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/kern/sys_generic.c

Version: -  FREEBSD  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-2  -  FREEBSD-11-1  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-4  -  FREEBSD-10-3  -  FREEBSD-10-2  -  FREEBSD-10-1  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-3  -  FREEBSD-9-2  -  FREEBSD-9-1  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-4  -  FREEBSD-8-3  -  FREEBSD-8-2  -  FREEBSD-8-1  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-4  -  FREEBSD-7-3  -  FREEBSD-7-2  -  FREEBSD-7-1  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-4  -  FREEBSD-6-3  -  FREEBSD-6-2  -  FREEBSD-6-1  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-5  -  FREEBSD-5-4  -  FREEBSD-5-3  -  FREEBSD-5-2  -  FREEBSD-5-1  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  linux-2.6  -  linux-2.4.22  -  MK83  -  MK84  -  PLAN9  -  DFBSD  -  NETBSD  -  NETBSD5  -  NETBSD4  -  NETBSD3  -  NETBSD20  -  OPENBSD  -  xnu-517  -  xnu-792  -  xnu-792.6.70  -  xnu-1228  -  xnu-1456.1.26  -  xnu-1699.24.8  -  xnu-2050.18.24  -  OPENSOLARIS  -  minix-3-1-1 
SearchContext: -  none  -  3  -  10 

    1 /*-
    2  * Copyright (c) 1982, 1986, 1989, 1993
    3  *      The Regents of the University of California.  All rights reserved.
    4  * (c) UNIX System Laboratories, Inc.
    5  * All or some portions of this file are derived from material licensed
    6  * to the University of California by American Telephone and Telegraph
    7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
    8  * the permission of UNIX System Laboratories, Inc.
    9  *
   10  * Redistribution and use in source and binary forms, with or without
   11  * modification, are permitted provided that the following conditions
   12  * are met:
   13  * 1. Redistributions of source code must retain the above copyright
   14  *    notice, this list of conditions and the following disclaimer.
   15  * 2. Redistributions in binary form must reproduce the above copyright
   16  *    notice, this list of conditions and the following disclaimer in the
   17  *    documentation and/or other materials provided with the distribution.
   18  * 4. Neither the name of the University nor the names of its contributors
   19  *    may be used to endorse or promote products derived from this software
   20  *    without specific prior written permission.
   21  *
   22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
   23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
   26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   32  * SUCH DAMAGE.
   33  *
   34  *      @(#)sys_generic.c       8.5 (Berkeley) 1/21/94
   35  */
   36 
   37 #include <sys/cdefs.h>
   38 __FBSDID("$FreeBSD: releng/8.0/sys/kern/sys_generic.c 197255 2009-09-16 13:24:37Z kib $");
   39 
   40 #include "opt_compat.h"
   41 #include "opt_ktrace.h"
   42 
   43 #include <sys/param.h>
   44 #include <sys/systm.h>
   45 #include <sys/sysproto.h>
   46 #include <sys/filedesc.h>
   47 #include <sys/filio.h>
   48 #include <sys/fcntl.h>
   49 #include <sys/file.h>
   50 #include <sys/proc.h>
   51 #include <sys/signalvar.h>
   52 #include <sys/socketvar.h>
   53 #include <sys/uio.h>
   54 #include <sys/kernel.h>
   55 #include <sys/ktr.h>
   56 #include <sys/limits.h>
   57 #include <sys/malloc.h>
   58 #include <sys/poll.h>
   59 #include <sys/resourcevar.h>
   60 #include <sys/selinfo.h>
   61 #include <sys/sleepqueue.h>
   62 #include <sys/syscallsubr.h>
   63 #include <sys/sysctl.h>
   64 #include <sys/sysent.h>
   65 #include <sys/vnode.h>
   66 #include <sys/bio.h>
   67 #include <sys/buf.h>
   68 #include <sys/condvar.h>
   69 #ifdef KTRACE
   70 #include <sys/ktrace.h>
   71 #endif
   72 
   73 #include <security/audit/audit.h>
   74 
   75 static MALLOC_DEFINE(M_IOCTLOPS, "ioctlops", "ioctl data buffer");
   76 static MALLOC_DEFINE(M_SELECT, "select", "select() buffer");
   77 MALLOC_DEFINE(M_IOV, "iov", "large iov's");
   78 
   79 static int      pollout(struct pollfd *, struct pollfd *, u_int);
   80 static int      pollscan(struct thread *, struct pollfd *, u_int);
   81 static int      pollrescan(struct thread *);
   82 static int      selscan(struct thread *, fd_mask **, fd_mask **, int);
   83 static int      selrescan(struct thread *, fd_mask **, fd_mask **);
   84 static void     selfdalloc(struct thread *, void *);
   85 static void     selfdfree(struct seltd *, struct selfd *);
   86 static int      dofileread(struct thread *, int, struct file *, struct uio *,
   87                     off_t, int);
   88 static int      dofilewrite(struct thread *, int, struct file *, struct uio *,
   89                     off_t, int);
   90 static void     doselwakeup(struct selinfo *, int);
   91 static void     seltdinit(struct thread *);
   92 static int      seltdwait(struct thread *, int);
   93 static void     seltdclear(struct thread *);
   94 
   95 /*
   96  * One seltd per-thread allocated on demand as needed.
   97  *
   98  *      t - protected by st_mtx
   99  *      k - Only accessed by curthread or read-only
  100  */
  101 struct seltd {
  102         STAILQ_HEAD(, selfd)    st_selq;        /* (k) List of selfds. */
  103         struct selfd            *st_free1;      /* (k) free fd for read set. */
  104         struct selfd            *st_free2;      /* (k) free fd for write set. */
  105         struct mtx              st_mtx;         /* Protects struct seltd */
  106         struct cv               st_wait;        /* (t) Wait channel. */
  107         int                     st_flags;       /* (t) SELTD_ flags. */
  108 };
  109 
  110 #define SELTD_PENDING   0x0001                  /* We have pending events. */
  111 #define SELTD_RESCAN    0x0002                  /* Doing a rescan. */
  112 
  113 /*
  114  * One selfd allocated per-thread per-file-descriptor.
  115  *      f - protected by sf_mtx
  116  */
  117 struct selfd {
  118         STAILQ_ENTRY(selfd)     sf_link;        /* (k) fds owned by this td. */
  119         TAILQ_ENTRY(selfd)      sf_threads;     /* (f) fds on this selinfo. */
  120         struct selinfo          *sf_si;         /* (f) selinfo when linked. */
  121         struct mtx              *sf_mtx;        /* Pointer to selinfo mtx. */
  122         struct seltd            *sf_td;         /* (k) owning seltd. */
  123         void                    *sf_cookie;     /* (k) fd or pollfd. */
  124 };
  125 
  126 static uma_zone_t selfd_zone;
  127 static struct mtx_pool *mtxpool_select;
  128 
  129 #ifndef _SYS_SYSPROTO_H_
  130 struct read_args {
  131         int     fd;
  132         void    *buf;
  133         size_t  nbyte;
  134 };
  135 #endif
  136 int
  137 read(td, uap)
  138         struct thread *td;
  139         struct read_args *uap;
  140 {
  141         struct uio auio;
  142         struct iovec aiov;
  143         int error;
  144 
  145         if (uap->nbyte > INT_MAX)
  146                 return (EINVAL);
  147         aiov.iov_base = uap->buf;
  148         aiov.iov_len = uap->nbyte;
  149         auio.uio_iov = &aiov;
  150         auio.uio_iovcnt = 1;
  151         auio.uio_resid = uap->nbyte;
  152         auio.uio_segflg = UIO_USERSPACE;
  153         error = kern_readv(td, uap->fd, &auio);
  154         return(error);
  155 }
  156 
  157 /*
  158  * Positioned read system call
  159  */
  160 #ifndef _SYS_SYSPROTO_H_
  161 struct pread_args {
  162         int     fd;
  163         void    *buf;
  164         size_t  nbyte;
  165         int     pad;
  166         off_t   offset;
  167 };
  168 #endif
  169 int
  170 pread(td, uap)
  171         struct thread *td;
  172         struct pread_args *uap;
  173 {
  174         struct uio auio;
  175         struct iovec aiov;
  176         int error;
  177 
  178         if (uap->nbyte > INT_MAX)
  179                 return (EINVAL);
  180         aiov.iov_base = uap->buf;
  181         aiov.iov_len = uap->nbyte;
  182         auio.uio_iov = &aiov;
  183         auio.uio_iovcnt = 1;
  184         auio.uio_resid = uap->nbyte;
  185         auio.uio_segflg = UIO_USERSPACE;
  186         error = kern_preadv(td, uap->fd, &auio, uap->offset);
  187         return(error);
  188 }
  189 
  190 int
  191 freebsd6_pread(td, uap)
  192         struct thread *td;
  193         struct freebsd6_pread_args *uap;
  194 {
  195         struct pread_args oargs;
  196 
  197         oargs.fd = uap->fd;
  198         oargs.buf = uap->buf;
  199         oargs.nbyte = uap->nbyte;
  200         oargs.offset = uap->offset;
  201         return (pread(td, &oargs));
  202 }
  203 
  204 /*
  205  * Scatter read system call.
  206  */
  207 #ifndef _SYS_SYSPROTO_H_
  208 struct readv_args {
  209         int     fd;
  210         struct  iovec *iovp;
  211         u_int   iovcnt;
  212 };
  213 #endif
  214 int
  215 readv(struct thread *td, struct readv_args *uap)
  216 {
  217         struct uio *auio;
  218         int error;
  219 
  220         error = copyinuio(uap->iovp, uap->iovcnt, &auio);
  221         if (error)
  222                 return (error);
  223         error = kern_readv(td, uap->fd, auio);
  224         free(auio, M_IOV);
  225         return (error);
  226 }
  227 
  228 int
  229 kern_readv(struct thread *td, int fd, struct uio *auio)
  230 {
  231         struct file *fp;
  232         int error;
  233 
  234         error = fget_read(td, fd, &fp);
  235         if (error)
  236                 return (error);
  237         error = dofileread(td, fd, fp, auio, (off_t)-1, 0);
  238         fdrop(fp, td);
  239         return (error);
  240 }
  241 
  242 /*
  243  * Scatter positioned read system call.
  244  */
  245 #ifndef _SYS_SYSPROTO_H_
  246 struct preadv_args {
  247         int     fd;
  248         struct  iovec *iovp;
  249         u_int   iovcnt;
  250         off_t   offset;
  251 };
  252 #endif
  253 int
  254 preadv(struct thread *td, struct preadv_args *uap)
  255 {
  256         struct uio *auio;
  257         int error;
  258 
  259         error = copyinuio(uap->iovp, uap->iovcnt, &auio);
  260         if (error)
  261                 return (error);
  262         error = kern_preadv(td, uap->fd, auio, uap->offset);
  263         free(auio, M_IOV);
  264         return (error);
  265 }
  266 
  267 int
  268 kern_preadv(td, fd, auio, offset)
  269         struct thread *td;
  270         int fd;
  271         struct uio *auio;
  272         off_t offset;
  273 {
  274         struct file *fp;
  275         int error;
  276 
  277         error = fget_read(td, fd, &fp);
  278         if (error)
  279                 return (error);
  280         if (!(fp->f_ops->fo_flags & DFLAG_SEEKABLE))
  281                 error = ESPIPE;
  282         else if (offset < 0 && fp->f_vnode->v_type != VCHR)
  283                 error = EINVAL;
  284         else
  285                 error = dofileread(td, fd, fp, auio, offset, FOF_OFFSET);
  286         fdrop(fp, td);
  287         return (error);
  288 }
  289 
  290 /*
  291  * Common code for readv and preadv that reads data in
  292  * from a file using the passed in uio, offset, and flags.
  293  */
  294 static int
  295 dofileread(td, fd, fp, auio, offset, flags)
  296         struct thread *td;
  297         int fd;
  298         struct file *fp;
  299         struct uio *auio;
  300         off_t offset;
  301         int flags;
  302 {
  303         ssize_t cnt;
  304         int error;
  305 #ifdef KTRACE
  306         struct uio *ktruio = NULL;
  307 #endif
  308 
  309         /* Finish zero length reads right here */
  310         if (auio->uio_resid == 0) {
  311                 td->td_retval[0] = 0;
  312                 return(0);
  313         }
  314         auio->uio_rw = UIO_READ;
  315         auio->uio_offset = offset;
  316         auio->uio_td = td;
  317 #ifdef KTRACE
  318         if (KTRPOINT(td, KTR_GENIO)) 
  319                 ktruio = cloneuio(auio);
  320 #endif
  321         cnt = auio->uio_resid;
  322         if ((error = fo_read(fp, auio, td->td_ucred, flags, td))) {
  323                 if (auio->uio_resid != cnt && (error == ERESTART ||
  324                     error == EINTR || error == EWOULDBLOCK))
  325                         error = 0;
  326         }
  327         cnt -= auio->uio_resid;
  328 #ifdef KTRACE
  329         if (ktruio != NULL) {
  330                 ktruio->uio_resid = cnt;
  331                 ktrgenio(fd, UIO_READ, ktruio, error);
  332         }
  333 #endif
  334         td->td_retval[0] = cnt;
  335         return (error);
  336 }
  337 
  338 #ifndef _SYS_SYSPROTO_H_
  339 struct write_args {
  340         int     fd;
  341         const void *buf;
  342         size_t  nbyte;
  343 };
  344 #endif
  345 int
  346 write(td, uap)
  347         struct thread *td;
  348         struct write_args *uap;
  349 {
  350         struct uio auio;
  351         struct iovec aiov;
  352         int error;
  353 
  354         if (uap->nbyte > INT_MAX)
  355                 return (EINVAL);
  356         aiov.iov_base = (void *)(uintptr_t)uap->buf;
  357         aiov.iov_len = uap->nbyte;
  358         auio.uio_iov = &aiov;
  359         auio.uio_iovcnt = 1;
  360         auio.uio_resid = uap->nbyte;
  361         auio.uio_segflg = UIO_USERSPACE;
  362         error = kern_writev(td, uap->fd, &auio);
  363         return(error);
  364 }
  365 
  366 /*
  367  * Positioned write system call.
  368  */
  369 #ifndef _SYS_SYSPROTO_H_
  370 struct pwrite_args {
  371         int     fd;
  372         const void *buf;
  373         size_t  nbyte;
  374         int     pad;
  375         off_t   offset;
  376 };
  377 #endif
  378 int
  379 pwrite(td, uap)
  380         struct thread *td;
  381         struct pwrite_args *uap;
  382 {
  383         struct uio auio;
  384         struct iovec aiov;
  385         int error;
  386 
  387         if (uap->nbyte > INT_MAX)
  388                 return (EINVAL);
  389         aiov.iov_base = (void *)(uintptr_t)uap->buf;
  390         aiov.iov_len = uap->nbyte;
  391         auio.uio_iov = &aiov;
  392         auio.uio_iovcnt = 1;
  393         auio.uio_resid = uap->nbyte;
  394         auio.uio_segflg = UIO_USERSPACE;
  395         error = kern_pwritev(td, uap->fd, &auio, uap->offset);
  396         return(error);
  397 }
  398 
  399 int
  400 freebsd6_pwrite(td, uap)
  401         struct thread *td;
  402         struct freebsd6_pwrite_args *uap;
  403 {
  404         struct pwrite_args oargs;
  405 
  406         oargs.fd = uap->fd;
  407         oargs.buf = uap->buf;
  408         oargs.nbyte = uap->nbyte;
  409         oargs.offset = uap->offset;
  410         return (pwrite(td, &oargs));
  411 }
  412 
  413 /*
  414  * Gather write system call.
  415  */
  416 #ifndef _SYS_SYSPROTO_H_
  417 struct writev_args {
  418         int     fd;
  419         struct  iovec *iovp;
  420         u_int   iovcnt;
  421 };
  422 #endif
  423 int
  424 writev(struct thread *td, struct writev_args *uap)
  425 {
  426         struct uio *auio;
  427         int error;
  428 
  429         error = copyinuio(uap->iovp, uap->iovcnt, &auio);
  430         if (error)
  431                 return (error);
  432         error = kern_writev(td, uap->fd, auio);
  433         free(auio, M_IOV);
  434         return (error);
  435 }
  436 
  437 int
  438 kern_writev(struct thread *td, int fd, struct uio *auio)
  439 {
  440         struct file *fp;
  441         int error;
  442 
  443         error = fget_write(td, fd, &fp);
  444         if (error)
  445                 return (error);
  446         error = dofilewrite(td, fd, fp, auio, (off_t)-1, 0);
  447         fdrop(fp, td);
  448         return (error);
  449 }
  450 
  451 /*
  452  * Gather positioned write system call.
  453  */
  454 #ifndef _SYS_SYSPROTO_H_
  455 struct pwritev_args {
  456         int     fd;
  457         struct  iovec *iovp;
  458         u_int   iovcnt;
  459         off_t   offset;
  460 };
  461 #endif
  462 int
  463 pwritev(struct thread *td, struct pwritev_args *uap)
  464 {
  465         struct uio *auio;
  466         int error;
  467 
  468         error = copyinuio(uap->iovp, uap->iovcnt, &auio);
  469         if (error)
  470                 return (error);
  471         error = kern_pwritev(td, uap->fd, auio, uap->offset);
  472         free(auio, M_IOV);
  473         return (error);
  474 }
  475 
  476 int
  477 kern_pwritev(td, fd, auio, offset)
  478         struct thread *td;
  479         struct uio *auio;
  480         int fd;
  481         off_t offset;
  482 {
  483         struct file *fp;
  484         int error;
  485 
  486         error = fget_write(td, fd, &fp);
  487         if (error)
  488                 return (error);
  489         if (!(fp->f_ops->fo_flags & DFLAG_SEEKABLE))
  490                 error = ESPIPE;
  491         else if (offset < 0 && fp->f_vnode->v_type != VCHR)
  492                 error = EINVAL;
  493         else
  494                 error = dofilewrite(td, fd, fp, auio, offset, FOF_OFFSET);
  495         fdrop(fp, td);
  496         return (error);
  497 }
  498 
  499 /*
  500  * Common code for writev and pwritev that writes data to
  501  * a file using the passed in uio, offset, and flags.
  502  */
  503 static int
  504 dofilewrite(td, fd, fp, auio, offset, flags)
  505         struct thread *td;
  506         int fd;
  507         struct file *fp;
  508         struct uio *auio;
  509         off_t offset;
  510         int flags;
  511 {
  512         ssize_t cnt;
  513         int error;
  514 #ifdef KTRACE
  515         struct uio *ktruio = NULL;
  516 #endif
  517 
  518         auio->uio_rw = UIO_WRITE;
  519         auio->uio_td = td;
  520         auio->uio_offset = offset;
  521 #ifdef KTRACE
  522         if (KTRPOINT(td, KTR_GENIO))
  523                 ktruio = cloneuio(auio);
  524 #endif
  525         cnt = auio->uio_resid;
  526         if (fp->f_type == DTYPE_VNODE)
  527                 bwillwrite();
  528         if ((error = fo_write(fp, auio, td->td_ucred, flags, td))) {
  529                 if (auio->uio_resid != cnt && (error == ERESTART ||
  530                     error == EINTR || error == EWOULDBLOCK))
  531                         error = 0;
  532                 /* Socket layer is responsible for issuing SIGPIPE. */
  533                 if (fp->f_type != DTYPE_SOCKET && error == EPIPE) {
  534                         PROC_LOCK(td->td_proc);
  535                         psignal(td->td_proc, SIGPIPE);
  536                         PROC_UNLOCK(td->td_proc);
  537                 }
  538         }
  539         cnt -= auio->uio_resid;
  540 #ifdef KTRACE
  541         if (ktruio != NULL) {
  542                 ktruio->uio_resid = cnt;
  543                 ktrgenio(fd, UIO_WRITE, ktruio, error);
  544         }
  545 #endif
  546         td->td_retval[0] = cnt;
  547         return (error);
  548 }
  549 
  550 /*
  551  * Truncate a file given a file descriptor.
  552  *
  553  * Can't use fget_write() here, since must return EINVAL and not EBADF if the
  554  * descriptor isn't writable.
  555  */
  556 int
  557 kern_ftruncate(td, fd, length)
  558         struct thread *td;
  559         int fd;
  560         off_t length;
  561 {
  562         struct file *fp;
  563         int error;
  564 
  565         AUDIT_ARG_FD(fd);
  566         if (length < 0)
  567                 return (EINVAL);
  568         error = fget(td, fd, &fp);
  569         if (error)
  570                 return (error);
  571         AUDIT_ARG_FILE(td->td_proc, fp);
  572         if (!(fp->f_flag & FWRITE)) {
  573                 fdrop(fp, td);
  574                 return (EINVAL);
  575         }
  576         error = fo_truncate(fp, length, td->td_ucred, td);
  577         fdrop(fp, td);
  578         return (error);
  579 }
  580 
  581 #ifndef _SYS_SYSPROTO_H_
  582 struct ftruncate_args {
  583         int     fd;
  584         int     pad;
  585         off_t   length;
  586 };
  587 #endif
  588 int
  589 ftruncate(td, uap)
  590         struct thread *td;
  591         struct ftruncate_args *uap;
  592 {
  593 
  594         return (kern_ftruncate(td, uap->fd, uap->length));
  595 }
  596 
  597 #if defined(COMPAT_43)
  598 #ifndef _SYS_SYSPROTO_H_
  599 struct oftruncate_args {
  600         int     fd;
  601         long    length;
  602 };
  603 #endif
  604 int
  605 oftruncate(td, uap)
  606         struct thread *td;
  607         struct oftruncate_args *uap;
  608 {
  609 
  610         return (kern_ftruncate(td, uap->fd, uap->length));
  611 }
  612 #endif /* COMPAT_43 */
  613 
  614 #ifndef _SYS_SYSPROTO_H_
  615 struct ioctl_args {
  616         int     fd;
  617         u_long  com;
  618         caddr_t data;
  619 };
  620 #endif
  621 /* ARGSUSED */
  622 int
  623 ioctl(struct thread *td, struct ioctl_args *uap)
  624 {
  625         u_long com;
  626         int arg, error;
  627         u_int size;
  628         caddr_t data;
  629 
  630         if (uap->com > 0xffffffff) {
  631                 printf(
  632                     "WARNING pid %d (%s): ioctl sign-extension ioctl %lx\n",
  633                     td->td_proc->p_pid, td->td_name, uap->com);
  634                 uap->com &= 0xffffffff;
  635         }
  636         com = uap->com;
  637 
  638         /*
  639          * Interpret high order word to find amount of data to be
  640          * copied to/from the user's address space.
  641          */
  642         size = IOCPARM_LEN(com);
  643         if ((size > IOCPARM_MAX) ||
  644             ((com & (IOC_VOID  | IOC_IN | IOC_OUT)) == 0) ||
  645 #if defined(COMPAT_FREEBSD5) || defined(COMPAT_FREEBSD4) || defined(COMPAT_43)
  646             ((com & IOC_OUT) && size == 0) ||
  647 #else
  648             ((com & (IOC_IN | IOC_OUT)) && size == 0) ||
  649 #endif
  650             ((com & IOC_VOID) && size > 0 && size != sizeof(int)))
  651                 return (ENOTTY);
  652 
  653         if (size > 0) {
  654                 if (com & IOC_VOID) {
  655                         /* Integer argument. */
  656                         arg = (intptr_t)uap->data;
  657                         data = (void *)&arg;
  658                         size = 0;
  659                 } else
  660                         data = malloc((u_long)size, M_IOCTLOPS, M_WAITOK);
  661         } else
  662                 data = (void *)&uap->data;
  663         if (com & IOC_IN) {
  664                 error = copyin(uap->data, data, (u_int)size);
  665                 if (error) {
  666                         if (size > 0)
  667                                 free(data, M_IOCTLOPS);
  668                         return (error);
  669                 }
  670         } else if (com & IOC_OUT) {
  671                 /*
  672                  * Zero the buffer so the user always
  673                  * gets back something deterministic.
  674                  */
  675                 bzero(data, size);
  676         }
  677 
  678         error = kern_ioctl(td, uap->fd, com, data);
  679 
  680         if (error == 0 && (com & IOC_OUT))
  681                 error = copyout(data, uap->data, (u_int)size);
  682 
  683         if (size > 0)
  684                 free(data, M_IOCTLOPS);
  685         return (error);
  686 }
  687 
  688 int
  689 kern_ioctl(struct thread *td, int fd, u_long com, caddr_t data)
  690 {
  691         struct file *fp;
  692         struct filedesc *fdp;
  693         int error;
  694         int tmp;
  695 
  696         AUDIT_ARG_FD(fd);
  697         AUDIT_ARG_CMD(com);
  698         if ((error = fget(td, fd, &fp)) != 0)
  699                 return (error);
  700         if ((fp->f_flag & (FREAD | FWRITE)) == 0) {
  701                 fdrop(fp, td);
  702                 return (EBADF);
  703         }
  704         fdp = td->td_proc->p_fd;
  705         switch (com) {
  706         case FIONCLEX:
  707                 FILEDESC_XLOCK(fdp);
  708                 fdp->fd_ofileflags[fd] &= ~UF_EXCLOSE;
  709                 FILEDESC_XUNLOCK(fdp);
  710                 goto out;
  711         case FIOCLEX:
  712                 FILEDESC_XLOCK(fdp);
  713                 fdp->fd_ofileflags[fd] |= UF_EXCLOSE;
  714                 FILEDESC_XUNLOCK(fdp);
  715                 goto out;
  716         case FIONBIO:
  717                 if ((tmp = *(int *)data))
  718                         atomic_set_int(&fp->f_flag, FNONBLOCK);
  719                 else
  720                         atomic_clear_int(&fp->f_flag, FNONBLOCK);
  721                 data = (void *)&tmp;
  722                 break;
  723         case FIOASYNC:
  724                 if ((tmp = *(int *)data))
  725                         atomic_set_int(&fp->f_flag, FASYNC);
  726                 else
  727                         atomic_clear_int(&fp->f_flag, FASYNC);
  728                 data = (void *)&tmp;
  729                 break;
  730         }
  731 
  732         error = fo_ioctl(fp, com, data, td->td_ucred, td);
  733 out:
  734         fdrop(fp, td);
  735         return (error);
  736 }
  737 
  738 int
  739 poll_no_poll(int events)
  740 {
  741         /*
  742          * Return true for read/write.  If the user asked for something
  743          * special, return POLLNVAL, so that clients have a way of
  744          * determining reliably whether or not the extended
  745          * functionality is present without hard-coding knowledge
  746          * of specific filesystem implementations.
  747          */
  748         if (events & ~POLLSTANDARD)
  749                 return (POLLNVAL);
  750 
  751         return (events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
  752 }
  753 
  754 #ifndef _SYS_SYSPROTO_H_
  755 struct select_args {
  756         int     nd;
  757         fd_set  *in, *ou, *ex;
  758         struct  timeval *tv;
  759 };
  760 #endif
  761 int
  762 select(td, uap)
  763         register struct thread *td;
  764         register struct select_args *uap;
  765 {
  766         struct timeval tv, *tvp;
  767         int error;
  768 
  769         if (uap->tv != NULL) {
  770                 error = copyin(uap->tv, &tv, sizeof(tv));
  771                 if (error)
  772                         return (error);
  773                 tvp = &tv;
  774         } else
  775                 tvp = NULL;
  776 
  777         return (kern_select(td, uap->nd, uap->in, uap->ou, uap->ex, tvp,
  778             NFDBITS));
  779 }
  780 
  781 int
  782 kern_select(struct thread *td, int nd, fd_set *fd_in, fd_set *fd_ou,
  783     fd_set *fd_ex, struct timeval *tvp, int abi_nfdbits)
  784 {
  785         struct filedesc *fdp;
  786         /*
  787          * The magic 2048 here is chosen to be just enough for FD_SETSIZE
  788          * infds with the new FD_SETSIZE of 1024, and more than enough for
  789          * FD_SETSIZE infds, outfds and exceptfds with the old FD_SETSIZE
  790          * of 256.
  791          */
  792         fd_mask s_selbits[howmany(2048, NFDBITS)];
  793         fd_mask *ibits[3], *obits[3], *selbits, *sbp;
  794         struct timeval atv, rtv, ttv;
  795         int error, timo;
  796         u_int nbufbytes, ncpbytes, ncpubytes, nfdbits;
  797 
  798         if (nd < 0)
  799                 return (EINVAL);
  800         fdp = td->td_proc->p_fd;
  801         if (nd > fdp->fd_lastfile + 1)
  802                 nd = fdp->fd_lastfile + 1;
  803 
  804         /*
  805          * Allocate just enough bits for the non-null fd_sets.  Use the
  806          * preallocated auto buffer if possible.
  807          */
  808         nfdbits = roundup(nd, NFDBITS);
  809         ncpbytes = nfdbits / NBBY;
  810         ncpubytes = roundup(nd, abi_nfdbits) / NBBY;
  811         nbufbytes = 0;
  812         if (fd_in != NULL)
  813                 nbufbytes += 2 * ncpbytes;
  814         if (fd_ou != NULL)
  815                 nbufbytes += 2 * ncpbytes;
  816         if (fd_ex != NULL)
  817                 nbufbytes += 2 * ncpbytes;
  818         if (nbufbytes <= sizeof s_selbits)
  819                 selbits = &s_selbits[0];
  820         else
  821                 selbits = malloc(nbufbytes, M_SELECT, M_WAITOK);
  822 
  823         /*
  824          * Assign pointers into the bit buffers and fetch the input bits.
  825          * Put the output buffers together so that they can be bzeroed
  826          * together.
  827          */
  828         sbp = selbits;
  829 #define getbits(name, x) \
  830         do {                                                            \
  831                 if (name == NULL)                                       \
  832                         ibits[x] = NULL;                                \
  833                 else {                                                  \
  834                         ibits[x] = sbp + nbufbytes / 2 / sizeof *sbp;   \
  835                         obits[x] = sbp;                                 \
  836                         sbp += ncpbytes / sizeof *sbp;                  \
  837                         error = copyin(name, ibits[x], ncpubytes);      \
  838                         if (error != 0)                                 \
  839                                 goto done;                              \
  840                         bzero((char *)ibits[x] + ncpubytes,             \
  841                             ncpbytes - ncpubytes);                      \
  842                 }                                                       \
  843         } while (0)
  844         getbits(fd_in, 0);
  845         getbits(fd_ou, 1);
  846         getbits(fd_ex, 2);
  847 #undef  getbits
  848         if (nbufbytes != 0)
  849                 bzero(selbits, nbufbytes / 2);
  850 
  851         if (tvp != NULL) {
  852                 atv = *tvp;
  853                 if (itimerfix(&atv)) {
  854                         error = EINVAL;
  855                         goto done;
  856                 }
  857                 getmicrouptime(&rtv);
  858                 timevaladd(&atv, &rtv);
  859         } else {
  860                 atv.tv_sec = 0;
  861                 atv.tv_usec = 0;
  862         }
  863         timo = 0;
  864         seltdinit(td);
  865         /* Iterate until the timeout expires or descriptors become ready. */
  866         for (;;) {
  867                 error = selscan(td, ibits, obits, nd);
  868                 if (error || td->td_retval[0] != 0)
  869                         break;
  870                 if (atv.tv_sec || atv.tv_usec) {
  871                         getmicrouptime(&rtv);
  872                         if (timevalcmp(&rtv, &atv, >=))
  873                                 break;
  874                         ttv = atv;
  875                         timevalsub(&ttv, &rtv);
  876                         timo = ttv.tv_sec > 24 * 60 * 60 ?
  877                             24 * 60 * 60 * hz : tvtohz(&ttv);
  878                 }
  879                 error = seltdwait(td, timo);
  880                 if (error)
  881                         break;
  882                 error = selrescan(td, ibits, obits);
  883                 if (error || td->td_retval[0] != 0)
  884                         break;
  885         }
  886         seltdclear(td);
  887 
  888 done:
  889         /* select is not restarted after signals... */
  890         if (error == ERESTART)
  891                 error = EINTR;
  892         if (error == EWOULDBLOCK)
  893                 error = 0;
  894 #define putbits(name, x) \
  895         if (name && (error2 = copyout(obits[x], name, ncpubytes))) \
  896                 error = error2;
  897         if (error == 0) {
  898                 int error2;
  899 
  900                 putbits(fd_in, 0);
  901                 putbits(fd_ou, 1);
  902                 putbits(fd_ex, 2);
  903 #undef putbits
  904         }
  905         if (selbits != &s_selbits[0])
  906                 free(selbits, M_SELECT);
  907 
  908         return (error);
  909 }
  910 /* 
  911  * Convert a select bit set to poll flags.
  912  *
  913  * The backend always returns POLLHUP/POLLERR if appropriate and we
  914  * return this as a set bit in any set.
  915  */
  916 static int select_flags[3] = {
  917     POLLRDNORM | POLLHUP | POLLERR,
  918     POLLWRNORM | POLLHUP | POLLERR,
  919     POLLRDBAND | POLLHUP | POLLERR
  920 };
  921 
  922 /*
  923  * Compute the fo_poll flags required for a fd given by the index and
  924  * bit position in the fd_mask array.
  925  */
  926 static __inline int
  927 selflags(fd_mask **ibits, int idx, fd_mask bit)
  928 {
  929         int flags;
  930         int msk;
  931 
  932         flags = 0;
  933         for (msk = 0; msk < 3; msk++) {
  934                 if (ibits[msk] == NULL)
  935                         continue;
  936                 if ((ibits[msk][idx] & bit) == 0)
  937                         continue;
  938                 flags |= select_flags[msk];
  939         }
  940         return (flags);
  941 }
  942 
  943 /*
  944  * Set the appropriate output bits given a mask of fired events and the
  945  * input bits originally requested.
  946  */
  947 static __inline int
  948 selsetbits(fd_mask **ibits, fd_mask **obits, int idx, fd_mask bit, int events)
  949 {
  950         int msk;
  951         int n;
  952 
  953         n = 0;
  954         for (msk = 0; msk < 3; msk++) {
  955                 if ((events & select_flags[msk]) == 0)
  956                         continue;
  957                 if (ibits[msk] == NULL)
  958                         continue;
  959                 if ((ibits[msk][idx] & bit) == 0)
  960                         continue;
  961                 /*
  962                  * XXX Check for a duplicate set.  This can occur because a
  963                  * socket calls selrecord() twice for each poll() call
  964                  * resulting in two selfds per real fd.  selrescan() will
  965                  * call selsetbits twice as a result.
  966                  */
  967                 if ((obits[msk][idx] & bit) != 0)
  968                         continue;
  969                 obits[msk][idx] |= bit;
  970                 n++;
  971         }
  972 
  973         return (n);
  974 }
  975 
  976 /*
  977  * Traverse the list of fds attached to this thread's seltd and check for
  978  * completion.
  979  */
  980 static int
  981 selrescan(struct thread *td, fd_mask **ibits, fd_mask **obits)
  982 {
  983         struct filedesc *fdp;
  984         struct selinfo *si;
  985         struct seltd *stp;
  986         struct selfd *sfp;
  987         struct selfd *sfn;
  988         struct file *fp;
  989         fd_mask bit;
  990         int fd, ev, n, idx;
  991 
  992         fdp = td->td_proc->p_fd;
  993         stp = td->td_sel;
  994         n = 0;
  995         STAILQ_FOREACH_SAFE(sfp, &stp->st_selq, sf_link, sfn) {
  996                 fd = (int)(uintptr_t)sfp->sf_cookie;
  997                 si = sfp->sf_si;
  998                 selfdfree(stp, sfp);
  999                 /* If the selinfo wasn't cleared the event didn't fire. */
 1000                 if (si != NULL)
 1001                         continue;
 1002                 if ((fp = fget_unlocked(fdp, fd)) == NULL)
 1003                         return (EBADF);
 1004                 idx = fd / NFDBITS;
 1005                 bit = (fd_mask)1 << (fd % NFDBITS);
 1006                 ev = fo_poll(fp, selflags(ibits, idx, bit), td->td_ucred, td);
 1007                 fdrop(fp, td);
 1008                 if (ev != 0)
 1009                         n += selsetbits(ibits, obits, idx, bit, ev);
 1010         }
 1011         stp->st_flags = 0;
 1012         td->td_retval[0] = n;
 1013         return (0);
 1014 }
 1015 
 1016 /*
 1017  * Perform the initial filedescriptor scan and register ourselves with
 1018  * each selinfo.
 1019  */
 1020 static int
 1021 selscan(td, ibits, obits, nfd)
 1022         struct thread *td;
 1023         fd_mask **ibits, **obits;
 1024         int nfd;
 1025 {
 1026         struct filedesc *fdp;
 1027         struct file *fp;
 1028         fd_mask bit;
 1029         int ev, flags, end, fd;
 1030         int n, idx;
 1031 
 1032         fdp = td->td_proc->p_fd;
 1033         n = 0;
 1034         for (idx = 0, fd = 0; fd < nfd; idx++) {
 1035                 end = imin(fd + NFDBITS, nfd);
 1036                 for (bit = 1; fd < end; bit <<= 1, fd++) {
 1037                         /* Compute the list of events we're interested in. */
 1038                         flags = selflags(ibits, idx, bit);
 1039                         if (flags == 0)
 1040                                 continue;
 1041                         if ((fp = fget_unlocked(fdp, fd)) == NULL)
 1042                                 return (EBADF);
 1043                         selfdalloc(td, (void *)(uintptr_t)fd);
 1044                         ev = fo_poll(fp, flags, td->td_ucred, td);
 1045                         fdrop(fp, td);
 1046                         if (ev != 0)
 1047                                 n += selsetbits(ibits, obits, idx, bit, ev);
 1048                 }
 1049         }
 1050 
 1051         td->td_retval[0] = n;
 1052         return (0);
 1053 }
 1054 
 1055 #ifndef _SYS_SYSPROTO_H_
 1056 struct poll_args {
 1057         struct pollfd *fds;
 1058         u_int   nfds;
 1059         int     timeout;
 1060 };
 1061 #endif
 1062 int
 1063 poll(td, uap)
 1064         struct thread *td;
 1065         struct poll_args *uap;
 1066 {
 1067         struct pollfd *bits;
 1068         struct pollfd smallbits[32];
 1069         struct timeval atv, rtv, ttv;
 1070         int error = 0, timo;
 1071         u_int nfds;
 1072         size_t ni;
 1073 
 1074         nfds = uap->nfds;
 1075         if (nfds > maxfilesperproc && nfds > FD_SETSIZE) 
 1076                 return (EINVAL);
 1077         ni = nfds * sizeof(struct pollfd);
 1078         if (ni > sizeof(smallbits))
 1079                 bits = malloc(ni, M_TEMP, M_WAITOK);
 1080         else
 1081                 bits = smallbits;
 1082         error = copyin(uap->fds, bits, ni);
 1083         if (error)
 1084                 goto done;
 1085         if (uap->timeout != INFTIM) {
 1086                 atv.tv_sec = uap->timeout / 1000;
 1087                 atv.tv_usec = (uap->timeout % 1000) * 1000;
 1088                 if (itimerfix(&atv)) {
 1089                         error = EINVAL;
 1090                         goto done;
 1091                 }
 1092                 getmicrouptime(&rtv);
 1093                 timevaladd(&atv, &rtv);
 1094         } else {
 1095                 atv.tv_sec = 0;
 1096                 atv.tv_usec = 0;
 1097         }
 1098         timo = 0;
 1099         seltdinit(td);
 1100         /* Iterate until the timeout expires or descriptors become ready. */
 1101         for (;;) {
 1102                 error = pollscan(td, bits, nfds);
 1103                 if (error || td->td_retval[0] != 0)
 1104                         break;
 1105                 if (atv.tv_sec || atv.tv_usec) {
 1106                         getmicrouptime(&rtv);
 1107                         if (timevalcmp(&rtv, &atv, >=))
 1108                                 break;
 1109                         ttv = atv;
 1110                         timevalsub(&ttv, &rtv);
 1111                         timo = ttv.tv_sec > 24 * 60 * 60 ?
 1112                             24 * 60 * 60 * hz : tvtohz(&ttv);
 1113                 }
 1114                 error = seltdwait(td, timo);
 1115                 if (error)
 1116                         break;
 1117                 error = pollrescan(td);
 1118                 if (error || td->td_retval[0] != 0)
 1119                         break;
 1120         }
 1121         seltdclear(td);
 1122 
 1123 done:
 1124         /* poll is not restarted after signals... */
 1125         if (error == ERESTART)
 1126                 error = EINTR;
 1127         if (error == EWOULDBLOCK)
 1128                 error = 0;
 1129         if (error == 0) {
 1130                 error = pollout(bits, uap->fds, nfds);
 1131                 if (error)
 1132                         goto out;
 1133         }
 1134 out:
 1135         if (ni > sizeof(smallbits))
 1136                 free(bits, M_TEMP);
 1137         return (error);
 1138 }
 1139 
 1140 static int
 1141 pollrescan(struct thread *td)
 1142 {
 1143         struct seltd *stp;
 1144         struct selfd *sfp;
 1145         struct selfd *sfn;
 1146         struct selinfo *si;
 1147         struct filedesc *fdp;
 1148         struct file *fp;
 1149         struct pollfd *fd;
 1150         int n;
 1151 
 1152         n = 0;
 1153         fdp = td->td_proc->p_fd;
 1154         stp = td->td_sel;
 1155         FILEDESC_SLOCK(fdp);
 1156         STAILQ_FOREACH_SAFE(sfp, &stp->st_selq, sf_link, sfn) {
 1157                 fd = (struct pollfd *)sfp->sf_cookie;
 1158                 si = sfp->sf_si;
 1159                 selfdfree(stp, sfp);
 1160                 /* If the selinfo wasn't cleared the event didn't fire. */
 1161                 if (si != NULL)
 1162                         continue;
 1163                 fp = fdp->fd_ofiles[fd->fd];
 1164                 if (fp == NULL) {
 1165                         fd->revents = POLLNVAL;
 1166                         n++;
 1167                         continue;
 1168                 }
 1169                 /*
 1170                  * Note: backend also returns POLLHUP and
 1171                  * POLLERR if appropriate.
 1172                  */
 1173                 fd->revents = fo_poll(fp, fd->events, td->td_ucred, td);
 1174                 if (fd->revents != 0)
 1175                         n++;
 1176         }
 1177         FILEDESC_SUNLOCK(fdp);
 1178         stp->st_flags = 0;
 1179         td->td_retval[0] = n;
 1180         return (0);
 1181 }
 1182 
 1183 
 1184 static int
 1185 pollout(fds, ufds, nfd)
 1186         struct pollfd *fds;
 1187         struct pollfd *ufds;
 1188         u_int nfd;
 1189 {
 1190         int error = 0;
 1191         u_int i = 0;
 1192 
 1193         for (i = 0; i < nfd; i++) {
 1194                 error = copyout(&fds->revents, &ufds->revents,
 1195                     sizeof(ufds->revents));
 1196                 if (error)
 1197                         return (error);
 1198                 fds++;
 1199                 ufds++;
 1200         }
 1201         return (0);
 1202 }
 1203 
 1204 static int
 1205 pollscan(td, fds, nfd)
 1206         struct thread *td;
 1207         struct pollfd *fds;
 1208         u_int nfd;
 1209 {
 1210         struct filedesc *fdp = td->td_proc->p_fd;
 1211         int i;
 1212         struct file *fp;
 1213         int n = 0;
 1214 
 1215         FILEDESC_SLOCK(fdp);
 1216         for (i = 0; i < nfd; i++, fds++) {
 1217                 if (fds->fd >= fdp->fd_nfiles) {
 1218                         fds->revents = POLLNVAL;
 1219                         n++;
 1220                 } else if (fds->fd < 0) {
 1221                         fds->revents = 0;
 1222                 } else {
 1223                         fp = fdp->fd_ofiles[fds->fd];
 1224                         if (fp == NULL) {
 1225                                 fds->revents = POLLNVAL;
 1226                                 n++;
 1227                         } else {
 1228                                 /*
 1229                                  * Note: backend also returns POLLHUP and
 1230                                  * POLLERR if appropriate.
 1231                                  */
 1232                                 selfdalloc(td, fds);
 1233                                 fds->revents = fo_poll(fp, fds->events,
 1234                                     td->td_ucred, td);
 1235                                 /*
 1236                                  * POSIX requires POLLOUT to be never
 1237                                  * set simultaneously with POLLHUP.
 1238                                  */
 1239                                 if ((fds->revents & POLLHUP) != 0)
 1240                                         fds->revents &= ~POLLOUT;
 1241 
 1242                                 if (fds->revents != 0)
 1243                                         n++;
 1244                         }
 1245                 }
 1246         }
 1247         FILEDESC_SUNLOCK(fdp);
 1248         td->td_retval[0] = n;
 1249         return (0);
 1250 }
 1251 
 1252 /*
 1253  * OpenBSD poll system call.
 1254  *
 1255  * XXX this isn't quite a true representation..  OpenBSD uses select ops.
 1256  */
 1257 #ifndef _SYS_SYSPROTO_H_
 1258 struct openbsd_poll_args {
 1259         struct pollfd *fds;
 1260         u_int   nfds;
 1261         int     timeout;
 1262 };
 1263 #endif
 1264 int
 1265 openbsd_poll(td, uap)
 1266         register struct thread *td;
 1267         register struct openbsd_poll_args *uap;
 1268 {
 1269         return (poll(td, (struct poll_args *)uap));
 1270 }
 1271 
 1272 /*
 1273  * XXX This was created specifically to support netncp and netsmb.  This
 1274  * allows the caller to specify a socket to wait for events on.  It returns
 1275  * 0 if any events matched and an error otherwise.  There is no way to
 1276  * determine which events fired.
 1277  */
 1278 int
 1279 selsocket(struct socket *so, int events, struct timeval *tvp, struct thread *td)
 1280 {
 1281         struct timeval atv, rtv, ttv;
 1282         int error, timo;
 1283 
 1284         if (tvp != NULL) {
 1285                 atv = *tvp;
 1286                 if (itimerfix(&atv))
 1287                         return (EINVAL);
 1288                 getmicrouptime(&rtv);
 1289                 timevaladd(&atv, &rtv);
 1290         } else {
 1291                 atv.tv_sec = 0;
 1292                 atv.tv_usec = 0;
 1293         }
 1294 
 1295         timo = 0;
 1296         seltdinit(td);
 1297         /*
 1298          * Iterate until the timeout expires or the socket becomes ready.
 1299          */
 1300         for (;;) {
 1301                 selfdalloc(td, NULL);
 1302                 error = sopoll(so, events, NULL, td);
 1303                 /* error here is actually the ready events. */
 1304                 if (error)
 1305                         return (0);
 1306                 if (atv.tv_sec || atv.tv_usec) {
 1307                         getmicrouptime(&rtv);
 1308                         if (timevalcmp(&rtv, &atv, >=)) {
 1309                                 seltdclear(td);
 1310                                 return (EWOULDBLOCK);
 1311                         }
 1312                         ttv = atv;
 1313                         timevalsub(&ttv, &rtv);
 1314                         timo = ttv.tv_sec > 24 * 60 * 60 ?
 1315                             24 * 60 * 60 * hz : tvtohz(&ttv);
 1316                 }
 1317                 error = seltdwait(td, timo);
 1318                 seltdclear(td);
 1319                 if (error)
 1320                         break;
 1321         }
 1322         /* XXX Duplicates ncp/smb behavior. */
 1323         if (error == ERESTART)
 1324                 error = 0;
 1325         return (error);
 1326 }
 1327 
 1328 /*
 1329  * Preallocate two selfds associated with 'cookie'.  Some fo_poll routines
 1330  * have two select sets, one for read and another for write.
 1331  */
 1332 static void
 1333 selfdalloc(struct thread *td, void *cookie)
 1334 {
 1335         struct seltd *stp;
 1336 
 1337         stp = td->td_sel;
 1338         if (stp->st_free1 == NULL)
 1339                 stp->st_free1 = uma_zalloc(selfd_zone, M_WAITOK|M_ZERO);
 1340         stp->st_free1->sf_td = stp;
 1341         stp->st_free1->sf_cookie = cookie;
 1342         if (stp->st_free2 == NULL)
 1343                 stp->st_free2 = uma_zalloc(selfd_zone, M_WAITOK|M_ZERO);
 1344         stp->st_free2->sf_td = stp;
 1345         stp->st_free2->sf_cookie = cookie;
 1346 }
 1347 
 1348 static void
 1349 selfdfree(struct seltd *stp, struct selfd *sfp)
 1350 {
 1351         STAILQ_REMOVE(&stp->st_selq, sfp, selfd, sf_link);
 1352         mtx_lock(sfp->sf_mtx);
 1353         if (sfp->sf_si)
 1354                 TAILQ_REMOVE(&sfp->sf_si->si_tdlist, sfp, sf_threads);
 1355         mtx_unlock(sfp->sf_mtx);
 1356         uma_zfree(selfd_zone, sfp);
 1357 }
 1358 
 1359 /*
 1360  * Record a select request.
 1361  */
 1362 void
 1363 selrecord(selector, sip)
 1364         struct thread *selector;
 1365         struct selinfo *sip;
 1366 {
 1367         struct selfd *sfp;
 1368         struct seltd *stp;
 1369         struct mtx *mtxp;
 1370 
 1371         stp = selector->td_sel;
 1372         /*
 1373          * Don't record when doing a rescan.
 1374          */
 1375         if (stp->st_flags & SELTD_RESCAN)
 1376                 return;
 1377         /*
 1378          * Grab one of the preallocated descriptors.
 1379          */
 1380         sfp = NULL;
 1381         if ((sfp = stp->st_free1) != NULL)
 1382                 stp->st_free1 = NULL;
 1383         else if ((sfp = stp->st_free2) != NULL)
 1384                 stp->st_free2 = NULL;
 1385         else
 1386                 panic("selrecord: No free selfd on selq");
 1387         mtxp = sip->si_mtx;
 1388         if (mtxp == NULL)
 1389                 mtxp = mtx_pool_find(mtxpool_select, sip);
 1390         /*
 1391          * Initialize the sfp and queue it in the thread.
 1392          */
 1393         sfp->sf_si = sip;
 1394         sfp->sf_mtx = mtxp;
 1395         STAILQ_INSERT_TAIL(&stp->st_selq, sfp, sf_link);
 1396         /*
 1397          * Now that we've locked the sip, check for initialization.
 1398          */
 1399         mtx_lock(mtxp);
 1400         if (sip->si_mtx == NULL) {
 1401                 sip->si_mtx = mtxp;
 1402                 TAILQ_INIT(&sip->si_tdlist);
 1403         }
 1404         /*
 1405          * Add this thread to the list of selfds listening on this selinfo.
 1406          */
 1407         TAILQ_INSERT_TAIL(&sip->si_tdlist, sfp, sf_threads);
 1408         mtx_unlock(sip->si_mtx);
 1409 }
 1410 
 1411 /* Wake up a selecting thread. */
 1412 void
 1413 selwakeup(sip)
 1414         struct selinfo *sip;
 1415 {
 1416         doselwakeup(sip, -1);
 1417 }
 1418 
 1419 /* Wake up a selecting thread, and set its priority. */
 1420 void
 1421 selwakeuppri(sip, pri)
 1422         struct selinfo *sip;
 1423         int pri;
 1424 {
 1425         doselwakeup(sip, pri);
 1426 }
 1427 
 1428 /*
 1429  * Do a wakeup when a selectable event occurs.
 1430  */
 1431 static void
 1432 doselwakeup(sip, pri)
 1433         struct selinfo *sip;
 1434         int pri;
 1435 {
 1436         struct selfd *sfp;
 1437         struct selfd *sfn;
 1438         struct seltd *stp;
 1439 
 1440         /* If it's not initialized there can't be any waiters. */
 1441         if (sip->si_mtx == NULL)
 1442                 return;
 1443         /*
 1444          * Locking the selinfo locks all selfds associated with it.
 1445          */
 1446         mtx_lock(sip->si_mtx);
 1447         TAILQ_FOREACH_SAFE(sfp, &sip->si_tdlist, sf_threads, sfn) {
 1448                 /*
 1449                  * Once we remove this sfp from the list and clear the
 1450                  * sf_si seltdclear will know to ignore this si.
 1451                  */
 1452                 TAILQ_REMOVE(&sip->si_tdlist, sfp, sf_threads);
 1453                 sfp->sf_si = NULL;
 1454                 stp = sfp->sf_td;
 1455                 mtx_lock(&stp->st_mtx);
 1456                 stp->st_flags |= SELTD_PENDING;
 1457                 cv_broadcastpri(&stp->st_wait, pri);
 1458                 mtx_unlock(&stp->st_mtx);
 1459         }
 1460         mtx_unlock(sip->si_mtx);
 1461 }
 1462 
 1463 static void
 1464 seltdinit(struct thread *td)
 1465 {
 1466         struct seltd *stp;
 1467 
 1468         if ((stp = td->td_sel) != NULL)
 1469                 goto out;
 1470         td->td_sel = stp = malloc(sizeof(*stp), M_SELECT, M_WAITOK|M_ZERO);
 1471         mtx_init(&stp->st_mtx, "sellck", NULL, MTX_DEF);
 1472         cv_init(&stp->st_wait, "select");
 1473 out:
 1474         stp->st_flags = 0;
 1475         STAILQ_INIT(&stp->st_selq);
 1476 }
 1477 
 1478 static int
 1479 seltdwait(struct thread *td, int timo)
 1480 {
 1481         struct seltd *stp;
 1482         int error;
 1483 
 1484         stp = td->td_sel;
 1485         /*
 1486          * An event of interest may occur while we do not hold the seltd
 1487          * locked so check the pending flag before we sleep.
 1488          */
 1489         mtx_lock(&stp->st_mtx);
 1490         /*
 1491          * Any further calls to selrecord will be a rescan.
 1492          */
 1493         stp->st_flags |= SELTD_RESCAN;
 1494         if (stp->st_flags & SELTD_PENDING) {
 1495                 mtx_unlock(&stp->st_mtx);
 1496                 return (0);
 1497         }
 1498         if (timo > 0)
 1499                 error = cv_timedwait_sig(&stp->st_wait, &stp->st_mtx, timo);
 1500         else
 1501                 error = cv_wait_sig(&stp->st_wait, &stp->st_mtx);
 1502         mtx_unlock(&stp->st_mtx);
 1503 
 1504         return (error);
 1505 }
 1506 
 1507 void
 1508 seltdfini(struct thread *td)
 1509 {
 1510         struct seltd *stp;
 1511 
 1512         stp = td->td_sel;
 1513         if (stp == NULL)
 1514                 return;
 1515         if (stp->st_free1)
 1516                 uma_zfree(selfd_zone, stp->st_free1);
 1517         if (stp->st_free2)
 1518                 uma_zfree(selfd_zone, stp->st_free2);
 1519         td->td_sel = NULL;
 1520         free(stp, M_SELECT);
 1521 }
 1522 
 1523 /*
 1524  * Remove the references to the thread from all of the objects we were
 1525  * polling.
 1526  */
 1527 static void
 1528 seltdclear(struct thread *td)
 1529 {
 1530         struct seltd *stp;
 1531         struct selfd *sfp;
 1532         struct selfd *sfn;
 1533 
 1534         stp = td->td_sel;
 1535         STAILQ_FOREACH_SAFE(sfp, &stp->st_selq, sf_link, sfn)
 1536                 selfdfree(stp, sfp);
 1537         stp->st_flags = 0;
 1538 }
 1539 
 1540 static void selectinit(void *);
 1541 SYSINIT(select, SI_SUB_SYSCALLS, SI_ORDER_ANY, selectinit, NULL);
 1542 static void
 1543 selectinit(void *dummy __unused)
 1544 {
 1545 
 1546         selfd_zone = uma_zcreate("selfd", sizeof(struct selfd), NULL, NULL,
 1547             NULL, NULL, UMA_ALIGN_PTR, 0);
 1548         mtxpool_select = mtx_pool_create("select mtxpool", 128, MTX_DEF);
 1549 }

Cache object: 194efe3d7cf461e95dc12c25fe9d6cb5


[ 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.