xref: /freebsd/sys/kern/subr_param.c (revision e17f5b1d)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1980, 1986, 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * (c) UNIX System Laboratories, Inc.
7  * All or some portions of this file are derived from material licensed
8  * to the University of California by American Telephone and Telegraph
9  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10  * the permission of UNIX System Laboratories, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	@(#)param.c	8.3 (Berkeley) 8/20/94
37  */
38 
39 #include <sys/cdefs.h>
40 __FBSDID("$FreeBSD$");
41 
42 #include "opt_param.h"
43 #include "opt_msgbuf.h"
44 #include "opt_maxusers.h"
45 
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/buf.h>
49 #include <sys/kernel.h>
50 #include <sys/limits.h>
51 #include <sys/msgbuf.h>
52 #include <sys/sysctl.h>
53 #include <sys/proc.h>
54 
55 #include <vm/vm.h>
56 #include <vm/vm_param.h>
57 #include <vm/pmap.h>
58 
59 /*
60  * System parameter formulae.
61  */
62 
63 #ifndef HZ
64 #  if defined(__mips__) || defined(__arm__)
65 #    define	HZ 100
66 #  else
67 #    define	HZ 1000
68 #  endif
69 #  ifndef HZ_VM
70 #    define	HZ_VM 100
71 #  endif
72 #else
73 #  ifndef HZ_VM
74 #    define	HZ_VM HZ
75 #  endif
76 #endif
77 #define	NPROC (20 + 16 * maxusers)
78 #ifndef NBUF
79 #define NBUF 0
80 #endif
81 #ifndef MAXFILES
82 #define	MAXFILES (40 + 32 * maxusers)
83 #endif
84 
85 static int sysctl_kern_vm_guest(SYSCTL_HANDLER_ARGS);
86 
87 int	hz;				/* system clock's frequency */
88 int	tick;				/* usec per tick (1000000 / hz) */
89 struct bintime tick_bt;			/* bintime per tick (1s / hz) */
90 sbintime_t tick_sbt;
91 int	maxusers;			/* base tunable */
92 int	maxproc;			/* maximum # of processes */
93 int	maxprocperuid;			/* max # of procs per user */
94 int	maxfiles;			/* sys. wide open files limit */
95 int	maxfilesperproc;		/* per-proc open files limit */
96 int	msgbufsize;			/* size of kernel message buffer */
97 int	nbuf;
98 int	bio_transient_maxcnt;
99 int	ngroups_max;			/* max # groups per process */
100 int	nswbuf;
101 pid_t	pid_max = PID_MAX;
102 long	maxswzone;			/* max swmeta KVA storage */
103 long	maxbcache;			/* max buffer cache KVA storage */
104 long	maxpipekva;			/* Limit on pipe KVA */
105 int	vm_guest = VM_GUEST_NO;		/* Running as virtual machine guest? */
106 u_long	maxtsiz;			/* max text size */
107 u_long	dfldsiz;			/* initial data size limit */
108 u_long	maxdsiz;			/* max data size */
109 u_long	dflssiz;			/* initial stack size limit */
110 u_long	maxssiz;			/* max stack size */
111 u_long	sgrowsiz;			/* amount to grow stack */
112 
113 SYSCTL_INT(_kern, OID_AUTO, hz, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &hz, 0,
114     "Number of clock ticks per second");
115 SYSCTL_INT(_kern, OID_AUTO, nbuf, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &nbuf, 0,
116     "Number of buffers in the buffer cache");
117 SYSCTL_INT(_kern, OID_AUTO, nswbuf, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &nswbuf, 0,
118     "Number of swap buffers");
119 SYSCTL_INT(_kern, OID_AUTO, msgbufsize, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &msgbufsize, 0,
120     "Size of the kernel message buffer");
121 SYSCTL_LONG(_kern, OID_AUTO, maxswzone, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &maxswzone, 0,
122     "Maximum memory for swap metadata");
123 SYSCTL_LONG(_kern, OID_AUTO, maxbcache, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &maxbcache, 0,
124     "Maximum value of vfs.maxbufspace");
125 SYSCTL_INT(_kern, OID_AUTO, bio_transient_maxcnt, CTLFLAG_RDTUN | CTLFLAG_NOFETCH,
126     &bio_transient_maxcnt, 0,
127     "Maximum number of transient BIOs mappings");
128 SYSCTL_ULONG(_kern, OID_AUTO, maxtsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &maxtsiz, 0,
129     "Maximum text size");
130 SYSCTL_ULONG(_kern, OID_AUTO, dfldsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &dfldsiz, 0,
131     "Initial data size limit");
132 SYSCTL_ULONG(_kern, OID_AUTO, maxdsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &maxdsiz, 0,
133     "Maximum data size");
134 SYSCTL_ULONG(_kern, OID_AUTO, dflssiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &dflssiz, 0,
135     "Initial stack size limit");
136 SYSCTL_ULONG(_kern, OID_AUTO, maxssiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &maxssiz, 0,
137     "Maximum stack size");
138 SYSCTL_ULONG(_kern, OID_AUTO, sgrowsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &sgrowsiz, 0,
139     "Amount to grow stack on a stack fault");
140 SYSCTL_PROC(_kern, OID_AUTO, vm_guest,
141     CTLFLAG_RD | CTLTYPE_STRING | CTLFLAG_MPSAFE, NULL, 0,
142     sysctl_kern_vm_guest, "A",
143     "Virtual machine guest detected?");
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 	[VM_GUEST_NO] = "none",
151 	[VM_GUEST_VM] = "generic",
152 	[VM_GUEST_XEN] = "xen",
153 	[VM_GUEST_HV] = "hv",
154 	[VM_GUEST_VMWARE] = "vmware",
155 	[VM_GUEST_KVM] = "kvm",
156 	[VM_GUEST_BHYVE] = "bhyve",
157 	[VM_GUEST_VBOX] = "vbox",
158 	[VM_GUEST_PARALLELS] = "parallels",
159 	[VM_LAST] = NULL
160 };
161 CTASSERT(nitems(vm_guest_sysctl_names) - 1 == VM_LAST);
162 
163 /*
164  * Boot time overrides that are not scaled against main memory
165  */
166 void
167 init_param1(void)
168 {
169 
170 #if !defined(__mips__) && !defined(__arm64__)
171 	TUNABLE_INT_FETCH("kern.kstack_pages", &kstack_pages);
172 #endif
173 	hz = -1;
174 	TUNABLE_INT_FETCH("kern.hz", &hz);
175 	if (hz == -1)
176 		hz = vm_guest > VM_GUEST_NO ? HZ_VM : HZ;
177 	tick = 1000000 / hz;
178 	tick_sbt = SBT_1S / hz;
179 	tick_bt = sbttobt(tick_sbt);
180 
181 	/*
182 	 * Arrange for ticks to wrap 10 minutes after boot to help catch
183 	 * sign problems sooner.
184 	 */
185 	ticks = INT_MAX - (hz * 10 * 60);
186 
187 #ifdef VM_SWZONE_SIZE_MAX
188 	maxswzone = VM_SWZONE_SIZE_MAX;
189 #endif
190 	TUNABLE_LONG_FETCH("kern.maxswzone", &maxswzone);
191 #ifdef VM_BCACHE_SIZE_MAX
192 	maxbcache = VM_BCACHE_SIZE_MAX;
193 #endif
194 	TUNABLE_LONG_FETCH("kern.maxbcache", &maxbcache);
195 	msgbufsize = MSGBUF_SIZE;
196 	TUNABLE_INT_FETCH("kern.msgbufsize", &msgbufsize);
197 
198 	maxtsiz = MAXTSIZ;
199 	TUNABLE_ULONG_FETCH("kern.maxtsiz", &maxtsiz);
200 	dfldsiz = DFLDSIZ;
201 	TUNABLE_ULONG_FETCH("kern.dfldsiz", &dfldsiz);
202 	maxdsiz = MAXDSIZ;
203 	TUNABLE_ULONG_FETCH("kern.maxdsiz", &maxdsiz);
204 	dflssiz = DFLSSIZ;
205 	TUNABLE_ULONG_FETCH("kern.dflssiz", &dflssiz);
206 	maxssiz = MAXSSIZ;
207 	TUNABLE_ULONG_FETCH("kern.maxssiz", &maxssiz);
208 	sgrowsiz = SGROWSIZ;
209 	TUNABLE_ULONG_FETCH("kern.sgrowsiz", &sgrowsiz);
210 
211 	/*
212 	 * Let the administrator set {NGROUPS_MAX}, but disallow values
213 	 * less than NGROUPS_MAX which would violate POSIX.1-2008 or
214 	 * greater than INT_MAX-1 which would result in overflow.
215 	 */
216 	ngroups_max = NGROUPS_MAX;
217 	TUNABLE_INT_FETCH("kern.ngroups", &ngroups_max);
218 	if (ngroups_max < NGROUPS_MAX)
219 		ngroups_max = NGROUPS_MAX;
220 
221 	/*
222 	 * Only allow to lower the maximal pid.
223 	 * Prevent setting up a non-bootable system if pid_max is too low.
224 	 */
225 	TUNABLE_INT_FETCH("kern.pid_max", &pid_max);
226 	if (pid_max > PID_MAX)
227 		pid_max = PID_MAX;
228 	else if (pid_max < 300)
229 		pid_max = 300;
230 
231 	TUNABLE_INT_FETCH("vfs.unmapped_buf_allowed", &unmapped_buf_allowed);
232 }
233 
234 /*
235  * Boot time overrides that are scaled against main memory
236  */
237 void
238 init_param2(long physpages)
239 {
240 
241 	/* Base parameters */
242 	maxusers = MAXUSERS;
243 	TUNABLE_INT_FETCH("kern.maxusers", &maxusers);
244 	if (maxusers == 0) {
245 		maxusers = physpages / (2 * 1024 * 1024 / PAGE_SIZE);
246 		if (maxusers < 32)
247 			maxusers = 32;
248 #ifdef VM_MAX_AUTOTUNE_MAXUSERS
249                 if (maxusers > VM_MAX_AUTOTUNE_MAXUSERS)
250                         maxusers = VM_MAX_AUTOTUNE_MAXUSERS;
251 #endif
252                 /*
253                  * Scales down the function in which maxusers grows once
254                  * we hit 384.
255                  */
256                 if (maxusers > 384)
257                         maxusers = 384 + ((maxusers - 384) / 8);
258         }
259 
260 	/*
261 	 * The following can be overridden after boot via sysctl.  Note:
262 	 * unless overriden, these macros are ultimately based on maxusers.
263 	 * Limit maxproc so that kmap entries cannot be exhausted by
264 	 * processes.
265 	 */
266 	maxproc = NPROC;
267 	TUNABLE_INT_FETCH("kern.maxproc", &maxproc);
268 	if (maxproc > (physpages / 12))
269 		maxproc = physpages / 12;
270 	if (maxproc > pid_max)
271 		maxproc = pid_max;
272 	maxprocperuid = (maxproc * 9) / 10;
273 
274 	/*
275 	 * The default limit for maxfiles is 1/12 of the number of
276 	 * physical page but not less than 16 times maxusers.
277 	 * At most it can be 1/6 the number of physical pages.
278 	 */
279 	maxfiles = imax(MAXFILES, physpages / 8);
280 	TUNABLE_INT_FETCH("kern.maxfiles", &maxfiles);
281 	if (maxfiles > (physpages / 4))
282 		maxfiles = physpages / 4;
283 	maxfilesperproc = (maxfiles / 10) * 9;
284 	TUNABLE_INT_FETCH("kern.maxfilesperproc", &maxfilesperproc);
285 
286 	/*
287 	 * Cannot be changed after boot.
288 	 */
289 	nbuf = NBUF;
290 	TUNABLE_INT_FETCH("kern.nbuf", &nbuf);
291 	TUNABLE_INT_FETCH("kern.bio_transient_maxcnt", &bio_transient_maxcnt);
292 
293 	/*
294 	 * Physical buffers are pre-allocated buffers (struct buf) that
295 	 * are used as temporary holders for I/O, such as paging I/O.
296 	 */
297 	TUNABLE_INT_FETCH("kern.nswbuf", &nswbuf);
298 
299 	/*
300 	 * The default for maxpipekva is min(1/64 of the kernel address space,
301 	 * max(1/64 of main memory, 512KB)).  See sys_pipe.c for more details.
302 	 */
303 	maxpipekva = (physpages / 64) * PAGE_SIZE;
304 	TUNABLE_LONG_FETCH("kern.ipc.maxpipekva", &maxpipekva);
305 	if (maxpipekva < 512 * 1024)
306 		maxpipekva = 512 * 1024;
307 	if (maxpipekva > (VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS) / 64)
308 		maxpipekva = (VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS) /
309 		    64;
310 }
311 
312 /*
313  * Sysctl stringifying handler for kern.vm_guest.
314  */
315 static int
316 sysctl_kern_vm_guest(SYSCTL_HANDLER_ARGS)
317 {
318 	return (SYSCTL_OUT_STR(req, vm_guest_sysctl_names[vm_guest]));
319 }
320