linux_machdep.c revision 161419
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 161419 2006-08-17 21:06:48Z 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; 319 320#ifdef DEBUG 321 if (ldebug(vfork)) 322 printf(ARGS(vfork, "")); 323#endif 324 325 if ((error = vfork(td, (struct vfork_args *)args)) != 0) 326 return (error); 327 /* Are we the child? */ 328 if (td->td_retval[1] == 1) 329 td->td_retval[0] = 0; 330 error = linux_proc_init(td, td->td_retval[0], 0); 331 if (error) 332 return (error); 333 return (0); 334} 335 336int 337linux_clone(struct thread *td, struct linux_clone_args *args) 338{ 339 int error, ff = RFPROC | RFSTOPPED; 340 struct proc *p2; 341 struct thread *td2; 342 int exit_signal; 343 struct linux_emuldata *em; 344 345#ifdef DEBUG 346 if (ldebug(clone)) { 347 printf(ARGS(clone, "flags %x, stack %x, parent tid: %x, child tid: %x"), 348 (unsigned int)args->flags, (unsigned int)args->stack, 349 (unsigned int)args->parent_tidptr, (unsigned int)args->child_tidptr); 350 } 351#endif 352 353 exit_signal = args->flags & 0x000000ff; 354 if (exit_signal >= LINUX_NSIG) 355 return (EINVAL); 356 357 if (exit_signal <= LINUX_SIGTBLSZ) 358 exit_signal = linux_to_bsd_signal[_SIG_IDX(exit_signal)]; 359 360 if (args->flags & CLONE_VM) 361 ff |= RFMEM; 362 if (args->flags & CLONE_SIGHAND) 363 ff |= RFSIGSHARE; 364 if (!(args->flags & CLONE_FILES)) 365 ff |= RFFDG; 366 367 /* 368 * Attempt to detect when linux_clone(2) is used for creating 369 * kernel threads. Unfortunately despite the existence of the 370 * CLONE_THREAD flag, version of linuxthreads package used in 371 * most popular distros as of beginning of 2005 doesn't make 372 * any use of it. Therefore, this detection relay fully on 373 * empirical observation that linuxthreads sets certain 374 * combination of flags, so that we can make more or less 375 * precise detection and notify the FreeBSD kernel that several 376 * processes are in fact part of the same threading group, so 377 * that special treatment is necessary for signal delivery 378 * between those processes and fd locking. 379 */ 380 if ((args->flags & 0xffffff00) == THREADING_FLAGS) 381 ff |= RFTHREAD; 382 383 error = fork1(td, ff, 0, &p2); 384 if (error) 385 return (error); 386 387 /* create the emuldata */ 388 error = linux_proc_init(td, p2->p_pid, args->flags); 389 /* reference it - no need to check this */ 390 em = em_find(p2, EMUL_UNLOCKED); 391 KASSERT(em != NULL, ("clone: emuldata not found.\n")); 392 /* and adjust it */ 393 if (args->flags & CLONE_PARENT_SETTID) { 394 if (args->parent_tidptr == NULL) { 395 EMUL_UNLOCK(&emul_lock); 396 return (EINVAL); 397 } 398 error = copyout(&p2->p_pid, args->parent_tidptr, sizeof(p2->p_pid)); 399 if (error) { 400 EMUL_UNLOCK(&emul_lock); 401 return (error); 402 } 403 } 404 405 if (args->flags & CLONE_PARENT) { 406#ifdef DEBUG 407 printf("linux_clone: CLONE_PARENT\n"); 408#endif 409 } 410 411 if (args->flags & CLONE_THREAD) { 412 /* XXX: linux mangles pgrp and pptr somehow 413 * I think it might be this but I am not sure. 414 */ 415#ifdef notyet 416 p2->p_pgrp = td->td_proc->p_pgrp; 417 p2->p_pptr = td->td_proc->p_pptr; 418#endif 419 exit_signal = 0; 420#ifdef DEBUG 421 printf("linux_clone: CLONE_THREADS\n"); 422#endif 423 } 424 425 if (args->flags & CLONE_CHILD_SETTID) 426 em->child_set_tid = args->child_tidptr; 427 else 428 em->child_set_tid = NULL; 429 430 if (args->flags & CLONE_CHILD_CLEARTID) 431 em->child_clear_tid = args->child_tidptr; 432 else 433 em->child_clear_tid = NULL; 434 EMUL_UNLOCK(&emul_lock); 435 436 PROC_LOCK(p2); 437 p2->p_sigparent = exit_signal; 438 PROC_UNLOCK(p2); 439 td2 = FIRST_THREAD_IN_PROC(p2); 440 /* 441 * in a case of stack = NULL we are supposed to COW calling process stack 442 * this is what normal fork() does so we just keep the tf_esp arg intact 443 */ 444 if (args->stack) 445 td2->td_frame->tf_esp = (unsigned int)args->stack; 446 447 if (args->flags & CLONE_SETTLS) { 448 struct l_user_desc info; 449 int idx; 450 int a[2]; 451 struct segment_descriptor sd; 452 453 error = copyin((void *)td->td_frame->tf_esi, &info, sizeof(struct l_user_desc)); 454 if (error) 455 return (error); 456 457 idx = info.entry_number; 458 459 /* 460 * looks like we're getting the idx we returned 461 * in the set_thread_area() syscall 462 */ 463 if (idx != 6 && idx != 3) 464 return (EINVAL); 465 466 /* this doesnt happen in practice */ 467 if (idx == 6) { 468 /* we might copy out the entry_number as 3 */ 469 info.entry_number = 3; 470 error = copyout(&info, (void *) td->td_frame->tf_esi, sizeof(struct l_user_desc)); 471 if (error) 472 return (error); 473 } 474 475 a[0] = LDT_entry_a(&info); 476 a[1] = LDT_entry_b(&info); 477 478 memcpy(&sd, &a, sizeof(a)); 479#ifdef DEBUG 480 if (ldebug(clone)) 481 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, 482 sd.sd_hibase, 483 sd.sd_lolimit, 484 sd.sd_hilimit, 485 sd.sd_type, 486 sd.sd_dpl, 487 sd.sd_p, 488 sd.sd_xx, 489 sd.sd_def32, 490 sd.sd_gran); 491#endif 492 493 /* this is taken from i386 version of cpu_set_user_tls() */ 494 critical_enter(); 495 /* set %gs */ 496 td2->td_pcb->pcb_gsd = sd; 497 PCPU_GET(fsgs_gdt)[1] = sd; 498 load_gs(GSEL(GUGS_SEL, SEL_UPL)); 499 critical_exit(); 500 } 501 502#ifdef DEBUG 503 if (ldebug(clone)) 504 printf(LMSG("clone: successful rfork to %ld, stack %p sig = %d"), 505 (long)p2->p_pid, args->stack, exit_signal); 506#endif 507 508 /* 509 * Make this runnable after we are finished with it. 510 */ 511 mtx_lock_spin(&sched_lock); 512 TD_SET_CAN_RUN(td2); 513 setrunqueue(td2, SRQ_BORING); 514 mtx_unlock_spin(&sched_lock); 515 516 td->td_retval[0] = p2->p_pid; 517 td->td_retval[1] = 0; 518 return (0); 519} 520 521/* XXX move */ 522struct l_mmap_argv { 523 l_caddr_t addr; 524 l_int len; 525 l_int prot; 526 l_int flags; 527 l_int fd; 528 l_int pos; 529}; 530 531#define STACK_SIZE (2 * 1024 * 1024) 532#define GUARD_SIZE (4 * PAGE_SIZE) 533 534static int linux_mmap_common(struct thread *, struct l_mmap_argv *); 535 536int 537linux_mmap2(struct thread *td, struct linux_mmap2_args *args) 538{ 539 struct l_mmap_argv linux_args; 540 541#ifdef DEBUG 542 if (ldebug(mmap2)) 543 printf(ARGS(mmap2, "%p, %d, %d, 0x%08x, %d, %d"), 544 (void *)args->addr, args->len, args->prot, 545 args->flags, args->fd, args->pgoff); 546#endif 547 548 linux_args.addr = (l_caddr_t)args->addr; 549 linux_args.len = args->len; 550 linux_args.prot = args->prot; 551 linux_args.flags = args->flags; 552 linux_args.fd = args->fd; 553 linux_args.pos = args->pgoff * PAGE_SIZE; 554 555 return (linux_mmap_common(td, &linux_args)); 556} 557 558int 559linux_mmap(struct thread *td, struct linux_mmap_args *args) 560{ 561 int error; 562 struct l_mmap_argv linux_args; 563 564 error = copyin(args->ptr, &linux_args, sizeof(linux_args)); 565 if (error) 566 return (error); 567 568#ifdef DEBUG 569 if (ldebug(mmap)) 570 printf(ARGS(mmap, "%p, %d, %d, 0x%08x, %d, %d"), 571 (void *)linux_args.addr, linux_args.len, linux_args.prot, 572 linux_args.flags, linux_args.fd, linux_args.pos); 573#endif 574 575 return (linux_mmap_common(td, &linux_args)); 576} 577 578static int 579linux_mmap_common(struct thread *td, struct l_mmap_argv *linux_args) 580{ 581 struct proc *p = td->td_proc; 582 struct mmap_args /* { 583 caddr_t addr; 584 size_t len; 585 int prot; 586 int flags; 587 int fd; 588 long pad; 589 off_t pos; 590 } */ bsd_args; 591 int error; 592 593 error = 0; 594 bsd_args.flags = 0; 595 if (linux_args->flags & LINUX_MAP_SHARED) 596 bsd_args.flags |= MAP_SHARED; 597 if (linux_args->flags & LINUX_MAP_PRIVATE) 598 bsd_args.flags |= MAP_PRIVATE; 599 if (linux_args->flags & LINUX_MAP_FIXED) 600 bsd_args.flags |= MAP_FIXED; 601 if (linux_args->flags & LINUX_MAP_ANON) 602 bsd_args.flags |= MAP_ANON; 603 else 604 bsd_args.flags |= MAP_NOSYNC; 605 if (linux_args->flags & LINUX_MAP_GROWSDOWN) { 606 bsd_args.flags |= MAP_STACK; 607 608 /* 609 * The linux MAP_GROWSDOWN option does not limit auto 610 * growth of the region. Linux mmap with this option 611 * takes as addr the inital BOS, and as len, the initial 612 * region size. It can then grow down from addr without 613 * limit. However, linux threads has an implicit internal 614 * limit to stack size of STACK_SIZE. Its just not 615 * enforced explicitly in linux. But, here we impose 616 * a limit of (STACK_SIZE - GUARD_SIZE) on the stack 617 * region, since we can do this with our mmap. 618 * 619 * Our mmap with MAP_STACK takes addr as the maximum 620 * downsize limit on BOS, and as len the max size of 621 * the region. It them maps the top SGROWSIZ bytes, 622 * and autgrows the region down, up to the limit 623 * in addr. 624 * 625 * If we don't use the MAP_STACK option, the effect 626 * of this code is to allocate a stack region of a 627 * fixed size of (STACK_SIZE - GUARD_SIZE). 628 */ 629 630 /* This gives us TOS */ 631 bsd_args.addr = linux_args->addr + linux_args->len; 632 633 if (bsd_args.addr > p->p_vmspace->vm_maxsaddr) { 634 /* 635 * Some linux apps will attempt to mmap 636 * thread stacks near the top of their 637 * address space. If their TOS is greater 638 * than vm_maxsaddr, vm_map_growstack() 639 * will confuse the thread stack with the 640 * process stack and deliver a SEGV if they 641 * attempt to grow the thread stack past their 642 * current stacksize rlimit. To avoid this, 643 * adjust vm_maxsaddr upwards to reflect 644 * the current stacksize rlimit rather 645 * than the maximum possible stacksize. 646 * It would be better to adjust the 647 * mmap'ed region, but some apps do not check 648 * mmap's return value. 649 */ 650 PROC_LOCK(p); 651 p->p_vmspace->vm_maxsaddr = (char *)USRSTACK - 652 lim_cur(p, RLIMIT_STACK); 653 PROC_UNLOCK(p); 654 } 655 656 /* This gives us our maximum stack size */ 657 if (linux_args->len > STACK_SIZE - GUARD_SIZE) 658 bsd_args.len = linux_args->len; 659 else 660 bsd_args.len = STACK_SIZE - GUARD_SIZE; 661 662 /* 663 * This gives us a new BOS. If we're using VM_STACK, then 664 * mmap will just map the top SGROWSIZ bytes, and let 665 * the stack grow down to the limit at BOS. If we're 666 * not using VM_STACK we map the full stack, since we 667 * don't have a way to autogrow it. 668 */ 669 bsd_args.addr -= bsd_args.len; 670 } else { 671 bsd_args.addr = linux_args->addr; 672 bsd_args.len = linux_args->len; 673 } 674 675 bsd_args.prot = linux_args->prot | PROT_READ; /* always required */ 676 if (linux_args->flags & LINUX_MAP_ANON) 677 bsd_args.fd = -1; 678 else 679 bsd_args.fd = linux_args->fd; 680 bsd_args.pos = linux_args->pos; 681 bsd_args.pad = 0; 682 683#ifdef DEBUG 684 if (ldebug(mmap)) 685 printf("-> %s(%p, %d, %d, 0x%08x, %d, 0x%x)\n", 686 __func__, 687 (void *)bsd_args.addr, bsd_args.len, bsd_args.prot, 688 bsd_args.flags, bsd_args.fd, (int)bsd_args.pos); 689#endif 690 error = mmap(td, &bsd_args); 691#ifdef DEBUG 692 if (ldebug(mmap)) 693 printf("-> %s() return: 0x%x (0x%08x)\n", 694 __func__, error, (u_int)td->td_retval[0]); 695#endif 696 return (error); 697} 698 699int 700linux_pipe(struct thread *td, struct linux_pipe_args *args) 701{ 702 int error; 703 int reg_edx; 704 705#ifdef DEBUG 706 if (ldebug(pipe)) 707 printf(ARGS(pipe, "*")); 708#endif 709 710 reg_edx = td->td_retval[1]; 711 error = pipe(td, 0); 712 if (error) { 713 td->td_retval[1] = reg_edx; 714 return (error); 715 } 716 717 error = copyout(td->td_retval, args->pipefds, 2*sizeof(int)); 718 if (error) { 719 td->td_retval[1] = reg_edx; 720 return (error); 721 } 722 723 td->td_retval[1] = reg_edx; 724 td->td_retval[0] = 0; 725 return (0); 726} 727 728int 729linux_ioperm(struct thread *td, struct linux_ioperm_args *args) 730{ 731 int error; 732 struct i386_ioperm_args iia; 733 734 iia.start = args->start; 735 iia.length = args->length; 736 iia.enable = args->enable; 737 mtx_lock(&Giant); 738 error = i386_set_ioperm(td, &iia); 739 mtx_unlock(&Giant); 740 return (error); 741} 742 743int 744linux_iopl(struct thread *td, struct linux_iopl_args *args) 745{ 746 int error; 747 748 if (args->level < 0 || args->level > 3) 749 return (EINVAL); 750 if ((error = suser(td)) != 0) 751 return (error); 752 if ((error = securelevel_gt(td->td_ucred, 0)) != 0) 753 return (error); 754 td->td_frame->tf_eflags = (td->td_frame->tf_eflags & ~PSL_IOPL) | 755 (args->level * (PSL_IOPL / 3)); 756 return (0); 757} 758 759int 760linux_modify_ldt(struct thread *td, struct linux_modify_ldt_args *uap) 761{ 762 int error; 763 struct i386_ldt_args ldt; 764 struct l_descriptor ld; 765 union descriptor desc; 766 767 if (uap->ptr == NULL) 768 return (EINVAL); 769 770 switch (uap->func) { 771 case 0x00: /* read_ldt */ 772 ldt.start = 0; 773 ldt.descs = uap->ptr; 774 ldt.num = uap->bytecount / sizeof(union descriptor); 775 mtx_lock(&Giant); 776 error = i386_get_ldt(td, &ldt); 777 td->td_retval[0] *= sizeof(union descriptor); 778 mtx_unlock(&Giant); 779 break; 780 case 0x01: /* write_ldt */ 781 case 0x11: /* write_ldt */ 782 if (uap->bytecount != sizeof(ld)) 783 return (EINVAL); 784 785 error = copyin(uap->ptr, &ld, sizeof(ld)); 786 if (error) 787 return (error); 788 789 ldt.start = ld.entry_number; 790 ldt.descs = &desc; 791 ldt.num = 1; 792 desc.sd.sd_lolimit = (ld.limit & 0x0000ffff); 793 desc.sd.sd_hilimit = (ld.limit & 0x000f0000) >> 16; 794 desc.sd.sd_lobase = (ld.base_addr & 0x00ffffff); 795 desc.sd.sd_hibase = (ld.base_addr & 0xff000000) >> 24; 796 desc.sd.sd_type = SDT_MEMRO | ((ld.read_exec_only ^ 1) << 1) | 797 (ld.contents << 2); 798 desc.sd.sd_dpl = 3; 799 desc.sd.sd_p = (ld.seg_not_present ^ 1); 800 desc.sd.sd_xx = 0; 801 desc.sd.sd_def32 = ld.seg_32bit; 802 desc.sd.sd_gran = ld.limit_in_pages; 803 mtx_lock(&Giant); 804 error = i386_set_ldt(td, &ldt, &desc); 805 mtx_unlock(&Giant); 806 break; 807 default: 808 error = EINVAL; 809 break; 810 } 811 812 if (error == EOPNOTSUPP) { 813 printf("linux: modify_ldt needs kernel option USER_LDT\n"); 814 error = ENOSYS; 815 } 816 817 return (error); 818} 819 820int 821linux_sigaction(struct thread *td, struct linux_sigaction_args *args) 822{ 823 l_osigaction_t osa; 824 l_sigaction_t act, oact; 825 int error; 826 827#ifdef DEBUG 828 if (ldebug(sigaction)) 829 printf(ARGS(sigaction, "%d, %p, %p"), 830 args->sig, (void *)args->nsa, (void *)args->osa); 831#endif 832 833 if (args->nsa != NULL) { 834 error = copyin(args->nsa, &osa, sizeof(l_osigaction_t)); 835 if (error) 836 return (error); 837 act.lsa_handler = osa.lsa_handler; 838 act.lsa_flags = osa.lsa_flags; 839 act.lsa_restorer = osa.lsa_restorer; 840 LINUX_SIGEMPTYSET(act.lsa_mask); 841 act.lsa_mask.__bits[0] = osa.lsa_mask; 842 } 843 844 error = linux_do_sigaction(td, args->sig, args->nsa ? &act : NULL, 845 args->osa ? &oact : NULL); 846 847 if (args->osa != NULL && !error) { 848 osa.lsa_handler = oact.lsa_handler; 849 osa.lsa_flags = oact.lsa_flags; 850 osa.lsa_restorer = oact.lsa_restorer; 851 osa.lsa_mask = oact.lsa_mask.__bits[0]; 852 error = copyout(&osa, args->osa, sizeof(l_osigaction_t)); 853 } 854 855 return (error); 856} 857 858/* 859 * Linux has two extra args, restart and oldmask. We dont use these, 860 * but it seems that "restart" is actually a context pointer that 861 * enables the signal to happen with a different register set. 862 */ 863int 864linux_sigsuspend(struct thread *td, struct linux_sigsuspend_args *args) 865{ 866 sigset_t sigmask; 867 l_sigset_t mask; 868 869#ifdef DEBUG 870 if (ldebug(sigsuspend)) 871 printf(ARGS(sigsuspend, "%08lx"), (unsigned long)args->mask); 872#endif 873 874 LINUX_SIGEMPTYSET(mask); 875 mask.__bits[0] = args->mask; 876 linux_to_bsd_sigset(&mask, &sigmask); 877 return (kern_sigsuspend(td, sigmask)); 878} 879 880int 881linux_rt_sigsuspend(struct thread *td, struct linux_rt_sigsuspend_args *uap) 882{ 883 l_sigset_t lmask; 884 sigset_t sigmask; 885 int error; 886 887#ifdef DEBUG 888 if (ldebug(rt_sigsuspend)) 889 printf(ARGS(rt_sigsuspend, "%p, %d"), 890 (void *)uap->newset, uap->sigsetsize); 891#endif 892 893 if (uap->sigsetsize != sizeof(l_sigset_t)) 894 return (EINVAL); 895 896 error = copyin(uap->newset, &lmask, sizeof(l_sigset_t)); 897 if (error) 898 return (error); 899 900 linux_to_bsd_sigset(&lmask, &sigmask); 901 return (kern_sigsuspend(td, sigmask)); 902} 903 904int 905linux_pause(struct thread *td, struct linux_pause_args *args) 906{ 907 struct proc *p = td->td_proc; 908 sigset_t sigmask; 909 910#ifdef DEBUG 911 if (ldebug(pause)) 912 printf(ARGS(pause, "")); 913#endif 914 915 PROC_LOCK(p); 916 sigmask = td->td_sigmask; 917 PROC_UNLOCK(p); 918 return (kern_sigsuspend(td, sigmask)); 919} 920 921int 922linux_sigaltstack(struct thread *td, struct linux_sigaltstack_args *uap) 923{ 924 stack_t ss, oss; 925 l_stack_t lss; 926 int error; 927 928#ifdef DEBUG 929 if (ldebug(sigaltstack)) 930 printf(ARGS(sigaltstack, "%p, %p"), uap->uss, uap->uoss); 931#endif 932 933 if (uap->uss != NULL) { 934 error = copyin(uap->uss, &lss, sizeof(l_stack_t)); 935 if (error) 936 return (error); 937 938 ss.ss_sp = lss.ss_sp; 939 ss.ss_size = lss.ss_size; 940 ss.ss_flags = linux_to_bsd_sigaltstack(lss.ss_flags); 941 } 942 error = kern_sigaltstack(td, (uap->uss != NULL) ? &ss : NULL, 943 (uap->uoss != NULL) ? &oss : NULL); 944 if (!error && uap->uoss != NULL) { 945 lss.ss_sp = oss.ss_sp; 946 lss.ss_size = oss.ss_size; 947 lss.ss_flags = bsd_to_linux_sigaltstack(oss.ss_flags); 948 error = copyout(&lss, uap->uoss, sizeof(l_stack_t)); 949 } 950 951 return (error); 952} 953 954int 955linux_ftruncate64(struct thread *td, struct linux_ftruncate64_args *args) 956{ 957 struct ftruncate_args sa; 958 959#ifdef DEBUG 960 if (ldebug(ftruncate64)) 961 printf(ARGS(ftruncate64, "%u, %jd"), args->fd, 962 (intmax_t)args->length); 963#endif 964 965 sa.fd = args->fd; 966 sa.pad = 0; 967 sa.length = args->length; 968 return ftruncate(td, &sa); 969} 970 971int 972linux_set_thread_area(struct thread *td, struct linux_set_thread_area_args *args) 973{ 974 struct l_user_desc info; 975 int error; 976 int idx; 977 int a[2]; 978 struct segment_descriptor sd; 979 980 error = copyin(args->desc, &info, sizeof(struct l_user_desc)); 981 if (error) 982 return (error); 983 984#ifdef DEBUG 985 if (ldebug(set_thread_area)) 986 printf(ARGS(set_thread_area, "%i, %x, %x, %i, %i, %i, %i, %i, %i\n"), 987 info.entry_number, 988 info.base_addr, 989 info.limit, 990 info.seg_32bit, 991 info.contents, 992 info.read_exec_only, 993 info.limit_in_pages, 994 info.seg_not_present, 995 info.useable); 996#endif 997 998 idx = info.entry_number; 999 /* 1000 * Semantics of linux version: every thread in the system has array 1001 * of 3 tls descriptors. 1st is GLIBC TLS, 2nd is WINE, 3rd unknown. This 1002 * syscall loads one of the selected tls decriptors with a value 1003 * and also loads GDT descriptors 6, 7 and 8 with the content of the per-thread 1004 * descriptors. 1005 * 1006 * Semantics of fbsd version: I think we can ignore that linux has 3 per-thread 1007 * descriptors and use just the 1st one. The tls_array[] is used only in 1008 * set/get-thread_area() syscalls and for loading the GDT descriptors. In fbsd 1009 * we use just one GDT descriptor for TLS so we will load just one. 1010 * XXX: this doesnt work when user-space process tries to use more then 1 TLS segment 1011 * comment in the linux sources says wine might do that. 1012 */ 1013 1014 /* 1015 * we support just GLIBC TLS now 1016 * we should let 3 proceed as well because we use this segment so 1017 * if code does two subsequent calls it should succeed 1018 */ 1019 if (idx != 6 && idx != -1 && idx != 3) 1020 return (EINVAL); 1021 1022 /* 1023 * we have to copy out the GDT entry we use 1024 * FreeBSD uses GDT entry #3 for storing %gs so load that 1025 * XXX: what if userspace program doesnt check this value and tries 1026 * to use 6, 7 or 8? 1027 */ 1028 idx = info.entry_number = 3; 1029 error = copyout(&info, args->desc, sizeof(struct l_user_desc)); 1030 if (error) 1031 return (error); 1032 1033 if (LDT_empty(&info)) { 1034 a[0] = 0; 1035 a[1] = 0; 1036 } else { 1037 a[0] = LDT_entry_a(&info); 1038 a[1] = LDT_entry_b(&info); 1039 } 1040 1041 memcpy(&sd, &a, sizeof(a)); 1042#ifdef DEBUG 1043 if (ldebug(set_thread_area)) 1044 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, 1045 sd.sd_hibase, 1046 sd.sd_lolimit, 1047 sd.sd_hilimit, 1048 sd.sd_type, 1049 sd.sd_dpl, 1050 sd.sd_p, 1051 sd.sd_xx, 1052 sd.sd_def32, 1053 sd.sd_gran); 1054#endif 1055 1056 /* this is taken from i386 version of cpu_set_user_tls() */ 1057 critical_enter(); 1058 /* set %gs */ 1059 td->td_pcb->pcb_gsd = sd; 1060 PCPU_GET(fsgs_gdt)[1] = sd; 1061 load_gs(GSEL(GUGS_SEL, SEL_UPL)); 1062 critical_exit(); 1063 1064 return (0); 1065} 1066 1067int 1068linux_get_thread_area(struct thread *td, struct linux_get_thread_area_args *args) 1069{ 1070 1071 struct l_user_desc info; 1072 int error; 1073 int idx; 1074 struct l_desc_struct desc; 1075 struct segment_descriptor sd; 1076 1077#ifdef DEBUG 1078 if (ldebug(get_thread_area)) 1079 printf(ARGS(get_thread_area, "%p"), args->desc); 1080#endif 1081 1082 error = copyin(args->desc, &info, sizeof(struct l_user_desc)); 1083 if (error) 1084 return (error); 1085 1086 idx = info.entry_number; 1087 /* XXX: I am not sure if we want 3 to be allowed too. */ 1088 if (idx != 6 && idx != 3) 1089 return (EINVAL); 1090 1091 idx = 3; 1092 1093 memset(&info, 0, sizeof(info)); 1094 1095 sd = PCPU_GET(fsgs_gdt)[1]; 1096 1097 memcpy(&desc, &sd, sizeof(desc)); 1098 1099 info.entry_number = idx; 1100 info.base_addr = GET_BASE(&desc); 1101 info.limit = GET_LIMIT(&desc); 1102 info.seg_32bit = GET_32BIT(&desc); 1103 info.contents = GET_CONTENTS(&desc); 1104 info.read_exec_only = !GET_WRITABLE(&desc); 1105 info.limit_in_pages = GET_LIMIT_PAGES(&desc); 1106 info.seg_not_present = !GET_PRESENT(&desc); 1107 info.useable = GET_USEABLE(&desc); 1108 1109 error = copyout(&info, args->desc, sizeof(struct l_user_desc)); 1110 if (error) 1111 return (EFAULT); 1112 1113 return (0); 1114} 1115 1116/* copied from kern/kern_time.c */ 1117int 1118linux_timer_create(struct thread *td, struct linux_timer_create_args *args) 1119{ 1120 return ktimer_create(td, (struct ktimer_create_args *) args); 1121} 1122 1123int 1124linux_timer_settime(struct thread *td, struct linux_timer_settime_args *args) 1125{ 1126 return ktimer_settime(td, (struct ktimer_settime_args *) args); 1127} 1128 1129int 1130linux_timer_gettime(struct thread *td, struct linux_timer_gettime_args *args) 1131{ 1132 return ktimer_gettime(td, (struct ktimer_gettime_args *) args); 1133} 1134 1135int 1136linux_timer_getoverrun(struct thread *td, struct linux_timer_getoverrun_args *args) 1137{ 1138 return ktimer_getoverrun(td, (struct ktimer_getoverrun_args *) args); 1139} 1140 1141int 1142linux_timer_delete(struct thread *td, struct linux_timer_delete_args *args) 1143{ 1144 return ktimer_delete(td, (struct ktimer_delete_args *) args); 1145} 1146 1147/* XXX: this wont work with module - convert it */ 1148int 1149linux_mq_open(struct thread *td, struct linux_mq_open_args *args) 1150{ 1151#ifdef P1003_1B_MQUEUE 1152 return kmq_open(td, (struct kmq_open_args *) args); 1153#else 1154 return (ENOSYS); 1155#endif 1156} 1157 1158int 1159linux_mq_unlink(struct thread *td, struct linux_mq_unlink_args *args) 1160{ 1161#ifdef P1003_1B_MQUEUE 1162 return kmq_unlink(td, (struct kmq_unlink_args *) args); 1163#else 1164 return (ENOSYS); 1165#endif 1166} 1167 1168int 1169linux_mq_timedsend(struct thread *td, struct linux_mq_timedsend_args *args) 1170{ 1171#ifdef P1003_1B_MQUEUE 1172 return kmq_timedsend(td, (struct kmq_timedsend_args *) args); 1173#else 1174 return (ENOSYS); 1175#endif 1176} 1177 1178int 1179linux_mq_timedreceive(struct thread *td, struct linux_mq_timedreceive_args *args) 1180{ 1181#ifdef P1003_1B_MQUEUE 1182 return kmq_timedreceive(td, (struct kmq_timedreceive_args *) args); 1183#else 1184 return (ENOSYS); 1185#endif 1186} 1187 1188int 1189linux_mq_notify(struct thread *td, struct linux_mq_notify_args *args) 1190{ 1191#ifdef P1003_1B_MQUEUE 1192 return kmq_notify(td, (struct kmq_notify_args *) args); 1193#else 1194 return (ENOSYS); 1195#endif 1196} 1197 1198int 1199linux_mq_getsetattr(struct thread *td, struct linux_mq_getsetattr_args *args) 1200{ 1201#ifdef P1003_1B_MQUEUE 1202 return kmq_setattr(td, (struct kmq_setattr_args *) args); 1203#else 1204 return (ENOSYS); 1205#endif 1206} 1207 1208