xref: /freebsd/sys/kern/subr_param.c (revision 955c8cbb4960e6cf3602de144b1b9154a5092968)
1 /*-
2  * Copyright (c) 1980, 1986, 1989, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  * (c) UNIX System Laboratories, Inc.
5  * All or some portions of this file are derived from material licensed
6  * to the University of California by American Telephone and Telegraph
7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8  * the permission of UNIX System Laboratories, Inc.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 4. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)param.c	8.3 (Berkeley) 8/20/94
35  */
36 
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
39 
40 #include "opt_param.h"
41 #include "opt_msgbuf.h"
42 #include "opt_maxusers.h"
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/kernel.h>
47 #include <sys/limits.h>
48 #include <sys/msgbuf.h>
49 #include <sys/sysctl.h>
50 #include <sys/proc.h>
51 
52 #include <vm/vm.h>
53 #include <vm/vm_param.h>
54 #include <vm/pmap.h>
55 
56 /*
57  * System parameter formulae.
58  */
59 
60 #ifndef HZ
61 #  if defined(__mips__) || defined(__arm__)
62 #    define	HZ 100
63 #  else
64 #    define	HZ 1000
65 #  endif
66 #  ifndef HZ_VM
67 #    define	HZ_VM 100
68 #  endif
69 #else
70 #  ifndef HZ_VM
71 #    define	HZ_VM HZ
72 #  endif
73 #endif
74 #define	NPROC (20 + 16 * maxusers)
75 #ifndef NBUF
76 #define NBUF 0
77 #endif
78 #ifndef MAXFILES
79 #define	MAXFILES (maxproc * 2)
80 #endif
81 
82 static int sysctl_kern_vm_guest(SYSCTL_HANDLER_ARGS);
83 
84 int	hz;
85 int	tick;
86 int	maxusers;			/* base tunable */
87 int	maxproc;			/* maximum # of processes */
88 int	maxprocperuid;			/* max # of procs per user */
89 int	maxfiles;			/* sys. wide open files limit */
90 int	maxfilesperproc;		/* per-proc open files limit */
91 int	msgbufsize;			/* size of kernel message buffer */
92 int	ncallout;			/* maximum # of timer events */
93 int	nbuf;
94 int	ngroups_max;			/* max # groups per process */
95 int	nswbuf;
96 quad_t	maxmbufmem;			/* max mbuf memory */
97 pid_t	pid_max = PID_MAX;
98 long	maxswzone;			/* max swmeta KVA storage */
99 long	maxbcache;			/* max buffer cache KVA storage */
100 long	maxpipekva;			/* Limit on pipe KVA */
101 int 	vm_guest;			/* Running as virtual machine guest? */
102 u_long	maxtsiz;			/* max text size */
103 u_long	dfldsiz;			/* initial data size limit */
104 u_long	maxdsiz;			/* max data size */
105 u_long	dflssiz;			/* initial stack size limit */
106 u_long	maxssiz;			/* max stack size */
107 u_long	sgrowsiz;			/* amount to grow stack */
108 
109 SYSCTL_INT(_kern, OID_AUTO, hz, CTLFLAG_RDTUN, &hz, 0,
110     "Number of clock ticks per second");
111 SYSCTL_INT(_kern, OID_AUTO, ncallout, CTLFLAG_RDTUN, &ncallout, 0,
112     "Number of pre-allocated timer events");
113 SYSCTL_INT(_kern, OID_AUTO, nbuf, CTLFLAG_RDTUN, &nbuf, 0,
114     "Number of buffers in the buffer cache");
115 SYSCTL_INT(_kern, OID_AUTO, nswbuf, CTLFLAG_RDTUN, &nswbuf, 0,
116     "Number of swap buffers");
117 SYSCTL_INT(_kern, OID_AUTO, msgbufsize, CTLFLAG_RDTUN, &msgbufsize, 0,
118     "Size of the kernel message buffer");
119 SYSCTL_LONG(_kern, OID_AUTO, maxswzone, CTLFLAG_RDTUN, &maxswzone, 0,
120     "Maximum memory for swap metadata");
121 SYSCTL_LONG(_kern, OID_AUTO, maxbcache, CTLFLAG_RDTUN, &maxbcache, 0,
122     "Maximum value of vfs.maxbufspace");
123 SYSCTL_ULONG(_kern, OID_AUTO, maxtsiz, CTLFLAG_RW | CTLFLAG_TUN, &maxtsiz, 0,
124     "Maximum text size");
125 SYSCTL_ULONG(_kern, OID_AUTO, dfldsiz, CTLFLAG_RW | CTLFLAG_TUN, &dfldsiz, 0,
126     "Initial data size limit");
127 SYSCTL_ULONG(_kern, OID_AUTO, maxdsiz, CTLFLAG_RW | CTLFLAG_TUN, &maxdsiz, 0,
128     "Maximum data size");
129 SYSCTL_ULONG(_kern, OID_AUTO, dflssiz, CTLFLAG_RW | CTLFLAG_TUN, &dflssiz, 0,
130     "Initial stack size limit");
131 SYSCTL_ULONG(_kern, OID_AUTO, maxssiz, CTLFLAG_RW | CTLFLAG_TUN, &maxssiz, 0,
132     "Maximum stack size");
133 SYSCTL_ULONG(_kern, OID_AUTO, sgrowsiz, CTLFLAG_RW | CTLFLAG_TUN, &sgrowsiz, 0,
134     "Amount to grow stack on a stack fault");
135 SYSCTL_PROC(_kern, OID_AUTO, vm_guest, CTLFLAG_RD | CTLTYPE_STRING,
136     NULL, 0, sysctl_kern_vm_guest, "A",
137     "Virtual machine guest detected? (none|generic|xen)");
138 
139 /*
140  * These have to be allocated somewhere; allocating
141  * them here forces loader errors if this file is omitted
142  * (if they've been externed everywhere else; hah!).
143  */
144 struct	buf *swbuf;
145 
146 /*
147  * The elements of this array are ordered based upon the values of the
148  * corresponding enum VM_GUEST members.
149  */
150 static const char *const vm_guest_sysctl_names[] = {
151 	"none",
152 	"generic",
153 	"xen",
154 	NULL
155 };
156 
157 #ifndef XEN
158 static const char *const vm_bnames[] = {
159 	"QEMU",				/* QEMU */
160 	"Plex86",			/* Plex86 */
161 	"Bochs",			/* Bochs */
162 	"Xen",				/* Xen */
163 	"BHYVE",			/* bhyve */
164 	NULL
165 };
166 
167 static const char *const vm_pnames[] = {
168 	"VMware Virtual Platform",	/* VMWare VM */
169 	"Virtual Machine",		/* Microsoft VirtualPC */
170 	"VirtualBox",			/* Sun xVM VirtualBox */
171 	"Parallels Virtual Platform",	/* Parallels VM */
172 	NULL
173 };
174 
175 
176 /*
177  * Detect known Virtual Machine hosts by inspecting the emulated BIOS.
178  */
179 static enum VM_GUEST
180 detect_virtual(void)
181 {
182 	char *sysenv;
183 	int i;
184 
185 	sysenv = getenv("smbios.bios.vendor");
186 	if (sysenv != NULL) {
187 		for (i = 0; vm_bnames[i] != NULL; i++)
188 			if (strcmp(sysenv, vm_bnames[i]) == 0) {
189 				freeenv(sysenv);
190 				return (VM_GUEST_VM);
191 			}
192 		freeenv(sysenv);
193 	}
194 	sysenv = getenv("smbios.system.product");
195 	if (sysenv != NULL) {
196 		for (i = 0; vm_pnames[i] != NULL; i++)
197 			if (strcmp(sysenv, vm_pnames[i]) == 0) {
198 				freeenv(sysenv);
199 				return (VM_GUEST_VM);
200 			}
201 		freeenv(sysenv);
202 	}
203 	return (VM_GUEST_NO);
204 }
205 #endif
206 
207 /*
208  * Boot time overrides that are not scaled against main memory
209  */
210 void
211 init_param1(void)
212 {
213 #ifndef XEN
214 	vm_guest = detect_virtual();
215 #else
216 	vm_guest = VM_GUEST_XEN;
217 #endif
218 	hz = -1;
219 	TUNABLE_INT_FETCH("kern.hz", &hz);
220 	if (hz == -1)
221 		hz = vm_guest > VM_GUEST_NO ? HZ_VM : HZ;
222 	tick = 1000000 / hz;
223 
224 #ifdef VM_SWZONE_SIZE_MAX
225 	maxswzone = VM_SWZONE_SIZE_MAX;
226 #endif
227 	TUNABLE_LONG_FETCH("kern.maxswzone", &maxswzone);
228 #ifdef VM_BCACHE_SIZE_MAX
229 	maxbcache = VM_BCACHE_SIZE_MAX;
230 #endif
231 	TUNABLE_LONG_FETCH("kern.maxbcache", &maxbcache);
232 	msgbufsize = MSGBUF_SIZE;
233 	TUNABLE_INT_FETCH("kern.msgbufsize", &msgbufsize);
234 
235 	maxtsiz = MAXTSIZ;
236 	TUNABLE_ULONG_FETCH("kern.maxtsiz", &maxtsiz);
237 	dfldsiz = DFLDSIZ;
238 	TUNABLE_ULONG_FETCH("kern.dfldsiz", &dfldsiz);
239 	maxdsiz = MAXDSIZ;
240 	TUNABLE_ULONG_FETCH("kern.maxdsiz", &maxdsiz);
241 	dflssiz = DFLSSIZ;
242 	TUNABLE_ULONG_FETCH("kern.dflssiz", &dflssiz);
243 	maxssiz = MAXSSIZ;
244 	TUNABLE_ULONG_FETCH("kern.maxssiz", &maxssiz);
245 	sgrowsiz = SGROWSIZ;
246 	TUNABLE_ULONG_FETCH("kern.sgrowsiz", &sgrowsiz);
247 
248 	/*
249 	 * Let the administrator set {NGROUPS_MAX}, but disallow values
250 	 * less than NGROUPS_MAX which would violate POSIX.1-2008 or
251 	 * greater than INT_MAX-1 which would result in overflow.
252 	 */
253 	ngroups_max = NGROUPS_MAX;
254 	TUNABLE_INT_FETCH("kern.ngroups", &ngroups_max);
255 	if (ngroups_max < NGROUPS_MAX)
256 		ngroups_max = NGROUPS_MAX;
257 
258 	/*
259 	 * Only allow to lower the maximal pid.
260 	 * Prevent setting up a non-bootable system if pid_max is too low.
261 	 */
262 	TUNABLE_INT_FETCH("kern.pid_max", &pid_max);
263 	if (pid_max > PID_MAX)
264 		pid_max = PID_MAX;
265 	else if (pid_max < 300)
266 		pid_max = 300;
267 }
268 
269 /*
270  * Boot time overrides that are scaled against main memory
271  */
272 void
273 init_param2(long physpages)
274 {
275 	quad_t realmem;
276 
277 	/* Base parameters */
278 	maxusers = MAXUSERS;
279 	TUNABLE_INT_FETCH("kern.maxusers", &maxusers);
280 	if (maxusers == 0) {
281 		maxusers = physpages / (2 * 1024 * 1024 / PAGE_SIZE);
282 		if (maxusers < 32)
283 			maxusers = 32;
284 #ifdef VM_MAX_AUTOTUNE_MAXUSERS
285                 if (maxusers > VM_MAX_AUTOTUNE_MAXUSERS)
286                         maxusers = VM_MAX_AUTOTUNE_MAXUSERS;
287 #endif
288                 /*
289                  * Scales down the function in which maxusers grows once
290                  * we hit 384.
291                  */
292                 if (maxusers > 384)
293                         maxusers = 384 + ((maxusers - 384) / 8);
294         }
295 
296 	/*
297 	 * The following can be overridden after boot via sysctl.  Note:
298 	 * unless overriden, these macros are ultimately based on maxusers.
299 	 * Limit maxproc so that kmap entries cannot be exhausted by
300 	 * processes.
301 	 */
302 	maxproc = NPROC;
303 	TUNABLE_INT_FETCH("kern.maxproc", &maxproc);
304 	if (maxproc > (physpages / 12))
305 		maxproc = physpages / 12;
306 	maxprocperuid = (maxproc * 9) / 10;
307 
308 	/*
309 	 * The default limit for maxfiles is 1/12 of the number of
310 	 * physical page but not less than 16 times maxusers.
311 	 * At most it can be 1/6 the number of physical pages.
312 	 */
313 	maxfiles = imax(MAXFILES, physpages / 8);
314 	TUNABLE_INT_FETCH("kern.maxfiles", &maxfiles);
315 	if (maxfiles > (physpages / 4))
316 		maxfiles = physpages / 4;
317 	maxfilesperproc = (maxfiles / 10) * 9;
318 
319 	/*
320 	 * Cannot be changed after boot.
321 	 */
322 	nbuf = NBUF;
323 	TUNABLE_INT_FETCH("kern.nbuf", &nbuf);
324 
325 	/*
326 	 * XXX: Does the callout wheel have to be so big?
327 	 */
328 	ncallout = 16 + maxproc + maxfiles;
329 	TUNABLE_INT_FETCH("kern.ncallout", &ncallout);
330 
331 	/*
332 	 * The default limit for all mbuf related memory is 1/2 of all
333 	 * available kernel memory (physical or kmem).
334 	 * At most it can be 3/4 of available kernel memory.
335 	 */
336 	realmem = qmin((quad_t)physpages * PAGE_SIZE,
337 	    VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS);
338 	maxmbufmem = realmem / 2;
339 	TUNABLE_QUAD_FETCH("kern.maxmbufmem", &maxmbufmem);
340 	if (maxmbufmem > (realmem / 4) * 3)
341 		maxmbufmem = (realmem / 4) * 3;
342 
343 	/*
344 	 * The default for maxpipekva is min(1/64 of the kernel address space,
345 	 * max(1/64 of main memory, 512KB)).  See sys_pipe.c for more details.
346 	 */
347 	maxpipekva = (physpages / 64) * PAGE_SIZE;
348 	TUNABLE_LONG_FETCH("kern.ipc.maxpipekva", &maxpipekva);
349 	if (maxpipekva < 512 * 1024)
350 		maxpipekva = 512 * 1024;
351 	if (maxpipekva > (VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS) / 64)
352 		maxpipekva = (VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS) /
353 		    64;
354 }
355 
356 /*
357  * Sysctl stringiying handler for kern.vm_guest.
358  */
359 static int
360 sysctl_kern_vm_guest(SYSCTL_HANDLER_ARGS)
361 {
362 	return (SYSCTL_OUT(req, vm_guest_sysctl_names[vm_guest],
363 	    strlen(vm_guest_sysctl_names[vm_guest])));
364 }
365