xref: /freebsd/sys/kern/subr_param.c (revision f05cddf9)
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;				/* system clock's frequency */
85 int	tick;				/* usec per tick (1000000 / hz) */
86 struct bintime tick_bt;			/* bintime per tick (1s / hz) */
87 sbintime_t tick_sbt;
88 int	maxusers;			/* base tunable */
89 int	maxproc;			/* maximum # of processes */
90 int	maxprocperuid;			/* max # of procs per user */
91 int	maxfiles;			/* sys. wide open files limit */
92 int	maxfilesperproc;		/* per-proc open files limit */
93 int	msgbufsize;			/* size of kernel message buffer */
94 int	nbuf;
95 int	bio_transient_maxcnt;
96 int	ngroups_max;			/* max # groups per process */
97 int	nswbuf;
98 pid_t	pid_max = PID_MAX;
99 long	maxswzone;			/* max swmeta KVA storage */
100 long	maxbcache;			/* max buffer cache KVA storage */
101 long	maxpipekva;			/* Limit on pipe KVA */
102 int 	vm_guest;			/* Running as virtual machine guest? */
103 u_long	maxtsiz;			/* max text size */
104 u_long	dfldsiz;			/* initial data size limit */
105 u_long	maxdsiz;			/* max data size */
106 u_long	dflssiz;			/* initial stack size limit */
107 u_long	maxssiz;			/* max stack size */
108 u_long	sgrowsiz;			/* amount to grow stack */
109 
110 SYSCTL_INT(_kern, OID_AUTO, hz, CTLFLAG_RDTUN, &hz, 0,
111     "Number of clock ticks per second");
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_INT(_kern, OID_AUTO, bio_transient_maxcnt, CTLFLAG_RDTUN,
123     &bio_transient_maxcnt, 0,
124     "Maximum number of transient BIOs mappings");
125 SYSCTL_ULONG(_kern, OID_AUTO, maxtsiz, CTLFLAG_RW | CTLFLAG_TUN, &maxtsiz, 0,
126     "Maximum text size");
127 SYSCTL_ULONG(_kern, OID_AUTO, dfldsiz, CTLFLAG_RW | CTLFLAG_TUN, &dfldsiz, 0,
128     "Initial data size limit");
129 SYSCTL_ULONG(_kern, OID_AUTO, maxdsiz, CTLFLAG_RW | CTLFLAG_TUN, &maxdsiz, 0,
130     "Maximum data size");
131 SYSCTL_ULONG(_kern, OID_AUTO, dflssiz, CTLFLAG_RW | CTLFLAG_TUN, &dflssiz, 0,
132     "Initial stack size limit");
133 SYSCTL_ULONG(_kern, OID_AUTO, maxssiz, CTLFLAG_RW | CTLFLAG_TUN, &maxssiz, 0,
134     "Maximum stack size");
135 SYSCTL_ULONG(_kern, OID_AUTO, sgrowsiz, CTLFLAG_RW | CTLFLAG_TUN, &sgrowsiz, 0,
136     "Amount to grow stack on a stack fault");
137 SYSCTL_PROC(_kern, OID_AUTO, vm_guest, CTLFLAG_RD | CTLTYPE_STRING,
138     NULL, 0, sysctl_kern_vm_guest, "A",
139     "Virtual machine guest detected? (none|generic|xen)");
140 
141 /*
142  * These have to be allocated somewhere; allocating
143  * them here forces loader errors if this file is omitted
144  * (if they've been externed everywhere else; hah!).
145  */
146 struct	buf *swbuf;
147 
148 /*
149  * The elements of this array are ordered based upon the values of the
150  * corresponding enum VM_GUEST members.
151  */
152 static const char *const vm_guest_sysctl_names[] = {
153 	"none",
154 	"generic",
155 	"xen",
156 	NULL
157 };
158 
159 #ifndef XEN
160 static const char *const vm_bnames[] = {
161 	"QEMU",				/* QEMU */
162 	"Plex86",			/* Plex86 */
163 	"Bochs",			/* Bochs */
164 	"Xen",				/* Xen */
165 	"BHYVE",			/* bhyve */
166 	"Seabios",			/* KVM */
167 	NULL
168 };
169 
170 static const char *const vm_pnames[] = {
171 	"VMware Virtual Platform",	/* VMWare VM */
172 	"Virtual Machine",		/* Microsoft VirtualPC */
173 	"VirtualBox",			/* Sun xVM VirtualBox */
174 	"Parallels Virtual Platform",	/* Parallels VM */
175 	"KVM",				/* KVM */
176 	NULL
177 };
178 
179 
180 /*
181  * Detect known Virtual Machine hosts by inspecting the emulated BIOS.
182  */
183 static enum VM_GUEST
184 detect_virtual(void)
185 {
186 	char *sysenv;
187 	int i;
188 
189 	sysenv = getenv("smbios.bios.vendor");
190 	if (sysenv != NULL) {
191 		for (i = 0; vm_bnames[i] != NULL; i++)
192 			if (strcmp(sysenv, vm_bnames[i]) == 0) {
193 				freeenv(sysenv);
194 				return (VM_GUEST_VM);
195 			}
196 		freeenv(sysenv);
197 	}
198 	sysenv = getenv("smbios.system.product");
199 	if (sysenv != NULL) {
200 		for (i = 0; vm_pnames[i] != NULL; i++)
201 			if (strcmp(sysenv, vm_pnames[i]) == 0) {
202 				freeenv(sysenv);
203 				return (VM_GUEST_VM);
204 			}
205 		freeenv(sysenv);
206 	}
207 	return (VM_GUEST_NO);
208 }
209 #endif
210 
211 /*
212  * Boot time overrides that are not scaled against main memory
213  */
214 void
215 init_param1(void)
216 {
217 #ifndef XEN
218 	vm_guest = detect_virtual();
219 #else
220 	vm_guest = VM_GUEST_XEN;
221 #endif
222 	hz = -1;
223 	TUNABLE_INT_FETCH("kern.hz", &hz);
224 	if (hz == -1)
225 		hz = vm_guest > VM_GUEST_NO ? HZ_VM : HZ;
226 	tick = 1000000 / hz;
227 	tick_sbt = SBT_1S / hz;
228 	tick_bt = sbttobt(tick_sbt);
229 
230 #ifdef VM_SWZONE_SIZE_MAX
231 	maxswzone = VM_SWZONE_SIZE_MAX;
232 #endif
233 	TUNABLE_LONG_FETCH("kern.maxswzone", &maxswzone);
234 #ifdef VM_BCACHE_SIZE_MAX
235 	maxbcache = VM_BCACHE_SIZE_MAX;
236 #endif
237 	TUNABLE_LONG_FETCH("kern.maxbcache", &maxbcache);
238 	msgbufsize = MSGBUF_SIZE;
239 	TUNABLE_INT_FETCH("kern.msgbufsize", &msgbufsize);
240 
241 	maxtsiz = MAXTSIZ;
242 	TUNABLE_ULONG_FETCH("kern.maxtsiz", &maxtsiz);
243 	dfldsiz = DFLDSIZ;
244 	TUNABLE_ULONG_FETCH("kern.dfldsiz", &dfldsiz);
245 	maxdsiz = MAXDSIZ;
246 	TUNABLE_ULONG_FETCH("kern.maxdsiz", &maxdsiz);
247 	dflssiz = DFLSSIZ;
248 	TUNABLE_ULONG_FETCH("kern.dflssiz", &dflssiz);
249 	maxssiz = MAXSSIZ;
250 	TUNABLE_ULONG_FETCH("kern.maxssiz", &maxssiz);
251 	sgrowsiz = SGROWSIZ;
252 	TUNABLE_ULONG_FETCH("kern.sgrowsiz", &sgrowsiz);
253 
254 	/*
255 	 * Let the administrator set {NGROUPS_MAX}, but disallow values
256 	 * less than NGROUPS_MAX which would violate POSIX.1-2008 or
257 	 * greater than INT_MAX-1 which would result in overflow.
258 	 */
259 	ngroups_max = NGROUPS_MAX;
260 	TUNABLE_INT_FETCH("kern.ngroups", &ngroups_max);
261 	if (ngroups_max < NGROUPS_MAX)
262 		ngroups_max = NGROUPS_MAX;
263 
264 	/*
265 	 * Only allow to lower the maximal pid.
266 	 * Prevent setting up a non-bootable system if pid_max is too low.
267 	 */
268 	TUNABLE_INT_FETCH("kern.pid_max", &pid_max);
269 	if (pid_max > PID_MAX)
270 		pid_max = PID_MAX;
271 	else if (pid_max < 300)
272 		pid_max = 300;
273 
274 	TUNABLE_INT_FETCH("vfs.unmapped_buf_allowed", &unmapped_buf_allowed);
275 }
276 
277 /*
278  * Boot time overrides that are scaled against main memory
279  */
280 void
281 init_param2(long physpages)
282 {
283 
284 	/* Base parameters */
285 	maxusers = MAXUSERS;
286 	TUNABLE_INT_FETCH("kern.maxusers", &maxusers);
287 	if (maxusers == 0) {
288 		maxusers = physpages / (2 * 1024 * 1024 / PAGE_SIZE);
289 		if (maxusers < 32)
290 			maxusers = 32;
291 #ifdef VM_MAX_AUTOTUNE_MAXUSERS
292                 if (maxusers > VM_MAX_AUTOTUNE_MAXUSERS)
293                         maxusers = VM_MAX_AUTOTUNE_MAXUSERS;
294 #endif
295                 /*
296                  * Scales down the function in which maxusers grows once
297                  * we hit 384.
298                  */
299                 if (maxusers > 384)
300                         maxusers = 384 + ((maxusers - 384) / 8);
301         }
302 
303 	/*
304 	 * The following can be overridden after boot via sysctl.  Note:
305 	 * unless overriden, these macros are ultimately based on maxusers.
306 	 * Limit maxproc so that kmap entries cannot be exhausted by
307 	 * processes.
308 	 */
309 	maxproc = NPROC;
310 	TUNABLE_INT_FETCH("kern.maxproc", &maxproc);
311 	if (maxproc > (physpages / 12))
312 		maxproc = physpages / 12;
313 	maxprocperuid = (maxproc * 9) / 10;
314 
315 	/*
316 	 * The default limit for maxfiles is 1/12 of the number of
317 	 * physical page but not less than 16 times maxusers.
318 	 * At most it can be 1/6 the number of physical pages.
319 	 */
320 	maxfiles = imax(MAXFILES, physpages / 8);
321 	TUNABLE_INT_FETCH("kern.maxfiles", &maxfiles);
322 	if (maxfiles > (physpages / 4))
323 		maxfiles = physpages / 4;
324 	maxfilesperproc = (maxfiles / 10) * 9;
325 
326 	/*
327 	 * Cannot be changed after boot.
328 	 */
329 	nbuf = NBUF;
330 	TUNABLE_INT_FETCH("kern.nbuf", &nbuf);
331 	TUNABLE_INT_FETCH("kern.bio_transient_maxcnt", &bio_transient_maxcnt);
332 
333 	/*
334 	 * The default for maxpipekva is min(1/64 of the kernel address space,
335 	 * max(1/64 of main memory, 512KB)).  See sys_pipe.c for more details.
336 	 */
337 	maxpipekva = (physpages / 64) * PAGE_SIZE;
338 	TUNABLE_LONG_FETCH("kern.ipc.maxpipekva", &maxpipekva);
339 	if (maxpipekva < 512 * 1024)
340 		maxpipekva = 512 * 1024;
341 	if (maxpipekva > (VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS) / 64)
342 		maxpipekva = (VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS) /
343 		    64;
344 }
345 
346 /*
347  * Sysctl stringiying handler for kern.vm_guest.
348  */
349 static int
350 sysctl_kern_vm_guest(SYSCTL_HANDLER_ARGS)
351 {
352 	return (SYSCTL_OUT(req, vm_guest_sysctl_names[vm_guest],
353 	    strlen(vm_guest_sysctl_names[vm_guest])));
354 }
355