rump.c revision 1.97
1 /* $NetBSD: rump.c,v 1.97 2009/03/18 10:22:44 cegger 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 1.97 2009/03/18 10:22:44 cegger 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/filedesc.h> 40 #include <sys/iostat.h> 41 #include <sys/kauth.h> 42 #include <sys/kernel.h> 43 #include <sys/kmem.h> 44 #include <sys/kprintf.h> 45 #include <sys/ksyms.h> 46 #include <sys/msgbuf.h> 47 #include <sys/module.h> 48 #include <sys/once.h> 49 #include <sys/percpu.h> 50 #include <sys/queue.h> 51 #include <sys/resourcevar.h> 52 #include <sys/select.h> 53 #include <sys/sysctl.h> 54 #include <sys/syscall.h> 55 #include <sys/tty.h> 56 #include <sys/uidinfo.h> 57 #include <sys/vmem.h> 58 59 #include <rump/rumpuser.h> 60 61 #include "rump_private.h" 62 #include "rump_net_private.h" 63 #include "rump_vfs_private.h" 64 65 struct proc proc0; 66 struct session rump_session = { 67 .s_count = 1, 68 .s_flags = 0, 69 .s_leader = &proc0, 70 .s_login = "rumphobo", 71 .s_sid = 0, 72 }; 73 struct pgrp rump_pgrp = { 74 .pg_members = LIST_HEAD_INITIALIZER(pg_members), 75 .pg_session = &rump_session, 76 .pg_jobc = 1, 77 }; 78 struct pstats rump_stats; 79 struct plimit rump_limits; 80 struct cpu_info rump_cpu; 81 struct filedesc rump_filedesc0; 82 struct proclist allproc; 83 char machine[] = "rump"; 84 static kauth_cred_t rump_susercred; 85 86 struct rumpuser_mtx *rump_giantlock; 87 88 sigset_t sigcantmask; 89 90 #ifdef RUMP_WITHOUT_THREADS 91 int rump_threads = 0; 92 #else 93 int rump_threads = 1; 94 #endif 95 96 static void 97 rump_aiodone_worker(struct work *wk, void *dummy) 98 { 99 struct buf *bp = (struct buf *)wk; 100 101 KASSERT(&bp->b_work == wk); 102 bp->b_iodone(bp); 103 } 104 105 static int rump_inited; 106 static struct emul emul_rump; 107 108 void rump__unavailable(void); 109 void rump__unavailable() {} 110 __weak_alias(rump_net_init,rump__unavailable); 111 __weak_alias(rump_vfs_init,rump__unavailable); 112 113 void rump__unavailable_vfs_panic(void); 114 void rump__unavailable_vfs_panic() {panic("vfs component not available");} 115 __weak_alias(vn_open,rump__unavailable_vfs_panic); 116 __weak_alias(vn_rdwr,rump__unavailable_vfs_panic); 117 __weak_alias(vn_close,rump__unavailable_vfs_panic); 118 119 static void 120 pvfsinit_nop(struct proc *p) 121 { 122 123 return; 124 } 125 126 static void 127 pvfsrele_nop(struct proc *p) 128 { 129 130 return; 131 } 132 133 rump_proc_vfs_init_fn rump_proc_vfs_init = pvfsinit_nop; 134 rump_proc_vfs_release_fn rump_proc_vfs_release = pvfsrele_nop; 135 136 int 137 rump__init(int rump_version) 138 { 139 char buf[256]; 140 struct proc *p; 141 struct lwp *l; 142 int error; 143 144 /* XXX */ 145 if (rump_inited) 146 return 0; 147 rump_inited = 1; 148 149 if (rump_version != RUMP_VERSION) { 150 printf("rump version mismatch, %d vs. %d\n", 151 rump_version, RUMP_VERSION); 152 return EPROGMISMATCH; 153 } 154 155 if (rumpuser_getenv("RUMP_THREADS", buf, sizeof(buf), &error) == 0) { 156 rump_threads = *buf != '0'; 157 } 158 rumpuser_thrinit(_kernel_lock, _kernel_unlock, rump_threads); 159 160 mutex_init(&tty_lock, MUTEX_DEFAULT, IPL_NONE); 161 rumpuser_mutex_recursive_init(&rump_giantlock); 162 ksyms_init(); 163 rumpvm_init(); 164 165 once_init(); 166 167 rump_sleepers_init(); 168 #ifdef RUMP_USE_REAL_ALLOCATORS 169 pool_subsystem_init(); 170 kmem_init(); 171 #endif 172 kprintf_init(); 173 loginit(); 174 175 kauth_init(); 176 rump_susercred = rump_cred_create(0, 0, 0, NULL); 177 178 l = &lwp0; 179 p = &proc0; 180 p->p_stats = &rump_stats; 181 p->p_limit = &rump_limits; 182 p->p_pgrp = &rump_pgrp; 183 p->p_pid = 0; 184 p->p_fd = &rump_filedesc0; 185 p->p_vmspace = &rump_vmspace; 186 p->p_emul = &emul_rump; 187 p->p_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 188 l->l_cred = rump_cred_suserget(); 189 l->l_proc = p; 190 l->l_lid = 1; 191 LIST_INSERT_HEAD(&allproc, p, p_list); 192 proc_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 193 194 rump_limits.pl_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY; 195 rump_limits.pl_rlimit[RLIMIT_NOFILE].rlim_cur = RLIM_INFINITY; 196 rump_limits.pl_rlimit[RLIMIT_SBSIZE].rlim_cur = RLIM_INFINITY; 197 198 callout_startup(); 199 callout_init_cpu(&rump_cpu); 200 201 iostat_init(); 202 uid_init(); 203 percpu_init(); 204 fd_sys_init(); 205 module_init(); 206 sysctl_init(); 207 softint_init(&rump_cpu); 208 cold = 0; 209 devsw_init(); 210 211 /* these do nothing if not present */ 212 rump_vfs_init(); 213 rump_net_init(); 214 215 /* aieeeedondest */ 216 if (rump_threads) { 217 if (workqueue_create(&uvm.aiodone_queue, "aiodoned", 218 rump_aiodone_worker, NULL, 0, 0, 0)) 219 panic("aiodoned"); 220 } 221 222 rumpuser_gethostname(hostname, MAXHOSTNAMELEN, &error); 223 hostnamelen = strlen(hostname); 224 225 sigemptyset(&sigcantmask); 226 227 lwp0.l_fd = proc0.p_fd = fd_init(&rump_filedesc0); 228 229 #ifdef RUMP_USE_REAL_ALLOCATORS 230 if (rump_threads) 231 vmem_rehash_start(); 232 #endif 233 234 return 0; 235 } 236 237 struct uio * 238 rump_uio_setup(void *buf, size_t bufsize, off_t offset, enum rump_uiorw rw) 239 { 240 struct uio *uio; 241 enum uio_rw uiorw; 242 243 switch (rw) { 244 case RUMPUIO_READ: 245 uiorw = UIO_READ; 246 break; 247 case RUMPUIO_WRITE: 248 uiorw = UIO_WRITE; 249 break; 250 default: 251 panic("%s: invalid rw %d", __func__, rw); 252 } 253 254 uio = kmem_alloc(sizeof(struct uio), KM_SLEEP); 255 uio->uio_iov = kmem_alloc(sizeof(struct iovec), KM_SLEEP); 256 257 uio->uio_iov->iov_base = buf; 258 uio->uio_iov->iov_len = bufsize; 259 260 uio->uio_iovcnt = 1; 261 uio->uio_offset = offset; 262 uio->uio_resid = bufsize; 263 uio->uio_rw = uiorw; 264 uio->uio_vmspace = UIO_VMSPACE_SYS; 265 266 return uio; 267 } 268 269 size_t 270 rump_uio_getresid(struct uio *uio) 271 { 272 273 return uio->uio_resid; 274 } 275 276 off_t 277 rump_uio_getoff(struct uio *uio) 278 { 279 280 return uio->uio_offset; 281 } 282 283 size_t 284 rump_uio_free(struct uio *uio) 285 { 286 size_t resid; 287 288 resid = uio->uio_resid; 289 kmem_free(uio->uio_iov, sizeof(*uio->uio_iov)); 290 kmem_free(uio, sizeof(*uio)); 291 292 return resid; 293 } 294 295 struct lwp * 296 rump_setup_curlwp(pid_t pid, lwpid_t lid, int set) 297 { 298 struct lwp *l; 299 struct proc *p; 300 301 l = kmem_zalloc(sizeof(struct lwp), KM_SLEEP); 302 if (pid != 0) { 303 p = kmem_zalloc(sizeof(struct proc), KM_SLEEP); 304 rump_proc_vfs_init(p); 305 p->p_stats = &rump_stats; 306 p->p_limit = &rump_limits; 307 p->p_pid = pid; 308 p->p_vmspace = &rump_vmspace; 309 p->p_fd = fd_init(NULL); 310 } else { 311 p = &proc0; 312 } 313 314 l->l_cred = rump_cred_suserget(); 315 l->l_proc = p; 316 l->l_lid = lid; 317 l->l_fd = p->p_fd; 318 l->l_mutex = RUMP_LMUTEX_MAGIC; 319 l->l_cpu = &rump_cpu; 320 321 if (set) 322 rumpuser_set_curlwp(l); 323 324 return l; 325 } 326 327 void 328 rump_clear_curlwp(void) 329 { 330 struct lwp *l; 331 332 l = rumpuser_get_curlwp(); 333 if (l->l_proc->p_pid != 0) { 334 fd_free(); 335 rump_proc_vfs_release(l->l_proc); 336 rump_cred_destroy(l->l_cred); 337 kmem_free(l->l_proc, sizeof(*l->l_proc)); 338 } 339 kmem_free(l, sizeof(*l)); 340 rumpuser_set_curlwp(NULL); 341 } 342 343 struct lwp * 344 rump_get_curlwp(void) 345 { 346 struct lwp *l; 347 348 l = rumpuser_get_curlwp(); 349 if (l == NULL) 350 l = &lwp0; 351 352 return l; 353 } 354 355 kauth_cred_t 356 rump_cred_create(uid_t uid, gid_t gid, size_t ngroups, gid_t *groups) 357 { 358 kauth_cred_t cred; 359 int rv; 360 361 cred = kauth_cred_alloc(); 362 kauth_cred_setuid(cred, uid); 363 kauth_cred_seteuid(cred, uid); 364 kauth_cred_setsvuid(cred, uid); 365 kauth_cred_setgid(cred, gid); 366 kauth_cred_setgid(cred, gid); 367 kauth_cred_setegid(cred, gid); 368 kauth_cred_setsvgid(cred, gid); 369 rv = kauth_cred_setgroups(cred, groups, ngroups, 0, UIO_SYSSPACE); 370 /* oh this is silly. and by "this" I mean kauth_cred_setgroups() */ 371 assert(rv == 0); 372 373 return cred; 374 } 375 376 void 377 rump_cred_destroy(kauth_cred_t cred) 378 { 379 380 kauth_cred_free(cred); 381 } 382 383 kauth_cred_t 384 rump_cred_suserget(void) 385 { 386 387 kauth_cred_hold(rump_susercred); 388 return rump_susercred; 389 } 390 391 /* 392 * Return the next system lwpid 393 */ 394 lwpid_t 395 rump_nextlid(void) 396 { 397 lwpid_t retid; 398 399 mutex_enter(proc0.p_lock); 400 /* 401 * Take next one, don't return 0 402 * XXX: most likely we'll have collisions in case this 403 * wraps around. 404 */ 405 if (++proc0.p_nlwpid == 0) 406 ++proc0.p_nlwpid; 407 retid = proc0.p_nlwpid; 408 mutex_exit(proc0.p_lock); 409 410 return retid; 411 } 412 413 int 414 rump_module_load(struct modinfo **mi) 415 { 416 417 if (!module_compatible((*mi)->mi_version, __NetBSD_Version__)) 418 return EPROGMISMATCH; 419 420 return (*mi)->mi_modcmd(MODULE_CMD_INIT, NULL); 421 } 422 423 int _syspuffs_stub(int, int *); 424 int 425 _syspuffs_stub(int fd, int *newfd) 426 { 427 428 return ENODEV; 429 } 430 __weak_alias(rump_syspuffs_glueinit,_syspuffs_stub); 431 432 static int 433 rump_sysproxy_local(int num, void *arg, uint8_t *data, size_t dlen, 434 register_t *retval) 435 { 436 struct lwp *l; 437 struct sysent *callp; 438 439 if (__predict_false(num >= SYS_NSYSENT)) 440 return ENOSYS; 441 442 l = curlwp; 443 callp = rump_sysent + num; 444 return callp->sy_call(l, (void *)data, retval); 445 } 446 447 rump_sysproxy_t rump_sysproxy = rump_sysproxy_local; 448 void *rump_sysproxy_arg; 449 450 /* 451 * This whole syscall-via-rpc is still taking form. For example, it 452 * may be necessary to set syscalls individually instead of lobbing 453 * them all to the same place. So don't think this interface is 454 * set in stone. 455 */ 456 int 457 rump_sysproxy_set(rump_sysproxy_t proxy, void *arg) 458 { 459 460 if (rump_sysproxy_arg) 461 return EBUSY; 462 463 rump_sysproxy_arg = arg; 464 rump_sysproxy = proxy; 465 466 return 0; 467 } 468
Indexes created Mon Sep 29 21:09:56 GMT 2025