bsd/kern/kern_physio.c

   1 /*
   2  * Copyright (c) 2006 Apple Computer, Inc. All Rights Reserved.
   3  *
   4  * @APPLE_LICENSE_OSREFERENCE_HEADER_START@
   5  *
   6  * This file contains Original Code and/or Modifications of Original Code
   7  * as defined in and that are subject to the Apple Public Source License
   8  * Version 2.0 (the 'License'). You may not use this file except in
   9  * compliance with the License.  The rights granted to you under the
  10  * License may not be used to create, or enable the creation or
  11  * redistribution of, unlawful or unlicensed copies of an Apple operating
  12  * system, or to circumvent, violate, or enable the circumvention or
  13  * violation of, any terms of an Apple operating system software license
  14  * agreement.
  15  *
  16  * Please obtain a copy of the License at
  17  * http://www.opensource.apple.com/apsl/ and read it before using this
  18  * file.
  19  *
  20  * The Original Code and all software distributed under the License are
  21  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  22  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  23  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  24  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
  25  * Please see the License for the specific language governing rights and
  26  * limitations under the License.
  27  *
  28  * @APPLE_LICENSE_OSREFERENCE_HEADER_END@
  29  */
  30 /*-
  31  * Copyright (c) 1982, 1986, 1990, 1993
  32  *      The Regents of the University of California.  All rights reserved.
  33  * (c) UNIX System Laboratories, Inc.
  34  * All or some portions of this file are derived from material licensed
  35  * to the University of California by American Telephone and Telegraph
  36  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
  37  * the permission of UNIX System Laboratories, Inc.
  38  *
  39  * Redistribution and use in source and binary forms, with or without
  40  * modification, are permitted provided that the following conditions
  41  * are met:
  42  * 1. Redistributions of source code must retain the above copyright
  43  *    notice, this list of conditions and the following disclaimer.
  44  * 2. Redistributions in binary form must reproduce the above copyright
  45  *    notice, this list of conditions and the following disclaimer in the
  46  *    documentation and/or other materials provided with the distribution.
  47  * 3. All advertising materials mentioning features or use of this software
  48  *    must display the following acknowledgement:
  49  *      This product includes software developed by the University of
  50  *      California, Berkeley and its contributors.
  51  * 4. Neither the name of the University nor the names of its contributors
  52  *    may be used to endorse or promote products derived from this software
  53  *    without specific prior written permission.
  54  *
  55  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  56  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  57  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  58  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  59  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  60  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  61  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  62  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  63  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  64  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  65  * SUCH DAMAGE.
  66  *
  67  *      from: @(#)kern_physio.c 8.1 (Berkeley) 6/10/93
  68  */
  69 /*
  70  * HISTORY
  71  * 27-July-97  Umesh Vaishampayan  (umeshv@apple.com)
  72  *      Allow physio() to kernel space.
  73  */
  74
  75 #include <sys/param.h>
  76 #include <sys/systm.h>
  77 #include <sys/buf_internal.h>
  78 #include <sys/conf.h>
  79 #include <sys/proc_internal.h>
  80 #include <sys/uio_internal.h>
  81
  82 int
  83 physio(strategy, bp, dev, flags, minphys, uio, blocksize)
  84         void (*strategy)();
  85         buf_t bp;
  86         dev_t dev;
  87         int flags;
  88         u_int (*minphys)();
  89         struct uio *uio;
  90         int blocksize;
  91 {
  92         struct proc *p = current_proc();
  93         int error, i, nobuf, todo, iosize;
  94 #if LP64KERN
  95         int64_t done;
  96 #else
  97         int done;
  98 #endif
  99
 100         error = 0;
 101         flags &= B_READ | B_WRITE;
 102
 103         /*
 104          * [check user read/write access to the data buffer]
 105          *
 106          * Check each iov one by one.  Note that we know if we're reading or
 107          * writing, so we ignore the uio's rw parameter.  Also note that if
 108          * we're doing a read, that's a *write* to user-space.
 109          */
 110         for (i = 0; i < uio->uio_iovcnt; i++) {
 111                 if(UIO_SEG_IS_USER_SPACE(uio->uio_segflg)) {
 112                         if (!useracc(uio_iov_base_at(uio, i),
 113                                         uio_iov_len_at(uio, i),
 114                                 (flags == B_READ) ? B_WRITE : B_READ))
 115                         return (EFAULT);
 116                 }
 117         }
 118         /* Make sure we have a buffer, creating one if necessary. */
 119         if (nobuf = (bp == NULL)) {
 120                         bp = buf_alloc((vnode_t)0);
 121                 }
 122
 123         /* [while the buffer is marked busy] */
 124         while (((error = (int)buf_acquire(bp, 0, 0, 0)) == EAGAIN));
 125
 126         if (error) {
 127                         if (nobuf)
 128                                 buf_free(bp);
 129                 return (error);
 130         }
 131
 132         /* [set up the fixed part of the buffer for a transfer] */
 133         bp->b_dev = dev;
 134         bp->b_proc = p;
 135
 136         buf_seterror(bp, 0);
 137         /*
 138          * [while there is data to transfer and no I/O error]
 139          * Note that I/O errors are handled with a 'goto' at the bottom
 140          * of the 'while' loop.
 141          */
 142         for (i = 0; i < uio->uio_iovcnt; i++) {
 143                 while (uio_iov_len_at(uio, i) > 0) {
 144                         /*
 145                          * [mark the buffer busy for physical I/O]
 146                          * (i.e. set B_PHYS (because it's an I/O to user
 147                          * memory, and B_RAW, because B_RAW is to be
 148                          * "Set by physio for raw transfers.", in addition
 149                          * to the read/write flag.)
 150                          */
 151                         buf_setflags(bp, B_PHYS | B_RAW | flags);
 152
 153                         if ( (iosize = uio_iov_len_at(uio, i)) > MAXPHYSIO_WIRED)
 154                                 iosize = MAXPHYSIO_WIRED;
 155
 156                         /* [set up the buffer for a maximum-sized transfer] */
 157                         buf_setblkno(bp, uio->uio_offset / blocksize);
 158                         buf_setcount(bp, iosize);
 159                         // LP64todo - fix this!
 160                         buf_setdataptr(bp, CAST_DOWN(caddr_t, uio_iov_base_at(uio, i)));
 161
 162                         /*
 163                          * [call minphys to bound the tranfer size]
 164                          * and remember the amount of data to transfer,
 165                          * for later comparison.
 166                          */
 167                         (*minphys)(bp);
 168                         todo = buf_count(bp);
 169
 170                         /*
 171                          * [lock the part of the user address space involved
 172                          *    in the transfer]
 173                          */
 174
 175                         if(UIO_SEG_IS_USER_SPACE(uio->uio_segflg))
 176                                 vslock(CAST_USER_ADDR_T(buf_dataptr(bp)),
 177                                        (user_size_t)todo);
 178
 179                         /* [call strategy to start the transfer] */
 180                         (*strategy)(bp);
 181
 182
 183                         /* [wait for the transfer to complete] */
 184                         error = (int)buf_biowait(bp);
 185
 186                         /*
 187                          * [unlock the part of the address space previously
 188                          *    locked]
 189                          */
 190                         if(UIO_SEG_IS_USER_SPACE(uio->uio_segflg))
 191                                 vsunlock(CAST_USER_ADDR_T(buf_dataptr(bp)),
 192                                          (user_size_t)todo,
 193                                          (flags & B_READ));
 194
 195                         /*
 196                          * [deduct the transfer size from the total number
 197                          *    of data to transfer]
 198                          */
 199                         done = buf_count(bp) - buf_resid(bp);
 200                         uio_iov_len_add_at(uio, -done, i);
 201                         uio_iov_base_add_at(uio, done, i);
 202                         uio->uio_offset += done;
 203                         uio_setresid(uio, (uio_resid(uio) - done));
 204
 205                         /*
 206                          * Now, check for an error.
 207                          * Also, handle weird end-of-disk semantics.
 208                          */
 209                         if (error || done < todo)
 210                                 goto done;
 211                 }
 212         }
 213
 214 done:
 215         /*
 216          * [clean up the state of the buffer]
 217          * Remember if somebody wants it, so we can wake them up below.
 218          * Also, if we had to steal it, give it back.
 219          */
 220
 221         buf_clearflags(bp, B_PHYS | B_RAW);
 222         if (nobuf)
 223                 buf_free(bp);
 224         else
 225                 {
 226                         buf_drop(bp);
 227                 }
 228
 229         return (error);
 230 }
 231
 232 /*
 233  * Leffler, et al., says on p. 231:
 234  * "The minphys() routine is called by physio() to adjust the
 235  * size of each I/O transfer before the latter is passed to
 236  * the strategy routine..."
 237  *
 238  * so, just adjust the buffer's count accounting to MAXPHYS here,
 239  * and return the new count;
 240  */
 241 u_int
 242 minphys(bp)
 243         struct buf *bp;
 244 {
 245
 246         buf_setcount(bp, min(MAXPHYS, buf_count(bp)));
 247         return buf_count(bp);
 248 }
 249
 250 /*
 251  * Do a read on a device for a user process.
 252  */
 253 rawread(dev, uio)
 254         dev_t dev;
 255         struct uio *uio;
 256 {
 257         return (physio(cdevsw[major(dev)].d_strategy, (struct buf *)NULL,
 258             dev, B_READ, minphys, uio, DEV_BSIZE));
 259 }
 260
 261 /*
 262  * Do a write on a device for a user process.
 263  */
 264 rawwrite(dev, uio)
 265         dev_t dev;
 266         struct uio *uio;
 267 {
 268         return (physio(cdevsw[major(dev)].d_strategy, (struct buf *)NULL,
 269             dev, B_WRITE, minphys, uio, DEV_BSIZE));
 270 }