rump.c revision 1.101
1 /* $NetBSD: rump.c,v 1.101 2009/04/16 14:07:18 pooka Exp $ */ 2 3 /* 4 * Copyright (c) 2007 Antti Kantee. All Rights Reserved. 5 * 6 * Development of this software was supported by Google Summer of Code. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS 18 * OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED 19 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE 20 * DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR 23 * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 */ 29 30 #include <sys/cdefs.h> 31 __KERNEL_RCSID(0, "$NetBSD: rump.c,v 1.101 2009/04/16 14:07:18 pooka Exp $"); 32 33 #include <sys/param.h> 34 #include <sys/atomic.h> 35 #include <sys/buf.h> 36 #include <sys/callout.h> 37 #include <sys/conf.h> 38 #include <sys/cpu.h> 39 #include <sys/evcnt.h> 40 #include <sys/event.h> 41 #include <sys/filedesc.h> 42 #include <sys/iostat.h> 43 #include <sys/kauth.h> 44 #include <sys/kernel.h> 45 #include <sys/kmem.h> 46 #include <sys/kprintf.h> 47 #include <sys/ksyms.h> 48 #include <sys/msgbuf.h> 49 #include <sys/module.h> 50 #include <sys/once.h> 51 #include <sys/percpu.h> 52 #include <sys/queue.h> 53 #include <sys/resourcevar.h> 54 #include <sys/select.h> 55 #include <sys/sysctl.h> 56 #include <sys/syscall.h> 57 #include <sys/tty.h> 58 #include <sys/uidinfo.h> 59 #include <sys/vmem.h> 60 61 #include <rump/rumpuser.h> 62 63 #include <secmodel/secmodel.h> 64 65 #include "rump_private.h" 66 #include "rump_net_private.h" 67 #include "rump_vfs_private.h" 68 69 struct proc proc0; 70 struct session rump_session = { 71 .s_count = 1, 72 .s_flags = 0, 73 .s_leader = &proc0, 74 .s_login = "rumphobo", 75 .s_sid = 0, 76 }; 77 struct pgrp rump_pgrp = { 78 .pg_members = LIST_HEAD_INITIALIZER(pg_members), 79 .pg_session = &rump_session, 80 .pg_jobc = 1, 81 }; 82 struct pstats rump_stats; 83 struct plimit rump_limits; 84 struct cpu_info rump_cpu; 85 struct filedesc rump_filedesc0; 86 struct proclist allproc; 87 char machine[] = "rump"; 88 static kauth_cred_t rump_susercred; 89 90 /* pretend the master rump proc is init */ 91 struct proc *initproc = &proc0; 92 93 struct rumpuser_mtx *rump_giantlock; 94 95 sigset_t sigcantmask; 96 97 #ifdef RUMP_WITHOUT_THREADS 98 int rump_threads = 0; 99 #else 100 int rump_threads = 1; 101 #endif 102 103 static void 104 rump_aiodone_worker(struct work *wk, void *dummy) 105 { 106 struct buf *bp = (struct buf *)wk; 107 108 KASSERT(&bp->b_work == wk); 109 bp->b_iodone(bp); 110 } 111 112 static int rump_inited; 113 static struct emul emul_rump; 114 115 void rump__unavailable(void); 116 void rump__unavailable() {} 117 __weak_alias(rump_net_init,rump__unavailable); 118 __weak_alias(rump_vfs_init,rump__unavailable); 119 120 void rump__unavailable_vfs_panic(void); 121 void rump__unavailable_vfs_panic() {panic("vfs component not available");} 122 __weak_alias(vn_open,rump__unavailable_vfs_panic); 123 __weak_alias(vn_rdwr,rump__unavailable_vfs_panic); 124 __weak_alias(vn_close,rump__unavailable_vfs_panic); 125 126 static void 127 pvfsinit_nop(struct proc *p) 128 { 129 130 return; 131 } 132 133 static void 134 pvfsrele_nop(struct proc *p) 135 { 136 137 return; 138 } 139 140 rump_proc_vfs_init_fn rump_proc_vfs_init = pvfsinit_nop; 141 rump_proc_vfs_release_fn rump_proc_vfs_release = pvfsrele_nop; 142 143 int 144 rump__init(int rump_version) 145 { 146 char buf[256]; 147 struct proc *p; 148 struct lwp *l; 149 int error; 150 151 /* XXX */ 152 if (rump_inited) 153 return 0; 154 rump_inited = 1; 155 156 if (rump_version != RUMP_VERSION) { 157 printf("rump version mismatch, %d vs. %d\n", 158 rump_version, RUMP_VERSION); 159 return EPROGMISMATCH; 160 } 161 162 if (rumpuser_getenv("RUMP_THREADS", buf, sizeof(buf), &error) == 0) { 163 rump_threads = *buf != '0'; 164 } 165 rumpuser_thrinit(_kernel_lock, _kernel_unlock, rump_threads); 166 167 mutex_init(&tty_lock, MUTEX_DEFAULT, IPL_NONE); 168 rumpuser_mutex_recursive_init(&rump_giantlock); 169 ksyms_init(); 170 rumpvm_init(); 171 evcnt_init(); 172 173 once_init(); 174 175 rump_sleepers_init(); 176 #ifdef RUMP_USE_REAL_ALLOCATORS 177 pool_subsystem_init(); 178 kmem_init(); 179 #endif 180 kprintf_init(); 181 loginit(); 182 183 kauth_init(); 184 rump_susercred = rump_cred_create(0, 0, 0, NULL); 185 186 l = &lwp0; 187 p = &proc0; 188 p->p_stats = &rump_stats; 189 p->p_limit = &rump_limits; 190 p->p_pgrp = &rump_pgrp; 191 p->p_pid = 0; 192 p->p_fd = &rump_filedesc0; 193 p->p_vmspace = &rump_vmspace; 194 p->p_emul = &emul_rump; 195 p->p_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 196 l->l_cred = rump_cred_suserget(); 197 l->l_proc = p; 198 l->l_lid = 1; 199 LIST_INIT(&allproc); 200 proc_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 201 202 rump_limits.pl_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY; 203 rump_limits.pl_rlimit[RLIMIT_NOFILE].rlim_cur = RLIM_INFINITY; 204 rump_limits.pl_rlimit[RLIMIT_SBSIZE].rlim_cur = RLIM_INFINITY; 205 206 callout_startup(); 207 callout_init_cpu(&rump_cpu); 208 209 kqueue_init(); 210 iostat_init(); 211 uid_init(); 212 percpu_init(); 213 fd_sys_init(); 214 module_init(); 215 sysctl_init(); 216 softint_init(&rump_cpu); 217 cold = 0; 218 devsw_init(); 219 secmodel_start(); 220 221 /* these do nothing if not present */ 222 rump_vfs_init(); 223 rump_net_init(); 224 225 /* aieeeedondest */ 226 if (rump_threads) { 227 if (workqueue_create(&uvm.aiodone_queue, "aiodoned", 228 rump_aiodone_worker, NULL, 0, 0, 0)) 229 panic("aiodoned"); 230 } 231 232 rumpuser_gethostname(hostname, MAXHOSTNAMELEN, &error); 233 hostnamelen = strlen(hostname); 234 235 sigemptyset(&sigcantmask); 236 237 lwp0.l_fd = proc0.p_fd = fd_init(&rump_filedesc0); 238 239 #ifdef RUMP_USE_REAL_ALLOCATORS 240 if (rump_threads) 241 vmem_rehash_start(); 242 #endif 243 244 return 0; 245 } 246 247 struct uio * 248 rump_uio_setup(void *buf, size_t bufsize, off_t offset, enum rump_uiorw rw) 249 { 250 struct uio *uio; 251 enum uio_rw uiorw; 252 253 switch (rw) { 254 case RUMPUIO_READ: 255 uiorw = UIO_READ; 256 break; 257 case RUMPUIO_WRITE: 258 uiorw = UIO_WRITE; 259 break; 260 default: 261 panic("%s: invalid rw %d", __func__, rw); 262 } 263 264 uio = kmem_alloc(sizeof(struct uio), KM_SLEEP); 265 uio->uio_iov = kmem_alloc(sizeof(struct iovec), KM_SLEEP); 266 267 uio->uio_iov->iov_base = buf; 268 uio->uio_iov->iov_len = bufsize; 269 270 uio->uio_iovcnt = 1; 271 uio->uio_offset = offset; 272 uio->uio_resid = bufsize; 273 uio->uio_rw = uiorw; 274 uio->uio_vmspace = UIO_VMSPACE_SYS; 275 276 return uio; 277 } 278 279 size_t 280 rump_uio_getresid(struct uio *uio) 281 { 282 283 return uio->uio_resid; 284 } 285 286 off_t 287 rump_uio_getoff(struct uio *uio) 288 { 289 290 return uio->uio_offset; 291 } 292 293 size_t 294 rump_uio_free(struct uio *uio) 295 { 296 size_t resid; 297 298 resid = uio->uio_resid; 299 kmem_free(uio->uio_iov, sizeof(*uio->uio_iov)); 300 kmem_free(uio, sizeof(*uio)); 301 302 return resid; 303 } 304 305 struct lwp * 306 rump_setup_curlwp(pid_t pid, lwpid_t lid, int set) 307 { 308 struct lwp *l; 309 struct proc *p; 310 311 l = kmem_zalloc(sizeof(struct lwp), KM_SLEEP); 312 if (pid != 0) { 313 p = kmem_zalloc(sizeof(struct proc), KM_SLEEP); 314 rump_proc_vfs_init(p); 315 p->p_stats = &rump_stats; 316 p->p_limit = &rump_limits; 317 p->p_pid = pid; 318 p->p_vmspace = &rump_vmspace; 319 p->p_fd = fd_init(NULL); 320 } else { 321 p = &proc0; 322 } 323 324 l->l_cred = rump_cred_suserget(); 325 l->l_proc = p; 326 l->l_lid = lid; 327 l->l_fd = p->p_fd; 328 l->l_mutex = RUMP_LMUTEX_MAGIC; 329 l->l_cpu = &rump_cpu; 330 331 if (set) 332 rumpuser_set_curlwp(l); 333 334 return l; 335 } 336 337 void 338 rump_clear_curlwp(void) 339 { 340 struct lwp *l; 341 342 l = rumpuser_get_curlwp(); 343 if (l->l_proc->p_pid != 0) { 344 fd_free(); 345 rump_proc_vfs_release(l->l_proc); 346 rump_cred_destroy(l->l_cred); 347 kmem_free(l->l_proc, sizeof(*l->l_proc)); 348 } 349 kmem_free(l, sizeof(*l)); 350 rumpuser_set_curlwp(NULL); 351 } 352 353 struct lwp * 354 rump_get_curlwp(void) 355 { 356 struct lwp *l; 357 358 l = rumpuser_get_curlwp(); 359 if (l == NULL) 360 l = &lwp0; 361 362 return l; 363 } 364 365 kauth_cred_t 366 rump_cred_create(uid_t uid, gid_t gid, size_t ngroups, gid_t *groups) 367 { 368 kauth_cred_t cred; 369 int rv; 370 371 cred = kauth_cred_alloc(); 372 kauth_cred_setuid(cred, uid); 373 kauth_cred_seteuid(cred, uid); 374 kauth_cred_setsvuid(cred, uid); 375 kauth_cred_setgid(cred, gid); 376 kauth_cred_setgid(cred, gid); 377 kauth_cred_setegid(cred, gid); 378 kauth_cred_setsvgid(cred, gid); 379 rv = kauth_cred_setgroups(cred, groups, ngroups, 0, UIO_SYSSPACE); 380 /* oh this is silly. and by "this" I mean kauth_cred_setgroups() */ 381 assert(rv == 0); 382 383 return cred; 384 } 385 386 void 387 rump_cred_destroy(kauth_cred_t cred) 388 { 389 390 kauth_cred_free(cred); 391 } 392 393 kauth_cred_t 394 rump_cred_suserget(void) 395 { 396 397 kauth_cred_hold(rump_susercred); 398 return rump_susercred; 399 } 400 401 /* 402 * Return the next system lwpid 403 */ 404 lwpid_t 405 rump_nextlid(void) 406 { 407 lwpid_t retid; 408 409 mutex_enter(proc0.p_lock); 410 /* 411 * Take next one, don't return 0 412 * XXX: most likely we'll have collisions in case this 413 * wraps around. 414 */ 415 if (++proc0.p_nlwpid == 0) 416 ++proc0.p_nlwpid; 417 retid = proc0.p_nlwpid; 418 mutex_exit(proc0.p_lock); 419 420 return retid; 421 } 422 423 int 424 rump_module_load(struct modinfo **mi) 425 { 426 427 if (!module_compatible((*mi)->mi_version, __NetBSD_Version__)) 428 return EPROGMISMATCH; 429 430 return (*mi)->mi_modcmd(MODULE_CMD_INIT, NULL); 431 } 432 433 int _syspuffs_stub(int, int *); 434 int 435 _syspuffs_stub(int fd, int *newfd) 436 { 437 438 return ENODEV; 439 } 440 __weak_alias(rump_syspuffs_glueinit,_syspuffs_stub); 441 442 static int 443 rump_sysproxy_local(int num, void *arg, uint8_t *data, size_t dlen, 444 register_t *retval) 445 { 446 struct lwp *l; 447 struct sysent *callp; 448 449 if (__predict_false(num >= SYS_NSYSENT)) 450 return ENOSYS; 451 452 l = curlwp; 453 callp = rump_sysent + num; 454 return callp->sy_call(l, (void *)data, retval); 455 } 456 457 rump_sysproxy_t rump_sysproxy = rump_sysproxy_local; 458 void *rump_sysproxy_arg; 459 460 /* 461 * This whole syscall-via-rpc is still taking form. For example, it 462 * may be necessary to set syscalls individually instead of lobbing 463 * them all to the same place. So don't think this interface is 464 * set in stone. 465 */ 466 int 467 rump_sysproxy_set(rump_sysproxy_t proxy, void *arg) 468 { 469 470 if (rump_sysproxy_arg) 471 return EBUSY; 472 473 rump_sysproxy_arg = arg; 474 rump_sysproxy = proxy; 475 476 return 0; 477 } 478
Indexes created Mon Sep 29 21:09:56 GMT 2025