linux_machdep.c revision 161611
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: head/sys/i386/linux/linux_machdep.c 161611 2006-08-25 11:59:56Z netchild $"); 31 32#include <sys/param.h> 33#include <sys/systm.h> 34#include <sys/imgact.h> 35#include <sys/lock.h> 36#include <sys/malloc.h> 37#include <sys/mman.h> 38#include <sys/mutex.h> 39#include <sys/sx.h> 40#include <sys/proc.h> 41#include <sys/queue.h> 42#include <sys/resource.h> 43#include <sys/resourcevar.h> 44#include <sys/signalvar.h> 45#include <sys/syscallsubr.h> 46#include <sys/sysproto.h> 47#include <sys/unistd.h> 48#include <sys/wait.h> 49 50#include <machine/frame.h> 51#include <machine/psl.h> 52#include <machine/segments.h> 53#include <machine/sysarch.h> 54 55#include <vm/vm.h> 56#include <vm/pmap.h> 57#include <vm/vm_map.h> 58 59#include <i386/linux/linux.h> 60#include <i386/linux/linux_proto.h> 61#include <compat/linux/linux_ipc.h> 62#include <compat/linux/linux_signal.h> 63#include <compat/linux/linux_util.h> 64#include <compat/linux/linux_emul.h> 65 66#include <i386/include/pcb.h> /* needed for pcb definition in linux_set_thread_area */ 67 68#include "opt_posix.h" 69 70extern struct sysentvec elf32_freebsd_sysvec; /* defined in i386/i386/elf_machdep.c */ 71 72struct l_descriptor { 73 l_uint entry_number; 74 l_ulong base_addr; 75 l_uint limit; 76 l_uint seg_32bit:1; 77 l_uint contents:2; 78 l_uint read_exec_only:1; 79 l_uint limit_in_pages:1; 80 l_uint seg_not_present:1; 81 l_uint useable:1; 82}; 83 84struct l_old_select_argv { 85 l_int nfds; 86 l_fd_set *readfds; 87 l_fd_set *writefds; 88 l_fd_set *exceptfds; 89 struct l_timeval *timeout; 90}; 91 92int 93linux_to_bsd_sigaltstack(int lsa) 94{ 95 int bsa = 0; 96 97 if (lsa & LINUX_SS_DISABLE) 98 bsa |= SS_DISABLE; 99 if (lsa & LINUX_SS_ONSTACK) 100 bsa |= SS_ONSTACK; 101 return (bsa); 102} 103 104int 105bsd_to_linux_sigaltstack(int bsa) 106{ 107 int lsa = 0; 108 109 if (bsa & SS_DISABLE) 110 lsa |= LINUX_SS_DISABLE; 111 if (bsa & SS_ONSTACK) 112 lsa |= LINUX_SS_ONSTACK; 113 return (lsa); 114} 115 116int 117linux_execve(struct thread *td, struct linux_execve_args *args) 118{ 119 int error; 120 char *newpath; 121 struct image_args eargs; 122 123 LCONVPATHEXIST(td, args->path, &newpath); 124 125#ifdef DEBUG 126 if (ldebug(execve)) 127 printf(ARGS(execve, "%s"), newpath); 128#endif 129 130 error = exec_copyin_args(&eargs, newpath, UIO_SYSSPACE, 131 args->argp, args->envp); 132 free(newpath, M_TEMP); 133 if (error == 0) 134 error = kern_execve(td, &eargs, NULL); 135 if (error == 0) 136 /* linux process can exec fbsd one, dont attempt 137 * to create emuldata for such process using 138 * linux_proc_init, this leads to a panic on KASSERT 139 * because such process has p->p_emuldata == NULL 140 */ 141 if (td->td_proc->p_sysent == &elf_linux_sysvec) 142 error = linux_proc_init(td, 0, 0); 143 return (error); 144} 145 146struct l_ipc_kludge { 147 struct l_msgbuf *msgp; 148 l_long msgtyp; 149}; 150 151int 152linux_ipc(struct thread *td, struct linux_ipc_args *args) 153{ 154 155 switch (args->what & 0xFFFF) { 156 case LINUX_SEMOP: { 157 struct linux_semop_args a; 158 159 a.semid = args->arg1; 160 a.tsops = args->ptr; 161 a.nsops = args->arg2; 162 return (linux_semop(td, &a)); 163 } 164 case LINUX_SEMGET: { 165 struct linux_semget_args a; 166 167 a.key = args->arg1; 168 a.nsems = args->arg2; 169 a.semflg = args->arg3; 170 return (linux_semget(td, &a)); 171 } 172 case LINUX_SEMCTL: { 173 struct linux_semctl_args a; 174 int error; 175 176 a.semid = args->arg1; 177 a.semnum = args->arg2; 178 a.cmd = args->arg3; 179 error = copyin(args->ptr, &a.arg, sizeof(a.arg)); 180 if (error) 181 return (error); 182 return (linux_semctl(td, &a)); 183 } 184 case LINUX_MSGSND: { 185 struct linux_msgsnd_args a; 186 187 a.msqid = args->arg1; 188 a.msgp = args->ptr; 189 a.msgsz = args->arg2; 190 a.msgflg = args->arg3; 191 return (linux_msgsnd(td, &a)); 192 } 193 case LINUX_MSGRCV: { 194 struct linux_msgrcv_args a; 195 196 a.msqid = args->arg1; 197 a.msgsz = args->arg2; 198 a.msgflg = args->arg3; 199 if ((args->what >> 16) == 0) { 200 struct l_ipc_kludge tmp; 201 int error; 202 203 if (args->ptr == NULL) 204 return (EINVAL); 205 error = copyin(args->ptr, &tmp, sizeof(tmp)); 206 if (error) 207 return (error); 208 a.msgp = tmp.msgp; 209 a.msgtyp = tmp.msgtyp; 210 } else { 211 a.msgp = args->ptr; 212 a.msgtyp = args->arg5; 213 } 214 return (linux_msgrcv(td, &a)); 215 } 216 case LINUX_MSGGET: { 217 struct linux_msgget_args a; 218 219 a.key = args->arg1; 220 a.msgflg = args->arg2; 221 return (linux_msgget(td, &a)); 222 } 223 case LINUX_MSGCTL: { 224 struct linux_msgctl_args a; 225 226 a.msqid = args->arg1; 227 a.cmd = args->arg2; 228 a.buf = args->ptr; 229 return (linux_msgctl(td, &a)); 230 } 231 case LINUX_SHMAT: { 232 struct linux_shmat_args a; 233 234 a.shmid = args->arg1; 235 a.shmaddr = args->ptr; 236 a.shmflg = args->arg2; 237 a.raddr = (l_ulong *)args->arg3; 238 return (linux_shmat(td, &a)); 239 } 240 case LINUX_SHMDT: { 241 struct linux_shmdt_args a; 242 243 a.shmaddr = args->ptr; 244 return (linux_shmdt(td, &a)); 245 } 246 case LINUX_SHMGET: { 247 struct linux_shmget_args a; 248 249 a.key = args->arg1; 250 a.size = args->arg2; 251 a.shmflg = args->arg3; 252 return (linux_shmget(td, &a)); 253 } 254 case LINUX_SHMCTL: { 255 struct linux_shmctl_args a; 256 257 a.shmid = args->arg1; 258 a.cmd = args->arg2; 259 a.buf = args->ptr; 260 return (linux_shmctl(td, &a)); 261 } 262 default: 263 break; 264 } 265 266 return (EINVAL); 267} 268 269int 270linux_old_select(struct thread *td, struct linux_old_select_args *args) 271{ 272 struct l_old_select_argv linux_args; 273 struct linux_select_args newsel; 274 int error; 275 276#ifdef DEBUG 277 if (ldebug(old_select)) 278 printf(ARGS(old_select, "%p"), args->ptr); 279#endif 280 281 error = copyin(args->ptr, &linux_args, sizeof(linux_args)); 282 if (error) 283 return (error); 284 285 newsel.nfds = linux_args.nfds; 286 newsel.readfds = linux_args.readfds; 287 newsel.writefds = linux_args.writefds; 288 newsel.exceptfds = linux_args.exceptfds; 289 newsel.timeout = linux_args.timeout; 290 return (linux_select(td, &newsel)); 291} 292 293int 294linux_fork(struct thread *td, struct linux_fork_args *args) 295{ 296 int error; 297 298#ifdef DEBUG 299 if (ldebug(fork)) 300 printf(ARGS(fork, "")); 301#endif 302 303 if ((error = fork(td, (struct fork_args *)args)) != 0) 304 return (error); 305 306 if (td->td_retval[1] == 1) 307 td->td_retval[0] = 0; 308 error = linux_proc_init(td, td->td_retval[0], 0); 309 if (error) 310 return (error); 311 312 return (0); 313} 314 315int 316linux_vfork(struct thread *td, struct linux_vfork_args *args) 317{ 318 int error; 319 struct proc *p2; 320 321#ifdef DEBUG 322 if (ldebug(vfork)) 323 printf(ARGS(vfork, "")); 324#endif 325 326 /* exclude RFPPWAIT */ 327 if ((error = fork1(td, RFFDG | RFPROC | RFMEM, 0, &p2)) != 0) 328 return (error); 329 if (error == 0) { 330 td->td_retval[0] = p2->p_pid; 331 td->td_retval[1] = 0; 332 } 333 /* Are we the child? */ 334 if (td->td_retval[1] == 1) 335 td->td_retval[0] = 0; 336 error = linux_proc_init(td, td->td_retval[0], 0); 337 if (error) 338 return (error); 339 /* wait for the children to exit, ie. emulate vfork */ 340 PROC_LOCK(p2); 341 while (p2->p_flag & P_PPWAIT) 342 msleep(td->td_proc, &p2->p_mtx, PWAIT, "ppwait", 0); 343 PROC_UNLOCK(p2); 344 345 return (0); 346} 347 348int 349linux_clone(struct thread *td, struct linux_clone_args *args) 350{ 351 int error, ff = RFPROC | RFSTOPPED; 352 struct proc *p2; 353 struct thread *td2; 354 int exit_signal; 355 struct linux_emuldata *em; 356 357#ifdef DEBUG 358 if (ldebug(clone)) { 359 printf(ARGS(clone, "flags %x, stack %x, parent tid: %x, child tid: %x"), 360 (unsigned int)args->flags, (unsigned int)args->stack, 361 (unsigned int)args->parent_tidptr, (unsigned int)args->child_tidptr); 362 } 363#endif 364 365 exit_signal = args->flags & 0x000000ff; 366 if (exit_signal >= LINUX_NSIG) 367 return (EINVAL); 368 369 if (exit_signal <= LINUX_SIGTBLSZ) 370 exit_signal = linux_to_bsd_signal[_SIG_IDX(exit_signal)]; 371 372 if (args->flags & CLONE_VM) 373 ff |= RFMEM; 374 if (args->flags & CLONE_SIGHAND) 375 ff |= RFSIGSHARE; 376 if (!(args->flags & CLONE_FILES)) 377 ff |= RFFDG; 378 379 /* 380 * Attempt to detect when linux_clone(2) is used for creating 381 * kernel threads. Unfortunately despite the existence of the 382 * CLONE_THREAD flag, version of linuxthreads package used in 383 * most popular distros as of beginning of 2005 doesn't make 384 * any use of it. Therefore, this detection relay fully on 385 * empirical observation that linuxthreads sets certain 386 * combination of flags, so that we can make more or less 387 * precise detection and notify the FreeBSD kernel that several 388 * processes are in fact part of the same threading group, so 389 * that special treatment is necessary for signal delivery 390 * between those processes and fd locking. 391 */ 392 if ((args->flags & 0xffffff00) == THREADING_FLAGS) 393 ff |= RFTHREAD; 394 395 error = fork1(td, ff, 0, &p2); 396 if (error) 397 return (error); 398 399 /* create the emuldata */ 400 error = linux_proc_init(td, p2->p_pid, args->flags); 401 /* reference it - no need to check this */ 402 em = em_find(p2, EMUL_UNLOCKED); 403 KASSERT(em != NULL, ("clone: emuldata not found.\n")); 404 /* and adjust it */ 405 if (args->flags & CLONE_PARENT_SETTID) { 406 if (args->parent_tidptr == NULL) { 407 EMUL_UNLOCK(&emul_lock); 408 return (EINVAL); 409 } 410 error = copyout(&p2->p_pid, args->parent_tidptr, sizeof(p2->p_pid)); 411 if (error) { 412 EMUL_UNLOCK(&emul_lock); 413 return (error); 414 } 415 } 416 417 if (args->flags & CLONE_PARENT) { 418#ifdef DEBUG 419 printf("linux_clone: CLONE_PARENT\n"); 420#endif 421 } 422 423 if (args->flags & CLONE_THREAD) { 424 /* XXX: linux mangles pgrp and pptr somehow 425 * I think it might be this but I am not sure. 426 */ 427#ifdef notyet 428 p2->p_pgrp = td->td_proc->p_pgrp; 429 p2->p_pptr = td->td_proc->p_pptr; 430#endif 431 exit_signal = 0; 432#ifdef DEBUG 433 printf("linux_clone: CLONE_THREADS\n"); 434#endif 435 } 436 437 if (args->flags & CLONE_CHILD_SETTID) 438 em->child_set_tid = args->child_tidptr; 439 else 440 em->child_set_tid = NULL; 441 442 if (args->flags & CLONE_CHILD_CLEARTID) 443 em->child_clear_tid = args->child_tidptr; 444 else 445 em->child_clear_tid = NULL; 446 EMUL_UNLOCK(&emul_lock); 447 448 PROC_LOCK(p2); 449 p2->p_sigparent = exit_signal; 450 PROC_UNLOCK(p2); 451 td2 = FIRST_THREAD_IN_PROC(p2); 452 /* 453 * in a case of stack = NULL we are supposed to COW calling process stack 454 * this is what normal fork() does so we just keep the tf_esp arg intact 455 */ 456 if (args->stack) 457 td2->td_frame->tf_esp = (unsigned int)args->stack; 458 459 if (args->flags & CLONE_SETTLS) { 460 struct l_user_desc info; 461 int idx; 462 int a[2]; 463 struct segment_descriptor sd; 464 465 error = copyin((void *)td->td_frame->tf_esi, &info, sizeof(struct l_user_desc)); 466 if (error) 467 return (error); 468 469 idx = info.entry_number; 470 471 /* 472 * looks like we're getting the idx we returned 473 * in the set_thread_area() syscall 474 */ 475 if (idx != 6 && idx != 3) 476 return (EINVAL); 477 478 /* this doesnt happen in practice */ 479 if (idx == 6) { 480 /* we might copy out the entry_number as 3 */ 481 info.entry_number = 3; 482 error = copyout(&info, (void *) td->td_frame->tf_esi, sizeof(struct l_user_desc)); 483 if (error) 484 return (error); 485 } 486 487 a[0] = LDT_entry_a(&info); 488 a[1] = LDT_entry_b(&info); 489 490 memcpy(&sd, &a, sizeof(a)); 491#ifdef DEBUG 492 if (ldebug(clone)) 493 printf("Segment created in clone with CLONE_SETTLS: lobase: %x, hibase: %x, lolimit: %x, hilimit: %x, type: %i, dpl: %i, p: %i, xx: %i, def32: %i, gran: %i\n", sd.sd_lobase, 494 sd.sd_hibase, 495 sd.sd_lolimit, 496 sd.sd_hilimit, 497 sd.sd_type, 498 sd.sd_dpl, 499 sd.sd_p, 500 sd.sd_xx, 501 sd.sd_def32, 502 sd.sd_gran); 503#endif 504 505 /* this is taken from i386 version of cpu_set_user_tls() */ 506 critical_enter(); 507 /* set %gs */ 508 td2->td_pcb->pcb_gsd = sd; 509 PCPU_GET(fsgs_gdt)[1] = sd; 510 load_gs(GSEL(GUGS_SEL, SEL_UPL)); 511 critical_exit(); 512 } 513 514#ifdef DEBUG 515 if (ldebug(clone)) 516 printf(LMSG("clone: successful rfork to %ld, stack %p sig = %d"), 517 (long)p2->p_pid, args->stack, exit_signal); 518#endif 519 520 /* 521 * Make this runnable after we are finished with it. 522 */ 523 mtx_lock_spin(&sched_lock); 524 TD_SET_CAN_RUN(td2); 525 setrunqueue(td2, SRQ_BORING); 526 mtx_unlock_spin(&sched_lock); 527 528 td->td_retval[0] = p2->p_pid; 529 td->td_retval[1] = 0; 530 return (0); 531} 532 533/* XXX move */ 534struct l_mmap_argv { 535 l_caddr_t addr; 536 l_int len; 537 l_int prot; 538 l_int flags; 539 l_int fd; 540 l_int pos; 541}; 542 543#define STACK_SIZE (2 * 1024 * 1024) 544#define GUARD_SIZE (4 * PAGE_SIZE) 545 546static int linux_mmap_common(struct thread *, struct l_mmap_argv *); 547 548int 549linux_mmap2(struct thread *td, struct linux_mmap2_args *args) 550{ 551 struct l_mmap_argv linux_args; 552 553#ifdef DEBUG 554 if (ldebug(mmap2)) 555 printf(ARGS(mmap2, "%p, %d, %d, 0x%08x, %d, %d"), 556 (void *)args->addr, args->len, args->prot, 557 args->flags, args->fd, args->pgoff); 558#endif 559 560 linux_args.addr = (l_caddr_t)args->addr; 561 linux_args.len = args->len; 562 linux_args.prot = args->prot; 563 linux_args.flags = args->flags; 564 linux_args.fd = args->fd; 565 linux_args.pos = args->pgoff * PAGE_SIZE; 566 567 return (linux_mmap_common(td, &linux_args)); 568} 569 570int 571linux_mmap(struct thread *td, struct linux_mmap_args *args) 572{ 573 int error; 574 struct l_mmap_argv linux_args; 575 576 error = copyin(args->ptr, &linux_args, sizeof(linux_args)); 577 if (error) 578 return (error); 579 580#ifdef DEBUG 581 if (ldebug(mmap)) 582 printf(ARGS(mmap, "%p, %d, %d, 0x%08x, %d, %d"), 583 (void *)linux_args.addr, linux_args.len, linux_args.prot, 584 linux_args.flags, linux_args.fd, linux_args.pos); 585#endif 586 587 return (linux_mmap_common(td, &linux_args)); 588} 589 590static int 591linux_mmap_common(struct thread *td, struct l_mmap_argv *linux_args) 592{ 593 struct proc *p = td->td_proc; 594 struct mmap_args /* { 595 caddr_t addr; 596 size_t len; 597 int prot; 598 int flags; 599 int fd; 600 long pad; 601 off_t pos; 602 } */ bsd_args; 603 int error; 604 605 error = 0; 606 bsd_args.flags = 0; 607 if (linux_args->flags & LINUX_MAP_SHARED) 608 bsd_args.flags |= MAP_SHARED; 609 if (linux_args->flags & LINUX_MAP_PRIVATE) 610 bsd_args.flags |= MAP_PRIVATE; 611 if (linux_args->flags & LINUX_MAP_FIXED) 612 bsd_args.flags |= MAP_FIXED; 613 if (linux_args->flags & LINUX_MAP_ANON) 614 bsd_args.flags |= MAP_ANON; 615 else 616 bsd_args.flags |= MAP_NOSYNC; 617 if (linux_args->flags & LINUX_MAP_GROWSDOWN) { 618 bsd_args.flags |= MAP_STACK; 619 620 /* 621 * The linux MAP_GROWSDOWN option does not limit auto 622 * growth of the region. Linux mmap with this option 623 * takes as addr the inital BOS, and as len, the initial 624 * region size. It can then grow down from addr without 625 * limit. However, linux threads has an implicit internal 626 * limit to stack size of STACK_SIZE. Its just not 627 * enforced explicitly in linux. But, here we impose 628 * a limit of (STACK_SIZE - GUARD_SIZE) on the stack 629 * region, since we can do this with our mmap. 630 * 631 * Our mmap with MAP_STACK takes addr as the maximum 632 * downsize limit on BOS, and as len the max size of 633 * the region. It them maps the top SGROWSIZ bytes, 634 * and autgrows the region down, up to the limit 635 * in addr. 636 * 637 * If we don't use the MAP_STACK option, the effect 638 * of this code is to allocate a stack region of a 639 * fixed size of (STACK_SIZE - GUARD_SIZE). 640 */ 641 642 /* This gives us TOS */ 643 bsd_args.addr = linux_args->addr + linux_args->len; 644 645 if (bsd_args.addr > p->p_vmspace->vm_maxsaddr) { 646 /* 647 * Some linux apps will attempt to mmap 648 * thread stacks near the top of their 649 * address space. If their TOS is greater 650 * than vm_maxsaddr, vm_map_growstack() 651 * will confuse the thread stack with the 652 * process stack and deliver a SEGV if they 653 * attempt to grow the thread stack past their 654 * current stacksize rlimit. To avoid this, 655 * adjust vm_maxsaddr upwards to reflect 656 * the current stacksize rlimit rather 657 * than the maximum possible stacksize. 658 * It would be better to adjust the 659 * mmap'ed region, but some apps do not check 660 * mmap's return value. 661 */ 662 PROC_LOCK(p); 663 p->p_vmspace->vm_maxsaddr = (char *)USRSTACK - 664 lim_cur(p, RLIMIT_STACK); 665 PROC_UNLOCK(p); 666 } 667 668 /* This gives us our maximum stack size */ 669 if (linux_args->len > STACK_SIZE - GUARD_SIZE) 670 bsd_args.len = linux_args->len; 671 else 672 bsd_args.len = STACK_SIZE - GUARD_SIZE; 673 674 /* 675 * This gives us a new BOS. If we're using VM_STACK, then 676 * mmap will just map the top SGROWSIZ bytes, and let 677 * the stack grow down to the limit at BOS. If we're 678 * not using VM_STACK we map the full stack, since we 679 * don't have a way to autogrow it. 680 */ 681 bsd_args.addr -= bsd_args.len; 682 } else { 683 bsd_args.addr = linux_args->addr; 684 bsd_args.len = linux_args->len; 685 } 686 687 bsd_args.prot = linux_args->prot | PROT_READ; /* always required */ 688 if (linux_args->flags & LINUX_MAP_ANON) 689 bsd_args.fd = -1; 690 else 691 bsd_args.fd = linux_args->fd; 692 bsd_args.pos = linux_args->pos; 693 bsd_args.pad = 0; 694 695#ifdef DEBUG 696 if (ldebug(mmap)) 697 printf("-> %s(%p, %d, %d, 0x%08x, %d, 0x%x)\n", 698 __func__, 699 (void *)bsd_args.addr, bsd_args.len, bsd_args.prot, 700 bsd_args.flags, bsd_args.fd, (int)bsd_args.pos); 701#endif 702 error = mmap(td, &bsd_args); 703#ifdef DEBUG 704 if (ldebug(mmap)) 705 printf("-> %s() return: 0x%x (0x%08x)\n", 706 __func__, error, (u_int)td->td_retval[0]); 707#endif 708 return (error); 709} 710 711int 712linux_pipe(struct thread *td, struct linux_pipe_args *args) 713{ 714 int error; 715 int reg_edx; 716 717#ifdef DEBUG 718 if (ldebug(pipe)) 719 printf(ARGS(pipe, "*")); 720#endif 721 722 reg_edx = td->td_retval[1]; 723 error = pipe(td, 0); 724 if (error) { 725 td->td_retval[1] = reg_edx; 726 return (error); 727 } 728 729 error = copyout(td->td_retval, args->pipefds, 2*sizeof(int)); 730 if (error) { 731 td->td_retval[1] = reg_edx; 732 return (error); 733 } 734 735 td->td_retval[1] = reg_edx; 736 td->td_retval[0] = 0; 737 return (0); 738} 739 740int 741linux_ioperm(struct thread *td, struct linux_ioperm_args *args) 742{ 743 int error; 744 struct i386_ioperm_args iia; 745 746 iia.start = args->start; 747 iia.length = args->length; 748 iia.enable = args->enable; 749 mtx_lock(&Giant); 750 error = i386_set_ioperm(td, &iia); 751 mtx_unlock(&Giant); 752 return (error); 753} 754 755int 756linux_iopl(struct thread *td, struct linux_iopl_args *args) 757{ 758 int error; 759 760 if (args->level < 0 || args->level > 3) 761 return (EINVAL); 762 if ((error = suser(td)) != 0) 763 return (error); 764 if ((error = securelevel_gt(td->td_ucred, 0)) != 0) 765 return (error); 766 td->td_frame->tf_eflags = (td->td_frame->tf_eflags & ~PSL_IOPL) | 767 (args->level * (PSL_IOPL / 3)); 768 return (0); 769} 770 771int 772linux_modify_ldt(struct thread *td, struct linux_modify_ldt_args *uap) 773{ 774 int error; 775 struct i386_ldt_args ldt; 776 struct l_descriptor ld; 777 union descriptor desc; 778 779 if (uap->ptr == NULL) 780 return (EINVAL); 781 782 switch (uap->func) { 783 case 0x00: /* read_ldt */ 784 ldt.start = 0; 785 ldt.descs = uap->ptr; 786 ldt.num = uap->bytecount / sizeof(union descriptor); 787 mtx_lock(&Giant); 788 error = i386_get_ldt(td, &ldt); 789 td->td_retval[0] *= sizeof(union descriptor); 790 mtx_unlock(&Giant); 791 break; 792 case 0x01: /* write_ldt */ 793 case 0x11: /* write_ldt */ 794 if (uap->bytecount != sizeof(ld)) 795 return (EINVAL); 796 797 error = copyin(uap->ptr, &ld, sizeof(ld)); 798 if (error) 799 return (error); 800 801 ldt.start = ld.entry_number; 802 ldt.descs = &desc; 803 ldt.num = 1; 804 desc.sd.sd_lolimit = (ld.limit & 0x0000ffff); 805 desc.sd.sd_hilimit = (ld.limit & 0x000f0000) >> 16; 806 desc.sd.sd_lobase = (ld.base_addr & 0x00ffffff); 807 desc.sd.sd_hibase = (ld.base_addr & 0xff000000) >> 24; 808 desc.sd.sd_type = SDT_MEMRO | ((ld.read_exec_only ^ 1) << 1) | 809 (ld.contents << 2); 810 desc.sd.sd_dpl = 3; 811 desc.sd.sd_p = (ld.seg_not_present ^ 1); 812 desc.sd.sd_xx = 0; 813 desc.sd.sd_def32 = ld.seg_32bit; 814 desc.sd.sd_gran = ld.limit_in_pages; 815 mtx_lock(&Giant); 816 error = i386_set_ldt(td, &ldt, &desc); 817 mtx_unlock(&Giant); 818 break; 819 default: 820 error = EINVAL; 821 break; 822 } 823 824 if (error == EOPNOTSUPP) { 825 printf("linux: modify_ldt needs kernel option USER_LDT\n"); 826 error = ENOSYS; 827 } 828 829 return (error); 830} 831 832int 833linux_sigaction(struct thread *td, struct linux_sigaction_args *args) 834{ 835 l_osigaction_t osa; 836 l_sigaction_t act, oact; 837 int error; 838 839#ifdef DEBUG 840 if (ldebug(sigaction)) 841 printf(ARGS(sigaction, "%d, %p, %p"), 842 args->sig, (void *)args->nsa, (void *)args->osa); 843#endif 844 845 if (args->nsa != NULL) { 846 error = copyin(args->nsa, &osa, sizeof(l_osigaction_t)); 847 if (error) 848 return (error); 849 act.lsa_handler = osa.lsa_handler; 850 act.lsa_flags = osa.lsa_flags; 851 act.lsa_restorer = osa.lsa_restorer; 852 LINUX_SIGEMPTYSET(act.lsa_mask); 853 act.lsa_mask.__bits[0] = osa.lsa_mask; 854 } 855 856 error = linux_do_sigaction(td, args->sig, args->nsa ? &act : NULL, 857 args->osa ? &oact : NULL); 858 859 if (args->osa != NULL && !error) { 860 osa.lsa_handler = oact.lsa_handler; 861 osa.lsa_flags = oact.lsa_flags; 862 osa.lsa_restorer = oact.lsa_restorer; 863 osa.lsa_mask = oact.lsa_mask.__bits[0]; 864 error = copyout(&osa, args->osa, sizeof(l_osigaction_t)); 865 } 866 867 return (error); 868} 869 870/* 871 * Linux has two extra args, restart and oldmask. We dont use these, 872 * but it seems that "restart" is actually a context pointer that 873 * enables the signal to happen with a different register set. 874 */ 875int 876linux_sigsuspend(struct thread *td, struct linux_sigsuspend_args *args) 877{ 878 sigset_t sigmask; 879 l_sigset_t mask; 880 881#ifdef DEBUG 882 if (ldebug(sigsuspend)) 883 printf(ARGS(sigsuspend, "%08lx"), (unsigned long)args->mask); 884#endif 885 886 LINUX_SIGEMPTYSET(mask); 887 mask.__bits[0] = args->mask; 888 linux_to_bsd_sigset(&mask, &sigmask); 889 return (kern_sigsuspend(td, sigmask)); 890} 891 892int 893linux_rt_sigsuspend(struct thread *td, struct linux_rt_sigsuspend_args *uap) 894{ 895 l_sigset_t lmask; 896 sigset_t sigmask; 897 int error; 898 899#ifdef DEBUG 900 if (ldebug(rt_sigsuspend)) 901 printf(ARGS(rt_sigsuspend, "%p, %d"), 902 (void *)uap->newset, uap->sigsetsize); 903#endif 904 905 if (uap->sigsetsize != sizeof(l_sigset_t)) 906 return (EINVAL); 907 908 error = copyin(uap->newset, &lmask, sizeof(l_sigset_t)); 909 if (error) 910 return (error); 911 912 linux_to_bsd_sigset(&lmask, &sigmask); 913 return (kern_sigsuspend(td, sigmask)); 914} 915 916int 917linux_pause(struct thread *td, struct linux_pause_args *args) 918{ 919 struct proc *p = td->td_proc; 920 sigset_t sigmask; 921 922#ifdef DEBUG 923 if (ldebug(pause)) 924 printf(ARGS(pause, "")); 925#endif 926 927 PROC_LOCK(p); 928 sigmask = td->td_sigmask; 929 PROC_UNLOCK(p); 930 return (kern_sigsuspend(td, sigmask)); 931} 932 933int 934linux_sigaltstack(struct thread *td, struct linux_sigaltstack_args *uap) 935{ 936 stack_t ss, oss; 937 l_stack_t lss; 938 int error; 939 940#ifdef DEBUG 941 if (ldebug(sigaltstack)) 942 printf(ARGS(sigaltstack, "%p, %p"), uap->uss, uap->uoss); 943#endif 944 945 if (uap->uss != NULL) { 946 error = copyin(uap->uss, &lss, sizeof(l_stack_t)); 947 if (error) 948 return (error); 949 950 ss.ss_sp = lss.ss_sp; 951 ss.ss_size = lss.ss_size; 952 ss.ss_flags = linux_to_bsd_sigaltstack(lss.ss_flags); 953 } 954 error = kern_sigaltstack(td, (uap->uss != NULL) ? &ss : NULL, 955 (uap->uoss != NULL) ? &oss : NULL); 956 if (!error && uap->uoss != NULL) { 957 lss.ss_sp = oss.ss_sp; 958 lss.ss_size = oss.ss_size; 959 lss.ss_flags = bsd_to_linux_sigaltstack(oss.ss_flags); 960 error = copyout(&lss, uap->uoss, sizeof(l_stack_t)); 961 } 962 963 return (error); 964} 965 966int 967linux_ftruncate64(struct thread *td, struct linux_ftruncate64_args *args) 968{ 969 struct ftruncate_args sa; 970 971#ifdef DEBUG 972 if (ldebug(ftruncate64)) 973 printf(ARGS(ftruncate64, "%u, %jd"), args->fd, 974 (intmax_t)args->length); 975#endif 976 977 sa.fd = args->fd; 978 sa.pad = 0; 979 sa.length = args->length; 980 return ftruncate(td, &sa); 981} 982 983int 984linux_set_thread_area(struct thread *td, struct linux_set_thread_area_args *args) 985{ 986 struct l_user_desc info; 987 int error; 988 int idx; 989 int a[2]; 990 struct segment_descriptor sd; 991 992 error = copyin(args->desc, &info, sizeof(struct l_user_desc)); 993 if (error) 994 return (error); 995 996#ifdef DEBUG 997 if (ldebug(set_thread_area)) 998 printf(ARGS(set_thread_area, "%i, %x, %x, %i, %i, %i, %i, %i, %i\n"), 999 info.entry_number, 1000 info.base_addr, 1001 info.limit, 1002 info.seg_32bit, 1003 info.contents, 1004 info.read_exec_only, 1005 info.limit_in_pages, 1006 info.seg_not_present, 1007 info.useable); 1008#endif 1009 1010 idx = info.entry_number; 1011 /* 1012 * Semantics of linux version: every thread in the system has array 1013 * of 3 tls descriptors. 1st is GLIBC TLS, 2nd is WINE, 3rd unknown. This 1014 * syscall loads one of the selected tls decriptors with a value 1015 * and also loads GDT descriptors 6, 7 and 8 with the content of the per-thread 1016 * descriptors. 1017 * 1018 * Semantics of fbsd version: I think we can ignore that linux has 3 per-thread 1019 * descriptors and use just the 1st one. The tls_array[] is used only in 1020 * set/get-thread_area() syscalls and for loading the GDT descriptors. In fbsd 1021 * we use just one GDT descriptor for TLS so we will load just one. 1022 * XXX: this doesnt work when user-space process tries to use more then 1 TLS segment 1023 * comment in the linux sources says wine might do that. 1024 */ 1025 1026 /* 1027 * we support just GLIBC TLS now 1028 * we should let 3 proceed as well because we use this segment so 1029 * if code does two subsequent calls it should succeed 1030 */ 1031 if (idx != 6 && idx != -1 && idx != 3) 1032 return (EINVAL); 1033 1034 /* 1035 * we have to copy out the GDT entry we use 1036 * FreeBSD uses GDT entry #3 for storing %gs so load that 1037 * XXX: what if userspace program doesnt check this value and tries 1038 * to use 6, 7 or 8? 1039 */ 1040 idx = info.entry_number = 3; 1041 error = copyout(&info, args->desc, sizeof(struct l_user_desc)); 1042 if (error) 1043 return (error); 1044 1045 if (LDT_empty(&info)) { 1046 a[0] = 0; 1047 a[1] = 0; 1048 } else { 1049 a[0] = LDT_entry_a(&info); 1050 a[1] = LDT_entry_b(&info); 1051 } 1052 1053 memcpy(&sd, &a, sizeof(a)); 1054#ifdef DEBUG 1055 if (ldebug(set_thread_area)) 1056 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, 1057 sd.sd_hibase, 1058 sd.sd_lolimit, 1059 sd.sd_hilimit, 1060 sd.sd_type, 1061 sd.sd_dpl, 1062 sd.sd_p, 1063 sd.sd_xx, 1064 sd.sd_def32, 1065 sd.sd_gran); 1066#endif 1067 1068 /* this is taken from i386 version of cpu_set_user_tls() */ 1069 critical_enter(); 1070 /* set %gs */ 1071 td->td_pcb->pcb_gsd = sd; 1072 PCPU_GET(fsgs_gdt)[1] = sd; 1073 load_gs(GSEL(GUGS_SEL, SEL_UPL)); 1074 critical_exit(); 1075 1076 return (0); 1077} 1078 1079int 1080linux_get_thread_area(struct thread *td, struct linux_get_thread_area_args *args) 1081{ 1082 1083 struct l_user_desc info; 1084 int error; 1085 int idx; 1086 struct l_desc_struct desc; 1087 struct segment_descriptor sd; 1088 1089#ifdef DEBUG 1090 if (ldebug(get_thread_area)) 1091 printf(ARGS(get_thread_area, "%p"), args->desc); 1092#endif 1093 1094 error = copyin(args->desc, &info, sizeof(struct l_user_desc)); 1095 if (error) 1096 return (error); 1097 1098 idx = info.entry_number; 1099 /* XXX: I am not sure if we want 3 to be allowed too. */ 1100 if (idx != 6 && idx != 3) 1101 return (EINVAL); 1102 1103 idx = 3; 1104 1105 memset(&info, 0, sizeof(info)); 1106 1107 sd = PCPU_GET(fsgs_gdt)[1]; 1108 1109 memcpy(&desc, &sd, sizeof(desc)); 1110 1111 info.entry_number = idx; 1112 info.base_addr = GET_BASE(&desc); 1113 info.limit = GET_LIMIT(&desc); 1114 info.seg_32bit = GET_32BIT(&desc); 1115 info.contents = GET_CONTENTS(&desc); 1116 info.read_exec_only = !GET_WRITABLE(&desc); 1117 info.limit_in_pages = GET_LIMIT_PAGES(&desc); 1118 info.seg_not_present = !GET_PRESENT(&desc); 1119 info.useable = GET_USEABLE(&desc); 1120 1121 error = copyout(&info, args->desc, sizeof(struct l_user_desc)); 1122 if (error) 1123 return (EFAULT); 1124 1125 return (0); 1126} 1127 1128/* copied from kern/kern_time.c */ 1129int 1130linux_timer_create(struct thread *td, struct linux_timer_create_args *args) 1131{ 1132 return ktimer_create(td, (struct ktimer_create_args *) args); 1133} 1134 1135int 1136linux_timer_settime(struct thread *td, struct linux_timer_settime_args *args) 1137{ 1138 return ktimer_settime(td, (struct ktimer_settime_args *) args); 1139} 1140 1141int 1142linux_timer_gettime(struct thread *td, struct linux_timer_gettime_args *args) 1143{ 1144 return ktimer_gettime(td, (struct ktimer_gettime_args *) args); 1145} 1146 1147int 1148linux_timer_getoverrun(struct thread *td, struct linux_timer_getoverrun_args *args) 1149{ 1150 return ktimer_getoverrun(td, (struct ktimer_getoverrun_args *) args); 1151} 1152 1153int 1154linux_timer_delete(struct thread *td, struct linux_timer_delete_args *args) 1155{ 1156 return ktimer_delete(td, (struct ktimer_delete_args *) args); 1157} 1158 1159/* XXX: this wont work with module - convert it */ 1160int 1161linux_mq_open(struct thread *td, struct linux_mq_open_args *args) 1162{ 1163#ifdef P1003_1B_MQUEUE 1164 return kmq_open(td, (struct kmq_open_args *) args); 1165#else 1166 return (ENOSYS); 1167#endif 1168} 1169 1170int 1171linux_mq_unlink(struct thread *td, struct linux_mq_unlink_args *args) 1172{ 1173#ifdef P1003_1B_MQUEUE 1174 return kmq_unlink(td, (struct kmq_unlink_args *) args); 1175#else 1176 return (ENOSYS); 1177#endif 1178} 1179 1180int 1181linux_mq_timedsend(struct thread *td, struct linux_mq_timedsend_args *args) 1182{ 1183#ifdef P1003_1B_MQUEUE 1184 return kmq_timedsend(td, (struct kmq_timedsend_args *) args); 1185#else 1186 return (ENOSYS); 1187#endif 1188} 1189 1190int 1191linux_mq_timedreceive(struct thread *td, struct linux_mq_timedreceive_args *args) 1192{ 1193#ifdef P1003_1B_MQUEUE 1194 return kmq_timedreceive(td, (struct kmq_timedreceive_args *) args); 1195#else 1196 return (ENOSYS); 1197#endif 1198} 1199 1200int 1201linux_mq_notify(struct thread *td, struct linux_mq_notify_args *args) 1202{ 1203#ifdef P1003_1B_MQUEUE 1204 return kmq_notify(td, (struct kmq_notify_args *) args); 1205#else 1206 return (ENOSYS); 1207#endif 1208} 1209 1210int 1211linux_mq_getsetattr(struct thread *td, struct linux_mq_getsetattr_args *args) 1212{ 1213#ifdef P1003_1B_MQUEUE 1214 return kmq_setattr(td, (struct kmq_setattr_args *) args); 1215#else 1216 return (ENOSYS); 1217#endif 1218} 1219 1220