xref: /titanic_41/usr/src/uts/common/os/subr.c (revision fd9cb95cbb2f626355a60efb9d02c5f0a33c10e6)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License, Version 1.0 only
6  * (the "License").  You may not use this file except in compliance
7  * with the License.
8  *
9  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
10  * or http://www.opensolaris.org/os/licensing.
11  * See the License for the specific language governing permissions
12  * and limitations under the License.
13  *
14  * When distributing Covered Code, include this CDDL HEADER in each
15  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
16  * If applicable, add the following below this CDDL HEADER, with the
17  * fields enclosed by brackets "[]" replaced with your own identifying
18  * information: Portions Copyright [yyyy] [name of copyright owner]
19  *
20  * CDDL HEADER END
21  */
22 /*
23  * Copyright 2005 Sun Microsystems, Inc.  All rights reserved.
24  * Use is subject to license terms.
25  */
26 
27 /*	Copyright (c) 1984, 1986, 1987, 1988, 1989 AT&T	*/
28 /*	  All Rights Reserved  	*/
29 
30 
31 #pragma ident	"%Z%%M%	%I%	%E% SMI"
32 
33 #include <sys/types.h>
34 #include <sys/sysmacros.h>
35 #include <sys/param.h>
36 #include <sys/vmparam.h>
37 #include <sys/systm.h>
38 #include <sys/cred.h>
39 #include <sys/user.h>
40 #include <sys/proc.h>
41 #include <sys/conf.h>
42 #include <sys/tuneable.h>
43 #include <sys/cpuvar.h>
44 #include <sys/archsystm.h>
45 #include <sys/vmem.h>
46 #include <vm/seg_kmem.h>
47 #include <sys/errno.h>
48 #include <sys/cmn_err.h>
49 #include <sys/debug.h>
50 #include <sys/atomic.h>
51 #include <sys/model.h>
52 #include <sys/kmem.h>
53 #include <sys/memlist.h>
54 #include <sys/autoconf.h>
55 #include <sys/ontrap.h>
56 #include <sys/utsname.h>
57 #include <sys/zone.h>
58 
59 #ifdef __sparc
60 #include <sys/membar.h>
61 #endif
62 
63 /*
64  * Routine which sets a user error; placed in
65  * illegal entries in the bdevsw and cdevsw tables.
66  */
67 
68 int
69 nodev()
70 {
71 	return (curthread->t_lwp ?
72 	    ttolwp(curthread)->lwp_error = ENXIO : ENXIO);
73 }
74 
75 /*
76  * Null routine; placed in insignificant entries
77  * in the bdevsw and cdevsw tables.
78  */
79 
80 int
81 nulldev()
82 {
83 	return (0);
84 }
85 
86 static kmutex_t udevlock;
87 
88 /*
89  * Generate an unused major device number.
90  */
91 major_t
92 getudev()
93 {
94 	static major_t next = 0;
95 	major_t ret;
96 
97 	/*
98 	 * Ensure that we start allocating major numbers above the 'devcnt'
99 	 * count.  The only limit we place on the number is that it should be a
100 	 * legal 32-bit SVR4 major number and be greater than or equal to devcnt
101 	 * in the current system).
102 	 */
103 	mutex_enter(&udevlock);
104 	if (next == 0)
105 		next = devcnt;
106 	if (next <= L_MAXMAJ32 && next >= devcnt)
107 		ret = next++;
108 	else {
109 		/*
110 		 * If we fail to allocate a major number because devcnt has
111 		 * reached L_MAXMAJ32, we may be the victim of a sparsely
112 		 * populated devnames array.  We scan the array backwards
113 		 * looking for an empty slot;  if we find one, mark it as
114 		 * DN_GETUDEV so it doesn't get taken by subsequent consumers
115 		 * users of the devnames array, and issue a warning.
116 		 * It is vital for this routine to take drastic measures to
117 		 * succeed, since the kernel really needs it to boot.
118 		 */
119 		int i;
120 		for (i = devcnt - 1; i >= 0; i--) {
121 			LOCK_DEV_OPS(&devnamesp[i].dn_lock);
122 			if (devnamesp[i].dn_name == NULL &&
123 			    ((devnamesp[i].dn_flags & DN_TAKEN_GETUDEV) == 0))
124 				break;
125 			UNLOCK_DEV_OPS(&devnamesp[i].dn_lock);
126 		}
127 		if (i != -1) {
128 			cmn_err(CE_WARN, "Reusing device major number %d.", i);
129 			ASSERT(i >= 0 && i < devcnt);
130 			devnamesp[i].dn_flags |= DN_TAKEN_GETUDEV;
131 			UNLOCK_DEV_OPS(&devnamesp[i].dn_lock);
132 			ret = (major_t)i;
133 		} else {
134 			ret = (major_t)-1;
135 		}
136 	}
137 	mutex_exit(&udevlock);
138 	return (ret);
139 }
140 
141 
142 /*
143  * Compress 'long' device number encoding to 32-bit device number
144  * encoding.  If it won't fit, we return failure, but set the
145  * device number to 32-bit NODEV for the sake of our callers.
146  */
147 int
148 cmpldev(dev32_t *dst, dev_t dev)
149 {
150 #if defined(_LP64)
151 	if (dev == NODEV) {
152 		*dst = NODEV32;
153 	} else {
154 		major_t major = dev >> L_BITSMINOR;
155 		minor_t minor = dev & L_MAXMIN;
156 
157 		if (major > L_MAXMAJ32 || minor > L_MAXMIN32) {
158 			*dst = NODEV32;
159 			return (0);
160 		}
161 
162 		*dst = (dev32_t)((major << L_BITSMINOR32) | minor);
163 	}
164 #else
165 	*dst = (dev32_t)dev;
166 #endif
167 	return (1);
168 }
169 
170 /*
171  * Expand 32-bit dev_t's to long dev_t's.  Expansion always "fits"
172  * into the return type, but we're careful to expand NODEV explicitly.
173  */
174 dev_t
175 expldev(dev32_t dev32)
176 {
177 #ifdef _LP64
178 	if (dev32 == NODEV32)
179 		return (NODEV);
180 	return (makedevice((dev32 >> L_BITSMINOR32) & L_MAXMAJ32,
181 	    dev32 & L_MAXMIN32));
182 #else
183 	return ((dev_t)dev32);
184 #endif
185 }
186 
187 #ifndef _LP64
188 /*
189  * Keep these entry points for 32-bit systems but enforce the use
190  * of MIN/MAX macros on 64-bit systems.  The DDI header files already
191  * define min/max as macros so drivers shouldn't need these functions.
192  */
193 
194 int
195 min(int a, int b)
196 {
197 	return (a < b ? a : b);
198 }
199 
200 int
201 max(int a, int b)
202 {
203 	return (a > b ? a : b);
204 }
205 
206 uint_t
207 umin(uint_t a, uint_t b)
208 {
209 	return (a < b ? a : b);
210 }
211 
212 uint_t
213 umax(uint_t a, uint_t b)
214 {
215 	return (a > b ? a : b);
216 }
217 
218 #endif /* !_LP64 */
219 
220 /*
221  * Return bit position of least significant bit set in mask,
222  * starting numbering from 1.
223  */
224 int
225 ffs(long mask)
226 {
227 	int i;
228 
229 	if (mask == 0)
230 		return (0);
231 	for (i = 1; i <= NBBY * sizeof (mask); i++) {
232 		if (mask & 1)
233 			return (i);
234 		mask >>= 1;
235 	}
236 	return (0);
237 }
238 
239 /*
240  * Parse suboptions from a string.
241  * Same as getsubopt(3C).
242  */
243 int
244 getsubopt(char **optionsp, char * const *tokens, char **valuep)
245 {
246 	char *s = *optionsp, *p;
247 	int i;
248 	size_t optlen;
249 
250 	*valuep = NULL;
251 	if (*s == '\0')
252 		return (-1);
253 	p = strchr(s, ',');		/* find next option */
254 	if (p == NULL) {
255 		p = s + strlen(s);
256 	} else {
257 		*p++ = '\0';		/* mark end and point to next */
258 	}
259 	*optionsp = p;			/* point to next option */
260 	p = strchr(s, '=');		/* find value */
261 	if (p == NULL) {
262 		optlen = strlen(s);
263 		*valuep = NULL;
264 	} else {
265 		optlen = p - s;
266 		*valuep = ++p;
267 	}
268 	for (i = 0; tokens[i] != NULL; i++) {
269 		if ((optlen == strlen(tokens[i])) &&
270 		    (strncmp(s, tokens[i], optlen) == 0))
271 			return (i);
272 	}
273 	/* no match, point value at option and return error */
274 	*valuep = s;
275 	return (-1);
276 }
277 
278 /*
279  * Append the suboption string 'opt' starting at the position 'str'
280  * within the buffer defined by 'buf' and 'len'. If 'buf' is not null,
281  * a comma is appended first.
282  * Return a pointer to the end of the resulting string (the null byte).
283  * Return NULL if there isn't enough space left to append 'opt'.
284  */
285 char *
286 append_subopt(const char *buf, size_t len, char *str, const char *opt)
287 {
288 	size_t l = strlen(opt);
289 
290 	/*
291 	 * Include a ',' if this is not the first option.
292 	 * Include space for the null byte.
293 	 */
294 	if (strlen(buf) + (buf[0] != '\0') + l + 1 > len)
295 		return (NULL);
296 
297 	if (buf[0] != '\0')
298 		*str++ = ',';
299 	(void) strcpy(str, opt);
300 	return (str + l);
301 }
302 
303 /*
304  * Tables to convert a single byte to/from binary-coded decimal (BCD).
305  */
306 uchar_t byte_to_bcd[256] = {
307 	0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08, 0x09,
308 	0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, 0x18, 0x19,
309 	0x20, 0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27, 0x28, 0x29,
310 	0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x36, 0x37, 0x38, 0x39,
311 	0x40, 0x41, 0x42, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48, 0x49,
312 	0x50, 0x51, 0x52, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58, 0x59,
313 	0x60, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69,
314 	0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79,
315 	0x80, 0x81, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89,
316 	0x90, 0x91, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98, 0x99,
317 };
318 
319 uchar_t bcd_to_byte[256] = {		/* CSTYLED */
320 	 0,  1,  2,  3,  4,  5,  6,  7,  8,  9,  0,  0,  0,  0,  0,  0,
321 	10, 11, 12, 13, 14, 15, 16, 17, 18, 19,  0,  0,  0,  0,  0,  0,
322 	20, 21, 22, 23, 24, 25, 26, 27, 28, 29,  0,  0,  0,  0,  0,  0,
323 	30, 31, 32, 33, 34, 35, 36, 37, 38, 39,  0,  0,  0,  0,  0,  0,
324 	40, 41, 42, 43, 44, 45, 46, 47, 48, 49,  0,  0,  0,  0,  0,  0,
325 	50, 51, 52, 53, 54, 55, 56, 57, 58, 59,  0,  0,  0,  0,  0,  0,
326 	60, 61, 62, 63, 64, 65, 66, 67, 68, 69,  0,  0,  0,  0,  0,  0,
327 	70, 71, 72, 73, 74, 75, 76, 77, 78, 79,  0,  0,  0,  0,  0,  0,
328 	80, 81, 82, 83, 84, 85, 86, 87, 88, 89,  0,  0,  0,  0,  0,  0,
329 	90, 91, 92, 93, 94, 95, 96, 97, 98, 99,
330 };
331 
332 /*
333  * Hot-patch a single instruction in the kernel's text.
334  * If you want to patch multiple instructions you must
335  * arrange to do it so that all intermediate stages are
336  * sane -- we don't stop other cpus while doing this.
337  * Size must be 1, 2, or 4 bytes with iaddr aligned accordingly.
338  */
339 void
340 hot_patch_kernel_text(caddr_t iaddr, uint32_t new_instr, uint_t size)
341 {
342 	caddr_t vaddr;
343 	page_t **ppp;
344 	uintptr_t off = (uintptr_t)iaddr & PAGEOFFSET;
345 
346 	vaddr = vmem_alloc(heap_arena, PAGESIZE, VM_SLEEP);
347 
348 	(void) as_pagelock(&kas, &ppp, iaddr - off, PAGESIZE, S_WRITE);
349 
350 	hat_devload(kas.a_hat, vaddr, PAGESIZE,
351 	    hat_getpfnum(kas.a_hat, iaddr - off),
352 	    PROT_READ | PROT_WRITE, HAT_LOAD_LOCK | HAT_LOAD_NOCONSIST);
353 
354 	switch (size) {
355 	case 1:
356 		*(uint8_t *)(vaddr + off) = new_instr;
357 		break;
358 	case 2:
359 		*(uint16_t *)(vaddr + off) = new_instr;
360 		break;
361 	case 4:
362 		*(uint32_t *)(vaddr + off) = new_instr;
363 		break;
364 	default:
365 		panic("illegal hot-patch");
366 	}
367 
368 	membar_enter();
369 	sync_icache(vaddr + off, size);
370 	sync_icache(iaddr, size);
371 	as_pageunlock(&kas, ppp, iaddr - off, PAGESIZE, S_WRITE);
372 	hat_unload(kas.a_hat, vaddr, PAGESIZE, HAT_UNLOAD_UNLOCK);
373 	vmem_free(heap_arena, vaddr, PAGESIZE);
374 }
375 
376 /*
377  * Routine to report an attempt to execute non-executable data.  If the
378  * address executed lies in the stack, explicitly say so.
379  */
380 void
381 report_stack_exec(proc_t *p, caddr_t addr)
382 {
383 	if (!noexec_user_stack_log)
384 		return;
385 
386 	if (addr < p->p_usrstack && addr >= (p->p_usrstack - p->p_stksize)) {
387 		cmn_err(CE_NOTE, "%s[%d] attempt to execute code "
388 		    "on stack by uid %d", p->p_user.u_comm,
389 		    p->p_pid, crgetruid(p->p_cred));
390 	} else {
391 		cmn_err(CE_NOTE, "%s[%d] attempt to execute non-executable "
392 		    "data at 0x%p by uid %d", p->p_user.u_comm,
393 		    p->p_pid, (void *) addr, crgetruid(p->p_cred));
394 	}
395 
396 	delay(hz / 50);
397 }
398 
399 /*
400  * Determine whether the address range [addr, addr + len) is in memlist mp.
401  */
402 int
403 address_in_memlist(struct memlist *mp, uint64_t addr, size_t len)
404 {
405 	while (mp != 0)	 {
406 		if ((addr >= mp->address) &&
407 		    (addr + len <= mp->address + mp->size))
408 			return (1);	 /* TRUE */
409 		mp = mp->next;
410 	}
411 	return (0);	/* FALSE */
412 }
413 
414 /*
415  * Pop the topmost element from the t_ontrap stack, removing the current set of
416  * on_trap() protections.  Refer to <sys/ontrap.h> for more info.  If the
417  * stack is already empty, no_trap() just returns.
418  */
419 void
420 no_trap(void)
421 {
422 	if (curthread->t_ontrap != NULL) {
423 #ifdef __sparc
424 		membar_sync(); /* deferred error barrier (see sparcv9_subr.s) */
425 #endif
426 		curthread->t_ontrap = curthread->t_ontrap->ot_prev;
427 	}
428 }
429 
430 /*
431  * Return utsname.nodename outside a zone, or the zone name within.
432  */
433 char *
434 uts_nodename(void)
435 {
436 	if (curproc == NULL)
437 		return (utsname.nodename);
438 	return (curproc->p_zone->zone_nodename);
439 }
440