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 sig;
 399
 400         sig = args->rval;
 401
 402         get_mplock();
 403
 404         EMUL_LOCK();
 405
 406         em = emuldata_get(curproc);
 407
 408         if (em->s->refs == 1) {
 409                 exit1(W_EXITCODE(0, sig));
 410                 /* notreached */
 411
 412                 EMUL_UNLOCK();
 413
 414                 rel_mplock();
 415                 return (0);
 416         }
 417         KKASSERT(em->proc == curproc);
 418         em->flags |= EMUL_DIDKILL;
 419         em->s->flags |= LINUX_LES_INEXITGROUP;
 420         em->s->xstat = W_EXITCODE(0, sig);
 421
 422         LIST_REMOVE(em, threads);
 423         LIST_INSERT_HEAD(&em->s->threads, em, threads);
 424
 425         while ((e = LIST_NEXT(em, threads)) != NULL) {
 426                 LIST_REMOVE(em, threads);
 427                 LIST_INSERT_AFTER(e, em, threads);
 428                 if ((e->flags & EMUL_DIDKILL) == 0) {
 429                         e->flags |= EMUL_DIDKILL;
 430                         KKASSERT(pfind(e->proc->p_pid) == e->proc);
 431                         ksignal(e->proc, SIGKILL);
 432                 }
 433         }
 434
 435
 436         EMUL_UNLOCK();
 437
 438         exit1(W_EXITCODE(0, sig));
 439         rel_mplock();
 440         /* notreached */
 441
 442         return (0);
 443 }
 444
 445 /*
 446  * MPSAFE
 447  */
 448 int
 449 sys_linux_vfork(struct linux_vfork_args *args)
 450 {
 451         struct lwp *lp = curthread->td_lwp;
 452         struct proc *p2;
 453         int error;
 454
 455         get_mplock();
 456         error = fork1(lp, RFFDG | RFPROC | RFPPWAIT | RFMEM | RFPGLOCK, &p2);
 457         if (error == 0) {
 458                 emuldata_init(curproc, p2, 0);
 459
 460                 start_forked_proc(lp, p2);
 461                 args->sysmsg_fds[0] = p2->p_pid;
 462                 args->sysmsg_fds[1] = 0;
 463         }
 464         rel_mplock();
 465
 466         if (args->sysmsg_iresult == 1)
 467                 args->sysmsg_iresult = 0;
 468
 469         return (error);
 470 }
 471
 472 /*
 473  * MPALMOSTSAFE
 474  */
 475 int
 476 sys_linux_clone(struct linux_clone_args *args)
 477 {
 478         struct segment_descriptor *desc;
 479         struct l_user_desc info;
 480         int idx;
 481         int a[2];
 482
 483         struct lwp *lp = curthread->td_lwp;
 484         int error, ff = RFPROC;
 485         struct proc *p2 = NULL;
 486         int exit_signal;
 487         vm_offset_t start;
 488
 489 #ifdef DEBUG
 490         if (ldebug(clone)) {
 491                 kprintf(ARGS(clone, "flags %x, stack %x"),
 492                     (unsigned int)args->flags, (unsigned int)args->stack);
 493                 if (args->flags & CLONE_PID)
 494                         kprintf(LMSG("CLONE_PID not yet supported"));
 495         }
 496 #endif
 497         exit_signal = args->flags & 0x000000ff;
 498         if (exit_signal >= LINUX_NSIG)
 499                 return (EINVAL);
 500         if (exit_signal <= LINUX_SIGTBLSZ)
 501                 exit_signal = linux_to_bsd_signal[_SIG_IDX(exit_signal)];
 502
 503         if (args->flags & LINUX_CLONE_VM)
 504                 ff |= RFMEM;
 505         if (args->flags & LINUX_CLONE_SIGHAND)
 506                 ff |= RFSIGSHARE;
 507         if (!(args->flags & (LINUX_CLONE_FILES | LINUX_CLONE_FS)))
 508                 ff |= RFFDG;
 509         if ((args->flags & 0xffffff00) == LINUX_THREADING_FLAGS)
 510                 ff |= RFTHREAD;
 511         if (args->flags & LINUX_CLONE_VFORK)
 512                 ff |= RFPPWAIT;
 513         if (args->flags & LINUX_CLONE_PARENT_SETTID) {
 514                 if (args->parent_tidptr == NULL)
 515                         return (EINVAL);
 516         }
 517
 518         error = 0;
 519         start = 0;
 520
 521         get_mplock();
 522         error = fork1(lp, ff | RFPGLOCK, &p2);
 523         if (error) {
 524                 rel_mplock();
 525                 return error;
 526         }
 527
 528         args->sysmsg_fds[0] = p2 ? p2->p_pid : 0;
 529         args->sysmsg_fds[1] = 0;
 530
 531         if (args->flags & (LINUX_CLONE_PARENT | LINUX_CLONE_THREAD))
 532                 proc_reparent(p2, curproc->p_pptr /* XXX */);
 533
 534         emuldata_init(curproc, p2, args->flags);
 535         linux_proc_fork(p2, curproc, args->child_tidptr);
 536         /*
 537          * XXX: this can't happen, p2 is never NULL, or else we'd have
 538          *      other problems, too (see p2->p_sigparent == ...,
 539          *      linux_proc_fork and emuldata_init.
 540          */
 541         if (p2 == NULL) {
 542                 error = ESRCH;
 543         } else {
 544                 if (args->flags & LINUX_CLONE_PARENT_SETTID) {
 545                         error = copyout(&p2->p_pid, args->parent_tidptr, sizeof(p2->p_pid));
 546                 }
 547         }
 548
 549         p2->p_sigparent = exit_signal;
 550         if (args->stack) {
 551                 ONLY_LWP_IN_PROC(p2)->lwp_md.md_regs->tf_esp =
 552                                         (unsigned long)args->stack;
 553         }
 554
 555         if (args->flags & LINUX_CLONE_SETTLS) {
 556                 error = copyin((void *)curthread->td_lwp->lwp_md.md_regs->tf_esi, &info, sizeof(struct l_user_desc));
 557                 if (error) {
 558                         kprintf("copyin of tf_esi to info failed\n");
 559                 } else {
 560                         idx = info.entry_number;
 561                         /*
 562                          * We understand both our own entries such as the ones
 563                          * we provide on linux_set_thread_area, as well as the
 564                          * linux-type entries 6-8.
 565                          */
 566                         if ((idx < 6 || idx > 8) && (idx < GTLS_START)) {
 567                                 kprintf("LINUX_CLONE_SETTLS, invalid idx requested: %d\n", idx);
 568                                 goto out;
 569                         }
 570                         if (idx < GTLS_START) {
 571                                 idx -= 6;
 572                         } else {
 573 #ifdef SMP
 574                                 idx -= (GTLS_START + mycpu->gd_cpuid * NGDT);
 575 #else
 576                                 idx -= GTLS_START;
 577 #endif
 578                         }
 579                         KKASSERT(idx >= 0);
 580
 581                         a[0] = LINUX_LDT_entry_a(&info);
 582                         a[1] = LINUX_LDT_entry_b(&info);
 583                         if (p2) {
 584                                 desc = &FIRST_LWP_IN_PROC(p2)->lwp_thread->td_tls.tls[idx];
 585                                 memcpy(desc, &a, sizeof(a));
 586                         } else {
 587                                 kprintf("linux_clone... we don't have a p2\n");
 588                         }
 589                 }
 590         }
 591 out:
 592         if (p2)
 593                 start_forked_proc(lp, p2);
 594
 595         rel_mplock();
 596 #ifdef DEBUG
 597         if (ldebug(clone))
 598                 kprintf(LMSG("clone: successful rfork to %ld"),
 599                     (long)p2->p_pid);
 600 #endif
 601
 602         return (error);
 603 }
 604
 605 /* XXX move */
 606 struct l_mmap_argv {
 607         l_caddr_t       addr;
 608         l_int           len;
 609         l_int           prot;
 610         l_int           flags;
 611         l_int           fd;
 612         l_int           pos;
 613 };
 614
 615 #define STACK_SIZE  (2 * 1024 * 1024)
 616 #define GUARD_SIZE  (4 * PAGE_SIZE)
 617
 618 /*
 619  * MPALMOSTSAFE
 620  */
 621 static int
 622 linux_mmap_common(caddr_t linux_addr, size_t linux_len, int linux_prot,
 623                   int linux_flags, int linux_fd, off_t pos, void **res)
 624 {
 625         struct thread *td = curthread;
 626         struct proc *p = td->td_proc;
 627         caddr_t addr;
 628         void *new;
 629         int error, flags, len, prot, fd;
 630
 631         flags = 0;
 632         if (linux_flags & LINUX_MAP_SHARED)
 633                 flags |= MAP_SHARED;
 634         if (linux_flags & LINUX_MAP_PRIVATE)
 635                 flags |= MAP_PRIVATE;
 636         if (linux_flags & LINUX_MAP_FIXED)
 637                 flags |= MAP_FIXED;
 638         if (linux_flags & LINUX_MAP_ANON) {
 639                 flags |= MAP_ANON;
 640         } else {
 641                 flags |= MAP_NOSYNC;
 642         }
 643
 644         get_mplock();
 645
 646         if (linux_flags & LINUX_MAP_GROWSDOWN) {
 647                 flags |= MAP_STACK;
 648                 /* The linux MAP_GROWSDOWN option does not limit auto
 649                  * growth of the region.  Linux mmap with this option
 650                  * takes as addr the inital BOS, and as len, the initial
 651                  * region size.  It can then grow down from addr without
 652                  * limit.  However, linux threads has an implicit internal
 653                  * limit to stack size of STACK_SIZE.  Its just not
 654                  * enforced explicitly in linux.  But, here we impose
 655                  * a limit of (STACK_SIZE - GUARD_SIZE) on the stack
 656                  * region, since we can do this with our mmap.
 657                  *
 658                  * Our mmap with MAP_STACK takes addr as the maximum
 659                  * downsize limit on BOS, and as len the max size of
 660                  * the region.  It them maps the top SGROWSIZ bytes,
 661                  * and autgrows the region down, up to the limit
 662                  * in addr.
 663                  *
 664                  * If we don't use the MAP_STACK option, the effect
 665                  * of this code is to allocate a stack region of a
 666                  * fixed size of (STACK_SIZE - GUARD_SIZE).
 667                  */
 668
 669                 /* This gives us TOS */
 670                 addr = linux_addr + linux_len;
 671
 672                 if (addr > p->p_vmspace->vm_maxsaddr) {
 673                         /* Some linux apps will attempt to mmap
 674                          * thread stacks near the top of their
 675                          * address space.  If their TOS is greater
 676                          * than vm_maxsaddr, vm_map_growstack()
 677                          * will confuse the thread stack with the
 678                          * process stack and deliver a SEGV if they
 679                          * attempt to grow the thread stack past their
 680                          * current stacksize rlimit.  To avoid this,
 681                          * adjust vm_maxsaddr upwards to reflect
 682                          * the current stacksize rlimit rather
 683                          * than the maximum possible stacksize.
 684                          * It would be better to adjust the
 685                          * mmap'ed region, but some apps do not check
 686                          * mmap's return value.
 687                          */
 688                         p->p_vmspace->vm_maxsaddr = (char *)USRSTACK -
 689                             p->p_rlimit[RLIMIT_STACK].rlim_cur;
 690                 }
 691
 692                 /* This gives us our maximum stack size */
 693                 if (linux_len > STACK_SIZE - GUARD_SIZE) {
 694                         len = linux_len;
 695                 } else {
 696                         len = STACK_SIZE - GUARD_SIZE;
 697                 }
 698                 /* This gives us a new BOS.  If we're using VM_STACK, then
 699                  * mmap will just map the top SGROWSIZ bytes, and let
 700                  * the stack grow down to the limit at BOS.  If we're
 701                  * not using VM_STACK we map the full stack, since we
 702                  * don't have a way to autogrow it.
 703                  */
 704                 addr -= len;
 705         } else {
 706                 addr = linux_addr;
 707                 len = linux_len;
 708         }
 709
 710         prot = linux_prot | PROT_READ;
 711         if (linux_flags & LINUX_MAP_ANON) {
 712                 fd = -1;
 713         } else {
 714                 fd = linux_fd;
 715         }
 716
 717 #ifdef DEBUG
 718         if (ldebug(mmap) || ldebug(mmap2))
 719                 kprintf("-> (%p, %d, %d, 0x%08x, %d, %lld)\n",
 720                     addr, len, prot, flags, fd, pos);
 721 #endif
 722         error = kern_mmap(curproc->p_vmspace, addr, len,
 723                           prot, flags, fd, pos, &new);
 724         rel_mplock();
 725
 726         if (error == 0)
 727                 *res = new;
 728         return (error);
 729 }
 730
 731 /*
 732  * MPSAFE
 733  */
 734 int
 735 sys_linux_mmap(struct linux_mmap_args *args)
 736 {
 737         struct l_mmap_argv linux_args;
 738         int error;
 739
 740         error = copyin((caddr_t)args->ptr, &linux_args, sizeof(linux_args));
 741         if (error)
 742                 return (error);
 743
 744 #ifdef DEBUG
 745         if (ldebug(mmap))
 746                 kprintf(ARGS(mmap, "%p, %d, %d, 0x%08x, %d, %d"),
 747                     (void *)linux_args.addr, linux_args.len, linux_args.prot,
 748                     linux_args.flags, linux_args.fd, linux_args.pos);
 749 #endif
 750         error = linux_mmap_common(linux_args.addr, linux_args.len,
 751             linux_args.prot, linux_args.flags, linux_args.fd,
 752             linux_args.pos, &args->sysmsg_resultp);
 753 #ifdef DEBUG
 754         if (ldebug(mmap))
 755                 kprintf("-> %p\n", args->sysmsg_resultp);
 756 #endif
 757         return(error);
 758 }
 759
 760 /*
 761  * MPSAFE
 762  */
 763 int
 764 sys_linux_mmap2(struct linux_mmap2_args *args)
 765 {
 766         int error;
 767
 768 #ifdef DEBUG
 769         if (ldebug(mmap2))
 770                 kprintf(ARGS(mmap2, "%p, %d, %d, 0x%08x, %d, %d"),
 771                     (void *)args->addr, args->len, args->prot, args->flags,
 772                     args->fd, args->pgoff);
 773 #endif
 774         error = linux_mmap_common((void *)args->addr, args->len, args->prot,
 775             args->flags, args->fd, args->pgoff * PAGE_SIZE,
 776             &args->sysmsg_resultp);
 777 #ifdef DEBUG
 778         if (ldebug(mmap2))
 779                 kprintf("-> %p\n", args->sysmsg_resultp);
 780 #endif
 781         return (error);
 782 }
 783
 784 /*
 785  * MPSAFE
 786  */
 787 int
 788 sys_linux_pipe(struct linux_pipe_args *args)
 789 {
 790         int error;
 791         int reg_edx;
 792         struct pipe_args bsd_args;
 793
 794 #ifdef DEBUG
 795         if (ldebug(pipe))
 796                 kprintf(ARGS(pipe, "*"));
 797 #endif
 798
 799         reg_edx = args->sysmsg_fds[1];
 800         error = sys_pipe(&bsd_args);
 801         if (error) {
 802                 args->sysmsg_fds[1] = reg_edx;
 803                 return (error);
 804         }
 805
 806         error = copyout(bsd_args.sysmsg_fds, args->pipefds, 2*sizeof(int));
 807         if (error) {
 808                 args->sysmsg_fds[1] = reg_edx;
 809                 return (error);
 810         }
 811
 812         args->sysmsg_fds[1] = reg_edx;
 813         args->sysmsg_fds[0] = 0;
 814         return (0);
 815 }
 816
 817 /*
 818  * MPSAFE
 819  */
 820 int
 821 sys_linux_ioperm(struct linux_ioperm_args *args)
 822 {
 823         struct sysarch_args sa;
 824         struct i386_ioperm_args *iia;
 825         caddr_t sg;
 826         int error;
 827
 828         sg = stackgap_init();
 829         iia = stackgap_alloc(&sg, sizeof(struct i386_ioperm_args));
 830         iia->start = args->start;
 831         iia->length = args->length;
 832         iia->enable = args->enable;
 833         sa.sysmsg_resultp = NULL;
 834         sa.op = I386_SET_IOPERM;
 835         sa.parms = (char *)iia;
 836         error = sys_sysarch(&sa);
 837         args->sysmsg_resultp = sa.sysmsg_resultp;
 838         return(error);
 839 }
 840
 841 /*
 842  * MPSAFE
 843  */
 844 int
 845 sys_linux_iopl(struct linux_iopl_args *args)
 846 {
 847         struct thread *td = curthread;
 848         struct lwp *lp = td->td_lwp;
 849         int error;
 850
 851         if (args->level < 0 || args->level > 3)
 852                 return (EINVAL);
 853         if ((error = priv_check(td, PRIV_ROOT)) != 0)
 854                 return (error);
 855         if (securelevel > 0)
 856                 return (EPERM);
 857         lp->lwp_md.md_regs->tf_eflags =
 858             (lp->lwp_md.md_regs->tf_eflags & ~PSL_IOPL) |
 859             (args->level * (PSL_IOPL / 3));
 860         return (0);
 861 }
 862
 863 /*
 864  * MPSAFE
 865  */
 866 int
 867 sys_linux_modify_ldt(struct linux_modify_ldt_args *uap)
 868 {
 869         int error;
 870         caddr_t sg;
 871         struct sysarch_args args;
 872         struct i386_ldt_args *ldt;
 873         struct l_descriptor ld;
 874         union descriptor *desc;
 875         int size, written;
 876
 877         sg = stackgap_init();
 878
 879         if (uap->ptr == NULL)
 880                 return (EINVAL);
 881
 882         switch (uap->func) {
 883         case 0x00: /* read_ldt */
 884                 ldt = stackgap_alloc(&sg, sizeof(*ldt));
 885                 ldt->start = 0;
 886                 ldt->descs = uap->ptr;
 887                 ldt->num = uap->bytecount / sizeof(union descriptor);
 888                 args.op = I386_GET_LDT;
 889                 args.parms = (char*)ldt;
 890                 args.sysmsg_iresult = 0;
 891                 error = sys_sysarch(&args);
 892                 uap->sysmsg_iresult = args.sysmsg_iresult *
 893                                       sizeof(union descriptor);
 894                 break;
 895         case 0x02: /* read_default_ldt = 0 */
 896                 size = 5*sizeof(struct l_desc_struct);
 897                 if (size > uap->bytecount)
 898                         size = uap->bytecount;
 899                 for (written = error = 0; written < size && error == 0; written++)
 900                         error = subyte((char *)uap->ptr + written, 0);
 901                 uap->sysmsg_iresult = written;
 902                 break;
 903         case 0x01: /* write_ldt */
 904         case 0x11: /* write_ldt */
 905                 if (uap->bytecount != sizeof(ld))
 906                         return (EINVAL);
 907
 908                 error = copyin(uap->ptr, &ld, sizeof(ld));
 909                 if (error)
 910                         return (error);
 911
 912                 ldt = stackgap_alloc(&sg, sizeof(*ldt));
 913                 desc = stackgap_alloc(&sg, sizeof(*desc));
 914                 ldt->start = ld.entry_number;
 915                 ldt->descs = desc;
 916                 ldt->num = 1;
 917                 desc->sd.sd_lolimit = (ld.limit & 0x0000ffff);
 918                 desc->sd.sd_hilimit = (ld.limit & 0x000f0000) >> 16;
 919                 desc->sd.sd_lobase = (ld.base_addr & 0x00ffffff);
 920                 desc->sd.sd_hibase = (ld.base_addr & 0xff000000) >> 24;
 921                 desc->sd.sd_type = SDT_MEMRO | ((ld.read_exec_only ^ 1) << 1) |
 922                         (ld.contents << 2);
 923                 desc->sd.sd_dpl = 3;
 924                 desc->sd.sd_p = (ld.seg_not_present ^ 1);
 925                 desc->sd.sd_xx = 0;
 926                 desc->sd.sd_def32 = ld.seg_32bit;
 927                 desc->sd.sd_gran = ld.limit_in_pages;
 928                 args.op = I386_SET_LDT;
 929                 args.parms = (char*)ldt;
 930                 args.sysmsg_iresult = 0;
 931                 error = sys_sysarch(&args);
 932                 uap->sysmsg_iresult = args.sysmsg_iresult;
 933                 break;
 934         default:
 935                 error = EINVAL;
 936                 break;
 937         }
 938
 939         return (error);
 940 }
 941
 942 /*
 943  * MPALMOSTSAFE
 944  */
 945 int
 946 sys_linux_sigaction(struct linux_sigaction_args *args)
 947 {
 948         l_osigaction_t osa;
 949         l_sigaction_t linux_act, linux_oact;
 950         struct sigaction act, oact;
 951         int error;
 952
 953 #ifdef DEBUG
 954         if (ldebug(sigaction))
 955                 kprintf(ARGS(sigaction, "%d, %p, %p"),
 956                     args->sig, (void *)args->nsa, (void *)args->osa);
 957 #endif
 958
 959         if (args->nsa) {
 960                 error = copyin(args->nsa, &osa, sizeof(l_osigaction_t));
 961                 if (error)
 962                         return (error);
 963                 linux_act.lsa_handler = osa.lsa_handler;
 964                 linux_act.lsa_flags = osa.lsa_flags;
 965                 linux_act.lsa_restorer = osa.lsa_restorer;
 966                 LINUX_SIGEMPTYSET(linux_act.lsa_mask);
 967                 linux_act.lsa_mask.__bits[0] = osa.lsa_mask;
 968                 linux_to_bsd_sigaction(&linux_act, &act);
 969         }
 970
 971         get_mplock();
 972         error = kern_sigaction(args->sig, args->nsa ? &act : NULL,
 973                                args->osa ? &oact : NULL);
 974         rel_mplock();
 975
 976         if (args->osa != NULL && !error) {
 977                 bsd_to_linux_sigaction(&oact, &linux_oact);
 978                 osa.lsa_handler = linux_oact.lsa_handler;
 979                 osa.lsa_flags = linux_oact.lsa_flags;
 980                 osa.lsa_restorer = linux_oact.lsa_restorer;
 981                 osa.lsa_mask = linux_oact.lsa_mask.__bits[0];
 982                 error = copyout(&osa, args->osa, sizeof(l_osigaction_t));
 983         }
 984         return (error);
 985 }
 986
 987 /*
 988  * Linux has two extra args, restart and oldmask.  We dont use these,
 989  * but it seems that "restart" is actually a context pointer that
 990  * enables the signal to happen with a different register set.
 991  *
 992  * MPALMOSTSAFE
 993  */
 994 int
 995 sys_linux_sigsuspend(struct linux_sigsuspend_args *args)
 996 {
 997         l_sigset_t linux_mask;
 998         sigset_t mask;
 999         int error;
1000
1001 #ifdef DEBUG
1002         if (ldebug(sigsuspend))
1003                 kprintf(ARGS(sigsuspend, "%08lx"), (unsigned long)args->mask);
1004 #endif
1005
1006         LINUX_SIGEMPTYSET(mask);
1007         mask.__bits[0] = args->mask;
1008         linux_to_bsd_sigset(&linux_mask, &mask);
1009
1010         get_mplock();
1011         error = kern_sigsuspend(&mask);
1012         rel_mplock();
1013
1014         return(error);
1015 }
1016
1017 /*
1018  * MPALMOSTSAFE
1019  */
1020 int
1021 sys_linux_rt_sigsuspend(struct linux_rt_sigsuspend_args *uap)
1022 {
1023         l_sigset_t linux_mask;
1024         sigset_t mask;
1025         int error;
1026
1027 #ifdef DEBUG
1028         if (ldebug(rt_sigsuspend))
1029                 kprintf(ARGS(rt_sigsuspend, "%p, %d"),
1030                     (void *)uap->newset, uap->sigsetsize);
1031 #endif
1032
1033         if (uap->sigsetsize != sizeof(l_sigset_t))
1034                 return (EINVAL);
1035
1036         error = copyin(uap->newset, &linux_mask, sizeof(l_sigset_t));
1037         if (error)
1038                 return (error);
1039
1040         linux_to_bsd_sigset(&linux_mask, &mask);
1041
1042         get_mplock();
1043         error = kern_sigsuspend(&mask);
1044         rel_mplock();
1045
1046         return(error);
1047 }
1048
1049 /*
1050  * MPALMOSTSAFE
1051  */
1052 int
1053 sys_linux_pause(struct linux_pause_args *args)
1054 {
1055         struct thread *td = curthread;
1056         struct lwp *lp = td->td_lwp;
1057         sigset_t mask;
1058         int error;
1059
1060 #ifdef DEBUG
1061         if (ldebug(pause))
1062                 kprintf(ARGS(pause, ""));
1063 #endif
1064
1065         mask = lp->lwp_sigmask;
1066
1067         get_mplock();
1068         error = kern_sigsuspend(&mask);
1069         rel_mplock();
1070
1071         return(error);
1072 }
1073
1074 /*
1075  * MPALMOSTSAFE
1076  */
1077 int
1078 sys_linux_sigaltstack(struct linux_sigaltstack_args *uap)
1079 {
1080         stack_t ss, oss;
1081         l_stack_t linux_ss;
1082         int error;
1083
1084 #ifdef DEBUG
1085         if (ldebug(sigaltstack))
1086                 kprintf(ARGS(sigaltstack, "%p, %p"), uap->uss, uap->uoss);
1087 #endif
1088
1089         if (uap->uss) {
1090                 error = copyin(uap->uss, &linux_ss, sizeof(l_stack_t));
1091                 if (error)
1092                         return (error);
1093
1094                 ss.ss_sp = linux_ss.ss_sp;
1095                 ss.ss_size = linux_ss.ss_size;
1096                 ss.ss_flags = linux_to_bsd_sigaltstack(linux_ss.ss_flags);
1097         }
1098
1099         get_mplock();
1100         error = kern_sigaltstack(uap->uss ? &ss : NULL,
1101                                  uap->uoss ? &oss : NULL);
1102         rel_mplock();
1103
1104         if (error == 0 && uap->uoss) {
1105                 linux_ss.ss_sp = oss.ss_sp;
1106                 linux_ss.ss_size = oss.ss_size;
1107                 linux_ss.ss_flags = bsd_to_linux_sigaltstack(oss.ss_flags);
1108                 error = copyout(&linux_ss, uap->uoss, sizeof(l_stack_t));
1109         }
1110
1111         return (error);
1112 }
1113
1114 int
1115 sys_linux_set_thread_area(struct linux_set_thread_area_args *args)
1116 {
1117         struct segment_descriptor *desc;
1118         struct l_user_desc info;
1119         int error;
1120         int idx;
1121         int a[2];
1122         int i;
1123
1124         error = copyin(args->desc, &info, sizeof(struct l_user_desc));
1125         if (error)
1126                 return (EFAULT);
1127
1128 #ifdef DEBUG
1129         if (ldebug(set_thread_area))
1130                 kprintf(ARGS(set_thread_area, "%i, %x, %x, %i, %i, %i, %i, %i, %i\n"),
1131                       info.entry_number,
1132                       info.base_addr,
1133                       info.limit,
1134                       info.seg_32bit,
1135                       info.contents,
1136                       info.read_exec_only,
1137                       info.limit_in_pages,
1138                       info.seg_not_present,
1139                       info.useable);
1140 #endif
1141
1142         idx = info.entry_number;
1143         if (idx != -1 && (idx < 6 || idx > 8))
1144                 return (EINVAL);
1145
1146         if (idx == -1) {
1147                 /* -1 means finding the first free TLS entry */
1148                 for (i = 0; i < NGTLS; i++) {
1149                         /*
1150                          * try to determine if the TLS entry is empty by looking
1151                          * at the lolimit entry.
1152                          */
1153                         if (curthread->td_tls.tls[idx].sd_lolimit == 0) {
1154                                 idx = i;
1155                                 break;
1156                         }
1157                 }
1158
1159                 if (idx == -1) {
1160                         /*
1161                          * By now we should have an index. If not, it means
1162                          * that no entry is free, so return ESRCH.
1163                          */
1164                         return (ESRCH);
1165                 }
1166         } else {
1167                 /* translate the index from Linux to ours */
1168                 idx -= 6;
1169                 KKASSERT(idx >= 0);
1170         }
1171
1172         /* Tell the caller about the allocated entry number */
1173 #if 0
1174         info.entry_number = idx;
1175 #endif
1176 #ifdef SMP
1177         info.entry_number = GTLS_START + mycpu->gd_cpuid * NGDT + idx;
1178 #else
1179         info.entry_number = GTLS_START + idx;
1180 #endif
1181
1182         error = copyout(&info, args->desc, sizeof(struct l_user_desc));
1183         if (error)
1184                 return (error);
1185
1186         if (LINUX_LDT_empty(&info)) {
1187                 a[0] = 0;
1188                 a[1] = 0;
1189         } else {
1190                 a[0] = LINUX_LDT_entry_a(&info);
1191                 a[1] = LINUX_LDT_entry_b(&info);
1192         }
1193
1194         /*
1195          * Update the TLS and the TLS entries in the GDT, but hold a critical
1196          * section as required by set_user_TLS().
1197          */
1198         crit_enter();
1199         desc = &curthread->td_tls.tls[idx];
1200         memcpy(desc, &a, sizeof(a));
1201         set_user_TLS();
1202         crit_exit();
1203
1204         return (0);
1205 }
1206
1207 int
1208 sys_linux_get_thread_area(struct linux_get_thread_area_args *args)
1209 {
1210         struct segment_descriptor *sd;
1211         struct l_desc_struct desc;
1212         struct l_user_desc info;
1213         int error;
1214         int idx;
1215
1216 #ifdef DEBUG
1217         if (ldebug(get_thread_area))
1218                 kprintf(ARGS(get_thread_area, "%p"), args->desc);
1219 #endif
1220
1221         error = copyin(args->desc, &info, sizeof(struct l_user_desc));
1222         if (error)
1223                 return (EFAULT);
1224
1225         idx = info.entry_number;
1226         if ((idx < 6 || idx > 8) && (idx < GTLS_START)) {
1227                 kprintf("sys_linux_get_thread_area, invalid idx requested: %d\n", idx);
1228                 return (EINVAL);
1229         }
1230
1231         memset(&info, 0, sizeof(info));
1232
1233         /* translate the index from Linux to ours */
1234         info.entry_number = idx;
1235         if (idx < GTLS_START) {
1236                 idx -= 6;
1237         } else {
1238 #ifdef SMP
1239                 idx -= (GTLS_START + mycpu->gd_cpuid * NGDT);
1240 #else
1241                 idx -= GTLS_START;
1242 #endif
1243         }
1244         KKASSERT(idx >= 0);
1245
1246         sd = &curthread->td_tls.tls[idx];
1247         memcpy(&desc, sd, sizeof(desc));
1248         info.base_addr = LINUX_GET_BASE(&desc);
1249         info.limit = LINUX_GET_LIMIT(&desc);
1250         info.seg_32bit = LINUX_GET_32BIT(&desc);
1251         info.contents = LINUX_GET_CONTENTS(&desc);
1252         info.read_exec_only = !LINUX_GET_WRITABLE(&desc);
1253         info.limit_in_pages = LINUX_GET_LIMIT_PAGES(&desc);
1254         info.seg_not_present = !LINUX_GET_PRESENT(&desc);
1255         info.useable = LINUX_GET_USEABLE(&desc);
1256
1257         error = copyout(&info, args->desc, sizeof(struct l_user_desc));
1258         if (error)
1259                 return (EFAULT);
1260
1261         return (0);
1262 }