1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD 3 * 4 * Copyright (c) 2004 Tim J. Robbins 5 * Copyright (c) 2002 Doug Rabson 6 * Copyright (c) 2000 Marcel Moolenaar 7 * All rights reserved. 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions 11 * are met: 12 * 1. Redistributions of source code must retain the above copyright 13 * notice, this list of conditions and the following disclaimer 14 * in this position and unchanged. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 20 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 21 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 22 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 23 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 24 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 25 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 26 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 27 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 28 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 29 */ 30 31 #include <sys/cdefs.h> 32 __FBSDID("$FreeBSD$"); 33 34 #include "opt_compat.h" 35 36 #include <sys/param.h> 37 #include <sys/systm.h> 38 #include <sys/imgact.h> 39 #include <sys/ktr.h> 40 #include <sys/lock.h> 41 #include <sys/mutex.h> 42 #include <sys/proc.h> 43 #include <sys/racct.h> 44 #include <sys/sched.h> 45 #include <sys/syscallsubr.h> 46 #include <sys/sx.h> 47 #include <sys/umtx.h> 48 #include <sys/unistd.h> 49 #include <sys/wait.h> 50 51 #include <vm/vm.h> 52 #include <vm/pmap.h> 53 #include <vm/vm_map.h> 54 55 #ifdef COMPAT_LINUX32 56 #include <machine/../linux32/linux.h> 57 #include <machine/../linux32/linux32_proto.h> 58 #else 59 #include <machine/../linux/linux.h> 60 #include <machine/../linux/linux_proto.h> 61 #endif 62 #include <compat/linux/linux_emul.h> 63 #include <compat/linux/linux_futex.h> 64 #include <compat/linux/linux_misc.h> 65 #include <compat/linux/linux_util.h> 66 67 int 68 linux_fork(struct thread *td, struct linux_fork_args *args) 69 { 70 struct fork_req fr; 71 int error; 72 struct proc *p2; 73 struct thread *td2; 74 75 #ifdef DEBUG 76 if (ldebug(fork)) 77 printf(ARGS(fork, "")); 78 #endif 79 80 bzero(&fr, sizeof(fr)); 81 fr.fr_flags = RFFDG | RFPROC | RFSTOPPED; 82 fr.fr_procp = &p2; 83 if ((error = fork1(td, &fr)) != 0) 84 return (error); 85 86 td2 = FIRST_THREAD_IN_PROC(p2); 87 88 linux_proc_init(td, td2, 0); 89 90 td->td_retval[0] = p2->p_pid; 91 92 /* 93 * Make this runnable after we are finished with it. 94 */ 95 thread_lock(td2); 96 TD_SET_CAN_RUN(td2); 97 sched_add(td2, SRQ_BORING); 98 thread_unlock(td2); 99 100 return (0); 101 } 102 103 int 104 linux_vfork(struct thread *td, struct linux_vfork_args *args) 105 { 106 struct fork_req fr; 107 int error; 108 struct proc *p2; 109 struct thread *td2; 110 111 #ifdef DEBUG 112 if (ldebug(vfork)) 113 printf(ARGS(vfork, "")); 114 #endif 115 116 bzero(&fr, sizeof(fr)); 117 fr.fr_flags = RFFDG | RFPROC | RFMEM | RFPPWAIT | RFSTOPPED; 118 fr.fr_procp = &p2; 119 if ((error = fork1(td, &fr)) != 0) 120 return (error); 121 122 td2 = FIRST_THREAD_IN_PROC(p2); 123 124 linux_proc_init(td, td2, 0); 125 126 td->td_retval[0] = p2->p_pid; 127 128 /* 129 * Make this runnable after we are finished with it. 130 */ 131 thread_lock(td2); 132 TD_SET_CAN_RUN(td2); 133 sched_add(td2, SRQ_BORING); 134 thread_unlock(td2); 135 136 return (0); 137 } 138 139 static int 140 linux_clone_proc(struct thread *td, struct linux_clone_args *args) 141 { 142 struct fork_req fr; 143 int error, ff = RFPROC | RFSTOPPED; 144 struct proc *p2; 145 struct thread *td2; 146 int exit_signal; 147 struct linux_emuldata *em; 148 149 #ifdef DEBUG 150 if (ldebug(clone)) { 151 printf(ARGS(clone, "flags %x, stack %p, parent tid: %p, " 152 "child tid: %p"), (unsigned)args->flags, 153 args->stack, args->parent_tidptr, args->child_tidptr); 154 } 155 #endif 156 157 exit_signal = args->flags & 0x000000ff; 158 if (LINUX_SIG_VALID(exit_signal)) { 159 exit_signal = linux_to_bsd_signal(exit_signal); 160 } else if (exit_signal != 0) 161 return (EINVAL); 162 163 if (args->flags & LINUX_CLONE_VM) 164 ff |= RFMEM; 165 if (args->flags & LINUX_CLONE_SIGHAND) 166 ff |= RFSIGSHARE; 167 /* 168 * XXX: In Linux, sharing of fs info (chroot/cwd/umask) 169 * and open files is independent. In FreeBSD, its in one 170 * structure but in reality it does not cause any problems 171 * because both of these flags are usually set together. 172 */ 173 if (!(args->flags & (LINUX_CLONE_FILES | LINUX_CLONE_FS))) 174 ff |= RFFDG; 175 176 if (args->flags & LINUX_CLONE_PARENT_SETTID) 177 if (args->parent_tidptr == NULL) 178 return (EINVAL); 179 180 if (args->flags & LINUX_CLONE_VFORK) 181 ff |= RFPPWAIT; 182 183 bzero(&fr, sizeof(fr)); 184 fr.fr_flags = ff; 185 fr.fr_procp = &p2; 186 error = fork1(td, &fr); 187 if (error) 188 return (error); 189 190 td2 = FIRST_THREAD_IN_PROC(p2); 191 192 /* create the emuldata */ 193 linux_proc_init(td, td2, args->flags); 194 195 em = em_find(td2); 196 KASSERT(em != NULL, ("clone_proc: emuldata not found.\n")); 197 198 if (args->flags & LINUX_CLONE_CHILD_SETTID) 199 em->child_set_tid = args->child_tidptr; 200 else 201 em->child_set_tid = NULL; 202 203 if (args->flags & LINUX_CLONE_CHILD_CLEARTID) 204 em->child_clear_tid = args->child_tidptr; 205 else 206 em->child_clear_tid = NULL; 207 208 if (args->flags & LINUX_CLONE_PARENT_SETTID) { 209 error = copyout(&p2->p_pid, args->parent_tidptr, 210 sizeof(p2->p_pid)); 211 if (error) 212 printf(LMSG("copyout failed!")); 213 } 214 215 PROC_LOCK(p2); 216 p2->p_sigparent = exit_signal; 217 PROC_UNLOCK(p2); 218 /* 219 * In a case of stack = NULL, we are supposed to COW calling process 220 * stack. This is what normal fork() does, so we just keep tf_rsp arg 221 * intact. 222 */ 223 linux_set_upcall_kse(td2, PTROUT(args->stack)); 224 225 if (args->flags & LINUX_CLONE_SETTLS) 226 linux_set_cloned_tls(td2, args->tls); 227 228 /* 229 * If CLONE_PARENT is set, then the parent of the new process will be 230 * the same as that of the calling process. 231 */ 232 if (args->flags & LINUX_CLONE_PARENT) { 233 sx_xlock(&proctree_lock); 234 PROC_LOCK(p2); 235 proc_reparent(p2, td->td_proc->p_pptr); 236 PROC_UNLOCK(p2); 237 sx_xunlock(&proctree_lock); 238 } 239 240 #ifdef DEBUG 241 if (ldebug(clone)) 242 printf(LMSG("clone: successful rfork to %d, " 243 "stack %p sig = %d"), (int)p2->p_pid, args->stack, 244 exit_signal); 245 #endif 246 247 /* 248 * Make this runnable after we are finished with it. 249 */ 250 thread_lock(td2); 251 TD_SET_CAN_RUN(td2); 252 sched_add(td2, SRQ_BORING); 253 thread_unlock(td2); 254 255 td->td_retval[0] = p2->p_pid; 256 257 return (0); 258 } 259 260 static int 261 linux_clone_thread(struct thread *td, struct linux_clone_args *args) 262 { 263 struct linux_emuldata *em; 264 struct thread *newtd; 265 struct proc *p; 266 int error; 267 268 #ifdef DEBUG 269 if (ldebug(clone)) { 270 printf(ARGS(clone, "thread: flags %x, stack %p, parent tid: %p, " 271 "child tid: %p"), (unsigned)args->flags, 272 args->stack, args->parent_tidptr, args->child_tidptr); 273 } 274 #endif 275 276 LINUX_CTR4(clone_thread, "thread(%d) flags %x ptid %p ctid %p", 277 td->td_tid, (unsigned)args->flags, 278 args->parent_tidptr, args->child_tidptr); 279 280 if (args->flags & LINUX_CLONE_PARENT_SETTID) 281 if (args->parent_tidptr == NULL) 282 return (EINVAL); 283 284 /* Threads should be created with own stack */ 285 if (args->stack == NULL) 286 return (EINVAL); 287 288 p = td->td_proc; 289 290 #ifdef RACCT 291 if (racct_enable) { 292 PROC_LOCK(p); 293 error = racct_add(p, RACCT_NTHR, 1); 294 PROC_UNLOCK(p); 295 if (error != 0) 296 return (EPROCLIM); 297 } 298 #endif 299 300 /* Initialize our td */ 301 error = kern_thr_alloc(p, 0, &newtd); 302 if (error) 303 goto fail; 304 305 cpu_copy_thread(newtd, td); 306 307 bzero(&newtd->td_startzero, 308 __rangeof(struct thread, td_startzero, td_endzero)); 309 bcopy(&td->td_startcopy, &newtd->td_startcopy, 310 __rangeof(struct thread, td_startcopy, td_endcopy)); 311 312 newtd->td_proc = p; 313 thread_cow_get(newtd, td); 314 315 /* create the emuldata */ 316 linux_proc_init(td, newtd, args->flags); 317 318 em = em_find(newtd); 319 KASSERT(em != NULL, ("clone_thread: emuldata not found.\n")); 320 321 if (args->flags & LINUX_CLONE_SETTLS) 322 linux_set_cloned_tls(newtd, args->tls); 323 324 if (args->flags & LINUX_CLONE_CHILD_SETTID) 325 em->child_set_tid = args->child_tidptr; 326 else 327 em->child_set_tid = NULL; 328 329 if (args->flags & LINUX_CLONE_CHILD_CLEARTID) 330 em->child_clear_tid = args->child_tidptr; 331 else 332 em->child_clear_tid = NULL; 333 334 cpu_thread_clean(newtd); 335 336 linux_set_upcall_kse(newtd, PTROUT(args->stack)); 337 338 PROC_LOCK(p); 339 p->p_flag |= P_HADTHREADS; 340 bcopy(p->p_comm, newtd->td_name, sizeof(newtd->td_name)); 341 342 if (args->flags & LINUX_CLONE_PARENT) 343 thread_link(newtd, p->p_pptr); 344 else 345 thread_link(newtd, p); 346 347 thread_lock(td); 348 /* let the scheduler know about these things. */ 349 sched_fork_thread(td, newtd); 350 thread_unlock(td); 351 if (P_SHOULDSTOP(p)) 352 newtd->td_flags |= TDF_ASTPENDING | TDF_NEEDSUSPCHK; 353 PROC_UNLOCK(p); 354 355 tidhash_add(newtd); 356 357 #ifdef DEBUG 358 if (ldebug(clone)) 359 printf(ARGS(clone, "successful clone to %d, stack %p"), 360 (int)newtd->td_tid, args->stack); 361 #endif 362 363 LINUX_CTR2(clone_thread, "thread(%d) successful clone to %d", 364 td->td_tid, newtd->td_tid); 365 366 if (args->flags & LINUX_CLONE_PARENT_SETTID) { 367 error = copyout(&newtd->td_tid, args->parent_tidptr, 368 sizeof(newtd->td_tid)); 369 if (error) 370 printf(LMSG("clone_thread: copyout failed!")); 371 } 372 373 /* 374 * Make this runnable after we are finished with it. 375 */ 376 thread_lock(newtd); 377 TD_SET_CAN_RUN(newtd); 378 sched_add(newtd, SRQ_BORING); 379 thread_unlock(newtd); 380 381 td->td_retval[0] = newtd->td_tid; 382 383 return (0); 384 385 fail: 386 #ifdef RACCT 387 if (racct_enable) { 388 PROC_LOCK(p); 389 racct_sub(p, RACCT_NTHR, 1); 390 PROC_UNLOCK(p); 391 } 392 #endif 393 return (error); 394 } 395 396 int 397 linux_clone(struct thread *td, struct linux_clone_args *args) 398 { 399 400 if (args->flags & LINUX_CLONE_THREAD) 401 return (linux_clone_thread(td, args)); 402 else 403 return (linux_clone_proc(td, args)); 404 } 405 406 int 407 linux_exit(struct thread *td, struct linux_exit_args *args) 408 { 409 struct linux_emuldata *em; 410 411 em = em_find(td); 412 KASSERT(em != NULL, ("exit: emuldata not found.\n")); 413 414 LINUX_CTR2(exit, "thread(%d) (%d)", em->em_tid, args->rval); 415 416 umtx_thread_exit(td); 417 418 linux_thread_detach(td); 419 420 /* 421 * XXX. When the last two threads of a process 422 * exit via pthread_exit() try thr_exit() first. 423 */ 424 kern_thr_exit(td); 425 exit1(td, args->rval, 0); 426 /* NOTREACHED */ 427 } 428 429 int 430 linux_set_tid_address(struct thread *td, struct linux_set_tid_address_args *args) 431 { 432 struct linux_emuldata *em; 433 434 em = em_find(td); 435 KASSERT(em != NULL, ("set_tid_address: emuldata not found.\n")); 436 437 em->child_clear_tid = args->tidptr; 438 439 td->td_retval[0] = em->em_tid; 440 441 LINUX_CTR3(set_tid_address, "tidptr(%d) %p, returns %d", 442 em->em_tid, args->tidptr, td->td_retval[0]); 443 444 return (0); 445 } 446 447 void 448 linux_thread_detach(struct thread *td) 449 { 450 struct linux_sys_futex_args cup; 451 struct linux_emuldata *em; 452 int *child_clear_tid; 453 int error; 454 455 em = em_find(td); 456 KASSERT(em != NULL, ("thread_detach: emuldata not found.\n")); 457 458 LINUX_CTR1(thread_detach, "thread(%d)", em->em_tid); 459 460 release_futexes(td, em); 461 462 child_clear_tid = em->child_clear_tid; 463 464 if (child_clear_tid != NULL) { 465 466 LINUX_CTR2(thread_detach, "thread(%d) %p", 467 em->em_tid, child_clear_tid); 468 469 error = suword32(child_clear_tid, 0); 470 if (error != 0) 471 return; 472 473 cup.uaddr = child_clear_tid; 474 cup.op = LINUX_FUTEX_WAKE; 475 cup.val = 1; /* wake one */ 476 cup.timeout = NULL; 477 cup.uaddr2 = NULL; 478 cup.val3 = 0; 479 error = linux_sys_futex(td, &cup); 480 /* 481 * this cannot happen at the moment and if this happens it 482 * probably means there is a user space bug 483 */ 484 if (error != 0) 485 linux_msg(td, "futex stuff in thread_detach failed."); 486 } 487 } 488