xref: /freebsd/sys/vm/vm_meter.c (revision ce4946daa5ce852d28008dac492029500ab2ee95)
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  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)vm_meter.c	8.4 (Berkeley) 1/4/94
34  * $FreeBSD$
35  */
36 
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/kernel.h>
40 #include <sys/lock.h>
41 #include <sys/mutex.h>
42 #include <sys/proc.h>
43 #include <sys/resource.h>
44 #include <sys/sx.h>
45 #include <sys/vmmeter.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 
56 struct loadavg averunnable =
57 	{ {0, 0, 0}, FSCALE };	/* load average, of runnable procs */
58 
59 struct vmmeter cnt;
60 
61 static int maxslp = MAXSLP;
62 
63 /*
64  * Constants for averages over 1, 5, and 15 minutes
65  * when sampling at 5 second intervals.
66  */
67 static fixpt_t cexp[3] = {
68 	0.9200444146293232 * FSCALE,	/* exp(-1/12) */
69 	0.9834714538216174 * FSCALE,	/* exp(-1/60) */
70 	0.9944598480048967 * FSCALE,	/* exp(-1/180) */
71 };
72 
73 /*
74  * Compute a tenex style load average of a quantity on
75  * 1, 5 and 15 minute intervals.
76  */
77 static void
78 loadav(struct loadavg *avg)
79 {
80 	register int i, nrun;
81 	register struct proc *p;
82 
83 	sx_slock(&allproc_lock);
84 	for (nrun = 0, p = LIST_FIRST(&allproc); p != 0; p = LIST_NEXT(p, p_list)) {
85 		switch (p->p_stat) {
86 		case SSLEEP:
87 			if (p->p_pri.pri_level > PZERO ||
88 			    p->p_slptime != 0)
89 				continue;
90 			/* FALLTHROUGH */
91 		case SRUN:
92 			if ((p->p_flag & P_NOLOAD) != 0)
93 				continue;
94 			/* FALLTHROUGH */
95 		case SIDL:
96 			nrun++;
97 		}
98 	}
99 	sx_sunlock(&allproc_lock);
100 	for (i = 0; i < 3; i++)
101 		avg->ldavg[i] = (cexp[i] * avg->ldavg[i] +
102 		    nrun * FSCALE * (FSCALE - cexp[i])) >> FSHIFT;
103 }
104 
105 void
106 vmmeter()
107 {
108 
109 	if (time_second % 5 == 0)
110 		loadav(&averunnable);
111 	if (proc0.p_slptime > maxslp / 2)
112 		wakeup(&proc0);
113 }
114 
115 SYSCTL_UINT(_vm, VM_V_FREE_MIN, v_free_min,
116 	CTLFLAG_RW, &cnt.v_free_min, 0, "");
117 SYSCTL_UINT(_vm, VM_V_FREE_TARGET, v_free_target,
118 	CTLFLAG_RW, &cnt.v_free_target, 0, "");
119 SYSCTL_UINT(_vm, VM_V_FREE_RESERVED, v_free_reserved,
120 	CTLFLAG_RW, &cnt.v_free_reserved, 0, "");
121 SYSCTL_UINT(_vm, VM_V_INACTIVE_TARGET, v_inactive_target,
122 	CTLFLAG_RW, &cnt.v_inactive_target, 0, "");
123 SYSCTL_UINT(_vm, VM_V_CACHE_MIN, v_cache_min,
124 	CTLFLAG_RW, &cnt.v_cache_min, 0, "");
125 SYSCTL_UINT(_vm, VM_V_CACHE_MAX, v_cache_max,
126 	CTLFLAG_RW, &cnt.v_cache_max, 0, "");
127 SYSCTL_UINT(_vm, VM_V_PAGEOUT_FREE_MIN, v_pageout_free_min,
128 	CTLFLAG_RW, &cnt.v_pageout_free_min, 0, "");
129 SYSCTL_UINT(_vm, OID_AUTO, v_free_severe,
130 	CTLFLAG_RW, &cnt.v_free_severe, 0, "");
131 
132 SYSCTL_STRUCT(_vm, VM_LOADAVG, loadavg, CTLFLAG_RD,
133     &averunnable, loadavg, "Machine loadaverage history");
134 
135 static int
136 vmtotal(SYSCTL_HANDLER_ARGS)
137 {
138 	struct proc *p;
139 	struct vmtotal total, *totalp;
140 	vm_map_entry_t entry;
141 	vm_object_t object;
142 	vm_map_t map;
143 	int paging;
144 
145 	totalp = &total;
146 	bzero(totalp, sizeof *totalp);
147 	/*
148 	 * Mark all objects as inactive.
149 	 */
150 	TAILQ_FOREACH(object, &vm_object_list, object_list)
151 		vm_object_clear_flag(object, OBJ_ACTIVE);
152 	/*
153 	 * Calculate process statistics.
154 	 */
155 	sx_slock(&allproc_lock);
156 	LIST_FOREACH(p, &allproc, p_list) {
157 		if (p->p_flag & P_SYSTEM)
158 			continue;
159 		mtx_lock_spin(&sched_lock);
160 		switch (p->p_stat) {
161 		case 0:
162 			mtx_unlock_spin(&sched_lock);
163 			continue;
164 
165 		case SMTX:
166 		case SSLEEP:
167 		case SSTOP:
168 			if (p->p_sflag & PS_INMEM) {
169 				if (p->p_pri.pri_level <= PZERO)
170 					totalp->t_dw++;
171 				else if (p->p_slptime < maxslp)
172 					totalp->t_sl++;
173 			} else if (p->p_slptime < maxslp)
174 				totalp->t_sw++;
175 			if (p->p_slptime >= maxslp) {
176 				mtx_unlock_spin(&sched_lock);
177 				continue;
178 			}
179 			break;
180 
181 		case SWAIT:
182 			totalp->t_sl++;
183 			continue;
184 
185 		case SRUN:
186 		case SIDL:
187 			if (p->p_sflag & PS_INMEM)
188 				totalp->t_rq++;
189 			else
190 				totalp->t_sw++;
191 			if (p->p_stat == SIDL) {
192 				mtx_unlock_spin(&sched_lock);
193 				continue;
194 			}
195 			break;
196 		}
197 		mtx_unlock_spin(&sched_lock);
198 		/*
199 		 * Note active objects.
200 		 */
201 		paging = 0;
202 		for (map = &p->p_vmspace->vm_map, entry = map->header.next;
203 		    entry != &map->header; entry = entry->next) {
204 			if ((entry->eflags & MAP_ENTRY_IS_SUB_MAP) ||
205 			    entry->object.vm_object == NULL)
206 				continue;
207 			vm_object_set_flag(entry->object.vm_object, OBJ_ACTIVE);
208 			paging |= entry->object.vm_object->paging_in_progress;
209 		}
210 		if (paging)
211 			totalp->t_pw++;
212 	}
213 	sx_sunlock(&allproc_lock);
214 	/*
215 	 * Calculate object memory usage statistics.
216 	 */
217 	TAILQ_FOREACH(object, &vm_object_list, object_list) {
218 		/*
219 		 * devices, like /dev/mem, will badly skew our totals
220 		 */
221 		if (object->type == OBJT_DEVICE)
222 			continue;
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 	totalp->t_free = cnt.v_free_count + cnt.v_cache_count;
240 	return (sysctl_handle_opaque(oidp, totalp, sizeof total, req));
241 }
242 
243 SYSCTL_PROC(_vm, VM_METER, vmmeter, CTLTYPE_OPAQUE|CTLFLAG_RD,
244     0, sizeof(struct vmtotal), vmtotal, "S,vmtotal",
245     "System virtual memory statistics");
246 SYSCTL_NODE(_vm, OID_AUTO, stats, CTLFLAG_RW, 0, "VM meter stats");
247 SYSCTL_NODE(_vm_stats, OID_AUTO, sys, CTLFLAG_RW, 0, "VM meter sys stats");
248 SYSCTL_NODE(_vm_stats, OID_AUTO, vm, CTLFLAG_RW, 0, "VM meter vm stats");
249 SYSCTL_NODE(_vm_stats, OID_AUTO, misc, CTLFLAG_RW, 0, "VM meter misc stats");
250 SYSCTL_UINT(_vm_stats_sys, OID_AUTO,
251 	v_swtch, CTLFLAG_RD, &cnt.v_swtch, 0, "Context switches");
252 SYSCTL_UINT(_vm_stats_sys, OID_AUTO,
253 	v_trap, CTLFLAG_RD, &cnt.v_trap, 0, "Traps");
254 SYSCTL_UINT(_vm_stats_sys, OID_AUTO,
255 	v_syscall, CTLFLAG_RD, &cnt.v_syscall, 0, "Syscalls");
256 SYSCTL_UINT(_vm_stats_sys, OID_AUTO, v_intr, CTLFLAG_RD,
257     &cnt.v_intr, 0, "Hardware interrupts");
258 SYSCTL_UINT(_vm_stats_sys, OID_AUTO, v_soft, CTLFLAG_RD,
259     &cnt.v_soft, 0, "Software interrupts");
260 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
261 	v_vm_faults, CTLFLAG_RD, &cnt.v_vm_faults, 0, "VM faults");
262 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
263 	v_cow_faults, CTLFLAG_RD, &cnt.v_cow_faults, 0, "COW faults");
264 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
265 	v_cow_optim, CTLFLAG_RD, &cnt.v_cow_optim, 0, "Optimized COW faults");
266 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
267 	v_zfod, CTLFLAG_RD, &cnt.v_zfod, 0, "Zero fill");
268 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
269 	v_ozfod, CTLFLAG_RD, &cnt.v_ozfod, 0, "Optimized zero fill");
270 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
271 	v_swapin, CTLFLAG_RD, &cnt.v_swapin, 0, "Swapin operations");
272 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
273 	v_swapout, CTLFLAG_RD, &cnt.v_swapout, 0, "Swapout operations");
274 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
275 	v_swappgsin, CTLFLAG_RD, &cnt.v_swappgsin, 0, "Swapin pages");
276 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
277 	v_swappgsout, CTLFLAG_RD, &cnt.v_swappgsout, 0, "Swapout pages");
278 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
279 	v_vnodein, CTLFLAG_RD, &cnt.v_vnodein, 0, "Vnodein operations");
280 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
281 	v_vnodeout, CTLFLAG_RD, &cnt.v_vnodeout, 0, "Vnodeout operations");
282 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
283 	v_vnodepgsin, CTLFLAG_RD, &cnt.v_vnodepgsin, 0, "Vnodein pages");
284 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
285 	v_vnodepgsout, CTLFLAG_RD, &cnt.v_vnodepgsout, 0, "Vnodeout pages");
286 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
287 	v_intrans, CTLFLAG_RD, &cnt.v_intrans, 0, "In transit page blocking");
288 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
289 	v_reactivated, CTLFLAG_RD, &cnt.v_reactivated, 0, "Reactivated pages");
290 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
291 	v_pdwakeups, CTLFLAG_RD, &cnt.v_pdwakeups, 0, "Pagedaemon wakeups");
292 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
293 	v_pdpages, CTLFLAG_RD, &cnt.v_pdpages, 0, "Pagedaemon page scans");
294 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
295 	v_dfree, CTLFLAG_RD, &cnt.v_dfree, 0, "");
296 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
297 	v_pfree, CTLFLAG_RD, &cnt.v_pfree, 0, "");
298 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
299 	v_tfree, CTLFLAG_RD, &cnt.v_tfree, 0, "");
300 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
301 	v_page_size, CTLFLAG_RD, &cnt.v_page_size, 0, "");
302 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
303 	v_page_count, CTLFLAG_RD, &cnt.v_page_count, 0, "");
304 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
305 	v_free_reserved, CTLFLAG_RD, &cnt.v_free_reserved, 0, "");
306 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
307 	v_free_target, CTLFLAG_RD, &cnt.v_free_target, 0, "");
308 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
309 	v_free_min, CTLFLAG_RD, &cnt.v_free_min, 0, "");
310 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
311 	v_free_count, CTLFLAG_RD, &cnt.v_free_count, 0, "");
312 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
313 	v_wire_count, CTLFLAG_RD, &cnt.v_wire_count, 0, "");
314 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
315 	v_active_count, CTLFLAG_RD, &cnt.v_active_count, 0, "");
316 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
317 	v_inactive_target, CTLFLAG_RD, &cnt.v_inactive_target, 0, "");
318 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
319 	v_inactive_count, CTLFLAG_RD, &cnt.v_inactive_count, 0, "");
320 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
321 	v_cache_count, CTLFLAG_RD, &cnt.v_cache_count, 0, "");
322 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
323 	v_cache_min, CTLFLAG_RD, &cnt.v_cache_min, 0, "");
324 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
325 	v_cache_max, CTLFLAG_RD, &cnt.v_cache_max, 0, "");
326 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
327 	v_pageout_free_min, CTLFLAG_RD, &cnt.v_pageout_free_min, 0, "");
328 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
329 	v_interrupt_free_min, CTLFLAG_RD, &cnt.v_interrupt_free_min, 0, "");
330 SYSCTL_INT(_vm_stats_misc, OID_AUTO,
331 	zero_page_count, CTLFLAG_RD, &vm_page_zero_count, 0, "");
332 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
333 	v_forks, CTLFLAG_RD, &cnt.v_forks, 0, "Number of fork() calls");
334 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
335 	v_vforks, CTLFLAG_RD, &cnt.v_vforks, 0, "Number of vfork() calls");
336 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
337 	v_rforks, CTLFLAG_RD, &cnt.v_rforks, 0, "Number of rfork() calls");
338 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
339 	v_kthreads, CTLFLAG_RD, &cnt.v_kthreads, 0, "Number of fork() calls by kernel");
340 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
341 	v_forkpages, CTLFLAG_RD, &cnt.v_forkpages, 0, "VM pages affected by fork()");
342 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
343 	v_vforkpages, CTLFLAG_RD, &cnt.v_vforkpages, 0, "VM pages affected by vfork()");
344 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
345 	v_rforkpages, CTLFLAG_RD, &cnt.v_rforkpages, 0, "VM pages affected by rfork()");
346 SYSCTL_UINT(_vm_stats_vm, OID_AUTO,
347 	v_kthreadpages, CTLFLAG_RD, &cnt.v_kthreadpages, 0, "VM pages affected by fork() by kernel");
348 #if 0
349 SYSCTL_INT(_vm_stats_misc, OID_AUTO,
350 	page_mask, CTLFLAG_RD, &page_mask, 0, "");
351 SYSCTL_INT(_vm_stats_misc, OID_AUTO,
352 	page_shift, CTLFLAG_RD, &page_shift, 0, "");
353 SYSCTL_INT(_vm_stats_misc, OID_AUTO,
354 	first_page, CTLFLAG_RD, &first_page, 0, "");
355 SYSCTL_INT(_vm_stats_misc, OID_AUTO,
356 	last_page, CTLFLAG_RD, &last_page, 0, "");
357 SYSCTL_INT(_vm_stats_misc, OID_AUTO,
358 	vm_page_bucket_count, CTLFLAG_RD, &vm_page_bucket_count, 0, "");
359 SYSCTL_INT(_vm_stats_misc, OID_AUTO,
360 	vm_page_hash_mask, CTLFLAG_RD, &vm_page_hash_mask, 0, "");
361 #endif
362 
363 /*
364  * Further sysctls used by systat: hw.nintr, hw.intrnames, hw.intrcnt.
365  * This does probably not really fit in here, but it is somehow connected.
366  * The definitions for this are machdep, but are currently defined for
367  * any architecture.
368  */
369 
370 /* include the machdep stuff */
371 #include <machine/intrcnt.h>
372 
373 int nintr = INTRCNT_COUNT;
374 SYSCTL_INT(_hw, OID_AUTO, nintr, CTLFLAG_RD, &nintr, 0, "Number of Interrupts");
375 
376 SYSCTL_OPAQUE(_hw, OID_AUTO, intrcnt, CTLFLAG_RD, &intrcnt,
377    sizeof(long) * INTRCNT_COUNT, "", "Interrupt Counts");
378 
379 /*
380  * We do not know the length in advance (in an MI fashion), so calculate things
381  * at run-time.
382  */
383 static int
384 sysctl_intrnames(SYSCTL_HANDLER_ARGS)
385 {
386 	return sysctl_handle_opaque(oidp, intrnames, eintrnames - intrnames,
387 	   req);
388 }
389 
390 SYSCTL_PROC(_hw, OID_AUTO, intrnames, CTLTYPE_OPAQUE | CTLFLAG_RD, NULL, 0,
391     sysctl_intrnames, "", "Interrupt Names");
392