xref: /freebsd/sys/kern/subr_param.c (revision ec965063070e5753c166cf592c9336444b74720a)
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 #  define	HZ 1000
67 #  ifndef HZ_VM
68 #    define	HZ_VM 100
69 #  endif
70 #else
71 #  ifndef HZ_VM
72 #    define	HZ_VM HZ
73 #  endif
74 #endif
75 #define	NPROC (20 + 16 * maxusers)
76 #ifndef NBUF
77 #define NBUF 0
78 #endif
79 #ifndef MAXFILES
80 #define	MAXFILES (40 + 32 * maxusers)
81 #endif
82 
83 static int sysctl_kern_vm_guest(SYSCTL_HANDLER_ARGS);
84 
85 int	hz;				/* system clock's frequency */
86 int	tick;				/* usec per tick (1000000 / hz) */
87 time_t	tick_seconds_max;		/* max hz * seconds an integer can hold */
88 struct bintime tick_bt;			/* bintime per tick (1s / hz) */
89 sbintime_t tick_sbt;
90 int	maxusers;			/* base tunable */
91 int	maxproc;			/* maximum # of processes */
92 int	maxprocperuid;			/* max # of procs per user */
93 int	maxfiles;			/* sys. wide open files limit */
94 int	maxfilesperproc;		/* per-proc open files limit */
95 int	msgbufsize;			/* size of kernel message buffer */
96 int	nbuf;				/* number of bcache bufs */
97 int	bio_transient_maxcnt;
98 int	ngroups_max;			/* max # groups per process */
99 int	nswbuf;
100 pid_t	pid_max = PID_MAX;
101 u_long	maxswzone;			/* max swmeta KVA storage */
102 u_long	maxbcache;			/* max buffer cache KVA storage */
103 u_long	maxpipekva;			/* Limit on pipe KVA */
104 u_long	maxphys;			/* max raw I/O transfer size */
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, hz_max, CTLFLAG_RD, SYSCTL_NULL_INT_PTR, HZ_MAXIMUM,
116     "Maximum hz value supported");
117 SYSCTL_INT(_kern, OID_AUTO, hz_min, CTLFLAG_RD, SYSCTL_NULL_INT_PTR, HZ_MINIMUM,
118     "Minimum hz value supported");
119 SYSCTL_INT(_kern, OID_AUTO, nbuf, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &nbuf, 0,
120     "Number of buffers in the buffer cache");
121 SYSCTL_INT(_kern, OID_AUTO, nswbuf, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &nswbuf, 0,
122     "Number of swap buffers");
123 SYSCTL_INT(_kern, OID_AUTO, msgbufsize, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &msgbufsize, 0,
124     "Size of the kernel message buffer");
125 SYSCTL_LONG(_kern, OID_AUTO, maxswzone, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &maxswzone, 0,
126     "Maximum memory for swap metadata");
127 SYSCTL_LONG(_kern, OID_AUTO, maxbcache, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &maxbcache, 0,
128     "Maximum value of vfs.maxbufspace");
129 SYSCTL_INT(_kern, OID_AUTO, bio_transient_maxcnt, CTLFLAG_RDTUN | CTLFLAG_NOFETCH,
130     &bio_transient_maxcnt, 0,
131     "Maximum number of transient BIOs mappings");
132 SYSCTL_ULONG(_kern, OID_AUTO, maxtsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &maxtsiz, 0,
133     "Maximum text size");
134 SYSCTL_ULONG(_kern, OID_AUTO, dfldsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &dfldsiz, 0,
135     "Initial data size limit");
136 SYSCTL_ULONG(_kern, OID_AUTO, maxdsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &maxdsiz, 0,
137     "Maximum data size");
138 SYSCTL_ULONG(_kern, OID_AUTO, dflssiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &dflssiz, 0,
139     "Initial stack size limit");
140 SYSCTL_ULONG(_kern, OID_AUTO, maxssiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &maxssiz, 0,
141     "Maximum stack size");
142 SYSCTL_ULONG(_kern, OID_AUTO, sgrowsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &sgrowsiz, 0,
143     "Amount to grow stack on a stack fault");
144 SYSCTL_PROC(_kern, OID_AUTO, vm_guest,
145     CTLFLAG_RD | CTLTYPE_STRING | CTLFLAG_MPSAFE, NULL, 0,
146     sysctl_kern_vm_guest, "A",
147     "Virtual machine guest detected?");
148 
149 /*
150  * The elements of this array are ordered based upon the values of the
151  * corresponding enum VM_GUEST members.
152  */
153 static const char *const vm_guest_sysctl_names[] = {
154 	[VM_GUEST_NO] = "none",
155 	[VM_GUEST_VM] = "generic",
156 	[VM_GUEST_XEN] = "xen",
157 	[VM_GUEST_HV] = "hv",
158 	[VM_GUEST_VMWARE] = "vmware",
159 	[VM_GUEST_KVM] = "kvm",
160 	[VM_GUEST_BHYVE] = "bhyve",
161 	[VM_GUEST_VBOX] = "vbox",
162 	[VM_GUEST_PARALLELS] = "parallels",
163 	[VM_LAST] = NULL
164 };
165 CTASSERT(nitems(vm_guest_sysctl_names) - 1 == VM_LAST);
166 
167 /*
168  * Boot time overrides that are not scaled against main memory
169  */
170 void
171 init_param1(void)
172 {
173 
174 #if !defined(__mips__) && !defined(__arm64__)
175 	TUNABLE_INT_FETCH("kern.kstack_pages", &kstack_pages);
176 #endif
177 	hz = -1;
178 	TUNABLE_INT_FETCH("kern.hz", &hz);
179 	if (hz == -1)
180 		hz = vm_guest > VM_GUEST_NO ? HZ_VM : HZ;
181 
182 	/* range check the "hz" value */
183 	if (__predict_false(hz < HZ_MINIMUM))
184 		hz = HZ_MINIMUM;
185 	else if (__predict_false(hz > HZ_MAXIMUM))
186 		hz = HZ_MAXIMUM;
187 
188 	tick = 1000000 / hz;
189 	tick_sbt = SBT_1S / hz;
190 	tick_bt = sbttobt(tick_sbt);
191 	tick_seconds_max = INT_MAX / hz;
192 
193 	/*
194 	 * Arrange for ticks to wrap 10 minutes after boot to help catch
195 	 * sign problems sooner.
196 	 */
197 	ticks = INT_MAX - (hz * 10 * 60);
198 
199 	vn_lock_pair_pause_max = hz / 100;
200 	if (vn_lock_pair_pause_max == 0)
201 		vn_lock_pair_pause_max = 1;
202 
203 #ifdef VM_SWZONE_SIZE_MAX
204 	maxswzone = VM_SWZONE_SIZE_MAX;
205 #endif
206 	TUNABLE_LONG_FETCH("kern.maxswzone", &maxswzone);
207 #ifdef VM_BCACHE_SIZE_MAX
208 	maxbcache = VM_BCACHE_SIZE_MAX;
209 #endif
210 	TUNABLE_LONG_FETCH("kern.maxbcache", &maxbcache);
211 	msgbufsize = MSGBUF_SIZE;
212 	TUNABLE_INT_FETCH("kern.msgbufsize", &msgbufsize);
213 
214 	maxtsiz = MAXTSIZ;
215 	TUNABLE_ULONG_FETCH("kern.maxtsiz", &maxtsiz);
216 	dfldsiz = DFLDSIZ;
217 	TUNABLE_ULONG_FETCH("kern.dfldsiz", &dfldsiz);
218 	maxdsiz = MAXDSIZ;
219 	TUNABLE_ULONG_FETCH("kern.maxdsiz", &maxdsiz);
220 	dflssiz = DFLSSIZ;
221 	TUNABLE_ULONG_FETCH("kern.dflssiz", &dflssiz);
222 	maxssiz = MAXSSIZ;
223 	TUNABLE_ULONG_FETCH("kern.maxssiz", &maxssiz);
224 	sgrowsiz = SGROWSIZ;
225 	TUNABLE_ULONG_FETCH("kern.sgrowsiz", &sgrowsiz);
226 
227 	/*
228 	 * Let the administrator set {NGROUPS_MAX}, but disallow values
229 	 * less than NGROUPS_MAX which would violate POSIX.1-2008 or
230 	 * greater than INT_MAX-1 which would result in overflow.
231 	 */
232 	ngroups_max = NGROUPS_MAX;
233 	TUNABLE_INT_FETCH("kern.ngroups", &ngroups_max);
234 	if (ngroups_max < NGROUPS_MAX)
235 		ngroups_max = NGROUPS_MAX;
236 
237 	/*
238 	 * Only allow to lower the maximal pid.
239 	 * Prevent setting up a non-bootable system if pid_max is too low.
240 	 */
241 	TUNABLE_INT_FETCH("kern.pid_max", &pid_max);
242 	if (pid_max > PID_MAX)
243 		pid_max = PID_MAX;
244 	else if (pid_max < 300)
245 		pid_max = 300;
246 
247 	TUNABLE_INT_FETCH("vfs.unmapped_buf_allowed", &unmapped_buf_allowed);
248 }
249 
250 /*
251  * Boot time overrides that are scaled against main memory
252  */
253 void
254 init_param2(long physpages)
255 {
256 
257 	/* Base parameters */
258 	maxusers = MAXUSERS;
259 	TUNABLE_INT_FETCH("kern.maxusers", &maxusers);
260 	if (maxusers == 0) {
261 		maxusers = physpages / (2 * 1024 * 1024 / PAGE_SIZE);
262 		if (maxusers < 32)
263 			maxusers = 32;
264 #ifdef VM_MAX_AUTOTUNE_MAXUSERS
265                 if (maxusers > VM_MAX_AUTOTUNE_MAXUSERS)
266                         maxusers = VM_MAX_AUTOTUNE_MAXUSERS;
267 #endif
268                 /*
269                  * Scales down the function in which maxusers grows once
270                  * we hit 384.
271                  */
272                 if (maxusers > 384)
273                         maxusers = 384 + ((maxusers - 384) / 8);
274         }
275 
276 	/*
277 	 * The following can be overridden after boot via sysctl.  Note:
278 	 * unless overridden, these macros are ultimately based on maxusers.
279 	 * Limit maxproc so that kmap entries cannot be exhausted by
280 	 * processes.
281 	 */
282 	maxproc = NPROC;
283 	TUNABLE_INT_FETCH("kern.maxproc", &maxproc);
284 	if (maxproc > (physpages / 12))
285 		maxproc = physpages / 12;
286 	if (maxproc > pid_max)
287 		maxproc = pid_max;
288 	maxprocperuid = (maxproc * 9) / 10;
289 
290 	/*
291 	 * The default limit for maxfiles is 1/12 of the number of
292 	 * physical page but not less than 16 times maxusers.
293 	 * At most it can be 1/6 the number of physical pages.
294 	 */
295 	maxfiles = imax(MAXFILES, physpages / 8);
296 	TUNABLE_INT_FETCH("kern.maxfiles", &maxfiles);
297 	if (maxfiles > (physpages / 4))
298 		maxfiles = physpages / 4;
299 	maxfilesperproc = (maxfiles / 10) * 9;
300 	TUNABLE_INT_FETCH("kern.maxfilesperproc", &maxfilesperproc);
301 
302 	/*
303 	 * Cannot be changed after boot.
304 	 */
305 	nbuf = NBUF;
306 	TUNABLE_INT_FETCH("kern.nbuf", &nbuf);
307 	TUNABLE_INT_FETCH("kern.bio_transient_maxcnt", &bio_transient_maxcnt);
308 	maxphys = MAXPHYS;
309 	TUNABLE_ULONG_FETCH("kern.maxphys", &maxphys);
310 	if (maxphys == 0) {
311 		maxphys = MAXPHYS;
312 	} else if (__bitcountl(maxphys) != 1) {	/* power of two */
313 		if (flsl(maxphys) == NBBY * sizeof(maxphys))
314 			maxphys = MAXPHYS;
315 		else
316 			maxphys = 1UL << flsl(maxphys);
317 	}
318 	if (maxphys < PAGE_SIZE)
319 		maxphys = MAXPHYS;
320 
321 	/*
322 	 * Physical buffers are pre-allocated buffers (struct buf) that
323 	 * are used as temporary holders for I/O, such as paging I/O.
324 	 */
325 	TUNABLE_INT_FETCH("kern.nswbuf", &nswbuf);
326 
327 	/*
328 	 * The default for maxpipekva is min(1/64 of the kernel address space,
329 	 * max(1/64 of main memory, 512KB)).  See sys_pipe.c for more details.
330 	 */
331 	maxpipekva = ptoa(physpages / 64);
332 	TUNABLE_LONG_FETCH("kern.ipc.maxpipekva", &maxpipekva);
333 	if (maxpipekva < 512 * 1024)
334 		maxpipekva = 512 * 1024;
335 	if (maxpipekva > (VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS) / 64)
336 		maxpipekva = (VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS) /
337 		    64;
338 }
339 
340 /*
341  * Sysctl stringifying handler for kern.vm_guest.
342  */
343 static int
344 sysctl_kern_vm_guest(SYSCTL_HANDLER_ARGS)
345 {
346 	return (SYSCTL_OUT_STR(req, vm_guest_sysctl_names[vm_guest]));
347 }
348