xref: /freebsd-13-stable/sys/kern/subr_param.c (revision 50e3e905e409a8f6387328326c5dc37a4fbbcdd8)
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 #  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 time_t	tick_seconds_max;		/* max hz * seconds an integer can hold */
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;				/* number of bcache bufs */
99 int	bio_transient_maxcnt;
100 int	ngroups_max;			/* max # groups per process */
101 int	nswbuf;
102 pid_t	pid_max = PID_MAX;
103 u_long	maxswzone;			/* max swmeta KVA storage */
104 u_long	maxbcache;			/* max buffer cache KVA storage */
105 u_long	maxpipekva;			/* Limit on pipe KVA */
106 u_long	maxphys;			/* max raw I/O transfer size */
107 int	vm_guest = VM_GUEST_NO;		/* Running as virtual machine guest? */
108 u_long	maxtsiz;			/* max text size */
109 u_long	dfldsiz;			/* initial data size limit */
110 u_long	maxdsiz;			/* max data size */
111 u_long	dflssiz;			/* initial stack size limit */
112 u_long	maxssiz;			/* max stack size */
113 u_long	sgrowsiz;			/* amount to grow stack */
114 
115 SYSCTL_INT(_kern, OID_AUTO, hz, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &hz, 0,
116     "Number of clock ticks per second");
117 SYSCTL_INT(_kern, OID_AUTO, hz_max, CTLFLAG_RD, SYSCTL_NULL_INT_PTR, HZ_MAXIMUM,
118     "Maximum hz value supported");
119 SYSCTL_INT(_kern, OID_AUTO, hz_min, CTLFLAG_RD, SYSCTL_NULL_INT_PTR, HZ_MINIMUM,
120     "Minimum hz value supported");
121 SYSCTL_INT(_kern, OID_AUTO, nbuf, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &nbuf, 0,
122     "Number of buffers in the buffer cache");
123 SYSCTL_INT(_kern, OID_AUTO, nswbuf, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &nswbuf, 0,
124     "Number of swap buffers");
125 SYSCTL_INT(_kern, OID_AUTO, msgbufsize, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &msgbufsize, 0,
126     "Size of the kernel message buffer");
127 SYSCTL_LONG(_kern, OID_AUTO, maxswzone, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &maxswzone, 0,
128     "Maximum memory for swap metadata");
129 SYSCTL_LONG(_kern, OID_AUTO, maxbcache, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &maxbcache, 0,
130     "Maximum value of vfs.maxbufspace");
131 SYSCTL_INT(_kern, OID_AUTO, bio_transient_maxcnt, CTLFLAG_RDTUN | CTLFLAG_NOFETCH,
132     &bio_transient_maxcnt, 0,
133     "Maximum number of transient BIOs mappings");
134 SYSCTL_ULONG(_kern, OID_AUTO, maxtsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &maxtsiz, 0,
135     "Maximum text size");
136 SYSCTL_ULONG(_kern, OID_AUTO, dfldsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &dfldsiz, 0,
137     "Initial data size limit");
138 SYSCTL_ULONG(_kern, OID_AUTO, maxdsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &maxdsiz, 0,
139     "Maximum data size");
140 SYSCTL_ULONG(_kern, OID_AUTO, dflssiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &dflssiz, 0,
141     "Initial stack size limit");
142 SYSCTL_ULONG(_kern, OID_AUTO, maxssiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &maxssiz, 0,
143     "Maximum stack size");
144 SYSCTL_ULONG(_kern, OID_AUTO, sgrowsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &sgrowsiz, 0,
145     "Amount to grow stack on a stack fault");
146 SYSCTL_PROC(_kern, OID_AUTO, vm_guest,
147     CTLFLAG_RD | CTLTYPE_STRING | CTLFLAG_MPSAFE, NULL, 0,
148     sysctl_kern_vm_guest, "A",
149     "Virtual machine guest detected?");
150 
151 /*
152  * The elements of this array are ordered based upon the values of the
153  * corresponding enum VM_GUEST members.
154  */
155 static const char *const vm_guest_sysctl_names[] = {
156 	[VM_GUEST_NO] = "none",
157 	[VM_GUEST_VM] = "generic",
158 	[VM_GUEST_XEN] = "xen",
159 	[VM_GUEST_HV] = "hv",
160 	[VM_GUEST_VMWARE] = "vmware",
161 	[VM_GUEST_KVM] = "kvm",
162 	[VM_GUEST_BHYVE] = "bhyve",
163 	[VM_GUEST_VBOX] = "vbox",
164 	[VM_GUEST_PARALLELS] = "parallels",
165 	[VM_GUEST_NVMM] = "nvmm",
166 	[VM_LAST] = NULL
167 };
168 CTASSERT(nitems(vm_guest_sysctl_names) - 1 == VM_LAST);
169 
170 /*
171  * Boot time overrides that are not scaled against main memory
172  */
173 void
init_param1(void)174 init_param1(void)
175 {
176 
177 	TSENTER();
178 
179 	/*
180 	 * arm64 and riscv currently hard-code the thread0 kstack size
181 	 * to KSTACK_PAGES, ignoring the tunable.
182 	 */
183 #if !defined(__mips__)
184 	TUNABLE_INT_FETCH("kern.kstack_pages", &kstack_pages);
185 #endif
186 
187 	hz = -1;
188 	TUNABLE_INT_FETCH("kern.hz", &hz);
189 	if (hz == -1)
190 		hz = vm_guest > VM_GUEST_NO ? HZ_VM : HZ;
191 
192 	/* range check the "hz" value */
193 	if (__predict_false(hz < HZ_MINIMUM))
194 		hz = HZ_MINIMUM;
195 	else if (__predict_false(hz > HZ_MAXIMUM))
196 		hz = HZ_MAXIMUM;
197 
198 	tick = 1000000 / hz;
199 	tick_sbt = SBT_1S / hz;
200 	tick_bt = sbttobt(tick_sbt);
201 	tick_seconds_max = INT_MAX / hz;
202 
203 	/*
204 	 * Arrange for ticks to wrap 10 minutes after boot to help catch
205 	 * sign problems sooner.
206 	 */
207 	ticks = INT_MAX - (hz * 10 * 60);
208 
209 	vn_lock_pair_pause_max = hz / 100;
210 	if (vn_lock_pair_pause_max == 0)
211 		vn_lock_pair_pause_max = 1;
212 
213 #ifdef VM_SWZONE_SIZE_MAX
214 	maxswzone = VM_SWZONE_SIZE_MAX;
215 #endif
216 	TUNABLE_LONG_FETCH("kern.maxswzone", &maxswzone);
217 #ifdef VM_BCACHE_SIZE_MAX
218 	maxbcache = VM_BCACHE_SIZE_MAX;
219 #endif
220 	TUNABLE_LONG_FETCH("kern.maxbcache", &maxbcache);
221 	msgbufsize = MSGBUF_SIZE;
222 	TUNABLE_INT_FETCH("kern.msgbufsize", &msgbufsize);
223 
224 	maxtsiz = MAXTSIZ;
225 	TUNABLE_ULONG_FETCH("kern.maxtsiz", &maxtsiz);
226 	dfldsiz = DFLDSIZ;
227 	TUNABLE_ULONG_FETCH("kern.dfldsiz", &dfldsiz);
228 	maxdsiz = MAXDSIZ;
229 	TUNABLE_ULONG_FETCH("kern.maxdsiz", &maxdsiz);
230 	dflssiz = DFLSSIZ;
231 	TUNABLE_ULONG_FETCH("kern.dflssiz", &dflssiz);
232 	maxssiz = MAXSSIZ;
233 	TUNABLE_ULONG_FETCH("kern.maxssiz", &maxssiz);
234 	sgrowsiz = SGROWSIZ;
235 	TUNABLE_ULONG_FETCH("kern.sgrowsiz", &sgrowsiz);
236 
237 	/*
238 	 * Let the administrator set {NGROUPS_MAX}.
239 	 *
240 	 * Values less than NGROUPS_MAX would violate POSIX/SuS (see the
241 	 * specification for <limits.h>, paragraph "Runtime Increasable
242 	 * Values").
243 	 *
244 	 * On the other hand, INT_MAX would result in an overflow for the common
245 	 * 'ngroups_max + 1' computation (to obtain the size of the internal
246 	 * groups array, its first element being reserved for the effective
247 	 * GID).  Also, the number of allocated bytes for the group array must
248 	 * not overflow on 32-bit machines.  For all these reasons, we limit the
249 	 * number of supplementary groups to some very high number that we
250 	 * expect will never be reached in all practical uses and ensures we
251 	 * avoid the problems just exposed, even if 'gid_t' was to be enlarged
252 	 * by a magnitude.
253 	 */
254 	ngroups_max = NGROUPS_MAX;
255 	TUNABLE_INT_FETCH("kern.ngroups", &ngroups_max);
256 	if (ngroups_max < NGROUPS_MAX)
257 		ngroups_max = NGROUPS_MAX;
258 	else {
259 		const int ngroups_max_max = (1 << 24) - 1;
260 
261 		if (ngroups_max > ngroups_max_max)
262 			ngroups_max = ngroups_max_max;
263 	}
264 
265 	/*
266 	 * Only allow to lower the maximal pid.
267 	 * Prevent setting up a non-bootable system if pid_max is too low.
268 	 */
269 	TUNABLE_INT_FETCH("kern.pid_max", &pid_max);
270 	if (pid_max > PID_MAX)
271 		pid_max = PID_MAX;
272 	else if (pid_max < 300)
273 		pid_max = 300;
274 
275 	TUNABLE_INT_FETCH("vfs.unmapped_buf_allowed", &unmapped_buf_allowed);
276 }
277 
278 /*
279  * Boot time overrides that are scaled against main memory
280  */
281 void
init_param2(long physpages)282 init_param2(long physpages)
283 {
284 
285 	/* Base parameters */
286 	maxusers = MAXUSERS;
287 	TUNABLE_INT_FETCH("kern.maxusers", &maxusers);
288 	if (maxusers == 0) {
289 		maxusers = physpages / (2 * 1024 * 1024 / PAGE_SIZE);
290 		if (maxusers < 32)
291 			maxusers = 32;
292 #ifdef VM_MAX_AUTOTUNE_MAXUSERS
293                 if (maxusers > VM_MAX_AUTOTUNE_MAXUSERS)
294                         maxusers = VM_MAX_AUTOTUNE_MAXUSERS;
295 #endif
296                 /*
297                  * Scales down the function in which maxusers grows once
298                  * we hit 384.
299                  */
300                 if (maxusers > 384)
301                         maxusers = 384 + ((maxusers - 384) / 8);
302         }
303 
304 	/*
305 	 * The following can be overridden after boot via sysctl.  Note:
306 	 * unless overridden, these macros are ultimately based on maxusers.
307 	 * Limit maxproc so that kmap entries cannot be exhausted by
308 	 * processes.
309 	 */
310 	maxproc = NPROC;
311 	TUNABLE_INT_FETCH("kern.maxproc", &maxproc);
312 	if (maxproc > (physpages / 12))
313 		maxproc = physpages / 12;
314 	if (maxproc > pid_max)
315 		maxproc = pid_max;
316 	maxprocperuid = (maxproc * 9) / 10;
317 
318 	/*
319 	 * The default limit for maxfiles is 1/12 of the number of
320 	 * physical page but not less than 16 times maxusers.
321 	 * At most it can be 1/6 the number of physical pages.
322 	 */
323 	maxfiles = imax(MAXFILES, physpages / 8);
324 	TUNABLE_INT_FETCH("kern.maxfiles", &maxfiles);
325 	if (maxfiles > (physpages / 4))
326 		maxfiles = physpages / 4;
327 	maxfilesperproc = (maxfiles / 10) * 9;
328 	TUNABLE_INT_FETCH("kern.maxfilesperproc", &maxfilesperproc);
329 
330 	/*
331 	 * Cannot be changed after boot.
332 	 */
333 	nbuf = NBUF;
334 	TUNABLE_INT_FETCH("kern.nbuf", &nbuf);
335 	TUNABLE_INT_FETCH("kern.bio_transient_maxcnt", &bio_transient_maxcnt);
336 	maxphys = MAXPHYS;
337 	TUNABLE_ULONG_FETCH("kern.maxphys", &maxphys);
338 	if (maxphys == 0) {
339 		maxphys = MAXPHYS;
340 	} else if (__bitcountl(maxphys) != 1) {	/* power of two */
341 		if (flsl(maxphys) == NBBY * sizeof(maxphys))
342 			maxphys = MAXPHYS;
343 		else
344 			maxphys = 1UL << flsl(maxphys);
345 	}
346 	if (maxphys < PAGE_SIZE)
347 		maxphys = MAXPHYS;
348 
349 	/*
350 	 * Physical buffers are pre-allocated buffers (struct buf) that
351 	 * are used as temporary holders for I/O, such as paging I/O.
352 	 */
353 	TUNABLE_INT_FETCH("kern.nswbuf", &nswbuf);
354 
355 	/*
356 	 * The default for maxpipekva is min(1/64 of the kernel address space,
357 	 * max(1/64 of main memory, 512KB)).  See sys_pipe.c for more details.
358 	 */
359 	maxpipekva = ptoa(physpages / 64);
360 	TUNABLE_LONG_FETCH("kern.ipc.maxpipekva", &maxpipekva);
361 	if (maxpipekva < 512 * 1024)
362 		maxpipekva = 512 * 1024;
363 	if (maxpipekva > (VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS) / 64)
364 		maxpipekva = (VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS) /
365 		    64;
366 }
367 
368 /*
369  * Sysctl stringifying handler for kern.vm_guest.
370  */
371 static int
sysctl_kern_vm_guest(SYSCTL_HANDLER_ARGS)372 sysctl_kern_vm_guest(SYSCTL_HANDLER_ARGS)
373 {
374 	return (SYSCTL_OUT_STR(req, vm_guest_sysctl_names[vm_guest]));
375 }
376