sys/kern/kern_subr.c

   1 /*
   2  * Copyright (c) 1982, 1986, 1991, 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  * 3. All advertising materials mentioning features or use of this software
  19  *    must display the following acknowledgement:
  20  *      This product includes software developed by the University of
  21  *      California, Berkeley and its contributors.
  22  * 4. Neither the name of the University nor the names of its contributors
  23  *    may be used to endorse or promote products derived from this software
  24  *    without specific prior written permission.
  25  *
  26  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  27  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  28  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  29  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  30  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  31  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  32  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  33  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  34  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  35  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  36  * SUCH DAMAGE.
  37  *
  38  *      @(#)kern_subr.c 8.3 (Berkeley) 1/21/94
  39  * $FreeBSD: src/sys/kern/kern_subr.c,v 1.31.2.2 2002/04/21 08:09:37 bde Exp $
  40  * $DragonFly: src/sys/kern/kern_subr.c,v 1.27 2007/01/29 20:44:02 tgen Exp $
  41  */
  42
  43 #include "opt_ddb.h"
  44
  45 #include <sys/param.h>
  46 #include <sys/systm.h>
  47 #include <sys/kernel.h>
  48 #include <sys/proc.h>
  49 #include <sys/malloc.h>
  50 #include <sys/lock.h>
  51 #include <sys/resourcevar.h>
  52 #include <sys/sysctl.h>
  53 #include <sys/uio.h>
  54 #include <sys/vnode.h>
  55 #include <sys/sfbuf.h>
  56 #include <sys/thread2.h>
  57 #include <machine/limits.h>
  58
  59 #include <vm/vm.h>
  60 #include <vm/vm_page.h>
  61 #include <vm/vm_map.h>
  62
  63 SYSCTL_INT(_kern, KERN_IOV_MAX, iov_max, CTLFLAG_RD, NULL, UIO_MAXIOV,
  64         "Maximum number of elements in an I/O vector; sysconf(_SC_IOV_MAX)");
  65
  66 /*
  67  * UIO_READ:    copy the kernelspace cp to the user or kernelspace UIO
  68  * UIO_WRITE:   copy the user or kernelspace UIO to the kernelspace cp
  69  *
  70  * For userspace UIO's, uio_td must be the current thread.
  71  *
  72  * The syscall interface is responsible for limiting the length to
  73  * ssize_t for things like read() or write() which return the bytes
  74  * read or written as ssize_t.  These functions work with unsigned
  75  * lengths.
  76  */
  77 int
  78 uiomove(caddr_t cp, size_t n, struct uio *uio)
  79 {
  80         struct iovec *iov;
  81         size_t cnt;
  82         int error = 0;
  83         int save = 0;
  84         int baseticks = ticks;
  85
  86         KASSERT(uio->uio_rw == UIO_READ || uio->uio_rw == UIO_WRITE,
  87             ("uiomove: mode"));
  88         KASSERT(uio->uio_segflg != UIO_USERSPACE || uio->uio_td == curthread,
  89             ("uiomove proc"));
  90
  91         if (curproc) {
  92                 save = curproc->p_flag & P_DEADLKTREAT;
  93                 curproc->p_flag |= P_DEADLKTREAT;
  94         }
  95
  96         while (n > 0 && uio->uio_resid) {
  97                 iov = uio->uio_iov;
  98                 cnt = iov->iov_len;
  99                 if (cnt == 0) {
 100                         uio->uio_iov++;
 101                         uio->uio_iovcnt--;
 102                         continue;
 103                 }
 104                 if (cnt > n)
 105                         cnt = n;
 106
 107                 switch (uio->uio_segflg) {
 108
 109                 case UIO_USERSPACE:
 110                         if (ticks - baseticks >= hogticks) {
 111                                 uio_yield();
 112                                 baseticks = ticks;
 113                         }
 114                         if (uio->uio_rw == UIO_READ)
 115                                 error = copyout(cp, iov->iov_base, cnt);
 116                         else
 117                                 error = copyin(iov->iov_base, cp, cnt);
 118                         if (error)
 119                                 break;
 120                         break;
 121
 122                 case UIO_SYSSPACE:
 123                         if (uio->uio_rw == UIO_READ)
 124                                 bcopy((caddr_t)cp, iov->iov_base, cnt);
 125                         else
 126                                 bcopy(iov->iov_base, (caddr_t)cp, cnt);
 127                         break;
 128                 case UIO_NOCOPY:
 129                         break;
 130                 }
 131                 iov->iov_base = (char *)iov->iov_base + cnt;
 132                 iov->iov_len -= cnt;
 133                 uio->uio_resid -= cnt;
 134                 uio->uio_offset += cnt;
 135                 cp += cnt;
 136                 n -= cnt;
 137         }
 138         if (curproc)
 139                 curproc->p_flag = (curproc->p_flag & ~P_DEADLKTREAT) | save;
 140         return (error);
 141 }
 142
 143 /*
 144  * Like uiomove() but copies zero-fill.  Only allowed for UIO_READ,
 145  * for obvious reasons.
 146  */
 147 int
 148 uiomovez(size_t n, struct uio *uio)
 149 {
 150         struct iovec *iov;
 151         size_t cnt;
 152         int error = 0;
 153
 154         KASSERT(uio->uio_rw == UIO_READ, ("uiomovez: mode"));
 155         KASSERT(uio->uio_segflg != UIO_USERSPACE || uio->uio_td == curthread,
 156                 ("uiomove proc"));
 157
 158         while (n > 0 && uio->uio_resid) {
 159                 iov = uio->uio_iov;
 160                 cnt = iov->iov_len;
 161                 if (cnt == 0) {
 162                         uio->uio_iov++;
 163                         uio->uio_iovcnt--;
 164                         continue;
 165                 }
 166                 if (cnt > n)
 167                         cnt = n;
 168
 169                 switch (uio->uio_segflg) {
 170                 case UIO_USERSPACE:
 171                         error = copyout(ZeroPage, iov->iov_base, cnt);
 172                         if (error)
 173                                 break;
 174                         break;
 175                 case UIO_SYSSPACE:
 176                         bzero(iov->iov_base, cnt);
 177                         break;
 178                 case UIO_NOCOPY:
 179                         break;
 180                 }
 181                 iov->iov_base = (char *)iov->iov_base + cnt;
 182                 iov->iov_len -= cnt;
 183                 uio->uio_resid -= cnt;
 184                 uio->uio_offset += cnt;
 185                 n -= cnt;
 186         }
 187         return (error);
 188 }
 189
 190 /*
 191  * Wrapper for uiomove() that validates the arguments against a known-good
 192  * kernel buffer.
 193  */
 194 int
 195 uiomove_frombuf(void *buf, size_t buflen, struct uio *uio)
 196 {
 197         size_t offset;
 198
 199         offset = (size_t)uio->uio_offset;
 200         if ((off_t)offset != uio->uio_offset)
 201                 return (EINVAL);
 202         if (buflen == 0 || offset >= buflen)
 203                 return (0);
 204         return (uiomove((char *)buf + offset, buflen - offset, uio));
 205 }
 206
 207 /*
 208  * Give next character to user as result of read.
 209  */
 210 int
 211 ureadc(int c, struct uio *uio)
 212 {
 213         struct iovec *iov;
 214         char *iov_base;
 215
 216 again:
 217         if (uio->uio_iovcnt == 0 || uio->uio_resid == 0)
 218                 panic("ureadc");
 219         iov = uio->uio_iov;
 220         if (iov->iov_len == 0) {
 221                 uio->uio_iovcnt--;
 222                 uio->uio_iov++;
 223                 goto again;
 224         }
 225         switch (uio->uio_segflg) {
 226
 227         case UIO_USERSPACE:
 228                 if (subyte(iov->iov_base, c) < 0)
 229                         return (EFAULT);
 230                 break;
 231
 232         case UIO_SYSSPACE:
 233                 iov_base = iov->iov_base;
 234                 *iov_base = c;
 235                 iov->iov_base = iov_base;
 236                 break;
 237
 238         case UIO_NOCOPY:
 239                 break;
 240         }
 241         iov->iov_base = (char *)iov->iov_base + 1;
 242         iov->iov_len--;
 243         uio->uio_resid--;
 244         uio->uio_offset++;
 245         return (0);
 246 }
 247
 248 /*
 249  * General routine to allocate a hash table.  Make the hash table size a
 250  * power of 2 greater or equal to the number of elements requested, and
 251  * store the masking value in *hashmask.
 252  */
 253 void *
 254 hashinit(int elements, struct malloc_type *type, u_long *hashmask)
 255 {
 256         long hashsize;
 257         LIST_HEAD(generic, generic) *hashtbl;
 258         int i;
 259
 260         if (elements <= 0)
 261                 panic("hashinit: bad elements");
 262         for (hashsize = 2; hashsize < elements; hashsize <<= 1)
 263                 continue;
 264         hashtbl = kmalloc((u_long)hashsize * sizeof(*hashtbl), type, M_WAITOK);
 265         for (i = 0; i < hashsize; i++)
 266                 LIST_INIT(&hashtbl[i]);
 267         *hashmask = hashsize - 1;
 268         return (hashtbl);
 269 }
 270
 271 static int primes[] = { 1, 13, 31, 61, 127, 251, 509, 761, 1021, 1531, 2039,
 272                         2557, 3067, 3583, 4093, 4603, 5119, 5623, 6143, 6653,
 273                         7159, 7673, 8191, 12281, 16381, 24571, 32749 };
 274 #define NPRIMES (sizeof(primes) / sizeof(primes[0]))
 275
 276 /*
 277  * General routine to allocate a prime number sized hash table.
 278  */
 279 void *
 280 phashinit(int elements, struct malloc_type *type, u_long *nentries)
 281 {
 282         long hashsize;
 283         LIST_HEAD(generic, generic) *hashtbl;
 284         int i;
 285
 286         if (elements <= 0)
 287                 panic("phashinit: bad elements");
 288         for (i = 1, hashsize = primes[1]; hashsize <= elements;) {
 289                 i++;
 290                 if (i == NPRIMES)
 291                         break;
 292                 hashsize = primes[i];
 293         }
 294         hashsize = primes[i - 1];
 295         hashtbl = kmalloc((u_long)hashsize * sizeof(*hashtbl), type, M_WAITOK);
 296         for (i = 0; i < hashsize; i++)
 297                 LIST_INIT(&hashtbl[i]);
 298         *nentries = hashsize;
 299         return (hashtbl);
 300 }
 301
 302 /*
 303  * Copyin an iovec.  If the iovec array fits, use the preallocated small
 304  * iovec structure.  If it is too big, dynamically allocate an iovec array
 305  * of sufficient size.
 306  *
 307  * MPSAFE
 308  */
 309 int
 310 iovec_copyin(struct iovec *uiov, struct iovec **kiov, struct iovec *siov,
 311              size_t iov_cnt, size_t *iov_len)
 312 {
 313         struct iovec *iovp;
 314         int error, i;
 315         size_t len;
 316
 317         if (iov_cnt > UIO_MAXIOV)
 318                 return EMSGSIZE;
 319         if (iov_cnt > UIO_SMALLIOV) {
 320                 MALLOC(*kiov, struct iovec *, sizeof(struct iovec) * iov_cnt,
 321                     M_IOV, M_WAITOK);
 322         } else {
 323                 *kiov = siov;
 324         }
 325         error = copyin(uiov, *kiov, iov_cnt * sizeof(struct iovec));
 326         if (error == 0) {
 327                 *iov_len = 0;
 328                 for (i = 0, iovp = *kiov; i < iov_cnt; i++, iovp++) {
 329                         /*
 330                          * Check for both *iov_len overflows and out of
 331                          * range iovp->iov_len's.  We limit to the
 332                          * capabilities of signed integers.
 333                          *
 334                          * GCC4 - overflow check opt requires assign/test.
 335                          */
 336                         len = *iov_len + iovp->iov_len;
 337                         if (len < *iov_len)
 338                                 error = EINVAL;
 339                         *iov_len = len;
 340                 }
 341         }
 342
 343         /*
 344          * From userland disallow iovec's which exceed the sized size
 345          * limit as the system calls return ssize_t.
 346          *
 347          * NOTE: Internal kernel interfaces can handle the unsigned
 348          *       limit.
 349          */
 350         if (error == 0 && (ssize_t)*iov_len < 0)
 351                 error = EINVAL;
 352
 353         if (error)
 354                 iovec_free(kiov, siov);
 355         return (error);
 356 }
 357
 358
 359 /*
 360  * Copyright (c) 2004 Alan L. Cox <alc@cs.rice.edu>
 361  * Copyright (c) 1982, 1986, 1991, 1993
 362  *      The Regents of the University of California.  All rights reserved.
 363  * (c) UNIX System Laboratories, Inc.
 364  * All or some portions of this file are derived from material licensed
 365  * to the University of California by American Telephone and Telegraph
 366  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
 367  * the permission of UNIX System Laboratories, Inc.
 368  *
 369  * Redistribution and use in source and binary forms, with or without
 370  * modification, are permitted provided that the following conditions
 371  * are met:
 372  * 1. Redistributions of source code must retain the above copyright
 373  *    notice, this list of conditions and the following disclaimer.
 374  * 2. Redistributions in binary form must reproduce the above copyright
 375  *    notice, this list of conditions and the following disclaimer in the
 376  *    documentation and/or other materials provided with the distribution.
 377  * 4. Neither the name of the University nor the names of its contributors
 378  *    may be used to endorse or promote products derived from this software
 379  *    without specific prior written permission.
 380  *
 381  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 382  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 383  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 384  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 385  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 386  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 387  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 388  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 389  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 390  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 391  * SUCH DAMAGE.
 392  *
 393  * @(#)kern_subr.c      8.3 (Berkeley) 1/21/94
 394  * $FreeBSD: src/sys/i386/i386/uio_machdep.c,v 1.1 2004/03/21 20:28:36 alc Exp $
 395  */
 396
 397 /*
 398  * Implement uiomove(9) from physical memory using sf_bufs to reduce
 399  * the creation and destruction of ephemeral mappings.
 400  */
 401 int
 402 uiomove_fromphys(vm_page_t *ma, vm_offset_t offset, size_t n, struct uio *uio)
 403 {
 404         struct sf_buf *sf;
 405         struct thread *td = curthread;
 406         struct iovec *iov;
 407         void *cp;
 408         vm_offset_t page_offset;
 409         vm_page_t m;
 410         size_t cnt;
 411         int error = 0;
 412         int save = 0;
 413
 414         KASSERT(uio->uio_rw == UIO_READ || uio->uio_rw == UIO_WRITE,
 415             ("uiomove_fromphys: mode"));
 416         KASSERT(uio->uio_segflg != UIO_USERSPACE || uio->uio_td == curthread,
 417             ("uiomove_fromphys proc"));
 418
 419         crit_enter();
 420         save = td->td_flags & TDF_DEADLKTREAT;
 421         td->td_flags |= TDF_DEADLKTREAT;
 422         crit_exit();
 423
 424         while (n > 0 && uio->uio_resid) {
 425                 iov = uio->uio_iov;
 426                 cnt = iov->iov_len;
 427                 if (cnt == 0) {
 428                         uio->uio_iov++;
 429                         uio->uio_iovcnt--;
 430                         continue;
 431                 }
 432                 if (cnt > n)
 433                         cnt = n;
 434                 page_offset = offset & PAGE_MASK;
 435                 cnt = min(cnt, PAGE_SIZE - page_offset);
 436                 m = ma[offset >> PAGE_SHIFT];
 437                 sf = sf_buf_alloc(m, SFB_CPUPRIVATE);
 438                 cp = (char *)sf_buf_kva(sf) + page_offset;
 439                 switch (uio->uio_segflg) {
 440                 case UIO_USERSPACE:
 441                         /*
 442                          * note: removed uioyield (it was the wrong place to
 443                          * put it).
 444                          */
 445                         if (uio->uio_rw == UIO_READ)
 446                                 error = copyout(cp, iov->iov_base, cnt);
 447                         else
 448                                 error = copyin(iov->iov_base, cp, cnt);
 449                         if (error) {
 450                                 sf_buf_free(sf);
 451                                 goto out;
 452                         }
 453                         break;
 454                 case UIO_SYSSPACE:
 455                         if (uio->uio_rw == UIO_READ)
 456                                 bcopy(cp, iov->iov_base, cnt);
 457                         else
 458                                 bcopy(iov->iov_base, cp, cnt);
 459                         break;
 460                 case UIO_NOCOPY:
 461                         break;
 462                 }
 463                 sf_buf_free(sf);
 464                 iov->iov_base = (char *)iov->iov_base + cnt;
 465                 iov->iov_len -= cnt;
 466                 uio->uio_resid -= cnt;
 467                 uio->uio_offset += cnt;
 468                 offset += cnt;
 469                 n -= cnt;
 470         }
 471 out:
 472         if (save == 0) {
 473                 crit_enter();
 474                 td->td_flags &= ~TDF_DEADLKTREAT;
 475                 crit_exit();
 476         }
 477         return (error);
 478 }
 479