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  */
  41
  42 #include <sys/param.h>
  43 #include <sys/systm.h>
  44 #include <sys/kernel.h>
  45 #include <sys/proc.h>
  46 #include <sys/malloc.h>
  47 #include <sys/lock.h>
  48 #include <sys/resourcevar.h>
  49 #include <sys/sysctl.h>
  50 #include <sys/uio.h>
  51 #include <sys/vnode.h>
  52 #include <sys/thread2.h>
  53 #include <machine/limits.h>
  54
  55 #include <cpu/lwbuf.h>
  56
  57 #include <vm/vm.h>
  58 #include <vm/vm_page.h>
  59 #include <vm/vm_map.h>
  60
  61 SYSCTL_INT(_kern, KERN_IOV_MAX, iov_max, CTLFLAG_RD, NULL, UIO_MAXIOV,
  62         "Maximum number of elements in an I/O vector; sysconf(_SC_IOV_MAX)");
  63
  64 /*
  65  * UIO_READ:    copy the kernelspace cp to the user or kernelspace UIO
  66  * UIO_WRITE:   copy the user or kernelspace UIO to the kernelspace cp
  67  *
  68  * For userspace UIO's, uio_td must be the current thread.
  69  *
  70  * The syscall interface is responsible for limiting the length to
  71  * ssize_t for things like read() or write() which return the bytes
  72  * read or written as ssize_t.  These functions work with unsigned
  73  * lengths.
  74  */
  75 int
  76 uiomove(caddr_t cp, size_t n, struct uio *uio)
  77 {
  78         thread_t td = curthread;
  79         struct iovec *iov;
  80         size_t cnt;
  81         size_t tot;
  82         int error = 0;
  83         int save = 0;
  84
  85         KASSERT(uio->uio_rw == UIO_READ || uio->uio_rw == UIO_WRITE,
  86             ("uiomove: mode"));
  87         KASSERT(uio->uio_segflg != UIO_USERSPACE || uio->uio_td == td,
  88             ("uiomove proc"));
  89
  90         crit_enter();
  91         save = td->td_flags & TDF_DEADLKTREAT;
  92         td->td_flags |= TDF_DEADLKTREAT;
  93         crit_exit();
  94
  95         tot = 0;
  96
  97         while (n > 0 && uio->uio_resid) {
  98                 iov = uio->uio_iov;
  99                 cnt = iov->iov_len;
 100                 if (cnt == 0) {
 101                         uio->uio_iov++;
 102                         uio->uio_iovcnt--;
 103                         continue;
 104                 }
 105                 if (cnt > n)
 106                         cnt = n;
 107                 tot += cnt;
 108
 109                 switch (uio->uio_segflg) {
 110                 case UIO_USERSPACE:
 111                         if (tot > 1024*1024)
 112                                 lwkt_user_yield();
 113                         if (uio->uio_rw == UIO_READ)
 114                                 error = copyout(cp, iov->iov_base, cnt);
 115                         else
 116                                 error = copyin(iov->iov_base, cp, cnt);
 117                         if (error)
 118                                 break;
 119                         break;
 120
 121                 case UIO_SYSSPACE:
 122                         if (uio->uio_rw == UIO_READ)
 123                                 bcopy(cp, iov->iov_base, cnt);
 124                         else
 125                                 bcopy(iov->iov_base, cp, cnt);
 126                         break;
 127                 case UIO_NOCOPY:
 128                         break;
 129                 }
 130                 iov->iov_base = (char *)iov->iov_base + cnt;
 131                 iov->iov_len -= cnt;
 132                 uio->uio_resid -= cnt;
 133                 uio->uio_offset += cnt;
 134                 cp += cnt;
 135                 n -= cnt;
 136         }
 137         crit_enter();
 138         td->td_flags = (td->td_flags & ~TDF_DEADLKTREAT) | save;
 139         crit_exit();
 140         return (error);
 141 }
 142
 143 /*
 144  * This is the same as uiomove() except (cp, n) is within the bounds of
 145  * the passed, locked buffer.  Under certain circumstances a VM fault
 146  * occuring with a locked buffer held can result in a deadlock or an
 147  * attempt to recursively lock the buffer.
 148  *
 149  * This procedure deals with these cases.
 150  *
 151  * If the buffer represents a regular file, is B_CACHE, but the last VM page
 152  * is not fully valid we fix-up the last VM page.  This should handle the
 153  * recursive lock issue.
 154  *
 155  * Deadlocks are another issue.  We are holding the vp and the bp locked
 156  * and could deadlock against a different vp and/or bp if another thread is
 157  * trying to access us while we accessing it.  The only solution here is
 158  * to release the bp and vnode lock and do the uio to/from a system buffer,
 159  * then regain the locks and copyback (if applicable).  XXX TODO.
 160  */
 161 int
 162 uiomovebp(struct buf *bp, caddr_t cp, size_t n, struct uio *uio)
 163 {
 164         int count;
 165         vm_page_t m;
 166
 167         if (bp->b_vp && bp->b_vp->v_type == VREG &&
 168             (bp->b_flags & B_CACHE) &&
 169             (count = bp->b_xio.xio_npages) != 0 &&
 170             (m = bp->b_xio.xio_pages[count-1])->valid != VM_PAGE_BITS_ALL) {
 171                 vm_page_zero_invalid(m, TRUE);
 172         }
 173         return (uiomove(cp, n, uio));
 174 }
 175
 176 /*
 177  * Like uiomove() but copies zero-fill.  Only allowed for UIO_READ,
 178  * for obvious reasons.
 179  */
 180 int
 181 uiomovez(size_t n, struct uio *uio)
 182 {
 183         struct iovec *iov;
 184         size_t cnt;
 185         int error = 0;
 186
 187         KASSERT(uio->uio_rw == UIO_READ, ("uiomovez: mode"));
 188         KASSERT(uio->uio_segflg != UIO_USERSPACE || uio->uio_td == curthread,
 189                 ("uiomove proc"));
 190
 191         while (n > 0 && uio->uio_resid) {
 192                 iov = uio->uio_iov;
 193                 cnt = iov->iov_len;
 194                 if (cnt == 0) {
 195                         uio->uio_iov++;
 196                         uio->uio_iovcnt--;
 197                         continue;
 198                 }
 199                 if (cnt > n)
 200                         cnt = n;
 201
 202                 switch (uio->uio_segflg) {
 203                 case UIO_USERSPACE:
 204                         error = copyout(ZeroPage, iov->iov_base, cnt);
 205                         if (error)
 206                                 break;
 207                         break;
 208                 case UIO_SYSSPACE:
 209                         bzero(iov->iov_base, cnt);
 210                         break;
 211                 case UIO_NOCOPY:
 212                         break;
 213                 }
 214                 iov->iov_base = (char *)iov->iov_base + cnt;
 215                 iov->iov_len -= cnt;
 216                 uio->uio_resid -= cnt;
 217                 uio->uio_offset += cnt;
 218                 n -= cnt;
 219         }
 220         return (error);
 221 }
 222
 223 /*
 224  * Wrapper for uiomove() that validates the arguments against a known-good
 225  * kernel buffer.  This function automatically indexes the buffer by
 226  * uio_offset and handles all range checking.
 227  */
 228 int
 229 uiomove_frombuf(void *buf, size_t buflen, struct uio *uio)
 230 {
 231         size_t offset;
 232
 233         offset = (size_t)uio->uio_offset;
 234         if ((off_t)offset != uio->uio_offset)
 235                 return (EINVAL);
 236         if (buflen == 0 || offset >= buflen)
 237                 return (0);
 238         return (uiomove((char *)buf + offset, buflen - offset, uio));
 239 }
 240
 241 /*
 242  * Give next character to user as result of read.
 243  */
 244 int
 245 ureadc(int c, struct uio *uio)
 246 {
 247         struct iovec *iov;
 248         char *iov_base;
 249
 250 again:
 251         if (uio->uio_iovcnt == 0 || uio->uio_resid == 0)
 252                 panic("ureadc");
 253         iov = uio->uio_iov;
 254         if (iov->iov_len == 0) {
 255                 uio->uio_iovcnt--;
 256                 uio->uio_iov++;
 257                 goto again;
 258         }
 259         switch (uio->uio_segflg) {
 260
 261         case UIO_USERSPACE:
 262                 if (subyte(iov->iov_base, c) < 0)
 263                         return (EFAULT);
 264                 break;
 265
 266         case UIO_SYSSPACE:
 267                 iov_base = iov->iov_base;
 268                 *iov_base = c;
 269                 iov->iov_base = iov_base;
 270                 break;
 271
 272         case UIO_NOCOPY:
 273                 break;
 274         }
 275         iov->iov_base = (char *)iov->iov_base + 1;
 276         iov->iov_len--;
 277         uio->uio_resid--;
 278         uio->uio_offset++;
 279         return (0);
 280 }
 281
 282 /*
 283  * General routine to allocate a hash table.  Make the hash table size a
 284  * power of 2 greater or equal to the number of elements requested, and
 285  * store the masking value in *hashmask.
 286  */
 287 void *
 288 hashinit(int elements, struct malloc_type *type, u_long *hashmask)
 289 {
 290         long hashsize;
 291         LIST_HEAD(generic, generic) *hashtbl;
 292         int i;
 293
 294         if (elements <= 0)
 295                 panic("hashinit: bad elements");
 296         for (hashsize = 2; hashsize < elements; hashsize <<= 1)
 297                 continue;
 298         hashtbl = kmalloc((u_long)hashsize * sizeof(*hashtbl), type, M_WAITOK);
 299         for (i = 0; i < hashsize; i++)
 300                 LIST_INIT(&hashtbl[i]);
 301         *hashmask = hashsize - 1;
 302         return (hashtbl);
 303 }
 304
 305 void
 306 hashdestroy(void *vhashtbl, struct malloc_type *type, u_long hashmask)
 307 {
 308         LIST_HEAD(generic, generic) *hashtbl, *hp;
 309
 310         hashtbl = vhashtbl;
 311         for (hp = hashtbl; hp <= &hashtbl[hashmask]; hp++)
 312                 KASSERT(LIST_EMPTY(hp), ("%s: hash not empty", __func__));
 313         kfree(hashtbl, type);
 314 }
 315
 316 /*
 317  * This is a newer version which allocates a hash table of structures.
 318  *
 319  * The returned array will be zero'd.  The caller is responsible for
 320  * initializing the structures.
 321  */
 322 void *
 323 hashinit_ext(int elements, size_t size, struct malloc_type *type,
 324              u_long *hashmask)
 325 {
 326         long hashsize;
 327         void *hashtbl;
 328
 329         if (elements <= 0)
 330                 panic("hashinit: bad elements");
 331         for (hashsize = 2; hashsize < elements; hashsize <<= 1)
 332                 continue;
 333         hashtbl = kmalloc((size_t)hashsize * size, type, M_WAITOK | M_ZERO);
 334         *hashmask = hashsize - 1;
 335         return (hashtbl);
 336 }
 337
 338 static int primes[] = { 1, 13, 31, 61, 127, 251, 509, 761, 1021, 1531, 2039,
 339                         2557, 3067, 3583, 4093, 4603, 5119, 5623, 6143, 6653,
 340                         7159, 7673, 8191, 12281, 16381, 24571, 32749 };
 341 #define NPRIMES NELEM(primes)
 342
 343 /*
 344  * General routine to allocate a prime number sized hash table.
 345  */
 346 void *
 347 phashinit(int elements, struct malloc_type *type, u_long *nentries)
 348 {
 349         long hashsize;
 350         LIST_HEAD(generic, generic) *hashtbl;
 351         int i;
 352
 353         if (elements <= 0)
 354                 panic("phashinit: bad elements");
 355         for (i = 1, hashsize = primes[1]; hashsize <= elements;) {
 356                 i++;
 357                 if (i == NPRIMES)
 358                         break;
 359                 hashsize = primes[i];
 360         }
 361         hashsize = primes[i - 1];
 362         hashtbl = kmalloc((u_long)hashsize * sizeof(*hashtbl), type, M_WAITOK);
 363         for (i = 0; i < hashsize; i++)
 364                 LIST_INIT(&hashtbl[i]);
 365         *nentries = hashsize;
 366         return (hashtbl);
 367 }
 368
 369 /*
 370  * This is a newer version which allocates a hash table of structures
 371  * in a prime-number size.
 372  *
 373  * The returned array will be zero'd.  The caller is responsible for
 374  * initializing the structures.
 375  */
 376 void *
 377 phashinit_ext(int elements, size_t size, struct malloc_type *type,
 378               u_long *nentries)
 379 {
 380         long hashsize;
 381         void *hashtbl;
 382         int i;
 383
 384         if (elements <= 0)
 385                 panic("phashinit: bad elements");
 386         for (i = 1, hashsize = primes[1]; hashsize <= elements;) {
 387                 i++;
 388                 if (i == NPRIMES)
 389                         break;
 390                 hashsize = primes[i];
 391         }
 392         hashsize = primes[i - 1];
 393         hashtbl = kmalloc((size_t)hashsize * size, type, M_WAITOK | M_ZERO);
 394         *nentries = hashsize;
 395         return (hashtbl);
 396 }
 397
 398 /*
 399  * Copyin an iovec.  If the iovec array fits, use the preallocated small
 400  * iovec structure.  If it is too big, dynamically allocate an iovec array
 401  * of sufficient size.
 402  *
 403  * MPSAFE
 404  */
 405 int
 406 iovec_copyin(struct iovec *uiov, struct iovec **kiov, struct iovec *siov,
 407              size_t iov_cnt, size_t *iov_len)
 408 {
 409         struct iovec *iovp;
 410         int error, i;
 411         size_t len;
 412
 413         if (iov_cnt > UIO_MAXIOV)
 414                 return EMSGSIZE;
 415         if (iov_cnt > UIO_SMALLIOV) {
 416                 *kiov = kmalloc(sizeof(struct iovec) * iov_cnt, M_IOV,
 417                                 M_WAITOK);
 418         } else {
 419                 *kiov = siov;
 420         }
 421         error = copyin(uiov, *kiov, iov_cnt * sizeof(struct iovec));
 422         if (error == 0) {
 423                 *iov_len = 0;
 424                 for (i = 0, iovp = *kiov; i < iov_cnt; i++, iovp++) {
 425                         /*
 426                          * Check for both *iov_len overflows and out of
 427                          * range iovp->iov_len's.  We limit to the
 428                          * capabilities of signed integers.
 429                          *
 430                          * GCC4 - overflow check opt requires assign/test.
 431                          */
 432                         len = *iov_len + iovp->iov_len;
 433                         if (len < *iov_len)
 434                                 error = EINVAL;
 435                         *iov_len = len;
 436                 }
 437         }
 438
 439         /*
 440          * From userland disallow iovec's which exceed the sized size
 441          * limit as the system calls return ssize_t.
 442          *
 443          * NOTE: Internal kernel interfaces can handle the unsigned
 444          *       limit.
 445          */
 446         if (error == 0 && (ssize_t)*iov_len < 0)
 447                 error = EINVAL;
 448
 449         if (error)
 450                 iovec_free(kiov, siov);
 451         return (error);
 452 }
 453
 454
 455 /*
 456  * Copyright (c) 2004 Alan L. Cox <alc@cs.rice.edu>
 457  * Copyright (c) 1982, 1986, 1991, 1993
 458  *      The Regents of the University of California.  All rights reserved.
 459  * (c) UNIX System Laboratories, Inc.
 460  * All or some portions of this file are derived from material licensed
 461  * to the University of California by American Telephone and Telegraph
 462  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
 463  * the permission of UNIX System Laboratories, Inc.
 464  *
 465  * Redistribution and use in source and binary forms, with or without
 466  * modification, are permitted provided that the following conditions
 467  * are met:
 468  * 1. Redistributions of source code must retain the above copyright
 469  *    notice, this list of conditions and the following disclaimer.
 470  * 2. Redistributions in binary form must reproduce the above copyright
 471  *    notice, this list of conditions and the following disclaimer in the
 472  *    documentation and/or other materials provided with the distribution.
 473  * 4. Neither the name of the University nor the names of its contributors
 474  *    may be used to endorse or promote products derived from this software
 475  *    without specific prior written permission.
 476  *
 477  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 478  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 479  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 480  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 481  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 482  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 483  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 484  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 485  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 486  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 487  * SUCH DAMAGE.
 488  *
 489  * @(#)kern_subr.c      8.3 (Berkeley) 1/21/94
 490  * $FreeBSD: src/sys/i386/i386/uio_machdep.c,v 1.1 2004/03/21 20:28:36 alc Exp $
 491  */
 492
 493 /*
 494  * Implement uiomove(9) from physical memory using lwbuf's to reduce
 495  * the creation and destruction of ephemeral mappings.
 496  */
 497 int
 498 uiomove_fromphys(vm_page_t *ma, vm_offset_t offset, size_t n, struct uio *uio)
 499 {
 500         struct lwbuf lwb_cache;
 501         struct lwbuf *lwb;
 502         struct thread *td = curthread;
 503         struct iovec *iov;
 504         void *cp;
 505         vm_offset_t page_offset;
 506         vm_page_t m;
 507         size_t cnt;
 508         int error = 0;
 509         int save = 0;
 510
 511         KASSERT(uio->uio_rw == UIO_READ || uio->uio_rw == UIO_WRITE,
 512             ("uiomove_fromphys: mode"));
 513         KASSERT(uio->uio_segflg != UIO_USERSPACE || uio->uio_td == curthread,
 514             ("uiomove_fromphys proc"));
 515
 516         crit_enter();
 517         save = td->td_flags & TDF_DEADLKTREAT;
 518         td->td_flags |= TDF_DEADLKTREAT;
 519         crit_exit();
 520
 521         while (n > 0 && uio->uio_resid) {
 522                 iov = uio->uio_iov;
 523                 cnt = iov->iov_len;
 524                 if (cnt == 0) {
 525                         uio->uio_iov++;
 526                         uio->uio_iovcnt--;
 527                         continue;
 528                 }
 529                 if (cnt > n)
 530                         cnt = n;
 531                 page_offset = offset & PAGE_MASK;
 532                 cnt = min(cnt, PAGE_SIZE - page_offset);
 533                 m = ma[offset >> PAGE_SHIFT];
 534                 lwb = lwbuf_alloc(m, &lwb_cache);
 535                 cp = (char *)lwbuf_kva(lwb) + page_offset;
 536                 switch (uio->uio_segflg) {
 537                 case UIO_USERSPACE:
 538                         /*
 539                          * note: removed uioyield (it was the wrong place to
 540                          * put it).
 541                          */
 542                         if (uio->uio_rw == UIO_READ)
 543                                 error = copyout(cp, iov->iov_base, cnt);
 544                         else
 545                                 error = copyin(iov->iov_base, cp, cnt);
 546                         if (error) {
 547                                 lwbuf_free(lwb);
 548                                 goto out;
 549                         }
 550                         break;
 551                 case UIO_SYSSPACE:
 552                         if (uio->uio_rw == UIO_READ)
 553                                 bcopy(cp, iov->iov_base, cnt);
 554                         else
 555                                 bcopy(iov->iov_base, cp, cnt);
 556                         break;
 557                 case UIO_NOCOPY:
 558                         break;
 559                 }
 560                 lwbuf_free(lwb);
 561                 iov->iov_base = (char *)iov->iov_base + cnt;
 562                 iov->iov_len -= cnt;
 563                 uio->uio_resid -= cnt;
 564                 uio->uio_offset += cnt;
 565                 offset += cnt;
 566                 n -= cnt;
 567         }
 568 out:
 569         if (save == 0) {
 570                 crit_enter();
 571                 td->td_flags &= ~TDF_DEADLKTREAT;
 572                 crit_exit();
 573         }
 574         return (error);
 575 }
 576