xref: /freebsd/sys/kern/subr_param.c (revision a36179937e7e768d1840896ac1ea13ba6f118a09)
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 #include <sys/vnode.h>
55 
56 #include <vm/vm.h>
57 #include <vm/vm_param.h>
58 #include <vm/pmap.h>
59 
60 /*
61  * System parameter formulae.
62  */
63 
64 #ifndef HZ
65 #  if defined(__mips__) || defined(__arm__)
66 #    define	HZ 100
67 #  else
68 #    define	HZ 1000
69 #  endif
70 #  ifndef HZ_VM
71 #    define	HZ_VM 100
72 #  endif
73 #else
74 #  ifndef HZ_VM
75 #    define	HZ_VM HZ
76 #  endif
77 #endif
78 #define	NPROC (20 + 16 * maxusers)
79 #ifndef NBUF
80 #define NBUF 0
81 #endif
82 #ifndef MAXFILES
83 #define	MAXFILES (40 + 32 * maxusers)
84 #endif
85 
86 static int sysctl_kern_vm_guest(SYSCTL_HANDLER_ARGS);
87 
88 int	hz;				/* system clock's frequency */
89 int	tick;				/* usec per tick (1000000 / hz) */
90 struct bintime tick_bt;			/* bintime per tick (1s / hz) */
91 sbintime_t tick_sbt;
92 int	maxusers;			/* base tunable */
93 int	maxproc;			/* maximum # of processes */
94 int	maxprocperuid;			/* max # of procs per user */
95 int	maxfiles;			/* sys. wide open files limit */
96 int	maxfilesperproc;		/* per-proc open files limit */
97 int	msgbufsize;			/* size of kernel message buffer */
98 int	nbuf;
99 int	bio_transient_maxcnt;
100 int	ngroups_max;			/* max # groups per process */
101 int	nswbuf;
102 pid_t	pid_max = PID_MAX;
103 long	maxswzone;			/* max swmeta KVA storage */
104 long	maxbcache;			/* max buffer cache KVA storage */
105 long	maxpipekva;			/* Limit on pipe KVA */
106 int	vm_guest = VM_GUEST_NO;		/* Running as virtual machine guest? */
107 u_long	maxtsiz;			/* max text size */
108 u_long	dfldsiz;			/* initial data size limit */
109 u_long	maxdsiz;			/* max data size */
110 u_long	dflssiz;			/* initial stack size limit */
111 u_long	maxssiz;			/* max stack size */
112 u_long	sgrowsiz;			/* amount to grow stack */
113 
114 SYSCTL_INT(_kern, OID_AUTO, hz, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &hz, 0,
115     "Number of clock ticks per second");
116 SYSCTL_INT(_kern, OID_AUTO, nbuf, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &nbuf, 0,
117     "Number of buffers in the buffer cache");
118 SYSCTL_INT(_kern, OID_AUTO, nswbuf, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &nswbuf, 0,
119     "Number of swap buffers");
120 SYSCTL_INT(_kern, OID_AUTO, msgbufsize, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &msgbufsize, 0,
121     "Size of the kernel message buffer");
122 SYSCTL_LONG(_kern, OID_AUTO, maxswzone, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &maxswzone, 0,
123     "Maximum memory for swap metadata");
124 SYSCTL_LONG(_kern, OID_AUTO, maxbcache, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &maxbcache, 0,
125     "Maximum value of vfs.maxbufspace");
126 SYSCTL_INT(_kern, OID_AUTO, bio_transient_maxcnt, CTLFLAG_RDTUN | CTLFLAG_NOFETCH,
127     &bio_transient_maxcnt, 0,
128     "Maximum number of transient BIOs mappings");
129 SYSCTL_ULONG(_kern, OID_AUTO, maxtsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &maxtsiz, 0,
130     "Maximum text size");
131 SYSCTL_ULONG(_kern, OID_AUTO, dfldsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &dfldsiz, 0,
132     "Initial data size limit");
133 SYSCTL_ULONG(_kern, OID_AUTO, maxdsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &maxdsiz, 0,
134     "Maximum data size");
135 SYSCTL_ULONG(_kern, OID_AUTO, dflssiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &dflssiz, 0,
136     "Initial stack size limit");
137 SYSCTL_ULONG(_kern, OID_AUTO, maxssiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &maxssiz, 0,
138     "Maximum stack size");
139 SYSCTL_ULONG(_kern, OID_AUTO, sgrowsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &sgrowsiz, 0,
140     "Amount to grow stack on a stack fault");
141 SYSCTL_PROC(_kern, OID_AUTO, vm_guest,
142     CTLFLAG_RD | CTLTYPE_STRING | CTLFLAG_MPSAFE, NULL, 0,
143     sysctl_kern_vm_guest, "A",
144     "Virtual machine guest detected?");
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 	[VM_GUEST_NO] = "none",
152 	[VM_GUEST_VM] = "generic",
153 	[VM_GUEST_XEN] = "xen",
154 	[VM_GUEST_HV] = "hv",
155 	[VM_GUEST_VMWARE] = "vmware",
156 	[VM_GUEST_KVM] = "kvm",
157 	[VM_GUEST_BHYVE] = "bhyve",
158 	[VM_GUEST_VBOX] = "vbox",
159 	[VM_GUEST_PARALLELS] = "parallels",
160 	[VM_LAST] = NULL
161 };
162 CTASSERT(nitems(vm_guest_sysctl_names) - 1 == VM_LAST);
163 
164 /*
165  * Boot time overrides that are not scaled against main memory
166  */
167 void
168 init_param1(void)
169 {
170 
171 #if !defined(__mips__) && !defined(__arm64__)
172 	TUNABLE_INT_FETCH("kern.kstack_pages", &kstack_pages);
173 #endif
174 	hz = -1;
175 	TUNABLE_INT_FETCH("kern.hz", &hz);
176 	if (hz == -1)
177 		hz = vm_guest > VM_GUEST_NO ? HZ_VM : HZ;
178 	tick = 1000000 / hz;
179 	tick_sbt = SBT_1S / hz;
180 	tick_bt = sbttobt(tick_sbt);
181 
182 	/*
183 	 * Arrange for ticks to wrap 10 minutes after boot to help catch
184 	 * sign problems sooner.
185 	 */
186 	ticks = INT_MAX - (hz * 10 * 60);
187 
188 	vn_lock_pair_pause_max = hz / 100;
189 	if (vn_lock_pair_pause_max == 0)
190 		vn_lock_pair_pause_max = 1;
191 
192 #ifdef VM_SWZONE_SIZE_MAX
193 	maxswzone = VM_SWZONE_SIZE_MAX;
194 #endif
195 	TUNABLE_LONG_FETCH("kern.maxswzone", &maxswzone);
196 #ifdef VM_BCACHE_SIZE_MAX
197 	maxbcache = VM_BCACHE_SIZE_MAX;
198 #endif
199 	TUNABLE_LONG_FETCH("kern.maxbcache", &maxbcache);
200 	msgbufsize = MSGBUF_SIZE;
201 	TUNABLE_INT_FETCH("kern.msgbufsize", &msgbufsize);
202 
203 	maxtsiz = MAXTSIZ;
204 	TUNABLE_ULONG_FETCH("kern.maxtsiz", &maxtsiz);
205 	dfldsiz = DFLDSIZ;
206 	TUNABLE_ULONG_FETCH("kern.dfldsiz", &dfldsiz);
207 	maxdsiz = MAXDSIZ;
208 	TUNABLE_ULONG_FETCH("kern.maxdsiz", &maxdsiz);
209 	dflssiz = DFLSSIZ;
210 	TUNABLE_ULONG_FETCH("kern.dflssiz", &dflssiz);
211 	maxssiz = MAXSSIZ;
212 	TUNABLE_ULONG_FETCH("kern.maxssiz", &maxssiz);
213 	sgrowsiz = SGROWSIZ;
214 	TUNABLE_ULONG_FETCH("kern.sgrowsiz", &sgrowsiz);
215 
216 	/*
217 	 * Let the administrator set {NGROUPS_MAX}, but disallow values
218 	 * less than NGROUPS_MAX which would violate POSIX.1-2008 or
219 	 * greater than INT_MAX-1 which would result in overflow.
220 	 */
221 	ngroups_max = NGROUPS_MAX;
222 	TUNABLE_INT_FETCH("kern.ngroups", &ngroups_max);
223 	if (ngroups_max < NGROUPS_MAX)
224 		ngroups_max = NGROUPS_MAX;
225 
226 	/*
227 	 * Only allow to lower the maximal pid.
228 	 * Prevent setting up a non-bootable system if pid_max is too low.
229 	 */
230 	TUNABLE_INT_FETCH("kern.pid_max", &pid_max);
231 	if (pid_max > PID_MAX)
232 		pid_max = PID_MAX;
233 	else if (pid_max < 300)
234 		pid_max = 300;
235 
236 	TUNABLE_INT_FETCH("vfs.unmapped_buf_allowed", &unmapped_buf_allowed);
237 }
238 
239 /*
240  * Boot time overrides that are scaled against main memory
241  */
242 void
243 init_param2(long physpages)
244 {
245 
246 	/* Base parameters */
247 	maxusers = MAXUSERS;
248 	TUNABLE_INT_FETCH("kern.maxusers", &maxusers);
249 	if (maxusers == 0) {
250 		maxusers = physpages / (2 * 1024 * 1024 / PAGE_SIZE);
251 		if (maxusers < 32)
252 			maxusers = 32;
253 #ifdef VM_MAX_AUTOTUNE_MAXUSERS
254                 if (maxusers > VM_MAX_AUTOTUNE_MAXUSERS)
255                         maxusers = VM_MAX_AUTOTUNE_MAXUSERS;
256 #endif
257                 /*
258                  * Scales down the function in which maxusers grows once
259                  * we hit 384.
260                  */
261                 if (maxusers > 384)
262                         maxusers = 384 + ((maxusers - 384) / 8);
263         }
264 
265 	/*
266 	 * The following can be overridden after boot via sysctl.  Note:
267 	 * unless overriden, these macros are ultimately based on maxusers.
268 	 * Limit maxproc so that kmap entries cannot be exhausted by
269 	 * processes.
270 	 */
271 	maxproc = NPROC;
272 	TUNABLE_INT_FETCH("kern.maxproc", &maxproc);
273 	if (maxproc > (physpages / 12))
274 		maxproc = physpages / 12;
275 	if (maxproc > pid_max)
276 		maxproc = pid_max;
277 	maxprocperuid = (maxproc * 9) / 10;
278 
279 	/*
280 	 * The default limit for maxfiles is 1/12 of the number of
281 	 * physical page but not less than 16 times maxusers.
282 	 * At most it can be 1/6 the number of physical pages.
283 	 */
284 	maxfiles = imax(MAXFILES, physpages / 8);
285 	TUNABLE_INT_FETCH("kern.maxfiles", &maxfiles);
286 	if (maxfiles > (physpages / 4))
287 		maxfiles = physpages / 4;
288 	maxfilesperproc = (maxfiles / 10) * 9;
289 	TUNABLE_INT_FETCH("kern.maxfilesperproc", &maxfilesperproc);
290 
291 	/*
292 	 * Cannot be changed after boot.
293 	 */
294 	nbuf = NBUF;
295 	TUNABLE_INT_FETCH("kern.nbuf", &nbuf);
296 	TUNABLE_INT_FETCH("kern.bio_transient_maxcnt", &bio_transient_maxcnt);
297 
298 	/*
299 	 * Physical buffers are pre-allocated buffers (struct buf) that
300 	 * are used as temporary holders for I/O, such as paging I/O.
301 	 */
302 	TUNABLE_INT_FETCH("kern.nswbuf", &nswbuf);
303 
304 	/*
305 	 * The default for maxpipekva is min(1/64 of the kernel address space,
306 	 * max(1/64 of main memory, 512KB)).  See sys_pipe.c for more details.
307 	 */
308 	maxpipekva = (physpages / 64) * PAGE_SIZE;
309 	TUNABLE_LONG_FETCH("kern.ipc.maxpipekva", &maxpipekva);
310 	if (maxpipekva < 512 * 1024)
311 		maxpipekva = 512 * 1024;
312 	if (maxpipekva > (VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS) / 64)
313 		maxpipekva = (VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS) /
314 		    64;
315 }
316 
317 /*
318  * Sysctl stringifying handler for kern.vm_guest.
319  */
320 static int
321 sysctl_kern_vm_guest(SYSCTL_HANDLER_ARGS)
322 {
323 	return (SYSCTL_OUT_STR(req, vm_guest_sysctl_names[vm_guest]));
324 }
325