_sys, OID_AUTO, v_intr, CTLTYPE_UINT|CTLFLAG_RD, 295 &cnt.v_intr, 0, vcnt, "IU", "Hardware interrupts"); 296SYSCTL_PROC(_vm_stats_sys, OID_AUTO, v_soft, CTLTYPE_UINT|CTLFLAG_RD, 297 &cnt.v_soft, 0, vcnt, "IU", "Software interrupts"); 298SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_vm_faults, CTLTYPE_UINT|CTLFLAG_RD, 299 &cnt.v_vm_faults, 0, vcnt, "IU", "VM faults"); 300SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_cow_faults, CTLTYPE_UINT|CTLFLAG_RD, 301 &cnt.v_cow_faults, 0, vcnt, "IU", "COW faults"); 302SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_cow_optim, CTLTYPE_UINT|CTLFLAG_RD, 303 &cnt.v_cow_optim, 0, vcnt, "IU", "Optimized COW faults"); 304SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_zfod, CTLTYPE_UINT|CTLFLAG_RD, 305 &cnt.v_zfod, 0, vcnt, "IU", "Zero fill"); 306SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_ozfod, CTLTYPE_UINT|CTLFLAG_RD, 307 &cnt.v_ozfod, 0, vcnt, "IU", "Optimized zero fill"); 308SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_swapin, CTLTYPE_UINT|CTLFLAG_RD, 309 &cnt.v_swapin, 0, vcnt, "IU", "Swapin operations"); 310SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_swapout, CTLTYPE_UINT|CTLFLAG_RD, 311 &cnt.v_swapout, 0, vcnt, "IU", "Swapout operations"); 312SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_swappgsin, CTLTYPE_UINT|CTLFLAG_RD, 313 &cnt.v_swappgsin, 0, vcnt, "IU", "Swapin pages"); 314SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_swappgsout, CTLTYPE_UINT|CTLFLAG_RD, 315 &cnt.v_swappgsout, 0, vcnt, "IU", "Swapout pages"); 316SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_vnodein, CTLTYPE_UINT|CTLFLAG_RD, 317 &cnt.v_vnodein, 0, vcnt, "IU", "Vnodein operations"); 318SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_vnodeout, CTLTYPE_UINT|CTLFLAG_RD, 319 &cnt.v_vnodeout, 0, vcnt, "IU", "Vnodeout operations"); 320SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_vnodepgsin, CTLTYPE_UINT|CTLFLAG_RD, 321 &cnt.v_vnodepgsin, 0, vcnt, "IU", "Vnodein pages"); 322SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_vnodepgsout, CTLTYPE_UINT|CTLFLAG_RD, 323 &cnt.v_vnodepgsout, 0, vcnt, "IU", "Vnodeout pages"); 324SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_intrans, CTLTYPE_UINT|CTLFLAG_RD, 325 &cnt.v_intrans, 0, vcnt, "IU", "In transit page blocking"); 326SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_reactivated, CTLTYPE_UINT|CTLFLAG_RD, 327 &cnt.v_reactivated, 0, vcnt, "IU", "Reactivated pages"); 328SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_pdwakeups, CTLTYPE_UINT|CTLFLAG_RD, 329 &cnt.v_pdwakeups, 0, vcnt, "IU", "Pagedaemon wakeups"); 330SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_pdpages, CTLTYPE_UINT|CTLFLAG_RD, 331 &cnt.v_pdpages, 0, vcnt, "IU", "Pagedaemon page scans"); 332SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_dfree, CTLTYPE_UINT|CTLFLAG_RD, 333 &cnt.v_dfree, 0, vcnt, "IU", ""); 334SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_pfree, CTLTYPE_UINT|CTLFLAG_RD, 335 &cnt.v_pfree, 0, vcnt, "IU", ""); 336SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_tfree, CTLTYPE_UINT|CTLFLAG_RD, 337 &cnt.v_tfree, 0, vcnt, "IU", ""); 338SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_page_size, CTLTYPE_UINT|CTLFLAG_RD, 339 &cnt.v_page_size, 0, vcnt, "IU", ""); 340SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_page_count, CTLTYPE_UINT|CTLFLAG_RD, 341 &cnt.v_page_count, 0, vcnt, "IU", ""); 342SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_free_reserved, CTLTYPE_UINT|CTLFLAG_RD, 343 &cnt.v_free_reserved, 0, vcnt, "IU", ""); 344SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_free_target, CTLTYPE_UINT|CTLFLAG_RD, 345 &cnt.v_free_target, 0, vcnt, "IU", ""); 346SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_free_min, CTLTYPE_UINT|CTLFLAG_RD, 347 &cnt.v_free_min, 0, vcnt, "IU", ""); 348SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_free_count, CTLTYPE_UINT|CTLFLAG_RD, 349 &cnt.v_free_count, 0, vcnt, "IU", ""); 350SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_wire_count, CTLTYPE_UINT|CTLFLAG_RD, 351 &cnt.v_wire_count, 0, vcnt, "IU", ""); 352SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_active_count, CTLTYPE_UINT|CTLFLAG_RD, 353 &cnt.v_active_count, 0, vcnt, "IU", ""); 354SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_inactive_target, CTLTYPE_UINT|CTLFLAG_RD, 355 &cnt.v_inactive_target, 0, vcnt, "IU", ""); 356SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_inactive_count, CTLTYPE_UINT|CTLFLAG_RD, 357 &cnt.v_inactive_count, 0, vcnt, "IU", ""); 358SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_cache_count, CTLTYPE_UINT|CTLFLAG_RD, 359 &cnt.v_cache_count, 0, vcnt, "IU", ""); 360SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_cache_min, CTLTYPE_UINT|CTLFLAG_RD, 361 &cnt.v_cache_min, 0, vcnt, "IU", ""); 362SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_cache_max, CTLTYPE_UINT|CTLFLAG_RD, 363 &cnt.v_cache_max, 0, vcnt, "IU", ""); 364SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_pageout_free_min, CTLTYPE_UINT|CTLFLAG_RD, 365 &cnt.v_pageout_free_min, 0, vcnt, "IU", ""); 366SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_interrupt_free_min, CTLTYPE_UINT|CTLFLAG_RD, 367 &cnt.v_interrupt_free_min, 0, vcnt, "IU", ""); 368SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_forks, CTLTYPE_UINT|CTLFLAG_RD, 369 &cnt.v_forks, 0, vcnt, "IU", "Number of fork() calls"); 370SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_vforks, CTLTYPE_UINT|CTLFLAG_RD, 371 &cnt.v_vforks, 0, vcnt, "IU", "Number of vfork() calls"); 372SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_rforks, CTLTYPE_UINT|CTLFLAG_RD, 373 &cnt.v_rforks, 0, vcnt, "IU", "Number of rfork() calls"); 374SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_kthreads, CTLTYPE_UINT|CTLFLAG_RD, 375 &cnt.v_kthreads, 0, vcnt, "IU", "Number of fork() calls by kernel"); 376SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_forkpages, CTLTYPE_UINT|CTLFLAG_RD, 377 &cnt.v_forkpages, 0, vcnt, "IU", "VM pages affected by fork()"); 378SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_vforkpages, CTLTYPE_UINT|CTLFLAG_RD, 379 &cnt.v_vforkpages, 0, vcnt, "IU", "VM pages affected by vfork()"); 380SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_rforkpages, CTLTYPE_UINT|CTLFLAG_RD, 381 &cnt.v_rforkpages, 0, vcnt, "IU", "VM pages affected by rfork()"); 382SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_kthreadpages, CTLTYPE_UINT|CTLFLAG_RD, 383 &cnt.v_kthreadpages, 0, vcnt, "IU", "VM pages affected by fork() by kernel"); 384 385SYSCTL_INT(_vm_stats_misc, OID_AUTO, 386 zero_page_count, CTLFLAG_RD, &vm_page_zero_count, 0, ""); 387#if 0 388SYSCTL_INT(_vm_stats_misc, OID_AUTO, 389 page_mask, CTLFLAG_RD, &page_mask, 0, ""); 390SYSCTL_INT(_vm_stats_misc, OID_AUTO, 391 page_shift, CTLFLAG_RD, &page_shift, 0, ""); 392SYSCTL_INT(_vm_stats_misc, OID_AUTO, 393 first_page, CTLFLAG_RD, &first_page, 0, ""); 394SYSCTL_INT(_vm_stats_misc, OID_AUTO, 395 last_page, CTLFLAG_RD, &last_page, 0, ""); 396#endif 397