xref: /freebsd/sys/vm/vm_meter.c (revision 70e0bbedef95258a4dadc996d641a9bebd3f107d)
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 SYSCTL_UINT(_vm, VM_V_FREE_MIN, v_free_min,
58 	CTLFLAG_RW, &cnt.v_free_min, 0, "Minimum low-free-pages threshold");
59 SYSCTL_UINT(_vm, VM_V_FREE_TARGET, v_free_target,
60 	CTLFLAG_RW, &cnt.v_free_target, 0, "Desired free pages");
61 SYSCTL_UINT(_vm, VM_V_FREE_RESERVED, v_free_reserved,
62 	CTLFLAG_RW, &cnt.v_free_reserved, 0, "Pages reserved for deadlock");
63 SYSCTL_UINT(_vm, VM_V_INACTIVE_TARGET, v_inactive_target,
64 	CTLFLAG_RW, &cnt.v_inactive_target, 0, "Pages desired inactive");
65 SYSCTL_UINT(_vm, VM_V_CACHE_MIN, v_cache_min,
66 	CTLFLAG_RW, &cnt.v_cache_min, 0, "Min pages on cache queue");
67 SYSCTL_UINT(_vm, VM_V_CACHE_MAX, v_cache_max,
68 	CTLFLAG_RW, &cnt.v_cache_max, 0, "Max pages on cache queue");
69 SYSCTL_UINT(_vm, VM_V_PAGEOUT_FREE_MIN, v_pageout_free_min,
70 	CTLFLAG_RW, &cnt.v_pageout_free_min, 0, "Min pages reserved for kernel");
71 SYSCTL_UINT(_vm, OID_AUTO, v_free_severe,
72 	CTLFLAG_RW, &cnt.v_free_severe, 0, "Severe page depletion point");
73 
74 static int
75 sysctl_vm_loadavg(SYSCTL_HANDLER_ARGS)
76 {
77 
78 #ifdef SCTL_MASK32
79 	u_int32_t la[4];
80 
81 	if (req->flags & SCTL_MASK32) {
82 		la[0] = averunnable.ldavg[0];
83 		la[1] = averunnable.ldavg[1];
84 		la[2] = averunnable.ldavg[2];
85 		la[3] = averunnable.fscale;
86 		return SYSCTL_OUT(req, la, sizeof(la));
87 	} else
88 #endif
89 		return SYSCTL_OUT(req, &averunnable, sizeof(averunnable));
90 }
91 SYSCTL_PROC(_vm, VM_LOADAVG, loadavg, CTLTYPE_STRUCT | CTLFLAG_RD |
92     CTLFLAG_MPSAFE, NULL, 0, sysctl_vm_loadavg, "S,loadavg",
93     "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 	mtx_lock(&vm_object_list_mtx);
112 	TAILQ_FOREACH(object, &vm_object_list, object_list) {
113 		if (!VM_OBJECT_TRYLOCK(object)) {
114 			/*
115 			 * Avoid a lock-order reversal.  Consequently,
116 			 * the reported number of active pages may be
117 			 * greater than the actual number.
118 			 */
119 			continue;
120 		}
121 		vm_object_clear_flag(object, OBJ_ACTIVE);
122 		VM_OBJECT_UNLOCK(object);
123 	}
124 	mtx_unlock(&vm_object_list_mtx);
125 	/*
126 	 * Calculate process statistics.
127 	 */
128 	sx_slock(&allproc_lock);
129 	FOREACH_PROC_IN_SYSTEM(p) {
130 		if (p->p_flag & P_SYSTEM)
131 			continue;
132 		PROC_LOCK(p);
133 		switch (p->p_state) {
134 		case PRS_NEW:
135 			PROC_UNLOCK(p);
136 			continue;
137 			break;
138 		default:
139 			FOREACH_THREAD_IN_PROC(p, td) {
140 				thread_lock(td);
141 				switch (td->td_state) {
142 				case TDS_INHIBITED:
143 					if (TD_IS_SWAPPED(td))
144 						total.t_sw++;
145 					else if (TD_IS_SLEEPING(td) &&
146 					    td->td_priority <= PZERO)
147 						total.t_dw++;
148 					else
149 						total.t_sl++;
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 		 * Note active objects.
169 		 */
170 		paging = 0;
171 		vm = vmspace_acquire_ref(p);
172 		if (vm == NULL)
173 			continue;
174 		map = &vm->vm_map;
175 		vm_map_lock_read(map);
176 		for (entry = map->header.next;
177 		    entry != &map->header; entry = entry->next) {
178 			if ((entry->eflags & MAP_ENTRY_IS_SUB_MAP) ||
179 			    (object = entry->object.vm_object) == NULL)
180 				continue;
181 			VM_OBJECT_LOCK(object);
182 			vm_object_set_flag(object, OBJ_ACTIVE);
183 			paging |= object->paging_in_progress;
184 			VM_OBJECT_UNLOCK(object);
185 		}
186 		vm_map_unlock_read(map);
187 		vmspace_free(vm);
188 		if (paging)
189 			total.t_pw++;
190 	}
191 	sx_sunlock(&allproc_lock);
192 	/*
193 	 * Calculate object memory usage statistics.
194 	 */
195 	mtx_lock(&vm_object_list_mtx);
196 	TAILQ_FOREACH(object, &vm_object_list, object_list) {
197 		/*
198 		 * Perform unsynchronized reads on the object to avoid
199 		 * a lock-order reversal.  In this case, the lack of
200 		 * synchronization should not impair the accuracy of
201 		 * the reported statistics.
202 		 */
203 		if (object->type == OBJT_DEVICE || object->type == OBJT_SG) {
204 			/*
205 			 * Devices, like /dev/mem, will badly skew our totals.
206 			 */
207 			continue;
208 		}
209 		if (object->ref_count == 0) {
210 			/*
211 			 * Also skip unreferenced objects, including
212 			 * vnodes representing mounted file systems.
213 			 */
214 			continue;
215 		}
216 		total.t_vm += object->size;
217 		total.t_rm += object->resident_page_count;
218 		if (object->flags & OBJ_ACTIVE) {
219 			total.t_avm += object->size;
220 			total.t_arm += object->resident_page_count;
221 		}
222 		if (object->shadow_count > 1) {
223 			/* shared object */
224 			total.t_vmshr += object->size;
225 			total.t_rmshr += object->resident_page_count;
226 			if (object->flags & OBJ_ACTIVE) {
227 				total.t_avmshr += object->size;
228 				total.t_armshr += object->resident_page_count;
229 			}
230 		}
231 	}
232 	mtx_unlock(&vm_object_list_mtx);
233 	total.t_free = cnt.v_free_count + cnt.v_cache_count;
234 	return (sysctl_handle_opaque(oidp, &total, sizeof(total), req));
235 }
236 
237 /*
238  * vcnt() -	accumulate statistics from all cpus and the global cnt
239  *		structure.
240  *
241  *	The vmmeter structure is now per-cpu as well as global.  Those
242  *	statistics which can be kept on a per-cpu basis (to avoid cache
243  *	stalls between cpus) can be moved to the per-cpu vmmeter.  Remaining
244  *	statistics, such as v_free_reserved, are left in the global
245  *	structure.
246  *
247  * (sysctl_oid *oidp, void *arg1, int arg2, struct sysctl_req *req)
248  */
249 static int
250 vcnt(SYSCTL_HANDLER_ARGS)
251 {
252 	int count = *(int *)arg1;
253 	int offset = (char *)arg1 - (char *)&cnt;
254 	int i;
255 
256 	CPU_FOREACH(i) {
257 		struct pcpu *pcpu = pcpu_find(i);
258 		count += *(int *)((char *)&pcpu->pc_cnt + offset);
259 	}
260 	return (SYSCTL_OUT(req, &count, sizeof(int)));
261 }
262 
263 SYSCTL_PROC(_vm, VM_TOTAL, vmtotal, CTLTYPE_OPAQUE|CTLFLAG_RD|CTLFLAG_MPSAFE,
264     0, sizeof(struct vmtotal), vmtotal, "S,vmtotal",
265     "System virtual memory statistics");
266 SYSCTL_NODE(_vm, OID_AUTO, stats, CTLFLAG_RW, 0, "VM meter stats");
267 static SYSCTL_NODE(_vm_stats, OID_AUTO, sys, CTLFLAG_RW, 0,
268 	"VM meter sys stats");
269 static SYSCTL_NODE(_vm_stats, OID_AUTO, vm, CTLFLAG_RW, 0,
270 	"VM meter vm stats");
271 SYSCTL_NODE(_vm_stats, OID_AUTO, misc, CTLFLAG_RW, 0, "VM meter misc stats");
272 
273 #define	VM_STATS(parent, var, descr) \
274 	SYSCTL_PROC(parent, OID_AUTO, var, \
275 	    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_MPSAFE, &cnt.var, 0, vcnt, \
276 	    "IU", descr)
277 #define	VM_STATS_VM(var, descr)		VM_STATS(_vm_stats_vm, var, descr)
278 #define	VM_STATS_SYS(var, descr)	VM_STATS(_vm_stats_sys, var, descr)
279 
280 VM_STATS_SYS(v_swtch, "Context switches");
281 VM_STATS_SYS(v_trap, "Traps");
282 VM_STATS_SYS(v_syscall, "System calls");
283 VM_STATS_SYS(v_intr, "Device interrupts");
284 VM_STATS_SYS(v_soft, "Software interrupts");
285 VM_STATS_VM(v_vm_faults, "Address memory faults");
286 VM_STATS_VM(v_cow_faults, "Copy-on-write faults");
287 VM_STATS_VM(v_cow_optim, "Optimized COW faults");
288 VM_STATS_VM(v_zfod, "Pages zero-filled on demand");
289 VM_STATS_VM(v_ozfod, "Optimized zero fill pages");
290 VM_STATS_VM(v_swapin, "Swap pager pageins");
291 VM_STATS_VM(v_swapout, "Swap pager pageouts");
292 VM_STATS_VM(v_swappgsin, "Swap pages swapped in");
293 VM_STATS_VM(v_swappgsout, "Swap pages swapped out");
294 VM_STATS_VM(v_vnodein, "Vnode pager pageins");
295 VM_STATS_VM(v_vnodeout, "Vnode pager pageouts");
296 VM_STATS_VM(v_vnodepgsin, "Vnode pages paged in");
297 VM_STATS_VM(v_vnodepgsout, "Vnode pages paged out");
298 VM_STATS_VM(v_intrans, "In transit page faults");
299 VM_STATS_VM(v_reactivated, "Pages reactivated from free list");
300 VM_STATS_VM(v_pdwakeups, "Pagedaemon wakeups");
301 VM_STATS_VM(v_pdpages, "Pages analyzed by pagedaemon");
302 VM_STATS_VM(v_tcached, "Total pages cached");
303 VM_STATS_VM(v_dfree, "Pages freed by pagedaemon");
304 VM_STATS_VM(v_pfree, "Pages freed by exiting processes");
305 VM_STATS_VM(v_tfree, "Total pages freed");
306 VM_STATS_VM(v_page_size, "Page size in bytes");
307 VM_STATS_VM(v_page_count, "Total number of pages in system");
308 VM_STATS_VM(v_free_reserved, "Pages reserved for deadlock");
309 VM_STATS_VM(v_free_target, "Pages desired free");
310 VM_STATS_VM(v_free_min, "Minimum low-free-pages threshold");
311 VM_STATS_VM(v_free_count, "Free pages");
312 VM_STATS_VM(v_wire_count, "Wired pages");
313 VM_STATS_VM(v_active_count, "Active pages");
314 VM_STATS_VM(v_inactive_target, "Desired inactive pages");
315 VM_STATS_VM(v_inactive_count, "Inactive pages");
316 VM_STATS_VM(v_cache_count, "Pages on cache queue");
317 VM_STATS_VM(v_cache_min, "Min pages on cache queue");
318 VM_STATS_VM(v_cache_max, "Max pages on cached queue");
319 VM_STATS_VM(v_pageout_free_min, "Min pages reserved for kernel");
320 VM_STATS_VM(v_interrupt_free_min, "Reserved pages for interrupt code");
321 VM_STATS_VM(v_forks, "Number of fork() calls");
322 VM_STATS_VM(v_vforks, "Number of vfork() calls");
323 VM_STATS_VM(v_rforks, "Number of rfork() calls");
324 VM_STATS_VM(v_kthreads, "Number of fork() calls by kernel");
325 VM_STATS_VM(v_forkpages, "VM pages affected by fork()");
326 VM_STATS_VM(v_vforkpages, "VM pages affected by vfork()");
327 VM_STATS_VM(v_rforkpages, "VM pages affected by rfork()");
328 VM_STATS_VM(v_kthreadpages, "VM pages affected by fork() by kernel");
329 
330 SYSCTL_INT(_vm_stats_misc, OID_AUTO, zero_page_count, CTLFLAG_RD,
331 	&vm_page_zero_count, 0, "Number of zero-ed free pages");
332