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 (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 22 /* 23 * Copyright 2007 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 #include <sys/param.h> 30 #include <sys/types.h> 31 #include <sys/sysmacros.h> 32 #include <sys/systm.h> 33 #include <sys/errno.h> 34 #include <sys/vfs.h> 35 #include <sys/vnode.h> 36 #include <sys/swap.h> 37 #include <sys/file.h> 38 #include <sys/proc.h> 39 #include <sys/var.h> 40 #include <sys/uadmin.h> 41 #include <sys/signal.h> 42 #include <sys/time.h> 43 #include <vm/seg_kmem.h> 44 #include <sys/modctl.h> 45 #include <sys/callb.h> 46 #include <sys/dumphdr.h> 47 #include <sys/debug.h> 48 #include <sys/ftrace.h> 49 #include <sys/cmn_err.h> 50 #include <sys/panic.h> 51 #include <sys/ddi.h> 52 #include <sys/sunddi.h> 53 #include <sys/policy.h> 54 #include <sys/zone.h> 55 #include <sys/condvar.h> 56 #include <sys/thread.h> 57 58 /* 59 * Administrivia system call. We provide this in two flavors: one for calling 60 * from the system call path (uadmin), and the other for calling from elsewhere 61 * within the kernel (kadmin). Callers must beware that certain uadmin cmd 62 * values (specifically A_SWAPCTL) are only supported by uadmin and not kadmin. 63 */ 64 65 extern ksema_t fsflush_sema; 66 kmutex_t ualock; 67 kcondvar_t uacond; 68 kthread_t *ua_shutdown_thread = NULL; 69 70 int sys_shutdown = 0; 71 72 /* 73 * Kill all user processes in said zone. A special argument of ALL_ZONES is 74 * passed in when the system as a whole is shutting down. The lack of per-zone 75 * process lists is likely to make the following a performance bottleneck on a 76 * system with many zones. 77 */ 78 void 79 killall(zoneid_t zoneid) 80 { 81 proc_t *p; 82 83 ASSERT(zoneid != GLOBAL_ZONEID); 84 /* 85 * Kill all processes except kernel daemons and ourself. 86 * Make a first pass to stop all processes so they won't 87 * be trying to restart children as we kill them. 88 */ 89 mutex_enter(&pidlock); 90 for (p = practive; p != NULL; p = p->p_next) { 91 if ((zoneid == ALL_ZONES || p->p_zone->zone_id == zoneid) && 92 p->p_exec != NULLVP && /* kernel daemons */ 93 p->p_as != &kas && 94 p->p_stat != SZOMB) { 95 mutex_enter(&p->p_lock); 96 p->p_flag |= SNOWAIT; 97 sigtoproc(p, NULL, SIGSTOP); 98 mutex_exit(&p->p_lock); 99 } 100 } 101 p = practive; 102 while (p != NULL) { 103 if ((zoneid == ALL_ZONES || p->p_zone->zone_id == zoneid) && 104 p->p_exec != NULLVP && /* kernel daemons */ 105 p->p_as != &kas && 106 p->p_stat != SIDL && 107 p->p_stat != SZOMB) { 108 mutex_enter(&p->p_lock); 109 if (sigismember(&p->p_sig, SIGKILL)) { 110 mutex_exit(&p->p_lock); 111 p = p->p_next; 112 } else { 113 sigtoproc(p, NULL, SIGKILL); 114 mutex_exit(&p->p_lock); 115 (void) cv_timedwait(&p->p_srwchan_cv, &pidlock, 116 lbolt + hz); 117 p = practive; 118 } 119 } else { 120 p = p->p_next; 121 } 122 } 123 mutex_exit(&pidlock); 124 } 125 126 int 127 kadmin(int cmd, int fcn, void *mdep, cred_t *credp) 128 { 129 int error = 0; 130 char *buf; 131 size_t buflen = 0; 132 boolean_t invoke_cb = B_FALSE; 133 134 /* 135 * We might be called directly by the kernel's fault-handling code, so 136 * we can't assert that the caller is in the global zone. 137 */ 138 139 /* 140 * Make sure that cmd is one of the valid <sys/uadmin.h> command codes 141 * and that we have appropriate privileges for this action. 142 */ 143 switch (cmd) { 144 case A_FTRACE: 145 case A_SHUTDOWN: 146 case A_REBOOT: 147 case A_REMOUNT: 148 case A_FREEZE: 149 case A_DUMP: 150 if (secpolicy_sys_config(credp, B_FALSE) != 0) 151 return (EPERM); 152 break; 153 154 default: 155 return (EINVAL); 156 } 157 158 /* 159 * Serialize these operations on ualock. If it is held, the 160 * system should shutdown, reboot, or remount shortly, unless there is 161 * an error. We need a cv rather than just a mutex because proper 162 * functioning of A_REBOOT relies on being able to interrupt blocked 163 * userland callers. 164 * 165 * We only clear ua_shutdown_thread after A_REMOUNT, because A_SHUTDOWN 166 * and A_REBOOT should never return. 167 */ 168 if (cmd == A_SHUTDOWN || cmd == A_REBOOT || cmd == A_REMOUNT) { 169 mutex_enter(&ualock); 170 while (ua_shutdown_thread != NULL) { 171 if (cv_wait_sig(&uacond, &ualock) == 0) { 172 /* 173 * If we were interrupted, leave, and handle 174 * the signal (or exit, depending on what 175 * happened) 176 */ 177 mutex_exit(&ualock); 178 return (EINTR); 179 } 180 } 181 ua_shutdown_thread = curthread; 182 mutex_exit(&ualock); 183 } 184 185 switch (cmd) { 186 case A_SHUTDOWN: 187 { 188 proc_t *p = ttoproc(curthread); 189 190 /* 191 * Release (almost) all of our own resources if we are called 192 * from a user context, however if we are calling kadmin() from 193 * a kernel context then we do not release these resources. 194 */ 195 if (p != &p0) { 196 proc_is_exiting(p); 197 if ((error = exitlwps(0)) != 0) { 198 /* 199 * Another thread in this process also called 200 * exitlwps(). 201 */ 202 mutex_enter(&ualock); 203 ua_shutdown_thread = NULL; 204 cv_signal(&uacond); 205 mutex_exit(&ualock); 206 return (error); 207 } 208 mutex_enter(&p->p_lock); 209 p->p_flag |= SNOWAIT; 210 sigfillset(&p->p_ignore); 211 curthread->t_lwp->lwp_cursig = 0; 212 curthread->t_lwp->lwp_extsig = 0; 213 if (p->p_exec) { 214 vnode_t *exec_vp = p->p_exec; 215 p->p_exec = NULLVP; 216 mutex_exit(&p->p_lock); 217 VN_RELE(exec_vp); 218 } else { 219 mutex_exit(&p->p_lock); 220 } 221 222 pollcleanup(); 223 closeall(P_FINFO(curproc)); 224 relvm(); 225 226 } else { 227 /* 228 * Reset t_cred if not set because much of the 229 * filesystem code depends on CRED() being valid. 230 */ 231 if (curthread->t_cred == NULL) 232 curthread->t_cred = kcred; 233 } 234 235 /* indicate shutdown in progress */ 236 sys_shutdown = 1; 237 238 /* 239 * Communcate that init shouldn't be restarted. 240 */ 241 zone_shutdown_global(); 242 243 killall(ALL_ZONES); 244 /* 245 * If we are calling kadmin() from a kernel context then we 246 * do not release these resources. 247 */ 248 if (ttoproc(curthread) != &p0) { 249 VN_RELE(PTOU(curproc)->u_cdir); 250 if (PTOU(curproc)->u_rdir) 251 VN_RELE(PTOU(curproc)->u_rdir); 252 if (PTOU(curproc)->u_cwd) 253 refstr_rele(PTOU(curproc)->u_cwd); 254 255 PTOU(curproc)->u_cdir = rootdir; 256 PTOU(curproc)->u_rdir = NULL; 257 PTOU(curproc)->u_cwd = NULL; 258 } 259 260 /* 261 * Allow the reboot/halt/poweroff code a chance to do 262 * anything it needs to whilst we still have filesystems 263 * mounted, like loading any modules necessary for later 264 * performing the actual poweroff. 265 */ 266 if ((mdep != NULL) && (*(char *)mdep == '/')) { 267 buf = i_convert_boot_device_name(mdep, NULL, &buflen); 268 mdpreboot(cmd, fcn, buf); 269 } else 270 mdpreboot(cmd, fcn, mdep); 271 272 /* 273 * Allow fsflush to finish running and then prevent it 274 * from ever running again so that vfs_unmountall() and 275 * vfs_syncall() can acquire the vfs locks they need. 276 */ 277 sema_p(&fsflush_sema); 278 (void) callb_execute_class(CB_CL_UADMIN_PRE_VFS, NULL); 279 280 vfs_unmountall(); 281 (void) VFS_MOUNTROOT(rootvfs, ROOT_UNMOUNT); 282 vfs_syncall(); 283 284 dump_ereports(); 285 dump_messages(); 286 287 invoke_cb = B_TRUE; 288 289 /* FALLTHROUGH */ 290 } 291 292 case A_REBOOT: 293 if ((mdep != NULL) && (*(char *)mdep == '/')) { 294 buf = i_convert_boot_device_name(mdep, NULL, &buflen); 295 mdboot(cmd, fcn, buf, invoke_cb); 296 } else 297 mdboot(cmd, fcn, mdep, invoke_cb); 298 /* no return expected */ 299 break; 300 301 case A_REMOUNT: 302 (void) VFS_MOUNTROOT(rootvfs, ROOT_REMOUNT); 303 /* Let other threads enter the shutdown path now */ 304 mutex_enter(&ualock); 305 ua_shutdown_thread = NULL; 306 cv_signal(&uacond); 307 mutex_exit(&ualock); 308 break; 309 310 case A_FREEZE: 311 { 312 /* 313 * This is the entrypoint for all suspend/resume actions. 314 */ 315 extern int cpr(int, void *); 316 317 if (modload("misc", "cpr") == -1) 318 return (ENOTSUP); 319 /* Let the CPR module decide what to do with mdep */ 320 error = cpr(fcn, mdep); 321 break; 322 } 323 324 case A_FTRACE: 325 { 326 switch (fcn) { 327 case AD_FTRACE_START: 328 (void) FTRACE_START(); 329 break; 330 case AD_FTRACE_STOP: 331 (void) FTRACE_STOP(); 332 break; 333 default: 334 error = EINVAL; 335 } 336 break; 337 } 338 339 case A_DUMP: 340 { 341 if (fcn == AD_NOSYNC) { 342 in_sync = 1; 343 break; 344 } 345 346 panic_bootfcn = fcn; 347 panic_forced = 1; 348 349 if ((mdep != NULL) && (*(char *)mdep == '/')) { 350 panic_bootstr = i_convert_boot_device_name(mdep, 351 NULL, &buflen); 352 } else 353 panic_bootstr = mdep; 354 355 panic("forced crash dump initiated at user request"); 356 /*NOTREACHED*/ 357 } 358 359 default: 360 error = EINVAL; 361 } 362 363 return (error); 364 } 365 366 int 367 uadmin(int cmd, int fcn, uintptr_t mdep) 368 { 369 int error = 0, rv = 0; 370 size_t nbytes = 0; 371 cred_t *credp = CRED(); 372 char *bootargs = NULL; 373 374 /* 375 * The swapctl system call doesn't have its own entry point: it uses 376 * uadmin as a wrapper so we just call it directly from here. 377 */ 378 if (cmd == A_SWAPCTL) { 379 if (get_udatamodel() == DATAMODEL_NATIVE) 380 error = swapctl(fcn, (void *)mdep, &rv); 381 #if defined(_SYSCALL32_IMPL) 382 else 383 error = swapctl32(fcn, (void *)mdep, &rv); 384 #endif /* _SYSCALL32_IMPL */ 385 return (error ? set_errno(error) : rv); 386 } 387 388 /* 389 * Certain subcommands intepret a non-NULL mdep value as a pointer to 390 * a boot string. We pull that in as bootargs, if applicable. 391 */ 392 if (mdep != NULL && 393 (cmd == A_SHUTDOWN || cmd == A_REBOOT || cmd == A_DUMP || 394 cmd == A_FREEZE)) { 395 bootargs = kmem_zalloc(BOOTARGS_MAX, KM_SLEEP); 396 if ((error = copyinstr((const char *)mdep, bootargs, 397 BOOTARGS_MAX, &nbytes)) != 0) { 398 kmem_free(bootargs, BOOTARGS_MAX); 399 return (set_errno(error)); 400 } 401 } 402 403 /* 404 * Invoke the appropriate kadmin() routine. 405 */ 406 if (getzoneid() != GLOBAL_ZONEID) 407 error = zone_kadmin(cmd, fcn, bootargs, credp); 408 else 409 error = kadmin(cmd, fcn, bootargs, credp); 410 411 if (bootargs != NULL) 412 kmem_free(bootargs, BOOTARGS_MAX); 413 return (error ? set_errno(error) : 0); 414 } 415