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/11/sys/vm/vm_meter.c 320605 2017-07-03 16:40:05Z markj $");
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 vm_cnt;
57
58SYSCTL_UINT(_vm, VM_V_FREE_MIN, v_free_min,
59	CTLFLAG_RW, &vm_cnt.v_free_min, 0, "Minimum low-free-pages threshold");
60SYSCTL_UINT(_vm, VM_V_FREE_TARGET, v_free_target,
61	CTLFLAG_RW, &vm_cnt.v_free_target, 0, "Desired free pages");
62SYSCTL_UINT(_vm, VM_V_FREE_RESERVED, v_free_reserved,
63	CTLFLAG_RW, &vm_cnt.v_free_reserved, 0, "Pages reserved for deadlock");
64SYSCTL_UINT(_vm, VM_V_INACTIVE_TARGET, v_inactive_target,
65	CTLFLAG_RW, &vm_cnt.v_inactive_target, 0, "Pages desired inactive");
66SYSCTL_UINT(_vm, VM_V_PAGEOUT_FREE_MIN, v_pageout_free_min,
67	CTLFLAG_RW, &vm_cnt.v_pageout_free_min, 0, "Min pages reserved for kernel");
68SYSCTL_UINT(_vm, OID_AUTO, v_free_severe,
69	CTLFLAG_RW, &vm_cnt.v_free_severe, 0, "Severe page depletion point");
70
71static int
72sysctl_vm_loadavg(SYSCTL_HANDLER_ARGS)
73{
74
75#ifdef SCTL_MASK32
76	u_int32_t la[4];
77
78	if (req->flags & SCTL_MASK32) {
79		la[0] = averunnable.ldavg[0];
80		la[1] = averunnable.ldavg[1];
81		la[2] = averunnable.ldavg[2];
82		la[3] = averunnable.fscale;
83		return SYSCTL_OUT(req, la, sizeof(la));
84	} else
85#endif
86		return SYSCTL_OUT(req, &averunnable, sizeof(averunnable));
87}
88SYSCTL_PROC(_vm, VM_LOADAVG, loadavg, CTLTYPE_STRUCT | CTLFLAG_RD |
89    CTLFLAG_MPSAFE, NULL, 0, sysctl_vm_loadavg, "S,loadavg",
90    "Machine loadaverage history");
91
92/*
93 * This function aims to determine if the object is mapped,
94 * specifically, if it is referenced by a vm_map_entry.  Because
95 * objects occasionally acquire transient references that do not
96 * represent a mapping, the method used here is inexact.  However, it
97 * has very low overhead and is good enough for the advisory
98 * vm.vmtotal sysctl.
99 */
100static bool
101is_object_active(vm_object_t obj)
102{
103
104	return (obj->ref_count > obj->shadow_count);
105}
106
107static int
108vmtotal(SYSCTL_HANDLER_ARGS)
109{
110	struct vmtotal total;
111	vm_object_t object;
112	struct proc *p;
113	struct thread *td;
114
115	bzero(&total, sizeof(total));
116
117	/*
118	 * Calculate process statistics.
119	 */
120	sx_slock(&allproc_lock);
121	FOREACH_PROC_IN_SYSTEM(p) {
122		if ((p->p_flag & P_SYSTEM) != 0)
123			continue;
124		PROC_LOCK(p);
125		if (p->p_state != PRS_NEW) {
126			FOREACH_THREAD_IN_PROC(p, td) {
127				thread_lock(td);
128				switch (td->td_state) {
129				case TDS_INHIBITED:
130					if (TD_IS_SWAPPED(td))
131						total.t_sw++;
132					else if (TD_IS_SLEEPING(td)) {
133						if (td->td_priority <= PZERO)
134							total.t_dw++;
135						else
136							total.t_sl++;
137						if (td->td_wchan ==
138						    &vm_cnt.v_free_count)
139							total.t_pw++;
140					}
141					break;
142				case TDS_CAN_RUN:
143					total.t_sw++;
144					break;
145				case TDS_RUNQ:
146				case TDS_RUNNING:
147					total.t_rq++;
148					break;
149				default:
150					break;
151				}
152				thread_unlock(td);
153			}
154		}
155		PROC_UNLOCK(p);
156	}
157	sx_sunlock(&allproc_lock);
158	/*
159	 * Calculate object memory usage statistics.
160	 */
161	mtx_lock(&vm_object_list_mtx);
162	TAILQ_FOREACH(object, &vm_object_list, object_list) {
163		/*
164		 * Perform unsynchronized reads on the object.  In
165		 * this case, the lack of synchronization should not
166		 * impair the accuracy of the reported statistics.
167		 */
168		if ((object->flags & OBJ_FICTITIOUS) != 0) {
169			/*
170			 * Devices, like /dev/mem, will badly skew our totals.
171			 */
172			continue;
173		}
174		if (object->ref_count == 0) {
175			/*
176			 * Also skip unreferenced objects, including
177			 * vnodes representing mounted file systems.
178			 */
179			continue;
180		}
181		if (object->ref_count == 1 &&
182		    (object->flags & OBJ_NOSPLIT) != 0) {
183			/*
184			 * Also skip otherwise unreferenced swap
185			 * objects backing tmpfs vnodes, and POSIX or
186			 * SysV shared memory.
187			 */
188			continue;
189		}
190		total.t_vm += object->size;
191		total.t_rm += object->resident_page_count;
192		if (is_object_active(object)) {
193			total.t_avm += object->size;
194			total.t_arm += object->resident_page_count;
195		}
196		if (object->shadow_count > 1) {
197			/* shared object */
198			total.t_vmshr += object->size;
199			total.t_rmshr += object->resident_page_count;
200			if (is_object_active(object)) {
201				total.t_avmshr += object->size;
202				total.t_armshr += object->resident_page_count;
203			}
204		}
205	}
206	mtx_unlock(&vm_object_list_mtx);
207	total.t_free = vm_cnt.v_free_count;
208	return (sysctl_handle_opaque(oidp, &total, sizeof(total), req));
209}
210
211/*
212 * vm_meter_cnt() -	accumulate statistics from all cpus and the global cnt
213 *			structure.
214 *
215 *	The vmmeter structure is now per-cpu as well as global.  Those
216 *	statistics which can be kept on a per-cpu basis (to avoid cache
217 *	stalls between cpus) can be moved to the per-cpu vmmeter.  Remaining
218 *	statistics, such as v_free_reserved, are left in the global
219 *	structure.
220 */
221u_int
222vm_meter_cnt(size_t offset)
223{
224	struct pcpu *pcpu;
225	u_int count;
226	int i;
227
228	count = *(u_int *)((char *)&vm_cnt + offset);
229	CPU_FOREACH(i) {
230		pcpu = pcpu_find(i);
231		count += *(u_int *)((char *)&pcpu->pc_cnt + offset);
232	}
233	return (count);
234}
235
236static int
237cnt_sysctl(SYSCTL_HANDLER_ARGS)
238{
239	u_int count;
240
241	count = vm_meter_cnt((char *)arg1 - (char *)&vm_cnt);
242	return (SYSCTL_OUT(req, &count, sizeof(count)));
243}
244
245SYSCTL_PROC(_vm, VM_TOTAL, vmtotal, CTLTYPE_OPAQUE|CTLFLAG_RD|CTLFLAG_MPSAFE,
246    0, sizeof(struct vmtotal), vmtotal, "S,vmtotal",
247    "System virtual memory statistics");
248SYSCTL_NODE(_vm, OID_AUTO, stats, CTLFLAG_RW, 0, "VM meter stats");
249static SYSCTL_NODE(_vm_stats, OID_AUTO, sys, CTLFLAG_RW, 0,
250	"VM meter sys stats");
251static SYSCTL_NODE(_vm_stats, OID_AUTO, vm, CTLFLAG_RW, 0,
252	"VM meter vm stats");
253SYSCTL_NODE(_vm_stats, OID_AUTO, misc, CTLFLAG_RW, 0, "VM meter misc stats");
254
255#define	VM_STATS(parent, var, descr) \
256	SYSCTL_PROC(parent, OID_AUTO, var, \
257	    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_MPSAFE, &vm_cnt.var, 0,	\
258	    cnt_sysctl, "IU", descr)
259#define	VM_STATS_VM(var, descr)		VM_STATS(_vm_stats_vm, var, descr)
260#define	VM_STATS_SYS(var, descr)	VM_STATS(_vm_stats_sys, var, descr)
261
262VM_STATS_SYS(v_swtch, "Context switches");
263VM_STATS_SYS(v_trap, "Traps");
264VM_STATS_SYS(v_syscall, "System calls");
265VM_STATS_SYS(v_intr, "Device interrupts");
266VM_STATS_SYS(v_soft, "Software interrupts");
267VM_STATS_VM(v_vm_faults, "Address memory faults");
268VM_STATS_VM(v_io_faults, "Page faults requiring I/O");
269VM_STATS_VM(v_cow_faults, "Copy-on-write faults");
270VM_STATS_VM(v_cow_optim, "Optimized COW faults");
271VM_STATS_VM(v_zfod, "Pages zero-filled on demand");
272VM_STATS_VM(v_ozfod, "Optimized zero fill pages");
273VM_STATS_VM(v_swapin, "Swap pager pageins");
274VM_STATS_VM(v_swapout, "Swap pager pageouts");
275VM_STATS_VM(v_swappgsin, "Swap pages swapped in");
276VM_STATS_VM(v_swappgsout, "Swap pages swapped out");
277VM_STATS_VM(v_vnodein, "Vnode pager pageins");
278VM_STATS_VM(v_vnodeout, "Vnode pager pageouts");
279VM_STATS_VM(v_vnodepgsin, "Vnode pages paged in");
280VM_STATS_VM(v_vnodepgsout, "Vnode pages paged out");
281VM_STATS_VM(v_intrans, "In transit page faults");
282VM_STATS_VM(v_reactivated, "Pages reactivated by pagedaemon");
283VM_STATS_VM(v_pdwakeups, "Pagedaemon wakeups");
284VM_STATS_VM(v_pdpages, "Pages analyzed by pagedaemon");
285VM_STATS_VM(v_pdshortfalls, "Page reclamation shortfalls");
286VM_STATS_VM(v_tcached, "Total pages cached");
287VM_STATS_VM(v_dfree, "Pages freed by pagedaemon");
288VM_STATS_VM(v_pfree, "Pages freed by exiting processes");
289VM_STATS_VM(v_tfree, "Total pages freed");
290VM_STATS_VM(v_page_size, "Page size in bytes");
291VM_STATS_VM(v_page_count, "Total number of pages in system");
292VM_STATS_VM(v_free_reserved, "Pages reserved for deadlock");
293VM_STATS_VM(v_free_target, "Pages desired free");
294VM_STATS_VM(v_free_min, "Minimum low-free-pages threshold");
295VM_STATS_VM(v_free_count, "Free pages");
296VM_STATS_VM(v_wire_count, "Wired pages");
297VM_STATS_VM(v_active_count, "Active pages");
298VM_STATS_VM(v_inactive_target, "Desired inactive pages");
299VM_STATS_VM(v_inactive_count, "Inactive pages");
300VM_STATS_VM(v_laundry_count, "Pages eligible for laundering");
301VM_STATS_VM(v_cache_count, "Pages on cache queue");
302VM_STATS_VM(v_pageout_free_min, "Min pages reserved for kernel");
303VM_STATS_VM(v_interrupt_free_min, "Reserved pages for interrupt code");
304VM_STATS_VM(v_forks, "Number of fork() calls");
305VM_STATS_VM(v_vforks, "Number of vfork() calls");
306VM_STATS_VM(v_rforks, "Number of rfork() calls");
307VM_STATS_VM(v_kthreads, "Number of fork() calls by kernel");
308VM_STATS_VM(v_forkpages, "VM pages affected by fork()");
309VM_STATS_VM(v_vforkpages, "VM pages affected by vfork()");
310VM_STATS_VM(v_rforkpages, "VM pages affected by rfork()");
311VM_STATS_VM(v_kthreadpages, "VM pages affected by fork() by kernel");
312
313SYSCTL_INT(_vm_stats_misc, OID_AUTO, zero_page_count, CTLFLAG_RD,
314	&vm_page_zero_count, 0, "Number of zero-ed free pages");
315