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 2004 Sun Microsystems, Inc. All rights reserved. 24 * Use is subject to license terms. 25 */ 26 27 28 #pragma ident "%Z%%M% %I% %E% SMI" 29 30 #include <sys/param.h> 31 #include <sys/types.h> 32 #include <sys/sysmacros.h> 33 #include <sys/systm.h> 34 #include <sys/errno.h> 35 #include <sys/vfs.h> 36 #include <sys/vnode.h> 37 #include <sys/swap.h> 38 #include <sys/file.h> 39 #include <sys/proc.h> 40 #include <sys/var.h> 41 #include <sys/uadmin.h> 42 #include <sys/signal.h> 43 #include <sys/time.h> 44 #include <vm/seg_kmem.h> 45 #include <sys/modctl.h> 46 #include <sys/callb.h> 47 #include <sys/dumphdr.h> 48 #include <sys/debug.h> 49 #include <sys/ftrace.h> 50 #include <sys/cmn_err.h> 51 #include <sys/panic.h> 52 #include <sys/ddi.h> 53 #include <sys/sunddi.h> 54 #include <sys/policy.h> 55 #include <sys/zone.h> 56 57 /* 58 * Administrivia system call. We provide this in two flavors: one for calling 59 * from the system call path (uadmin), and the other for calling from elsewhere 60 * within the kernel (kadmin). Callers must beware that certain uadmin cmd 61 * values (specifically A_SWAPCTL) are only supported by uadmin and not kadmin. 62 */ 63 64 extern ksema_t fsflush_sema; 65 kmutex_t ualock; 66 67 68 /* 69 * Kill all user processes in said zone. A special argument of ALL_ZONES is 70 * passed in when the system as a whole is shutting down. The lack of per-zone 71 * process lists is likely to make the following a performance bottleneck on a 72 * system with many zones. 73 */ 74 void 75 killall(zoneid_t zoneid) 76 { 77 proc_t *p; 78 79 ASSERT(zoneid != GLOBAL_ZONEID); 80 /* 81 * Kill all processes except kernel daemons and ourself. 82 * Make a first pass to stop all processes so they won't 83 * be trying to restart children as we kill them. 84 */ 85 mutex_enter(&pidlock); 86 for (p = practive; p != NULL; p = p->p_next) { 87 if ((zoneid == ALL_ZONES || p->p_zone->zone_id == zoneid) && 88 p->p_exec != NULLVP && /* kernel daemons */ 89 p->p_as != &kas && 90 p->p_stat != SZOMB) { 91 mutex_enter(&p->p_lock); 92 p->p_flag |= SNOWAIT; 93 sigtoproc(p, NULL, SIGSTOP); 94 mutex_exit(&p->p_lock); 95 } 96 } 97 p = practive; 98 while (p != NULL) { 99 if ((zoneid == ALL_ZONES || p->p_zone->zone_id == zoneid) && 100 p->p_exec != NULLVP && /* kernel daemons */ 101 p->p_as != &kas && 102 p->p_stat != SIDL && 103 p->p_stat != SZOMB) { 104 mutex_enter(&p->p_lock); 105 if (sigismember(&p->p_sig, SIGKILL)) { 106 mutex_exit(&p->p_lock); 107 p = p->p_next; 108 } else { 109 sigtoproc(p, NULL, SIGKILL); 110 mutex_exit(&p->p_lock); 111 (void) cv_timedwait(&p->p_srwchan_cv, 112 &pidlock, lbolt + hz); 113 p = practive; 114 } 115 } else { 116 p = p->p_next; 117 } 118 } 119 mutex_exit(&pidlock); 120 } 121 122 int 123 kadmin(int cmd, int fcn, void *mdep, cred_t *credp) 124 { 125 int error = 0; 126 int locked = 0; 127 char *buf; 128 size_t buflen = 0; 129 130 /* 131 * We might be called directly by the kernel's fault-handling code, so 132 * we can't assert that the caller is in the global zone. 133 */ 134 135 /* 136 * Make sure that cmd is one of the valid <sys/uadmin.h> command codes 137 * and that we have appropriate privileges for this action. 138 */ 139 switch (cmd) { 140 case A_FTRACE: 141 case A_SHUTDOWN: 142 case A_REBOOT: 143 case A_REMOUNT: 144 case A_FREEZE: 145 case A_DUMP: 146 if (secpolicy_sys_config(credp, B_FALSE) != 0) 147 return (EPERM); 148 break; 149 150 default: 151 return (EINVAL); 152 } 153 154 /* 155 * Serialize these operations on ualock. If it is held, just return 156 * as if successful since the system will soon reset or remount. 157 */ 158 if (cmd == A_SHUTDOWN || cmd == A_REBOOT || cmd == A_REMOUNT) { 159 if (!mutex_tryenter(&ualock)) 160 return (0); 161 locked = 1; 162 } 163 164 switch (cmd) { 165 case A_SHUTDOWN: 166 { 167 proc_t *p = ttoproc(curthread); 168 169 /* 170 * Release (almost) all of our own resources if we are called 171 * from a user context, however if we are calling kadmin() from 172 * a kernel context then we do not release these resources. 173 */ 174 if (ttoproc(curthread) != &p0) { 175 if ((error = exitlwps(0)) != 0) 176 return (error); 177 mutex_enter(&p->p_lock); 178 p->p_flag |= SNOWAIT; 179 sigfillset(&p->p_ignore); 180 curthread->t_lwp->lwp_cursig = 0; 181 curthread->t_lwp->lwp_extsig = 0; 182 if (p->p_exec) { 183 vnode_t *exec_vp = p->p_exec; 184 p->p_exec = NULLVP; 185 mutex_exit(&p->p_lock); 186 VN_RELE(exec_vp); 187 } else { 188 mutex_exit(&p->p_lock); 189 } 190 191 pollcleanup(); 192 closeall(P_FINFO(curproc)); 193 relvm(); 194 195 } else { 196 /* 197 * Reset t_cred if not set because much of the 198 * filesystem code depends on CRED() being valid. 199 */ 200 if (curthread->t_cred == NULL) 201 curthread->t_cred = kcred; 202 } 203 204 /* 205 * Communcate that init shouldn't be restarted. 206 */ 207 zone_shutdown_global(); 208 209 killall(ALL_ZONES); 210 /* 211 * If we are calling kadmin() from a kernel context then we 212 * do not release these resources. 213 */ 214 if (ttoproc(curthread) != &p0) { 215 VN_RELE(u.u_cdir); 216 if (u.u_rdir) 217 VN_RELE(u.u_rdir); 218 if (u.u_cwd) 219 refstr_rele(u.u_cwd); 220 221 u.u_cdir = rootdir; 222 u.u_rdir = NULL; 223 u.u_cwd = NULL; 224 } 225 226 /* 227 * Allow the reboot/halt/poweroff code a chance to do 228 * anything it needs to whilst we still have filesystems 229 * mounted, like loading any modules necessary for later 230 * performing the actual poweroff. 231 */ 232 if ((mdep != NULL) && (*(char *)mdep == '/')) { 233 buf = i_convert_boot_device_name(mdep, NULL, &buflen); 234 mdpreboot(cmd, fcn, buf); 235 } else 236 mdpreboot(cmd, fcn, mdep); 237 238 /* 239 * Allow fsflush to finish running and then prevent it 240 * from ever running again so that vfs_unmountall() and 241 * vfs_syncall() can acquire the vfs locks they need. 242 */ 243 sema_p(&fsflush_sema); 244 (void) callb_execute_class(CB_CL_UADMIN_PRE_VFS, NULL); 245 246 vfs_unmountall(); 247 (void) VFS_MOUNTROOT(rootvfs, ROOT_UNMOUNT); 248 vfs_syncall(); 249 250 (void) callb_execute_class(CB_CL_UADMIN_POST_VFS, NULL); 251 dump_ereports(); 252 dump_messages(); 253 254 /* FALLTHROUGH */ 255 } 256 257 case A_REBOOT: 258 if ((mdep != NULL) && (*(char *)mdep == '/')) { 259 buf = i_convert_boot_device_name(mdep, NULL, &buflen); 260 mdboot(cmd, fcn, buf); 261 } else 262 mdboot(cmd, fcn, mdep); 263 /* no return expected */ 264 break; 265 266 case A_REMOUNT: 267 (void) VFS_MOUNTROOT(rootvfs, ROOT_REMOUNT); 268 break; 269 270 case A_FREEZE: 271 { 272 /* XXX: declare in some header file */ 273 extern int cpr(int); 274 275 if (modload("misc", "cpr") == -1) 276 return (ENOTSUP); 277 error = cpr(fcn); 278 break; 279 } 280 281 case A_FTRACE: 282 { 283 switch (fcn) { 284 case AD_FTRACE_START: 285 (void) FTRACE_START(); 286 break; 287 case AD_FTRACE_STOP: 288 (void) FTRACE_STOP(); 289 break; 290 default: 291 error = EINVAL; 292 } 293 break; 294 } 295 296 case A_DUMP: 297 { 298 if (fcn == AD_NOSYNC) { 299 in_sync = 1; 300 break; 301 } 302 303 panic_bootfcn = fcn; 304 panic_forced = 1; 305 306 if ((mdep != NULL) && (*(char *)mdep == '/')) { 307 panic_bootstr = i_convert_boot_device_name(mdep, 308 NULL, &buflen); 309 } else 310 panic_bootstr = mdep; 311 312 panic("forced crash dump initiated at user request"); 313 /*NOTREACHED*/ 314 } 315 316 default: 317 error = EINVAL; 318 } 319 320 if (locked) 321 mutex_exit(&ualock); 322 323 return (error); 324 } 325 326 int 327 uadmin(int cmd, int fcn, uintptr_t mdep) 328 { 329 int error = 0, rv = 0; 330 size_t nbytes = 0; 331 char buf[257]; 332 cred_t *credp = CRED(); 333 334 /* 335 * The swapctl system call doesn't have its own entry point: it uses 336 * uadmin as a wrapper so we just call it directly from here. 337 */ 338 if (cmd == A_SWAPCTL) { 339 if (get_udatamodel() == DATAMODEL_NATIVE) 340 error = swapctl(fcn, (void *)mdep, &rv); 341 #if defined(_SYSCALL32_IMPL) 342 else 343 error = swapctl32(fcn, (void *)mdep, &rv); 344 #endif /* _SYSCALL32_IMPL */ 345 return (error ? set_errno(error) : rv); 346 } 347 348 /* 349 * Handle zones. 350 */ 351 if (getzoneid() != GLOBAL_ZONEID) { 352 error = zone_uadmin(cmd, fcn, credp); 353 return (error ? set_errno(error) : 0); 354 } 355 356 /* 357 * Certain subcommands intepret a non-NULL mdep value as a pointer to 358 * a boot string. Attempt to copy it in now, or reset mdep to NULL. 359 */ 360 if (cmd == A_SHUTDOWN || cmd == A_REBOOT || cmd == A_DUMP) { 361 if (mdep != NULL && copyinstr((const char *)mdep, buf, 362 sizeof (buf) - 1, &nbytes) == 0) { 363 buf[nbytes] = '\0'; 364 mdep = (uintptr_t)buf; 365 } else 366 mdep = NULL; 367 } 368 369 if ((error = kadmin(cmd, fcn, (void *)mdep, credp)) != 0) 370 return (set_errno(error)); 371 372 return (0); 373 } 374