1/*- 2 * Copyright (c) 2000 Marcel Moolenaar 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer 10 * in this position and unchanged. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 3. The name of the author may not be used to endorse or promote products 15 * derived from this software without specific prior written permission. 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 <sys/param.h> 33#include <sys/systm.h> 34#include <sys/capability.h> 35#include <sys/file.h> 36#include <sys/fcntl.h> 37#include <sys/imgact.h> 38#include <sys/lock.h> 39#include <sys/malloc.h> 40#include <sys/mman.h> 41#include <sys/mutex.h> 42#include <sys/sx.h> 43#include <sys/priv.h> 44#include <sys/proc.h> 45#include <sys/queue.h> 46#include <sys/resource.h> 47#include <sys/resourcevar.h> 48#include <sys/signalvar.h> 49#include <sys/syscallsubr.h> 50#include <sys/sysproto.h> 51#include <sys/unistd.h> 52#include <sys/wait.h> 53#include <sys/sched.h> 54 55#include <machine/frame.h> 56#include <machine/psl.h> 57#include <machine/segments.h> 58#include <machine/sysarch.h> 59 60#include <vm/vm.h> 61#include <vm/pmap.h> 62#include <vm/vm_map.h> 63 64#include <i386/linux/linux.h> 65#include <i386/linux/linux_proto.h> 66#include <compat/linux/linux_ipc.h> 67#include <compat/linux/linux_misc.h> 68#include <compat/linux/linux_signal.h> 69#include <compat/linux/linux_util.h> 70#include <compat/linux/linux_emul.h> 71 72#include <i386/include/pcb.h> /* needed for pcb definition in linux_set_thread_area */ 73 74#include "opt_posix.h" 75 76extern struct sysentvec elf32_freebsd_sysvec; /* defined in i386/i386/elf_machdep.c */ 77 78struct l_descriptor { 79 l_uint entry_number; 80 l_ulong base_addr; 81 l_uint limit; 82 l_uint seg_32bit:1; 83 l_uint contents:2; 84 l_uint read_exec_only:1; 85 l_uint limit_in_pages:1; 86 l_uint seg_not_present:1; 87 l_uint useable:1; 88}; 89 90struct l_old_select_argv { 91 l_int nfds; 92 l_fd_set *readfds; 93 l_fd_set *writefds; 94 l_fd_set *exceptfds; 95 struct l_timeval *timeout; 96}; 97 98static int linux_mmap_common(struct thread *td, l_uintptr_t addr, 99 l_size_t len, l_int prot, l_int flags, l_int fd, 100 l_loff_t pos); 101 102int 103linux_to_bsd_sigaltstack(int lsa) 104{ 105 int bsa = 0; 106 107 if (lsa & LINUX_SS_DISABLE) 108 bsa |= SS_DISABLE; 109 if (lsa & LINUX_SS_ONSTACK) 110 bsa |= SS_ONSTACK; 111 return (bsa); 112} 113 114int 115bsd_to_linux_sigaltstack(int bsa) 116{ 117 int lsa = 0; 118 119 if (bsa & SS_DISABLE) 120 lsa |= LINUX_SS_DISABLE; 121 if (bsa & SS_ONSTACK) 122 lsa |= LINUX_SS_ONSTACK; 123 return (lsa); 124} 125 126int 127linux_execve(struct thread *td, struct linux_execve_args *args) 128{ 129 int error; 130 char *newpath; 131 struct image_args eargs; 132 133 LCONVPATHEXIST(td, args->path, &newpath); 134 135#ifdef DEBUG 136 if (ldebug(execve)) 137 printf(ARGS(execve, "%s"), newpath); 138#endif 139 140 error = exec_copyin_args(&eargs, newpath, UIO_SYSSPACE, 141 args->argp, args->envp); 142 free(newpath, M_TEMP); 143 if (error == 0) 144 error = kern_execve(td, &eargs, NULL); 145 if (error == 0) 146 /* linux process can exec fbsd one, dont attempt 147 * to create emuldata for such process using 148 * linux_proc_init, this leads to a panic on KASSERT 149 * because such process has p->p_emuldata == NULL 150 */ 151 if (SV_PROC_ABI(td->td_proc) == SV_ABI_LINUX) 152 error = linux_proc_init(td, 0, 0); 153 return (error); 154} 155 156struct l_ipc_kludge { 157 struct l_msgbuf *msgp; 158 l_long msgtyp; 159}; 160 161int 162linux_ipc(struct thread *td, struct linux_ipc_args *args) 163{ 164 165 switch (args->what & 0xFFFF) { 166 case LINUX_SEMOP: { 167 struct linux_semop_args a; 168 169 a.semid = args->arg1; 170 a.tsops = args->ptr; 171 a.nsops = args->arg2; 172 return (linux_semop(td, &a)); 173 } 174 case LINUX_SEMGET: { 175 struct linux_semget_args a; 176 177 a.key = args->arg1; 178 a.nsems = args->arg2; 179 a.semflg = args->arg3; 180 return (linux_semget(td, &a)); 181 } 182 case LINUX_SEMCTL: { 183 struct linux_semctl_args a; 184 int error; 185 186 a.semid = args->arg1; 187 a.semnum = args->arg2; 188 a.cmd = args->arg3; 189 error = copyin(args->ptr, &a.arg, sizeof(a.arg)); 190 if (error) 191 return (error); 192 return (linux_semctl(td, &a)); 193 } 194 case LINUX_MSGSND: { 195 struct linux_msgsnd_args a; 196 197 a.msqid = args->arg1; 198 a.msgp = args->ptr; 199 a.msgsz = args->arg2; 200 a.msgflg = args->arg3; 201 return (linux_msgsnd(td, &a)); 202 } 203 case LINUX_MSGRCV: { 204 struct linux_msgrcv_args a; 205 206 a.msqid = args->arg1; 207 a.msgsz = args->arg2; 208 a.msgflg = args->arg3; 209 if ((args->what >> 16) == 0) { 210 struct l_ipc_kludge tmp; 211 int error; 212 213 if (args->ptr == NULL) 214 return (EINVAL); 215 error = copyin(args->ptr, &tmp, sizeof(tmp)); 216 if (error) 217 return (error); 218 a.msgp = tmp.msgp; 219 a.msgtyp = tmp.msgtyp; 220 } else { 221 a.msgp = args->ptr; 222 a.msgtyp = args->arg5; 223 } 224 return (linux_msgrcv(td, &a)); 225 } 226 case LINUX_MSGGET: { 227 struct linux_msgget_args a; 228 229 a.key = args->arg1; 230 a.msgflg = args->arg2; 231 return (linux_msgget(td, &a)); 232 } 233 case LINUX_MSGCTL: { 234 struct linux_msgctl_args a; 235 236 a.msqid = args->arg1; 237 a.cmd = args->arg2; 238 a.buf = args->ptr; 239 return (linux_msgctl(td, &a)); 240 } 241 case LINUX_SHMAT: { 242 struct linux_shmat_args a; 243 244 a.shmid = args->arg1; 245 a.shmaddr = args->ptr; 246 a.shmflg = args->arg2; 247 a.raddr = (l_ulong *)args->arg3; 248 return (linux_shmat(td, &a)); 249 } 250 case LINUX_SHMDT: { 251 struct linux_shmdt_args a; 252 253 a.shmaddr = args->ptr; 254 return (linux_shmdt(td, &a)); 255 } 256 case LINUX_SHMGET: { 257 struct linux_shmget_args a; 258 259 a.key = args->arg1; 260 a.size = args->arg2; 261 a.shmflg = args->arg3; 262 return (linux_shmget(td, &a)); 263 } 264 case LINUX_SHMCTL: { 265 struct linux_shmctl_args a; 266 267 a.shmid = args->arg1; 268 a.cmd = args->arg2; 269 a.buf = args->ptr; 270 return (linux_shmctl(td, &a)); 271 } 272 default: 273 break; 274 } 275 276 return (EINVAL); 277} 278 279int 280linux_old_select(struct thread *td, struct linux_old_select_args *args) 281{ 282 struct l_old_select_argv linux_args; 283 struct linux_select_args newsel; 284 int error; 285 286#ifdef DEBUG 287 if (ldebug(old_select)) 288 printf(ARGS(old_select, "%p"), args->ptr); 289#endif 290 291 error = copyin(args->ptr, &linux_args, sizeof(linux_args)); 292 if (error) 293 return (error); 294 295 newsel.nfds = linux_args.nfds; 296 newsel.readfds = linux_args.readfds; 297 newsel.writefds = linux_args.writefds; 298 newsel.exceptfds = linux_args.exceptfds; 299 newsel.timeout = linux_args.timeout; 300 return (linux_select(td, &newsel)); 301} 302 303int 304linux_set_cloned_tls(struct thread *td, void *desc) 305{ 306 struct segment_descriptor sd; 307 struct l_user_desc info; 308 int idx, error; 309 int a[2]; 310 311 error = copyin(desc, &info, sizeof(struct l_user_desc)); 312 if (error) { 313 printf(LMSG("copyin failed!")); 314 } else { 315 idx = info.entry_number; 316 317 /* 318 * looks like we're getting the idx we returned 319 * in the set_thread_area() syscall 320 */ 321 if (idx != 6 && idx != 3) { 322 printf(LMSG("resetting idx!")); 323 idx = 3; 324 } 325 326 /* this doesnt happen in practice */ 327 if (idx == 6) { 328 /* we might copy out the entry_number as 3 */ 329 info.entry_number = 3; 330 error = copyout(&info, desc, sizeof(struct l_user_desc)); 331 if (error) 332 printf(LMSG("copyout failed!")); 333 } 334 335 a[0] = LINUX_LDT_entry_a(&info); 336 a[1] = LINUX_LDT_entry_b(&info); 337 338 memcpy(&sd, &a, sizeof(a)); 339#ifdef DEBUG 340 if (ldebug(clone)) 341 printf("Segment created in clone with " 342 "CLONE_SETTLS: lobase: %x, hibase: %x, " 343 "lolimit: %x, hilimit: %x, type: %i, " 344 "dpl: %i, p: %i, xx: %i, def32: %i, " 345 "gran: %i\n", sd.sd_lobase, sd.sd_hibase, 346 sd.sd_lolimit, sd.sd_hilimit, sd.sd_type, 347 sd.sd_dpl, sd.sd_p, sd.sd_xx, 348 sd.sd_def32, sd.sd_gran); 349#endif 350 351 /* set %gs */ 352 td->td_pcb->pcb_gsd = sd; 353 td->td_pcb->pcb_gs = GSEL(GUGS_SEL, SEL_UPL); 354 } 355 356 return (error); 357} 358 359int 360linux_set_upcall_kse(struct thread *td, register_t stack) 361{ 362 363 td->td_frame->tf_esp = stack; 364 365 return (0); 366} 367 368#define STACK_SIZE (2 * 1024 * 1024) 369#define GUARD_SIZE (4 * PAGE_SIZE) 370 371int 372linux_mmap2(struct thread *td, struct linux_mmap2_args *args) 373{ 374 375#ifdef DEBUG 376 if (ldebug(mmap2)) 377 printf(ARGS(mmap2, "%p, %d, %d, 0x%08x, %d, %d"), 378 (void *)args->addr, args->len, args->prot, 379 args->flags, args->fd, args->pgoff); 380#endif 381 382 return (linux_mmap_common(td, args->addr, args->len, args->prot, 383 args->flags, args->fd, (uint64_t)(uint32_t)args->pgoff * 384 PAGE_SIZE)); 385} 386 387int 388linux_mmap(struct thread *td, struct linux_mmap_args *args) 389{ 390 int error; 391 struct l_mmap_argv linux_args; 392 393 error = copyin(args->ptr, &linux_args, sizeof(linux_args)); 394 if (error) 395 return (error); 396 397#ifdef DEBUG 398 if (ldebug(mmap)) 399 printf(ARGS(mmap, "%p, %d, %d, 0x%08x, %d, %d"), 400 (void *)linux_args.addr, linux_args.len, linux_args.prot, 401 linux_args.flags, linux_args.fd, linux_args.pgoff); 402#endif 403 404 return (linux_mmap_common(td, linux_args.addr, linux_args.len, 405 linux_args.prot, linux_args.flags, linux_args.fd, 406 (uint32_t)linux_args.pgoff)); 407} 408 409static int 410linux_mmap_common(struct thread *td, l_uintptr_t addr, l_size_t len, l_int prot, 411 l_int flags, l_int fd, l_loff_t pos) 412{ 413 struct proc *p = td->td_proc; 414 struct mmap_args /* { 415 caddr_t addr; 416 size_t len; 417 int prot; 418 int flags; 419 int fd; 420 long pad; 421 off_t pos; 422 } */ bsd_args; 423 int error; 424 struct file *fp; 425 cap_rights_t rights; 426 427 error = 0; 428 bsd_args.flags = 0; 429 fp = NULL; 430 431 /* 432 * Linux mmap(2): 433 * You must specify exactly one of MAP_SHARED and MAP_PRIVATE 434 */ 435 if (!((flags & LINUX_MAP_SHARED) ^ (flags & LINUX_MAP_PRIVATE))) 436 return (EINVAL); 437 438 if (flags & LINUX_MAP_SHARED) 439 bsd_args.flags |= MAP_SHARED; 440 if (flags & LINUX_MAP_PRIVATE) 441 bsd_args.flags |= MAP_PRIVATE; 442 if (flags & LINUX_MAP_FIXED) 443 bsd_args.flags |= MAP_FIXED; 444 if (flags & LINUX_MAP_ANON) { 445 /* Enforce pos to be on page boundary, then ignore. */ 446 if ((pos & PAGE_MASK) != 0) 447 return (EINVAL); 448 pos = 0; 449 bsd_args.flags |= MAP_ANON; 450 } else 451 bsd_args.flags |= MAP_NOSYNC; 452 if (flags & LINUX_MAP_GROWSDOWN) 453 bsd_args.flags |= MAP_STACK; 454 455 /* 456 * PROT_READ, PROT_WRITE, or PROT_EXEC implies PROT_READ and PROT_EXEC 457 * on Linux/i386. We do this to ensure maximum compatibility. 458 * Linux/ia64 does the same in i386 emulation mode. 459 */ 460 bsd_args.prot = prot; 461 if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC)) 462 bsd_args.prot |= PROT_READ | PROT_EXEC; 463 464 /* Linux does not check file descriptor when MAP_ANONYMOUS is set. */ 465 bsd_args.fd = (bsd_args.flags & MAP_ANON) ? -1 : fd; 466 if (bsd_args.fd != -1) { 467 /* 468 * Linux follows Solaris mmap(2) description: 469 * The file descriptor fildes is opened with 470 * read permission, regardless of the 471 * protection options specified. 472 * 473 * Checking just CAP_MMAP is fine here, since the real work 474 * is done in the FreeBSD mmap(). 475 */ 476 477 error = fget(td, bsd_args.fd, 478 cap_rights_init(&rights, CAP_MMAP), &fp); 479 if (error != 0) 480 return (error); 481 if (fp->f_type != DTYPE_VNODE) { 482 fdrop(fp, td); 483 return (EINVAL); 484 } 485 486 /* Linux mmap() just fails for O_WRONLY files */ 487 if (!(fp->f_flag & FREAD)) { 488 fdrop(fp, td); 489 return (EACCES); 490 } 491 492 fdrop(fp, td); 493 } 494 495 if (flags & LINUX_MAP_GROWSDOWN) { 496 /* 497 * The Linux MAP_GROWSDOWN option does not limit auto 498 * growth of the region. Linux mmap with this option 499 * takes as addr the inital BOS, and as len, the initial 500 * region size. It can then grow down from addr without 501 * limit. However, linux threads has an implicit internal 502 * limit to stack size of STACK_SIZE. Its just not 503 * enforced explicitly in linux. But, here we impose 504 * a limit of (STACK_SIZE - GUARD_SIZE) on the stack 505 * region, since we can do this with our mmap. 506 * 507 * Our mmap with MAP_STACK takes addr as the maximum 508 * downsize limit on BOS, and as len the max size of 509 * the region. It them maps the top SGROWSIZ bytes, 510 * and auto grows the region down, up to the limit 511 * in addr. 512 * 513 * If we don't use the MAP_STACK option, the effect 514 * of this code is to allocate a stack region of a 515 * fixed size of (STACK_SIZE - GUARD_SIZE). 516 */ 517 518 if ((caddr_t)PTRIN(addr) + len > p->p_vmspace->vm_maxsaddr) { 519 /* 520 * Some linux apps will attempt to mmap 521 * thread stacks near the top of their 522 * address space. If their TOS is greater 523 * than vm_maxsaddr, vm_map_growstack() 524 * will confuse the thread stack with the 525 * process stack and deliver a SEGV if they 526 * attempt to grow the thread stack past their 527 * current stacksize rlimit. To avoid this, 528 * adjust vm_maxsaddr upwards to reflect 529 * the current stacksize rlimit rather 530 * than the maximum possible stacksize. 531 * It would be better to adjust the 532 * mmap'ed region, but some apps do not check 533 * mmap's return value. 534 */ 535 PROC_LOCK(p); 536 p->p_vmspace->vm_maxsaddr = (char *)USRSTACK - 537 lim_cur(p, RLIMIT_STACK); 538 PROC_UNLOCK(p); 539 } 540 541 /* 542 * This gives us our maximum stack size and a new BOS. 543 * If we're using VM_STACK, then mmap will just map 544 * the top SGROWSIZ bytes, and let the stack grow down 545 * to the limit at BOS. If we're not using VM_STACK 546 * we map the full stack, since we don't have a way 547 * to autogrow it. 548 */ 549 if (len > STACK_SIZE - GUARD_SIZE) { 550 bsd_args.addr = (caddr_t)PTRIN(addr); 551 bsd_args.len = len; 552 } else { 553 bsd_args.addr = (caddr_t)PTRIN(addr) - 554 (STACK_SIZE - GUARD_SIZE - len); 555 bsd_args.len = STACK_SIZE - GUARD_SIZE; 556 } 557 } else { 558 bsd_args.addr = (caddr_t)PTRIN(addr); 559 bsd_args.len = len; 560 } 561 bsd_args.pos = pos; 562 563#ifdef DEBUG 564 if (ldebug(mmap)) 565 printf("-> %s(%p, %d, %d, 0x%08x, %d, 0x%x)\n", 566 __func__, 567 (void *)bsd_args.addr, bsd_args.len, bsd_args.prot, 568 bsd_args.flags, bsd_args.fd, (int)bsd_args.pos); 569#endif 570 error = sys_mmap(td, &bsd_args); 571#ifdef DEBUG 572 if (ldebug(mmap)) 573 printf("-> %s() return: 0x%x (0x%08x)\n", 574 __func__, error, (u_int)td->td_retval[0]); 575#endif 576 return (error); 577} 578 579int 580linux_mprotect(struct thread *td, struct linux_mprotect_args *uap) 581{ 582 struct mprotect_args bsd_args; 583 584 bsd_args.addr = uap->addr; 585 bsd_args.len = uap->len; 586 bsd_args.prot = uap->prot; 587 if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC)) 588 bsd_args.prot |= PROT_READ | PROT_EXEC; 589 return (sys_mprotect(td, &bsd_args)); 590} 591 592int 593linux_ioperm(struct thread *td, struct linux_ioperm_args *args) 594{ 595 int error; 596 struct i386_ioperm_args iia; 597 598 iia.start = args->start; 599 iia.length = args->length; 600 iia.enable = args->enable; 601 error = i386_set_ioperm(td, &iia); 602 return (error); 603} 604 605int 606linux_iopl(struct thread *td, struct linux_iopl_args *args) 607{ 608 int error; 609 610 if (args->level < 0 || args->level > 3) 611 return (EINVAL); 612 if ((error = priv_check(td, PRIV_IO)) != 0) 613 return (error); 614 if ((error = securelevel_gt(td->td_ucred, 0)) != 0) 615 return (error); 616 td->td_frame->tf_eflags = (td->td_frame->tf_eflags & ~PSL_IOPL) | 617 (args->level * (PSL_IOPL / 3)); 618 return (0); 619} 620 621int 622linux_modify_ldt(struct thread *td, struct linux_modify_ldt_args *uap) 623{ 624 int error; 625 struct i386_ldt_args ldt; 626 struct l_descriptor ld; 627 union descriptor desc; 628 int size, written; 629 630 switch (uap->func) { 631 case 0x00: /* read_ldt */ 632 ldt.start = 0; 633 ldt.descs = uap->ptr; 634 ldt.num = uap->bytecount / sizeof(union descriptor); 635 error = i386_get_ldt(td, &ldt); 636 td->td_retval[0] *= sizeof(union descriptor); 637 break; 638 case 0x02: /* read_default_ldt = 0 */ 639 size = 5*sizeof(struct l_desc_struct); 640 if (size > uap->bytecount) 641 size = uap->bytecount; 642 for (written = error = 0; written < size && error == 0; written++) 643 error = subyte((char *)uap->ptr + written, 0); 644 td->td_retval[0] = written; 645 break; 646 case 0x01: /* write_ldt */ 647 case 0x11: /* write_ldt */ 648 if (uap->bytecount != sizeof(ld)) 649 return (EINVAL); 650 651 error = copyin(uap->ptr, &ld, sizeof(ld)); 652 if (error) 653 return (error); 654 655 ldt.start = ld.entry_number; 656 ldt.descs = &desc; 657 ldt.num = 1; 658 desc.sd.sd_lolimit = (ld.limit & 0x0000ffff); 659 desc.sd.sd_hilimit = (ld.limit & 0x000f0000) >> 16; 660 desc.sd.sd_lobase = (ld.base_addr & 0x00ffffff); 661 desc.sd.sd_hibase = (ld.base_addr & 0xff000000) >> 24; 662 desc.sd.sd_type = SDT_MEMRO | ((ld.read_exec_only ^ 1) << 1) | 663 (ld.contents << 2); 664 desc.sd.sd_dpl = 3; 665 desc.sd.sd_p = (ld.seg_not_present ^ 1); 666 desc.sd.sd_xx = 0; 667 desc.sd.sd_def32 = ld.seg_32bit; 668 desc.sd.sd_gran = ld.limit_in_pages; 669 error = i386_set_ldt(td, &ldt, &desc); 670 break; 671 default: 672 error = ENOSYS; 673 break; 674 } 675 676 if (error == EOPNOTSUPP) { 677 printf("linux: modify_ldt needs kernel option USER_LDT\n"); 678 error = ENOSYS; 679 } 680 681 return (error); 682} 683 684int 685linux_sigaction(struct thread *td, struct linux_sigaction_args *args) 686{ 687 l_osigaction_t osa; 688 l_sigaction_t act, oact; 689 int error; 690 691#ifdef DEBUG 692 if (ldebug(sigaction)) 693 printf(ARGS(sigaction, "%d, %p, %p"), 694 args->sig, (void *)args->nsa, (void *)args->osa); 695#endif 696 697 if (args->nsa != NULL) { 698 error = copyin(args->nsa, &osa, sizeof(l_osigaction_t)); 699 if (error) 700 return (error); 701 act.lsa_handler = osa.lsa_handler; 702 act.lsa_flags = osa.lsa_flags; 703 act.lsa_restorer = osa.lsa_restorer; 704 LINUX_SIGEMPTYSET(act.lsa_mask); 705 act.lsa_mask.__bits[0] = osa.lsa_mask; 706 } 707 708 error = linux_do_sigaction(td, args->sig, args->nsa ? &act : NULL, 709 args->osa ? &oact : NULL); 710 711 if (args->osa != NULL && !error) { 712 osa.lsa_handler = oact.lsa_handler; 713 osa.lsa_flags = oact.lsa_flags; 714 osa.lsa_restorer = oact.lsa_restorer; 715 osa.lsa_mask = oact.lsa_mask.__bits[0]; 716 error = copyout(&osa, args->osa, sizeof(l_osigaction_t)); 717 } 718 719 return (error); 720} 721 722/* 723 * Linux has two extra args, restart and oldmask. We dont use these, 724 * but it seems that "restart" is actually a context pointer that 725 * enables the signal to happen with a different register set. 726 */ 727int 728linux_sigsuspend(struct thread *td, struct linux_sigsuspend_args *args) 729{ 730 sigset_t sigmask; 731 l_sigset_t mask; 732 733#ifdef DEBUG 734 if (ldebug(sigsuspend)) 735 printf(ARGS(sigsuspend, "%08lx"), (unsigned long)args->mask); 736#endif 737 738 LINUX_SIGEMPTYSET(mask); 739 mask.__bits[0] = args->mask; 740 linux_to_bsd_sigset(&mask, &sigmask); 741 return (kern_sigsuspend(td, sigmask)); 742} 743 744int 745linux_rt_sigsuspend(struct thread *td, struct linux_rt_sigsuspend_args *uap) 746{ 747 l_sigset_t lmask; 748 sigset_t sigmask; 749 int error; 750 751#ifdef DEBUG 752 if (ldebug(rt_sigsuspend)) 753 printf(ARGS(rt_sigsuspend, "%p, %d"), 754 (void *)uap->newset, uap->sigsetsize); 755#endif 756 757 if (uap->sigsetsize != sizeof(l_sigset_t)) 758 return (EINVAL); 759 760 error = copyin(uap->newset, &lmask, sizeof(l_sigset_t)); 761 if (error) 762 return (error); 763 764 linux_to_bsd_sigset(&lmask, &sigmask); 765 return (kern_sigsuspend(td, sigmask)); 766} 767 768int 769linux_pause(struct thread *td, struct linux_pause_args *args) 770{ 771 struct proc *p = td->td_proc; 772 sigset_t sigmask; 773 774#ifdef DEBUG 775 if (ldebug(pause)) 776 printf(ARGS(pause, "")); 777#endif 778 779 PROC_LOCK(p); 780 sigmask = td->td_sigmask; 781 PROC_UNLOCK(p); 782 return (kern_sigsuspend(td, sigmask)); 783} 784 785int 786linux_sigaltstack(struct thread *td, struct linux_sigaltstack_args *uap) 787{ 788 stack_t ss, oss; 789 l_stack_t lss; 790 int error; 791 792#ifdef DEBUG 793 if (ldebug(sigaltstack)) 794 printf(ARGS(sigaltstack, "%p, %p"), uap->uss, uap->uoss); 795#endif 796 797 if (uap->uss != NULL) { 798 error = copyin(uap->uss, &lss, sizeof(l_stack_t)); 799 if (error) 800 return (error); 801 802 ss.ss_sp = lss.ss_sp; 803 ss.ss_size = lss.ss_size; 804 ss.ss_flags = linux_to_bsd_sigaltstack(lss.ss_flags); 805 } 806 error = kern_sigaltstack(td, (uap->uss != NULL) ? &ss : NULL, 807 (uap->uoss != NULL) ? &oss : NULL); 808 if (!error && uap->uoss != NULL) { 809 lss.ss_sp = oss.ss_sp; 810 lss.ss_size = oss.ss_size; 811 lss.ss_flags = bsd_to_linux_sigaltstack(oss.ss_flags); 812 error = copyout(&lss, uap->uoss, sizeof(l_stack_t)); 813 } 814 815 return (error); 816} 817 818int 819linux_ftruncate64(struct thread *td, struct linux_ftruncate64_args *args) 820{ 821 struct ftruncate_args sa; 822 823#ifdef DEBUG 824 if (ldebug(ftruncate64)) 825 printf(ARGS(ftruncate64, "%u, %jd"), args->fd, 826 (intmax_t)args->length); 827#endif 828 829 sa.fd = args->fd; 830 sa.length = args->length; 831 return sys_ftruncate(td, &sa); 832} 833 834int 835linux_set_thread_area(struct thread *td, struct linux_set_thread_area_args *args) 836{ 837 struct l_user_desc info; 838 int error; 839 int idx; 840 int a[2]; 841 struct segment_descriptor sd; 842 843 error = copyin(args->desc, &info, sizeof(struct l_user_desc)); 844 if (error) 845 return (error); 846 847#ifdef DEBUG 848 if (ldebug(set_thread_area)) 849 printf(ARGS(set_thread_area, "%i, %x, %x, %i, %i, %i, %i, %i, %i\n"), 850 info.entry_number, 851 info.base_addr, 852 info.limit, 853 info.seg_32bit, 854 info.contents, 855 info.read_exec_only, 856 info.limit_in_pages, 857 info.seg_not_present, 858 info.useable); 859#endif 860 861 idx = info.entry_number; 862 /* 863 * Semantics of linux version: every thread in the system has array of 864 * 3 tls descriptors. 1st is GLIBC TLS, 2nd is WINE, 3rd unknown. This 865 * syscall loads one of the selected tls decriptors with a value and 866 * also loads GDT descriptors 6, 7 and 8 with the content of the 867 * per-thread descriptors. 868 * 869 * Semantics of fbsd version: I think we can ignore that linux has 3 870 * per-thread descriptors and use just the 1st one. The tls_array[] 871 * is used only in set/get-thread_area() syscalls and for loading the 872 * GDT descriptors. In fbsd we use just one GDT descriptor for TLS so 873 * we will load just one. 874 * 875 * XXX: this doesn't work when a user space process tries to use more 876 * than 1 TLS segment. Comment in the linux sources says wine might do 877 * this. 878 */ 879 880 /* 881 * we support just GLIBC TLS now 882 * we should let 3 proceed as well because we use this segment so 883 * if code does two subsequent calls it should succeed 884 */ 885 if (idx != 6 && idx != -1 && idx != 3) 886 return (EINVAL); 887 888 /* 889 * we have to copy out the GDT entry we use 890 * FreeBSD uses GDT entry #3 for storing %gs so load that 891 * 892 * XXX: what if a user space program doesn't check this value and tries 893 * to use 6, 7 or 8? 894 */ 895 idx = info.entry_number = 3; 896 error = copyout(&info, args->desc, sizeof(struct l_user_desc)); 897 if (error) 898 return (error); 899 900 if (LINUX_LDT_empty(&info)) { 901 a[0] = 0; 902 a[1] = 0; 903 } else { 904 a[0] = LINUX_LDT_entry_a(&info); 905 a[1] = LINUX_LDT_entry_b(&info); 906 } 907 908 memcpy(&sd, &a, sizeof(a)); 909#ifdef DEBUG 910 if (ldebug(set_thread_area)) 911 printf("Segment created in set_thread_area: lobase: %x, hibase: %x, lolimit: %x, hilimit: %x, type: %i, dpl: %i, p: %i, xx: %i, def32: %i, gran: %i\n", sd.sd_lobase, 912 sd.sd_hibase, 913 sd.sd_lolimit, 914 sd.sd_hilimit, 915 sd.sd_type, 916 sd.sd_dpl, 917 sd.sd_p, 918 sd.sd_xx, 919 sd.sd_def32, 920 sd.sd_gran); 921#endif 922 923 /* this is taken from i386 version of cpu_set_user_tls() */ 924 critical_enter(); 925 /* set %gs */ 926 td->td_pcb->pcb_gsd = sd; 927 PCPU_GET(fsgs_gdt)[1] = sd; 928 load_gs(GSEL(GUGS_SEL, SEL_UPL)); 929 critical_exit(); 930 931 return (0); 932} 933 934int 935linux_get_thread_area(struct thread *td, struct linux_get_thread_area_args *args) 936{ 937 938 struct l_user_desc info; 939 int error; 940 int idx; 941 struct l_desc_struct desc; 942 struct segment_descriptor sd; 943 944#ifdef DEBUG 945 if (ldebug(get_thread_area)) 946 printf(ARGS(get_thread_area, "%p"), args->desc); 947#endif 948 949 error = copyin(args->desc, &info, sizeof(struct l_user_desc)); 950 if (error) 951 return (error); 952 953 idx = info.entry_number; 954 /* XXX: I am not sure if we want 3 to be allowed too. */ 955 if (idx != 6 && idx != 3) 956 return (EINVAL); 957 958 idx = 3; 959 960 memset(&info, 0, sizeof(info)); 961 962 sd = PCPU_GET(fsgs_gdt)[1]; 963 964 memcpy(&desc, &sd, sizeof(desc)); 965 966 info.entry_number = idx; 967 info.base_addr = LINUX_GET_BASE(&desc); 968 info.limit = LINUX_GET_LIMIT(&desc); 969 info.seg_32bit = LINUX_GET_32BIT(&desc); 970 info.contents = LINUX_GET_CONTENTS(&desc); 971 info.read_exec_only = !LINUX_GET_WRITABLE(&desc); 972 info.limit_in_pages = LINUX_GET_LIMIT_PAGES(&desc); 973 info.seg_not_present = !LINUX_GET_PRESENT(&desc); 974 info.useable = LINUX_GET_USEABLE(&desc); 975 976 error = copyout(&info, args->desc, sizeof(struct l_user_desc)); 977 if (error) 978 return (EFAULT); 979 980 return (0); 981} 982 983/* XXX: this wont work with module - convert it */ 984int 985linux_mq_open(struct thread *td, struct linux_mq_open_args *args) 986{ 987#ifdef P1003_1B_MQUEUE 988 return sys_kmq_open(td, (struct kmq_open_args *) args); 989#else 990 return (ENOSYS); 991#endif 992} 993 994int 995linux_mq_unlink(struct thread *td, struct linux_mq_unlink_args *args) 996{ 997#ifdef P1003_1B_MQUEUE 998 return sys_kmq_unlink(td, (struct kmq_unlink_args *) args); 999#else 1000 return (ENOSYS); 1001#endif 1002} 1003 1004int 1005linux_mq_timedsend(struct thread *td, struct linux_mq_timedsend_args *args) 1006{ 1007#ifdef P1003_1B_MQUEUE 1008 return sys_kmq_timedsend(td, (struct kmq_timedsend_args *) args); 1009#else 1010 return (ENOSYS); 1011#endif 1012} 1013 1014int 1015linux_mq_timedreceive(struct thread *td, struct linux_mq_timedreceive_args *args) 1016{ 1017#ifdef P1003_1B_MQUEUE 1018 return sys_kmq_timedreceive(td, (struct kmq_timedreceive_args *) args); 1019#else 1020 return (ENOSYS); 1021#endif 1022} 1023 1024int 1025linux_mq_notify(struct thread *td, struct linux_mq_notify_args *args) 1026{ 1027#ifdef P1003_1B_MQUEUE 1028 return sys_kmq_notify(td, (struct kmq_notify_args *) args); 1029#else 1030 return (ENOSYS); 1031#endif 1032} 1033 1034int 1035linux_mq_getsetattr(struct thread *td, struct linux_mq_getsetattr_args *args) 1036{ 1037#ifdef P1003_1B_MQUEUE 1038 return sys_kmq_setattr(td, (struct kmq_setattr_args *) args); 1039#else 1040 return (ENOSYS); 1041#endif 1042} 1043 1044int 1045linux_wait4(struct thread *td, struct linux_wait4_args *args) 1046{ 1047 int error, options; 1048 struct rusage ru, *rup; 1049 1050#ifdef DEBUG 1051 if (ldebug(wait4)) 1052 printf(ARGS(wait4, "%d, %p, %d, %p"), 1053 args->pid, (void *)args->status, args->options, 1054 (void *)args->rusage); 1055#endif 1056 1057 options = (args->options & (WNOHANG | WUNTRACED)); 1058 /* WLINUXCLONE should be equal to __WCLONE, but we make sure */ 1059 if (args->options & __WCLONE) 1060 options |= WLINUXCLONE; 1061 1062 if (args->rusage != NULL) 1063 rup = &ru; 1064 else 1065 rup = NULL; 1066 error = linux_common_wait(td, args->pid, args->status, options, rup); 1067 if (error) 1068 return (error); 1069 if (args->rusage != NULL) 1070 error = copyout(&ru, args->rusage, sizeof(ru)); 1071 1072 return (error); 1073} 1074