vm_meter.c revision 302243
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: stable/10/sys/vm/vm_meter.c 302243 2016-06-28 00:37:34Z kib $"); 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/rwlock.h> 43#include <sys/sx.h> 44#include <sys/vmmeter.h> 45#include <sys/smp.h> 46 47#include <vm/vm.h> 48#include <vm/vm_page.h> 49#include <vm/vm_extern.h> 50#include <vm/vm_param.h> 51#include <vm/pmap.h> 52#include <vm/vm_map.h> 53#include <vm/vm_object.h> 54#include <sys/sysctl.h> 55 56struct vmmeter cnt; 57 58SYSCTL_UINT(_vm, VM_V_FREE_MIN, v_free_min, 59 CTLFLAG_RW, &cnt.v_free_min, 0, "Minimum low-free-pages threshold"); 60SYSCTL_UINT(_vm, VM_V_FREE_TARGET, v_free_target, 61 CTLFLAG_RW, &cnt.v_free_target, 0, "Desired free pages"); 62SYSCTL_UINT(_vm, VM_V_FREE_RESERVED, v_free_reserved, 63 CTLFLAG_RW, &cnt.v_free_reserved, 0, "Pages reserved for deadlock"); 64SYSCTL_UINT(_vm, VM_V_INACTIVE_TARGET, v_inactive_target, 65 CTLFLAG_RW, &cnt.v_inactive_target, 0, "Pages desired inactive"); 66SYSCTL_UINT(_vm, VM_V_CACHE_MIN, v_cache_min, 67 CTLFLAG_RW, &cnt.v_cache_min, 0, "Min pages on cache queue"); 68SYSCTL_UINT(_vm, VM_V_CACHE_MAX, v_cache_max, 69 CTLFLAG_RW, &cnt.v_cache_max, 0, "Max pages on cache queue"); 70SYSCTL_UINT(_vm, VM_V_PAGEOUT_FREE_MIN, v_pageout_free_min, 71 CTLFLAG_RW, &cnt.v_pageout_free_min, 0, "Min pages reserved for kernel"); 72SYSCTL_UINT(_vm, OID_AUTO, v_free_severe, 73 CTLFLAG_RW, &cnt.v_free_severe, 0, "Severe page depletion point"); 74 75static int 76sysctl_vm_loadavg(SYSCTL_HANDLER_ARGS) 77{ 78 79#ifdef SCTL_MASK32 80 u_int32_t la[4]; 81 82 if (req->flags & SCTL_MASK32) { 83 la[0] = averunnable.ldavg[0]; 84 la[1] = averunnable.ldavg[1]; 85 la[2] = averunnable.ldavg[2]; 86 la[3] = averunnable.fscale; 87 return SYSCTL_OUT(req, la, sizeof(la)); 88 } else 89#endif 90 return SYSCTL_OUT(req, &averunnable, sizeof(averunnable)); 91} 92SYSCTL_PROC(_vm, VM_LOADAVG, loadavg, CTLTYPE_STRUCT | CTLFLAG_RD | 93 CTLFLAG_MPSAFE, NULL, 0, sysctl_vm_loadavg, "S,loadavg", 94 "Machine loadaverage history"); 95 96/* 97 * This function aims to determine if the object is mapped, 98 * specifically, if it is referenced by a vm_map_entry. Because 99 * objects occasionally acquire transient references that do not 100 * represent a mapping, the method used here is inexact. However, it 101 * has very low overhead and is good enough for the advisory 102 * vm.vmtotal sysctl. 103 */ 104static bool 105is_object_active(vm_object_t obj) 106{ 107 108 return (obj->ref_count > obj->shadow_count); 109} 110 111static int 112vmtotal(SYSCTL_HANDLER_ARGS) 113{ 114 struct vmtotal total; 115 vm_object_t object; 116 struct proc *p; 117 struct thread *td; 118 119 bzero(&total, sizeof(total)); 120 121 /* 122 * Calculate process statistics. 123 */ 124 sx_slock(&allproc_lock); 125 FOREACH_PROC_IN_SYSTEM(p) { 126 if (p->p_flag & P_SYSTEM) 127 continue; 128 PROC_LOCK(p); 129 switch (p->p_state) { 130 case PRS_NEW: 131 PROC_UNLOCK(p); 132 continue; 133 break; 134 default: 135 FOREACH_THREAD_IN_PROC(p, td) { 136 thread_lock(td); 137 switch (td->td_state) { 138 case TDS_INHIBITED: 139 if (TD_IS_SWAPPED(td)) 140 total.t_sw++; 141 else if (TD_IS_SLEEPING(td)) { 142 if (td->td_priority <= PZERO) 143 total.t_dw++; 144 else 145 total.t_sl++; 146 if (td->td_wchan == 147 &cnt.v_free_count) 148 total.t_pw++; 149 } 150 break; 151 152 case TDS_CAN_RUN: 153 total.t_sw++; 154 break; 155 case TDS_RUNQ: 156 case TDS_RUNNING: 157 total.t_rq++; 158 thread_unlock(td); 159 continue; 160 default: 161 break; 162 } 163 thread_unlock(td); 164 } 165 } 166 PROC_UNLOCK(p); 167 } 168 sx_sunlock(&allproc_lock); 169 /* 170 * Calculate object memory usage statistics. 171 */ 172 mtx_lock(&vm_object_list_mtx); 173 TAILQ_FOREACH(object, &vm_object_list, object_list) { 174 /* 175 * Perform unsynchronized reads on the object. In 176 * this case, the lack of synchronization should not 177 * impair the accuracy of the reported statistics. 178 */ 179 if ((object->flags & OBJ_FICTITIOUS) != 0) { 180 /* 181 * Devices, like /dev/mem, will badly skew our totals. 182 */ 183 continue; 184 } 185 if (object->ref_count == 0) { 186 /* 187 * Also skip unreferenced objects, including 188 * vnodes representing mounted file systems. 189 */ 190 continue; 191 } 192 if (object->ref_count == 1 && 193 (object->flags & OBJ_NOSPLIT) != 0) { 194 /* 195 * Also skip otherwise unreferenced swap 196 * objects backing tmpfs vnodes, and POSIX or 197 * SysV shared memory. 198 */ 199 continue; 200 } 201 total.t_vm += object->size; 202 total.t_rm += object->resident_page_count; 203 if (is_object_active(object)) { 204 total.t_avm += object->size; 205 total.t_arm += object->resident_page_count; 206 } 207 if (object->shadow_count > 1) { 208 /* shared object */ 209 total.t_vmshr += object->size; 210 total.t_rmshr += object->resident_page_count; 211 if (is_object_active(object)) { 212 total.t_avmshr += object->size; 213 total.t_armshr += object->resident_page_count; 214 } 215 } 216 } 217 mtx_unlock(&vm_object_list_mtx); 218 total.t_free = cnt.v_free_count + cnt.v_cache_count; 219 return (sysctl_handle_opaque(oidp, &total, sizeof(total), req)); 220} 221 222/* 223 * vcnt() - accumulate statistics from all cpus and the global cnt 224 * structure. 225 * 226 * The vmmeter structure is now per-cpu as well as global. Those 227 * statistics which can be kept on a per-cpu basis (to avoid cache 228 * stalls between cpus) can be moved to the per-cpu vmmeter. Remaining 229 * statistics, such as v_free_reserved, are left in the global 230 * structure. 231 * 232 * (sysctl_oid *oidp, void *arg1, int arg2, struct sysctl_req *req) 233 */ 234static int 235vcnt(SYSCTL_HANDLER_ARGS) 236{ 237 int count = *(int *)arg1; 238 int offset = (char *)arg1 - (char *)&cnt; 239 int i; 240 241 CPU_FOREACH(i) { 242 struct pcpu *pcpu = pcpu_find(i); 243 count += *(int *)((char *)&pcpu->pc_cnt + offset); 244 } 245 return (SYSCTL_OUT(req, &count, sizeof(int))); 246} 247 248SYSCTL_PROC(_vm, VM_TOTAL, vmtotal, CTLTYPE_OPAQUE|CTLFLAG_RD|CTLFLAG_MPSAFE, 249 0, sizeof(struct vmtotal), vmtotal, "S,vmtotal", 250 "System virtual memory statistics"); 251SYSCTL_NODE(_vm, OID_AUTO, stats, CTLFLAG_RW, 0, "VM meter stats"); 252static SYSCTL_NODE(_vm_stats, OID_AUTO, sys, CTLFLAG_RW, 0, 253 "VM meter sys stats"); 254static SYSCTL_NODE(_vm_stats, OID_AUTO, vm, CTLFLAG_RW, 0, 255 "VM meter vm stats"); 256SYSCTL_NODE(_vm_stats, OID_AUTO, misc, CTLFLAG_RW, 0, "VM meter misc stats"); 257 258#define VM_STATS(parent, var, descr) \ 259 SYSCTL_PROC(parent, OID_AUTO, var, \ 260 CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_MPSAFE, &cnt.var, 0, vcnt, \ 261 "IU", descr) 262#define VM_STATS_VM(var, descr) VM_STATS(_vm_stats_vm, var, descr) 263#define VM_STATS_SYS(var, descr) VM_STATS(_vm_stats_sys, var, descr) 264 265VM_STATS_SYS(v_swtch, "Context switches"); 266VM_STATS_SYS(v_trap, "Traps"); 267VM_STATS_SYS(v_syscall, "System calls"); 268VM_STATS_SYS(v_intr, "Device interrupts"); 269VM_STATS_SYS(v_soft, "Software interrupts"); 270VM_STATS_VM(v_vm_faults, "Address memory faults"); 271VM_STATS_VM(v_io_faults, "Page faults requiring I/O"); 272VM_STATS_VM(v_cow_faults, "Copy-on-write faults"); 273VM_STATS_VM(v_cow_optim, "Optimized COW faults"); 274VM_STATS_VM(v_zfod, "Pages zero-filled on demand"); 275VM_STATS_VM(v_ozfod, "Optimized zero fill pages"); 276VM_STATS_VM(v_swapin, "Swap pager pageins"); 277VM_STATS_VM(v_swapout, "Swap pager pageouts"); 278VM_STATS_VM(v_swappgsin, "Swap pages swapped in"); 279VM_STATS_VM(v_swappgsout, "Swap pages swapped out"); 280VM_STATS_VM(v_vnodein, "Vnode pager pageins"); 281VM_STATS_VM(v_vnodeout, "Vnode pager pageouts"); 282VM_STATS_VM(v_vnodepgsin, "Vnode pages paged in"); 283VM_STATS_VM(v_vnodepgsout, "Vnode pages paged out"); 284VM_STATS_VM(v_intrans, "In transit page faults"); 285VM_STATS_VM(v_reactivated, "Pages reactivated from free list"); 286VM_STATS_VM(v_pdwakeups, "Pagedaemon wakeups"); 287VM_STATS_VM(v_pdpages, "Pages analyzed by pagedaemon"); 288VM_STATS_VM(v_tcached, "Total pages cached"); 289VM_STATS_VM(v_dfree, "Pages freed by pagedaemon"); 290VM_STATS_VM(v_pfree, "Pages freed by exiting processes"); 291VM_STATS_VM(v_tfree, "Total pages freed"); 292VM_STATS_VM(v_page_size, "Page size in bytes"); 293VM_STATS_VM(v_page_count, "Total number of pages in system"); 294VM_STATS_VM(v_free_reserved, "Pages reserved for deadlock"); 295VM_STATS_VM(v_free_target, "Pages desired free"); 296VM_STATS_VM(v_free_min, "Minimum low-free-pages threshold"); 297VM_STATS_VM(v_free_count, "Free pages"); 298VM_STATS_VM(v_wire_count, "Wired pages"); 299VM_STATS_VM(v_active_count, "Active pages"); 300VM_STATS_VM(v_inactive_target, "Desired inactive pages"); 301VM_STATS_VM(v_inactive_count, "Inactive pages"); 302VM_STATS_VM(v_cache_count, "Pages on cache queue"); 303VM_STATS_VM(v_cache_min, "Min pages on cache queue"); 304VM_STATS_VM(v_cache_max, "Max pages on cached queue"); 305VM_STATS_VM(v_pageout_free_min, "Min pages reserved for kernel"); 306VM_STATS_VM(v_interrupt_free_min, "Reserved pages for interrupt code"); 307VM_STATS_VM(v_forks, "Number of fork() calls"); 308VM_STATS_VM(v_vforks, "Number of vfork() calls"); 309VM_STATS_VM(v_rforks, "Number of rfork() calls"); 310VM_STATS_VM(v_kthreads, "Number of fork() calls by kernel"); 311VM_STATS_VM(v_forkpages, "VM pages affected by fork()"); 312VM_STATS_VM(v_vforkpages, "VM pages affected by vfork()"); 313VM_STATS_VM(v_rforkpages, "VM pages affected by rfork()"); 314VM_STATS_VM(v_kthreadpages, "VM pages affected by fork() by kernel"); 315 316SYSCTL_INT(_vm_stats_misc, OID_AUTO, zero_page_count, CTLFLAG_RD, 317 &vm_page_zero_count, 0, "Number of zero-ed free pages"); 318