xref: /netbsd-src/sys/kern/kern_ras.c (revision 4b71a66d0f279143147d63ebfcfd8a59499a3684)
1 /*	$NetBSD: kern_ras.c,v 1.31 2008/05/27 17:42:14 ad Exp $	*/
2 
3 /*-
4  * Copyright (c) 2002, 2006, 2007, 2008 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Gregory McGarry, and by Andrew Doran.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  */
31 
32 #include <sys/cdefs.h>
33 __KERNEL_RCSID(0, "$NetBSD: kern_ras.c,v 1.31 2008/05/27 17:42:14 ad Exp $");
34 
35 #include <sys/param.h>
36 #include <sys/systm.h>
37 #include <sys/kernel.h>
38 #include <sys/kmem.h>
39 #include <sys/proc.h>
40 #include <sys/ras.h>
41 #include <sys/xcall.h>
42 #include <sys/syscallargs.h>
43 
44 #include <uvm/uvm_extern.h>
45 
46 #define MAX_RAS_PER_PROC	16
47 
48 u_int ras_per_proc = MAX_RAS_PER_PROC;
49 
50 #ifdef DEBUG
51 int ras_debug = 0;
52 #define DPRINTF(x)	if (ras_debug) printf x
53 #else
54 #define DPRINTF(x)	/* nothing */
55 #endif
56 
57 /*
58  * Force all CPUs through cpu_switchto(), waiting until complete.
59  * Context switching will drain the write buffer on the calling
60  * CPU.
61  */
62 static void
63 ras_sync(void)
64 {
65 
66 	/* No need to sync if exiting or single threaded. */
67 	if (curproc->p_nlwps > 1 && ncpu > 1) {
68 #ifdef NO_SOFTWARE_PATENTS
69 		uint64_t where;
70 		where = xc_broadcast(0, (xcfunc_t)nullop, NULL, NULL);
71 		xc_wait(where);
72 #else
73 		/*
74 		 * Assumptions:
75 		 *
76 		 * o preemption is disabled by the thread in
77 		 *   ras_lookup().
78 		 * o proc::p_raslist is only inspected with
79 		 *   preemption disabled.
80 		 * o ras_lookup() plus loads reordered in advance
81 		 *   will take no longer than 1/8s to complete.
82 		 */
83 		const int delta = hz >> 3;
84 		int target = hardclock_ticks + delta;
85 		do {
86 			kpause("ras", false, delta, NULL);
87 		} while (hardclock_ticks < target);
88 #endif
89 	}
90 }
91 
92 /*
93  * Check the specified address to see if it is within the
94  * sequence.  If it is found, we return the restart address,
95  * otherwise we return -1.  If we do perform a restart, we
96  * mark the sequence as hit.
97  *
98  * No locking required: we disable preemption and ras_sync()
99  * guarantees that individual entries are valid while we still
100  * have visibility of them.
101  */
102 void *
103 ras_lookup(struct proc *p, void *addr)
104 {
105 	struct ras *rp;
106 	void *startaddr;
107 	lwp_t *l;
108 
109 	startaddr = (void *)-1;
110 	l = curlwp;
111 
112 	KPREEMPT_DISABLE(l);
113 	for (rp = p->p_raslist; rp != NULL; rp = rp->ras_next) {
114 		if (addr > rp->ras_startaddr && addr < rp->ras_endaddr) {
115 			startaddr = rp->ras_startaddr;
116 			DPRINTF(("RAS hit: p=%p %p\n", p, addr));
117 			break;
118 		}
119 	}
120 	KPREEMPT_ENABLE(l);
121 
122 	return startaddr;
123 }
124 
125 /*
126  * During a fork, we copy all of the sequences from parent p1 to
127  * the child p2.
128  *
129  * No locking required as the parent must be paused.
130  */
131 int
132 ras_fork(struct proc *p1, struct proc *p2)
133 {
134 	struct ras *rp, *nrp;
135 
136 	for (rp = p1->p_raslist; rp != NULL; rp = rp->ras_next) {
137 		nrp = kmem_alloc(sizeof(*nrp), KM_SLEEP);
138 		nrp->ras_startaddr = rp->ras_startaddr;
139 		nrp->ras_endaddr = rp->ras_endaddr;
140 		nrp->ras_next = p2->p_raslist;
141 		p2->p_raslist = nrp;
142 	}
143 
144 	DPRINTF(("ras_fork: p1=%p, p2=%p\n", p1, p2));
145 
146 	return 0;
147 }
148 
149 /*
150  * Nuke all sequences for this process.
151  */
152 int
153 ras_purgeall(void)
154 {
155 	struct ras *rp, *nrp;
156 	proc_t *p;
157 
158 	p = curproc;
159 
160 	mutex_enter(&p->p_auxlock);
161 	if ((rp = p->p_raslist) != NULL) {
162 		p->p_raslist = NULL;
163 		ras_sync();
164 		for(; rp != NULL; rp = nrp) {
165 			nrp = rp->ras_next;
166 			kmem_free(rp, sizeof(*rp));
167 		}
168 	}
169 	mutex_exit(&p->p_auxlock);
170 
171 	return 0;
172 }
173 
174 #if defined(__HAVE_RAS)
175 
176 /*
177  * Install the new sequence.  If it already exists, return
178  * an error.
179  */
180 static int
181 ras_install(void *addr, size_t len)
182 {
183 	struct ras *rp;
184 	struct ras *newrp;
185 	void *endaddr;
186 	int nras, error;
187 	proc_t *p;
188 
189 	endaddr = (char *)addr + len;
190 
191 	if (addr < (void *)VM_MIN_ADDRESS ||
192 	    endaddr > (void *)VM_MAXUSER_ADDRESS)
193 		return (EINVAL);
194 
195 	if (len <= 0)
196 		return (EINVAL);
197 
198 	newrp = kmem_alloc(sizeof(*newrp), KM_SLEEP);
199 	newrp->ras_startaddr = addr;
200 	newrp->ras_endaddr = endaddr;
201 	error = 0;
202 	nras = 0;
203 	p = curproc;
204 
205 	mutex_enter(&p->p_auxlock);
206 	for (rp = p->p_raslist; rp != NULL; rp = rp->ras_next) {
207 		if (++nras >= ras_per_proc) {
208 			error = EINVAL;
209 			break;
210 		}
211 		if (addr < rp->ras_endaddr && endaddr > rp->ras_startaddr) {
212 			error = EEXIST;
213 			break;
214 		}
215 	}
216 	if (rp == NULL) {
217 		newrp->ras_next = p->p_raslist;
218 		p->p_raslist = newrp;
219 		ras_sync();
220 	 	mutex_exit(&p->p_auxlock);
221 	} else {
222 	 	mutex_exit(&p->p_auxlock);
223  		kmem_free(newrp, sizeof(*newrp));
224 	}
225 
226 	return error;
227 }
228 
229 /*
230  * Nuke the specified sequence.  Both address and len must
231  * match, otherwise we return an error.
232  */
233 static int
234 ras_purge(void *addr, size_t len)
235 {
236 	struct ras *rp, **link;
237 	void *endaddr;
238 	proc_t *p;
239 
240 	endaddr = (char *)addr + len;
241 	p = curproc;
242 
243 	mutex_enter(&p->p_auxlock);
244 	link = &p->p_raslist;
245 	for (rp = *link; rp != NULL; link = &rp->ras_next, rp = *link) {
246 		if (addr == rp->ras_startaddr && endaddr == rp->ras_endaddr)
247 			break;
248 	}
249 	if (rp != NULL) {
250 		*link = rp->ras_next;
251 		ras_sync();
252 		mutex_exit(&p->p_auxlock);
253 		kmem_free(rp, sizeof(*rp));
254 		return 0;
255 	} else {
256 		mutex_exit(&p->p_auxlock);
257 		return ESRCH;
258 	}
259 }
260 
261 #endif /* defined(__HAVE_RAS) */
262 
263 /*ARGSUSED*/
264 int
265 sys_rasctl(struct lwp *l, const struct sys_rasctl_args *uap, register_t *retval)
266 {
267 
268 #if defined(__HAVE_RAS)
269 	/* {
270 		syscallarg(void *) addr;
271 		syscallarg(size_t) len;
272 		syscallarg(int) op;
273 	} */
274 	void *addr;
275 	size_t len;
276 	int op;
277 	int error;
278 
279 	/*
280 	 * first, extract syscall args from the uap.
281 	 */
282 
283 	addr = (void *)SCARG(uap, addr);
284 	len = (size_t)SCARG(uap, len);
285 	op = SCARG(uap, op);
286 
287 	DPRINTF(("sys_rasctl: p=%p addr=%p, len=%ld, op=0x%x\n",
288 	    curproc, addr, (long)len, op));
289 
290 	switch (op) {
291 	case RAS_INSTALL:
292 		error = ras_install(addr, len);
293 		break;
294 	case RAS_PURGE:
295 		error = ras_purge(addr, len);
296 		break;
297 	case RAS_PURGE_ALL:
298 		error = ras_purgeall();
299 		break;
300 	default:
301 		error = EINVAL;
302 		break;
303 	}
304 
305 	return (error);
306 
307 #else
308 
309 	return (EOPNOTSUPP);
310 
311 #endif
312 
313 }
314