linux_fork.c revision 293607
1/*- 2 * Copyright (c) 2004 Tim J. Robbins 3 * Copyright (c) 2002 Doug Rabson 4 * Copyright (c) 2000 Marcel Moolenaar 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer 12 * in this position and unchanged. 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 OR 18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 27 */ 28 29#include <sys/cdefs.h> 30__FBSDID("$FreeBSD: stable/10/sys/compat/linux/linux_fork.c 293607 2016-01-09 18:24:53Z dchagin $"); 31 32#include "opt_compat.h" 33#include "opt_kdtrace.h" 34 35#include <sys/param.h> 36#include <sys/systm.h> 37#include <sys/imgact.h> 38#include <sys/ktr.h> 39#include <sys/lock.h> 40#include <sys/mutex.h> 41#include <sys/proc.h> 42#include <sys/racct.h> 43#include <sys/sched.h> 44#include <sys/syscallsubr.h> 45#include <sys/sx.h> 46#include <sys/unistd.h> 47#include <sys/wait.h> 48 49#include <vm/vm.h> 50#include <vm/pmap.h> 51#include <vm/vm_map.h> 52 53#ifdef COMPAT_LINUX32 54#include <machine/../linux32/linux.h> 55#include <machine/../linux32/linux32_proto.h> 56#else 57#include <machine/../linux/linux.h> 58#include <machine/../linux/linux_proto.h> 59#endif 60#include <compat/linux/linux_emul.h> 61#include <compat/linux/linux_futex.h> 62#include <compat/linux/linux_misc.h> 63#include <compat/linux/linux_util.h> 64 65int 66linux_fork(struct thread *td, struct linux_fork_args *args) 67{ 68 int error; 69 struct proc *p2; 70 struct thread *td2; 71 72#ifdef DEBUG 73 if (ldebug(fork)) 74 printf(ARGS(fork, "")); 75#endif 76 77 if ((error = fork1(td, RFFDG | RFPROC | RFSTOPPED, 0, &p2, NULL, 0)) 78 != 0) 79 return (error); 80 81 td2 = FIRST_THREAD_IN_PROC(p2); 82 83 linux_proc_init(td, td2, 0); 84 85 td->td_retval[0] = p2->p_pid; 86 87 /* 88 * Make this runnable after we are finished with it. 89 */ 90 thread_lock(td2); 91 TD_SET_CAN_RUN(td2); 92 sched_add(td2, SRQ_BORING); 93 thread_unlock(td2); 94 95 return (0); 96} 97 98int 99linux_vfork(struct thread *td, struct linux_vfork_args *args) 100{ 101 int error; 102 struct proc *p2; 103 struct thread *td2; 104 105#ifdef DEBUG 106 if (ldebug(vfork)) 107 printf(ARGS(vfork, "")); 108#endif 109 110 /* Exclude RFPPWAIT */ 111 if ((error = fork1(td, RFFDG | RFPROC | RFMEM | RFSTOPPED, 0, &p2, 112 NULL, 0)) != 0) 113 return (error); 114 115 116 td2 = FIRST_THREAD_IN_PROC(p2); 117 118 linux_proc_init(td, td2, 0); 119 120 PROC_LOCK(p2); 121 p2->p_flag |= P_PPWAIT; 122 PROC_UNLOCK(p2); 123 124 td->td_retval[0] = p2->p_pid; 125 126 /* 127 * Make this runnable after we are finished with it. 128 */ 129 thread_lock(td2); 130 TD_SET_CAN_RUN(td2); 131 sched_add(td2, SRQ_BORING); 132 thread_unlock(td2); 133 134 /* wait for the children to exit, ie. emulate vfork */ 135 PROC_LOCK(p2); 136 while (p2->p_flag & P_PPWAIT) 137 cv_wait(&p2->p_pwait, &p2->p_mtx); 138 PROC_UNLOCK(p2); 139 140 return (0); 141} 142 143static int 144linux_clone_proc(struct thread *td, struct linux_clone_args *args) 145{ 146 int error, ff = RFPROC | RFSTOPPED; 147 struct proc *p2; 148 struct thread *td2; 149 int exit_signal; 150 struct linux_emuldata *em; 151 152#ifdef DEBUG 153 if (ldebug(clone)) { 154 printf(ARGS(clone, "flags %x, stack %p, parent tid: %p, " 155 "child tid: %p"), (unsigned)args->flags, 156 args->stack, args->parent_tidptr, args->child_tidptr); 157 } 158#endif 159 160 exit_signal = args->flags & 0x000000ff; 161 if (LINUX_SIG_VALID(exit_signal)) { 162 exit_signal = linux_to_bsd_signal(exit_signal); 163 } else if (exit_signal != 0) 164 return (EINVAL); 165 166 if (args->flags & LINUX_CLONE_VM) 167 ff |= RFMEM; 168 if (args->flags & LINUX_CLONE_SIGHAND) 169 ff |= RFSIGSHARE; 170 /* 171 * XXX: In Linux, sharing of fs info (chroot/cwd/umask) 172 * and open files is independant. In FreeBSD, its in one 173 * structure but in reality it does not cause any problems 174 * because both of these flags are usually set together. 175 */ 176 if (!(args->flags & (LINUX_CLONE_FILES | LINUX_CLONE_FS))) 177 ff |= RFFDG; 178 179 if (args->flags & LINUX_CLONE_PARENT_SETTID) 180 if (args->parent_tidptr == NULL) 181 return (EINVAL); 182 183 error = fork1(td, ff, 0, &p2, NULL, 0); 184 if (error) 185 return (error); 186 187 td2 = FIRST_THREAD_IN_PROC(p2); 188 189 /* create the emuldata */ 190 linux_proc_init(td, td2, args->flags); 191 192 em = em_find(td2); 193 KASSERT(em != NULL, ("clone_proc: emuldata not found.\n")); 194 195 if (args->flags & LINUX_CLONE_CHILD_SETTID) 196 em->child_set_tid = args->child_tidptr; 197 else 198 em->child_set_tid = NULL; 199 200 if (args->flags & LINUX_CLONE_CHILD_CLEARTID) 201 em->child_clear_tid = args->child_tidptr; 202 else 203 em->child_clear_tid = NULL; 204 205 if (args->flags & LINUX_CLONE_PARENT_SETTID) { 206 error = copyout(&p2->p_pid, args->parent_tidptr, 207 sizeof(p2->p_pid)); 208 if (error) 209 printf(LMSG("copyout failed!")); 210 } 211 212 PROC_LOCK(p2); 213 p2->p_sigparent = exit_signal; 214 PROC_UNLOCK(p2); 215 /* 216 * In a case of stack = NULL, we are supposed to COW calling process 217 * stack. This is what normal fork() does, so we just keep tf_rsp arg 218 * intact. 219 */ 220 linux_set_upcall_kse(td2, PTROUT(args->stack)); 221 222 if (args->flags & LINUX_CLONE_SETTLS) 223 linux_set_cloned_tls(td2, args->tls); 224 225#ifdef DEBUG 226 if (ldebug(clone)) 227 printf(LMSG("clone: successful rfork to %d, " 228 "stack %p sig = %d"), (int)p2->p_pid, args->stack, 229 exit_signal); 230#endif 231 232 if (args->flags & LINUX_CLONE_VFORK) { 233 PROC_LOCK(p2); 234 p2->p_flag |= P_PPWAIT; 235 PROC_UNLOCK(p2); 236 } 237 238 /* 239 * Make this runnable after we are finished with it. 240 */ 241 thread_lock(td2); 242 TD_SET_CAN_RUN(td2); 243 sched_add(td2, SRQ_BORING); 244 thread_unlock(td2); 245 246 td->td_retval[0] = p2->p_pid; 247 248 if (args->flags & LINUX_CLONE_VFORK) { 249 /* wait for the children to exit, ie. emulate vfork */ 250 PROC_LOCK(p2); 251 while (p2->p_flag & P_PPWAIT) 252 cv_wait(&p2->p_pwait, &p2->p_mtx); 253 PROC_UNLOCK(p2); 254 } 255 256 return (0); 257} 258 259static int 260linux_clone_thread(struct thread *td, struct linux_clone_args *args) 261{ 262 struct linux_emuldata *em; 263 struct thread *newtd; 264 struct proc *p; 265 int error; 266 267#ifdef DEBUG 268 if (ldebug(clone)) { 269 printf(ARGS(clone, "thread: flags %x, stack %p, parent tid: %p, " 270 "child tid: %p"), (unsigned)args->flags, 271 args->stack, args->parent_tidptr, args->child_tidptr); 272 } 273#endif 274 275 LINUX_CTR4(clone_thread, "thread(%d) flags %x ptid %p ctid %p", 276 td->td_tid, (unsigned)args->flags, 277 args->parent_tidptr, args->child_tidptr); 278 279 if (args->flags & LINUX_CLONE_PARENT_SETTID) 280 if (args->parent_tidptr == NULL) 281 return (EINVAL); 282 283 /* Threads should be created with own stack */ 284 if (args->stack == NULL) 285 return (EINVAL); 286 287 p = td->td_proc; 288 289#ifdef RACCT 290 if (racct_enable) { 291 PROC_LOCK(p); 292 error = racct_add(p, RACCT_NTHR, 1); 293 PROC_UNLOCK(p); 294 if (error != 0) 295 return (EPROCLIM); 296 } 297#endif 298 299 /* Initialize our td */ 300 error = kern_thr_alloc(p, 0, &newtd); 301 if (error) 302 goto fail; 303 304 cpu_set_upcall(newtd, td); 305 306 bzero(&newtd->td_startzero, 307 __rangeof(struct thread, td_startzero, td_endzero)); 308 bcopy(&td->td_startcopy, &newtd->td_startcopy, 309 __rangeof(struct thread, td_startcopy, td_endcopy)); 310 311 newtd->td_proc = p; 312 newtd->td_ucred = crhold(td->td_ucred); 313 314 /* create the emuldata */ 315 linux_proc_init(td, newtd, args->flags); 316 317 em = em_find(newtd); 318 KASSERT(em != NULL, ("clone_thread: emuldata not found.\n")); 319 320 if (args->flags & LINUX_CLONE_SETTLS) 321 linux_set_cloned_tls(newtd, args->tls); 322 323 if (args->flags & LINUX_CLONE_CHILD_SETTID) 324 em->child_set_tid = args->child_tidptr; 325 else 326 em->child_set_tid = NULL; 327 328 if (args->flags & LINUX_CLONE_CHILD_CLEARTID) 329 em->child_clear_tid = args->child_tidptr; 330 else 331 em->child_clear_tid = NULL; 332 333 cpu_thread_clean(newtd); 334 335 linux_set_upcall_kse(newtd, PTROUT(args->stack)); 336 337 PROC_LOCK(p); 338 p->p_flag |= P_HADTHREADS; 339 bcopy(p->p_comm, newtd->td_name, sizeof(newtd->td_name)); 340 341 if (args->flags & LINUX_CLONE_PARENT) 342 thread_link(newtd, p->p_pptr); 343 else 344 thread_link(newtd, p); 345 346 thread_lock(td); 347 /* let the scheduler know about these things. */ 348 sched_fork_thread(td, newtd); 349 thread_unlock(td); 350 if (P_SHOULDSTOP(p)) 351 newtd->td_flags |= TDF_ASTPENDING | TDF_NEEDSUSPCHK; 352 PROC_UNLOCK(p); 353 354 tidhash_add(newtd); 355 356#ifdef DEBUG 357 if (ldebug(clone)) 358 printf(ARGS(clone, "successful clone to %d, stack %p"), 359 (int)newtd->td_tid, args->stack); 360#endif 361 362 LINUX_CTR2(clone_thread, "thread(%d) successful clone to %d", 363 td->td_tid, newtd->td_tid); 364 365 if (args->flags & LINUX_CLONE_PARENT_SETTID) { 366 error = copyout(&newtd->td_tid, args->parent_tidptr, 367 sizeof(newtd->td_tid)); 368 if (error) 369 printf(LMSG("clone_thread: copyout failed!")); 370 } 371 372 /* 373 * Make this runnable after we are finished with it. 374 */ 375 thread_lock(newtd); 376 TD_SET_CAN_RUN(newtd); 377 sched_add(newtd, SRQ_BORING); 378 thread_unlock(newtd); 379 380 td->td_retval[0] = newtd->td_tid; 381 382 return (0); 383 384fail: 385#ifdef RACCT 386 if (racct_enable) { 387 PROC_LOCK(p); 388 racct_sub(p, RACCT_NTHR, 1); 389 PROC_UNLOCK(p); 390 } 391#endif 392 return (error); 393} 394 395int 396linux_clone(struct thread *td, struct linux_clone_args *args) 397{ 398 399 if (args->flags & LINUX_CLONE_THREAD) 400 return (linux_clone_thread(td, args)); 401 else 402 return (linux_clone_proc(td, args)); 403} 404 405int 406linux_exit(struct thread *td, struct linux_exit_args *args) 407{ 408 struct linux_emuldata *em; 409 410 em = em_find(td); 411 KASSERT(em != NULL, ("exit: emuldata not found.\n")); 412 413 LINUX_CTR2(exit, "thread(%d) (%d)", em->em_tid, args->rval); 414 415 linux_thread_detach(td); 416 417 /* 418 * XXX. When the last two threads of a process 419 * exit via pthread_exit() try thr_exit() first. 420 */ 421 kern_thr_exit(td); 422 exit1(td, W_EXITCODE(args->rval, 0)); 423 /* NOTREACHED */ 424} 425 426int 427linux_set_tid_address(struct thread *td, struct linux_set_tid_address_args *args) 428{ 429 struct linux_emuldata *em; 430 431 em = em_find(td); 432 KASSERT(em != NULL, ("set_tid_address: emuldata not found.\n")); 433 434 em->child_clear_tid = args->tidptr; 435 436 td->td_retval[0] = em->em_tid; 437 438 LINUX_CTR3(set_tid_address, "tidptr(%d) %p, returns %d", 439 em->em_tid, args->tidptr, td->td_retval[0]); 440 441 return (0); 442} 443 444void 445linux_thread_detach(struct thread *td) 446{ 447 struct linux_sys_futex_args cup; 448 struct linux_emuldata *em; 449 int *child_clear_tid; 450 int error; 451 452 em = em_find(td); 453 KASSERT(em != NULL, ("thread_detach: emuldata not found.\n")); 454 455 LINUX_CTR1(thread_detach, "thread(%d)", em->em_tid); 456 457 release_futexes(td, em); 458 459 child_clear_tid = em->child_clear_tid; 460 461 if (child_clear_tid != NULL) { 462 463 LINUX_CTR2(thread_detach, "thread(%d) %p", 464 em->em_tid, child_clear_tid); 465 466 error = suword32(child_clear_tid, 0); 467 if (error != 0) 468 return; 469 470 cup.uaddr = child_clear_tid; 471 cup.op = LINUX_FUTEX_WAKE; 472 cup.val = 1; /* wake one */ 473 cup.timeout = NULL; 474 cup.uaddr2 = NULL; 475 cup.val3 = 0; 476 error = linux_sys_futex(td, &cup); 477 /* 478 * this cannot happen at the moment and if this happens it 479 * probably means there is a user space bug 480 */ 481 if (error != 0) 482 linux_msg(td, "futex stuff in thread_detach failed."); 483 } 484} 485