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/ptrace.h>
  44 #include <sys/sysproto.h>
  45 #include <sys/thread2.h>
  46 #include <sys/unistd.h>
  47 #include <sys/wait.h>
  48
  49 #include <machine/frame.h>
  50 #include <machine/psl.h>
  51 #include <machine/segments.h>
  52 #include <machine/sysarch.h>
  53
  54 #include <vm/vm.h>
  55 #include <vm/pmap.h>
  56 #include <vm/vm_map.h>
  57
  58 #include <sys/mplock2.h>
  59
  60 #include "linux.h"
  61 #include "linux_proto.h"
  62 #include "../linux_ipc.h"
  63 #include "../linux_signal.h"
  64 #include "../linux_util.h"
  65 #include "../linux_emuldata.h"
  66
  67 struct l_descriptor {
  68         l_uint          entry_number;
  69         l_ulong         base_addr;
  70         l_uint          limit;
  71         l_uint          seg_32bit:1;
  72         l_uint          contents:2;
  73         l_uint          read_exec_only:1;
  74         l_uint          limit_in_pages:1;
  75         l_uint          seg_not_present:1;
  76         l_uint          useable:1;
  77 };
  78
  79 struct l_old_select_argv {
  80         l_int           nfds;
  81         l_fd_set        *readfds;
  82         l_fd_set        *writefds;
  83         l_fd_set        *exceptfds;
  84         struct l_timeval        *timeout;
  85 };
  86
  87 int
  88 linux_to_bsd_sigaltstack(int lsa)
  89 {
  90         int bsa = 0;
  91
  92         if (lsa & LINUX_SS_DISABLE)
  93                 bsa |= SS_DISABLE;
  94         if (lsa & LINUX_SS_ONSTACK)
  95                 bsa |= SS_ONSTACK;
  96         return (bsa);
  97 }
  98
  99 int
 100 bsd_to_linux_sigaltstack(int bsa)
 101 {
 102         int lsa = 0;
 103
 104         if (bsa & SS_DISABLE)
 105                 lsa |= LINUX_SS_DISABLE;
 106         if (bsa & SS_ONSTACK)
 107                 lsa |= LINUX_SS_ONSTACK;
 108         return (lsa);
 109 }
 110
 111 /*
 112  * MPALMOSTSAFE
 113  */
 114 int
 115 sys_linux_execve(struct linux_execve_args *args)
 116 {
 117         struct nlookupdata nd;
 118         struct image_args exec_args;
 119         char *path;
 120         int error;
 121
 122         error = linux_copyin_path(args->path, &path, LINUX_PATH_EXISTS);
 123         if (error)
 124                 return (error);
 125 #ifdef DEBUG
 126         if (ldebug(execve))
 127                 kprintf(ARGS(execve, "%s"), path);
 128 #endif
 129         get_mplock();
 130         error = nlookup_init(&nd, path, UIO_SYSSPACE, NLC_FOLLOW);
 131         bzero(&exec_args, sizeof(exec_args));
 132         if (error == 0) {
 133                 error = exec_copyin_args(&exec_args, path, PATH_SYSSPACE,
 134                                         args->argp, args->envp);
 135         }
 136         if (error == 0)
 137                 error = kern_execve(&nd, &exec_args);
 138         nlookup_done(&nd);
 139
 140         /*
 141          * The syscall result is returned in registers to the new program.
 142          * Linux will register %edx as an atexit function and we must be
 143          * sure to set it to 0.  XXX
 144          */
 145         if (error == 0) {
 146                 args->sysmsg_result64 = 0;
 147                 if (curproc->p_sysent == &elf_linux_sysvec)
 148                         error = emuldata_init(curproc, NULL, 0);
 149         }
 150
 151         exec_free_args(&exec_args);
 152         linux_free_path(&path);
 153
 154         if (error < 0) {
 155                 /* We hit a lethal error condition.  Let's die now. */
 156                 exit1(W_EXITCODE(0, SIGABRT));
 157                 /* NOTREACHED */
 158         }
 159         rel_mplock();
 160
 161         return(error);
 162 }
 163
 164 struct l_ipc_kludge {
 165         struct l_msgbuf *msgp;
 166         l_long msgtyp;
 167 };
 168
 169 /*
 170  * MPALMOSTSAFE
 171  */
 172 int
 173 sys_linux_ipc(struct linux_ipc_args *args)
 174 {
 175         int error = 0;
 176
 177         get_mplock();
 178
 179         switch (args->what & 0xFFFF) {
 180         case LINUX_SEMOP: {
 181                 struct linux_semop_args a;
 182
 183                 a.semid = args->arg1;
 184                 a.tsops = args->ptr;
 185                 a.nsops = args->arg2;
 186                 a.sysmsg_lresult = 0;
 187                 error = linux_semop(&a);
 188                 args->sysmsg_lresult = a.sysmsg_lresult;
 189                 break;
 190         }
 191         case LINUX_SEMGET: {
 192                 struct linux_semget_args a;
 193
 194                 a.key = args->arg1;
 195                 a.nsems = args->arg2;
 196                 a.semflg = args->arg3;
 197                 a.sysmsg_lresult = 0;
 198                 error = linux_semget(&a);
 199                 args->sysmsg_lresult = a.sysmsg_lresult;
 200                 break;
 201         }
 202         case LINUX_SEMCTL: {
 203                 struct linux_semctl_args a;
 204                 int error;
 205
 206                 a.semid = args->arg1;
 207                 a.semnum = args->arg2;
 208                 a.cmd = args->arg3;
 209                 a.sysmsg_lresult = 0;
 210                 error = copyin((caddr_t)args->ptr, &a.arg, sizeof(a.arg));
 211                 if (error)
 212                         break;
 213                 error = linux_semctl(&a);
 214                 args->sysmsg_lresult = a.sysmsg_lresult;
 215                 break;
 216         }
 217         case LINUX_MSGSND: {
 218                 struct linux_msgsnd_args a;
 219
 220                 a.msqid = args->arg1;
 221                 a.msgp = args->ptr;
 222                 a.msgsz = args->arg2;
 223                 a.msgflg = args->arg3;
 224                 a.sysmsg_lresult = 0;
 225                 error = linux_msgsnd(&a);
 226                 args->sysmsg_lresult = a.sysmsg_lresult;
 227                 break;
 228         }
 229         case LINUX_MSGRCV: {
 230                 struct linux_msgrcv_args a;
 231
 232                 a.msqid = args->arg1;
 233                 a.msgsz = args->arg2;
 234                 if (a.msgsz < 0) {
 235                         error = EINVAL;
 236                         break;
 237                 }
 238                 a.msgflg = args->arg3;
 239                 a.sysmsg_lresult = 0;
 240                 if ((args->what >> 16) == 0) {
 241                         struct l_ipc_kludge tmp;
 242                         int error;
 243
 244                         if (args->ptr == NULL) {
 245                                 error = EINVAL;
 246                                 break;
 247                         }
 248                         error = copyin((caddr_t)args->ptr, &tmp, sizeof(tmp));
 249                         if (error)
 250                                 break;
 251                         a.msgp = tmp.msgp;
 252                         a.msgtyp = tmp.msgtyp;
 253                 } else {
 254                         a.msgp = args->ptr;
 255                         a.msgtyp = args->arg5;
 256                 }
 257                 error = linux_msgrcv(&a);
 258                 args->sysmsg_lresult = a.sysmsg_lresult;
 259                 break;
 260         }
 261         case LINUX_MSGGET: {
 262                 struct linux_msgget_args a;
 263
 264                 a.key = args->arg1;
 265                 a.msgflg = args->arg2;
 266                 a.sysmsg_lresult = 0;
 267                 error = linux_msgget(&a);
 268                 args->sysmsg_lresult = a.sysmsg_lresult;
 269                 break;
 270         }
 271         case LINUX_MSGCTL: {
 272                 struct linux_msgctl_args a;
 273
 274                 a.msqid = args->arg1;
 275                 a.cmd = args->arg2;
 276                 a.buf = args->ptr;
 277                 a.sysmsg_lresult = 0;
 278                 error = linux_msgctl(&a);
 279                 args->sysmsg_lresult = a.sysmsg_lresult;
 280                 break;
 281         }
 282         case LINUX_SHMAT: {
 283                 struct linux_shmat_args a;
 284
 285                 a.shmid = args->arg1;
 286                 a.shmaddr = args->ptr;
 287                 a.shmflg = args->arg2;
 288                 a.raddr = (l_ulong *)args->arg3;
 289                 a.sysmsg_lresult = 0;
 290                 error = linux_shmat(&a);
 291                 args->sysmsg_lresult = a.sysmsg_lresult;
 292                 break;
 293         }
 294         case LINUX_SHMDT: {
 295                 struct linux_shmdt_args a;
 296
 297                 a.shmaddr = args->ptr;
 298                 a.sysmsg_lresult = 0;
 299                 error = linux_shmdt(&a);
 300                 args->sysmsg_lresult = a.sysmsg_lresult;
 301                 break;
 302         }
 303         case LINUX_SHMGET: {
 304                 struct linux_shmget_args a;
 305
 306                 a.key = args->arg1;
 307                 a.size = args->arg2;
 308                 a.shmflg = args->arg3;
 309                 a.sysmsg_lresult = 0;
 310                 error = linux_shmget(&a);
 311                 args->sysmsg_lresult = a.sysmsg_lresult;
 312                 break;
 313         }
 314         case LINUX_SHMCTL: {
 315                 struct linux_shmctl_args a;
 316
 317                 a.shmid = args->arg1;
 318                 a.cmd = args->arg2;
 319                 a.buf = args->ptr;
 320                 a.sysmsg_lresult = 0;
 321                 error = linux_shmctl(&a);
 322                 args->sysmsg_lresult = a.sysmsg_lresult;
 323                 break;
 324         }
 325         default:
 326                 error = EINVAL;
 327                 break;
 328         }
 329         rel_mplock();
 330         return(error);
 331 }
 332
 333 /*
 334  * MPSAFE
 335  */
 336 int
 337 sys_linux_old_select(struct linux_old_select_args *args)
 338 {
 339         struct l_old_select_argv linux_args;
 340         struct linux_select_args newsel;
 341         int error;
 342
 343 #ifdef DEBUG
 344         if (ldebug(old_select))
 345                 kprintf(ARGS(old_select, "%p"), args->ptr);
 346 #endif
 347
 348         error = copyin((caddr_t)args->ptr, &linux_args, sizeof(linux_args));
 349         if (error)
 350                 return (error);
 351
 352         newsel.sysmsg_iresult = 0;
 353         newsel.nfds = linux_args.nfds;
 354         newsel.readfds = linux_args.readfds;
 355         newsel.writefds = linux_args.writefds;
 356         newsel.exceptfds = linux_args.exceptfds;
 357         newsel.timeout = linux_args.timeout;
 358         error = sys_linux_select(&newsel);
 359         args->sysmsg_iresult = newsel.sysmsg_iresult;
 360         return(error);
 361 }
 362
 363 /*
 364  * MPSAFE
 365  */
 366 int
 367 sys_linux_fork(struct linux_fork_args *args)
 368 {
 369         struct lwp *lp = curthread->td_lwp;
 370         struct proc *p2;
 371         int error;
 372
 373         get_mplock();
 374         error = fork1(lp, RFFDG | RFPROC | RFPGLOCK, &p2);
 375         if (error == 0) {
 376                 emuldata_init(curproc, p2, 0);
 377
 378                 start_forked_proc(lp, p2);
 379                 args->sysmsg_fds[0] = p2->p_pid;
 380                 args->sysmsg_fds[1] = 0;
 381         }
 382         rel_mplock();
 383
 384         /* Are we the child? */
 385         if (args->sysmsg_iresult == 1)
 386                 args->sysmsg_iresult = 0;
 387
 388         return (error);
 389 }
 390
 391 /*
 392  * MPALMOSTSAFE
 393  */
 394 int
 395 sys_linux_exit_group(struct linux_exit_group_args *args)
 396 {
 397         struct linux_emuldata *em, *e;
 398         int rval;
 399
 400         rval = args->rval;
 401         EMUL_LOCK();
 402
 403         em = emuldata_get(curproc);
 404
 405         if (em->s->refs == 1) {
 406                 EMUL_UNLOCK();
 407                 exit1(W_EXITCODE(rval, 0));
 408                 /* NOTREACHED */
 409                 return (0);
 410         }
 411         KKASSERT(em->proc == curproc);
 412         em->flags |= EMUL_DIDKILL;
 413         em->s->flags |= LINUX_LES_INEXITGROUP;
 414         em->s->xstat = W_EXITCODE(rval, 0);
 415
 416         LIST_REMOVE(em, threads);
 417         LIST_INSERT_HEAD(&em->s->threads, em, threads);
 418
 419         while ((e = LIST_NEXT(em, threads)) != NULL) {
 420                 LIST_REMOVE(em, threads);
 421                 LIST_INSERT_AFTER(e, em, threads);
 422                 if ((e->flags & EMUL_DIDKILL) == 0) {
 423                         e->flags |= EMUL_DIDKILL;
 424                         KKASSERT(pfind(e->proc->p_pid) == e->proc);
 425                         get_mplock();
 426                         ksignal(e->proc, SIGKILL);
 427                         rel_mplock();
 428                 }
 429         }
 430
 431         EMUL_UNLOCK();
 432         exit1(W_EXITCODE(rval, 0));
 433         /* NOTREACHED */
 434
 435         return (0);
 436 }
 437
 438 /*
 439  * MPSAFE
 440  */
 441 int
 442 sys_linux_vfork(struct linux_vfork_args *args)
 443 {
 444         struct lwp *lp = curthread->td_lwp;
 445         struct proc *p2;
 446         int error;
 447
 448         get_mplock();
 449         error = fork1(lp, RFFDG | RFPROC | RFPPWAIT | RFMEM | RFPGLOCK, &p2);
 450         if (error == 0) {
 451                 emuldata_init(curproc, p2, 0);
 452
 453                 start_forked_proc(lp, p2);
 454                 args->sysmsg_fds[0] = p2->p_pid;
 455                 args->sysmsg_fds[1] = 0;
 456         }
 457         rel_mplock();
 458
 459         if (args->sysmsg_iresult == 1)
 460                 args->sysmsg_iresult = 0;
 461
 462         return (error);
 463 }
 464
 465 /*
 466  * MPALMOSTSAFE
 467  */
 468 int
 469 sys_linux_clone(struct linux_clone_args *args)
 470 {
 471         struct segment_descriptor *desc;
 472         struct l_user_desc info;
 473         int idx;
 474         int a[2];
 475
 476         struct lwp *lp = curthread->td_lwp;
 477         int error, ff = RFPROC;
 478         struct proc *p2 = NULL;
 479         int exit_signal;
 480         vm_offset_t start;
 481
 482         exit_signal = args->flags & 0x000000ff;
 483         if (exit_signal >= LINUX_NSIG)
 484                 return (EINVAL);
 485         if (exit_signal <= LINUX_SIGTBLSZ)
 486                 exit_signal = linux_to_bsd_signal[_SIG_IDX(exit_signal)];
 487
 488         if (args->flags & LINUX_CLONE_VM)
 489                 ff |= RFMEM;
 490         if (args->flags & LINUX_CLONE_SIGHAND)
 491                 ff |= RFSIGSHARE;
 492         if (!(args->flags & (LINUX_CLONE_FILES | LINUX_CLONE_FS)))
 493                 ff |= RFFDG;
 494         if ((args->flags & 0xffffff00) == LINUX_THREADING_FLAGS)
 495                 ff |= RFTHREAD;
 496         if (args->flags & LINUX_CLONE_VFORK)
 497                 ff |= RFPPWAIT;
 498         if (args->flags & LINUX_CLONE_PARENT_SETTID) {
 499                 if (args->parent_tidptr == NULL)
 500                         return (EINVAL);
 501         }
 502
 503         error = 0;
 504         start = 0;
 505
 506         get_mplock();
 507         error = fork1(lp, ff | RFPGLOCK, &p2);
 508         if (error) {
 509                 rel_mplock();
 510                 return error;
 511         }
 512
 513         args->sysmsg_fds[0] = p2 ? p2->p_pid : 0;
 514         args->sysmsg_fds[1] = 0;
 515
 516         if (args->flags & (LINUX_CLONE_PARENT | LINUX_CLONE_THREAD))
 517                 proc_reparent(p2, curproc->p_pptr /* XXX */);
 518
 519         emuldata_init(curproc, p2, args->flags);
 520         linux_proc_fork(p2, curproc, args->child_tidptr);
 521         /*
 522          * XXX: this can't happen, p2 is never NULL, or else we'd have
 523          *      other problems, too (see p2->p_sigparent == ...,
 524          *      linux_proc_fork and emuldata_init.
 525          */
 526         if (p2 == NULL) {
 527                 error = ESRCH;
 528         } else {
 529                 if (args->flags & LINUX_CLONE_PARENT_SETTID) {
 530                         error = copyout(&p2->p_pid, args->parent_tidptr, sizeof(p2->p_pid));
 531                 }
 532         }
 533
 534         p2->p_sigparent = exit_signal;
 535         if (args->stack) {
 536                 ONLY_LWP_IN_PROC(p2)->lwp_md.md_regs->tf_esp =
 537                                         (unsigned long)args->stack;
 538         }
 539
 540         if (args->flags & LINUX_CLONE_SETTLS) {
 541                 error = copyin((void *)curthread->td_lwp->lwp_md.md_regs->tf_esi, &info, sizeof(struct l_user_desc));
 542                 if (error) {
 543                         kprintf("copyin of tf_esi to info failed\n");
 544                 } else {
 545                         idx = info.entry_number;
 546                         /*
 547                          * We understand both our own entries such as the ones
 548                          * we provide on linux_set_thread_area, as well as the
 549                          * linux-type entries 6-8.
 550                          */
 551                         if ((idx < 6 || idx > 8) && (idx < GTLS_START)) {
 552                                 kprintf("LINUX_CLONE_SETTLS, invalid idx requested: %d\n", idx);
 553                                 goto out;
 554                         }
 555                         if (idx < GTLS_START) {
 556                                 idx -= 6;
 557                         } else {
 558 #if 0 /* was SMP */
 559                                 idx -= (GTLS_START + mycpu->gd_cpuid * NGDT);
 560 #endif
 561                                 idx -= GTLS_START;
 562                         }
 563                         KKASSERT(idx >= 0);
 564
 565                         a[0] = LINUX_LDT_entry_a(&info);
 566                         a[1] = LINUX_LDT_entry_b(&info);
 567                         if (p2) {
 568                                 desc = &FIRST_LWP_IN_PROC(p2)->lwp_thread->td_tls.tls[idx];
 569                                 memcpy(desc, &a, sizeof(a));
 570                         } else {
 571                                 kprintf("linux_clone... we don't have a p2\n");
 572                         }
 573                 }
 574         }
 575 out:
 576         if (p2)
 577                 start_forked_proc(lp, p2);
 578
 579         rel_mplock();
 580 #ifdef DEBUG
 581         if (ldebug(clone))
 582                 kprintf(LMSG("clone: successful rfork to %ld"),
 583                     (long)p2->p_pid);
 584 #endif
 585
 586         return (error);
 587 }
 588
 589 /* XXX move */
 590 struct l_mmap_argv {
 591         l_caddr_t       addr;
 592         l_int           len;
 593         l_int           prot;
 594         l_int           flags;
 595         l_int           fd;
 596         l_int           pos;
 597 };
 598
 599 #define STACK_SIZE  (2 * 1024 * 1024)
 600 #define GUARD_SIZE  (4 * PAGE_SIZE)
 601
 602 /*
 603  * MPALMOSTSAFE
 604  */
 605 static int
 606 linux_mmap_common(caddr_t linux_addr, size_t linux_len, int linux_prot,
 607                   int linux_flags, int linux_fd, off_t pos, void **res)
 608 {
 609         struct thread *td = curthread;
 610         struct proc *p = td->td_proc;
 611         caddr_t addr;
 612         void *new;
 613         int error, flags, len, prot, fd;
 614
 615         flags = 0;
 616         if (linux_flags & LINUX_MAP_SHARED)
 617                 flags |= MAP_SHARED;
 618         if (linux_flags & LINUX_MAP_PRIVATE)
 619                 flags |= MAP_PRIVATE;
 620         if (linux_flags & LINUX_MAP_FIXED)
 621                 flags |= MAP_FIXED;
 622         if (linux_flags & LINUX_MAP_ANON) {
 623                 flags |= MAP_ANON;
 624         } else {
 625                 flags |= MAP_NOSYNC;
 626         }
 627
 628         lwkt_gettoken(&vm_token);
 629         lwkt_gettoken(&vmspace_token);
 630
 631         if (linux_flags & LINUX_MAP_GROWSDOWN) {
 632                 flags |= MAP_STACK;
 633                 /* The linux MAP_GROWSDOWN option does not limit auto
 634                  * growth of the region.  Linux mmap with this option
 635                  * takes as addr the inital BOS, and as len, the initial
 636                  * region size.  It can then grow down from addr without
 637                  * limit.  However, linux threads has an implicit internal
 638                  * limit to stack size of STACK_SIZE.  Its just not
 639                  * enforced explicitly in linux.  But, here we impose
 640                  * a limit of (STACK_SIZE - GUARD_SIZE) on the stack
 641                  * region, since we can do this with our mmap.
 642                  *
 643                  * Our mmap with MAP_STACK takes addr as the maximum
 644                  * downsize limit on BOS, and as len the max size of
 645                  * the region.  It them maps the top SGROWSIZ bytes,
 646                  * and autgrows the region down, up to the limit
 647                  * in addr.
 648                  *
 649                  * If we don't use the MAP_STACK option, the effect
 650                  * of this code is to allocate a stack region of a
 651                  * fixed size of (STACK_SIZE - GUARD_SIZE).
 652                  */
 653
 654                 /* This gives us TOS */
 655                 addr = linux_addr + linux_len;
 656
 657                 if (addr > p->p_vmspace->vm_maxsaddr) {
 658                         /* Some linux apps will attempt to mmap
 659                          * thread stacks near the top of their
 660                          * address space.  If their TOS is greater
 661                          * than vm_maxsaddr, vm_map_growstack()
 662                          * will confuse the thread stack with the
 663                          * process stack and deliver a SEGV if they
 664                          * attempt to grow the thread stack past their
 665                          * current stacksize rlimit.  To avoid this,
 666                          * adjust vm_maxsaddr upwards to reflect
 667                          * the current stacksize rlimit rather
 668                          * than the maximum possible stacksize.
 669                          * It would be better to adjust the
 670                          * mmap'ed region, but some apps do not check
 671                          * mmap's return value.
 672                          */
 673                         p->p_vmspace->vm_maxsaddr = (char *)USRSTACK -
 674                             p->p_rlimit[RLIMIT_STACK].rlim_cur;
 675                 }
 676
 677                 /* This gives us our maximum stack size */
 678                 if (linux_len > STACK_SIZE - GUARD_SIZE) {
 679                         len = linux_len;
 680                 } else {
 681                         len = STACK_SIZE - GUARD_SIZE;
 682                 }
 683                 /* This gives us a new BOS.  If we're using VM_STACK, then
 684                  * mmap will just map the top SGROWSIZ bytes, and let
 685                  * the stack grow down to the limit at BOS.  If we're
 686                  * not using VM_STACK we map the full stack, since we
 687                  * don't have a way to autogrow it.
 688                  */
 689                 addr -= len;
 690         } else {
 691                 addr = linux_addr;
 692                 len = linux_len;
 693         }
 694
 695         prot = linux_prot;
 696
 697         if (prot & (PROT_READ | PROT_WRITE | PROT_EXEC))
 698                 prot |= PROT_READ | PROT_EXEC;
 699
 700         if (linux_flags & LINUX_MAP_ANON) {
 701                 fd = -1;
 702         } else {
 703                 fd = linux_fd;
 704         }
 705
 706 #ifdef DEBUG
 707         if (ldebug(mmap) || ldebug(mmap2))
 708                 kprintf("-> (%p, %d, %d, 0x%08x, %d, %lld)\n",
 709                     addr, len, prot, flags, fd, pos);
 710 #endif
 711         error = kern_mmap(curproc->p_vmspace, addr, len,
 712                           prot, flags, fd, pos, &new);
 713
 714         lwkt_reltoken(&vmspace_token);
 715         lwkt_reltoken(&vm_token);
 716
 717         if (error == 0)
 718                 *res = new;
 719         return (error);
 720 }
 721
 722 /*
 723  * MPSAFE
 724  */
 725 int
 726 sys_linux_mmap(struct linux_mmap_args *args)
 727 {
 728         struct l_mmap_argv linux_args;
 729         int error;
 730
 731         error = copyin((caddr_t)args->ptr, &linux_args, sizeof(linux_args));
 732         if (error)
 733                 return (error);
 734
 735 #ifdef DEBUG
 736         if (ldebug(mmap))
 737                 kprintf(ARGS(mmap, "%p, %d, %d, 0x%08x, %d, %d"),
 738                     (void *)linux_args.addr, linux_args.len, linux_args.prot,
 739                     linux_args.flags, linux_args.fd, linux_args.pos);
 740 #endif
 741         error = linux_mmap_common(linux_args.addr, linux_args.len,
 742             linux_args.prot, linux_args.flags, linux_args.fd,
 743             linux_args.pos, &args->sysmsg_resultp);
 744 #ifdef DEBUG
 745         if (ldebug(mmap))
 746                 kprintf("-> %p\n", args->sysmsg_resultp);
 747 #endif
 748         return(error);
 749 }
 750
 751 /*
 752  * MPSAFE
 753  */
 754 int
 755 sys_linux_mmap2(struct linux_mmap2_args *args)
 756 {
 757         int error;
 758
 759 #ifdef DEBUG
 760         if (ldebug(mmap2))
 761                 kprintf(ARGS(mmap2, "%p, %d, %d, 0x%08x, %d, %d"),
 762                     (void *)args->addr, args->len, args->prot, args->flags,
 763                     args->fd, args->pgoff);
 764 #endif
 765         error = linux_mmap_common((void *)args->addr, args->len, args->prot,
 766             args->flags, args->fd, args->pgoff * PAGE_SIZE,
 767             &args->sysmsg_resultp);
 768 #ifdef DEBUG
 769         if (ldebug(mmap2))
 770                 kprintf("-> %p\n", args->sysmsg_resultp);
 771 #endif
 772         return (error);
 773 }
 774
 775 /*
 776  * MPSAFE
 777  */
 778 int
 779 sys_linux_pipe(struct linux_pipe_args *args)
 780 {
 781         int error;
 782         int reg_edx;
 783         struct pipe_args bsd_args;
 784
 785 #ifdef DEBUG
 786         if (ldebug(pipe))
 787                 kprintf(ARGS(pipe, "*"));
 788 #endif
 789
 790         reg_edx = args->sysmsg_fds[1];
 791         error = sys_pipe(&bsd_args);
 792         if (error) {
 793                 args->sysmsg_fds[1] = reg_edx;
 794                 return (error);
 795         }
 796
 797         error = copyout(bsd_args.sysmsg_fds, args->pipefds, 2*sizeof(int));
 798         if (error) {
 799                 args->sysmsg_fds[1] = reg_edx;
 800                 return (error);
 801         }
 802
 803         args->sysmsg_fds[1] = reg_edx;
 804         args->sysmsg_fds[0] = 0;
 805         return (0);
 806 }
 807
 808 /*
 809  * XXX: Preliminary
 810  */
 811 int
 812 sys_linux_pipe2(struct linux_pipe2_args *args)
 813 {
 814         struct thread *td = curthread;
 815         int error;
 816         int reg_edx;
 817         struct pipe_args bsd_args;
 818         union fcntl_dat dat;
 819
 820         reg_edx = args->sysmsg_fds[1];
 821         error = sys_pipe(&bsd_args);
 822         if (error) {
 823                 args->sysmsg_fds[1] = reg_edx;
 824                 return (error);
 825         }
 826
 827 //      if (args->flags & LINUX_O_CLOEXEC) {
 828 //      }
 829
 830         if (args->flags & LINUX_O_NONBLOCK) {
 831                 dat.fc_flags = O_NONBLOCK;
 832                 kern_fcntl(bsd_args.sysmsg_fds[0], F_SETFL, &dat, td->td_ucred);
 833                 kern_fcntl(bsd_args.sysmsg_fds[1], F_SETFL, &dat, td->td_ucred);
 834         }
 835
 836         error = copyout(bsd_args.sysmsg_fds, args->pipefds, 2*sizeof(int));
 837         if (error) {
 838                 args->sysmsg_fds[1] = reg_edx;
 839                 return (error);
 840         }
 841
 842         args->sysmsg_fds[1] = reg_edx;
 843         args->sysmsg_fds[0] = 0;
 844         return (0);
 845 }
 846
 847 /*
 848  * MPSAFE
 849  */
 850 int
 851 sys_linux_ioperm(struct linux_ioperm_args *args)
 852 {
 853         struct sysarch_args sa;
 854         struct i386_ioperm_args *iia;
 855         caddr_t sg;
 856         int error;
 857
 858         sg = stackgap_init();
 859         iia = stackgap_alloc(&sg, sizeof(struct i386_ioperm_args));
 860         iia->start = args->start;
 861         iia->length = args->length;
 862         iia->enable = args->enable;
 863         sa.sysmsg_resultp = NULL;
 864         sa.op = I386_SET_IOPERM;
 865         sa.parms = (char *)iia;
 866         error = sys_sysarch(&sa);
 867         args->sysmsg_resultp = sa.sysmsg_resultp;
 868         return(error);
 869 }
 870
 871 /*
 872  * MPSAFE
 873  */
 874 int
 875 sys_linux_iopl(struct linux_iopl_args *args)
 876 {
 877         struct thread *td = curthread;
 878         struct lwp *lp = td->td_lwp;
 879         int error;
 880
 881         if (args->level < 0 || args->level > 3)
 882                 return (EINVAL);
 883         if ((error = priv_check(td, PRIV_ROOT)) != 0)
 884                 return (error);
 885         if (securelevel > 0)
 886                 return (EPERM);
 887         lp->lwp_md.md_regs->tf_eflags =
 888             (lp->lwp_md.md_regs->tf_eflags & ~PSL_IOPL) |
 889             (args->level * (PSL_IOPL / 3));
 890         return (0);
 891 }
 892
 893 /*
 894  * MPSAFE
 895  */
 896 int
 897 sys_linux_modify_ldt(struct linux_modify_ldt_args *uap)
 898 {
 899         int error;
 900         caddr_t sg;
 901         struct sysarch_args args;
 902         struct i386_ldt_args *ldt;
 903         struct l_descriptor ld;
 904         union descriptor *desc;
 905         int size, written;
 906
 907         sg = stackgap_init();
 908
 909         if (uap->ptr == NULL)
 910                 return (EINVAL);
 911
 912         switch (uap->func) {
 913         case 0x00: /* read_ldt */
 914                 ldt = stackgap_alloc(&sg, sizeof(*ldt));
 915                 ldt->start = 0;
 916                 ldt->descs = uap->ptr;
 917                 ldt->num = uap->bytecount / sizeof(union descriptor);
 918                 args.op = I386_GET_LDT;
 919                 args.parms = (char*)ldt;
 920                 args.sysmsg_iresult = 0;
 921                 error = sys_sysarch(&args);
 922                 uap->sysmsg_iresult = args.sysmsg_iresult *
 923                                       sizeof(union descriptor);
 924                 break;
 925         case 0x02: /* read_default_ldt = 0 */
 926                 size = 5*sizeof(struct l_desc_struct);
 927                 if (size > uap->bytecount)
 928                         size = uap->bytecount;
 929                 for (written = error = 0; written < size && error == 0; written++)
 930                         error = subyte((char *)uap->ptr + written, 0);
 931                 uap->sysmsg_iresult = written;
 932                 break;
 933         case 0x01: /* write_ldt */
 934         case 0x11: /* write_ldt */
 935                 if (uap->bytecount != sizeof(ld))
 936                         return (EINVAL);
 937
 938                 error = copyin(uap->ptr, &ld, sizeof(ld));
 939                 if (error)
 940                         return (error);
 941
 942                 ldt = stackgap_alloc(&sg, sizeof(*ldt));
 943                 desc = stackgap_alloc(&sg, sizeof(*desc));
 944                 ldt->start = ld.entry_number;
 945                 ldt->descs = desc;
 946                 ldt->num = 1;
 947                 desc->sd.sd_lolimit = (ld.limit & 0x0000ffff);
 948                 desc->sd.sd_hilimit = (ld.limit & 0x000f0000) >> 16;
 949                 desc->sd.sd_lobase = (ld.base_addr & 0x00ffffff);
 950                 desc->sd.sd_hibase = (ld.base_addr & 0xff000000) >> 24;
 951                 desc->sd.sd_type = SDT_MEMRO | ((ld.read_exec_only ^ 1) << 1) |
 952                         (ld.contents << 2);
 953                 desc->sd.sd_dpl = 3;
 954                 desc->sd.sd_p = (ld.seg_not_present ^ 1);
 955                 desc->sd.sd_xx = 0;
 956                 desc->sd.sd_def32 = ld.seg_32bit;
 957                 desc->sd.sd_gran = ld.limit_in_pages;
 958                 args.op = I386_SET_LDT;
 959                 args.parms = (char*)ldt;
 960                 args.sysmsg_iresult = 0;
 961                 error = sys_sysarch(&args);
 962                 uap->sysmsg_iresult = args.sysmsg_iresult;
 963                 break;
 964         default:
 965                 error = EINVAL;
 966                 break;
 967         }
 968
 969         return (error);
 970 }
 971
 972 /*
 973  * MPALMOSTSAFE
 974  */
 975 int
 976 sys_linux_sigaction(struct linux_sigaction_args *args)
 977 {
 978         l_osigaction_t osa;
 979         l_sigaction_t linux_act, linux_oact;
 980         struct sigaction act, oact;
 981         int error, sig;
 982
 983 #ifdef DEBUG
 984         if (ldebug(sigaction))
 985                 kprintf(ARGS(sigaction, "%d, %p, %p"),
 986                     args->sig, (void *)args->nsa, (void *)args->osa);
 987 #endif
 988
 989         if (args->nsa) {
 990                 error = copyin(args->nsa, &osa, sizeof(l_osigaction_t));
 991                 if (error)
 992                         return (error);
 993                 linux_act.lsa_handler = osa.lsa_handler;
 994                 linux_act.lsa_flags = osa.lsa_flags;
 995                 linux_act.lsa_restorer = osa.lsa_restorer;
 996                 LINUX_SIGEMPTYSET(linux_act.lsa_mask);
 997                 linux_act.lsa_mask.__bits[0] = osa.lsa_mask;
 998                 linux_to_bsd_sigaction(&linux_act, &act);
 999         }
1000
1001         if (args->sig <= LINUX_SIGTBLSZ)
1002                 sig = linux_to_bsd_signal[_SIG_IDX(args->sig)];
1003         else
1004                 sig = args->sig;
1005
1006         get_mplock();
1007         error = kern_sigaction(sig, args->nsa ? &act : NULL,
1008                                args->osa ? &oact : NULL);
1009         rel_mplock();
1010
1011         if (args->osa != NULL && !error) {
1012                 bsd_to_linux_sigaction(&oact, &linux_oact);
1013                 osa.lsa_handler = linux_oact.lsa_handler;
1014                 osa.lsa_flags = linux_oact.lsa_flags;
1015                 osa.lsa_restorer = linux_oact.lsa_restorer;
1016                 osa.lsa_mask = linux_oact.lsa_mask.__bits[0];
1017                 error = copyout(&osa, args->osa, sizeof(l_osigaction_t));
1018         }
1019         return (error);
1020 }
1021
1022 /*
1023  * Linux has two extra args, restart and oldmask.  We dont use these,
1024  * but it seems that "restart" is actually a context pointer that
1025  * enables the signal to happen with a different register set.
1026  *
1027  * MPALMOSTSAFE
1028  */
1029 int
1030 sys_linux_sigsuspend(struct linux_sigsuspend_args *args)
1031 {
1032         l_sigset_t linux_mask;
1033         sigset_t mask;
1034         int error;
1035
1036 #ifdef DEBUG
1037         if (ldebug(sigsuspend))
1038                 kprintf(ARGS(sigsuspend, "%08lx"), (unsigned long)args->mask);
1039 #endif
1040
1041         LINUX_SIGEMPTYSET(mask);
1042         mask.__bits[0] = args->mask;
1043         linux_to_bsd_sigset(&linux_mask, &mask);
1044
1045         get_mplock();
1046         error = kern_sigsuspend(&mask);
1047         rel_mplock();
1048
1049         return(error);
1050 }
1051
1052 /*
1053  * MPALMOSTSAFE
1054  */
1055 int
1056 sys_linux_rt_sigsuspend(struct linux_rt_sigsuspend_args *uap)
1057 {
1058         l_sigset_t linux_mask;
1059         sigset_t mask;
1060         int error;
1061
1062 #ifdef DEBUG
1063         if (ldebug(rt_sigsuspend))
1064                 kprintf(ARGS(rt_sigsuspend, "%p, %d"),
1065                     (void *)uap->newset, uap->sigsetsize);
1066 #endif
1067
1068         if (uap->sigsetsize != sizeof(l_sigset_t))
1069                 return (EINVAL);
1070
1071         error = copyin(uap->newset, &linux_mask, sizeof(l_sigset_t));
1072         if (error)
1073                 return (error);
1074
1075         linux_to_bsd_sigset(&linux_mask, &mask);
1076
1077         get_mplock();
1078         error = kern_sigsuspend(&mask);
1079         rel_mplock();
1080
1081         return(error);
1082 }
1083
1084 /*
1085  * MPALMOSTSAFE
1086  */
1087 int
1088 sys_linux_pause(struct linux_pause_args *args)
1089 {
1090         struct thread *td = curthread;
1091         struct lwp *lp = td->td_lwp;
1092         sigset_t mask;
1093         int error;
1094
1095 #ifdef DEBUG
1096         if (ldebug(pause))
1097                 kprintf(ARGS(pause, ""));
1098 #endif
1099
1100         mask = lp->lwp_sigmask;
1101
1102         get_mplock();
1103         error = kern_sigsuspend(&mask);
1104         rel_mplock();
1105
1106         return(error);
1107 }
1108
1109 /*
1110  * MPALMOSTSAFE
1111  */
1112 int
1113 sys_linux_sigaltstack(struct linux_sigaltstack_args *uap)
1114 {
1115         stack_t ss, oss;
1116         l_stack_t linux_ss;
1117         int error;
1118
1119 #ifdef DEBUG
1120         if (ldebug(sigaltstack))
1121                 kprintf(ARGS(sigaltstack, "%p, %p"), uap->uss, uap->uoss);
1122 #endif
1123
1124         if (uap->uss) {
1125                 error = copyin(uap->uss, &linux_ss, sizeof(l_stack_t));
1126                 if (error)
1127                         return (error);
1128
1129                 ss.ss_sp = linux_ss.ss_sp;
1130                 ss.ss_size = linux_ss.ss_size;
1131                 ss.ss_flags = linux_to_bsd_sigaltstack(linux_ss.ss_flags);
1132         }
1133
1134         get_mplock();
1135         error = kern_sigaltstack(uap->uss ? &ss : NULL,
1136                                  uap->uoss ? &oss : NULL);
1137         rel_mplock();
1138
1139         if (error == 0 && uap->uoss) {
1140                 linux_ss.ss_sp = oss.ss_sp;
1141                 linux_ss.ss_size = oss.ss_size;
1142                 linux_ss.ss_flags = bsd_to_linux_sigaltstack(oss.ss_flags);
1143                 error = copyout(&linux_ss, uap->uoss, sizeof(l_stack_t));
1144         }
1145
1146         return (error);
1147 }
1148
1149 int
1150 sys_linux_set_thread_area(struct linux_set_thread_area_args *args)
1151 {
1152         struct segment_descriptor *desc;
1153         struct l_user_desc info;
1154         int error;
1155         int idx;
1156         int a[2];
1157         int i;
1158
1159         error = copyin(args->desc, &info, sizeof(struct l_user_desc));
1160         if (error)
1161                 return (EFAULT);
1162
1163 #ifdef DEBUG
1164         if (ldebug(set_thread_area))
1165                 kprintf(ARGS(set_thread_area, "%i, %x, %x, %i, %i, %i, %i, %i, %i\n"),
1166                       info.entry_number,
1167                       info.base_addr,
1168                       info.limit,
1169                       info.seg_32bit,
1170                       info.contents,
1171                       info.read_exec_only,
1172                       info.limit_in_pages,
1173                       info.seg_not_present,
1174                       info.useable);
1175 #endif
1176
1177         idx = info.entry_number;
1178         if (idx != -1 && (idx < 6 || idx > 8))
1179                 return (EINVAL);
1180
1181         if (idx == -1) {
1182                 /* -1 means finding the first free TLS entry */
1183                 for (i = 0; i < NGTLS; i++) {
1184                         /*
1185                          * try to determine if the TLS entry is empty by looking
1186                          * at the lolimit entry.
1187                          */
1188                         if (curthread->td_tls.tls[idx].sd_lolimit == 0) {
1189                                 idx = i;
1190                                 break;
1191                         }
1192                 }
1193
1194                 if (idx == -1) {
1195                         /*
1196                          * By now we should have an index. If not, it means
1197                          * that no entry is free, so return ESRCH.
1198                          */
1199                         return (ESRCH);
1200                 }
1201         } else {
1202                 /* translate the index from Linux to ours */
1203                 idx -= 6;
1204                 KKASSERT(idx >= 0);
1205         }
1206
1207         /* Tell the caller about the allocated entry number */
1208 #if 0 /* was SMP */
1209         info.entry_number = GTLS_START + mycpu->gd_cpuid * NGDT + idx;
1210 #endif
1211         info.entry_number = GTLS_START + idx;
1212
1213
1214         error = copyout(&info, args->desc, sizeof(struct l_user_desc));
1215         if (error)
1216                 return (error);
1217
1218         if (LINUX_LDT_empty(&info)) {
1219                 a[0] = 0;
1220                 a[1] = 0;
1221         } else {
1222                 a[0] = LINUX_LDT_entry_a(&info);
1223                 a[1] = LINUX_LDT_entry_b(&info);
1224         }
1225
1226         /*
1227          * Update the TLS and the TLS entries in the GDT, but hold a critical
1228          * section as required by set_user_TLS().
1229          */
1230         crit_enter();
1231         desc = &curthread->td_tls.tls[idx];
1232         memcpy(desc, &a, sizeof(a));
1233         set_user_TLS();
1234         crit_exit();
1235
1236         return (0);
1237 }
1238
1239 int
1240 sys_linux_get_thread_area(struct linux_get_thread_area_args *args)
1241 {
1242         struct segment_descriptor *sd;
1243         struct l_desc_struct desc;
1244         struct l_user_desc info;
1245         int error;
1246         int idx;
1247
1248 #ifdef DEBUG
1249         if (ldebug(get_thread_area))
1250                 kprintf(ARGS(get_thread_area, "%p"), args->desc);
1251 #endif
1252
1253         error = copyin(args->desc, &info, sizeof(struct l_user_desc));
1254         if (error)
1255                 return (EFAULT);
1256
1257         idx = info.entry_number;
1258         if ((idx < 6 || idx > 8) && (idx < GTLS_START)) {
1259                 kprintf("sys_linux_get_thread_area, invalid idx requested: %d\n", idx);
1260                 return (EINVAL);
1261         }
1262
1263         memset(&info, 0, sizeof(info));
1264
1265         /* translate the index from Linux to ours */
1266         info.entry_number = idx;
1267         if (idx < GTLS_START) {
1268                 idx -= 6;
1269         } else {
1270 #if 0 /* was SMP */
1271                 idx -= (GTLS_START + mycpu->gd_cpuid * NGDT);
1272 #endif
1273                 idx -= GTLS_START;
1274
1275         }
1276         KKASSERT(idx >= 0);
1277
1278         sd = &curthread->td_tls.tls[idx];
1279         memcpy(&desc, sd, sizeof(desc));
1280         info.base_addr = LINUX_GET_BASE(&desc);
1281         info.limit = LINUX_GET_LIMIT(&desc);
1282         info.seg_32bit = LINUX_GET_32BIT(&desc);
1283         info.contents = LINUX_GET_CONTENTS(&desc);
1284         info.read_exec_only = !LINUX_GET_WRITABLE(&desc);
1285         info.limit_in_pages = LINUX_GET_LIMIT_PAGES(&desc);
1286         info.seg_not_present = !LINUX_GET_PRESENT(&desc);
1287         info.useable = LINUX_GET_USEABLE(&desc);
1288
1289         error = copyout(&info, args->desc, sizeof(struct l_user_desc));
1290         if (error)
1291                 return (EFAULT);
1292
1293         return (0);
1294 }