linux_fork.c revision 293608
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 293608 2016-01-09 18:28:15Z 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 if ((error = fork1(td, RFFDG | RFPROC | RFMEM | RFPPWAIT | RFSTOPPED, 111 0, &p2, NULL, 0)) != 0) 112 return (error); 113 114 td2 = FIRST_THREAD_IN_PROC(p2); 115 116 linux_proc_init(td, td2, 0); 117 118 td->td_retval[0] = p2->p_pid; 119 120 /* 121 * Make this runnable after we are finished with it. 122 */ 123 thread_lock(td2); 124 TD_SET_CAN_RUN(td2); 125 sched_add(td2, SRQ_BORING); 126 thread_unlock(td2); 127 128 return (0); 129} 130 131static int 132linux_clone_proc(struct thread *td, struct linux_clone_args *args) 133{ 134 int error, ff = RFPROC | RFSTOPPED; 135 struct proc *p2; 136 struct thread *td2; 137 int exit_signal; 138 struct linux_emuldata *em; 139 140#ifdef DEBUG 141 if (ldebug(clone)) { 142 printf(ARGS(clone, "flags %x, stack %p, parent tid: %p, " 143 "child tid: %p"), (unsigned)args->flags, 144 args->stack, args->parent_tidptr, args->child_tidptr); 145 } 146#endif 147 148 exit_signal = args->flags & 0x000000ff; 149 if (LINUX_SIG_VALID(exit_signal)) { 150 exit_signal = linux_to_bsd_signal(exit_signal); 151 } else if (exit_signal != 0) 152 return (EINVAL); 153 154 if (args->flags & LINUX_CLONE_VM) 155 ff |= RFMEM; 156 if (args->flags & LINUX_CLONE_SIGHAND) 157 ff |= RFSIGSHARE; 158 /* 159 * XXX: In Linux, sharing of fs info (chroot/cwd/umask) 160 * and open files is independant. In FreeBSD, its in one 161 * structure but in reality it does not cause any problems 162 * because both of these flags are usually set together. 163 */ 164 if (!(args->flags & (LINUX_CLONE_FILES | LINUX_CLONE_FS))) 165 ff |= RFFDG; 166 167 if (args->flags & LINUX_CLONE_PARENT_SETTID) 168 if (args->parent_tidptr == NULL) 169 return (EINVAL); 170 171 if (args->flags & LINUX_CLONE_VFORK) 172 ff |= RFPPWAIT; 173 174 error = fork1(td, ff, 0, &p2, NULL, 0); 175 if (error) 176 return (error); 177 178 td2 = FIRST_THREAD_IN_PROC(p2); 179 180 /* create the emuldata */ 181 linux_proc_init(td, td2, args->flags); 182 183 em = em_find(td2); 184 KASSERT(em != NULL, ("clone_proc: emuldata not found.\n")); 185 186 if (args->flags & LINUX_CLONE_CHILD_SETTID) 187 em->child_set_tid = args->child_tidptr; 188 else 189 em->child_set_tid = NULL; 190 191 if (args->flags & LINUX_CLONE_CHILD_CLEARTID) 192 em->child_clear_tid = args->child_tidptr; 193 else 194 em->child_clear_tid = NULL; 195 196 if (args->flags & LINUX_CLONE_PARENT_SETTID) { 197 error = copyout(&p2->p_pid, args->parent_tidptr, 198 sizeof(p2->p_pid)); 199 if (error) 200 printf(LMSG("copyout failed!")); 201 } 202 203 PROC_LOCK(p2); 204 p2->p_sigparent = exit_signal; 205 PROC_UNLOCK(p2); 206 /* 207 * In a case of stack = NULL, we are supposed to COW calling process 208 * stack. This is what normal fork() does, so we just keep tf_rsp arg 209 * intact. 210 */ 211 linux_set_upcall_kse(td2, PTROUT(args->stack)); 212 213 if (args->flags & LINUX_CLONE_SETTLS) 214 linux_set_cloned_tls(td2, args->tls); 215 216#ifdef DEBUG 217 if (ldebug(clone)) 218 printf(LMSG("clone: successful rfork to %d, " 219 "stack %p sig = %d"), (int)p2->p_pid, args->stack, 220 exit_signal); 221#endif 222 223 /* 224 * Make this runnable after we are finished with it. 225 */ 226 thread_lock(td2); 227 TD_SET_CAN_RUN(td2); 228 sched_add(td2, SRQ_BORING); 229 thread_unlock(td2); 230 231 td->td_retval[0] = p2->p_pid; 232 233 return (0); 234} 235 236static int 237linux_clone_thread(struct thread *td, struct linux_clone_args *args) 238{ 239 struct linux_emuldata *em; 240 struct thread *newtd; 241 struct proc *p; 242 int error; 243 244#ifdef DEBUG 245 if (ldebug(clone)) { 246 printf(ARGS(clone, "thread: flags %x, stack %p, parent tid: %p, " 247 "child tid: %p"), (unsigned)args->flags, 248 args->stack, args->parent_tidptr, args->child_tidptr); 249 } 250#endif 251 252 LINUX_CTR4(clone_thread, "thread(%d) flags %x ptid %p ctid %p", 253 td->td_tid, (unsigned)args->flags, 254 args->parent_tidptr, args->child_tidptr); 255 256 if (args->flags & LINUX_CLONE_PARENT_SETTID) 257 if (args->parent_tidptr == NULL) 258 return (EINVAL); 259 260 /* Threads should be created with own stack */ 261 if (args->stack == NULL) 262 return (EINVAL); 263 264 p = td->td_proc; 265 266#ifdef RACCT 267 if (racct_enable) { 268 PROC_LOCK(p); 269 error = racct_add(p, RACCT_NTHR, 1); 270 PROC_UNLOCK(p); 271 if (error != 0) 272 return (EPROCLIM); 273 } 274#endif 275 276 /* Initialize our td */ 277 error = kern_thr_alloc(p, 0, &newtd); 278 if (error) 279 goto fail; 280 281 cpu_set_upcall(newtd, td); 282 283 bzero(&newtd->td_startzero, 284 __rangeof(struct thread, td_startzero, td_endzero)); 285 bcopy(&td->td_startcopy, &newtd->td_startcopy, 286 __rangeof(struct thread, td_startcopy, td_endcopy)); 287 288 newtd->td_proc = p; 289 newtd->td_ucred = crhold(td->td_ucred); 290 291 /* create the emuldata */ 292 linux_proc_init(td, newtd, args->flags); 293 294 em = em_find(newtd); 295 KASSERT(em != NULL, ("clone_thread: emuldata not found.\n")); 296 297 if (args->flags & LINUX_CLONE_SETTLS) 298 linux_set_cloned_tls(newtd, args->tls); 299 300 if (args->flags & LINUX_CLONE_CHILD_SETTID) 301 em->child_set_tid = args->child_tidptr; 302 else 303 em->child_set_tid = NULL; 304 305 if (args->flags & LINUX_CLONE_CHILD_CLEARTID) 306 em->child_clear_tid = args->child_tidptr; 307 else 308 em->child_clear_tid = NULL; 309 310 cpu_thread_clean(newtd); 311 312 linux_set_upcall_kse(newtd, PTROUT(args->stack)); 313 314 PROC_LOCK(p); 315 p->p_flag |= P_HADTHREADS; 316 bcopy(p->p_comm, newtd->td_name, sizeof(newtd->td_name)); 317 318 if (args->flags & LINUX_CLONE_PARENT) 319 thread_link(newtd, p->p_pptr); 320 else 321 thread_link(newtd, p); 322 323 thread_lock(td); 324 /* let the scheduler know about these things. */ 325 sched_fork_thread(td, newtd); 326 thread_unlock(td); 327 if (P_SHOULDSTOP(p)) 328 newtd->td_flags |= TDF_ASTPENDING | TDF_NEEDSUSPCHK; 329 PROC_UNLOCK(p); 330 331 tidhash_add(newtd); 332 333#ifdef DEBUG 334 if (ldebug(clone)) 335 printf(ARGS(clone, "successful clone to %d, stack %p"), 336 (int)newtd->td_tid, args->stack); 337#endif 338 339 LINUX_CTR2(clone_thread, "thread(%d) successful clone to %d", 340 td->td_tid, newtd->td_tid); 341 342 if (args->flags & LINUX_CLONE_PARENT_SETTID) { 343 error = copyout(&newtd->td_tid, args->parent_tidptr, 344 sizeof(newtd->td_tid)); 345 if (error) 346 printf(LMSG("clone_thread: copyout failed!")); 347 } 348 349 /* 350 * Make this runnable after we are finished with it. 351 */ 352 thread_lock(newtd); 353 TD_SET_CAN_RUN(newtd); 354 sched_add(newtd, SRQ_BORING); 355 thread_unlock(newtd); 356 357 td->td_retval[0] = newtd->td_tid; 358 359 return (0); 360 361fail: 362#ifdef RACCT 363 if (racct_enable) { 364 PROC_LOCK(p); 365 racct_sub(p, RACCT_NTHR, 1); 366 PROC_UNLOCK(p); 367 } 368#endif 369 return (error); 370} 371 372int 373linux_clone(struct thread *td, struct linux_clone_args *args) 374{ 375 376 if (args->flags & LINUX_CLONE_THREAD) 377 return (linux_clone_thread(td, args)); 378 else 379 return (linux_clone_proc(td, args)); 380} 381 382int 383linux_exit(struct thread *td, struct linux_exit_args *args) 384{ 385 struct linux_emuldata *em; 386 387 em = em_find(td); 388 KASSERT(em != NULL, ("exit: emuldata not found.\n")); 389 390 LINUX_CTR2(exit, "thread(%d) (%d)", em->em_tid, args->rval); 391 392 linux_thread_detach(td); 393 394 /* 395 * XXX. When the last two threads of a process 396 * exit via pthread_exit() try thr_exit() first. 397 */ 398 kern_thr_exit(td); 399 exit1(td, W_EXITCODE(args->rval, 0)); 400 /* NOTREACHED */ 401} 402 403int 404linux_set_tid_address(struct thread *td, struct linux_set_tid_address_args *args) 405{ 406 struct linux_emuldata *em; 407 408 em = em_find(td); 409 KASSERT(em != NULL, ("set_tid_address: emuldata not found.\n")); 410 411 em->child_clear_tid = args->tidptr; 412 413 td->td_retval[0] = em->em_tid; 414 415 LINUX_CTR3(set_tid_address, "tidptr(%d) %p, returns %d", 416 em->em_tid, args->tidptr, td->td_retval[0]); 417 418 return (0); 419} 420 421void 422linux_thread_detach(struct thread *td) 423{ 424 struct linux_sys_futex_args cup; 425 struct linux_emuldata *em; 426 int *child_clear_tid; 427 int error; 428 429 em = em_find(td); 430 KASSERT(em != NULL, ("thread_detach: emuldata not found.\n")); 431 432 LINUX_CTR1(thread_detach, "thread(%d)", em->em_tid); 433 434 release_futexes(td, em); 435 436 child_clear_tid = em->child_clear_tid; 437 438 if (child_clear_tid != NULL) { 439 440 LINUX_CTR2(thread_detach, "thread(%d) %p", 441 em->em_tid, child_clear_tid); 442 443 error = suword32(child_clear_tid, 0); 444 if (error != 0) 445 return; 446 447 cup.uaddr = child_clear_tid; 448 cup.op = LINUX_FUTEX_WAKE; 449 cup.val = 1; /* wake one */ 450 cup.timeout = NULL; 451 cup.uaddr2 = NULL; 452 cup.val3 = 0; 453 error = linux_sys_futex(td, &cup); 454 /* 455 * this cannot happen at the moment and if this happens it 456 * probably means there is a user space bug 457 */ 458 if (error != 0) 459 linux_msg(td, "futex stuff in thread_detach failed."); 460 } 461} 462