linux_fork.c revision 321343
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/11/sys/compat/linux/linux_fork.c 321343 2017-07-21 18:06:57Z kib $"); 31 32#include "opt_compat.h" 33 34#include <sys/param.h> 35#include <sys/systm.h> 36#include <sys/imgact.h> 37#include <sys/ktr.h> 38#include <sys/lock.h> 39#include <sys/mutex.h> 40#include <sys/proc.h> 41#include <sys/racct.h> 42#include <sys/sched.h> 43#include <sys/syscallsubr.h> 44#include <sys/sx.h> 45#include <sys/umtx.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 struct fork_req fr; 69 int error; 70 struct proc *p2; 71 struct thread *td2; 72 73#ifdef DEBUG 74 if (ldebug(fork)) 75 printf(ARGS(fork, "")); 76#endif 77 78 bzero(&fr, sizeof(fr)); 79 fr.fr_flags = RFFDG | RFPROC | RFSTOPPED; 80 fr.fr_procp = &p2; 81 if ((error = fork1(td, &fr)) != 0) 82 return (error); 83 84 td2 = FIRST_THREAD_IN_PROC(p2); 85 86 linux_proc_init(td, td2, 0); 87 88 td->td_retval[0] = p2->p_pid; 89 90 /* 91 * Make this runnable after we are finished with it. 92 */ 93 thread_lock(td2); 94 TD_SET_CAN_RUN(td2); 95 sched_add(td2, SRQ_BORING); 96 thread_unlock(td2); 97 98 return (0); 99} 100 101int 102linux_vfork(struct thread *td, struct linux_vfork_args *args) 103{ 104 struct fork_req fr; 105 int error; 106 struct proc *p2; 107 struct thread *td2; 108 109#ifdef DEBUG 110 if (ldebug(vfork)) 111 printf(ARGS(vfork, "")); 112#endif 113 114 bzero(&fr, sizeof(fr)); 115 fr.fr_flags = RFFDG | RFPROC | RFMEM | RFPPWAIT | RFSTOPPED; 116 fr.fr_procp = &p2; 117 if ((error = fork1(td, &fr)) != 0) 118 return (error); 119 120 td2 = FIRST_THREAD_IN_PROC(p2); 121 122 linux_proc_init(td, td2, 0); 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 return (0); 135} 136 137static int 138linux_clone_proc(struct thread *td, struct linux_clone_args *args) 139{ 140 struct fork_req fr; 141 int error, ff = RFPROC | RFSTOPPED; 142 struct proc *p2; 143 struct thread *td2; 144 int exit_signal; 145 struct linux_emuldata *em; 146 147#ifdef DEBUG 148 if (ldebug(clone)) { 149 printf(ARGS(clone, "flags %x, stack %p, parent tid: %p, " 150 "child tid: %p"), (unsigned)args->flags, 151 args->stack, args->parent_tidptr, args->child_tidptr); 152 } 153#endif 154 155 exit_signal = args->flags & 0x000000ff; 156 if (LINUX_SIG_VALID(exit_signal)) { 157 exit_signal = linux_to_bsd_signal(exit_signal); 158 } else if (exit_signal != 0) 159 return (EINVAL); 160 161 if (args->flags & LINUX_CLONE_VM) 162 ff |= RFMEM; 163 if (args->flags & LINUX_CLONE_SIGHAND) 164 ff |= RFSIGSHARE; 165 /* 166 * XXX: In Linux, sharing of fs info (chroot/cwd/umask) 167 * and open files is independent. In FreeBSD, its in one 168 * structure but in reality it does not cause any problems 169 * because both of these flags are usually set together. 170 */ 171 if (!(args->flags & (LINUX_CLONE_FILES | LINUX_CLONE_FS))) 172 ff |= RFFDG; 173 174 if (args->flags & LINUX_CLONE_PARENT_SETTID) 175 if (args->parent_tidptr == NULL) 176 return (EINVAL); 177 178 if (args->flags & LINUX_CLONE_VFORK) 179 ff |= RFPPWAIT; 180 181 bzero(&fr, sizeof(fr)); 182 fr.fr_flags = ff; 183 fr.fr_procp = &p2; 184 error = fork1(td, &fr); 185 if (error) 186 return (error); 187 188 td2 = FIRST_THREAD_IN_PROC(p2); 189 190 /* create the emuldata */ 191 linux_proc_init(td, td2, args->flags); 192 193 em = em_find(td2); 194 KASSERT(em != NULL, ("clone_proc: emuldata not found.\n")); 195 196 if (args->flags & LINUX_CLONE_CHILD_SETTID) 197 em->child_set_tid = args->child_tidptr; 198 else 199 em->child_set_tid = NULL; 200 201 if (args->flags & LINUX_CLONE_CHILD_CLEARTID) 202 em->child_clear_tid = args->child_tidptr; 203 else 204 em->child_clear_tid = NULL; 205 206 if (args->flags & LINUX_CLONE_PARENT_SETTID) { 207 error = copyout(&p2->p_pid, args->parent_tidptr, 208 sizeof(p2->p_pid)); 209 if (error) 210 printf(LMSG("copyout failed!")); 211 } 212 213 PROC_LOCK(p2); 214 p2->p_sigparent = exit_signal; 215 PROC_UNLOCK(p2); 216 /* 217 * In a case of stack = NULL, we are supposed to COW calling process 218 * stack. This is what normal fork() does, so we just keep tf_rsp arg 219 * intact. 220 */ 221 linux_set_upcall_kse(td2, PTROUT(args->stack)); 222 223 if (args->flags & LINUX_CLONE_SETTLS) 224 linux_set_cloned_tls(td2, args->tls); 225 226 /* 227 * If CLONE_PARENT is set, then the parent of the new process will be 228 * the same as that of the calling process. 229 */ 230 if (args->flags & LINUX_CLONE_PARENT) { 231 sx_xlock(&proctree_lock); 232 PROC_LOCK(p2); 233 proc_reparent(p2, td->td_proc->p_pptr); 234 PROC_UNLOCK(p2); 235 sx_xunlock(&proctree_lock); 236 } 237 238#ifdef DEBUG 239 if (ldebug(clone)) 240 printf(LMSG("clone: successful rfork to %d, " 241 "stack %p sig = %d"), (int)p2->p_pid, args->stack, 242 exit_signal); 243#endif 244 245 /* 246 * Make this runnable after we are finished with it. 247 */ 248 thread_lock(td2); 249 TD_SET_CAN_RUN(td2); 250 sched_add(td2, SRQ_BORING); 251 thread_unlock(td2); 252 253 td->td_retval[0] = p2->p_pid; 254 255 return (0); 256} 257 258static int 259linux_clone_thread(struct thread *td, struct linux_clone_args *args) 260{ 261 struct linux_emuldata *em; 262 struct thread *newtd; 263 struct proc *p; 264 int error; 265 266#ifdef DEBUG 267 if (ldebug(clone)) { 268 printf(ARGS(clone, "thread: flags %x, stack %p, parent tid: %p, " 269 "child tid: %p"), (unsigned)args->flags, 270 args->stack, args->parent_tidptr, args->child_tidptr); 271 } 272#endif 273 274 LINUX_CTR4(clone_thread, "thread(%d) flags %x ptid %p ctid %p", 275 td->td_tid, (unsigned)args->flags, 276 args->parent_tidptr, args->child_tidptr); 277 278 if (args->flags & LINUX_CLONE_PARENT_SETTID) 279 if (args->parent_tidptr == NULL) 280 return (EINVAL); 281 282 /* Threads should be created with own stack */ 283 if (args->stack == NULL) 284 return (EINVAL); 285 286 p = td->td_proc; 287 288#ifdef RACCT 289 if (racct_enable) { 290 PROC_LOCK(p); 291 error = racct_add(p, RACCT_NTHR, 1); 292 PROC_UNLOCK(p); 293 if (error != 0) 294 return (EPROCLIM); 295 } 296#endif 297 298 /* Initialize our td */ 299 error = kern_thr_alloc(p, 0, &newtd); 300 if (error) 301 goto fail; 302 303 cpu_copy_thread(newtd, td); 304 305 bzero(&newtd->td_startzero, 306 __rangeof(struct thread, td_startzero, td_endzero)); 307 bcopy(&td->td_startcopy, &newtd->td_startcopy, 308 __rangeof(struct thread, td_startcopy, td_endcopy)); 309 newtd->td_sa = td->td_sa; 310 311 newtd->td_proc = p; 312 thread_cow_get(newtd, td); 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 umtx_thread_exit(td); 416 417 linux_thread_detach(td); 418 419 /* 420 * XXX. When the last two threads of a process 421 * exit via pthread_exit() try thr_exit() first. 422 */ 423 kern_thr_exit(td); 424 exit1(td, args->rval, 0); 425 /* NOTREACHED */ 426} 427 428int 429linux_set_tid_address(struct thread *td, struct linux_set_tid_address_args *args) 430{ 431 struct linux_emuldata *em; 432 433 em = em_find(td); 434 KASSERT(em != NULL, ("set_tid_address: emuldata not found.\n")); 435 436 em->child_clear_tid = args->tidptr; 437 438 td->td_retval[0] = em->em_tid; 439 440 LINUX_CTR3(set_tid_address, "tidptr(%d) %p, returns %d", 441 em->em_tid, args->tidptr, td->td_retval[0]); 442 443 return (0); 444} 445 446void 447linux_thread_detach(struct thread *td) 448{ 449 struct linux_sys_futex_args cup; 450 struct linux_emuldata *em; 451 int *child_clear_tid; 452 int error; 453 454 em = em_find(td); 455 KASSERT(em != NULL, ("thread_detach: emuldata not found.\n")); 456 457 LINUX_CTR1(thread_detach, "thread(%d)", em->em_tid); 458 459 release_futexes(td, em); 460 461 child_clear_tid = em->child_clear_tid; 462 463 if (child_clear_tid != NULL) { 464 465 LINUX_CTR2(thread_detach, "thread(%d) %p", 466 em->em_tid, child_clear_tid); 467 468 error = suword32(child_clear_tid, 0); 469 if (error != 0) 470 return; 471 472 cup.uaddr = child_clear_tid; 473 cup.op = LINUX_FUTEX_WAKE; 474 cup.val = 1; /* wake one */ 475 cup.timeout = NULL; 476 cup.uaddr2 = NULL; 477 cup.val3 = 0; 478 error = linux_sys_futex(td, &cup); 479 /* 480 * this cannot happen at the moment and if this happens it 481 * probably means there is a user space bug 482 */ 483 if (error != 0) 484 linux_msg(td, "futex stuff in thread_detach failed."); 485 } 486} 487