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_emul.h> 60 #include <compat/linux/linux_futex.h> 61 #include <compat/linux/linux_misc.h> 62 #include <compat/linux/linux_util.h> 63 64 int 65 linux_fork(struct thread *td, struct linux_fork_args *args) 66 { 67 int error; 68 struct proc *p2; 69 struct thread *td2; 70 71 #ifdef DEBUG 72 if (ldebug(fork)) 73 printf(ARGS(fork, "")); 74 #endif 75 76 if ((error = fork1(td, RFFDG | RFPROC | RFSTOPPED, 0, &p2, NULL, 0, 77 NULL)) != 0) 78 return (error); 79 80 td2 = FIRST_THREAD_IN_PROC(p2); 81 82 linux_proc_init(td, td2, 0); 83 84 td->td_retval[0] = p2->p_pid; 85 86 /* 87 * Make this runnable after we are finished with it. 88 */ 89 thread_lock(td2); 90 TD_SET_CAN_RUN(td2); 91 sched_add(td2, SRQ_BORING); 92 thread_unlock(td2); 93 94 return (0); 95 } 96 97 int 98 linux_vfork(struct thread *td, struct linux_vfork_args *args) 99 { 100 int error; 101 struct proc *p2; 102 struct thread *td2; 103 104 #ifdef DEBUG 105 if (ldebug(vfork)) 106 printf(ARGS(vfork, "")); 107 #endif 108 109 if ((error = fork1(td, RFFDG | RFPROC | RFMEM | RFPPWAIT | RFSTOPPED, 110 0, &p2, NULL, 0, NULL)) != 0) 111 return (error); 112 113 td2 = FIRST_THREAD_IN_PROC(p2); 114 115 linux_proc_init(td, td2, 0); 116 117 td->td_retval[0] = p2->p_pid; 118 119 /* 120 * Make this runnable after we are finished with it. 121 */ 122 thread_lock(td2); 123 TD_SET_CAN_RUN(td2); 124 sched_add(td2, SRQ_BORING); 125 thread_unlock(td2); 126 127 return (0); 128 } 129 130 static int 131 linux_clone_proc(struct thread *td, struct linux_clone_args *args) 132 { 133 int error, ff = RFPROC | RFSTOPPED; 134 struct proc *p2; 135 struct thread *td2; 136 int exit_signal; 137 struct linux_emuldata *em; 138 139 #ifdef DEBUG 140 if (ldebug(clone)) { 141 printf(ARGS(clone, "flags %x, stack %p, parent tid: %p, " 142 "child tid: %p"), (unsigned)args->flags, 143 args->stack, args->parent_tidptr, args->child_tidptr); 144 } 145 #endif 146 147 exit_signal = args->flags & 0x000000ff; 148 if (LINUX_SIG_VALID(exit_signal)) { 149 exit_signal = linux_to_bsd_signal(exit_signal); 150 } else if (exit_signal != 0) 151 return (EINVAL); 152 153 if (args->flags & LINUX_CLONE_VM) 154 ff |= RFMEM; 155 if (args->flags & LINUX_CLONE_SIGHAND) 156 ff |= RFSIGSHARE; 157 /* 158 * XXX: In Linux, sharing of fs info (chroot/cwd/umask) 159 * and open files is independant. In FreeBSD, its in one 160 * structure but in reality it does not cause any problems 161 * because both of these flags are usually set together. 162 */ 163 if (!(args->flags & (LINUX_CLONE_FILES | LINUX_CLONE_FS))) 164 ff |= RFFDG; 165 166 if (args->flags & LINUX_CLONE_PARENT_SETTID) 167 if (args->parent_tidptr == NULL) 168 return (EINVAL); 169 170 if (args->flags & LINUX_CLONE_VFORK) 171 ff |= RFPPWAIT; 172 173 error = fork1(td, ff, 0, &p2, NULL, 0, NULL); 174 if (error) 175 return (error); 176 177 td2 = FIRST_THREAD_IN_PROC(p2); 178 179 /* create the emuldata */ 180 linux_proc_init(td, td2, args->flags); 181 182 em = em_find(td2); 183 KASSERT(em != NULL, ("clone_proc: emuldata not found.\n")); 184 185 if (args->flags & LINUX_CLONE_CHILD_SETTID) 186 em->child_set_tid = args->child_tidptr; 187 else 188 em->child_set_tid = NULL; 189 190 if (args->flags & LINUX_CLONE_CHILD_CLEARTID) 191 em->child_clear_tid = args->child_tidptr; 192 else 193 em->child_clear_tid = NULL; 194 195 if (args->flags & LINUX_CLONE_PARENT_SETTID) { 196 error = copyout(&p2->p_pid, args->parent_tidptr, 197 sizeof(p2->p_pid)); 198 if (error) 199 printf(LMSG("copyout failed!")); 200 } 201 202 PROC_LOCK(p2); 203 p2->p_sigparent = exit_signal; 204 PROC_UNLOCK(p2); 205 /* 206 * In a case of stack = NULL, we are supposed to COW calling process 207 * stack. This is what normal fork() does, so we just keep tf_rsp arg 208 * intact. 209 */ 210 linux_set_upcall_kse(td2, PTROUT(args->stack)); 211 212 if (args->flags & LINUX_CLONE_SETTLS) 213 linux_set_cloned_tls(td2, args->tls); 214 215 #ifdef DEBUG 216 if (ldebug(clone)) 217 printf(LMSG("clone: successful rfork to %d, " 218 "stack %p sig = %d"), (int)p2->p_pid, args->stack, 219 exit_signal); 220 #endif 221 222 /* 223 * Make this runnable after we are finished with it. 224 */ 225 thread_lock(td2); 226 TD_SET_CAN_RUN(td2); 227 sched_add(td2, SRQ_BORING); 228 thread_unlock(td2); 229 230 td->td_retval[0] = p2->p_pid; 231 232 return (0); 233 } 234 235 static int 236 linux_clone_thread(struct thread *td, struct linux_clone_args *args) 237 { 238 struct linux_emuldata *em; 239 struct thread *newtd; 240 struct proc *p; 241 int error; 242 243 #ifdef DEBUG 244 if (ldebug(clone)) { 245 printf(ARGS(clone, "thread: flags %x, stack %p, parent tid: %p, " 246 "child tid: %p"), (unsigned)args->flags, 247 args->stack, args->parent_tidptr, args->child_tidptr); 248 } 249 #endif 250 251 LINUX_CTR4(clone_thread, "thread(%d) flags %x ptid %p ctid %p", 252 td->td_tid, (unsigned)args->flags, 253 args->parent_tidptr, args->child_tidptr); 254 255 if (args->flags & LINUX_CLONE_PARENT_SETTID) 256 if (args->parent_tidptr == NULL) 257 return (EINVAL); 258 259 /* Threads should be created with own stack */ 260 if (args->stack == NULL) 261 return (EINVAL); 262 263 p = td->td_proc; 264 265 #ifdef RACCT 266 if (racct_enable) { 267 PROC_LOCK(p); 268 error = racct_add(p, RACCT_NTHR, 1); 269 PROC_UNLOCK(p); 270 if (error != 0) 271 return (EPROCLIM); 272 } 273 #endif 274 275 /* Initialize our td */ 276 error = kern_thr_alloc(p, 0, &newtd); 277 if (error) 278 goto fail; 279 280 cpu_set_upcall(newtd, td); 281 282 bzero(&newtd->td_startzero, 283 __rangeof(struct thread, td_startzero, td_endzero)); 284 bcopy(&td->td_startcopy, &newtd->td_startcopy, 285 __rangeof(struct thread, td_startcopy, td_endcopy)); 286 287 newtd->td_proc = p; 288 thread_cow_get(newtd, td); 289 290 /* create the emuldata */ 291 linux_proc_init(td, newtd, args->flags); 292 293 em = em_find(newtd); 294 KASSERT(em != NULL, ("clone_thread: emuldata not found.\n")); 295 296 if (args->flags & LINUX_CLONE_SETTLS) 297 linux_set_cloned_tls(newtd, args->tls); 298 299 if (args->flags & LINUX_CLONE_CHILD_SETTID) 300 em->child_set_tid = args->child_tidptr; 301 else 302 em->child_set_tid = NULL; 303 304 if (args->flags & LINUX_CLONE_CHILD_CLEARTID) 305 em->child_clear_tid = args->child_tidptr; 306 else 307 em->child_clear_tid = NULL; 308 309 cpu_thread_clean(newtd); 310 311 linux_set_upcall_kse(newtd, PTROUT(args->stack)); 312 313 PROC_LOCK(p); 314 p->p_flag |= P_HADTHREADS; 315 bcopy(p->p_comm, newtd->td_name, sizeof(newtd->td_name)); 316 317 if (args->flags & LINUX_CLONE_PARENT) 318 thread_link(newtd, p->p_pptr); 319 else 320 thread_link(newtd, p); 321 322 thread_lock(td); 323 /* let the scheduler know about these things. */ 324 sched_fork_thread(td, newtd); 325 thread_unlock(td); 326 if (P_SHOULDSTOP(p)) 327 newtd->td_flags |= TDF_ASTPENDING | TDF_NEEDSUSPCHK; 328 PROC_UNLOCK(p); 329 330 tidhash_add(newtd); 331 332 #ifdef DEBUG 333 if (ldebug(clone)) 334 printf(ARGS(clone, "successful clone to %d, stack %p"), 335 (int)newtd->td_tid, args->stack); 336 #endif 337 338 LINUX_CTR2(clone_thread, "thread(%d) successful clone to %d", 339 td->td_tid, newtd->td_tid); 340 341 if (args->flags & LINUX_CLONE_PARENT_SETTID) { 342 error = copyout(&newtd->td_tid, args->parent_tidptr, 343 sizeof(newtd->td_tid)); 344 if (error) 345 printf(LMSG("clone_thread: copyout failed!")); 346 } 347 348 /* 349 * Make this runnable after we are finished with it. 350 */ 351 thread_lock(newtd); 352 TD_SET_CAN_RUN(newtd); 353 sched_add(newtd, SRQ_BORING); 354 thread_unlock(newtd); 355 356 td->td_retval[0] = newtd->td_tid; 357 358 return (0); 359 360 fail: 361 #ifdef RACCT 362 if (racct_enable) { 363 PROC_LOCK(p); 364 racct_sub(p, RACCT_NTHR, 1); 365 PROC_UNLOCK(p); 366 } 367 #endif 368 return (error); 369 } 370 371 int 372 linux_clone(struct thread *td, struct linux_clone_args *args) 373 { 374 375 if (args->flags & LINUX_CLONE_THREAD) 376 return (linux_clone_thread(td, args)); 377 else 378 return (linux_clone_proc(td, args)); 379 } 380 381 int 382 linux_exit(struct thread *td, struct linux_exit_args *args) 383 { 384 struct linux_emuldata *em; 385 386 em = em_find(td); 387 KASSERT(em != NULL, ("exit: emuldata not found.\n")); 388 389 LINUX_CTR2(exit, "thread(%d) (%d)", em->em_tid, args->rval); 390 391 linux_thread_detach(td); 392 393 /* 394 * XXX. When the last two threads of a process 395 * exit via pthread_exit() try thr_exit() first. 396 */ 397 kern_thr_exit(td); 398 exit1(td, args->rval, 0); 399 /* NOTREACHED */ 400 } 401 402 int 403 linux_set_tid_address(struct thread *td, struct linux_set_tid_address_args *args) 404 { 405 struct linux_emuldata *em; 406 407 em = em_find(td); 408 KASSERT(em != NULL, ("set_tid_address: emuldata not found.\n")); 409 410 em->child_clear_tid = args->tidptr; 411 412 td->td_retval[0] = em->em_tid; 413 414 LINUX_CTR3(set_tid_address, "tidptr(%d) %p, returns %d", 415 em->em_tid, args->tidptr, td->td_retval[0]); 416 417 return (0); 418 } 419 420 void 421 linux_thread_detach(struct thread *td) 422 { 423 struct linux_sys_futex_args cup; 424 struct linux_emuldata *em; 425 int *child_clear_tid; 426 int error; 427 428 em = em_find(td); 429 KASSERT(em != NULL, ("thread_detach: emuldata not found.\n")); 430 431 LINUX_CTR1(thread_detach, "thread(%d)", em->em_tid); 432 433 release_futexes(td, em); 434 435 child_clear_tid = em->child_clear_tid; 436 437 if (child_clear_tid != NULL) { 438 439 LINUX_CTR2(thread_detach, "thread(%d) %p", 440 em->em_tid, child_clear_tid); 441 442 error = suword32(child_clear_tid, 0); 443 if (error != 0) 444 return; 445 446 cup.uaddr = child_clear_tid; 447 cup.op = LINUX_FUTEX_WAKE; 448 cup.val = 1; /* wake one */ 449 cup.timeout = NULL; 450 cup.uaddr2 = NULL; 451 cup.val3 = 0; 452 error = linux_sys_futex(td, &cup); 453 /* 454 * this cannot happen at the moment and if this happens it 455 * probably means there is a user space bug 456 */ 457 if (error != 0) 458 linux_msg(td, "futex stuff in thread_detach failed."); 459 } 460 } 461