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 /*
 272  * This is a newer version which allocates a hash table of structures.
 273  *
 274  * The returned array will be zero'd.  The caller is responsible for
 275  * initializing the structures.
 276  */
 277 void *
 278 hashinit_ext(int elements, size_t size, struct malloc_type *type,
 279              u_long *hashmask)
 280 {
 281         long hashsize;
 282         void *hashtbl;
 283
 284         if (elements <= 0)
 285                 panic("hashinit: bad elements");
 286         for (hashsize = 2; hashsize < elements; hashsize <<= 1)
 287                 continue;
 288         hashtbl = kmalloc((size_t)hashsize * size, type, M_WAITOK | M_ZERO);
 289         *hashmask = hashsize - 1;
 290         return (hashtbl);
 291 }
 292
 293 static int primes[] = { 1, 13, 31, 61, 127, 251, 509, 761, 1021, 1531, 2039,
 294                         2557, 3067, 3583, 4093, 4603, 5119, 5623, 6143, 6653,
 295                         7159, 7673, 8191, 12281, 16381, 24571, 32749 };
 296 #define NPRIMES (sizeof(primes) / sizeof(primes[0]))
 297
 298 /*
 299  * General routine to allocate a prime number sized hash table.
 300  */
 301 void *
 302 phashinit(int elements, struct malloc_type *type, u_long *nentries)
 303 {
 304         long hashsize;
 305         LIST_HEAD(generic, generic) *hashtbl;
 306         int i;
 307
 308         if (elements <= 0)
 309                 panic("phashinit: bad elements");
 310         for (i = 1, hashsize = primes[1]; hashsize <= elements;) {
 311                 i++;
 312                 if (i == NPRIMES)
 313                         break;
 314                 hashsize = primes[i];
 315         }
 316         hashsize = primes[i - 1];
 317         hashtbl = kmalloc((u_long)hashsize * sizeof(*hashtbl), type, M_WAITOK);
 318         for (i = 0; i < hashsize; i++)
 319                 LIST_INIT(&hashtbl[i]);
 320         *nentries = hashsize;
 321         return (hashtbl);
 322 }
 323
 324 /*
 325  * This is a newer version which allocates a hash table of structures
 326  * in a prime-number size.
 327  *
 328  * The returned array will be zero'd.  The caller is responsible for
 329  * initializing the structures.
 330  */
 331 void *
 332 phashinit_ext(int elements, size_t size, struct malloc_type *type,
 333               u_long *nentries)
 334 {
 335         long hashsize;
 336         void *hashtbl;
 337         int i;
 338
 339         if (elements <= 0)
 340                 panic("phashinit: bad elements");
 341         for (i = 1, hashsize = primes[1]; hashsize <= elements;) {
 342                 i++;
 343                 if (i == NPRIMES)
 344                         break;
 345                 hashsize = primes[i];
 346         }
 347         hashsize = primes[i - 1];
 348         hashtbl = kmalloc((size_t)hashsize * size, type, M_WAITOK | M_ZERO);
 349         *nentries = hashsize;
 350         return (hashtbl);
 351 }
 352
 353 /*
 354  * Copyin an iovec.  If the iovec array fits, use the preallocated small
 355  * iovec structure.  If it is too big, dynamically allocate an iovec array
 356  * of sufficient size.
 357  *
 358  * MPSAFE
 359  */
 360 int
 361 iovec_copyin(struct iovec *uiov, struct iovec **kiov, struct iovec *siov,
 362              size_t iov_cnt, size_t *iov_len)
 363 {
 364         struct iovec *iovp;
 365         int error, i;
 366         size_t len;
 367
 368         if (iov_cnt > UIO_MAXIOV)
 369                 return EMSGSIZE;
 370         if (iov_cnt > UIO_SMALLIOV) {
 371                 MALLOC(*kiov, struct iovec *, sizeof(struct iovec) * iov_cnt,
 372                     M_IOV, M_WAITOK);
 373         } else {
 374                 *kiov = siov;
 375         }
 376         error = copyin(uiov, *kiov, iov_cnt * sizeof(struct iovec));
 377         if (error == 0) {
 378                 *iov_len = 0;
 379                 for (i = 0, iovp = *kiov; i < iov_cnt; i++, iovp++) {
 380                         /*
 381                          * Check for both *iov_len overflows and out of
 382                          * range iovp->iov_len's.  We limit to the
 383                          * capabilities of signed integers.
 384                          *
 385                          * GCC4 - overflow check opt requires assign/test.
 386                          */
 387                         len = *iov_len + iovp->iov_len;
 388                         if (len < *iov_len)
 389                                 error = EINVAL;
 390                         *iov_len = len;
 391                 }
 392         }
 393
 394         /*
 395          * From userland disallow iovec's which exceed the sized size
 396          * limit as the system calls return ssize_t.
 397          *
 398          * NOTE: Internal kernel interfaces can handle the unsigned
 399          *       limit.
 400          */
 401         if (error == 0 && (ssize_t)*iov_len < 0)
 402                 error = EINVAL;
 403
 404         if (error)
 405                 iovec_free(kiov, siov);
 406         return (error);
 407 }
 408
 409
 410 /*
 411  * Copyright (c) 2004 Alan L. Cox <alc@cs.rice.edu>
 412  * Copyright (c) 1982, 1986, 1991, 1993
 413  *      The Regents of the University of California.  All rights reserved.
 414  * (c) UNIX System Laboratories, Inc.
 415  * All or some portions of this file are derived from material licensed
 416  * to the University of California by American Telephone and Telegraph
 417  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
 418  * the permission of UNIX System Laboratories, Inc.
 419  *
 420  * Redistribution and use in source and binary forms, with or without
 421  * modification, are permitted provided that the following conditions
 422  * are met:
 423  * 1. Redistributions of source code must retain the above copyright
 424  *    notice, this list of conditions and the following disclaimer.
 425  * 2. Redistributions in binary form must reproduce the above copyright
 426  *    notice, this list of conditions and the following disclaimer in the
 427  *    documentation and/or other materials provided with the distribution.
 428  * 4. Neither the name of the University nor the names of its contributors
 429  *    may be used to endorse or promote products derived from this software
 430  *    without specific prior written permission.
 431  *
 432  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 433  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 434  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 435  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 436  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 437  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 438  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 439  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 440  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 441  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 442  * SUCH DAMAGE.
 443  *
 444  * @(#)kern_subr.c      8.3 (Berkeley) 1/21/94
 445  * $FreeBSD: src/sys/i386/i386/uio_machdep.c,v 1.1 2004/03/21 20:28:36 alc Exp $
 446  */
 447
 448 /*
 449  * Implement uiomove(9) from physical memory using sf_bufs to reduce
 450  * the creation and destruction of ephemeral mappings.
 451  */
 452 int
 453 uiomove_fromphys(vm_page_t *ma, vm_offset_t offset, size_t n, struct uio *uio)
 454 {
 455         struct sf_buf *sf;
 456         struct thread *td = curthread;
 457         struct iovec *iov;
 458         void *cp;
 459         vm_offset_t page_offset;
 460         vm_page_t m;
 461         size_t cnt;
 462         int error = 0;
 463         int save = 0;
 464
 465         KASSERT(uio->uio_rw == UIO_READ || uio->uio_rw == UIO_WRITE,
 466             ("uiomove_fromphys: mode"));
 467         KASSERT(uio->uio_segflg != UIO_USERSPACE || uio->uio_td == curthread,
 468             ("uiomove_fromphys proc"));
 469
 470         crit_enter();
 471         save = td->td_flags & TDF_DEADLKTREAT;
 472         td->td_flags |= TDF_DEADLKTREAT;
 473         crit_exit();
 474
 475         while (n > 0 && uio->uio_resid) {
 476                 iov = uio->uio_iov;
 477                 cnt = iov->iov_len;
 478                 if (cnt == 0) {
 479                         uio->uio_iov++;
 480                         uio->uio_iovcnt--;
 481                         continue;
 482                 }
 483                 if (cnt > n)
 484                         cnt = n;
 485                 page_offset = offset & PAGE_MASK;
 486                 cnt = min(cnt, PAGE_SIZE - page_offset);
 487                 m = ma[offset >> PAGE_SHIFT];
 488                 sf = sf_buf_alloc(m, SFB_CPUPRIVATE);
 489                 cp = (char *)sf_buf_kva(sf) + page_offset;
 490                 switch (uio->uio_segflg) {
 491                 case UIO_USERSPACE:
 492                         /*
 493                          * note: removed uioyield (it was the wrong place to
 494                          * put it).
 495                          */
 496                         if (uio->uio_rw == UIO_READ)
 497                                 error = copyout(cp, iov->iov_base, cnt);
 498                         else
 499                                 error = copyin(iov->iov_base, cp, cnt);
 500                         if (error) {
 501                                 sf_buf_free(sf);
 502                                 goto out;
 503                         }
 504                         break;
 505                 case UIO_SYSSPACE:
 506                         if (uio->uio_rw == UIO_READ)
 507                                 bcopy(cp, iov->iov_base, cnt);
 508                         else
 509                                 bcopy(iov->iov_base, cp, cnt);
 510                         break;
 511                 case UIO_NOCOPY:
 512                         break;
 513                 }
 514                 sf_buf_free(sf);
 515                 iov->iov_base = (char *)iov->iov_base + cnt;
 516                 iov->iov_len -= cnt;
 517                 uio->uio_resid -= cnt;
 518                 uio->uio_offset += cnt;
 519                 offset += cnt;
 520                 n -= cnt;
 521         }
 522 out:
 523         if (save == 0) {
 524                 crit_enter();
 525                 td->td_flags &= ~TDF_DEADLKTREAT;
 526                 crit_exit();
 527         }
 528         return (error);
 529 }
 530