xref: /freebsd/sys/kern/subr_param.c (revision 2008043f386721d58158e37e0d7e50df8095942d)
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 #include "opt_param.h"
41 #include "opt_msgbuf.h"
42 #include "opt_maxphys.h"
43 #include "opt_maxusers.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/buf.h>
48 #include <sys/kernel.h>
49 #include <sys/limits.h>
50 #include <sys/msgbuf.h>
51 #include <sys/sysctl.h>
52 #include <sys/proc.h>
53 #include <sys/vnode.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 #  define	HZ 1000
65 #  ifndef HZ_VM
66 #    define	HZ_VM 100
67 #  endif
68 #else
69 #  ifndef HZ_VM
70 #    define	HZ_VM HZ
71 #  endif
72 #endif
73 #define	NPROC (20 + 16 * maxusers)
74 #ifndef NBUF
75 #define NBUF 0
76 #endif
77 #ifndef MAXFILES
78 #define	MAXFILES (40 + 32 * maxusers)
79 #endif
80 
81 static int sysctl_kern_vm_guest(SYSCTL_HANDLER_ARGS);
82 
83 int	hz;				/* system clock's frequency */
84 int	tick;				/* usec per tick (1000000 / hz) */
85 time_t	tick_seconds_max;		/* max hz * seconds an integer can hold */
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;				/* number of bcache bufs */
95 int	bio_transient_maxcnt;
96 int	ngroups_max;			/* max # groups per process */
97 int	nswbuf;
98 pid_t	pid_max = PID_MAX;
99 u_long	maxswzone;			/* max swmeta KVA storage */
100 u_long	maxbcache;			/* max buffer cache KVA storage */
101 u_long	maxpipekva;			/* Limit on pipe KVA */
102 u_long	maxphys;			/* max raw I/O transfer size */
103 int	vm_guest = VM_GUEST_NO;		/* Running as virtual machine guest? */
104 u_long	maxtsiz;			/* max text size */
105 u_long	dfldsiz;			/* initial data size limit */
106 u_long	maxdsiz;			/* max data size */
107 u_long	dflssiz;			/* initial stack size limit */
108 u_long	maxssiz;			/* max stack size */
109 u_long	sgrowsiz;			/* amount to grow stack */
110 
111 SYSCTL_INT(_kern, OID_AUTO, hz, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &hz, 0,
112     "Number of clock ticks per second");
113 SYSCTL_INT(_kern, OID_AUTO, hz_max, CTLFLAG_RD, SYSCTL_NULL_INT_PTR, HZ_MAXIMUM,
114     "Maximum hz value supported");
115 SYSCTL_INT(_kern, OID_AUTO, hz_min, CTLFLAG_RD, SYSCTL_NULL_INT_PTR, HZ_MINIMUM,
116     "Minimum hz value supported");
117 SYSCTL_INT(_kern, OID_AUTO, nbuf, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &nbuf, 0,
118     "Number of buffers in the buffer cache");
119 SYSCTL_INT(_kern, OID_AUTO, nswbuf, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &nswbuf, 0,
120     "Number of swap buffers");
121 SYSCTL_INT(_kern, OID_AUTO, msgbufsize, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &msgbufsize, 0,
122     "Size of the kernel message buffer");
123 SYSCTL_LONG(_kern, OID_AUTO, maxswzone, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &maxswzone, 0,
124     "Maximum memory for swap metadata");
125 SYSCTL_LONG(_kern, OID_AUTO, maxbcache, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &maxbcache, 0,
126     "Maximum value of vfs.maxbufspace");
127 SYSCTL_INT(_kern, OID_AUTO, bio_transient_maxcnt, CTLFLAG_RDTUN | CTLFLAG_NOFETCH,
128     &bio_transient_maxcnt, 0,
129     "Maximum number of transient BIOs mappings");
130 SYSCTL_ULONG(_kern, OID_AUTO, maxtsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &maxtsiz, 0,
131     "Maximum text size");
132 SYSCTL_ULONG(_kern, OID_AUTO, dfldsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &dfldsiz, 0,
133     "Initial data size limit");
134 SYSCTL_ULONG(_kern, OID_AUTO, maxdsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &maxdsiz, 0,
135     "Maximum data size");
136 SYSCTL_ULONG(_kern, OID_AUTO, dflssiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &dflssiz, 0,
137     "Initial stack size limit");
138 SYSCTL_ULONG(_kern, OID_AUTO, maxssiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &maxssiz, 0,
139     "Maximum stack size");
140 SYSCTL_ULONG(_kern, OID_AUTO, sgrowsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &sgrowsiz, 0,
141     "Amount to grow stack on a stack fault");
142 SYSCTL_PROC(_kern, OID_AUTO, vm_guest,
143     CTLFLAG_RD | CTLTYPE_STRING | CTLFLAG_MPSAFE, NULL, 0,
144     sysctl_kern_vm_guest, "A",
145     "Virtual machine guest detected?");
146 
147 /*
148  * The elements of this array are ordered based upon the values of the
149  * corresponding enum VM_GUEST members.
150  */
151 static const char *const vm_guest_sysctl_names[] = {
152 	[VM_GUEST_NO] = "none",
153 	[VM_GUEST_VM] = "generic",
154 	[VM_GUEST_XEN] = "xen",
155 	[VM_GUEST_HV] = "hv",
156 	[VM_GUEST_VMWARE] = "vmware",
157 	[VM_GUEST_KVM] = "kvm",
158 	[VM_GUEST_BHYVE] = "bhyve",
159 	[VM_GUEST_VBOX] = "vbox",
160 	[VM_GUEST_PARALLELS] = "parallels",
161 	[VM_LAST] = NULL
162 };
163 CTASSERT(nitems(vm_guest_sysctl_names) - 1 == VM_LAST);
164 
165 /*
166  * Boot time overrides that are not scaled against main memory
167  */
168 void
169 init_param1(void)
170 {
171 
172 	TSENTER();
173 
174 	/*
175 	 * arm64 and riscv currently hard-code the thread0 kstack size
176 	 * to KSTACK_PAGES, ignoring the tunable.
177 	 */
178 	TUNABLE_INT_FETCH("kern.kstack_pages", &kstack_pages);
179 
180 	hz = -1;
181 	TUNABLE_INT_FETCH("kern.hz", &hz);
182 	if (hz == -1)
183 		hz = vm_guest > VM_GUEST_NO ? HZ_VM : HZ;
184 
185 	/* range check the "hz" value */
186 	if (__predict_false(hz < HZ_MINIMUM))
187 		hz = HZ_MINIMUM;
188 	else if (__predict_false(hz > HZ_MAXIMUM))
189 		hz = HZ_MAXIMUM;
190 
191 	tick = 1000000 / hz;
192 	tick_sbt = SBT_1S / hz;
193 	tick_bt = sbttobt(tick_sbt);
194 	tick_seconds_max = INT_MAX / hz;
195 
196 	/*
197 	 * Arrange for ticks to wrap 10 minutes after boot to help catch
198 	 * sign problems sooner.
199 	 */
200 	ticks = INT_MAX - (hz * 10 * 60);
201 
202 	vn_lock_pair_pause_max = hz / 100;
203 	if (vn_lock_pair_pause_max == 0)
204 		vn_lock_pair_pause_max = 1;
205 
206 #ifdef VM_SWZONE_SIZE_MAX
207 	maxswzone = VM_SWZONE_SIZE_MAX;
208 #endif
209 	TUNABLE_LONG_FETCH("kern.maxswzone", &maxswzone);
210 #ifdef VM_BCACHE_SIZE_MAX
211 	maxbcache = VM_BCACHE_SIZE_MAX;
212 #endif
213 	TUNABLE_LONG_FETCH("kern.maxbcache", &maxbcache);
214 	msgbufsize = MSGBUF_SIZE;
215 	TUNABLE_INT_FETCH("kern.msgbufsize", &msgbufsize);
216 
217 	maxtsiz = MAXTSIZ;
218 	TUNABLE_ULONG_FETCH("kern.maxtsiz", &maxtsiz);
219 	dfldsiz = DFLDSIZ;
220 	TUNABLE_ULONG_FETCH("kern.dfldsiz", &dfldsiz);
221 	maxdsiz = MAXDSIZ;
222 	TUNABLE_ULONG_FETCH("kern.maxdsiz", &maxdsiz);
223 	dflssiz = DFLSSIZ;
224 	TUNABLE_ULONG_FETCH("kern.dflssiz", &dflssiz);
225 	maxssiz = MAXSSIZ;
226 	TUNABLE_ULONG_FETCH("kern.maxssiz", &maxssiz);
227 	sgrowsiz = SGROWSIZ;
228 	TUNABLE_ULONG_FETCH("kern.sgrowsiz", &sgrowsiz);
229 
230 	/*
231 	 * Let the administrator set {NGROUPS_MAX}, but disallow values
232 	 * less than NGROUPS_MAX which would violate POSIX.1-2008 or
233 	 * greater than INT_MAX-1 which would result in overflow.
234 	 */
235 	ngroups_max = NGROUPS_MAX;
236 	TUNABLE_INT_FETCH("kern.ngroups", &ngroups_max);
237 	if (ngroups_max < NGROUPS_MAX)
238 		ngroups_max = NGROUPS_MAX;
239 
240 	/*
241 	 * Only allow to lower the maximal pid.
242 	 * Prevent setting up a non-bootable system if pid_max is too low.
243 	 */
244 	TUNABLE_INT_FETCH("kern.pid_max", &pid_max);
245 	if (pid_max > PID_MAX)
246 		pid_max = PID_MAX;
247 	else if (pid_max < 300)
248 		pid_max = 300;
249 
250 	TUNABLE_INT_FETCH("vfs.unmapped_buf_allowed", &unmapped_buf_allowed);
251 	TSEXIT();
252 }
253 
254 /*
255  * Boot time overrides that are scaled against main memory
256  */
257 void
258 init_param2(long physpages)
259 {
260 
261 	TSENTER();
262 	/* Base parameters */
263 	maxusers = MAXUSERS;
264 	TUNABLE_INT_FETCH("kern.maxusers", &maxusers);
265 	if (maxusers == 0) {
266 		maxusers = physpages / (2 * 1024 * 1024 / PAGE_SIZE);
267 		if (maxusers < 32)
268 			maxusers = 32;
269 #ifdef VM_MAX_AUTOTUNE_MAXUSERS
270                 if (maxusers > VM_MAX_AUTOTUNE_MAXUSERS)
271                         maxusers = VM_MAX_AUTOTUNE_MAXUSERS;
272 #endif
273                 /*
274                  * Scales down the function in which maxusers grows once
275                  * we hit 384.
276                  */
277                 if (maxusers > 384)
278                         maxusers = 384 + ((maxusers - 384) / 8);
279         }
280 
281 	/*
282 	 * The following can be overridden after boot via sysctl.  Note:
283 	 * unless overridden, these macros are ultimately based on maxusers.
284 	 * Limit maxproc so that kmap entries cannot be exhausted by
285 	 * processes.
286 	 */
287 	maxproc = NPROC;
288 	TUNABLE_INT_FETCH("kern.maxproc", &maxproc);
289 	if (maxproc > (physpages / 12))
290 		maxproc = physpages / 12;
291 	if (maxproc > pid_max)
292 		maxproc = pid_max;
293 	maxprocperuid = (maxproc * 9) / 10;
294 
295 	/*
296 	 * The default limit for maxfiles is 1/12 of the number of
297 	 * physical page but not less than 16 times maxusers.
298 	 * At most it can be 1/6 the number of physical pages.
299 	 */
300 	maxfiles = imax(MAXFILES, physpages / 8);
301 	TUNABLE_INT_FETCH("kern.maxfiles", &maxfiles);
302 	if (maxfiles > (physpages / 4))
303 		maxfiles = physpages / 4;
304 	maxfilesperproc = (maxfiles / 10) * 9;
305 	TUNABLE_INT_FETCH("kern.maxfilesperproc", &maxfilesperproc);
306 
307 	/*
308 	 * Cannot be changed after boot.
309 	 */
310 	nbuf = NBUF;
311 	TUNABLE_INT_FETCH("kern.nbuf", &nbuf);
312 	TUNABLE_INT_FETCH("kern.bio_transient_maxcnt", &bio_transient_maxcnt);
313 	maxphys = MAXPHYS;
314 	TUNABLE_ULONG_FETCH("kern.maxphys", &maxphys);
315 	if (maxphys == 0) {
316 		maxphys = MAXPHYS;
317 	} else if (__bitcountl(maxphys) != 1) {	/* power of two */
318 		if (flsl(maxphys) == NBBY * sizeof(maxphys))
319 			maxphys = MAXPHYS;
320 		else
321 			maxphys = 1UL << flsl(maxphys);
322 	}
323 	if (maxphys < PAGE_SIZE)
324 		maxphys = MAXPHYS;
325 
326 	/*
327 	 * Physical buffers are pre-allocated buffers (struct buf) that
328 	 * are used as temporary holders for I/O, such as paging I/O.
329 	 */
330 	TUNABLE_INT_FETCH("kern.nswbuf", &nswbuf);
331 
332 	/*
333 	 * The default for maxpipekva is min(1/64 of the kernel address space,
334 	 * max(1/64 of main memory, 512KB)).  See sys_pipe.c for more details.
335 	 */
336 	maxpipekva = ptoa(physpages / 64);
337 	TUNABLE_LONG_FETCH("kern.ipc.maxpipekva", &maxpipekva);
338 	if (maxpipekva < 512 * 1024)
339 		maxpipekva = 512 * 1024;
340 	if (maxpipekva > (VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS) / 64)
341 		maxpipekva = (VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS) /
342 		    64;
343 	TSEXIT();
344 }
345 
346 /*
347  * Sysctl stringifying handler for kern.vm_guest.
348  */
349 static int
350 sysctl_kern_vm_guest(SYSCTL_HANDLER_ARGS)
351 {
352 	return (SYSCTL_OUT_STR(req, vm_guest_sysctl_names[vm_guest]));
353 }
354