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("$FreeBSD$"); 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 int debug_mpsafevm = 1; 59 TUNABLE_INT("debug.mpsafevm", &debug_mpsafevm); 60 SYSCTL_INT(_debug, OID_AUTO, mpsafevm, CTLFLAG_RD, &debug_mpsafevm, 0, 61 "Enable/disable MPSAFE virtual memory support"); 62 63 struct vmmeter cnt; 64 65 int maxslp = MAXSLP; 66 67 SYSCTL_UINT(_vm, VM_V_FREE_MIN, v_free_min, 68 CTLFLAG_RW, &cnt.v_free_min, 0, ""); 69 SYSCTL_UINT(_vm, VM_V_FREE_TARGET, v_free_target, 70 CTLFLAG_RW, &cnt.v_free_target, 0, ""); 71 SYSCTL_UINT(_vm, VM_V_FREE_RESERVED, v_free_reserved, 72 CTLFLAG_RW, &cnt.v_free_reserved, 0, ""); 73 SYSCTL_UINT(_vm, VM_V_INACTIVE_TARGET, v_inactive_target, 74 CTLFLAG_RW, &cnt.v_inactive_target, 0, ""); 75 SYSCTL_UINT(_vm, VM_V_CACHE_MIN, v_cache_min, 76 CTLFLAG_RW, &cnt.v_cache_min, 0, ""); 77 SYSCTL_UINT(_vm, VM_V_CACHE_MAX, v_cache_max, 78 CTLFLAG_RW, &cnt.v_cache_max, 0, ""); 79 SYSCTL_UINT(_vm, VM_V_PAGEOUT_FREE_MIN, v_pageout_free_min, 80 CTLFLAG_RW, &cnt.v_pageout_free_min, 0, ""); 81 SYSCTL_UINT(_vm, OID_AUTO, v_free_severe, 82 CTLFLAG_RW, &cnt.v_free_severe, 0, ""); 83 84 static int 85 sysctl_vm_loadavg(SYSCTL_HANDLER_ARGS) 86 { 87 #ifdef SCTL_MASK32 88 u_int32_t la[4]; 89 90 if (req->flags & SCTL_MASK32) { 91 la[0] = averunnable.ldavg[0]; 92 la[1] = averunnable.ldavg[1]; 93 la[2] = averunnable.ldavg[2]; 94 la[3] = averunnable.fscale; 95 return SYSCTL_OUT(req, la, sizeof(la)); 96 } else 97 #endif 98 return SYSCTL_OUT(req, &averunnable, sizeof(averunnable)); 99 } 100 SYSCTL_PROC(_vm, VM_LOADAVG, loadavg, CTLTYPE_STRUCT|CTLFLAG_RD, 101 NULL, 0, sysctl_vm_loadavg, "S,loadavg", "Machine loadaverage history"); 102 103 static int 104 vmtotal(SYSCTL_HANDLER_ARGS) 105 { 106 /* XXXKSE almost completely broken */ 107 struct proc *p; 108 struct vmtotal total, *totalp; 109 vm_map_entry_t entry; 110 vm_object_t object; 111 vm_map_t map; 112 int paging; 113 struct thread *td; 114 struct vmspace *vm; 115 116 totalp = &total; 117 bzero(totalp, sizeof *totalp); 118 /* 119 * Mark all objects as inactive. 120 */ 121 GIANT_REQUIRED; 122 mtx_lock(&vm_object_list_mtx); 123 TAILQ_FOREACH(object, &vm_object_list, object_list) { 124 if (!VM_OBJECT_TRYLOCK(object)) { 125 /* 126 * Avoid a lock-order reversal. Consequently, 127 * the reported number of active pages may be 128 * greater than the actual number. 129 */ 130 continue; 131 } 132 vm_object_clear_flag(object, OBJ_ACTIVE); 133 VM_OBJECT_UNLOCK(object); 134 } 135 mtx_unlock(&vm_object_list_mtx); 136 /* 137 * Calculate process statistics. 138 */ 139 sx_slock(&allproc_lock); 140 FOREACH_PROC_IN_SYSTEM(p) { 141 if (p->p_flag & P_SYSTEM) 142 continue; 143 mtx_lock_spin(&sched_lock); 144 switch (p->p_state) { 145 case PRS_NEW: 146 mtx_unlock_spin(&sched_lock); 147 continue; 148 break; 149 default: 150 FOREACH_THREAD_IN_PROC(p, td) { 151 /* Need new statistics XXX */ 152 switch (td->td_state) { 153 case TDS_INHIBITED: 154 if (TD_ON_LOCK(td) || 155 (td->td_inhibitors == 156 TDI_SWAPPED)) { 157 totalp->t_sw++; 158 } else if (TD_IS_SLEEPING(td) || 159 TD_AWAITING_INTR(td) || 160 TD_IS_SUSPENDED(td)) { 161 if (td->td_priority <= PZERO) 162 totalp->t_dw++; 163 else 164 totalp->t_sl++; 165 } 166 break; 167 168 case TDS_CAN_RUN: 169 totalp->t_sw++; 170 break; 171 case TDS_RUNQ: 172 case TDS_RUNNING: 173 totalp->t_rq++; 174 continue; 175 default: 176 break; 177 } 178 } 179 } 180 mtx_unlock_spin(&sched_lock); 181 /* 182 * Note active objects. 183 */ 184 paging = 0; 185 vm = vmspace_acquire_ref(p); 186 if (vm == NULL) 187 continue; 188 map = &vm->vm_map; 189 vm_map_lock_read(map); 190 for (entry = map->header.next; 191 entry != &map->header; entry = entry->next) { 192 if ((entry->eflags & MAP_ENTRY_IS_SUB_MAP) || 193 (object = entry->object.vm_object) == NULL) 194 continue; 195 VM_OBJECT_LOCK(object); 196 vm_object_set_flag(object, OBJ_ACTIVE); 197 paging |= object->paging_in_progress; 198 VM_OBJECT_UNLOCK(object); 199 } 200 vm_map_unlock_read(map); 201 vmspace_free(vm); 202 if (paging) 203 totalp->t_pw++; 204 } 205 sx_sunlock(&allproc_lock); 206 /* 207 * Calculate object memory usage statistics. 208 */ 209 mtx_lock(&vm_object_list_mtx); 210 TAILQ_FOREACH(object, &vm_object_list, object_list) { 211 /* 212 * Perform unsynchronized reads on the object to avoid 213 * a lock-order reversal. In this case, the lack of 214 * synchronization should not impair the accuracy of 215 * the reported statistics. 216 */ 217 if (object->type == OBJT_DEVICE) { 218 /* 219 * Devices, like /dev/mem, will badly skew our totals. 220 */ 221 continue; 222 } 223 totalp->t_vm += object->size; 224 totalp->t_rm += object->resident_page_count; 225 if (object->flags & OBJ_ACTIVE) { 226 totalp->t_avm += object->size; 227 totalp->t_arm += object->resident_page_count; 228 } 229 if (object->shadow_count > 1) { 230 /* shared object */ 231 totalp->t_vmshr += object->size; 232 totalp->t_rmshr += object->resident_page_count; 233 if (object->flags & OBJ_ACTIVE) { 234 totalp->t_avmshr += object->size; 235 totalp->t_armshr += object->resident_page_count; 236 } 237 } 238 } 239 mtx_unlock(&vm_object_list_mtx); 240 totalp->t_free = cnt.v_free_count + cnt.v_cache_count; 241 return (sysctl_handle_opaque(oidp, totalp, sizeof total, req)); 242 } 243 244 /* 245 * vcnt() - accumulate statistics from all cpus and the global cnt 246 * structure. 247 * 248 * The vmmeter structure is now per-cpu as well as global. Those 249 * statistics which can be kept on a per-cpu basis (to avoid cache 250 * stalls between cpus) can be moved to the per-cpu vmmeter. Remaining 251 * statistics, such as v_free_reserved, are left in the global 252 * structure. 253 * 254 * (sysctl_oid *oidp, void *arg1, int arg2, struct sysctl_req *req) 255 */ 256 static int 257 vcnt(SYSCTL_HANDLER_ARGS) 258 { 259 int count = *(int *)arg1; 260 int offset = (char *)arg1 - (char *)&cnt; 261 #ifdef SMP 262 int i; 263 264 for (i = 0; i < mp_ncpus; ++i) { 265 struct pcpu *pcpu = pcpu_find(i); 266 count += *(int *)((char *)&pcpu->pc_cnt + offset); 267 } 268 #else 269 count += *(int *)((char *)PCPU_PTR(cnt) + offset); 270 #endif 271 return (SYSCTL_OUT(req, &count, sizeof(int))); 272 } 273 274 SYSCTL_PROC(_vm, VM_TOTAL, vmtotal, CTLTYPE_OPAQUE|CTLFLAG_RD, 275 0, sizeof(struct vmtotal), vmtotal, "S,vmtotal", 276 "System virtual memory statistics"); 277 SYSCTL_NODE(_vm, OID_AUTO, stats, CTLFLAG_RW, 0, "VM meter stats"); 278 static SYSCTL_NODE(_vm_stats, OID_AUTO, sys, CTLFLAG_RW, 0, 279 "VM meter sys stats"); 280 static SYSCTL_NODE(_vm_stats, OID_AUTO, vm, CTLFLAG_RW, 0, 281 "VM meter vm stats"); 282 SYSCTL_NODE(_vm_stats, OID_AUTO, misc, CTLFLAG_RW, 0, "VM meter misc stats"); 283 284 SYSCTL_PROC(_vm_stats_sys, OID_AUTO, v_swtch, CTLTYPE_UINT|CTLFLAG_RD, 285 &cnt.v_swtch, 0, vcnt, "IU", "Context switches"); 286 SYSCTL_PROC(_vm_stats_sys, OID_AUTO, v_trap, CTLTYPE_UINT|CTLFLAG_RD, 287 &cnt.v_trap, 0, vcnt, "IU", "Traps"); 288 SYSCTL_PROC(_vm_stats_sys, OID_AUTO, v_syscall, CTLTYPE_UINT|CTLFLAG_RD, 289 &cnt.v_syscall, 0, vcnt, "IU", "Syscalls"); 290 SYSCTL_PROC(_vm_stats_sys, OID_AUTO, v_intr, CTLTYPE_UINT|CTLFLAG_RD, 291 &cnt.v_intr, 0, vcnt, "IU", "Hardware interrupts"); 292 SYSCTL_PROC(_vm_stats_sys, OID_AUTO, v_soft, CTLTYPE_UINT|CTLFLAG_RD, 293 &cnt.v_soft, 0, vcnt, "IU", "Software interrupts"); 294 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_vm_faults, CTLTYPE_UINT|CTLFLAG_RD, 295 &cnt.v_vm_faults, 0, vcnt, "IU", "VM faults"); 296 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_cow_faults, CTLTYPE_UINT|CTLFLAG_RD, 297 &cnt.v_cow_faults, 0, vcnt, "IU", "COW faults"); 298 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_cow_optim, CTLTYPE_UINT|CTLFLAG_RD, 299 &cnt.v_cow_optim, 0, vcnt, "IU", "Optimized COW faults"); 300 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_zfod, CTLTYPE_UINT|CTLFLAG_RD, 301 &cnt.v_zfod, 0, vcnt, "IU", "Zero fill"); 302 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_ozfod, CTLTYPE_UINT|CTLFLAG_RD, 303 &cnt.v_ozfod, 0, vcnt, "IU", "Optimized zero fill"); 304 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_swapin, CTLTYPE_UINT|CTLFLAG_RD, 305 &cnt.v_swapin, 0, vcnt, "IU", "Swapin operations"); 306 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_swapout, CTLTYPE_UINT|CTLFLAG_RD, 307 &cnt.v_swapout, 0, vcnt, "IU", "Swapout operations"); 308 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_swappgsin, CTLTYPE_UINT|CTLFLAG_RD, 309 &cnt.v_swappgsin, 0, vcnt, "IU", "Swapin pages"); 310 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_swappgsout, CTLTYPE_UINT|CTLFLAG_RD, 311 &cnt.v_swappgsout, 0, vcnt, "IU", "Swapout pages"); 312 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_vnodein, CTLTYPE_UINT|CTLFLAG_RD, 313 &cnt.v_vnodein, 0, vcnt, "IU", "Vnodein operations"); 314 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_vnodeout, CTLTYPE_UINT|CTLFLAG_RD, 315 &cnt.v_vnodeout, 0, vcnt, "IU", "Vnodeout operations"); 316 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_vnodepgsin, CTLTYPE_UINT|CTLFLAG_RD, 317 &cnt.v_vnodepgsin, 0, vcnt, "IU", "Vnodein pages"); 318 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_vnodepgsout, CTLTYPE_UINT|CTLFLAG_RD, 319 &cnt.v_vnodepgsout, 0, vcnt, "IU", "Vnodeout pages"); 320 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_intrans, CTLTYPE_UINT|CTLFLAG_RD, 321 &cnt.v_intrans, 0, vcnt, "IU", "In transit page blocking"); 322 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_reactivated, CTLTYPE_UINT|CTLFLAG_RD, 323 &cnt.v_reactivated, 0, vcnt, "IU", "Reactivated pages"); 324 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_pdwakeups, CTLTYPE_UINT|CTLFLAG_RD, 325 &cnt.v_pdwakeups, 0, vcnt, "IU", "Pagedaemon wakeups"); 326 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_pdpages, CTLTYPE_UINT|CTLFLAG_RD, 327 &cnt.v_pdpages, 0, vcnt, "IU", "Pagedaemon page scans"); 328 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_dfree, CTLTYPE_UINT|CTLFLAG_RD, 329 &cnt.v_dfree, 0, vcnt, "IU", ""); 330 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_pfree, CTLTYPE_UINT|CTLFLAG_RD, 331 &cnt.v_pfree, 0, vcnt, "IU", ""); 332 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_tfree, CTLTYPE_UINT|CTLFLAG_RD, 333 &cnt.v_tfree, 0, vcnt, "IU", ""); 334 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_page_size, CTLTYPE_UINT|CTLFLAG_RD, 335 &cnt.v_page_size, 0, vcnt, "IU", ""); 336 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_page_count, CTLTYPE_UINT|CTLFLAG_RD, 337 &cnt.v_page_count, 0, vcnt, "IU", ""); 338 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_free_reserved, CTLTYPE_UINT|CTLFLAG_RD, 339 &cnt.v_free_reserved, 0, vcnt, "IU", ""); 340 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_free_target, CTLTYPE_UINT|CTLFLAG_RD, 341 &cnt.v_free_target, 0, vcnt, "IU", ""); 342 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_free_min, CTLTYPE_UINT|CTLFLAG_RD, 343 &cnt.v_free_min, 0, vcnt, "IU", ""); 344 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_free_count, CTLTYPE_UINT|CTLFLAG_RD, 345 &cnt.v_free_count, 0, vcnt, "IU", ""); 346 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_wire_count, CTLTYPE_UINT|CTLFLAG_RD, 347 &cnt.v_wire_count, 0, vcnt, "IU", ""); 348 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_active_count, CTLTYPE_UINT|CTLFLAG_RD, 349 &cnt.v_active_count, 0, vcnt, "IU", ""); 350 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_inactive_target, CTLTYPE_UINT|CTLFLAG_RD, 351 &cnt.v_inactive_target, 0, vcnt, "IU", ""); 352 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_inactive_count, CTLTYPE_UINT|CTLFLAG_RD, 353 &cnt.v_inactive_count, 0, vcnt, "IU", ""); 354 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_cache_count, CTLTYPE_UINT|CTLFLAG_RD, 355 &cnt.v_cache_count, 0, vcnt, "IU", ""); 356 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_cache_min, CTLTYPE_UINT|CTLFLAG_RD, 357 &cnt.v_cache_min, 0, vcnt, "IU", ""); 358 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_cache_max, CTLTYPE_UINT|CTLFLAG_RD, 359 &cnt.v_cache_max, 0, vcnt, "IU", ""); 360 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_pageout_free_min, CTLTYPE_UINT|CTLFLAG_RD, 361 &cnt.v_pageout_free_min, 0, vcnt, "IU", ""); 362 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_interrupt_free_min, CTLTYPE_UINT|CTLFLAG_RD, 363 &cnt.v_interrupt_free_min, 0, vcnt, "IU", ""); 364 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_forks, CTLTYPE_UINT|CTLFLAG_RD, 365 &cnt.v_forks, 0, vcnt, "IU", "Number of fork() calls"); 366 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_vforks, CTLTYPE_UINT|CTLFLAG_RD, 367 &cnt.v_vforks, 0, vcnt, "IU", "Number of vfork() calls"); 368 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_rforks, CTLTYPE_UINT|CTLFLAG_RD, 369 &cnt.v_rforks, 0, vcnt, "IU", "Number of rfork() calls"); 370 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_kthreads, CTLTYPE_UINT|CTLFLAG_RD, 371 &cnt.v_kthreads, 0, vcnt, "IU", "Number of fork() calls by kernel"); 372 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_forkpages, CTLTYPE_UINT|CTLFLAG_RD, 373 &cnt.v_forkpages, 0, vcnt, "IU", "VM pages affected by fork()"); 374 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_vforkpages, CTLTYPE_UINT|CTLFLAG_RD, 375 &cnt.v_vforkpages, 0, vcnt, "IU", "VM pages affected by vfork()"); 376 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_rforkpages, CTLTYPE_UINT|CTLFLAG_RD, 377 &cnt.v_rforkpages, 0, vcnt, "IU", "VM pages affected by rfork()"); 378 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_kthreadpages, CTLTYPE_UINT|CTLFLAG_RD, 379 &cnt.v_kthreadpages, 0, vcnt, "IU", "VM pages affected by fork() by kernel"); 380 381 SYSCTL_INT(_vm_stats_misc, OID_AUTO, 382 zero_page_count, CTLFLAG_RD, &vm_page_zero_count, 0, ""); 383 #if 0 384 SYSCTL_INT(_vm_stats_misc, OID_AUTO, 385 page_mask, CTLFLAG_RD, &page_mask, 0, ""); 386 SYSCTL_INT(_vm_stats_misc, OID_AUTO, 387 page_shift, CTLFLAG_RD, &page_shift, 0, ""); 388 SYSCTL_INT(_vm_stats_misc, OID_AUTO, 389 first_page, CTLFLAG_RD, &first_page, 0, ""); 390 SYSCTL_INT(_vm_stats_misc, OID_AUTO, 391 last_page, CTLFLAG_RD, &last_page, 0, ""); 392 #endif 393