xref: /freebsd/sys/kern/subr_param.c (revision 89e3d5671ba13dceca272d5b159c9bd805f3f504)
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_maxphys.h"
45 #include "opt_maxusers.h"
46 
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/buf.h>
50 #include <sys/kernel.h>
51 #include <sys/limits.h>
52 #include <sys/msgbuf.h>
53 #include <sys/sysctl.h>
54 #include <sys/proc.h>
55 #include <sys/vnode.h>
56 
57 #include <vm/vm.h>
58 #include <vm/vm_param.h>
59 #include <vm/pmap.h>
60 
61 /*
62  * System parameter formulae.
63  */
64 
65 #ifndef HZ
66 #  if defined(__mips__) || defined(__arm__)
67 #    define	HZ 100
68 #  else
69 #    define	HZ 1000
70 #  endif
71 #  ifndef HZ_VM
72 #    define	HZ_VM 100
73 #  endif
74 #else
75 #  ifndef HZ_VM
76 #    define	HZ_VM HZ
77 #  endif
78 #endif
79 #define	NPROC (20 + 16 * maxusers)
80 #ifndef NBUF
81 #define NBUF 0
82 #endif
83 #ifndef MAXFILES
84 #define	MAXFILES (40 + 32 * maxusers)
85 #endif
86 
87 static int sysctl_kern_vm_guest(SYSCTL_HANDLER_ARGS);
88 
89 int	hz;				/* system clock's frequency */
90 int	tick;				/* usec per tick (1000000 / hz) */
91 struct bintime tick_bt;			/* bintime per tick (1s / hz) */
92 sbintime_t tick_sbt;
93 int	maxusers;			/* base tunable */
94 int	maxproc;			/* maximum # of processes */
95 int	maxprocperuid;			/* max # of procs per user */
96 int	maxfiles;			/* sys. wide open files limit */
97 int	maxfilesperproc;		/* per-proc open files limit */
98 int	msgbufsize;			/* size of kernel message buffer */
99 int	nbuf;				/* number of bcache bufs */
100 int	bio_transient_maxcnt;
101 int	ngroups_max;			/* max # groups per process */
102 int	nswbuf;
103 pid_t	pid_max = PID_MAX;
104 u_long	maxswzone;			/* max swmeta KVA storage */
105 u_long	maxbcache;			/* max buffer cache KVA storage */
106 u_long	maxpipekva;			/* Limit on pipe KVA */
107 u_long	maxphys;			/* max raw I/O transfer size */
108 int	vm_guest = VM_GUEST_NO;		/* Running as virtual machine guest? */
109 u_long	maxtsiz;			/* max text size */
110 u_long	dfldsiz;			/* initial data size limit */
111 u_long	maxdsiz;			/* max data size */
112 u_long	dflssiz;			/* initial stack size limit */
113 u_long	maxssiz;			/* max stack size */
114 u_long	sgrowsiz;			/* amount to grow stack */
115 
116 SYSCTL_INT(_kern, OID_AUTO, hz, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &hz, 0,
117     "Number of clock ticks per second");
118 SYSCTL_INT(_kern, OID_AUTO, nbuf, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &nbuf, 0,
119     "Number of buffers in the buffer cache");
120 SYSCTL_INT(_kern, OID_AUTO, nswbuf, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &nswbuf, 0,
121     "Number of swap buffers");
122 SYSCTL_INT(_kern, OID_AUTO, msgbufsize, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &msgbufsize, 0,
123     "Size of the kernel message buffer");
124 SYSCTL_LONG(_kern, OID_AUTO, maxswzone, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &maxswzone, 0,
125     "Maximum memory for swap metadata");
126 SYSCTL_LONG(_kern, OID_AUTO, maxbcache, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &maxbcache, 0,
127     "Maximum value of vfs.maxbufspace");
128 SYSCTL_INT(_kern, OID_AUTO, bio_transient_maxcnt, CTLFLAG_RDTUN | CTLFLAG_NOFETCH,
129     &bio_transient_maxcnt, 0,
130     "Maximum number of transient BIOs mappings");
131 SYSCTL_ULONG(_kern, OID_AUTO, maxtsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &maxtsiz, 0,
132     "Maximum text size");
133 SYSCTL_ULONG(_kern, OID_AUTO, dfldsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &dfldsiz, 0,
134     "Initial data size limit");
135 SYSCTL_ULONG(_kern, OID_AUTO, maxdsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &maxdsiz, 0,
136     "Maximum data size");
137 SYSCTL_ULONG(_kern, OID_AUTO, dflssiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &dflssiz, 0,
138     "Initial stack size limit");
139 SYSCTL_ULONG(_kern, OID_AUTO, maxssiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &maxssiz, 0,
140     "Maximum stack size");
141 SYSCTL_ULONG(_kern, OID_AUTO, sgrowsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &sgrowsiz, 0,
142     "Amount to grow stack on a stack fault");
143 SYSCTL_PROC(_kern, OID_AUTO, vm_guest,
144     CTLFLAG_RD | CTLTYPE_STRING | CTLFLAG_MPSAFE, NULL, 0,
145     sysctl_kern_vm_guest, "A",
146     "Virtual machine guest detected?");
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 	[VM_GUEST_NO] = "none",
154 	[VM_GUEST_VM] = "generic",
155 	[VM_GUEST_XEN] = "xen",
156 	[VM_GUEST_HV] = "hv",
157 	[VM_GUEST_VMWARE] = "vmware",
158 	[VM_GUEST_KVM] = "kvm",
159 	[VM_GUEST_BHYVE] = "bhyve",
160 	[VM_GUEST_VBOX] = "vbox",
161 	[VM_GUEST_PARALLELS] = "parallels",
162 	[VM_LAST] = NULL
163 };
164 CTASSERT(nitems(vm_guest_sysctl_names) - 1 == VM_LAST);
165 
166 /*
167  * Boot time overrides that are not scaled against main memory
168  */
169 void
170 init_param1(void)
171 {
172 
173 #if !defined(__mips__) && !defined(__arm64__)
174 	TUNABLE_INT_FETCH("kern.kstack_pages", &kstack_pages);
175 #endif
176 	hz = -1;
177 	TUNABLE_INT_FETCH("kern.hz", &hz);
178 	if (hz == -1)
179 		hz = vm_guest > VM_GUEST_NO ? HZ_VM : HZ;
180 	tick = 1000000 / hz;
181 	tick_sbt = SBT_1S / hz;
182 	tick_bt = sbttobt(tick_sbt);
183 
184 	/*
185 	 * Arrange for ticks to wrap 10 minutes after boot to help catch
186 	 * sign problems sooner.
187 	 */
188 	ticks = INT_MAX - (hz * 10 * 60);
189 
190 	vn_lock_pair_pause_max = hz / 100;
191 	if (vn_lock_pair_pause_max == 0)
192 		vn_lock_pair_pause_max = 1;
193 
194 #ifdef VM_SWZONE_SIZE_MAX
195 	maxswzone = VM_SWZONE_SIZE_MAX;
196 #endif
197 	TUNABLE_LONG_FETCH("kern.maxswzone", &maxswzone);
198 #ifdef VM_BCACHE_SIZE_MAX
199 	maxbcache = VM_BCACHE_SIZE_MAX;
200 #endif
201 	TUNABLE_LONG_FETCH("kern.maxbcache", &maxbcache);
202 	msgbufsize = MSGBUF_SIZE;
203 	TUNABLE_INT_FETCH("kern.msgbufsize", &msgbufsize);
204 
205 	maxtsiz = MAXTSIZ;
206 	TUNABLE_ULONG_FETCH("kern.maxtsiz", &maxtsiz);
207 	dfldsiz = DFLDSIZ;
208 	TUNABLE_ULONG_FETCH("kern.dfldsiz", &dfldsiz);
209 	maxdsiz = MAXDSIZ;
210 	TUNABLE_ULONG_FETCH("kern.maxdsiz", &maxdsiz);
211 	dflssiz = DFLSSIZ;
212 	TUNABLE_ULONG_FETCH("kern.dflssiz", &dflssiz);
213 	maxssiz = MAXSSIZ;
214 	TUNABLE_ULONG_FETCH("kern.maxssiz", &maxssiz);
215 	sgrowsiz = SGROWSIZ;
216 	TUNABLE_ULONG_FETCH("kern.sgrowsiz", &sgrowsiz);
217 
218 	/*
219 	 * Let the administrator set {NGROUPS_MAX}, but disallow values
220 	 * less than NGROUPS_MAX which would violate POSIX.1-2008 or
221 	 * greater than INT_MAX-1 which would result in overflow.
222 	 */
223 	ngroups_max = NGROUPS_MAX;
224 	TUNABLE_INT_FETCH("kern.ngroups", &ngroups_max);
225 	if (ngroups_max < NGROUPS_MAX)
226 		ngroups_max = NGROUPS_MAX;
227 
228 	/*
229 	 * Only allow to lower the maximal pid.
230 	 * Prevent setting up a non-bootable system if pid_max is too low.
231 	 */
232 	TUNABLE_INT_FETCH("kern.pid_max", &pid_max);
233 	if (pid_max > PID_MAX)
234 		pid_max = PID_MAX;
235 	else if (pid_max < 300)
236 		pid_max = 300;
237 
238 	TUNABLE_INT_FETCH("vfs.unmapped_buf_allowed", &unmapped_buf_allowed);
239 }
240 
241 /*
242  * Boot time overrides that are scaled against main memory
243  */
244 void
245 init_param2(long physpages)
246 {
247 
248 	/* Base parameters */
249 	maxusers = MAXUSERS;
250 	TUNABLE_INT_FETCH("kern.maxusers", &maxusers);
251 	if (maxusers == 0) {
252 		maxusers = physpages / (2 * 1024 * 1024 / PAGE_SIZE);
253 		if (maxusers < 32)
254 			maxusers = 32;
255 #ifdef VM_MAX_AUTOTUNE_MAXUSERS
256                 if (maxusers > VM_MAX_AUTOTUNE_MAXUSERS)
257                         maxusers = VM_MAX_AUTOTUNE_MAXUSERS;
258 #endif
259                 /*
260                  * Scales down the function in which maxusers grows once
261                  * we hit 384.
262                  */
263                 if (maxusers > 384)
264                         maxusers = 384 + ((maxusers - 384) / 8);
265         }
266 
267 	/*
268 	 * The following can be overridden after boot via sysctl.  Note:
269 	 * unless overriden, these macros are ultimately based on maxusers.
270 	 * Limit maxproc so that kmap entries cannot be exhausted by
271 	 * processes.
272 	 */
273 	maxproc = NPROC;
274 	TUNABLE_INT_FETCH("kern.maxproc", &maxproc);
275 	if (maxproc > (physpages / 12))
276 		maxproc = physpages / 12;
277 	if (maxproc > pid_max)
278 		maxproc = pid_max;
279 	maxprocperuid = (maxproc * 9) / 10;
280 
281 	/*
282 	 * The default limit for maxfiles is 1/12 of the number of
283 	 * physical page but not less than 16 times maxusers.
284 	 * At most it can be 1/6 the number of physical pages.
285 	 */
286 	maxfiles = imax(MAXFILES, physpages / 8);
287 	TUNABLE_INT_FETCH("kern.maxfiles", &maxfiles);
288 	if (maxfiles > (physpages / 4))
289 		maxfiles = physpages / 4;
290 	maxfilesperproc = (maxfiles / 10) * 9;
291 	TUNABLE_INT_FETCH("kern.maxfilesperproc", &maxfilesperproc);
292 
293 	/*
294 	 * Cannot be changed after boot.
295 	 */
296 	nbuf = NBUF;
297 	TUNABLE_INT_FETCH("kern.nbuf", &nbuf);
298 	TUNABLE_INT_FETCH("kern.bio_transient_maxcnt", &bio_transient_maxcnt);
299 	maxphys = MAXPHYS;
300 	TUNABLE_ULONG_FETCH("kern.maxphys", &maxphys);
301 	if (maxphys == 0) {
302 		maxphys = MAXPHYS;
303 	} else if (__bitcountl(maxphys) != 1) {	/* power of two */
304 		if (flsl(maxphys) == NBBY * sizeof(maxphys))
305 			maxphys = MAXPHYS;
306 		else
307 			maxphys = 1UL << flsl(maxphys);
308 	}
309 	if (maxphys < PAGE_SIZE)
310 		maxphys = MAXPHYS;
311 
312 	/*
313 	 * Physical buffers are pre-allocated buffers (struct buf) that
314 	 * are used as temporary holders for I/O, such as paging I/O.
315 	 */
316 	TUNABLE_INT_FETCH("kern.nswbuf", &nswbuf);
317 
318 	/*
319 	 * The default for maxpipekva is min(1/64 of the kernel address space,
320 	 * max(1/64 of main memory, 512KB)).  See sys_pipe.c for more details.
321 	 */
322 	maxpipekva = ptoa(physpages / 64);
323 	TUNABLE_LONG_FETCH("kern.ipc.maxpipekva", &maxpipekva);
324 	if (maxpipekva < 512 * 1024)
325 		maxpipekva = 512 * 1024;
326 	if (maxpipekva > (VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS) / 64)
327 		maxpipekva = (VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS) /
328 		    64;
329 }
330 
331 /*
332  * Sysctl stringifying handler for kern.vm_guest.
333  */
334 static int
335 sysctl_kern_vm_guest(SYSCTL_HANDLER_ARGS)
336 {
337 	return (SYSCTL_OUT_STR(req, vm_guest_sysctl_names[vm_guest]));
338 }
339