The Design and Implementation of the FreeBSD Operating System, Second Edition
Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]

FreeBSD/Linux Kernel Cross Reference
sys/amd64/linux32/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/10.3/sys/amd64/linux32/linux32_machdep.c 293600 2016-01-09 18:07:48Z dchagin $");
   33 
   34 #include "opt_compat.h"
   35 
   36 #include <sys/param.h>
   37 #include <sys/kernel.h>
   38 #include <sys/systm.h>
   39 #include <sys/capsicum.h>
   40 #include <sys/file.h>
   41 #include <sys/fcntl.h>
   42 #include <sys/clock.h>
   43 #include <sys/imgact.h>
   44 #include <sys/limits.h>
   45 #include <sys/lock.h>
   46 #include <sys/malloc.h>
   47 #include <sys/mman.h>
   48 #include <sys/mutex.h>
   49 #include <sys/priv.h>
   50 #include <sys/proc.h>
   51 #include <sys/resource.h>
   52 #include <sys/resourcevar.h>
   53 #include <sys/syscallsubr.h>
   54 #include <sys/sysproto.h>
   55 #include <sys/unistd.h>
   56 #include <sys/wait.h>
   57 
   58 #include <machine/frame.h>
   59 #include <machine/pcb.h>
   60 #include <machine/psl.h>
   61 #include <machine/segments.h>
   62 #include <machine/specialreg.h>
   63 
   64 #include <vm/vm.h>
   65 #include <vm/pmap.h>
   66 #include <vm/vm_map.h>
   67 
   68 #include <compat/freebsd32/freebsd32_util.h>
   69 #include <amd64/linux32/linux.h>
   70 #include <amd64/linux32/linux32_proto.h>
   71 #include <compat/linux/linux_ipc.h>
   72 #include <compat/linux/linux_misc.h>
   73 #include <compat/linux/linux_signal.h>
   74 #include <compat/linux/linux_util.h>
   75 #include <compat/linux/linux_emul.h>
   76 
   77 static void     bsd_to_linux_rusage(struct rusage *ru, struct l_rusage *lru);
   78 
   79 struct l_old_select_argv {
   80         l_int           nfds;
   81         l_uintptr_t     readfds;
   82         l_uintptr_t     writefds;
   83         l_uintptr_t     exceptfds;
   84         l_uintptr_t     timeout;
   85 } __packed;
   86 
   87 static int      linux_mmap_common(struct thread *td, l_uintptr_t addr,
   88                     l_size_t len, l_int prot, l_int flags, l_int fd,
   89                     l_loff_t pos);
   90 
   91 static void
   92 bsd_to_linux_rusage(struct rusage *ru, struct l_rusage *lru)
   93 {
   94 
   95         lru->ru_utime.tv_sec = ru->ru_utime.tv_sec;
   96         lru->ru_utime.tv_usec = ru->ru_utime.tv_usec;
   97         lru->ru_stime.tv_sec = ru->ru_stime.tv_sec;
   98         lru->ru_stime.tv_usec = ru->ru_stime.tv_usec;
   99         lru->ru_maxrss = ru->ru_maxrss;
  100         lru->ru_ixrss = ru->ru_ixrss;
  101         lru->ru_idrss = ru->ru_idrss;
  102         lru->ru_isrss = ru->ru_isrss;
  103         lru->ru_minflt = ru->ru_minflt;
  104         lru->ru_majflt = ru->ru_majflt;
  105         lru->ru_nswap = ru->ru_nswap;
  106         lru->ru_inblock = ru->ru_inblock;
  107         lru->ru_oublock = ru->ru_oublock;
  108         lru->ru_msgsnd = ru->ru_msgsnd;
  109         lru->ru_msgrcv = ru->ru_msgrcv;
  110         lru->ru_nsignals = ru->ru_nsignals;
  111         lru->ru_nvcsw = ru->ru_nvcsw;
  112         lru->ru_nivcsw = ru->ru_nivcsw;
  113 }
  114 
  115 int
  116 linux_copyout_rusage(struct rusage *ru, void *uaddr)
  117 {
  118         struct l_rusage lru;
  119 
  120         bsd_to_linux_rusage(ru, &lru);
  121 
  122         return (copyout(&lru, uaddr, sizeof(struct l_rusage)));
  123 }
  124 
  125 int
  126 linux_execve(struct thread *td, struct linux_execve_args *args)
  127 {
  128         struct image_args eargs;
  129         char *path;
  130         int error;
  131 
  132         LCONVPATHEXIST(td, args->path, &path);
  133 
  134 #ifdef DEBUG
  135         if (ldebug(execve))
  136                 printf(ARGS(execve, "%s"), path);
  137 #endif
  138 
  139         error = freebsd32_exec_copyin_args(&eargs, path, UIO_SYSSPACE,
  140             args->argp, args->envp);
  141         free(path, M_TEMP);
  142         if (error == 0)
  143                 error = linux_common_execve(td, &eargs);
  144         return (error);
  145 }
  146 
  147 CTASSERT(sizeof(struct l_iovec32) == 8);
  148 
  149 static int
  150 linux32_copyinuio(struct l_iovec32 *iovp, l_ulong iovcnt, struct uio **uiop)
  151 {
  152         struct l_iovec32 iov32;
  153         struct iovec *iov;
  154         struct uio *uio;
  155         uint32_t iovlen;
  156         int error, i;
  157 
  158         *uiop = NULL;
  159         if (iovcnt > UIO_MAXIOV)
  160                 return (EINVAL);
  161         iovlen = iovcnt * sizeof(struct iovec);
  162         uio = malloc(iovlen + sizeof(*uio), M_IOV, M_WAITOK);
  163         iov = (struct iovec *)(uio + 1);
  164         for (i = 0; i < iovcnt; i++) {
  165                 error = copyin(&iovp[i], &iov32, sizeof(struct l_iovec32));
  166                 if (error) {
  167                         free(uio, M_IOV);
  168                         return (error);
  169                 }
  170                 iov[i].iov_base = PTRIN(iov32.iov_base);
  171                 iov[i].iov_len = iov32.iov_len;
  172         }
  173         uio->uio_iov = iov;
  174         uio->uio_iovcnt = iovcnt;
  175         uio->uio_segflg = UIO_USERSPACE;
  176         uio->uio_offset = -1;
  177         uio->uio_resid = 0;
  178         for (i = 0; i < iovcnt; i++) {
  179                 if (iov->iov_len > INT_MAX - uio->uio_resid) {
  180                         free(uio, M_IOV);
  181                         return (EINVAL);
  182                 }
  183                 uio->uio_resid += iov->iov_len;
  184                 iov++;
  185         }
  186         *uiop = uio;
  187         return (0);
  188 }
  189 
  190 int
  191 linux32_copyiniov(struct l_iovec32 *iovp32, l_ulong iovcnt, struct iovec **iovp,
  192     int error)
  193 {
  194         struct l_iovec32 iov32;
  195         struct iovec *iov;
  196         uint32_t iovlen;
  197         int i;
  198 
  199         *iovp = NULL;
  200         if (iovcnt > UIO_MAXIOV)
  201                 return (error);
  202         iovlen = iovcnt * sizeof(struct iovec);
  203         iov = malloc(iovlen, M_IOV, M_WAITOK);
  204         for (i = 0; i < iovcnt; i++) {
  205                 error = copyin(&iovp32[i], &iov32, sizeof(struct l_iovec32));
  206                 if (error) {
  207                         free(iov, M_IOV);
  208                         return (error);
  209                 }
  210                 iov[i].iov_base = PTRIN(iov32.iov_base);
  211                 iov[i].iov_len = iov32.iov_len;
  212         }
  213         *iovp = iov;
  214         return(0);
  215 
  216 }
  217 
  218 int
  219 linux_readv(struct thread *td, struct linux_readv_args *uap)
  220 {
  221         struct uio *auio;
  222         int error;
  223 
  224         error = linux32_copyinuio(uap->iovp, uap->iovcnt, &auio);
  225         if (error)
  226                 return (error);
  227         error = kern_readv(td, uap->fd, auio);
  228         free(auio, M_IOV);
  229         return (error);
  230 }
  231 
  232 int
  233 linux_writev(struct thread *td, struct linux_writev_args *uap)
  234 {
  235         struct uio *auio;
  236         int error;
  237 
  238         error = linux32_copyinuio(uap->iovp, uap->iovcnt, &auio);
  239         if (error)
  240                 return (error);
  241         error = kern_writev(td, uap->fd, auio);
  242         free(auio, M_IOV);
  243         return (error);
  244 }
  245 
  246 struct l_ipc_kludge {
  247         l_uintptr_t msgp;
  248         l_long msgtyp;
  249 } __packed;
  250 
  251 int
  252 linux_ipc(struct thread *td, struct linux_ipc_args *args)
  253 {
  254 
  255         switch (args->what & 0xFFFF) {
  256         case LINUX_SEMOP: {
  257                 struct linux_semop_args a;
  258 
  259                 a.semid = args->arg1;
  260                 a.tsops = args->ptr;
  261                 a.nsops = args->arg2;
  262                 return (linux_semop(td, &a));
  263         }
  264         case LINUX_SEMGET: {
  265                 struct linux_semget_args a;
  266 
  267                 a.key = args->arg1;
  268                 a.nsems = args->arg2;
  269                 a.semflg = args->arg3;
  270                 return (linux_semget(td, &a));
  271         }
  272         case LINUX_SEMCTL: {
  273                 struct linux_semctl_args a;
  274                 int error;
  275 
  276                 a.semid = args->arg1;
  277                 a.semnum = args->arg2;
  278                 a.cmd = args->arg3;
  279                 error = copyin(args->ptr, &a.arg, sizeof(a.arg));
  280                 if (error)
  281                         return (error);
  282                 return (linux_semctl(td, &a));
  283         }
  284         case LINUX_MSGSND: {
  285                 struct linux_msgsnd_args a;
  286 
  287                 a.msqid = args->arg1;
  288                 a.msgp = args->ptr;
  289                 a.msgsz = args->arg2;
  290                 a.msgflg = args->arg3;
  291                 return (linux_msgsnd(td, &a));
  292         }
  293         case LINUX_MSGRCV: {
  294                 struct linux_msgrcv_args a;
  295 
  296                 a.msqid = args->arg1;
  297                 a.msgsz = args->arg2;
  298                 a.msgflg = args->arg3;
  299                 if ((args->what >> 16) == 0) {
  300                         struct l_ipc_kludge tmp;
  301                         int error;
  302 
  303                         if (args->ptr == 0)
  304                                 return (EINVAL);
  305                         error = copyin(args->ptr, &tmp, sizeof(tmp));
  306                         if (error)
  307                                 return (error);
  308                         a.msgp = PTRIN(tmp.msgp);
  309                         a.msgtyp = tmp.msgtyp;
  310                 } else {
  311                         a.msgp = args->ptr;
  312                         a.msgtyp = args->arg5;
  313                 }
  314                 return (linux_msgrcv(td, &a));
  315         }
  316         case LINUX_MSGGET: {
  317                 struct linux_msgget_args a;
  318 
  319                 a.key = args->arg1;
  320                 a.msgflg = args->arg2;
  321                 return (linux_msgget(td, &a));
  322         }
  323         case LINUX_MSGCTL: {
  324                 struct linux_msgctl_args a;
  325 
  326                 a.msqid = args->arg1;
  327                 a.cmd = args->arg2;
  328                 a.buf = args->ptr;
  329                 return (linux_msgctl(td, &a));
  330         }
  331         case LINUX_SHMAT: {
  332                 struct linux_shmat_args a;
  333 
  334                 a.shmid = args->arg1;
  335                 a.shmaddr = args->ptr;
  336                 a.shmflg = args->arg2;
  337                 a.raddr = PTRIN((l_uint)args->arg3);
  338                 return (linux_shmat(td, &a));
  339         }
  340         case LINUX_SHMDT: {
  341                 struct linux_shmdt_args a;
  342 
  343                 a.shmaddr = args->ptr;
  344                 return (linux_shmdt(td, &a));
  345         }
  346         case LINUX_SHMGET: {
  347                 struct linux_shmget_args a;
  348 
  349                 a.key = args->arg1;
  350                 a.size = args->arg2;
  351                 a.shmflg = args->arg3;
  352                 return (linux_shmget(td, &a));
  353         }
  354         case LINUX_SHMCTL: {
  355                 struct linux_shmctl_args a;
  356 
  357                 a.shmid = args->arg1;
  358                 a.cmd = args->arg2;
  359                 a.buf = args->ptr;
  360                 return (linux_shmctl(td, &a));
  361         }
  362         default:
  363                 break;
  364         }
  365 
  366         return (EINVAL);
  367 }
  368 
  369 int
  370 linux_old_select(struct thread *td, struct linux_old_select_args *args)
  371 {
  372         struct l_old_select_argv linux_args;
  373         struct linux_select_args newsel;
  374         int error;
  375 
  376 #ifdef DEBUG
  377         if (ldebug(old_select))
  378                 printf(ARGS(old_select, "%p"), args->ptr);
  379 #endif
  380 
  381         error = copyin(args->ptr, &linux_args, sizeof(linux_args));
  382         if (error)
  383                 return (error);
  384 
  385         newsel.nfds = linux_args.nfds;
  386         newsel.readfds = PTRIN(linux_args.readfds);
  387         newsel.writefds = PTRIN(linux_args.writefds);
  388         newsel.exceptfds = PTRIN(linux_args.exceptfds);
  389         newsel.timeout = PTRIN(linux_args.timeout);
  390         return (linux_select(td, &newsel));
  391 }
  392 
  393 int
  394 linux_set_cloned_tls(struct thread *td, void *desc)
  395 {
  396         struct user_segment_descriptor sd;
  397         struct l_user_desc info;
  398         struct pcb *pcb;
  399         int error;
  400         int a[2];
  401 
  402         error = copyin(desc, &info, sizeof(struct l_user_desc));
  403         if (error) {
  404                 printf(LMSG("copyin failed!"));
  405         } else {
  406                 /* We might copy out the entry_number as GUGS32_SEL. */
  407                 info.entry_number = GUGS32_SEL;
  408                 error = copyout(&info, desc, sizeof(struct l_user_desc));
  409                 if (error)
  410                         printf(LMSG("copyout failed!"));
  411 
  412                 a[0] = LINUX_LDT_entry_a(&info);
  413                 a[1] = LINUX_LDT_entry_b(&info);
  414 
  415                 memcpy(&sd, &a, sizeof(a));
  416 #ifdef DEBUG
  417                 if (ldebug(clone))
  418                         printf("Segment created in clone with "
  419                             "CLONE_SETTLS: lobase: %x, hibase: %x, "
  420                             "lolimit: %x, hilimit: %x, type: %i, "
  421                             "dpl: %i, p: %i, xx: %i, long: %i, "
  422                             "def32: %i, gran: %i\n", sd.sd_lobase,
  423                             sd.sd_hibase, sd.sd_lolimit, sd.sd_hilimit,
  424                             sd.sd_type, sd.sd_dpl, sd.sd_p, sd.sd_xx,
  425                             sd.sd_long, sd.sd_def32, sd.sd_gran);
  426 #endif
  427                 pcb = td->td_pcb;
  428                 pcb->pcb_gsbase = (register_t)info.base_addr;
  429                 td->td_frame->tf_gs = GSEL(GUGS32_SEL, SEL_UPL);
  430                 set_pcb_flags(pcb, PCB_32BIT);
  431         }
  432 
  433         return (error);
  434 }
  435 
  436 int
  437 linux_set_upcall_kse(struct thread *td, register_t stack)
  438 {
  439 
  440         if (stack)
  441                 td->td_frame->tf_rsp = stack;
  442 
  443         /*
  444          * The newly created Linux thread returns
  445          * to the user space by the same path that a parent do.
  446          */
  447         td->td_frame->tf_rax = 0;
  448         return (0);
  449 }
  450 
  451 #define STACK_SIZE  (2 * 1024 * 1024)
  452 #define GUARD_SIZE  (4 * PAGE_SIZE)
  453 
  454 int
  455 linux_mmap2(struct thread *td, struct linux_mmap2_args *args)
  456 {
  457 
  458 #ifdef DEBUG
  459         if (ldebug(mmap2))
  460                 printf(ARGS(mmap2, "0x%08x, %d, %d, 0x%08x, %d, %d"),
  461                     args->addr, args->len, args->prot,
  462                     args->flags, args->fd, args->pgoff);
  463 #endif
  464 
  465         return (linux_mmap_common(td, PTROUT(args->addr), args->len, args->prot,
  466                 args->flags, args->fd, (uint64_t)(uint32_t)args->pgoff *
  467                 PAGE_SIZE));
  468 }
  469 
  470 int
  471 linux_mmap(struct thread *td, struct linux_mmap_args *args)
  472 {
  473         int error;
  474         struct l_mmap_argv linux_args;
  475 
  476         error = copyin(args->ptr, &linux_args, sizeof(linux_args));
  477         if (error)
  478                 return (error);
  479 
  480 #ifdef DEBUG
  481         if (ldebug(mmap))
  482                 printf(ARGS(mmap, "0x%08x, %d, %d, 0x%08x, %d, %d"),
  483                     linux_args.addr, linux_args.len, linux_args.prot,
  484                     linux_args.flags, linux_args.fd, linux_args.pgoff);
  485 #endif
  486 
  487         return (linux_mmap_common(td, linux_args.addr, linux_args.len,
  488             linux_args.prot, linux_args.flags, linux_args.fd,
  489             (uint32_t)linux_args.pgoff));
  490 }
  491 
  492 static int
  493 linux_mmap_common(struct thread *td, l_uintptr_t addr, l_size_t len, l_int prot,
  494     l_int flags, l_int fd, l_loff_t pos)
  495 {
  496         struct proc *p = td->td_proc;
  497         struct mmap_args /* {
  498                 caddr_t addr;
  499                 size_t len;
  500                 int prot;
  501                 int flags;
  502                 int fd;
  503                 long pad;
  504                 off_t pos;
  505         } */ bsd_args;
  506         int error;
  507         struct file *fp;
  508         cap_rights_t rights;
  509 
  510         error = 0;
  511         bsd_args.flags = 0;
  512         fp = NULL;
  513 
  514         /*
  515          * Linux mmap(2):
  516          * You must specify exactly one of MAP_SHARED and MAP_PRIVATE
  517          */
  518         if (!((flags & LINUX_MAP_SHARED) ^ (flags & LINUX_MAP_PRIVATE)))
  519                 return (EINVAL);
  520 
  521         if (flags & LINUX_MAP_SHARED)
  522                 bsd_args.flags |= MAP_SHARED;
  523         if (flags & LINUX_MAP_PRIVATE)
  524                 bsd_args.flags |= MAP_PRIVATE;
  525         if (flags & LINUX_MAP_FIXED)
  526                 bsd_args.flags |= MAP_FIXED;
  527         if (flags & LINUX_MAP_ANON) {
  528                 /* Enforce pos to be on page boundary, then ignore. */
  529                 if ((pos & PAGE_MASK) != 0)
  530                         return (EINVAL);
  531                 pos = 0;
  532                 bsd_args.flags |= MAP_ANON;
  533         } else
  534                 bsd_args.flags |= MAP_NOSYNC;
  535         if (flags & LINUX_MAP_GROWSDOWN)
  536                 bsd_args.flags |= MAP_STACK;
  537 
  538         /*
  539          * PROT_READ, PROT_WRITE, or PROT_EXEC implies PROT_READ and PROT_EXEC
  540          * on Linux/i386. We do this to ensure maximum compatibility.
  541          * Linux/ia64 does the same in i386 emulation mode.
  542          */
  543         bsd_args.prot = prot;
  544         if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC))
  545                 bsd_args.prot |= PROT_READ | PROT_EXEC;
  546 
  547         /* Linux does not check file descriptor when MAP_ANONYMOUS is set. */
  548         bsd_args.fd = (bsd_args.flags & MAP_ANON) ? -1 : fd;
  549         if (bsd_args.fd != -1) {
  550                 /*
  551                  * Linux follows Solaris mmap(2) description:
  552                  * The file descriptor fildes is opened with
  553                  * read permission, regardless of the
  554                  * protection options specified.
  555                  */
  556 
  557                 error = fget(td, bsd_args.fd,
  558                     cap_rights_init(&rights, CAP_MMAP), &fp);
  559                 if (error != 0)
  560                         return (error);
  561                 if (fp->f_type != DTYPE_VNODE) {
  562                         fdrop(fp, td);
  563                         return (EINVAL);
  564                 }
  565 
  566                 /* Linux mmap() just fails for O_WRONLY files */
  567                 if (!(fp->f_flag & FREAD)) {
  568                         fdrop(fp, td);
  569                         return (EACCES);
  570                 }
  571 
  572                 fdrop(fp, td);
  573         }
  574 
  575         if (flags & LINUX_MAP_GROWSDOWN) {
  576                 /*
  577                  * The Linux MAP_GROWSDOWN option does not limit auto
  578                  * growth of the region.  Linux mmap with this option
  579                  * takes as addr the inital BOS, and as len, the initial
  580                  * region size.  It can then grow down from addr without
  581                  * limit.  However, Linux threads has an implicit internal
  582                  * limit to stack size of STACK_SIZE.  Its just not
  583                  * enforced explicitly in Linux.  But, here we impose
  584                  * a limit of (STACK_SIZE - GUARD_SIZE) on the stack
  585                  * region, since we can do this with our mmap.
  586                  *
  587                  * Our mmap with MAP_STACK takes addr as the maximum
  588                  * downsize limit on BOS, and as len the max size of
  589                  * the region.  It then maps the top SGROWSIZ bytes,
  590                  * and auto grows the region down, up to the limit
  591                  * in addr.
  592                  *
  593                  * If we don't use the MAP_STACK option, the effect
  594                  * of this code is to allocate a stack region of a
  595                  * fixed size of (STACK_SIZE - GUARD_SIZE).
  596                  */
  597 
  598                 if ((caddr_t)PTRIN(addr) + len > p->p_vmspace->vm_maxsaddr) {
  599                         /*
  600                          * Some Linux apps will attempt to mmap
  601                          * thread stacks near the top of their
  602                          * address space.  If their TOS is greater
  603                          * than vm_maxsaddr, vm_map_growstack()
  604                          * will confuse the thread stack with the
  605                          * process stack and deliver a SEGV if they
  606                          * attempt to grow the thread stack past their
  607                          * current stacksize rlimit.  To avoid this,
  608                          * adjust vm_maxsaddr upwards to reflect
  609                          * the current stacksize rlimit rather
  610                          * than the maximum possible stacksize.
  611                          * It would be better to adjust the
  612                          * mmap'ed region, but some apps do not check
  613                          * mmap's return value.
  614                          */
  615                         PROC_LOCK(p);
  616                         p->p_vmspace->vm_maxsaddr = (char *)LINUX32_USRSTACK -
  617                             lim_cur(p, RLIMIT_STACK);
  618                         PROC_UNLOCK(p);
  619                 }
  620 
  621                 /*
  622                  * This gives us our maximum stack size and a new BOS.
  623                  * If we're using VM_STACK, then mmap will just map
  624                  * the top SGROWSIZ bytes, and let the stack grow down
  625                  * to the limit at BOS.  If we're not using VM_STACK
  626                  * we map the full stack, since we don't have a way
  627                  * to autogrow it.
  628                  */
  629                 if (len > STACK_SIZE - GUARD_SIZE) {
  630                         bsd_args.addr = (caddr_t)PTRIN(addr);
  631                         bsd_args.len = len;
  632                 } else {
  633                         bsd_args.addr = (caddr_t)PTRIN(addr) -
  634                             (STACK_SIZE - GUARD_SIZE - len);
  635                         bsd_args.len = STACK_SIZE - GUARD_SIZE;
  636                 }
  637         } else {
  638                 bsd_args.addr = (caddr_t)PTRIN(addr);
  639                 bsd_args.len  = len;
  640         }
  641         bsd_args.pos = pos;
  642 
  643 #ifdef DEBUG
  644         if (ldebug(mmap))
  645                 printf("-> %s(%p, %d, %d, 0x%08x, %d, 0x%x)\n",
  646                     __func__,
  647                     (void *)bsd_args.addr, (int)bsd_args.len, bsd_args.prot,
  648                     bsd_args.flags, bsd_args.fd, (int)bsd_args.pos);
  649 #endif
  650         error = sys_mmap(td, &bsd_args);
  651 #ifdef DEBUG
  652         if (ldebug(mmap))
  653                 printf("-> %s() return: 0x%x (0x%08x)\n",
  654                         __func__, error, (u_int)td->td_retval[0]);
  655 #endif
  656         return (error);
  657 }
  658 
  659 int
  660 linux_mprotect(struct thread *td, struct linux_mprotect_args *uap)
  661 {
  662         struct mprotect_args bsd_args;
  663 
  664         bsd_args.addr = uap->addr;
  665         bsd_args.len = uap->len;
  666         bsd_args.prot = uap->prot;
  667         if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC))
  668                 bsd_args.prot |= PROT_READ | PROT_EXEC;
  669         return (sys_mprotect(td, &bsd_args));
  670 }
  671 
  672 int
  673 linux_iopl(struct thread *td, struct linux_iopl_args *args)
  674 {
  675         int error;
  676 
  677         if (args->level < 0 || args->level > 3)
  678                 return (EINVAL);
  679         if ((error = priv_check(td, PRIV_IO)) != 0)
  680                 return (error);
  681         if ((error = securelevel_gt(td->td_ucred, 0)) != 0)
  682                 return (error);
  683         td->td_frame->tf_rflags = (td->td_frame->tf_rflags & ~PSL_IOPL) |
  684             (args->level * (PSL_IOPL / 3));
  685 
  686         return (0);
  687 }
  688 
  689 int
  690 linux_sigaction(struct thread *td, struct linux_sigaction_args *args)
  691 {
  692         l_osigaction_t osa;
  693         l_sigaction_t act, oact;
  694         int error;
  695 
  696 #ifdef DEBUG
  697         if (ldebug(sigaction))
  698                 printf(ARGS(sigaction, "%d, %p, %p"),
  699                     args->sig, (void *)args->nsa, (void *)args->osa);
  700 #endif
  701 
  702         if (args->nsa != NULL) {
  703                 error = copyin(args->nsa, &osa, sizeof(l_osigaction_t));
  704                 if (error)
  705                         return (error);
  706                 act.lsa_handler = osa.lsa_handler;
  707                 act.lsa_flags = osa.lsa_flags;
  708                 act.lsa_restorer = osa.lsa_restorer;
  709                 LINUX_SIGEMPTYSET(act.lsa_mask);
  710                 act.lsa_mask.__mask = osa.lsa_mask;
  711         }
  712 
  713         error = linux_do_sigaction(td, args->sig, args->nsa ? &act : NULL,
  714             args->osa ? &oact : NULL);
  715 
  716         if (args->osa != NULL && !error) {
  717                 osa.lsa_handler = oact.lsa_handler;
  718                 osa.lsa_flags = oact.lsa_flags;
  719                 osa.lsa_restorer = oact.lsa_restorer;
  720                 osa.lsa_mask = oact.lsa_mask.__mask;
  721                 error = copyout(&osa, args->osa, sizeof(l_osigaction_t));
  722         }
  723 
  724         return (error);
  725 }
  726 
  727 /*
  728  * Linux has two extra args, restart and oldmask.  We don't use these,
  729  * but it seems that "restart" is actually a context pointer that
  730  * enables the signal to happen with a different register set.
  731  */
  732 int
  733 linux_sigsuspend(struct thread *td, struct linux_sigsuspend_args *args)
  734 {
  735         sigset_t sigmask;
  736         l_sigset_t mask;
  737 
  738 #ifdef DEBUG
  739         if (ldebug(sigsuspend))
  740                 printf(ARGS(sigsuspend, "%08lx"), (unsigned long)args->mask);
  741 #endif
  742 
  743         LINUX_SIGEMPTYSET(mask);
  744         mask.__mask = args->mask;
  745         linux_to_bsd_sigset(&mask, &sigmask);
  746         return (kern_sigsuspend(td, sigmask));
  747 }
  748 
  749 int
  750 linux_rt_sigsuspend(struct thread *td, struct linux_rt_sigsuspend_args *uap)
  751 {
  752         l_sigset_t lmask;
  753         sigset_t sigmask;
  754         int error;
  755 
  756 #ifdef DEBUG
  757         if (ldebug(rt_sigsuspend))
  758                 printf(ARGS(rt_sigsuspend, "%p, %d"),
  759                     (void *)uap->newset, uap->sigsetsize);
  760 #endif
  761 
  762         if (uap->sigsetsize != sizeof(l_sigset_t))
  763                 return (EINVAL);
  764 
  765         error = copyin(uap->newset, &lmask, sizeof(l_sigset_t));
  766         if (error)
  767                 return (error);
  768 
  769         linux_to_bsd_sigset(&lmask, &sigmask);
  770         return (kern_sigsuspend(td, sigmask));
  771 }
  772 
  773 int
  774 linux_pause(struct thread *td, struct linux_pause_args *args)
  775 {
  776         struct proc *p = td->td_proc;
  777         sigset_t sigmask;
  778 
  779 #ifdef DEBUG
  780         if (ldebug(pause))
  781                 printf(ARGS(pause, ""));
  782 #endif
  783 
  784         PROC_LOCK(p);
  785         sigmask = td->td_sigmask;
  786         PROC_UNLOCK(p);
  787         return (kern_sigsuspend(td, sigmask));
  788 }
  789 
  790 int
  791 linux_sigaltstack(struct thread *td, struct linux_sigaltstack_args *uap)
  792 {
  793         stack_t ss, oss;
  794         l_stack_t lss;
  795         int error;
  796 
  797 #ifdef DEBUG
  798         if (ldebug(sigaltstack))
  799                 printf(ARGS(sigaltstack, "%p, %p"), uap->uss, uap->uoss);
  800 #endif
  801 
  802         if (uap->uss != NULL) {
  803                 error = copyin(uap->uss, &lss, sizeof(l_stack_t));
  804                 if (error)
  805                         return (error);
  806 
  807                 ss.ss_sp = PTRIN(lss.ss_sp);
  808                 ss.ss_size = lss.ss_size;
  809                 ss.ss_flags = linux_to_bsd_sigaltstack(lss.ss_flags);
  810         }
  811         error = kern_sigaltstack(td, (uap->uss != NULL) ? &ss : NULL,
  812             (uap->uoss != NULL) ? &oss : NULL);
  813         if (!error && uap->uoss != NULL) {
  814                 lss.ss_sp = PTROUT(oss.ss_sp);
  815                 lss.ss_size = oss.ss_size;
  816                 lss.ss_flags = bsd_to_linux_sigaltstack(oss.ss_flags);
  817                 error = copyout(&lss, uap->uoss, sizeof(l_stack_t));
  818         }
  819 
  820         return (error);
  821 }
  822 
  823 int
  824 linux_ftruncate64(struct thread *td, struct linux_ftruncate64_args *args)
  825 {
  826         struct ftruncate_args sa;
  827 
  828 #ifdef DEBUG
  829         if (ldebug(ftruncate64))
  830                 printf(ARGS(ftruncate64, "%u, %jd"), args->fd,
  831                     (intmax_t)args->length);
  832 #endif
  833 
  834         sa.fd = args->fd;
  835         sa.length = args->length;
  836         return sys_ftruncate(td, &sa);
  837 }
  838 
  839 int
  840 linux_gettimeofday(struct thread *td, struct linux_gettimeofday_args *uap)
  841 {
  842         struct timeval atv;
  843         l_timeval atv32;
  844         struct timezone rtz;
  845         int error = 0;
  846 
  847         if (uap->tp) {
  848                 microtime(&atv);
  849                 atv32.tv_sec = atv.tv_sec;
  850                 atv32.tv_usec = atv.tv_usec;
  851                 error = copyout(&atv32, uap->tp, sizeof(atv32));
  852         }
  853         if (error == 0 && uap->tzp != NULL) {
  854                 rtz.tz_minuteswest = tz_minuteswest;
  855                 rtz.tz_dsttime = tz_dsttime;
  856                 error = copyout(&rtz, uap->tzp, sizeof(rtz));
  857         }
  858         return (error);
  859 }
  860 
  861 int
  862 linux_settimeofday(struct thread *td, struct linux_settimeofday_args *uap)
  863 {
  864         l_timeval atv32;
  865         struct timeval atv, *tvp;
  866         struct timezone atz, *tzp;
  867         int error;
  868 
  869         if (uap->tp) {
  870                 error = copyin(uap->tp, &atv32, sizeof(atv32));
  871                 if (error)
  872                         return (error);
  873                 atv.tv_sec = atv32.tv_sec;
  874                 atv.tv_usec = atv32.tv_usec;
  875                 tvp = &atv;
  876         } else
  877                 tvp = NULL;
  878         if (uap->tzp) {
  879                 error = copyin(uap->tzp, &atz, sizeof(atz));
  880                 if (error)
  881                         return (error);
  882                 tzp = &atz;
  883         } else
  884                 tzp = NULL;
  885         return (kern_settimeofday(td, tvp, tzp));
  886 }
  887 
  888 int
  889 linux_getrusage(struct thread *td, struct linux_getrusage_args *uap)
  890 {
  891         struct rusage s;
  892         int error;
  893 
  894         error = kern_getrusage(td, uap->who, &s);
  895         if (error != 0)
  896                 return (error);
  897         if (uap->rusage != NULL)
  898                 error = linux_copyout_rusage(&s, uap->rusage);
  899         return (error);
  900 }
  901 
  902 int
  903 linux_set_thread_area(struct thread *td,
  904     struct linux_set_thread_area_args *args)
  905 {
  906         struct l_user_desc info;
  907         struct user_segment_descriptor sd;
  908         struct pcb *pcb;
  909         int a[2];
  910         int error;
  911 
  912         error = copyin(args->desc, &info, sizeof(struct l_user_desc));
  913         if (error)
  914                 return (error);
  915 
  916 #ifdef DEBUG
  917         if (ldebug(set_thread_area))
  918                 printf(ARGS(set_thread_area, "%i, %x, %x, %i, %i, %i, "
  919                     "%i, %i, %i"), info.entry_number, info.base_addr,
  920                     info.limit, info.seg_32bit, info.contents,
  921                     info.read_exec_only, info.limit_in_pages,
  922                     info.seg_not_present, info.useable);
  923 #endif
  924 
  925         /*
  926          * Semantics of Linux version: every thread in the system has array
  927          * of three TLS descriptors. 1st is GLIBC TLS, 2nd is WINE, 3rd unknown.
  928          * This syscall loads one of the selected TLS decriptors with a value
  929          * and also loads GDT descriptors 6, 7 and 8 with the content of
  930          * the per-thread descriptors.
  931          *
  932          * Semantics of FreeBSD version: I think we can ignore that Linux has
  933          * three per-thread descriptors and use just the first one.
  934          * The tls_array[] is used only in [gs]et_thread_area() syscalls and
  935          * for loading the GDT descriptors. We use just one GDT descriptor
  936          * for TLS, so we will load just one.
  937          *
  938          * XXX: This doesn't work when a user space process tries to use more
  939          * than one TLS segment. Comment in the Linux source says wine might
  940          * do this.
  941          */
  942 
  943         /*
  944          * GLIBC reads current %gs and call set_thread_area() with it.
  945          * We should let GUDATA_SEL and GUGS32_SEL proceed as well because
  946          * we use these segments.
  947          */
  948         switch (info.entry_number) {
  949         case GUGS32_SEL:
  950         case GUDATA_SEL:
  951         case 6:
  952         case -1:
  953                 info.entry_number = GUGS32_SEL;
  954                 break;
  955         default:
  956                 return (EINVAL);
  957         }
  958 
  959         /*
  960          * We have to copy out the GDT entry we use.
  961          *
  962          * XXX: What if a user space program does not check the return value
  963          * and tries to use 6, 7 or 8?
  964          */
  965         error = copyout(&info, args->desc, sizeof(struct l_user_desc));
  966         if (error)
  967                 return (error);
  968 
  969         if (LINUX_LDT_empty(&info)) {
  970                 a[0] = 0;
  971                 a[1] = 0;
  972         } else {
  973                 a[0] = LINUX_LDT_entry_a(&info);
  974                 a[1] = LINUX_LDT_entry_b(&info);
  975         }
  976 
  977         memcpy(&sd, &a, sizeof(a));
  978 #ifdef DEBUG
  979         if (ldebug(set_thread_area))
  980                 printf("Segment created in set_thread_area: "
  981                     "lobase: %x, hibase: %x, lolimit: %x, hilimit: %x, "
  982                     "type: %i, dpl: %i, p: %i, xx: %i, long: %i, "
  983                     "def32: %i, gran: %i\n",
  984                     sd.sd_lobase,
  985                     sd.sd_hibase,
  986                     sd.sd_lolimit,
  987                     sd.sd_hilimit,
  988                     sd.sd_type,
  989                     sd.sd_dpl,
  990                     sd.sd_p,
  991                     sd.sd_xx,
  992                     sd.sd_long,
  993                     sd.sd_def32,
  994                     sd.sd_gran);
  995 #endif
  996 
  997         pcb = td->td_pcb;
  998         pcb->pcb_gsbase = (register_t)info.base_addr;
  999         set_pcb_flags(pcb, PCB_32BIT);
 1000         update_gdt_gsbase(td, info.base_addr);
 1001 
 1002         return (0);
 1003 }

Cache object: 81adc9c9fa8b18e52c7a7f98572d9eb1


[ 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.