xref: /freebsd/sys/kern/subr_param.c (revision 8ec81e392654c03ebba7792e47bf3744d623cf54)
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 pid_t	pid_max = PID_MAX;
97 long	maxswzone;			/* max swmeta KVA storage */
98 long	maxbcache;			/* max buffer cache KVA storage */
99 long	maxpipekva;			/* Limit on pipe KVA */
100 int 	vm_guest;			/* Running as virtual machine guest? */
101 u_long	maxtsiz;			/* max text size */
102 u_long	dfldsiz;			/* initial data size limit */
103 u_long	maxdsiz;			/* max data size */
104 u_long	dflssiz;			/* initial stack size limit */
105 u_long	maxssiz;			/* max stack size */
106 u_long	sgrowsiz;			/* amount to grow stack */
107 
108 SYSCTL_INT(_kern, OID_AUTO, hz, CTLFLAG_RDTUN, &hz, 0,
109     "Number of clock ticks per second");
110 SYSCTL_INT(_kern, OID_AUTO, ncallout, CTLFLAG_RDTUN, &ncallout, 0,
111     "Number of pre-allocated timer events");
112 SYSCTL_INT(_kern, OID_AUTO, nbuf, CTLFLAG_RDTUN, &nbuf, 0,
113     "Number of buffers in the buffer cache");
114 SYSCTL_INT(_kern, OID_AUTO, nswbuf, CTLFLAG_RDTUN, &nswbuf, 0,
115     "Number of swap buffers");
116 SYSCTL_INT(_kern, OID_AUTO, msgbufsize, CTLFLAG_RDTUN, &msgbufsize, 0,
117     "Size of the kernel message buffer");
118 SYSCTL_LONG(_kern, OID_AUTO, maxswzone, CTLFLAG_RDTUN, &maxswzone, 0,
119     "Maximum memory for swap metadata");
120 SYSCTL_LONG(_kern, OID_AUTO, maxbcache, CTLFLAG_RDTUN, &maxbcache, 0,
121     "Maximum value of vfs.maxbufspace");
122 SYSCTL_ULONG(_kern, OID_AUTO, maxtsiz, CTLFLAG_RW | CTLFLAG_TUN, &maxtsiz, 0,
123     "Maximum text size");
124 SYSCTL_ULONG(_kern, OID_AUTO, dfldsiz, CTLFLAG_RW | CTLFLAG_TUN, &dfldsiz, 0,
125     "Initial data size limit");
126 SYSCTL_ULONG(_kern, OID_AUTO, maxdsiz, CTLFLAG_RW | CTLFLAG_TUN, &maxdsiz, 0,
127     "Maximum data size");
128 SYSCTL_ULONG(_kern, OID_AUTO, dflssiz, CTLFLAG_RW | CTLFLAG_TUN, &dflssiz, 0,
129     "Initial stack size limit");
130 SYSCTL_ULONG(_kern, OID_AUTO, maxssiz, CTLFLAG_RW | CTLFLAG_TUN, &maxssiz, 0,
131     "Maximum stack size");
132 SYSCTL_ULONG(_kern, OID_AUTO, sgrowsiz, CTLFLAG_RW | CTLFLAG_TUN, &sgrowsiz, 0,
133     "Amount to grow stack on a stack fault");
134 SYSCTL_PROC(_kern, OID_AUTO, vm_guest, CTLFLAG_RD | CTLTYPE_STRING,
135     NULL, 0, sysctl_kern_vm_guest, "A",
136     "Virtual machine guest detected? (none|generic|xen)");
137 
138 /*
139  * These have to be allocated somewhere; allocating
140  * them here forces loader errors if this file is omitted
141  * (if they've been externed everywhere else; hah!).
142  */
143 struct	buf *swbuf;
144 
145 /*
146  * The elements of this array are ordered based upon the values of the
147  * corresponding enum VM_GUEST members.
148  */
149 static const char *const vm_guest_sysctl_names[] = {
150 	"none",
151 	"generic",
152 	"xen",
153 	NULL
154 };
155 
156 #ifndef XEN
157 static const char *const vm_bnames[] = {
158 	"QEMU",				/* QEMU */
159 	"Plex86",			/* Plex86 */
160 	"Bochs",			/* Bochs */
161 	"Xen",				/* Xen */
162 	"BHYVE",			/* bhyve */
163 	"Seabios",			/* KVM */
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 	"KVM",				/* KVM */
173 	NULL
174 };
175 
176 
177 /*
178  * Detect known Virtual Machine hosts by inspecting the emulated BIOS.
179  */
180 static enum VM_GUEST
181 detect_virtual(void)
182 {
183 	char *sysenv;
184 	int i;
185 
186 	sysenv = getenv("smbios.bios.vendor");
187 	if (sysenv != NULL) {
188 		for (i = 0; vm_bnames[i] != NULL; i++)
189 			if (strcmp(sysenv, vm_bnames[i]) == 0) {
190 				freeenv(sysenv);
191 				return (VM_GUEST_VM);
192 			}
193 		freeenv(sysenv);
194 	}
195 	sysenv = getenv("smbios.system.product");
196 	if (sysenv != NULL) {
197 		for (i = 0; vm_pnames[i] != NULL; i++)
198 			if (strcmp(sysenv, vm_pnames[i]) == 0) {
199 				freeenv(sysenv);
200 				return (VM_GUEST_VM);
201 			}
202 		freeenv(sysenv);
203 	}
204 	return (VM_GUEST_NO);
205 }
206 #endif
207 
208 /*
209  * Boot time overrides that are not scaled against main memory
210  */
211 void
212 init_param1(void)
213 {
214 #ifndef XEN
215 	vm_guest = detect_virtual();
216 #else
217 	vm_guest = VM_GUEST_XEN;
218 #endif
219 	hz = -1;
220 	TUNABLE_INT_FETCH("kern.hz", &hz);
221 	if (hz == -1)
222 		hz = vm_guest > VM_GUEST_NO ? HZ_VM : HZ;
223 	tick = 1000000 / hz;
224 
225 #ifdef VM_SWZONE_SIZE_MAX
226 	maxswzone = VM_SWZONE_SIZE_MAX;
227 #endif
228 	TUNABLE_LONG_FETCH("kern.maxswzone", &maxswzone);
229 #ifdef VM_BCACHE_SIZE_MAX
230 	maxbcache = VM_BCACHE_SIZE_MAX;
231 #endif
232 	TUNABLE_LONG_FETCH("kern.maxbcache", &maxbcache);
233 	msgbufsize = MSGBUF_SIZE;
234 	TUNABLE_INT_FETCH("kern.msgbufsize", &msgbufsize);
235 
236 	maxtsiz = MAXTSIZ;
237 	TUNABLE_ULONG_FETCH("kern.maxtsiz", &maxtsiz);
238 	dfldsiz = DFLDSIZ;
239 	TUNABLE_ULONG_FETCH("kern.dfldsiz", &dfldsiz);
240 	maxdsiz = MAXDSIZ;
241 	TUNABLE_ULONG_FETCH("kern.maxdsiz", &maxdsiz);
242 	dflssiz = DFLSSIZ;
243 	TUNABLE_ULONG_FETCH("kern.dflssiz", &dflssiz);
244 	maxssiz = MAXSSIZ;
245 	TUNABLE_ULONG_FETCH("kern.maxssiz", &maxssiz);
246 	sgrowsiz = SGROWSIZ;
247 	TUNABLE_ULONG_FETCH("kern.sgrowsiz", &sgrowsiz);
248 
249 	/*
250 	 * Let the administrator set {NGROUPS_MAX}, but disallow values
251 	 * less than NGROUPS_MAX which would violate POSIX.1-2008 or
252 	 * greater than INT_MAX-1 which would result in overflow.
253 	 */
254 	ngroups_max = NGROUPS_MAX;
255 	TUNABLE_INT_FETCH("kern.ngroups", &ngroups_max);
256 	if (ngroups_max < NGROUPS_MAX)
257 		ngroups_max = NGROUPS_MAX;
258 
259 	/*
260 	 * Only allow to lower the maximal pid.
261 	 * Prevent setting up a non-bootable system if pid_max is too low.
262 	 */
263 	TUNABLE_INT_FETCH("kern.pid_max", &pid_max);
264 	if (pid_max > PID_MAX)
265 		pid_max = PID_MAX;
266 	else if (pid_max < 300)
267 		pid_max = 300;
268 }
269 
270 /*
271  * Boot time overrides that are scaled against main memory
272  */
273 void
274 init_param2(long physpages)
275 {
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 	 * Clip callout to result of previous function of maxusers maximum
329 	 * 384.  This is still huge, but acceptable.
330 	 */
331 	ncallout = imin(16 + maxproc + maxfiles, 18508);
332 	TUNABLE_INT_FETCH("kern.ncallout", &ncallout);
333 
334 	/*
335 	 * The default for maxpipekva is min(1/64 of the kernel address space,
336 	 * max(1/64 of main memory, 512KB)).  See sys_pipe.c for more details.
337 	 */
338 	maxpipekva = (physpages / 64) * PAGE_SIZE;
339 	TUNABLE_LONG_FETCH("kern.ipc.maxpipekva", &maxpipekva);
340 	if (maxpipekva < 512 * 1024)
341 		maxpipekva = 512 * 1024;
342 	if (maxpipekva > (VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS) / 64)
343 		maxpipekva = (VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS) /
344 		    64;
345 }
346 
347 /*
348  * Sysctl stringiying handler for kern.vm_guest.
349  */
350 static int
351 sysctl_kern_vm_guest(SYSCTL_HANDLER_ARGS)
352 {
353 	return (SYSCTL_OUT(req, vm_guest_sysctl_names[vm_guest],
354 	    strlen(vm_guest_sysctl_names[vm_guest])));
355 }
356