xref: /illumos-gate/usr/src/uts/common/cpr/cpr_mod.c (revision 88f8b78a88cbdc6d8c1af5c3e54bc49d25095c98)
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 #pragma ident	"%Z%%M%	%I%	%E% SMI"
28 
29 /*
30  * System call to checkpoint and resume the currently running kernel
31  */
32 #include <sys/types.h>
33 #include <sys/errno.h>
34 #include <sys/modctl.h>
35 #include <sys/syscall.h>
36 #include <sys/cred.h>
37 #include <sys/uadmin.h>
38 #include <sys/cmn_err.h>
39 #include <sys/systm.h>
40 #include <sys/cpr.h>
41 #include <sys/swap.h>
42 #include <sys/vfs.h>
43 #include <sys/autoconf.h>
44 #include <sys/machsystm.h>
45 
46 extern int i_cpr_is_supported(void);
47 extern int cpr_is_ufs(struct vfs *);
48 extern int cpr_check_spec_statefile(void);
49 extern int cpr_reusable_mount_check(void);
50 extern void cpr_forget_cprconfig(void);
51 extern int i_cpr_reusable_supported(void);
52 extern int i_cpr_reusefini(void);
53 
54 extern struct mod_ops mod_miscops;
55 
56 static struct modlmisc modlmisc = {
57 	&mod_miscops, "checkpoint resume"
58 };
59 
60 static struct modlinkage modlinkage = {
61 	MODREV_1, (void *)&modlmisc, NULL
62 };
63 
64 char _depends_on[] = "misc/bootdev";	/* i_devname_to_promname() */
65 
66 int cpr_reusable_mode;
67 
68 kmutex_t	cpr_slock;	/* cpr serial lock */
69 cpr_t		cpr_state;
70 int		cpr_debug;
71 int		cpr_test_mode; /* true if called via uadmin testmode */
72 
73 /*
74  * All the loadable module related code follows
75  */
76 int
77 _init(void)
78 {
79 	register int e;
80 
81 	if ((e = mod_install(&modlinkage)) == 0) {
82 		mutex_init(&cpr_slock, NULL, MUTEX_DEFAULT, NULL);
83 	}
84 	return (e);
85 }
86 
87 int
88 _fini(void)
89 {
90 	register int e;
91 
92 	if ((e = mod_remove(&modlinkage)) == 0) {
93 		mutex_destroy(&cpr_slock);
94 	}
95 	return (e);
96 }
97 
98 int
99 _info(struct modinfo *modinfop)
100 {
101 	return (mod_info(&modlinkage, modinfop));
102 }
103 
104 int
105 cpr(int fcn)
106 {
107 	static const char noswapstr[] = "reusable statefile requires "
108 	    "that no swap area be configured.\n";
109 	static const char blockstr[] = "reusable statefile must be "
110 	    "a block device.  See power.conf(4) and pmconfig(1M).\n";
111 	static const char normalfmt[] = "cannot run normal "
112 	    "checkpoint/resume when in reusable statefile mode. "
113 	    "use uadmin A_FREEZE AD_REUSEFINI (uadmin %d %d) "
114 	    "to exit reusable statefile mode.\n";
115 	static const char modefmt[] = "%s in reusable mode.\n";
116 	register int rc = 0;
117 	extern int cpr_init(int);
118 	extern void cpr_done(void);
119 
120 	/*
121 	 * Need to know if we're in reusable mode, but we will likely have
122 	 * rebooted since REUSEINIT, so we have to get the info from the
123 	 * file system
124 	 */
125 	if (!cpr_reusable_mode)
126 		cpr_reusable_mode = cpr_get_reusable_mode();
127 
128 	cpr_forget_cprconfig();
129 	switch (fcn) {
130 
131 	case AD_CPR_REUSEINIT:
132 		if (!i_cpr_reusable_supported())
133 			return (ENOTSUP);
134 		if (!cpr_statefile_is_spec()) {
135 			cpr_err(CE_CONT, blockstr);
136 			return (EINVAL);
137 		}
138 		if ((rc = cpr_check_spec_statefile()) != 0)
139 			return (rc);
140 		if (swapinfo) {
141 			cpr_err(CE_CONT, noswapstr);
142 			return (EINVAL);
143 		}
144 		cpr_test_mode = 0;
145 		break;
146 
147 	case AD_CPR_NOCOMPRESS:
148 	case AD_CPR_COMPRESS:
149 	case AD_CPR_FORCE:
150 		if (cpr_reusable_mode) {
151 			cpr_err(CE_CONT, normalfmt, A_FREEZE, AD_REUSEFINI);
152 			return (ENOTSUP);
153 		}
154 		cpr_test_mode = 0;
155 		break;
156 
157 	case AD_CPR_REUSABLE:
158 		if (!i_cpr_reusable_supported())
159 			return (ENOTSUP);
160 		if (!cpr_statefile_is_spec()) {
161 			cpr_err(CE_CONT, blockstr);
162 			return (EINVAL);
163 		}
164 		if ((rc = cpr_check_spec_statefile()) != 0)
165 			return (rc);
166 		if (swapinfo) {
167 			cpr_err(CE_CONT, noswapstr);
168 			return (EINVAL);
169 		}
170 		if ((rc = cpr_reusable_mount_check()) != 0)
171 			return (rc);
172 		cpr_test_mode = 0;
173 		break;
174 
175 	case AD_CPR_REUSEFINI:
176 		if (!i_cpr_reusable_supported())
177 			return (ENOTSUP);
178 		cpr_test_mode = 0;
179 		break;
180 
181 	case AD_CPR_TESTZ:
182 	case AD_CPR_TESTNOZ:
183 	case AD_CPR_TESTHALT:
184 		if (cpr_reusable_mode) {
185 			cpr_err(CE_CONT, normalfmt, A_FREEZE, AD_REUSEFINI);
186 			return (ENOTSUP);
187 		}
188 		cpr_test_mode = 1;
189 		break;
190 
191 	case AD_CPR_CHECK:
192 		if (!i_cpr_is_supported() || cpr_reusable_mode)
193 			return (ENOTSUP);
194 		return (0);
195 
196 	case AD_CPR_PRINT:
197 		CPR_STAT_EVENT_END("POST CPR DELAY");
198 		cpr_stat_event_print();
199 		return (0);
200 
201 	case AD_CPR_DEBUG0:
202 		cpr_debug = 0;
203 		return (0);
204 
205 	case AD_CPR_DEBUG1:
206 	case AD_CPR_DEBUG2:
207 	case AD_CPR_DEBUG3:
208 	case AD_CPR_DEBUG4:
209 	case AD_CPR_DEBUG5:
210 	case AD_CPR_DEBUG7:
211 	case AD_CPR_DEBUG8:
212 		cpr_debug |= CPR_DEBUG_BIT(fcn);
213 		return (0);
214 
215 	case AD_CPR_DEBUG9:
216 		cpr_debug |= LEVEL6;
217 		return (0);
218 
219 	default:
220 		return (ENOTSUP);
221 	}
222 
223 	if (!i_cpr_is_supported() || !cpr_is_ufs(rootvfs))
224 		return (ENOTSUP);
225 
226 	if (fcn == AD_CPR_REUSEINIT) {
227 		if (mutex_tryenter(&cpr_slock) == 0)
228 			return (EBUSY);
229 		if (cpr_reusable_mode) {
230 			cpr_err(CE_CONT, modefmt, "already");
231 			mutex_exit(&cpr_slock);
232 			return (EBUSY);
233 		}
234 		rc = i_cpr_reuseinit();
235 		mutex_exit(&cpr_slock);
236 		return (rc);
237 	}
238 
239 	if (fcn == AD_CPR_REUSEFINI) {
240 		if (mutex_tryenter(&cpr_slock) == 0)
241 			return (EBUSY);
242 		if (!cpr_reusable_mode) {
243 			cpr_err(CE_CONT, modefmt, "not");
244 			mutex_exit(&cpr_slock);
245 			return (EINVAL);
246 		}
247 		rc = i_cpr_reusefini();
248 		mutex_exit(&cpr_slock);
249 		return (rc);
250 	}
251 
252 	/*
253 	 * acquire cpr serial lock and init cpr state structure.
254 	 */
255 	if (rc = cpr_init(fcn))
256 		return (rc);
257 
258 	if (fcn == AD_CPR_REUSABLE) {
259 		if ((rc = i_cpr_check_cprinfo()) != 0)  {
260 			mutex_exit(&cpr_slock);
261 			return (rc);
262 		}
263 	}
264 
265 	/*
266 	 * Call the main cpr routine. If we are successful, we will be coming
267 	 * down from the resume side, otherwise we are still in suspend.
268 	 */
269 	cpr_err(CE_CONT, "System is being suspended");
270 	if (rc = cpr_main()) {
271 		CPR->c_flags |= C_ERROR;
272 		cpr_err(CE_NOTE, "Suspend operation failed.");
273 	} else if (CPR->c_flags & C_SUSPENDING) {
274 		extern void cpr_power_down();
275 		/*
276 		 * Back from a successful checkpoint
277 		 */
278 		if (fcn == AD_CPR_TESTZ || fcn == AD_CPR_TESTNOZ) {
279 			mdboot(0, AD_BOOT, "", B_FALSE);
280 			/* NOTREACHED */
281 		}
282 
283 		/* make sure there are no more changes to the device tree */
284 		devtree_freeze();
285 
286 		/*
287 		 * stop other cpus and raise our priority.  since there is only
288 		 * one active cpu after this, and our priority will be too high
289 		 * for us to be preempted, we're essentially single threaded
290 		 * from here on out.
291 		 */
292 		stop_other_cpus();
293 		(void) spl6();
294 
295 		/*
296 		 * try and reset leaf devices.  reset_leaves() should only
297 		 * be called when there are no other threads that could be
298 		 * accessing devices
299 		 */
300 		reset_leaves();
301 
302 		/*
303 		 * If cpr_power_down() succeeds, it'll not return.
304 		 *
305 		 * Drives with write-cache enabled need to flush
306 		 * their cache.
307 		 */
308 		if (fcn != AD_CPR_TESTHALT)
309 			cpr_power_down();
310 
311 		errp("(Done. Please Switch Off)\n");
312 		halt(NULL);
313 		/* NOTREACHED */
314 	}
315 	/*
316 	 * For resuming: release resources and the serial lock.
317 	 */
318 	cpr_done();
319 	return (rc);
320 }
321