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/9.1/sys/amd64/linux32/linux32_machdep.c 225617 2011-09-16 13:58:51Z kmacy $");
   33 
   34 #include <sys/param.h>
   35 #include <sys/kernel.h>
   36 #include <sys/systm.h>
   37 #include <sys/capability.h>
   38 #include <sys/file.h>
   39 #include <sys/fcntl.h>
   40 #include <sys/clock.h>
   41 #include <sys/imgact.h>
   42 #include <sys/limits.h>
   43 #include <sys/lock.h>
   44 #include <sys/malloc.h>
   45 #include <sys/mman.h>
   46 #include <sys/mutex.h>
   47 #include <sys/priv.h>
   48 #include <sys/proc.h>
   49 #include <sys/resource.h>
   50 #include <sys/resourcevar.h>
   51 #include <sys/sched.h>
   52 #include <sys/syscallsubr.h>
   53 #include <sys/sysproto.h>
   54 #include <sys/unistd.h>
   55 #include <sys/wait.h>
   56 
   57 #include <machine/frame.h>
   58 #include <machine/pcb.h>
   59 #include <machine/psl.h>
   60 #include <machine/segments.h>
   61 #include <machine/specialreg.h>
   62 
   63 #include <vm/vm.h>
   64 #include <vm/pmap.h>
   65 #include <vm/vm_map.h>
   66 
   67 #include <compat/freebsd32/freebsd32_util.h>
   68 #include <amd64/linux32/linux.h>
   69 #include <amd64/linux32/linux32_proto.h>
   70 #include <compat/linux/linux_ipc.h>
   71 #include <compat/linux/linux_misc.h>
   72 #include <compat/linux/linux_signal.h>
   73 #include <compat/linux/linux_util.h>
   74 #include <compat/linux/linux_emul.h>
   75 
   76 struct l_old_select_argv {
   77         l_int           nfds;
   78         l_uintptr_t     readfds;
   79         l_uintptr_t     writefds;
   80         l_uintptr_t     exceptfds;
   81         l_uintptr_t     timeout;
   82 } __packed;
   83 
   84 int
   85 linux_to_bsd_sigaltstack(int lsa)
   86 {
   87         int bsa = 0;
   88 
   89         if (lsa & LINUX_SS_DISABLE)
   90                 bsa |= SS_DISABLE;
   91         if (lsa & LINUX_SS_ONSTACK)
   92                 bsa |= SS_ONSTACK;
   93         return (bsa);
   94 }
   95 
   96 static int      linux_mmap_common(struct thread *td, l_uintptr_t addr,
   97                     l_size_t len, l_int prot, l_int flags, l_int fd,
   98                     l_loff_t pos);
   99 
  100 int
  101 bsd_to_linux_sigaltstack(int bsa)
  102 {
  103         int lsa = 0;
  104 
  105         if (bsa & SS_DISABLE)
  106                 lsa |= LINUX_SS_DISABLE;
  107         if (bsa & SS_ONSTACK)
  108                 lsa |= LINUX_SS_ONSTACK;
  109         return (lsa);
  110 }
  111 
  112 static void
  113 bsd_to_linux_rusage(struct rusage *ru, struct l_rusage *lru)
  114 {
  115 
  116         lru->ru_utime.tv_sec = ru->ru_utime.tv_sec;
  117         lru->ru_utime.tv_usec = ru->ru_utime.tv_usec;
  118         lru->ru_stime.tv_sec = ru->ru_stime.tv_sec;
  119         lru->ru_stime.tv_usec = ru->ru_stime.tv_usec;
  120         lru->ru_maxrss = ru->ru_maxrss;
  121         lru->ru_ixrss = ru->ru_ixrss;
  122         lru->ru_idrss = ru->ru_idrss;
  123         lru->ru_isrss = ru->ru_isrss;
  124         lru->ru_minflt = ru->ru_minflt;
  125         lru->ru_majflt = ru->ru_majflt;
  126         lru->ru_nswap = ru->ru_nswap;
  127         lru->ru_inblock = ru->ru_inblock;
  128         lru->ru_oublock = ru->ru_oublock;
  129         lru->ru_msgsnd = ru->ru_msgsnd;
  130         lru->ru_msgrcv = ru->ru_msgrcv;
  131         lru->ru_nsignals = ru->ru_nsignals;
  132         lru->ru_nvcsw = ru->ru_nvcsw;
  133         lru->ru_nivcsw = ru->ru_nivcsw;
  134 }
  135 
  136 int
  137 linux_execve(struct thread *td, struct linux_execve_args *args)
  138 {
  139         struct image_args eargs;
  140         char *path;
  141         int error;
  142 
  143         LCONVPATHEXIST(td, args->path, &path);
  144 
  145 #ifdef DEBUG
  146         if (ldebug(execve))
  147                 printf(ARGS(execve, "%s"), path);
  148 #endif
  149 
  150         error = freebsd32_exec_copyin_args(&eargs, path, UIO_SYSSPACE,
  151             args->argp, args->envp);
  152         free(path, M_TEMP);
  153         if (error == 0)
  154                 error = kern_execve(td, &eargs, NULL);
  155         if (error == 0)
  156                 /* Linux process can execute FreeBSD one, do not attempt
  157                  * to create emuldata for such process using
  158                  * linux_proc_init, this leads to a panic on KASSERT
  159                  * because such process has p->p_emuldata == NULL.
  160                  */
  161                 if (SV_PROC_ABI(td->td_proc) == SV_ABI_LINUX)
  162                         error = linux_proc_init(td, 0, 0);
  163         return (error);
  164 }
  165 
  166 CTASSERT(sizeof(struct l_iovec32) == 8);
  167 
  168 static int
  169 linux32_copyinuio(struct l_iovec32 *iovp, l_ulong iovcnt, struct uio **uiop)
  170 {
  171         struct l_iovec32 iov32;
  172         struct iovec *iov;
  173         struct uio *uio;
  174         uint32_t iovlen;
  175         int error, i;
  176 
  177         *uiop = NULL;
  178         if (iovcnt > UIO_MAXIOV)
  179                 return (EINVAL);
  180         iovlen = iovcnt * sizeof(struct iovec);
  181         uio = malloc(iovlen + sizeof(*uio), M_IOV, M_WAITOK);
  182         iov = (struct iovec *)(uio + 1);
  183         for (i = 0; i < iovcnt; i++) {
  184                 error = copyin(&iovp[i], &iov32, sizeof(struct l_iovec32));
  185                 if (error) {
  186                         free(uio, M_IOV);
  187                         return (error);
  188                 }
  189                 iov[i].iov_base = PTRIN(iov32.iov_base);
  190                 iov[i].iov_len = iov32.iov_len;
  191         }
  192         uio->uio_iov = iov;
  193         uio->uio_iovcnt = iovcnt;
  194         uio->uio_segflg = UIO_USERSPACE;
  195         uio->uio_offset = -1;
  196         uio->uio_resid = 0;
  197         for (i = 0; i < iovcnt; i++) {
  198                 if (iov->iov_len > INT_MAX - uio->uio_resid) {
  199                         free(uio, M_IOV);
  200                         return (EINVAL);
  201                 }
  202                 uio->uio_resid += iov->iov_len;
  203                 iov++;
  204         }
  205         *uiop = uio;
  206         return (0);
  207 }
  208 
  209 int
  210 linux32_copyiniov(struct l_iovec32 *iovp32, l_ulong iovcnt, struct iovec **iovp,
  211     int error)
  212 {
  213         struct l_iovec32 iov32;
  214         struct iovec *iov;
  215         uint32_t iovlen;
  216         int i;
  217 
  218         *iovp = NULL;
  219         if (iovcnt > UIO_MAXIOV)
  220                 return (error);
  221         iovlen = iovcnt * sizeof(struct iovec);
  222         iov = malloc(iovlen, M_IOV, M_WAITOK);
  223         for (i = 0; i < iovcnt; i++) {
  224                 error = copyin(&iovp32[i], &iov32, sizeof(struct l_iovec32));
  225                 if (error) {
  226                         free(iov, M_IOV);
  227                         return (error);
  228                 }
  229                 iov[i].iov_base = PTRIN(iov32.iov_base);
  230                 iov[i].iov_len = iov32.iov_len;
  231         }
  232         *iovp = iov;
  233         return(0);
  234 
  235 }
  236 
  237 int
  238 linux_readv(struct thread *td, struct linux_readv_args *uap)
  239 {
  240         struct uio *auio;
  241         int error;
  242 
  243         error = linux32_copyinuio(uap->iovp, uap->iovcnt, &auio);
  244         if (error)
  245                 return (error);
  246         error = kern_readv(td, uap->fd, auio);
  247         free(auio, M_IOV);
  248         return (error);
  249 }
  250 
  251 int
  252 linux_writev(struct thread *td, struct linux_writev_args *uap)
  253 {
  254         struct uio *auio;
  255         int error;
  256 
  257         error = linux32_copyinuio(uap->iovp, uap->iovcnt, &auio);
  258         if (error)
  259                 return (error);
  260         error = kern_writev(td, uap->fd, auio);
  261         free(auio, M_IOV);
  262         return (error);
  263 }
  264 
  265 struct l_ipc_kludge {
  266         l_uintptr_t msgp;
  267         l_long msgtyp;
  268 } __packed;
  269 
  270 int
  271 linux_ipc(struct thread *td, struct linux_ipc_args *args)
  272 {
  273 
  274         switch (args->what & 0xFFFF) {
  275         case LINUX_SEMOP: {
  276                 struct linux_semop_args a;
  277 
  278                 a.semid = args->arg1;
  279                 a.tsops = args->ptr;
  280                 a.nsops = args->arg2;
  281                 return (linux_semop(td, &a));
  282         }
  283         case LINUX_SEMGET: {
  284                 struct linux_semget_args a;
  285 
  286                 a.key = args->arg1;
  287                 a.nsems = args->arg2;
  288                 a.semflg = args->arg3;
  289                 return (linux_semget(td, &a));
  290         }
  291         case LINUX_SEMCTL: {
  292                 struct linux_semctl_args a;
  293                 int error;
  294 
  295                 a.semid = args->arg1;
  296                 a.semnum = args->arg2;
  297                 a.cmd = args->arg3;
  298                 error = copyin(args->ptr, &a.arg, sizeof(a.arg));
  299                 if (error)
  300                         return (error);
  301                 return (linux_semctl(td, &a));
  302         }
  303         case LINUX_MSGSND: {
  304                 struct linux_msgsnd_args a;
  305 
  306                 a.msqid = args->arg1;
  307                 a.msgp = args->ptr;
  308                 a.msgsz = args->arg2;
  309                 a.msgflg = args->arg3;
  310                 return (linux_msgsnd(td, &a));
  311         }
  312         case LINUX_MSGRCV: {
  313                 struct linux_msgrcv_args a;
  314 
  315                 a.msqid = args->arg1;
  316                 a.msgsz = args->arg2;
  317                 a.msgflg = args->arg3;
  318                 if ((args->what >> 16) == 0) {
  319                         struct l_ipc_kludge tmp;
  320                         int error;
  321 
  322                         if (args->ptr == 0)
  323                                 return (EINVAL);
  324                         error = copyin(args->ptr, &tmp, sizeof(tmp));
  325                         if (error)
  326                                 return (error);
  327                         a.msgp = PTRIN(tmp.msgp);
  328                         a.msgtyp = tmp.msgtyp;
  329                 } else {
  330                         a.msgp = args->ptr;
  331                         a.msgtyp = args->arg5;
  332                 }
  333                 return (linux_msgrcv(td, &a));
  334         }
  335         case LINUX_MSGGET: {
  336                 struct linux_msgget_args a;
  337 
  338                 a.key = args->arg1;
  339                 a.msgflg = args->arg2;
  340                 return (linux_msgget(td, &a));
  341         }
  342         case LINUX_MSGCTL: {
  343                 struct linux_msgctl_args a;
  344 
  345                 a.msqid = args->arg1;
  346                 a.cmd = args->arg2;
  347                 a.buf = args->ptr;
  348                 return (linux_msgctl(td, &a));
  349         }
  350         case LINUX_SHMAT: {
  351                 struct linux_shmat_args a;
  352 
  353                 a.shmid = args->arg1;
  354                 a.shmaddr = args->ptr;
  355                 a.shmflg = args->arg2;
  356                 a.raddr = PTRIN((l_uint)args->arg3);
  357                 return (linux_shmat(td, &a));
  358         }
  359         case LINUX_SHMDT: {
  360                 struct linux_shmdt_args a;
  361 
  362                 a.shmaddr = args->ptr;
  363                 return (linux_shmdt(td, &a));
  364         }
  365         case LINUX_SHMGET: {
  366                 struct linux_shmget_args a;
  367 
  368                 a.key = args->arg1;
  369                 a.size = args->arg2;
  370                 a.shmflg = args->arg3;
  371                 return (linux_shmget(td, &a));
  372         }
  373         case LINUX_SHMCTL: {
  374                 struct linux_shmctl_args a;
  375 
  376                 a.shmid = args->arg1;
  377                 a.cmd = args->arg2;
  378                 a.buf = args->ptr;
  379                 return (linux_shmctl(td, &a));
  380         }
  381         default:
  382                 break;
  383         }
  384 
  385         return (EINVAL);
  386 }
  387 
  388 int
  389 linux_old_select(struct thread *td, struct linux_old_select_args *args)
  390 {
  391         struct l_old_select_argv linux_args;
  392         struct linux_select_args newsel;
  393         int error;
  394 
  395 #ifdef DEBUG
  396         if (ldebug(old_select))
  397                 printf(ARGS(old_select, "%p"), args->ptr);
  398 #endif
  399 
  400         error = copyin(args->ptr, &linux_args, sizeof(linux_args));
  401         if (error)
  402                 return (error);
  403 
  404         newsel.nfds = linux_args.nfds;
  405         newsel.readfds = PTRIN(linux_args.readfds);
  406         newsel.writefds = PTRIN(linux_args.writefds);
  407         newsel.exceptfds = PTRIN(linux_args.exceptfds);
  408         newsel.timeout = PTRIN(linux_args.timeout);
  409         return (linux_select(td, &newsel));
  410 }
  411 
  412 int
  413 linux_set_cloned_tls(struct thread *td, void *desc)
  414 {
  415         struct user_segment_descriptor sd;
  416         struct l_user_desc info;
  417         struct pcb *pcb;
  418         int error;
  419         int a[2];
  420 
  421         error = copyin(desc, &info, sizeof(struct l_user_desc));
  422         if (error) {
  423                 printf(LMSG("copyin failed!"));
  424         } else {
  425                 /* We might copy out the entry_number as GUGS32_SEL. */
  426                 info.entry_number = GUGS32_SEL;
  427                 error = copyout(&info, desc, sizeof(struct l_user_desc));
  428                 if (error)
  429                         printf(LMSG("copyout failed!"));
  430 
  431                 a[0] = LINUX_LDT_entry_a(&info);
  432                 a[1] = LINUX_LDT_entry_b(&info);
  433 
  434                 memcpy(&sd, &a, sizeof(a));
  435 #ifdef DEBUG
  436                 if (ldebug(clone))
  437                         printf("Segment created in clone with "
  438                             "CLONE_SETTLS: lobase: %x, hibase: %x, "
  439                             "lolimit: %x, hilimit: %x, type: %i, "
  440                             "dpl: %i, p: %i, xx: %i, long: %i, "
  441                             "def32: %i, gran: %i\n", sd.sd_lobase,
  442                             sd.sd_hibase, sd.sd_lolimit, sd.sd_hilimit,
  443                             sd.sd_type, sd.sd_dpl, sd.sd_p, sd.sd_xx,
  444                             sd.sd_long, sd.sd_def32, sd.sd_gran);
  445 #endif
  446                 pcb = td->td_pcb;
  447                 pcb->pcb_gsbase = (register_t)info.base_addr;
  448 /* XXXKIB       pcb->pcb_gs32sd = sd; */
  449                 td->td_frame->tf_gs = GSEL(GUGS32_SEL, SEL_UPL);
  450                 set_pcb_flags(pcb, PCB_GS32BIT | PCB_32BIT);
  451         }
  452 
  453         return (error);
  454 }
  455 
  456 int
  457 linux_set_upcall_kse(struct thread *td, register_t stack)
  458 {
  459 
  460         td->td_frame->tf_rsp = stack;
  461 
  462         return (0);
  463 }
  464 
  465 #define STACK_SIZE  (2 * 1024 * 1024)
  466 #define GUARD_SIZE  (4 * PAGE_SIZE)
  467 
  468 int
  469 linux_mmap2(struct thread *td, struct linux_mmap2_args *args)
  470 {
  471 
  472 #ifdef DEBUG
  473         if (ldebug(mmap2))
  474                 printf(ARGS(mmap2, "0x%08x, %d, %d, 0x%08x, %d, %d"),
  475                     args->addr, args->len, args->prot,
  476                     args->flags, args->fd, args->pgoff);
  477 #endif
  478 
  479         return (linux_mmap_common(td, PTROUT(args->addr), args->len, args->prot,
  480                 args->flags, args->fd, (uint64_t)(uint32_t)args->pgoff *
  481                 PAGE_SIZE));
  482 }
  483 
  484 int
  485 linux_mmap(struct thread *td, struct linux_mmap_args *args)
  486 {
  487         int error;
  488         struct l_mmap_argv linux_args;
  489 
  490         error = copyin(args->ptr, &linux_args, sizeof(linux_args));
  491         if (error)
  492                 return (error);
  493 
  494 #ifdef DEBUG
  495         if (ldebug(mmap))
  496                 printf(ARGS(mmap, "0x%08x, %d, %d, 0x%08x, %d, %d"),
  497                     linux_args.addr, linux_args.len, linux_args.prot,
  498                     linux_args.flags, linux_args.fd, linux_args.pgoff);
  499 #endif
  500 
  501         return (linux_mmap_common(td, linux_args.addr, linux_args.len,
  502             linux_args.prot, linux_args.flags, linux_args.fd,
  503             (uint32_t)linux_args.pgoff));
  504 }
  505 
  506 static int
  507 linux_mmap_common(struct thread *td, l_uintptr_t addr, l_size_t len, l_int prot,
  508     l_int flags, l_int fd, l_loff_t pos)
  509 {
  510         struct proc *p = td->td_proc;
  511         struct mmap_args /* {
  512                 caddr_t addr;
  513                 size_t len;
  514                 int prot;
  515                 int flags;
  516                 int fd;
  517                 long pad;
  518                 off_t pos;
  519         } */ bsd_args;
  520         int error;
  521         struct file *fp;
  522 
  523         error = 0;
  524         bsd_args.flags = 0;
  525         fp = NULL;
  526 
  527         /*
  528          * Linux mmap(2):
  529          * You must specify exactly one of MAP_SHARED and MAP_PRIVATE
  530          */
  531         if (!((flags & LINUX_MAP_SHARED) ^ (flags & LINUX_MAP_PRIVATE)))
  532                 return (EINVAL);
  533 
  534         if (flags & LINUX_MAP_SHARED)
  535                 bsd_args.flags |= MAP_SHARED;
  536         if (flags & LINUX_MAP_PRIVATE)
  537                 bsd_args.flags |= MAP_PRIVATE;
  538         if (flags & LINUX_MAP_FIXED)
  539                 bsd_args.flags |= MAP_FIXED;
  540         if (flags & LINUX_MAP_ANON) {
  541                 /* Enforce pos to be on page boundary, then ignore. */
  542                 if ((pos & PAGE_MASK) != 0)
  543                         return (EINVAL);
  544                 pos = 0;
  545                 bsd_args.flags |= MAP_ANON;
  546         } else
  547                 bsd_args.flags |= MAP_NOSYNC;
  548         if (flags & LINUX_MAP_GROWSDOWN)
  549                 bsd_args.flags |= MAP_STACK;
  550 
  551         /*
  552          * PROT_READ, PROT_WRITE, or PROT_EXEC implies PROT_READ and PROT_EXEC
  553          * on Linux/i386. We do this to ensure maximum compatibility.
  554          * Linux/ia64 does the same in i386 emulation mode.
  555          */
  556         bsd_args.prot = prot;
  557         if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC))
  558                 bsd_args.prot |= PROT_READ | PROT_EXEC;
  559 
  560         /* Linux does not check file descriptor when MAP_ANONYMOUS is set. */
  561         bsd_args.fd = (bsd_args.flags & MAP_ANON) ? -1 : fd;
  562         if (bsd_args.fd != -1) {
  563                 /*
  564                  * Linux follows Solaris mmap(2) description:
  565                  * The file descriptor fildes is opened with
  566                  * read permission, regardless of the
  567                  * protection options specified.
  568                  */
  569 
  570                 if ((error = fget(td, bsd_args.fd, CAP_MMAP, &fp)) != 0)
  571                         return (error);
  572                 if (fp->f_type != DTYPE_VNODE) {
  573                         fdrop(fp, td);
  574                         return (EINVAL);
  575                 }
  576 
  577                 /* Linux mmap() just fails for O_WRONLY files */
  578                 if (!(fp->f_flag & FREAD)) {
  579                         fdrop(fp, td);
  580                         return (EACCES);
  581                 }
  582 
  583                 fdrop(fp, td);
  584         }
  585 
  586         if (flags & LINUX_MAP_GROWSDOWN) {
  587                 /*
  588                  * The Linux MAP_GROWSDOWN option does not limit auto
  589                  * growth of the region.  Linux mmap with this option
  590                  * takes as addr the inital BOS, and as len, the initial
  591                  * region size.  It can then grow down from addr without
  592                  * limit.  However, Linux threads has an implicit internal
  593                  * limit to stack size of STACK_SIZE.  Its just not
  594                  * enforced explicitly in Linux.  But, here we impose
  595                  * a limit of (STACK_SIZE - GUARD_SIZE) on the stack
  596                  * region, since we can do this with our mmap.
  597                  *
  598                  * Our mmap with MAP_STACK takes addr as the maximum
  599                  * downsize limit on BOS, and as len the max size of
  600                  * the region.  It then maps the top SGROWSIZ bytes,
  601                  * and auto grows the region down, up to the limit
  602                  * in addr.
  603                  *
  604                  * If we don't use the MAP_STACK option, the effect
  605                  * of this code is to allocate a stack region of a
  606                  * fixed size of (STACK_SIZE - GUARD_SIZE).
  607                  */
  608 
  609                 if ((caddr_t)PTRIN(addr) + len > p->p_vmspace->vm_maxsaddr) {
  610                         /*
  611                          * Some Linux apps will attempt to mmap
  612                          * thread stacks near the top of their
  613                          * address space.  If their TOS is greater
  614                          * than vm_maxsaddr, vm_map_growstack()
  615                          * will confuse the thread stack with the
  616                          * process stack and deliver a SEGV if they
  617                          * attempt to grow the thread stack past their
  618                          * current stacksize rlimit.  To avoid this,
  619                          * adjust vm_maxsaddr upwards to reflect
  620                          * the current stacksize rlimit rather
  621                          * than the maximum possible stacksize.
  622                          * It would be better to adjust the
  623                          * mmap'ed region, but some apps do not check
  624                          * mmap's return value.
  625                          */
  626                         PROC_LOCK(p);
  627                         p->p_vmspace->vm_maxsaddr = (char *)LINUX32_USRSTACK -
  628                             lim_cur(p, RLIMIT_STACK);
  629                         PROC_UNLOCK(p);
  630                 }
  631 
  632                 /*
  633                  * This gives us our maximum stack size and a new BOS.
  634                  * If we're using VM_STACK, then mmap will just map
  635                  * the top SGROWSIZ bytes, and let the stack grow down
  636                  * to the limit at BOS.  If we're not using VM_STACK
  637                  * we map the full stack, since we don't have a way
  638                  * to autogrow it.
  639                  */
  640                 if (len > STACK_SIZE - GUARD_SIZE) {
  641                         bsd_args.addr = (caddr_t)PTRIN(addr);
  642                         bsd_args.len = len;
  643                 } else {
  644                         bsd_args.addr = (caddr_t)PTRIN(addr) -
  645                             (STACK_SIZE - GUARD_SIZE - len);
  646                         bsd_args.len = STACK_SIZE - GUARD_SIZE;
  647                 }
  648         } else {
  649                 bsd_args.addr = (caddr_t)PTRIN(addr);
  650                 bsd_args.len  = len;
  651         }
  652         bsd_args.pos = pos;
  653 
  654 #ifdef DEBUG
  655         if (ldebug(mmap))
  656                 printf("-> %s(%p, %d, %d, 0x%08x, %d, 0x%x)\n",
  657                     __func__,
  658                     (void *)bsd_args.addr, (int)bsd_args.len, bsd_args.prot,
  659                     bsd_args.flags, bsd_args.fd, (int)bsd_args.pos);
  660 #endif
  661         error = sys_mmap(td, &bsd_args);
  662 #ifdef DEBUG
  663         if (ldebug(mmap))
  664                 printf("-> %s() return: 0x%x (0x%08x)\n",
  665                         __func__, error, (u_int)td->td_retval[0]);
  666 #endif
  667         return (error);
  668 }
  669 
  670 int
  671 linux_mprotect(struct thread *td, struct linux_mprotect_args *uap)
  672 {
  673         struct mprotect_args bsd_args;
  674 
  675         bsd_args.addr = uap->addr;
  676         bsd_args.len = uap->len;
  677         bsd_args.prot = uap->prot;
  678         if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC))
  679                 bsd_args.prot |= PROT_READ | PROT_EXEC;
  680         return (sys_mprotect(td, &bsd_args));
  681 }
  682 
  683 int
  684 linux_iopl(struct thread *td, struct linux_iopl_args *args)
  685 {
  686         int error;
  687 
  688         if (args->level < 0 || args->level > 3)
  689                 return (EINVAL);
  690         if ((error = priv_check(td, PRIV_IO)) != 0)
  691                 return (error);
  692         if ((error = securelevel_gt(td->td_ucred, 0)) != 0)
  693                 return (error);
  694         td->td_frame->tf_rflags = (td->td_frame->tf_rflags & ~PSL_IOPL) |
  695             (args->level * (PSL_IOPL / 3));
  696 
  697         return (0);
  698 }
  699 
  700 int
  701 linux_pipe(struct thread *td, struct linux_pipe_args *args)
  702 {
  703         int error;
  704         int fildes[2];
  705 
  706 #ifdef DEBUG
  707         if (ldebug(pipe))
  708                 printf(ARGS(pipe, "*"));
  709 #endif
  710 
  711         error = kern_pipe(td, fildes);
  712         if (error)
  713                 return (error);
  714 
  715         /* XXX: Close descriptors on error. */
  716         return (copyout(fildes, args->pipefds, sizeof fildes));
  717 }
  718 
  719 int
  720 linux_sigaction(struct thread *td, struct linux_sigaction_args *args)
  721 {
  722         l_osigaction_t osa;
  723         l_sigaction_t act, oact;
  724         int error;
  725 
  726 #ifdef DEBUG
  727         if (ldebug(sigaction))
  728                 printf(ARGS(sigaction, "%d, %p, %p"),
  729                     args->sig, (void *)args->nsa, (void *)args->osa);
  730 #endif
  731 
  732         if (args->nsa != NULL) {
  733                 error = copyin(args->nsa, &osa, sizeof(l_osigaction_t));
  734                 if (error)
  735                         return (error);
  736                 act.lsa_handler = osa.lsa_handler;
  737                 act.lsa_flags = osa.lsa_flags;
  738                 act.lsa_restorer = osa.lsa_restorer;
  739                 LINUX_SIGEMPTYSET(act.lsa_mask);
  740                 act.lsa_mask.__bits[0] = osa.lsa_mask;
  741         }
  742 
  743         error = linux_do_sigaction(td, args->sig, args->nsa ? &act : NULL,
  744             args->osa ? &oact : NULL);
  745 
  746         if (args->osa != NULL && !error) {
  747                 osa.lsa_handler = oact.lsa_handler;
  748                 osa.lsa_flags = oact.lsa_flags;
  749                 osa.lsa_restorer = oact.lsa_restorer;
  750                 osa.lsa_mask = oact.lsa_mask.__bits[0];
  751                 error = copyout(&osa, args->osa, sizeof(l_osigaction_t));
  752         }
  753 
  754         return (error);
  755 }
  756 
  757 /*
  758  * Linux has two extra args, restart and oldmask.  We don't use these,
  759  * but it seems that "restart" is actually a context pointer that
  760  * enables the signal to happen with a different register set.
  761  */
  762 int
  763 linux_sigsuspend(struct thread *td, struct linux_sigsuspend_args *args)
  764 {
  765         sigset_t sigmask;
  766         l_sigset_t mask;
  767 
  768 #ifdef DEBUG
  769         if (ldebug(sigsuspend))
  770                 printf(ARGS(sigsuspend, "%08lx"), (unsigned long)args->mask);
  771 #endif
  772 
  773         LINUX_SIGEMPTYSET(mask);
  774         mask.__bits[0] = args->mask;
  775         linux_to_bsd_sigset(&mask, &sigmask);
  776         return (kern_sigsuspend(td, sigmask));
  777 }
  778 
  779 int
  780 linux_rt_sigsuspend(struct thread *td, struct linux_rt_sigsuspend_args *uap)
  781 {
  782         l_sigset_t lmask;
  783         sigset_t sigmask;
  784         int error;
  785 
  786 #ifdef DEBUG
  787         if (ldebug(rt_sigsuspend))
  788                 printf(ARGS(rt_sigsuspend, "%p, %d"),
  789                     (void *)uap->newset, uap->sigsetsize);
  790 #endif
  791 
  792         if (uap->sigsetsize != sizeof(l_sigset_t))
  793                 return (EINVAL);
  794 
  795         error = copyin(uap->newset, &lmask, sizeof(l_sigset_t));
  796         if (error)
  797                 return (error);
  798 
  799         linux_to_bsd_sigset(&lmask, &sigmask);
  800         return (kern_sigsuspend(td, sigmask));
  801 }
  802 
  803 int
  804 linux_pause(struct thread *td, struct linux_pause_args *args)
  805 {
  806         struct proc *p = td->td_proc;
  807         sigset_t sigmask;
  808 
  809 #ifdef DEBUG
  810         if (ldebug(pause))
  811                 printf(ARGS(pause, ""));
  812 #endif
  813 
  814         PROC_LOCK(p);
  815         sigmask = td->td_sigmask;
  816         PROC_UNLOCK(p);
  817         return (kern_sigsuspend(td, sigmask));
  818 }
  819 
  820 int
  821 linux_sigaltstack(struct thread *td, struct linux_sigaltstack_args *uap)
  822 {
  823         stack_t ss, oss;
  824         l_stack_t lss;
  825         int error;
  826 
  827 #ifdef DEBUG
  828         if (ldebug(sigaltstack))
  829                 printf(ARGS(sigaltstack, "%p, %p"), uap->uss, uap->uoss);
  830 #endif
  831 
  832         if (uap->uss != NULL) {
  833                 error = copyin(uap->uss, &lss, sizeof(l_stack_t));
  834                 if (error)
  835                         return (error);
  836 
  837                 ss.ss_sp = PTRIN(lss.ss_sp);
  838                 ss.ss_size = lss.ss_size;
  839                 ss.ss_flags = linux_to_bsd_sigaltstack(lss.ss_flags);
  840         }
  841         error = kern_sigaltstack(td, (uap->uss != NULL) ? &ss : NULL,
  842             (uap->uoss != NULL) ? &oss : NULL);
  843         if (!error && uap->uoss != NULL) {
  844                 lss.ss_sp = PTROUT(oss.ss_sp);
  845                 lss.ss_size = oss.ss_size;
  846                 lss.ss_flags = bsd_to_linux_sigaltstack(oss.ss_flags);
  847                 error = copyout(&lss, uap->uoss, sizeof(l_stack_t));
  848         }
  849 
  850         return (error);
  851 }
  852 
  853 int
  854 linux_ftruncate64(struct thread *td, struct linux_ftruncate64_args *args)
  855 {
  856         struct ftruncate_args sa;
  857 
  858 #ifdef DEBUG
  859         if (ldebug(ftruncate64))
  860                 printf(ARGS(ftruncate64, "%u, %jd"), args->fd,
  861                     (intmax_t)args->length);
  862 #endif
  863 
  864         sa.fd = args->fd;
  865         sa.length = args->length;
  866         return sys_ftruncate(td, &sa);
  867 }
  868 
  869 int
  870 linux_gettimeofday(struct thread *td, struct linux_gettimeofday_args *uap)
  871 {
  872         struct timeval atv;
  873         l_timeval atv32;
  874         struct timezone rtz;
  875         int error = 0;
  876 
  877         if (uap->tp) {
  878                 microtime(&atv);
  879                 atv32.tv_sec = atv.tv_sec;
  880                 atv32.tv_usec = atv.tv_usec;
  881                 error = copyout(&atv32, uap->tp, sizeof(atv32));
  882         }
  883         if (error == 0 && uap->tzp != NULL) {
  884                 rtz.tz_minuteswest = tz_minuteswest;
  885                 rtz.tz_dsttime = tz_dsttime;
  886                 error = copyout(&rtz, uap->tzp, sizeof(rtz));
  887         }
  888         return (error);
  889 }
  890 
  891 int
  892 linux_settimeofday(struct thread *td, struct linux_settimeofday_args *uap)
  893 {
  894         l_timeval atv32;
  895         struct timeval atv, *tvp;
  896         struct timezone atz, *tzp;
  897         int error;
  898 
  899         if (uap->tp) {
  900                 error = copyin(uap->tp, &atv32, sizeof(atv32));
  901                 if (error)
  902                         return (error);
  903                 atv.tv_sec = atv32.tv_sec;
  904                 atv.tv_usec = atv32.tv_usec;
  905                 tvp = &atv;
  906         } else
  907                 tvp = NULL;
  908         if (uap->tzp) {
  909                 error = copyin(uap->tzp, &atz, sizeof(atz));
  910                 if (error)
  911                         return (error);
  912                 tzp = &atz;
  913         } else
  914                 tzp = NULL;
  915         return (kern_settimeofday(td, tvp, tzp));
  916 }
  917 
  918 int
  919 linux_getrusage(struct thread *td, struct linux_getrusage_args *uap)
  920 {
  921         struct l_rusage s32;
  922         struct rusage s;
  923         int error;
  924 
  925         error = kern_getrusage(td, uap->who, &s);
  926         if (error != 0)
  927                 return (error);
  928         if (uap->rusage != NULL) {
  929                 bsd_to_linux_rusage(&s, &s32);
  930                 error = copyout(&s32, uap->rusage, sizeof(s32));
  931         }
  932         return (error);
  933 }
  934 
  935 int
  936 linux_sched_rr_get_interval(struct thread *td,
  937     struct linux_sched_rr_get_interval_args *uap)
  938 {
  939         struct timespec ts;
  940         struct l_timespec ts32;
  941         int error;
  942 
  943         error = kern_sched_rr_get_interval(td, uap->pid, &ts);
  944         if (error != 0)
  945                 return (error);
  946         ts32.tv_sec = ts.tv_sec;
  947         ts32.tv_nsec = ts.tv_nsec;
  948         return (copyout(&ts32, uap->interval, sizeof(ts32)));
  949 }
  950 
  951 int
  952 linux_set_thread_area(struct thread *td,
  953     struct linux_set_thread_area_args *args)
  954 {
  955         struct l_user_desc info;
  956         struct user_segment_descriptor sd;
  957         struct pcb *pcb;
  958         int a[2];
  959         int error;
  960 
  961         error = copyin(args->desc, &info, sizeof(struct l_user_desc));
  962         if (error)
  963                 return (error);
  964 
  965 #ifdef DEBUG
  966         if (ldebug(set_thread_area))
  967                 printf(ARGS(set_thread_area, "%i, %x, %x, %i, %i, %i, "
  968                     "%i, %i, %i"), info.entry_number, info.base_addr,
  969                     info.limit, info.seg_32bit, info.contents,
  970                     info.read_exec_only, info.limit_in_pages,
  971                     info.seg_not_present, info.useable);
  972 #endif
  973 
  974         /*
  975          * Semantics of Linux version: every thread in the system has array
  976          * of three TLS descriptors. 1st is GLIBC TLS, 2nd is WINE, 3rd unknown.
  977          * This syscall loads one of the selected TLS decriptors with a value
  978          * and also loads GDT descriptors 6, 7 and 8 with the content of
  979          * the per-thread descriptors.
  980          *
  981          * Semantics of FreeBSD version: I think we can ignore that Linux has
  982          * three per-thread descriptors and use just the first one.
  983          * The tls_array[] is used only in [gs]et_thread_area() syscalls and
  984          * for loading the GDT descriptors. We use just one GDT descriptor
  985          * for TLS, so we will load just one.
  986          *
  987          * XXX: This doesn't work when a user space process tries to use more
  988          * than one TLS segment. Comment in the Linux source says wine might
  989          * do this.
  990          */
  991 
  992         /*
  993          * GLIBC reads current %gs and call set_thread_area() with it.
  994          * We should let GUDATA_SEL and GUGS32_SEL proceed as well because
  995          * we use these segments.
  996          */
  997         switch (info.entry_number) {
  998         case GUGS32_SEL:
  999         case GUDATA_SEL:
 1000         case 6:
 1001         case -1:
 1002                 info.entry_number = GUGS32_SEL;
 1003                 break;
 1004         default:
 1005                 return (EINVAL);
 1006         }
 1007 
 1008         /*
 1009          * We have to copy out the GDT entry we use.
 1010          *
 1011          * XXX: What if a user space program does not check the return value
 1012          * and tries to use 6, 7 or 8?
 1013          */
 1014         error = copyout(&info, args->desc, sizeof(struct l_user_desc));
 1015         if (error)
 1016                 return (error);
 1017 
 1018         if (LINUX_LDT_empty(&info)) {
 1019                 a[0] = 0;
 1020                 a[1] = 0;
 1021         } else {
 1022                 a[0] = LINUX_LDT_entry_a(&info);
 1023                 a[1] = LINUX_LDT_entry_b(&info);
 1024         }
 1025 
 1026         memcpy(&sd, &a, sizeof(a));
 1027 #ifdef DEBUG
 1028         if (ldebug(set_thread_area))
 1029                 printf("Segment created in set_thread_area: "
 1030                     "lobase: %x, hibase: %x, lolimit: %x, hilimit: %x, "
 1031                     "type: %i, dpl: %i, p: %i, xx: %i, long: %i, "
 1032                     "def32: %i, gran: %i\n",
 1033                     sd.sd_lobase,
 1034                     sd.sd_hibase,
 1035                     sd.sd_lolimit,
 1036                     sd.sd_hilimit,
 1037                     sd.sd_type,
 1038                     sd.sd_dpl,
 1039                     sd.sd_p,
 1040                     sd.sd_xx,
 1041                     sd.sd_long,
 1042                     sd.sd_def32,
 1043                     sd.sd_gran);
 1044 #endif
 1045 
 1046         pcb = td->td_pcb;
 1047         pcb->pcb_gsbase = (register_t)info.base_addr;
 1048         set_pcb_flags(pcb, PCB_32BIT | PCB_GS32BIT);
 1049         update_gdt_gsbase(td, info.base_addr);
 1050 
 1051         return (0);
 1052 }
 1053 
 1054 int
 1055 linux_wait4(struct thread *td, struct linux_wait4_args *args)
 1056 {
 1057         int error, options;
 1058         struct rusage ru, *rup;
 1059         struct l_rusage lru;
 1060 
 1061 #ifdef DEBUG
 1062         if (ldebug(wait4))
 1063                 printf(ARGS(wait4, "%d, %p, %d, %p"),
 1064                     args->pid, (void *)args->status, args->options,
 1065                     (void *)args->rusage);
 1066 #endif
 1067 
 1068         options = (args->options & (WNOHANG | WUNTRACED));
 1069         /* WLINUXCLONE should be equal to __WCLONE, but we make sure */
 1070         if (args->options & __WCLONE)
 1071                 options |= WLINUXCLONE;
 1072 
 1073         if (args->rusage != NULL)
 1074                 rup = &ru;
 1075         else
 1076                 rup = NULL;
 1077         error = linux_common_wait(td, args->pid, args->status, options, rup);
 1078         if (error)
 1079                 return (error);
 1080         if (args->rusage != NULL) {
 1081                 bsd_to_linux_rusage(rup, &lru);
 1082                 error = copyout(&lru, args->rusage, sizeof(lru));
 1083         }
 1084 
 1085         return (error);
 1086 }
Cache object: 2cecfe2426f1a99589bb3befa2520308
[ 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
