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("$FreeBSD$"); 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/unistd.h> 46 #include <sys/wait.h> 47 48 #include <vm/vm.h> 49 #include <vm/pmap.h> 50 #include <vm/vm_map.h> 51 52 #ifdef COMPAT_LINUX32 53 #include <machine/../linux32/linux.h> 54 #include <machine/../linux32/linux32_proto.h> 55 #else 56 #include <machine/../linux/linux.h> 57 #include <machine/../linux/linux_proto.h> 58 #endif 59 #include <compat/linux/linux_signal.h> 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 65 int 66 linux_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 98 int 99 linux_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 } 142 143 static int 144 linux_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 if (exit_signal <= LINUX_SIGTBLSZ) 163 exit_signal = 164 linux_to_bsd_signal[_SIG_IDX(exit_signal)]; 165 } else if (exit_signal != 0) 166 return (EINVAL); 167 168 if (args->flags & LINUX_CLONE_VM) 169 ff |= RFMEM; 170 if (args->flags & LINUX_CLONE_SIGHAND) 171 ff |= RFSIGSHARE; 172 /* 173 * XXX: In Linux, sharing of fs info (chroot/cwd/umask) 174 * and open files is independant. In FreeBSD, its in one 175 * structure but in reality it does not cause any problems 176 * because both of these flags are usually set together. 177 */ 178 if (!(args->flags & (LINUX_CLONE_FILES | LINUX_CLONE_FS))) 179 ff |= RFFDG; 180 181 if (args->flags & LINUX_CLONE_PARENT_SETTID) 182 if (args->parent_tidptr == NULL) 183 return (EINVAL); 184 185 error = fork1(td, ff, 0, &p2, NULL, 0); 186 if (error) 187 return (error); 188 189 td2 = FIRST_THREAD_IN_PROC(p2); 190 191 /* create the emuldata */ 192 linux_proc_init(td, td2, args->flags); 193 194 em = em_find(td2); 195 KASSERT(em != NULL, ("clone_proc: emuldata not found.\n")); 196 197 if (args->flags & LINUX_CLONE_CHILD_SETTID) 198 em->child_set_tid = args->child_tidptr; 199 else 200 em->child_set_tid = NULL; 201 202 if (args->flags & LINUX_CLONE_CHILD_CLEARTID) 203 em->child_clear_tid = args->child_tidptr; 204 else 205 em->child_clear_tid = NULL; 206 207 if (args->flags & LINUX_CLONE_PARENT_SETTID) { 208 error = copyout(&p2->p_pid, args->parent_tidptr, 209 sizeof(p2->p_pid)); 210 if (error) 211 printf(LMSG("copyout failed!")); 212 } 213 214 PROC_LOCK(p2); 215 p2->p_sigparent = exit_signal; 216 PROC_UNLOCK(p2); 217 /* 218 * In a case of stack = NULL, we are supposed to COW calling process 219 * stack. This is what normal fork() does, so we just keep tf_rsp arg 220 * intact. 221 */ 222 linux_set_upcall_kse(td2, PTROUT(args->stack)); 223 224 if (args->flags & LINUX_CLONE_SETTLS) 225 linux_set_cloned_tls(td2, args->tls); 226 227 #ifdef DEBUG 228 if (ldebug(clone)) 229 printf(LMSG("clone: successful rfork to %d, " 230 "stack %p sig = %d"), (int)p2->p_pid, args->stack, 231 exit_signal); 232 #endif 233 234 if (args->flags & LINUX_CLONE_VFORK) { 235 PROC_LOCK(p2); 236 p2->p_flag |= P_PPWAIT; 237 PROC_UNLOCK(p2); 238 } 239 240 /* 241 * Make this runnable after we are finished with it. 242 */ 243 thread_lock(td2); 244 TD_SET_CAN_RUN(td2); 245 sched_add(td2, SRQ_BORING); 246 thread_unlock(td2); 247 248 td->td_retval[0] = p2->p_pid; 249 250 if (args->flags & LINUX_CLONE_VFORK) { 251 /* wait for the children to exit, ie. emulate vfork */ 252 PROC_LOCK(p2); 253 while (p2->p_flag & P_PPWAIT) 254 cv_wait(&p2->p_pwait, &p2->p_mtx); 255 PROC_UNLOCK(p2); 256 } 257 258 return (0); 259 } 260 261 static int 262 linux_clone_thread(struct thread *td, struct linux_clone_args *args) 263 { 264 struct linux_emuldata *em; 265 struct thread *newtd; 266 struct proc *p; 267 int error; 268 269 #ifdef DEBUG 270 if (ldebug(clone)) { 271 printf(ARGS(clone, "thread: flags %x, stack %p, parent tid: %p, " 272 "child tid: %p"), (unsigned)args->flags, 273 args->stack, args->parent_tidptr, args->child_tidptr); 274 } 275 #endif 276 277 LINUX_CTR4(clone, "thread(%d) flags %x ptid %p ctid %p", 278 td->td_tid, (unsigned)args->flags, 279 args->parent_tidptr, args->child_tidptr); 280 281 if (args->flags & LINUX_CLONE_PARENT_SETTID) 282 if (args->parent_tidptr == NULL) 283 return (EINVAL); 284 285 /* Threads should be created with own stack */ 286 if (args->stack == NULL) 287 return (EINVAL); 288 289 p = td->td_proc; 290 291 /* Initialize our td */ 292 error = kern_thr_alloc(p, 0, &newtd); 293 if (error) 294 return (error); 295 296 cpu_set_upcall(newtd, td); 297 298 bzero(&newtd->td_startzero, 299 __rangeof(struct thread, td_startzero, td_endzero)); 300 bcopy(&td->td_startcopy, &newtd->td_startcopy, 301 __rangeof(struct thread, td_startcopy, td_endcopy)); 302 303 newtd->td_proc = p; 304 newtd->td_ucred = crhold(td->td_ucred); 305 306 /* create the emuldata */ 307 linux_proc_init(td, newtd, args->flags); 308 309 em = em_find(newtd); 310 KASSERT(em != NULL, ("clone_thread: emuldata not found.\n")); 311 312 if (args->flags & LINUX_CLONE_SETTLS) 313 linux_set_cloned_tls(newtd, args->tls); 314 315 if (args->flags & LINUX_CLONE_CHILD_SETTID) 316 em->child_set_tid = args->child_tidptr; 317 else 318 em->child_set_tid = NULL; 319 320 if (args->flags & LINUX_CLONE_CHILD_CLEARTID) 321 em->child_clear_tid = args->child_tidptr; 322 else 323 em->child_clear_tid = NULL; 324 325 cpu_thread_clean(newtd); 326 327 linux_set_upcall_kse(newtd, PTROUT(args->stack)); 328 329 PROC_LOCK(p); 330 p->p_flag |= P_HADTHREADS; 331 newtd->td_sigmask = td->td_sigmask; 332 bcopy(p->p_comm, newtd->td_name, sizeof(newtd->td_name)); 333 334 if (args->flags & LINUX_CLONE_PARENT) 335 thread_link(newtd, p->p_pptr); 336 else 337 thread_link(newtd, p); 338 339 thread_lock(td); 340 /* let the scheduler know about these things. */ 341 sched_fork_thread(td, newtd); 342 thread_unlock(td); 343 if (P_SHOULDSTOP(p)) 344 newtd->td_flags |= TDF_ASTPENDING | TDF_NEEDSUSPCHK; 345 PROC_UNLOCK(p); 346 347 tidhash_add(newtd); 348 349 #ifdef DEBUG 350 if (ldebug(clone)) 351 printf(ARGS(clone, "successful clone to %d, stack %p"), 352 (int)newtd->td_tid, args->stack); 353 #endif 354 355 LINUX_CTR2(clone, "thread(%d) successful clone to %d", 356 td->td_tid, newtd->td_tid); 357 358 if (args->flags & LINUX_CLONE_PARENT_SETTID) { 359 error = copyout(&newtd->td_tid, args->parent_tidptr, 360 sizeof(newtd->td_tid)); 361 if (error) 362 printf(LMSG("clone_thread: copyout failed!")); 363 } 364 365 /* 366 * Make this runnable after we are finished with it. 367 */ 368 thread_lock(newtd); 369 TD_SET_CAN_RUN(newtd); 370 sched_add(newtd, SRQ_BORING); 371 thread_unlock(newtd); 372 373 td->td_retval[0] = newtd->td_tid; 374 375 return (0); 376 } 377 378 int 379 linux_clone(struct thread *td, struct linux_clone_args *args) 380 { 381 382 if (args->flags & LINUX_CLONE_THREAD) 383 return (linux_clone_thread(td, args)); 384 else 385 return (linux_clone_proc(td, args)); 386 } 387 388 int 389 linux_exit(struct thread *td, struct linux_exit_args *args) 390 { 391 struct linux_emuldata *em; 392 393 em = em_find(td); 394 KASSERT(em != NULL, ("exit: emuldata not found.\n")); 395 396 LINUX_CTR2(exit, "thread(%d) (%d)", em->em_tid, args->rval); 397 398 linux_thread_detach(td); 399 400 /* 401 * XXX. When the last two threads of a process 402 * exit via pthread_exit() try thr_exit() first. 403 */ 404 kern_thr_exit(td); 405 exit1(td, W_EXITCODE(args->rval, 0)); 406 /* NOTREACHED */ 407 } 408 409 int 410 linux_set_tid_address(struct thread *td, struct linux_set_tid_address_args *args) 411 { 412 struct linux_emuldata *em; 413 414 em = em_find(td); 415 KASSERT(em != NULL, ("set_tid_address: emuldata not found.\n")); 416 417 em->child_clear_tid = args->tidptr; 418 419 td->td_retval[0] = em->em_tid; 420 421 LINUX_CTR3(set_tid_address, "tidptr(%d) %p, returns %d", 422 em->em_tid, args->tidptr, td->td_retval[0]); 423 424 return (0); 425 } 426 427 void 428 linux_thread_detach(struct thread *td) 429 { 430 struct linux_sys_futex_args cup; 431 struct linux_emuldata *em; 432 int *child_clear_tid; 433 int error; 434 435 em = em_find(td); 436 KASSERT(em != NULL, ("thread_detach: emuldata not found.\n")); 437 438 LINUX_CTR1(exit, "thread detach(%d)", em->em_tid); 439 440 release_futexes(td, em); 441 442 child_clear_tid = em->child_clear_tid; 443 444 if (child_clear_tid != NULL) { 445 446 LINUX_CTR2(exit, "thread detach(%d) %p", 447 em->em_tid, child_clear_tid); 448 449 error = suword32(child_clear_tid, 0); 450 if (error != 0) 451 return; 452 453 cup.uaddr = child_clear_tid; 454 cup.op = LINUX_FUTEX_WAKE; 455 cup.val = 1; /* wake one */ 456 cup.timeout = NULL; 457 cup.uaddr2 = NULL; 458 cup.val3 = 0; 459 error = linux_sys_futex(td, &cup); 460 /* 461 * this cannot happen at the moment and if this happens it 462 * probably means there is a user space bug 463 */ 464 if (error != 0) 465 linux_msg(td, "futex stuff in thread_detach failed."); 466 } 467 } 468