rump.c revision 1.100
1 /* $NetBSD: rump.c,v 1.100 2009/03/29 18:22:08 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 1.100 2009/03/29 18:22:08 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 "rump_private.h" 64 #include "rump_net_private.h" 65 #include "rump_vfs_private.h" 66 67 struct proc proc0; 68 struct session rump_session = { 69 .s_count = 1, 70 .s_flags = 0, 71 .s_leader = &proc0, 72 .s_login = "rumphobo", 73 .s_sid = 0, 74 }; 75 struct pgrp rump_pgrp = { 76 .pg_members = LIST_HEAD_INITIALIZER(pg_members), 77 .pg_session = &rump_session, 78 .pg_jobc = 1, 79 }; 80 struct pstats rump_stats; 81 struct plimit rump_limits; 82 struct cpu_info rump_cpu; 83 struct filedesc rump_filedesc0; 84 struct proclist allproc; 85 char machine[] = "rump"; 86 static kauth_cred_t rump_susercred; 87 88 struct rumpuser_mtx *rump_giantlock; 89 90 sigset_t sigcantmask; 91 92 #ifdef RUMP_WITHOUT_THREADS 93 int rump_threads = 0; 94 #else 95 int rump_threads = 1; 96 #endif 97 98 static void 99 rump_aiodone_worker(struct work *wk, void *dummy) 100 { 101 struct buf *bp = (struct buf *)wk; 102 103 KASSERT(&bp->b_work == wk); 104 bp->b_iodone(bp); 105 } 106 107 static int rump_inited; 108 static struct emul emul_rump; 109 110 void rump__unavailable(void); 111 void rump__unavailable() {} 112 __weak_alias(rump_net_init,rump__unavailable); 113 __weak_alias(rump_vfs_init,rump__unavailable); 114 115 void rump__unavailable_vfs_panic(void); 116 void rump__unavailable_vfs_panic() {panic("vfs component not available");} 117 __weak_alias(vn_open,rump__unavailable_vfs_panic); 118 __weak_alias(vn_rdwr,rump__unavailable_vfs_panic); 119 __weak_alias(vn_close,rump__unavailable_vfs_panic); 120 121 static void 122 pvfsinit_nop(struct proc *p) 123 { 124 125 return; 126 } 127 128 static void 129 pvfsrele_nop(struct proc *p) 130 { 131 132 return; 133 } 134 135 rump_proc_vfs_init_fn rump_proc_vfs_init = pvfsinit_nop; 136 rump_proc_vfs_release_fn rump_proc_vfs_release = pvfsrele_nop; 137 138 int 139 rump__init(int rump_version) 140 { 141 char buf[256]; 142 struct proc *p; 143 struct lwp *l; 144 int error; 145 146 /* XXX */ 147 if (rump_inited) 148 return 0; 149 rump_inited = 1; 150 151 if (rump_version != RUMP_VERSION) { 152 printf("rump version mismatch, %d vs. %d\n", 153 rump_version, RUMP_VERSION); 154 return EPROGMISMATCH; 155 } 156 157 if (rumpuser_getenv("RUMP_THREADS", buf, sizeof(buf), &error) == 0) { 158 rump_threads = *buf != '0'; 159 } 160 rumpuser_thrinit(_kernel_lock, _kernel_unlock, rump_threads); 161 162 mutex_init(&tty_lock, MUTEX_DEFAULT, IPL_NONE); 163 rumpuser_mutex_recursive_init(&rump_giantlock); 164 ksyms_init(); 165 rumpvm_init(); 166 evcnt_init(); 167 168 once_init(); 169 170 rump_sleepers_init(); 171 #ifdef RUMP_USE_REAL_ALLOCATORS 172 pool_subsystem_init(); 173 kmem_init(); 174 #endif 175 kprintf_init(); 176 loginit(); 177 178 kauth_init(); 179 rump_susercred = rump_cred_create(0, 0, 0, NULL); 180 181 l = &lwp0; 182 p = &proc0; 183 p->p_stats = &rump_stats; 184 p->p_limit = &rump_limits; 185 p->p_pgrp = &rump_pgrp; 186 p->p_pid = 0; 187 p->p_fd = &rump_filedesc0; 188 p->p_vmspace = &rump_vmspace; 189 p->p_emul = &emul_rump; 190 p->p_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 191 l->l_cred = rump_cred_suserget(); 192 l->l_proc = p; 193 l->l_lid = 1; 194 LIST_INIT(&allproc); 195 proc_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 196 197 rump_limits.pl_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY; 198 rump_limits.pl_rlimit[RLIMIT_NOFILE].rlim_cur = RLIM_INFINITY; 199 rump_limits.pl_rlimit[RLIMIT_SBSIZE].rlim_cur = RLIM_INFINITY; 200 201 callout_startup(); 202 callout_init_cpu(&rump_cpu); 203 204 kqueue_init(); 205 iostat_init(); 206 uid_init(); 207 percpu_init(); 208 fd_sys_init(); 209 module_init(); 210 sysctl_init(); 211 softint_init(&rump_cpu); 212 cold = 0; 213 devsw_init(); 214 215 /* these do nothing if not present */ 216 rump_vfs_init(); 217 rump_net_init(); 218 219 /* aieeeedondest */ 220 if (rump_threads) { 221 if (workqueue_create(&uvm.aiodone_queue, "aiodoned", 222 rump_aiodone_worker, NULL, 0, 0, 0)) 223 panic("aiodoned"); 224 } 225 226 rumpuser_gethostname(hostname, MAXHOSTNAMELEN, &error); 227 hostnamelen = strlen(hostname); 228 229 sigemptyset(&sigcantmask); 230 231 lwp0.l_fd = proc0.p_fd = fd_init(&rump_filedesc0); 232 233 #ifdef RUMP_USE_REAL_ALLOCATORS 234 if (rump_threads) 235 vmem_rehash_start(); 236 #endif 237 238 return 0; 239 } 240 241 struct uio * 242 rump_uio_setup(void *buf, size_t bufsize, off_t offset, enum rump_uiorw rw) 243 { 244 struct uio *uio; 245 enum uio_rw uiorw; 246 247 switch (rw) { 248 case RUMPUIO_READ: 249 uiorw = UIO_READ; 250 break; 251 case RUMPUIO_WRITE: 252 uiorw = UIO_WRITE; 253 break; 254 default: 255 panic("%s: invalid rw %d", __func__, rw); 256 } 257 258 uio = kmem_alloc(sizeof(struct uio), KM_SLEEP); 259 uio->uio_iov = kmem_alloc(sizeof(struct iovec), KM_SLEEP); 260 261 uio->uio_iov->iov_base = buf; 262 uio->uio_iov->iov_len = bufsize; 263 264 uio->uio_iovcnt = 1; 265 uio->uio_offset = offset; 266 uio->uio_resid = bufsize; 267 uio->uio_rw = uiorw; 268 uio->uio_vmspace = UIO_VMSPACE_SYS; 269 270 return uio; 271 } 272 273 size_t 274 rump_uio_getresid(struct uio *uio) 275 { 276 277 return uio->uio_resid; 278 } 279 280 off_t 281 rump_uio_getoff(struct uio *uio) 282 { 283 284 return uio->uio_offset; 285 } 286 287 size_t 288 rump_uio_free(struct uio *uio) 289 { 290 size_t resid; 291 292 resid = uio->uio_resid; 293 kmem_free(uio->uio_iov, sizeof(*uio->uio_iov)); 294 kmem_free(uio, sizeof(*uio)); 295 296 return resid; 297 } 298 299 struct lwp * 300 rump_setup_curlwp(pid_t pid, lwpid_t lid, int set) 301 { 302 struct lwp *l; 303 struct proc *p; 304 305 l = kmem_zalloc(sizeof(struct lwp), KM_SLEEP); 306 if (pid != 0) { 307 p = kmem_zalloc(sizeof(struct proc), KM_SLEEP); 308 rump_proc_vfs_init(p); 309 p->p_stats = &rump_stats; 310 p->p_limit = &rump_limits; 311 p->p_pid = pid; 312 p->p_vmspace = &rump_vmspace; 313 p->p_fd = fd_init(NULL); 314 } else { 315 p = &proc0; 316 } 317 318 l->l_cred = rump_cred_suserget(); 319 l->l_proc = p; 320 l->l_lid = lid; 321 l->l_fd = p->p_fd; 322 l->l_mutex = RUMP_LMUTEX_MAGIC; 323 l->l_cpu = &rump_cpu; 324 325 if (set) 326 rumpuser_set_curlwp(l); 327 328 return l; 329 } 330 331 void 332 rump_clear_curlwp(void) 333 { 334 struct lwp *l; 335 336 l = rumpuser_get_curlwp(); 337 if (l->l_proc->p_pid != 0) { 338 fd_free(); 339 rump_proc_vfs_release(l->l_proc); 340 rump_cred_destroy(l->l_cred); 341 kmem_free(l->l_proc, sizeof(*l->l_proc)); 342 } 343 kmem_free(l, sizeof(*l)); 344 rumpuser_set_curlwp(NULL); 345 } 346 347 struct lwp * 348 rump_get_curlwp(void) 349 { 350 struct lwp *l; 351 352 l = rumpuser_get_curlwp(); 353 if (l == NULL) 354 l = &lwp0; 355 356 return l; 357 } 358 359 kauth_cred_t 360 rump_cred_create(uid_t uid, gid_t gid, size_t ngroups, gid_t *groups) 361 { 362 kauth_cred_t cred; 363 int rv; 364 365 cred = kauth_cred_alloc(); 366 kauth_cred_setuid(cred, uid); 367 kauth_cred_seteuid(cred, uid); 368 kauth_cred_setsvuid(cred, uid); 369 kauth_cred_setgid(cred, gid); 370 kauth_cred_setgid(cred, gid); 371 kauth_cred_setegid(cred, gid); 372 kauth_cred_setsvgid(cred, gid); 373 rv = kauth_cred_setgroups(cred, groups, ngroups, 0, UIO_SYSSPACE); 374 /* oh this is silly. and by "this" I mean kauth_cred_setgroups() */ 375 assert(rv == 0); 376 377 return cred; 378 } 379 380 void 381 rump_cred_destroy(kauth_cred_t cred) 382 { 383 384 kauth_cred_free(cred); 385 } 386 387 kauth_cred_t 388 rump_cred_suserget(void) 389 { 390 391 kauth_cred_hold(rump_susercred); 392 return rump_susercred; 393 } 394 395 /* 396 * Return the next system lwpid 397 */ 398 lwpid_t 399 rump_nextlid(void) 400 { 401 lwpid_t retid; 402 403 mutex_enter(proc0.p_lock); 404 /* 405 * Take next one, don't return 0 406 * XXX: most likely we'll have collisions in case this 407 * wraps around. 408 */ 409 if (++proc0.p_nlwpid == 0) 410 ++proc0.p_nlwpid; 411 retid = proc0.p_nlwpid; 412 mutex_exit(proc0.p_lock); 413 414 return retid; 415 } 416 417 int 418 rump_module_load(struct modinfo **mi) 419 { 420 421 if (!module_compatible((*mi)->mi_version, __NetBSD_Version__)) 422 return EPROGMISMATCH; 423 424 return (*mi)->mi_modcmd(MODULE_CMD_INIT, NULL); 425 } 426 427 int _syspuffs_stub(int, int *); 428 int 429 _syspuffs_stub(int fd, int *newfd) 430 { 431 432 return ENODEV; 433 } 434 __weak_alias(rump_syspuffs_glueinit,_syspuffs_stub); 435 436 static int 437 rump_sysproxy_local(int num, void *arg, uint8_t *data, size_t dlen, 438 register_t *retval) 439 { 440 struct lwp *l; 441 struct sysent *callp; 442 443 if (__predict_false(num >= SYS_NSYSENT)) 444 return ENOSYS; 445 446 l = curlwp; 447 callp = rump_sysent + num; 448 return callp->sy_call(l, (void *)data, retval); 449 } 450 451 rump_sysproxy_t rump_sysproxy = rump_sysproxy_local; 452 void *rump_sysproxy_arg; 453 454 /* 455 * This whole syscall-via-rpc is still taking form. For example, it 456 * may be necessary to set syscalls individually instead of lobbing 457 * them all to the same place. So don't think this interface is 458 * set in stone. 459 */ 460 int 461 rump_sysproxy_set(rump_sysproxy_t proxy, void *arg) 462 { 463 464 if (rump_sysproxy_arg) 465 return EBUSY; 466 467 rump_sysproxy_arg = arg; 468 rump_sysproxy = proxy; 469 470 return 0; 471 } 472
Indexes created Sun Sep 28 16:09:52 GMT 2025