sys/emulation/linux/i386/linux_machdep.c

   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  * $FreeBSD: src/sys/i386/linux/linux_machdep.c,v 1.6.2.4 2001/11/05 19:08:23 marcel Exp $
  29  * $DragonFly: src/sys/emulation/linux/i386/linux_machdep.c,v 1.23 2007/07/30 17:41:23 pavalos Exp $
  30  */
  31
  32 #include <sys/param.h>
  33 #include <sys/systm.h>
  34 #include <sys/imgact.h>
  35 #include <sys/kern_syscall.h>
  36 #include <sys/lock.h>
  37 #include <sys/mman.h>
  38 #include <sys/nlookup.h>
  39 #include <sys/proc.h>
  40 #include <sys/priv.h>
  41 #include <sys/resource.h>
  42 #include <sys/resourcevar.h>
  43 #include <sys/sysproto.h>
  44 #include <sys/unistd.h>
  45 #include <sys/wait.h>
  46
  47 #include <machine/frame.h>
  48 #include <machine/psl.h>
  49 #include <machine/segments.h>
  50 #include <machine/sysarch.h>
  51
  52 #include <vm/vm.h>
  53 #include <vm/pmap.h>
  54 #include <vm/vm_map.h>
  55
  56 #include <sys/mplock2.h>
  57
  58 #include "linux.h"
  59 #include "linux_proto.h"
  60 #include "../linux_ipc.h"
  61 #include "../linux_signal.h"
  62 #include "../linux_util.h"
  63
  64 struct l_descriptor {
  65         l_uint          entry_number;
  66         l_ulong         base_addr;
  67         l_uint          limit;
  68         l_uint          seg_32bit:1;
  69         l_uint          contents:2;
  70         l_uint          read_exec_only:1;
  71         l_uint          limit_in_pages:1;
  72         l_uint          seg_not_present:1;
  73         l_uint          useable:1;
  74 };
  75
  76 struct l_old_select_argv {
  77         l_int           nfds;
  78         l_fd_set        *readfds;
  79         l_fd_set        *writefds;
  80         l_fd_set        *exceptfds;
  81         struct l_timeval        *timeout;
  82 };
  83
  84 int
  85 linux_to_bsd_sigaltstack(int lsa)
  86 {
  87         int bsa = 0;
  88
  89         if (lsa & LINUX_SS_DISABLE)
  90                 bsa |= SS_DISABLE;
  91         if (lsa & LINUX_SS_ONSTACK)
  92                 bsa |= SS_ONSTACK;
  93         return (bsa);
  94 }
  95
  96 int
  97 bsd_to_linux_sigaltstack(int bsa)
  98 {
  99         int lsa = 0;
 100
 101         if (bsa & SS_DISABLE)
 102                 lsa |= LINUX_SS_DISABLE;
 103         if (bsa & SS_ONSTACK)
 104                 lsa |= LINUX_SS_ONSTACK;
 105         return (lsa);
 106 }
 107
 108 /*
 109  * MPALMOSTSAFE
 110  */
 111 int
 112 sys_linux_execve(struct linux_execve_args *args)
 113 {
 114         struct nlookupdata nd;
 115         struct image_args exec_args;
 116         char *path;
 117         int error;
 118
 119         error = linux_copyin_path(args->path, &path, LINUX_PATH_EXISTS);
 120         if (error)
 121                 return (error);
 122 #ifdef DEBUG
 123         if (ldebug(execve))
 124                 kprintf(ARGS(execve, "%s"), path);
 125 #endif
 126         get_mplock();
 127         error = nlookup_init(&nd, path, UIO_SYSSPACE, NLC_FOLLOW);
 128         bzero(&exec_args, sizeof(exec_args));
 129         if (error == 0) {
 130                 error = exec_copyin_args(&exec_args, path, PATH_SYSSPACE,
 131                                         args->argp, args->envp);
 132         }
 133         if (error == 0)
 134                 error = kern_execve(&nd, &exec_args);
 135         nlookup_done(&nd);
 136
 137         /*
 138          * The syscall result is returned in registers to the new program.
 139          * Linux will register %edx as an atexit function and we must be
 140          * sure to set it to 0.  XXX
 141          */
 142         if (error == 0)
 143                 args->sysmsg_result64 = 0;
 144
 145         exec_free_args(&exec_args);
 146         linux_free_path(&path);
 147
 148         if (error < 0) {
 149                 /* We hit a lethal error condition.  Let's die now. */
 150                 exit1(W_EXITCODE(0, SIGABRT));
 151                 /* NOTREACHED */
 152         }
 153         rel_mplock();
 154
 155         return(error);
 156 }
 157
 158 struct l_ipc_kludge {
 159         struct l_msgbuf *msgp;
 160         l_long msgtyp;
 161 };
 162
 163 /*
 164  * MPALMOSTSAFE
 165  */
 166 int
 167 sys_linux_ipc(struct linux_ipc_args *args)
 168 {
 169         int error = 0;
 170
 171         get_mplock();
 172
 173         switch (args->what & 0xFFFF) {
 174         case LINUX_SEMOP: {
 175                 struct linux_semop_args a;
 176
 177                 a.semid = args->arg1;
 178                 a.tsops = args->ptr;
 179                 a.nsops = args->arg2;
 180                 a.sysmsg_lresult = 0;
 181                 error = linux_semop(&a);
 182                 args->sysmsg_lresult = a.sysmsg_lresult;
 183                 break;
 184         }
 185         case LINUX_SEMGET: {
 186                 struct linux_semget_args a;
 187
 188                 a.key = args->arg1;
 189                 a.nsems = args->arg2;
 190                 a.semflg = args->arg3;
 191                 a.sysmsg_lresult = 0;
 192                 error = linux_semget(&a);
 193                 args->sysmsg_lresult = a.sysmsg_lresult;
 194                 break;
 195         }
 196         case LINUX_SEMCTL: {
 197                 struct linux_semctl_args a;
 198                 int error;
 199
 200                 a.semid = args->arg1;
 201                 a.semnum = args->arg2;
 202                 a.cmd = args->arg3;
 203                 a.sysmsg_lresult = 0;
 204                 error = copyin((caddr_t)args->ptr, &a.arg, sizeof(a.arg));
 205                 if (error)
 206                         break;
 207                 error = linux_semctl(&a);
 208                 args->sysmsg_lresult = a.sysmsg_lresult;
 209                 break;
 210         }
 211         case LINUX_MSGSND: {
 212                 struct linux_msgsnd_args a;
 213
 214                 a.msqid = args->arg1;
 215                 a.msgp = args->ptr;
 216                 a.msgsz = args->arg2;
 217                 a.msgflg = args->arg3;
 218                 a.sysmsg_lresult = 0;
 219                 error = linux_msgsnd(&a);
 220                 args->sysmsg_lresult = a.sysmsg_lresult;
 221                 break;
 222         }
 223         case LINUX_MSGRCV: {
 224                 struct linux_msgrcv_args a;
 225
 226                 a.msqid = args->arg1;
 227                 a.msgsz = args->arg2;
 228                 a.msgflg = args->arg3;
 229                 a.sysmsg_lresult = 0;
 230                 if ((args->what >> 16) == 0) {
 231                         struct l_ipc_kludge tmp;
 232                         int error;
 233
 234                         if (args->ptr == NULL) {
 235                                 error = EINVAL;
 236                                 break;
 237                         }
 238                         error = copyin((caddr_t)args->ptr, &tmp, sizeof(tmp));
 239                         if (error)
 240                                 break;
 241                         a.msgp = tmp.msgp;
 242                         a.msgtyp = tmp.msgtyp;
 243                 } else {
 244                         a.msgp = args->ptr;
 245                         a.msgtyp = args->arg5;
 246                 }
 247                 error = linux_msgrcv(&a);
 248                 args->sysmsg_lresult = a.sysmsg_lresult;
 249                 break;
 250         }
 251         case LINUX_MSGGET: {
 252                 struct linux_msgget_args a;
 253
 254                 a.key = args->arg1;
 255                 a.msgflg = args->arg2;
 256                 a.sysmsg_lresult = 0;
 257                 error = linux_msgget(&a);
 258                 args->sysmsg_lresult = a.sysmsg_lresult;
 259                 break;
 260         }
 261         case LINUX_MSGCTL: {
 262                 struct linux_msgctl_args a;
 263
 264                 a.msqid = args->arg1;
 265                 a.cmd = args->arg2;
 266                 a.buf = args->ptr;
 267                 a.sysmsg_lresult = 0;
 268                 error = linux_msgctl(&a);
 269                 args->sysmsg_lresult = a.sysmsg_lresult;
 270                 break;
 271         }
 272         case LINUX_SHMAT: {
 273                 struct linux_shmat_args a;
 274
 275                 a.shmid = args->arg1;
 276                 a.shmaddr = args->ptr;
 277                 a.shmflg = args->arg2;
 278                 a.raddr = (l_ulong *)args->arg3;
 279                 a.sysmsg_lresult = 0;
 280                 error = linux_shmat(&a);
 281                 args->sysmsg_lresult = a.sysmsg_lresult;
 282                 break;
 283         }
 284         case LINUX_SHMDT: {
 285                 struct linux_shmdt_args a;
 286
 287                 a.shmaddr = args->ptr;
 288                 a.sysmsg_lresult = 0;
 289                 error = linux_shmdt(&a);
 290                 args->sysmsg_lresult = a.sysmsg_lresult;
 291                 break;
 292         }
 293         case LINUX_SHMGET: {
 294                 struct linux_shmget_args a;
 295
 296                 a.key = args->arg1;
 297                 a.size = args->arg2;
 298                 a.shmflg = args->arg3;
 299                 a.sysmsg_lresult = 0;
 300                 error = linux_shmget(&a);
 301                 args->sysmsg_lresult = a.sysmsg_lresult;
 302                 break;
 303         }
 304         case LINUX_SHMCTL: {
 305                 struct linux_shmctl_args a;
 306
 307                 a.shmid = args->arg1;
 308                 a.cmd = args->arg2;
 309                 a.buf = args->ptr;
 310                 a.sysmsg_lresult = 0;
 311                 error = linux_shmctl(&a);
 312                 args->sysmsg_lresult = a.sysmsg_lresult;
 313                 break;
 314         }
 315         default:
 316                 error = EINVAL;
 317                 break;
 318         }
 319         rel_mplock();
 320         return(error);
 321 }
 322
 323 /*
 324  * MPSAFE
 325  */
 326 int
 327 sys_linux_old_select(struct linux_old_select_args *args)
 328 {
 329         struct l_old_select_argv linux_args;
 330         struct linux_select_args newsel;
 331         int error;
 332
 333 #ifdef DEBUG
 334         if (ldebug(old_select))
 335                 kprintf(ARGS(old_select, "%p"), args->ptr);
 336 #endif
 337
 338         error = copyin((caddr_t)args->ptr, &linux_args, sizeof(linux_args));
 339         if (error)
 340                 return (error);
 341
 342         newsel.sysmsg_iresult = 0;
 343         newsel.nfds = linux_args.nfds;
 344         newsel.readfds = linux_args.readfds;
 345         newsel.writefds = linux_args.writefds;
 346         newsel.exceptfds = linux_args.exceptfds;
 347         newsel.timeout = linux_args.timeout;
 348         error = sys_linux_select(&newsel);
 349         args->sysmsg_iresult = newsel.sysmsg_iresult;
 350         return(error);
 351 }
 352
 353 /*
 354  * MPSAFE
 355  */
 356 int
 357 sys_linux_fork(struct linux_fork_args *args)
 358 {
 359         int error;
 360
 361 #ifdef DEBUG
 362         if (ldebug(fork))
 363                 kprintf(ARGS(fork, ""));
 364 #endif
 365
 366         if ((error = sys_fork((struct fork_args *)args)) != 0)
 367                 return (error);
 368
 369         if (args->sysmsg_iresult == 1)
 370                 args->sysmsg_iresult = 0;
 371         return (0);
 372 }
 373
 374 /*
 375  * MPSAFE
 376  */
 377 int
 378 sys_linux_exit_group(struct linux_exit_group_args *args)
 379 {
 380         struct exit_args newargs;
 381         int error;
 382
 383         newargs.sysmsg_iresult = 0;
 384         newargs.rval = args->rval;
 385         error = sys_exit(&newargs);
 386         args->sysmsg_iresult = newargs.sysmsg_iresult;
 387         return (error);
 388 }
 389
 390 /*
 391  * MPSAFE
 392  */
 393 int
 394 sys_linux_vfork(struct linux_vfork_args *args)
 395 {
 396         int error;
 397
 398 #ifdef DEBUG
 399         if (ldebug(vfork))
 400                 kprintf(ARGS(vfork, ""));
 401 #endif
 402
 403         if ((error = sys_vfork((struct vfork_args *)args)) != 0)
 404                 return (error);
 405         /* Are we the child? */
 406         if (args->sysmsg_iresult == 1)
 407                 args->sysmsg_iresult = 0;
 408         return (0);
 409 }
 410
 411 #define CLONE_VM        0x100
 412 #define CLONE_FS        0x200
 413 #define CLONE_FILES     0x400
 414 #define CLONE_SIGHAND   0x800
 415 #define CLONE_PID       0x1000
 416
 417 /*
 418  * MPALMOSTSAFE
 419  */
 420 int
 421 sys_linux_clone(struct linux_clone_args *args)
 422 {
 423         int error, ff = RFPROC;
 424         struct proc *p2 = NULL;
 425         int exit_signal;
 426         vm_offset_t start;
 427         struct rfork_args rf_args;
 428
 429 #ifdef DEBUG
 430         if (ldebug(clone)) {
 431                 kprintf(ARGS(clone, "flags %x, stack %x"),
 432                     (unsigned int)args->flags, (unsigned int)args->stack);
 433                 if (args->flags & CLONE_PID)
 434                         kprintf(LMSG("CLONE_PID not yet supported"));
 435         }
 436 #endif
 437
 438         if (!args->stack)
 439                 return (EINVAL);
 440
 441         exit_signal = args->flags & 0x000000ff;
 442         if (exit_signal >= LINUX_NSIG)
 443                 return (EINVAL);
 444
 445         if (exit_signal <= LINUX_SIGTBLSZ)
 446                 exit_signal = linux_to_bsd_signal[_SIG_IDX(exit_signal)];
 447
 448         /* RFTHREAD probably not necessary here, but it shouldn't hurt */
 449         ff |= RFTHREAD;
 450
 451         if (args->flags & CLONE_VM)
 452                 ff |= RFMEM;
 453         if (args->flags & CLONE_SIGHAND)
 454                 ff |= RFSIGSHARE;
 455         if (!(args->flags & CLONE_FILES))
 456                 ff |= RFFDG;
 457
 458         error = 0;
 459         start = 0;
 460
 461         rf_args.flags = ff;
 462         rf_args.sysmsg_iresult = 0;
 463         get_mplock();
 464         if ((error = sys_rfork(&rf_args)) == 0) {
 465                 args->sysmsg_iresult = rf_args.sysmsg_iresult;
 466
 467                 p2 = pfind(rf_args.sysmsg_iresult);
 468                 if (p2 == NULL)
 469                         error = ESRCH;
 470         }
 471         rel_mplock();
 472         if (error == 0) {
 473                 p2->p_sigparent = exit_signal;
 474                 ONLY_LWP_IN_PROC(p2)->lwp_md.md_regs->tf_esp =
 475                                                 (unsigned long)args->stack;
 476
 477 #ifdef DEBUG
 478                 if (ldebug(clone))
 479                         kprintf(LMSG("clone: successful rfork to %ld"),
 480                             (long)p2->p_pid);
 481 #endif
 482         }
 483         return (error);
 484 }
 485
 486 /* XXX move */
 487 struct l_mmap_argv {
 488         l_caddr_t       addr;
 489         l_int           len;
 490         l_int           prot;
 491         l_int           flags;
 492         l_int           fd;
 493         l_int           pos;
 494 };
 495
 496 #define STACK_SIZE  (2 * 1024 * 1024)
 497 #define GUARD_SIZE  (4 * PAGE_SIZE)
 498
 499 /*
 500  * MPALMOSTSAFE
 501  */
 502 static int
 503 linux_mmap_common(caddr_t linux_addr, size_t linux_len, int linux_prot,
 504                   int linux_flags, int linux_fd, off_t pos, void **res)
 505 {
 506         struct thread *td = curthread;
 507         struct proc *p = td->td_proc;
 508         caddr_t addr;
 509         void *new;
 510         int error, flags, len, prot, fd;
 511
 512         flags = 0;
 513         if (linux_flags & LINUX_MAP_SHARED)
 514                 flags |= MAP_SHARED;
 515         if (linux_flags & LINUX_MAP_PRIVATE)
 516                 flags |= MAP_PRIVATE;
 517         if (linux_flags & LINUX_MAP_FIXED)
 518                 flags |= MAP_FIXED;
 519         if (linux_flags & LINUX_MAP_ANON) {
 520                 flags |= MAP_ANON;
 521         } else {
 522                 flags |= MAP_NOSYNC;
 523         }
 524
 525         get_mplock();
 526
 527         if (linux_flags & LINUX_MAP_GROWSDOWN) {
 528                 flags |= MAP_STACK;
 529                 /* The linux MAP_GROWSDOWN option does not limit auto
 530                  * growth of the region.  Linux mmap with this option
 531                  * takes as addr the inital BOS, and as len, the initial
 532                  * region size.  It can then grow down from addr without
 533                  * limit.  However, linux threads has an implicit internal
 534                  * limit to stack size of STACK_SIZE.  Its just not
 535                  * enforced explicitly in linux.  But, here we impose
 536                  * a limit of (STACK_SIZE - GUARD_SIZE) on the stack
 537                  * region, since we can do this with our mmap.
 538                  *
 539                  * Our mmap with MAP_STACK takes addr as the maximum
 540                  * downsize limit on BOS, and as len the max size of
 541                  * the region.  It them maps the top SGROWSIZ bytes,
 542                  * and autgrows the region down, up to the limit
 543                  * in addr.
 544                  *
 545                  * If we don't use the MAP_STACK option, the effect
 546                  * of this code is to allocate a stack region of a
 547                  * fixed size of (STACK_SIZE - GUARD_SIZE).
 548                  */
 549
 550                 /* This gives us TOS */
 551                 addr = linux_addr + linux_len;
 552
 553                 if (addr > p->p_vmspace->vm_maxsaddr) {
 554                         /* Some linux apps will attempt to mmap
 555                          * thread stacks near the top of their
 556                          * address space.  If their TOS is greater
 557                          * than vm_maxsaddr, vm_map_growstack()
 558                          * will confuse the thread stack with the
 559                          * process stack and deliver a SEGV if they
 560                          * attempt to grow the thread stack past their
 561                          * current stacksize rlimit.  To avoid this,
 562                          * adjust vm_maxsaddr upwards to reflect
 563                          * the current stacksize rlimit rather
 564                          * than the maximum possible stacksize.
 565                          * It would be better to adjust the
 566                          * mmap'ed region, but some apps do not check
 567                          * mmap's return value.
 568                          */
 569                         p->p_vmspace->vm_maxsaddr = (char *)USRSTACK -
 570                             p->p_rlimit[RLIMIT_STACK].rlim_cur;
 571                 }
 572
 573                 /* This gives us our maximum stack size */
 574                 if (linux_len > STACK_SIZE - GUARD_SIZE) {
 575                         len = linux_len;
 576                 } else {
 577                         len = STACK_SIZE - GUARD_SIZE;
 578                 }
 579                 /* This gives us a new BOS.  If we're using VM_STACK, then
 580                  * mmap will just map the top SGROWSIZ bytes, and let
 581                  * the stack grow down to the limit at BOS.  If we're
 582                  * not using VM_STACK we map the full stack, since we
 583                  * don't have a way to autogrow it.
 584                  */
 585                 addr -= len;
 586         } else {
 587                 addr = linux_addr;
 588                 len = linux_len;
 589         }
 590
 591         prot = linux_prot | PROT_READ;
 592         if (linux_flags & LINUX_MAP_ANON) {
 593                 fd = -1;
 594         } else {
 595                 fd = linux_fd;
 596         }
 597
 598 #ifdef DEBUG
 599         if (ldebug(mmap) || ldebug(mmap2))
 600                 kprintf("-> (%p, %d, %d, 0x%08x, %d, %lld)\n",
 601                     addr, len, prot, flags, fd, pos);
 602 #endif
 603         error = kern_mmap(curproc->p_vmspace, addr, len,
 604                           prot, flags, fd, pos, &new);
 605         rel_mplock();
 606
 607         if (error == 0)
 608                 *res = new;
 609         return (error);
 610 }
 611
 612 /*
 613  * MPSAFE
 614  */
 615 int
 616 sys_linux_mmap(struct linux_mmap_args *args)
 617 {
 618         struct l_mmap_argv linux_args;
 619         int error;
 620
 621         error = copyin((caddr_t)args->ptr, &linux_args, sizeof(linux_args));
 622         if (error)
 623                 return (error);
 624
 625 #ifdef DEBUG
 626         if (ldebug(mmap))
 627                 kprintf(ARGS(mmap, "%p, %d, %d, 0x%08x, %d, %d"),
 628                     (void *)linux_args.addr, linux_args.len, linux_args.prot,
 629                     linux_args.flags, linux_args.fd, linux_args.pos);
 630 #endif
 631         error = linux_mmap_common(linux_args.addr, linux_args.len,
 632             linux_args.prot, linux_args.flags, linux_args.fd,
 633             linux_args.pos, &args->sysmsg_resultp);
 634 #ifdef DEBUG
 635         if (ldebug(mmap))
 636                 kprintf("-> %p\n", args->sysmsg_resultp);
 637 #endif
 638         return(error);
 639 }
 640
 641 /*
 642  * MPSAFE
 643  */
 644 int
 645 sys_linux_mmap2(struct linux_mmap2_args *args)
 646 {
 647         int error;
 648
 649 #ifdef DEBUG
 650         if (ldebug(mmap2))
 651                 kprintf(ARGS(mmap2, "%p, %d, %d, 0x%08x, %d, %d"),
 652                     (void *)args->addr, args->len, args->prot, args->flags,
 653                     args->fd, args->pgoff);
 654 #endif
 655         error = linux_mmap_common((void *)args->addr, args->len, args->prot,
 656             args->flags, args->fd, args->pgoff * PAGE_SIZE,
 657             &args->sysmsg_resultp);
 658 #ifdef DEBUG
 659         if (ldebug(mmap2))
 660                 kprintf("-> %p\n", args->sysmsg_resultp);
 661 #endif
 662         return (error);
 663 }
 664
 665 /*
 666  * MPSAFE
 667  */
 668 int
 669 sys_linux_pipe(struct linux_pipe_args *args)
 670 {
 671         int error;
 672         int reg_edx;
 673         struct pipe_args bsd_args;
 674
 675 #ifdef DEBUG
 676         if (ldebug(pipe))
 677                 kprintf(ARGS(pipe, "*"));
 678 #endif
 679
 680         reg_edx = args->sysmsg_fds[1];
 681         error = sys_pipe(&bsd_args);
 682         if (error) {
 683                 args->sysmsg_fds[1] = reg_edx;
 684                 return (error);
 685         }
 686
 687         error = copyout(bsd_args.sysmsg_fds, args->pipefds, 2*sizeof(int));
 688         if (error) {
 689                 args->sysmsg_fds[1] = reg_edx;
 690                 return (error);
 691         }
 692
 693         args->sysmsg_fds[1] = reg_edx;
 694         args->sysmsg_fds[0] = 0;
 695         return (0);
 696 }
 697
 698 /*
 699  * MPSAFE
 700  */
 701 int
 702 sys_linux_ioperm(struct linux_ioperm_args *args)
 703 {
 704         struct sysarch_args sa;
 705         struct i386_ioperm_args *iia;
 706         caddr_t sg;
 707         int error;
 708
 709         sg = stackgap_init();
 710         iia = stackgap_alloc(&sg, sizeof(struct i386_ioperm_args));
 711         iia->start = args->start;
 712         iia->length = args->length;
 713         iia->enable = args->enable;
 714         sa.sysmsg_resultp = NULL;
 715         sa.op = I386_SET_IOPERM;
 716         sa.parms = (char *)iia;
 717         error = sys_sysarch(&sa);
 718         args->sysmsg_resultp = sa.sysmsg_resultp;
 719         return(error);
 720 }
 721
 722 /*
 723  * MPSAFE
 724  */
 725 int
 726 sys_linux_iopl(struct linux_iopl_args *args)
 727 {
 728         struct thread *td = curthread;
 729         struct lwp *lp = td->td_lwp;
 730         int error;
 731
 732         if (args->level < 0 || args->level > 3)
 733                 return (EINVAL);
 734         if ((error = priv_check(td, PRIV_ROOT)) != 0)
 735                 return (error);
 736         if (securelevel > 0)
 737                 return (EPERM);
 738         lp->lwp_md.md_regs->tf_eflags =
 739             (lp->lwp_md.md_regs->tf_eflags & ~PSL_IOPL) |
 740             (args->level * (PSL_IOPL / 3));
 741         return (0);
 742 }
 743
 744 /*
 745  * MPSAFE
 746  */
 747 int
 748 sys_linux_modify_ldt(struct linux_modify_ldt_args *uap)
 749 {
 750         int error;
 751         caddr_t sg;
 752         struct sysarch_args args;
 753         struct i386_ldt_args *ldt;
 754         struct l_descriptor ld;
 755         union descriptor *desc;
 756
 757         sg = stackgap_init();
 758
 759         if (uap->ptr == NULL)
 760                 return (EINVAL);
 761
 762         switch (uap->func) {
 763         case 0x00: /* read_ldt */
 764                 ldt = stackgap_alloc(&sg, sizeof(*ldt));
 765                 ldt->start = 0;
 766                 ldt->descs = uap->ptr;
 767                 ldt->num = uap->bytecount / sizeof(union descriptor);
 768                 args.op = I386_GET_LDT;
 769                 args.parms = (char*)ldt;
 770                 args.sysmsg_iresult = 0;
 771                 error = sys_sysarch(&args);
 772                 uap->sysmsg_iresult = args.sysmsg_iresult *
 773                                       sizeof(union descriptor);
 774                 break;
 775         case 0x01: /* write_ldt */
 776         case 0x11: /* write_ldt */
 777                 if (uap->bytecount != sizeof(ld))
 778                         return (EINVAL);
 779
 780                 error = copyin(uap->ptr, &ld, sizeof(ld));
 781                 if (error)
 782                         return (error);
 783
 784                 ldt = stackgap_alloc(&sg, sizeof(*ldt));
 785                 desc = stackgap_alloc(&sg, sizeof(*desc));
 786                 ldt->start = ld.entry_number;
 787                 ldt->descs = desc;
 788                 ldt->num = 1;
 789                 desc->sd.sd_lolimit = (ld.limit & 0x0000ffff);
 790                 desc->sd.sd_hilimit = (ld.limit & 0x000f0000) >> 16;
 791                 desc->sd.sd_lobase = (ld.base_addr & 0x00ffffff);
 792                 desc->sd.sd_hibase = (ld.base_addr & 0xff000000) >> 24;
 793                 desc->sd.sd_type = SDT_MEMRO | ((ld.read_exec_only ^ 1) << 1) |
 794                         (ld.contents << 2);
 795                 desc->sd.sd_dpl = 3;
 796                 desc->sd.sd_p = (ld.seg_not_present ^ 1);
 797                 desc->sd.sd_xx = 0;
 798                 desc->sd.sd_def32 = ld.seg_32bit;
 799                 desc->sd.sd_gran = ld.limit_in_pages;
 800                 args.op = I386_SET_LDT;
 801                 args.parms = (char*)ldt;
 802                 args.sysmsg_iresult = 0;
 803                 error = sys_sysarch(&args);
 804                 uap->sysmsg_iresult = args.sysmsg_iresult;
 805                 break;
 806         default:
 807                 error = EINVAL;
 808                 break;
 809         }
 810
 811         return (error);
 812 }
 813
 814 /*
 815  * MPALMOSTSAFE
 816  */
 817 int
 818 sys_linux_sigaction(struct linux_sigaction_args *args)
 819 {
 820         l_osigaction_t osa;
 821         l_sigaction_t linux_act, linux_oact;
 822         struct sigaction act, oact;
 823         int error;
 824
 825 #ifdef DEBUG
 826         if (ldebug(sigaction))
 827                 kprintf(ARGS(sigaction, "%d, %p, %p"),
 828                     args->sig, (void *)args->nsa, (void *)args->osa);
 829 #endif
 830
 831         if (args->nsa) {
 832                 error = copyin(args->nsa, &osa, sizeof(l_osigaction_t));
 833                 if (error)
 834                         return (error);
 835                 linux_act.lsa_handler = osa.lsa_handler;
 836                 linux_act.lsa_flags = osa.lsa_flags;
 837                 linux_act.lsa_restorer = osa.lsa_restorer;
 838                 LINUX_SIGEMPTYSET(linux_act.lsa_mask);
 839                 linux_act.lsa_mask.__bits[0] = osa.lsa_mask;
 840                 linux_to_bsd_sigaction(&linux_act, &act);
 841         }
 842
 843         get_mplock();
 844         error = kern_sigaction(args->sig, args->nsa ? &act : NULL,
 845                                args->osa ? &oact : NULL);
 846         rel_mplock();
 847
 848         if (args->osa != NULL && !error) {
 849                 bsd_to_linux_sigaction(&oact, &linux_oact);
 850                 osa.lsa_handler = linux_oact.lsa_handler;
 851                 osa.lsa_flags = linux_oact.lsa_flags;
 852                 osa.lsa_restorer = linux_oact.lsa_restorer;
 853                 osa.lsa_mask = linux_oact.lsa_mask.__bits[0];
 854                 error = copyout(&osa, args->osa, sizeof(l_osigaction_t));
 855         }
 856         return (error);
 857 }
 858
 859 /*
 860  * Linux has two extra args, restart and oldmask.  We dont use these,
 861  * but it seems that "restart" is actually a context pointer that
 862  * enables the signal to happen with a different register set.
 863  *
 864  * MPALMOSTSAFE
 865  */
 866 int
 867 sys_linux_sigsuspend(struct linux_sigsuspend_args *args)
 868 {
 869         l_sigset_t linux_mask;
 870         sigset_t mask;
 871         int error;
 872
 873 #ifdef DEBUG
 874         if (ldebug(sigsuspend))
 875                 kprintf(ARGS(sigsuspend, "%08lx"), (unsigned long)args->mask);
 876 #endif
 877
 878         LINUX_SIGEMPTYSET(mask);
 879         mask.__bits[0] = args->mask;
 880         linux_to_bsd_sigset(&linux_mask, &mask);
 881
 882         get_mplock();
 883         error = kern_sigsuspend(&mask);
 884         rel_mplock();
 885
 886         return(error);
 887 }
 888
 889 /*
 890  * MPALMOSTSAFE
 891  */
 892 int
 893 sys_linux_rt_sigsuspend(struct linux_rt_sigsuspend_args *uap)
 894 {
 895         l_sigset_t linux_mask;
 896         sigset_t mask;
 897         int error;
 898
 899 #ifdef DEBUG
 900         if (ldebug(rt_sigsuspend))
 901                 kprintf(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, &linux_mask, sizeof(l_sigset_t));
 909         if (error)
 910                 return (error);
 911
 912         linux_to_bsd_sigset(&linux_mask, &mask);
 913
 914         get_mplock();
 915         error = kern_sigsuspend(&mask);
 916         rel_mplock();
 917
 918         return(error);
 919 }
 920
 921 /*
 922  * MPALMOSTSAFE
 923  */
 924 int
 925 sys_linux_pause(struct linux_pause_args *args)
 926 {
 927         struct thread *td = curthread;
 928         struct lwp *lp = td->td_lwp;
 929         sigset_t mask;
 930         int error;
 931
 932 #ifdef DEBUG
 933         if (ldebug(pause))
 934                 kprintf(ARGS(pause, ""));
 935 #endif
 936
 937         mask = lp->lwp_sigmask;
 938
 939         get_mplock();
 940         error = kern_sigsuspend(&mask);
 941         rel_mplock();
 942
 943         return(error);
 944 }
 945
 946 /*
 947  * MPALMOSTSAFE
 948  */
 949 int
 950 sys_linux_sigaltstack(struct linux_sigaltstack_args *uap)
 951 {
 952         stack_t ss, oss;
 953         l_stack_t linux_ss;
 954         int error;
 955
 956 #ifdef DEBUG
 957         if (ldebug(sigaltstack))
 958                 kprintf(ARGS(sigaltstack, "%p, %p"), uap->uss, uap->uoss);
 959 #endif
 960
 961         if (uap->uss) {
 962                 error = copyin(uap->uss, &linux_ss, sizeof(l_stack_t));
 963                 if (error)
 964                         return (error);
 965
 966                 ss.ss_sp = linux_ss.ss_sp;
 967                 ss.ss_size = linux_ss.ss_size;
 968                 ss.ss_flags = linux_to_bsd_sigaltstack(linux_ss.ss_flags);
 969         }
 970
 971         get_mplock();
 972         error = kern_sigaltstack(uap->uss ? &ss : NULL,
 973                                  uap->uoss ? &oss : NULL);
 974         rel_mplock();
 975
 976         if (error == 0 && uap->uoss) {
 977                 linux_ss.ss_sp = oss.ss_sp;
 978                 linux_ss.ss_size = oss.ss_size;
 979                 linux_ss.ss_flags = bsd_to_linux_sigaltstack(oss.ss_flags);
 980                 error = copyout(&linux_ss, uap->uoss, sizeof(l_stack_t));
 981         }
 982
 983         return (error);
 984 }