xref: /freebsd/sys/vm/vm_meter.c (revision 10f0bcab61ef441cb5af32fb706688d8cbd55dc0)
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 struct vmmeter cnt;
56 
57 int maxslp = MAXSLP;
58 
59 SYSCTL_UINT(_vm, VM_V_FREE_MIN, v_free_min,
60 	CTLFLAG_RW, &cnt.v_free_min, 0, "");
61 SYSCTL_UINT(_vm, VM_V_FREE_TARGET, v_free_target,
62 	CTLFLAG_RW, &cnt.v_free_target, 0, "");
63 SYSCTL_UINT(_vm, VM_V_FREE_RESERVED, v_free_reserved,
64 	CTLFLAG_RW, &cnt.v_free_reserved, 0, "");
65 SYSCTL_UINT(_vm, VM_V_INACTIVE_TARGET, v_inactive_target,
66 	CTLFLAG_RW, &cnt.v_inactive_target, 0, "");
67 SYSCTL_UINT(_vm, VM_V_CACHE_MIN, v_cache_min,
68 	CTLFLAG_RW, &cnt.v_cache_min, 0, "");
69 SYSCTL_UINT(_vm, VM_V_CACHE_MAX, v_cache_max,
70 	CTLFLAG_RW, &cnt.v_cache_max, 0, "");
71 SYSCTL_UINT(_vm, VM_V_PAGEOUT_FREE_MIN, v_pageout_free_min,
72 	CTLFLAG_RW, &cnt.v_pageout_free_min, 0, "");
73 SYSCTL_UINT(_vm, OID_AUTO, v_free_severe,
74 	CTLFLAG_RW, &cnt.v_free_severe, 0, "");
75 
76 static int
77 sysctl_vm_loadavg(SYSCTL_HANDLER_ARGS)
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 }
92 SYSCTL_PROC(_vm, VM_LOADAVG, loadavg, CTLTYPE_STRUCT|CTLFLAG_RD,
93     NULL, 0, sysctl_vm_loadavg, "S,loadavg", "Machine loadaverage history");
94 
95 static int
96 vmtotal(SYSCTL_HANDLER_ARGS)
97 {
98 	struct proc *p;
99 	struct vmtotal total;
100 	vm_map_entry_t entry;
101 	vm_object_t object;
102 	vm_map_t map;
103 	int paging;
104 	struct thread *td;
105 	struct vmspace *vm;
106 
107 	bzero(&total, sizeof(total));
108 	/*
109 	 * Mark all objects as inactive.
110 	 */
111 	GIANT_REQUIRED;
112 	mtx_lock(&vm_object_list_mtx);
113 	TAILQ_FOREACH(object, &vm_object_list, object_list) {
114 		if (!VM_OBJECT_TRYLOCK(object)) {
115 			/*
116 			 * Avoid a lock-order reversal.  Consequently,
117 			 * the reported number of active pages may be
118 			 * greater than the actual number.
119 			 */
120 			continue;
121 		}
122 		vm_object_clear_flag(object, OBJ_ACTIVE);
123 		VM_OBJECT_UNLOCK(object);
124 	}
125 	mtx_unlock(&vm_object_list_mtx);
126 	/*
127 	 * Calculate process statistics.
128 	 */
129 	sx_slock(&allproc_lock);
130 	FOREACH_PROC_IN_SYSTEM(p) {
131 		if (p->p_flag & P_SYSTEM)
132 			continue;
133 		PROC_LOCK(p);
134 		PROC_SLOCK(p);
135 		switch (p->p_state) {
136 		case PRS_NEW:
137 			PROC_SUNLOCK(p);
138 			PROC_UNLOCK(p);
139 			continue;
140 			break;
141 		default:
142 			PROC_SUNLOCK(p);
143 			FOREACH_THREAD_IN_PROC(p, td) {
144 				thread_lock(td);
145 				switch (td->td_state) {
146 				case TDS_INHIBITED:
147 					if (TD_IS_SWAPPED(td))
148 						total.t_sw++;
149 					else if (TD_IS_SLEEPING(td) &&
150 					    td->td_priority <= PZERO)
151 						total.t_dw++;
152 					else
153 						total.t_sl++;
154 					break;
155 
156 				case TDS_CAN_RUN:
157 					total.t_sw++;
158 					break;
159 				case TDS_RUNQ:
160 				case TDS_RUNNING:
161 					total.t_rq++;
162 					thread_unlock(td);
163 					continue;
164 				default:
165 					break;
166 				}
167 				thread_unlock(td);
168 			}
169 		}
170 		PROC_UNLOCK(p);
171 		/*
172 		 * Note active objects.
173 		 */
174 		paging = 0;
175 		vm = vmspace_acquire_ref(p);
176 		if (vm == NULL)
177 			continue;
178 		map = &vm->vm_map;
179 		vm_map_lock_read(map);
180 		for (entry = map->header.next;
181 		    entry != &map->header; entry = entry->next) {
182 			if ((entry->eflags & MAP_ENTRY_IS_SUB_MAP) ||
183 			    (object = entry->object.vm_object) == NULL)
184 				continue;
185 			VM_OBJECT_LOCK(object);
186 			vm_object_set_flag(object, OBJ_ACTIVE);
187 			paging |= object->paging_in_progress;
188 			VM_OBJECT_UNLOCK(object);
189 		}
190 		vm_map_unlock_read(map);
191 		vmspace_free(vm);
192 		if (paging)
193 			total.t_pw++;
194 	}
195 	sx_sunlock(&allproc_lock);
196 	/*
197 	 * Calculate object memory usage statistics.
198 	 */
199 	mtx_lock(&vm_object_list_mtx);
200 	TAILQ_FOREACH(object, &vm_object_list, object_list) {
201 		/*
202 		 * Perform unsynchronized reads on the object to avoid
203 		 * a lock-order reversal.  In this case, the lack of
204 		 * synchronization should not impair the accuracy of
205 		 * the reported statistics.
206 		 */
207 		if (object->type == OBJT_DEVICE) {
208 			/*
209 			 * Devices, like /dev/mem, will badly skew our totals.
210 			 */
211 			continue;
212 		}
213 		if (object->ref_count == 0) {
214 			/*
215 			 * Also skip unreferenced objects, including
216 			 * vnodes representing mounted file systems.
217 			 */
218 			continue;
219 		}
220 		total.t_vm += object->size;
221 		total.t_rm += object->resident_page_count;
222 		if (object->flags & OBJ_ACTIVE) {
223 			total.t_avm += object->size;
224 			total.t_arm += object->resident_page_count;
225 		}
226 		if (object->shadow_count > 1) {
227 			/* shared object */
228 			total.t_vmshr += object->size;
229 			total.t_rmshr += object->resident_page_count;
230 			if (object->flags & OBJ_ACTIVE) {
231 				total.t_avmshr += object->size;
232 				total.t_armshr += object->resident_page_count;
233 			}
234 		}
235 	}
236 	mtx_unlock(&vm_object_list_mtx);
237 	total.t_free = cnt.v_free_count + cnt.v_cache_count;
238 	return (sysctl_handle_opaque(oidp, &total, sizeof(total), req));
239 }
240 
241 /*
242  * vcnt() -	accumulate statistics from all cpus and the global cnt
243  *		structure.
244  *
245  *	The vmmeter structure is now per-cpu as well as global.  Those
246  *	statistics which can be kept on a per-cpu basis (to avoid cache
247  *	stalls between cpus) can be moved to the per-cpu vmmeter.  Remaining
248  *	statistics, such as v_free_reserved, are left in the global
249  *	structure.
250  *
251  * (sysctl_oid *oidp, void *arg1, int arg2, struct sysctl_req *req)
252  */
253 static int
254 vcnt(SYSCTL_HANDLER_ARGS)
255 {
256 	int count = *(int *)arg1;
257 	int offset = (char *)arg1 - (char *)&cnt;
258 #ifdef SMP
259 	int i;
260 
261 	for (i = 0; i < mp_ncpus; ++i) {
262 		struct pcpu *pcpu = pcpu_find(i);
263 		count += *(int *)((char *)&pcpu->pc_cnt + offset);
264 	}
265 #else
266 	count += *(int *)((char *)PCPU_PTR(cnt) + offset);
267 #endif
268 	return (SYSCTL_OUT(req, &count, sizeof(int)));
269 }
270 
271 SYSCTL_PROC(_vm, VM_TOTAL, vmtotal, CTLTYPE_OPAQUE|CTLFLAG_RD,
272     0, sizeof(struct vmtotal), vmtotal, "S,vmtotal",
273     "System virtual memory statistics");
274 SYSCTL_NODE(_vm, OID_AUTO, stats, CTLFLAG_RW, 0, "VM meter stats");
275 static SYSCTL_NODE(_vm_stats, OID_AUTO, sys, CTLFLAG_RW, 0,
276 	"VM meter sys stats");
277 static SYSCTL_NODE(_vm_stats, OID_AUTO, vm, CTLFLAG_RW, 0,
278 	"VM meter vm stats");
279 SYSCTL_NODE(_vm_stats, OID_AUTO, misc, CTLFLAG_RW, 0, "VM meter misc stats");
280 
281 SYSCTL_PROC(_vm_stats_sys, OID_AUTO, v_swtch, CTLTYPE_UINT|CTLFLAG_RD,
282 	&cnt.v_swtch, 0, vcnt, "IU", "Context switches");
283 SYSCTL_PROC(_vm_stats_sys, OID_AUTO, v_trap, CTLTYPE_UINT|CTLFLAG_RD,
284 	&cnt.v_trap, 0, vcnt, "IU", "Traps");
285 SYSCTL_PROC(_vm_stats_sys, OID_AUTO, v_syscall, CTLTYPE_UINT|CTLFLAG_RD,
286 	&cnt.v_syscall, 0, vcnt, "IU", "Syscalls");
287 SYSCTL_PROC(_vm_stats_sys, OID_AUTO, v_intr, CTLTYPE_UINT|CTLFLAG_RD,
288 	&cnt.v_intr, 0, vcnt, "IU", "Hardware interrupts");
289 SYSCTL_PROC(_vm_stats_sys, OID_AUTO, v_soft, CTLTYPE_UINT|CTLFLAG_RD,
290 	&cnt.v_soft, 0, vcnt, "IU", "Software interrupts");
291 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_vm_faults, CTLTYPE_UINT|CTLFLAG_RD,
292 	&cnt.v_vm_faults, 0, vcnt, "IU", "VM faults");
293 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_cow_faults, CTLTYPE_UINT|CTLFLAG_RD,
294 	&cnt.v_cow_faults, 0, vcnt, "IU", "COW faults");
295 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_cow_optim, CTLTYPE_UINT|CTLFLAG_RD,
296 	&cnt.v_cow_optim, 0, vcnt, "IU", "Optimized COW faults");
297 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_zfod, CTLTYPE_UINT|CTLFLAG_RD,
298 	&cnt.v_zfod, 0, vcnt, "IU", "Zero fill");
299 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_ozfod, CTLTYPE_UINT|CTLFLAG_RD,
300 	&cnt.v_ozfod, 0, vcnt, "IU", "Optimized zero fill");
301 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_swapin, CTLTYPE_UINT|CTLFLAG_RD,
302 	&cnt.v_swapin, 0, vcnt, "IU", "Swapin operations");
303 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_swapout, CTLTYPE_UINT|CTLFLAG_RD,
304 	&cnt.v_swapout, 0, vcnt, "IU", "Swapout operations");
305 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_swappgsin, CTLTYPE_UINT|CTLFLAG_RD,
306 	&cnt.v_swappgsin, 0, vcnt, "IU", "Swapin pages");
307 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_swappgsout, CTLTYPE_UINT|CTLFLAG_RD,
308 	&cnt.v_swappgsout, 0, vcnt, "IU", "Swapout pages");
309 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_vnodein, CTLTYPE_UINT|CTLFLAG_RD,
310 	&cnt.v_vnodein, 0, vcnt, "IU", "Vnodein operations");
311 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_vnodeout, CTLTYPE_UINT|CTLFLAG_RD,
312 	&cnt.v_vnodeout, 0, vcnt, "IU", "Vnodeout operations");
313 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_vnodepgsin, CTLTYPE_UINT|CTLFLAG_RD,
314 	&cnt.v_vnodepgsin, 0, vcnt, "IU", "Vnodein pages");
315 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_vnodepgsout, CTLTYPE_UINT|CTLFLAG_RD,
316 	&cnt.v_vnodepgsout, 0, vcnt, "IU", "Vnodeout pages");
317 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_intrans, CTLTYPE_UINT|CTLFLAG_RD,
318 	&cnt.v_intrans, 0, vcnt, "IU", "In transit page blocking");
319 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_reactivated, CTLTYPE_UINT|CTLFLAG_RD,
320 	&cnt.v_reactivated, 0, vcnt, "IU", "Reactivated pages");
321 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_pdwakeups, CTLTYPE_UINT|CTLFLAG_RD,
322 	&cnt.v_pdwakeups, 0, vcnt, "IU", "Pagedaemon wakeups");
323 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_pdpages, CTLTYPE_UINT|CTLFLAG_RD,
324 	&cnt.v_pdpages, 0, vcnt, "IU", "Pagedaemon page scans");
325 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_tcached, CTLTYPE_UINT|CTLFLAG_RD,
326 	&cnt.v_tcached, 0, vcnt, "IU", "Total pages cached");
327 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_dfree, CTLTYPE_UINT|CTLFLAG_RD,
328 	&cnt.v_dfree, 0, vcnt, "IU", "");
329 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_pfree, CTLTYPE_UINT|CTLFLAG_RD,
330 	&cnt.v_pfree, 0, vcnt, "IU", "");
331 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_tfree, CTLTYPE_UINT|CTLFLAG_RD,
332 	&cnt.v_tfree, 0, vcnt, "IU", "");
333 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_page_size, CTLTYPE_UINT|CTLFLAG_RD,
334 	&cnt.v_page_size, 0, vcnt, "IU", "");
335 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_page_count, CTLTYPE_UINT|CTLFLAG_RD,
336 	&cnt.v_page_count, 0, vcnt, "IU", "");
337 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_free_reserved, CTLTYPE_UINT|CTLFLAG_RD,
338 	&cnt.v_free_reserved, 0, vcnt, "IU", "");
339 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_free_target, CTLTYPE_UINT|CTLFLAG_RD,
340 	&cnt.v_free_target, 0, vcnt, "IU", "");
341 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_free_min, CTLTYPE_UINT|CTLFLAG_RD,
342 	&cnt.v_free_min, 0, vcnt, "IU", "");
343 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_free_count, CTLTYPE_UINT|CTLFLAG_RD,
344 	&cnt.v_free_count, 0, vcnt, "IU", "");
345 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_wire_count, CTLTYPE_UINT|CTLFLAG_RD,
346 	&cnt.v_wire_count, 0, vcnt, "IU", "");
347 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_active_count, CTLTYPE_UINT|CTLFLAG_RD,
348 	&cnt.v_active_count, 0, vcnt, "IU", "");
349 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_inactive_target, CTLTYPE_UINT|CTLFLAG_RD,
350 	&cnt.v_inactive_target, 0, vcnt, "IU", "");
351 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_inactive_count, CTLTYPE_UINT|CTLFLAG_RD,
352 	&cnt.v_inactive_count, 0, vcnt, "IU", "");
353 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_cache_count, CTLTYPE_UINT|CTLFLAG_RD,
354 	&cnt.v_cache_count, 0, vcnt, "IU", "");
355 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_cache_min, CTLTYPE_UINT|CTLFLAG_RD,
356 	&cnt.v_cache_min, 0, vcnt, "IU", "");
357 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_cache_max, CTLTYPE_UINT|CTLFLAG_RD,
358 	&cnt.v_cache_max, 0, vcnt, "IU", "");
359 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_pageout_free_min, CTLTYPE_UINT|CTLFLAG_RD,
360 	&cnt.v_pageout_free_min, 0, vcnt, "IU", "");
361 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_interrupt_free_min, CTLTYPE_UINT|CTLFLAG_RD,
362 	&cnt.v_interrupt_free_min, 0, vcnt, "IU", "");
363 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_forks, CTLTYPE_UINT|CTLFLAG_RD,
364 	&cnt.v_forks, 0, vcnt, "IU", "Number of fork() calls");
365 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_vforks, CTLTYPE_UINT|CTLFLAG_RD,
366 	&cnt.v_vforks, 0, vcnt, "IU", "Number of vfork() calls");
367 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_rforks, CTLTYPE_UINT|CTLFLAG_RD,
368 	&cnt.v_rforks, 0, vcnt, "IU", "Number of rfork() calls");
369 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_kthreads, CTLTYPE_UINT|CTLFLAG_RD,
370 	&cnt.v_kthreads, 0, vcnt, "IU", "Number of fork() calls by kernel");
371 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_forkpages, CTLTYPE_UINT|CTLFLAG_RD,
372 	&cnt.v_forkpages, 0, vcnt, "IU", "VM pages affected by fork()");
373 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_vforkpages, CTLTYPE_UINT|CTLFLAG_RD,
374 	&cnt.v_vforkpages, 0, vcnt, "IU", "VM pages affected by vfork()");
375 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_rforkpages, CTLTYPE_UINT|CTLFLAG_RD,
376 	&cnt.v_rforkpages, 0, vcnt, "IU", "VM pages affected by rfork()");
377 SYSCTL_PROC(_vm_stats_vm, OID_AUTO, v_kthreadpages, CTLTYPE_UINT|CTLFLAG_RD,
378 	&cnt.v_kthreadpages, 0, vcnt, "IU", "VM pages affected by fork() by kernel");
379 
380 SYSCTL_INT(_vm_stats_misc, OID_AUTO,
381 	zero_page_count, CTLFLAG_RD, &vm_page_zero_count, 0, "");
382