linux32_machdep.c

Version: - FREEBSD - FREEBSD-13-STABLE - FREEBSD-13-0 - FREEBSD-12-STABLE - FREEBSD-12-0 - FREEBSD-11-STABLE - FREEBSD-11-0 - FREEBSD-10-STABLE - FREEBSD-10-0 - FREEBSD-9-STABLE - FREEBSD-9-0 - FREEBSD-8-STABLE - FREEBSD-8-0 - FREEBSD-7-STABLE - FREEBSD-7-0 - FREEBSD-6-STABLE - FREEBSD-6-0 - FREEBSD-5-STABLE - FREEBSD-5-0 - FREEBSD-4-STABLE - FREEBSD-3-STABLE - FREEBSD22 - l41 - OPENBSD - linux-2.6 - MK84 - PLAN9 - xnu-8792
SearchContext: - none - 3 - 10
    1 /*-
    2  * Copyright (c) 2004 Tim J. Robbins
    3  * Copyright (c) 2002 Doug Rabson
    4  * Copyright (c) 2000 Marcel Moolenaar
    5  * All rights reserved.
    6  *
    7  * Redistribution and use in source and binary forms, with or without
    8  * modification, are permitted provided that the following conditions
    9  * are met:
   10  * 1. Redistributions of source code must retain the above copyright
   11  *    notice, this list of conditions and the following disclaimer
   12  *    in this position and unchanged.
   13  * 2. Redistributions in binary form must reproduce the above copyright
   14  *    notice, this list of conditions and the following disclaimer in the
   15  *    documentation and/or other materials provided with the distribution.
   16  * 3. The name of the author may not be used to endorse or promote products
   17  *    derived from this software without specific prior written permission.
   18  *
   19  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
   20  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
   21  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
   22  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
   23  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
   24  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
   25  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
   26  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
   27  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
   28  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
   29  */
   30 
   31 #include <sys/cdefs.h>
   32 __FBSDID("$FreeBSD: releng/6.1/sys/amd64/linux32/linux32_machdep.c 147588 2005-06-24 17:41:28Z jhb $");
   33 
   34 #include <sys/param.h>
   35 #include <sys/kernel.h>
   36 #include <sys/systm.h>
   37 #include <sys/imgact.h>
   38 #include <sys/lock.h>
   39 #include <sys/malloc.h>
   40 #include <sys/mman.h>
   41 #include <sys/mutex.h>
   42 #include <sys/proc.h>
   43 #include <sys/resource.h>
   44 #include <sys/resourcevar.h>
   45 #include <sys/syscallsubr.h>
   46 #include <sys/sysproto.h>
   47 #include <sys/unistd.h>
   48 
   49 #include <machine/frame.h>
   50 
   51 #include <vm/vm.h>
   52 #include <vm/pmap.h>
   53 #include <vm/vm_extern.h>
   54 #include <vm/vm_kern.h>
   55 #include <vm/vm_map.h>
   56 
   57 #include <amd64/linux32/linux.h>
   58 #include <amd64/linux32/linux32_proto.h>
   59 #include <compat/linux/linux_ipc.h>
   60 #include <compat/linux/linux_signal.h>
   61 #include <compat/linux/linux_util.h>
   62 
   63 struct l_old_select_argv {
   64         l_int           nfds;
   65         l_uintptr_t     readfds;
   66         l_uintptr_t     writefds;
   67         l_uintptr_t     exceptfds;
   68         l_uintptr_t     timeout;
   69 } __packed;
   70 
   71 int
   72 linux_to_bsd_sigaltstack(int lsa)
   73 {
   74         int bsa = 0;
   75 
   76         if (lsa & LINUX_SS_DISABLE)
   77                 bsa |= SS_DISABLE;
   78         if (lsa & LINUX_SS_ONSTACK)
   79                 bsa |= SS_ONSTACK;
   80         return (bsa);
   81 }
   82 
   83 int
   84 bsd_to_linux_sigaltstack(int bsa)
   85 {
   86         int lsa = 0;
   87 
   88         if (bsa & SS_DISABLE)
   89                 lsa |= LINUX_SS_DISABLE;
   90         if (bsa & SS_ONSTACK)
   91                 lsa |= LINUX_SS_ONSTACK;
   92         return (lsa);
   93 }
   94 
   95 /*
   96  * Custom version of exec_copyin_args() so that we can translate
   97  * the pointers.
   98  */
   99 static int
  100 linux_exec_copyin_args(struct image_args *args, char *fname,
  101     enum uio_seg segflg, char **argv, char **envv)
  102 {
  103         char *argp, *envp;
  104         u_int32_t *p32, arg;
  105         size_t length;
  106         int error;
  107 
  108         bzero(args, sizeof(*args));
  109         if (argv == NULL)
  110                 return (EFAULT);
  111 
  112         /*
  113          * Allocate temporary demand zeroed space for argument and
  114          *      environment strings
  115          */
  116         args->buf = (char *) kmem_alloc_wait(exec_map,
  117             PATH_MAX + ARG_MAX + MAXSHELLCMDLEN);
  118         if (args->buf == NULL)
  119                 return (ENOMEM);
  120         args->begin_argv = args->buf;
  121         args->endp = args->begin_argv;
  122         args->stringspace = ARG_MAX;
  123 
  124         args->fname = args->buf + ARG_MAX;
  125 
  126         /*
  127          * Copy the file name.
  128          */
  129         error = (segflg == UIO_SYSSPACE) ?
  130             copystr(fname, args->fname, PATH_MAX, &length) :
  131             copyinstr(fname, args->fname, PATH_MAX, &length);
  132         if (error != 0)
  133                 return (error);
  134 
  135         /*
  136          * extract arguments first
  137          */
  138         p32 = (u_int32_t *)argv;
  139         for (;;) {
  140                 error = copyin(p32++, &arg, sizeof(arg));
  141                 if (error)
  142                         return (error);
  143                 if (arg == 0)
  144                         break;
  145                 argp = PTRIN(arg);
  146                 error = copyinstr(argp, args->endp, args->stringspace, &length);
  147                 if (error) {
  148                         if (error == ENAMETOOLONG)
  149                                 return (E2BIG);
  150                         else
  151                                 return (error);
  152                 }
  153                 args->stringspace -= length;
  154                 args->endp += length;
  155                 args->argc++;
  156         }
  157                         
  158         args->begin_envv = args->endp;
  159 
  160         /*
  161          * extract environment strings
  162          */
  163         if (envv) {
  164                 p32 = (u_int32_t *)envv;
  165                 for (;;) {
  166                         error = copyin(p32++, &arg, sizeof(arg));
  167                         if (error)
  168                                 return (error);
  169                         if (arg == 0)
  170                                 break;
  171                         envp = PTRIN(arg);
  172                         error = copyinstr(envp, args->endp, args->stringspace,
  173                             &length);
  174                         if (error) {
  175                                 if (error == ENAMETOOLONG)
  176                                         return (E2BIG);
  177                                 else
  178                                         return (error);
  179                         }
  180                         args->stringspace -= length;
  181                         args->endp += length;
  182                         args->envc++;
  183                 }
  184         }
  185 
  186         return (0);
  187 }
  188 
  189 int
  190 linux_execve(struct thread *td, struct linux_execve_args *args)
  191 {
  192         struct image_args eargs;
  193         char *path;
  194         int error;
  195 
  196         LCONVPATHEXIST(td, args->path, &path);
  197 
  198 #ifdef DEBUG
  199         if (ldebug(execve))
  200                 printf(ARGS(execve, "%s"), path);
  201 #endif
  202 
  203         error = linux_exec_copyin_args(&eargs, path, UIO_SYSSPACE, args->argp,
  204             args->envp);
  205         free(path, M_TEMP);
  206         if (error == 0)
  207                 error = kern_execve(td, &eargs, NULL);
  208         exec_free_args(&eargs);
  209         return (error);
  210 }
  211 
  212 struct iovec32 {
  213         u_int32_t iov_base;
  214         int     iov_len;
  215 };
  216 
  217 CTASSERT(sizeof(struct iovec32) == 8);
  218 
  219 static int
  220 linux32_copyinuio(struct iovec32 *iovp, u_int iovcnt, struct uio **uiop)
  221 {
  222         struct iovec32 iov32;
  223         struct iovec *iov;
  224         struct uio *uio;
  225         u_int iovlen;
  226         int error, i;
  227 
  228         *uiop = NULL;
  229         if (iovcnt > UIO_MAXIOV)
  230                 return (EINVAL);
  231         iovlen = iovcnt * sizeof(struct iovec);
  232         uio = malloc(iovlen + sizeof *uio, M_IOV, M_WAITOK);
  233         iov = (struct iovec *)(uio + 1);
  234         for (i = 0; i < iovcnt; i++) {
  235                 error = copyin(&iovp[i], &iov32, sizeof(struct iovec32));
  236                 if (error) {
  237                         free(uio, M_IOV);
  238                         return (error);
  239                 }
  240                 iov[i].iov_base = PTRIN(iov32.iov_base);
  241                 iov[i].iov_len = iov32.iov_len;
  242         }
  243         uio->uio_iov = iov;
  244         uio->uio_iovcnt = iovcnt;
  245         uio->uio_segflg = UIO_USERSPACE;
  246         uio->uio_offset = -1;
  247         uio->uio_resid = 0;
  248         for (i = 0; i < iovcnt; i++) {
  249                 if (iov->iov_len > INT_MAX - uio->uio_resid) {
  250                         free(uio, M_IOV);
  251                         return (EINVAL);
  252                 }
  253                 uio->uio_resid += iov->iov_len;
  254                 iov++;
  255         }
  256         *uiop = uio;
  257         return (0);
  258 }
  259 
  260 int
  261 linux_readv(struct thread *td, struct linux_readv_args *uap)
  262 {
  263         struct uio *auio;
  264         int error;
  265 
  266         error = linux32_copyinuio(uap->iovp, uap->iovcnt, &auio);
  267         if (error)
  268                 return (error);
  269         error = kern_readv(td, uap->fd, auio);
  270         free(auio, M_IOV);
  271         return (error);
  272 }
  273 
  274 int
  275 linux_writev(struct thread *td, struct linux_writev_args *uap)
  276 {
  277         struct uio *auio;
  278         int error;
  279 
  280         error = linux32_copyinuio(uap->iovp, uap->iovcnt, &auio);
  281         if (error)
  282                 return (error);
  283         error = kern_writev(td, uap->fd, auio);
  284         free(auio, M_IOV);
  285         return (error);
  286 }
  287 
  288 struct l_ipc_kludge {
  289         l_uintptr_t msgp;
  290         l_long msgtyp;
  291 } __packed;
  292 
  293 int
  294 linux_ipc(struct thread *td, struct linux_ipc_args *args)
  295 {
  296 
  297         switch (args->what & 0xFFFF) {
  298         case LINUX_SEMOP: {
  299                 struct linux_semop_args a;
  300 
  301                 a.semid = args->arg1;
  302                 a.tsops = args->ptr;
  303                 a.nsops = args->arg2;
  304                 return (linux_semop(td, &a));
  305         }
  306         case LINUX_SEMGET: {
  307                 struct linux_semget_args a;
  308 
  309                 a.key = args->arg1;
  310                 a.nsems = args->arg2;
  311                 a.semflg = args->arg3;
  312                 return (linux_semget(td, &a));
  313         }
  314         case LINUX_SEMCTL: {
  315                 struct linux_semctl_args a;
  316                 int error;
  317 
  318                 a.semid = args->arg1;
  319                 a.semnum = args->arg2;
  320                 a.cmd = args->arg3;
  321                 error = copyin(args->ptr, &a.arg, sizeof(a.arg));
  322                 if (error)
  323                         return (error);
  324                 return (linux_semctl(td, &a));
  325         }
  326         case LINUX_MSGSND: {
  327                 struct linux_msgsnd_args a;
  328 
  329                 a.msqid = args->arg1;
  330                 a.msgp = args->ptr;
  331                 a.msgsz = args->arg2;
  332                 a.msgflg = args->arg3;
  333                 return (linux_msgsnd(td, &a));
  334         }
  335         case LINUX_MSGRCV: {
  336                 struct linux_msgrcv_args a;
  337 
  338                 a.msqid = args->arg1;
  339                 a.msgsz = args->arg2;
  340                 a.msgflg = args->arg3;
  341                 if ((args->what >> 16) == 0) {
  342                         struct l_ipc_kludge tmp;
  343                         int error;
  344 
  345                         if (args->ptr == 0)
  346                                 return (EINVAL);
  347                         error = copyin(args->ptr, &tmp, sizeof(tmp));
  348                         if (error)
  349                                 return (error);
  350                         a.msgp = PTRIN(tmp.msgp);
  351                         a.msgtyp = tmp.msgtyp;
  352                 } else {
  353                         a.msgp = args->ptr;
  354                         a.msgtyp = args->arg5;
  355                 }
  356                 return (linux_msgrcv(td, &a));
  357         }
  358         case LINUX_MSGGET: {
  359                 struct linux_msgget_args a;
  360 
  361                 a.key = args->arg1;
  362                 a.msgflg = args->arg2;
  363                 return (linux_msgget(td, &a));
  364         }
  365         case LINUX_MSGCTL: {
  366                 struct linux_msgctl_args a;
  367 
  368                 a.msqid = args->arg1;
  369                 a.cmd = args->arg2;
  370                 a.buf = args->ptr;
  371                 return (linux_msgctl(td, &a));
  372         }
  373         case LINUX_SHMAT: {
  374                 struct linux_shmat_args a;
  375 
  376                 a.shmid = args->arg1;
  377                 a.shmaddr = args->ptr;
  378                 a.shmflg = args->arg2;
  379                 a.raddr = PTRIN((l_uint)args->arg3);
  380                 return (linux_shmat(td, &a));
  381         }
  382         case LINUX_SHMDT: {
  383                 struct linux_shmdt_args a;
  384 
  385                 a.shmaddr = args->ptr;
  386                 return (linux_shmdt(td, &a));
  387         }
  388         case LINUX_SHMGET: {
  389                 struct linux_shmget_args a;
  390 
  391                 a.key = args->arg1;
  392                 a.size = args->arg2;
  393                 a.shmflg = args->arg3;
  394                 return (linux_shmget(td, &a));
  395         }
  396         case LINUX_SHMCTL: {
  397                 struct linux_shmctl_args a;
  398 
  399                 a.shmid = args->arg1;
  400                 a.cmd = args->arg2;
  401                 a.buf = args->ptr;
  402                 return (linux_shmctl(td, &a));
  403         }
  404         default:
  405                 break;
  406         }
  407 
  408         return (EINVAL);
  409 }
  410 
  411 int
  412 linux_old_select(struct thread *td, struct linux_old_select_args *args)
  413 {
  414         struct l_old_select_argv linux_args;
  415         struct linux_select_args newsel;
  416         int error;
  417 
  418 #ifdef DEBUG
  419         if (ldebug(old_select))
  420                 printf(ARGS(old_select, "%p"), args->ptr);
  421 #endif
  422 
  423         error = copyin(args->ptr, &linux_args, sizeof(linux_args));
  424         if (error)
  425                 return (error);
  426 
  427         newsel.nfds = linux_args.nfds;
  428         newsel.readfds = PTRIN(linux_args.readfds);
  429         newsel.writefds = PTRIN(linux_args.writefds);
  430         newsel.exceptfds = PTRIN(linux_args.exceptfds);
  431         newsel.timeout = PTRIN(linux_args.timeout);
  432         return (linux_select(td, &newsel));
  433 }
  434 
  435 int
  436 linux_fork(struct thread *td, struct linux_fork_args *args)
  437 {
  438         int error;
  439 
  440 #ifdef DEBUG
  441         if (ldebug(fork))
  442                 printf(ARGS(fork, ""));
  443 #endif
  444 
  445         if ((error = fork(td, (struct fork_args *)args)) != 0)
  446                 return (error);
  447 
  448         if (td->td_retval[1] == 1)
  449                 td->td_retval[0] = 0;
  450         return (0);
  451 }
  452 
  453 int
  454 linux_vfork(struct thread *td, struct linux_vfork_args *args)
  455 {
  456         int error;
  457 
  458 #ifdef DEBUG
  459         if (ldebug(vfork))
  460                 printf(ARGS(vfork, ""));
  461 #endif
  462 
  463         if ((error = vfork(td, (struct vfork_args *)args)) != 0)
  464                 return (error);
  465         /* Are we the child? */
  466         if (td->td_retval[1] == 1)
  467                 td->td_retval[0] = 0;
  468         return (0);
  469 }
  470 
  471 #define CLONE_VM        0x100
  472 #define CLONE_FS        0x200
  473 #define CLONE_FILES     0x400
  474 #define CLONE_SIGHAND   0x800
  475 #define CLONE_PID       0x1000
  476 
  477 int
  478 linux_clone(struct thread *td, struct linux_clone_args *args)
  479 {
  480         int error, ff = RFPROC | RFSTOPPED;
  481         struct proc *p2;
  482         struct thread *td2;
  483         int exit_signal;
  484 
  485 #ifdef DEBUG
  486         if (ldebug(clone)) {
  487                 printf(ARGS(clone, "flags %x, stack %x"),
  488                     (unsigned int)(uintptr_t)args->flags,
  489                     (unsigned int)(uintptr_t)args->stack);
  490                 if (args->flags & CLONE_PID)
  491                         printf(LMSG("CLONE_PID not yet supported"));
  492         }
  493 #endif
  494 
  495         if (!args->stack)
  496                 return (EINVAL);
  497 
  498         exit_signal = args->flags & 0x000000ff;
  499         if (exit_signal >= LINUX_NSIG)
  500                 return (EINVAL);
  501 
  502         if (exit_signal <= LINUX_SIGTBLSZ)
  503                 exit_signal = linux_to_bsd_signal[_SIG_IDX(exit_signal)];
  504 
  505         if (args->flags & CLONE_VM)
  506                 ff |= RFMEM;
  507         if (args->flags & CLONE_SIGHAND)
  508                 ff |= RFSIGSHARE;
  509         if (!(args->flags & CLONE_FILES))
  510                 ff |= RFFDG;
  511 
  512         error = fork1(td, ff, 0, &p2);
  513         if (error)
  514                 return (error);
  515         
  516 
  517         PROC_LOCK(p2);
  518         p2->p_sigparent = exit_signal;
  519         PROC_UNLOCK(p2);
  520         td2 = FIRST_THREAD_IN_PROC(p2);
  521         td2->td_frame->tf_rsp = PTROUT(args->stack);
  522 
  523 #ifdef DEBUG
  524         if (ldebug(clone))
  525                 printf(LMSG("clone: successful rfork to %ld, stack %p sig = %d"),
  526                     (long)p2->p_pid, args->stack, exit_signal);
  527 #endif
  528 
  529         /*
  530          * Make this runnable after we are finished with it.
  531          */
  532         mtx_lock_spin(&sched_lock);
  533         TD_SET_CAN_RUN(td2);
  534         setrunqueue(td2, SRQ_BORING);
  535         mtx_unlock_spin(&sched_lock);
  536 
  537         td->td_retval[0] = p2->p_pid;
  538         td->td_retval[1] = 0;
  539         return (0);
  540 }
  541 
  542 /* XXX move */
  543 struct l_mmap_argv {
  544         l_ulong         addr;
  545         l_ulong         len;
  546         l_ulong         prot;
  547         l_ulong         flags;
  548         l_ulong         fd;
  549         l_ulong         pgoff;
  550 };
  551 
  552 #define STACK_SIZE  (2 * 1024 * 1024)
  553 #define GUARD_SIZE  (4 * PAGE_SIZE)
  554 
  555 static int linux_mmap_common(struct thread *, struct l_mmap_argv *);
  556 
  557 int
  558 linux_mmap2(struct thread *td, struct linux_mmap2_args *args)
  559 {
  560         struct l_mmap_argv linux_args;
  561 
  562 #ifdef DEBUG
  563         if (ldebug(mmap2))
  564                 printf(ARGS(mmap2, "%p, %d, %d, 0x%08x, %d, %d"),
  565                     (void *)(intptr_t)args->addr, args->len, args->prot,
  566                     args->flags, args->fd, args->pgoff);
  567 #endif
  568 
  569         linux_args.addr = PTROUT(args->addr);
  570         linux_args.len = args->len;
  571         linux_args.prot = args->prot;
  572         linux_args.flags = args->flags;
  573         linux_args.fd = args->fd;
  574         linux_args.pgoff = args->pgoff;
  575 
  576         return (linux_mmap_common(td, &linux_args));
  577 }
  578 
  579 int
  580 linux_mmap(struct thread *td, struct linux_mmap_args *args)
  581 {
  582         int error;
  583         struct l_mmap_argv linux_args;
  584 
  585         error = copyin(args->ptr, &linux_args, sizeof(linux_args));
  586         if (error)
  587                 return (error);
  588 
  589 #ifdef DEBUG
  590         if (ldebug(mmap))
  591                 printf(ARGS(mmap, "%p, %d, %d, 0x%08x, %d, %d"),
  592                     (void *)(intptr_t)linux_args.addr, linux_args.len,
  593                     linux_args.prot, linux_args.flags, linux_args.fd,
  594                     linux_args.pgoff);
  595 #endif
  596         if ((linux_args.pgoff % PAGE_SIZE) != 0)
  597                 return (EINVAL);
  598         linux_args.pgoff /= PAGE_SIZE;
  599 
  600         return (linux_mmap_common(td, &linux_args));
  601 }
  602 
  603 static int
  604 linux_mmap_common(struct thread *td, struct l_mmap_argv *linux_args)
  605 {
  606         struct proc *p = td->td_proc;
  607         struct mmap_args /* {
  608                 caddr_t addr;
  609                 size_t len;
  610                 int prot;
  611                 int flags;
  612                 int fd;
  613                 long pad;
  614                 off_t pos;
  615         } */ bsd_args;
  616         int error;
  617 
  618         error = 0;
  619         bsd_args.flags = 0;
  620         if (linux_args->flags & LINUX_MAP_SHARED)
  621                 bsd_args.flags |= MAP_SHARED;
  622         if (linux_args->flags & LINUX_MAP_PRIVATE)
  623                 bsd_args.flags |= MAP_PRIVATE;
  624         if (linux_args->flags & LINUX_MAP_FIXED)
  625                 bsd_args.flags |= MAP_FIXED;
  626         if (linux_args->flags & LINUX_MAP_ANON)
  627                 bsd_args.flags |= MAP_ANON;
  628         else
  629                 bsd_args.flags |= MAP_NOSYNC;
  630         if (linux_args->flags & LINUX_MAP_GROWSDOWN) {
  631                 bsd_args.flags |= MAP_STACK;
  632 
  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                 bsd_args.addr = (caddr_t)PTRIN(linux_args->addr) +
  656                     linux_args->len;
  657 
  658                 if ((caddr_t)PTRIN(bsd_args.addr) >
  659                     p->p_vmspace->vm_maxsaddr) {
  660                         /* Some linux apps will attempt to mmap
  661                          * thread stacks near the top of their
  662                          * address space.  If their TOS is greater
  663                          * than vm_maxsaddr, vm_map_growstack()
  664                          * will confuse the thread stack with the
  665                          * process stack and deliver a SEGV if they
  666                          * attempt to grow the thread stack past their
  667                          * current stacksize rlimit.  To avoid this,
  668                          * adjust vm_maxsaddr upwards to reflect
  669                          * the current stacksize rlimit rather
  670                          * than the maximum possible stacksize.
  671                          * It would be better to adjust the
  672                          * mmap'ed region, but some apps do not check
  673                          * mmap's return value.
  674                          */
  675                         PROC_LOCK(p);
  676                         p->p_vmspace->vm_maxsaddr =
  677                             (char *)LINUX32_USRSTACK -
  678                             lim_cur(p, RLIMIT_STACK);
  679                         PROC_UNLOCK(p);
  680                 }
  681 
  682                 /* This gives us our maximum stack size */
  683                 if (linux_args->len > STACK_SIZE - GUARD_SIZE)
  684                         bsd_args.len = linux_args->len;
  685                 else
  686                         bsd_args.len  = STACK_SIZE - GUARD_SIZE;
  687 
  688                 /* This gives us a new BOS.  If we're using VM_STACK, then
  689                  * mmap will just map the top SGROWSIZ bytes, and let
  690                  * the stack grow down to the limit at BOS.  If we're
  691                  * not using VM_STACK we map the full stack, since we
  692                  * don't have a way to autogrow it.
  693                  */
  694                 bsd_args.addr -= bsd_args.len;
  695         } else {
  696                 bsd_args.addr = (caddr_t)PTRIN(linux_args->addr);
  697                 bsd_args.len  = linux_args->len;
  698         }
  699         /*
  700          * XXX i386 Linux always emulator forces PROT_READ on (why?)
  701          * so we do the same. We add PROT_EXEC to work around buggy
  702          * applications (e.g. Java) that take advantage of the fact
  703          * that execute permissions are not enforced by x86 CPUs.
  704          */
  705         bsd_args.prot = linux_args->prot | PROT_EXEC | PROT_READ;
  706         if (linux_args->flags & LINUX_MAP_ANON)
  707                 bsd_args.fd = -1;
  708         else
  709                 bsd_args.fd = linux_args->fd;
  710         bsd_args.pos = (off_t)linux_args->pgoff * PAGE_SIZE;
  711         bsd_args.pad = 0;
  712 
  713 #ifdef DEBUG
  714         if (ldebug(mmap))
  715                 printf("-> %s(%p, %d, %d, 0x%08x, %d, 0x%x)\n",
  716                     __func__,
  717                     (void *)bsd_args.addr, (int)bsd_args.len, bsd_args.prot,
  718                     bsd_args.flags, bsd_args.fd, (int)bsd_args.pos);
  719 #endif
  720         error = mmap(td, &bsd_args);
  721 #ifdef DEBUG
  722         if (ldebug(mmap))
  723                 printf("-> %s() return: 0x%x (0x%08x)\n",
  724                         __func__, error, (u_int)td->td_retval[0]);
  725 #endif
  726         return (error);
  727 }
  728 
  729 int
  730 linux_pipe(struct thread *td, struct linux_pipe_args *args)
  731 {
  732         int pip[2];
  733         int error;
  734         register_t reg_rdx;
  735 
  736 #ifdef DEBUG
  737         if (ldebug(pipe))
  738                 printf(ARGS(pipe, "*"));
  739 #endif
  740 
  741         reg_rdx = td->td_retval[1];
  742         error = pipe(td, 0);
  743         if (error) {
  744                 td->td_retval[1] = reg_rdx;
  745                 return (error);
  746         }
  747 
  748         pip[0] = td->td_retval[0];
  749         pip[1] = td->td_retval[1];
  750         error = copyout(pip, args->pipefds, 2 * sizeof(int));
  751         if (error) {
  752                 td->td_retval[1] = reg_rdx;
  753                 return (error);
  754         }
  755 
  756         td->td_retval[1] = reg_rdx;
  757         td->td_retval[0] = 0;
  758         return (0);
  759 }
  760 
  761 int
  762 linux_sigaction(struct thread *td, struct linux_sigaction_args *args)
  763 {
  764         l_osigaction_t osa;
  765         l_sigaction_t act, oact;
  766         int error;
  767 
  768 #ifdef DEBUG
  769         if (ldebug(sigaction))
  770                 printf(ARGS(sigaction, "%d, %p, %p"),
  771                     args->sig, (void *)args->nsa, (void *)args->osa);
  772 #endif
  773 
  774         if (args->nsa != NULL) {
  775                 error = copyin(args->nsa, &osa, sizeof(l_osigaction_t));
  776                 if (error)
  777                         return (error);
  778                 act.lsa_handler = osa.lsa_handler;
  779                 act.lsa_flags = osa.lsa_flags;
  780                 act.lsa_restorer = osa.lsa_restorer;
  781                 LINUX_SIGEMPTYSET(act.lsa_mask);
  782                 act.lsa_mask.__bits[0] = osa.lsa_mask;
  783         }
  784 
  785         error = linux_do_sigaction(td, args->sig, args->nsa ? &act : NULL,
  786             args->osa ? &oact : NULL);
  787 
  788         if (args->osa != NULL && !error) {
  789                 osa.lsa_handler = oact.lsa_handler;
  790                 osa.lsa_flags = oact.lsa_flags;
  791                 osa.lsa_restorer = oact.lsa_restorer;
  792                 osa.lsa_mask = oact.lsa_mask.__bits[0];
  793                 error = copyout(&osa, args->osa, sizeof(l_osigaction_t));
  794         }
  795 
  796         return (error);
  797 }
  798 
  799 /*
  800  * Linux has two extra args, restart and oldmask.  We dont use these,
  801  * but it seems that "restart" is actually a context pointer that
  802  * enables the signal to happen with a different register set.
  803  */
  804 int
  805 linux_sigsuspend(struct thread *td, struct linux_sigsuspend_args *args)
  806 {
  807         sigset_t sigmask;
  808         l_sigset_t mask;
  809 
  810 #ifdef DEBUG
  811         if (ldebug(sigsuspend))
  812                 printf(ARGS(sigsuspend, "%08lx"), (unsigned long)args->mask);
  813 #endif
  814 
  815         LINUX_SIGEMPTYSET(mask);
  816         mask.__bits[0] = args->mask;
  817         linux_to_bsd_sigset(&mask, &sigmask);
  818         return (kern_sigsuspend(td, sigmask));
  819 }
  820 
  821 int
  822 linux_rt_sigsuspend(struct thread *td, struct linux_rt_sigsuspend_args *uap)
  823 {
  824         l_sigset_t lmask;
  825         sigset_t sigmask;
  826         int error;
  827 
  828 #ifdef DEBUG
  829         if (ldebug(rt_sigsuspend))
  830                 printf(ARGS(rt_sigsuspend, "%p, %d"),
  831                     (void *)uap->newset, uap->sigsetsize);
  832 #endif
  833 
  834         if (uap->sigsetsize != sizeof(l_sigset_t))
  835                 return (EINVAL);
  836 
  837         error = copyin(uap->newset, &lmask, sizeof(l_sigset_t));
  838         if (error)
  839                 return (error);
  840 
  841         linux_to_bsd_sigset(&lmask, &sigmask);
  842         return (kern_sigsuspend(td, sigmask));
  843 }
  844 
  845 int
  846 linux_pause(struct thread *td, struct linux_pause_args *args)
  847 {
  848         struct proc *p = td->td_proc;
  849         sigset_t sigmask;
  850 
  851 #ifdef DEBUG
  852         if (ldebug(pause))
  853                 printf(ARGS(pause, ""));
  854 #endif
  855 
  856         PROC_LOCK(p);
  857         sigmask = td->td_sigmask;
  858         PROC_UNLOCK(p);
  859         return (kern_sigsuspend(td, sigmask));
  860 }
  861 
  862 int
  863 linux_sigaltstack(struct thread *td, struct linux_sigaltstack_args *uap)
  864 {
  865         stack_t ss, oss;
  866         l_stack_t lss;
  867         int error;
  868 
  869 #ifdef DEBUG
  870         if (ldebug(sigaltstack))
  871                 printf(ARGS(sigaltstack, "%p, %p"), uap->uss, uap->uoss);
  872 #endif
  873 
  874         if (uap->uss != NULL) {
  875                 error = copyin(uap->uss, &lss, sizeof(l_stack_t));
  876                 if (error)
  877                         return (error);
  878 
  879                 ss.ss_sp = PTRIN(lss.ss_sp);
  880                 ss.ss_size = lss.ss_size;
  881                 ss.ss_flags = linux_to_bsd_sigaltstack(lss.ss_flags);
  882         }
  883         error = kern_sigaltstack(td, (uap->uss != NULL) ? &ss : NULL,
  884             (uap->uoss != NULL) ? &oss : NULL);
  885         if (!error && uap->uoss != NULL) {
  886                 lss.ss_sp = PTROUT(oss.ss_sp);
  887                 lss.ss_size = oss.ss_size;
  888                 lss.ss_flags = bsd_to_linux_sigaltstack(oss.ss_flags);
  889                 error = copyout(&lss, uap->uoss, sizeof(l_stack_t));
  890         }
  891 
  892         return (error);
  893 }
  894 
  895 int
  896 linux_ftruncate64(struct thread *td, struct linux_ftruncate64_args *args)
  897 {
  898         struct ftruncate_args sa;
  899 
  900 #ifdef DEBUG
  901         if (ldebug(ftruncate64))
  902                 printf(ARGS(ftruncate64, "%u, %jd"), args->fd,
  903                     (intmax_t)args->length);
  904 #endif
  905 
  906         sa.fd = args->fd;
  907         sa.pad = 0;
  908         sa.length = args->length;
  909         return ftruncate(td, &sa);
  910 }
  911 
  912 int
  913 linux_gettimeofday(struct thread *td, struct linux_gettimeofday_args *uap)
  914 {
  915         struct timeval atv;
  916         l_timeval atv32;
  917         struct timezone rtz;
  918         int error = 0;
  919 
  920         if (uap->tp) {
  921                 microtime(&atv);
  922                 atv32.tv_sec = atv.tv_sec;
  923                 atv32.tv_usec = atv.tv_usec;
  924                 error = copyout(&atv32, uap->tp, sizeof (atv32));
  925         }
  926         if (error == 0 && uap->tzp != NULL) {
  927                 rtz.tz_minuteswest = tz_minuteswest;
  928                 rtz.tz_dsttime = tz_dsttime;
  929                 error = copyout(&rtz, uap->tzp, sizeof (rtz));
  930         }
  931         return (error);
  932 }
  933 
  934 int
  935 linux_nanosleep(struct thread *td, struct linux_nanosleep_args *uap)
  936 {
  937         struct timespec rqt, rmt;
  938         struct l_timespec ats32;
  939         int error;
  940 
  941         error = copyin(uap->rqtp, &ats32, sizeof(ats32));
  942         if (error != 0)
  943                 return (error);
  944         rqt.tv_sec = ats32.tv_sec;
  945         rqt.tv_nsec = ats32.tv_nsec;
  946         error = kern_nanosleep(td, &rqt, &rmt);
  947         if (uap->rmtp != NULL) {
  948                 ats32.tv_sec = rmt.tv_sec;
  949                 ats32.tv_nsec = rmt.tv_nsec;
  950                 error = copyout(&ats32, uap->rmtp, sizeof(ats32));
  951         }
  952         return (error);
  953 }
  954 
  955 int
  956 linux_getrusage(struct thread *td, struct linux_getrusage_args *uap)
  957 {
  958         struct l_rusage s32;
  959         struct rusage s;
  960         int error;
  961 
  962         error = kern_getrusage(td, uap->who, &s);
  963         if (error != 0)
  964                 return (error);
  965         if (uap->rusage != NULL) {
  966                 s32.ru_utime.tv_sec = s.ru_utime.tv_sec;
  967                 s32.ru_utime.tv_usec = s.ru_utime.tv_usec;
  968                 s32.ru_stime.tv_sec = s.ru_stime.tv_sec;
  969                 s32.ru_stime.tv_usec = s.ru_stime.tv_usec;
  970                 s32.ru_maxrss = s.ru_maxrss;
  971                 s32.ru_ixrss = s.ru_ixrss;
  972                 s32.ru_idrss = s.ru_idrss;
  973                 s32.ru_isrss = s.ru_isrss;
  974                 s32.ru_minflt = s.ru_minflt;
  975                 s32.ru_majflt = s.ru_majflt;
  976                 s32.ru_nswap = s.ru_nswap;
  977                 s32.ru_inblock = s.ru_inblock;
  978                 s32.ru_oublock = s.ru_oublock;
  979                 s32.ru_msgsnd = s.ru_msgsnd;
  980                 s32.ru_msgrcv = s.ru_msgrcv;
  981                 s32.ru_nsignals = s.ru_nsignals;
  982                 s32.ru_nvcsw = s.ru_nvcsw;
  983                 s32.ru_nivcsw = s.ru_nivcsw;
  984                 error = copyout(&s32, uap->rusage, sizeof(s32));
  985         }
  986         return (error);
  987 }
  988 
  989 int
  990 linux_sched_rr_get_interval(struct thread *td,
  991     struct linux_sched_rr_get_interval_args *uap)
  992 {
  993         struct timespec ts;
  994         struct l_timespec ts32;
  995         int error;
  996 
  997         error = kern_sched_rr_get_interval(td, uap->pid, &ts);
  998         if (error != 0)
  999                 return (error);
 1000         ts32.tv_sec = ts.tv_sec;
 1001         ts32.tv_nsec = ts.tv_nsec;
 1002         return (copyout(&ts32, uap->interval, sizeof(ts32)));
 1003 }
 1004 
 1005 int
 1006 linux_mprotect(struct thread *td, struct linux_mprotect_args *uap)
 1007 {
 1008         struct mprotect_args bsd_args;
 1009 
 1010         bsd_args.addr = uap->addr;
 1011         bsd_args.len = uap->len;
 1012         bsd_args.prot = uap->prot;
 1013         /* XXX PROT_READ implies PROT_EXEC; see linux_mmap_common(). */
 1014         if ((bsd_args.prot & PROT_READ) != 0)
 1015                 bsd_args.prot |= PROT_EXEC;
 1016         return (mprotect(td, &bsd_args));
 1017 }
Cache object: ca473c26795277a4c55c97a05caa5aea
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]
This page is part of the FreeBSD/Linux Linux Kernel Cross-Reference, and was automatically generated using a modified version of the LXR engine.
FreeBSD/Linux Kernel Cross Reference sys/amd64/linux32/linux32_machdep.c

FreeBSD/Linux Kernel Cross Reference
sys/amd64/linux32/linux32_machdep.c
