bsd/kern/kern_physio.c

   1 /*
   2  * Copyright (c) 2000-2004 Apple Computer, Inc. All rights reserved.
   3  *
   4  * @APPLE_LICENSE_HEADER_START@
   5  *
   6  * The contents of this file constitute Original Code as defined in and
   7  * are subject to the Apple Public Source License Version 1.1 (the
   8  * "License").  You may not use this file except in compliance with the
   9  * License.  Please obtain a copy of the License at
  10  * http://www.apple.com/publicsource and read it before using this file.
  11  *
  12  * This Original Code and all software distributed under the License are
  13  * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  14  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  15  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  16  * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT.  Please see the
  17  * License for the specific language governing rights and limitations
  18  * under the License.
  19  *
  20  * @APPLE_LICENSE_HEADER_END@
  21  */
  22 /*-
  23  * Copyright (c) 1982, 1986, 1990, 1993
  24  *      The Regents of the University of California.  All rights reserved.
  25  * (c) UNIX System Laboratories, Inc.
  26  * All or some portions of this file are derived from material licensed
  27  * to the University of California by American Telephone and Telegraph
  28  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
  29  * the permission of UNIX System Laboratories, Inc.
  30  *
  31  * Redistribution and use in source and binary forms, with or without
  32  * modification, are permitted provided that the following conditions
  33  * are met:
  34  * 1. Redistributions of source code must retain the above copyright
  35  *    notice, this list of conditions and the following disclaimer.
  36  * 2. Redistributions in binary form must reproduce the above copyright
  37  *    notice, this list of conditions and the following disclaimer in the
  38  *    documentation and/or other materials provided with the distribution.
  39  * 3. All advertising materials mentioning features or use of this software
  40  *    must display the following acknowledgement:
  41  *      This product includes software developed by the University of
  42  *      California, Berkeley and its contributors.
  43  * 4. Neither the name of the University nor the names of its contributors
  44  *    may be used to endorse or promote products derived from this software
  45  *    without specific prior written permission.
  46  *
  47  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  48  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  49  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  50  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  51  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  52  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  53  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  54  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  55  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  56  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  57  * SUCH DAMAGE.
  58  *
  59  *      from: @(#)kern_physio.c 8.1 (Berkeley) 6/10/93
  60  */
  61 /*
  62  * HISTORY
  63  * 27-July-97  Umesh Vaishampayan  (umeshv@apple.com)
  64  *      Allow physio() to kernel space.
  65  */
  66
  67 #include <sys/param.h>
  68 #include <sys/systm.h>
  69 #include <sys/buf_internal.h>
  70 #include <sys/conf.h>
  71 #include <sys/proc_internal.h>
  72 #include <sys/uio_internal.h>
  73
  74 int
  75 physio(strategy, bp, dev, flags, minphys, uio, blocksize)
  76         void (*strategy)();
  77         buf_t bp;
  78         dev_t dev;
  79         int flags;
  80         u_int (*minphys)();
  81         struct uio *uio;
  82         int blocksize;
  83 {
  84         struct proc *p = current_proc();
  85         int error, i, nobuf, todo, iosize;
  86 #if LP64KERN
  87         int64_t done;
  88 #else
  89         int done;
  90 #endif
  91
  92         error = 0;
  93         flags &= B_READ | B_WRITE;
  94
  95         /*
  96          * [check user read/write access to the data buffer]
  97          *
  98          * Check each iov one by one.  Note that we know if we're reading or
  99          * writing, so we ignore the uio's rw parameter.  Also note that if
 100          * we're doing a read, that's a *write* to user-space.
 101          */
 102         for (i = 0; i < uio->uio_iovcnt; i++) {
 103                 if(UIO_SEG_IS_USER_SPACE(uio->uio_segflg)) {
 104                         if (!useracc(uio_iov_base_at(uio, i),
 105                                         uio_iov_len_at(uio, i),
 106                                 (flags == B_READ) ? B_WRITE : B_READ))
 107                         return (EFAULT);
 108                 }
 109         }
 110         /* Make sure we have a buffer, creating one if necessary. */
 111         if (nobuf = (bp == NULL)) {
 112                         bp = buf_alloc((vnode_t)0);
 113                 }
 114
 115         /* [while the buffer is marked busy] */
 116         while (((error = (int)buf_acquire(bp, 0, 0, 0)) == EAGAIN));
 117
 118         if (error) {
 119                         if (nobuf)
 120                                 buf_free(bp);
 121                 return (error);
 122         }
 123
 124         /* [set up the fixed part of the buffer for a transfer] */
 125         bp->b_dev = dev;
 126         bp->b_proc = p;
 127
 128         buf_seterror(bp, 0);
 129         /*
 130          * [while there is data to transfer and no I/O error]
 131          * Note that I/O errors are handled with a 'goto' at the bottom
 132          * of the 'while' loop.
 133          */
 134         for (i = 0; i < uio->uio_iovcnt; i++) {
 135                 while (uio_iov_len_at(uio, i) > 0) {
 136                         /*
 137                          * [mark the buffer busy for physical I/O]
 138                          * (i.e. set B_PHYS (because it's an I/O to user
 139                          * memory, and B_RAW, because B_RAW is to be
 140                          * "Set by physio for raw transfers.", in addition
 141                          * to the read/write flag.)
 142                          */
 143                         buf_setflags(bp, B_PHYS | B_RAW | flags);
 144
 145                         if ( (iosize = uio_iov_len_at(uio, i)) > MAXPHYSIO_WIRED)
 146                                 iosize = MAXPHYSIO_WIRED;
 147
 148                         /* [set up the buffer for a maximum-sized transfer] */
 149                         buf_setblkno(bp, uio->uio_offset / blocksize);
 150                         buf_setcount(bp, iosize);
 151                         // LP64todo - fix this!
 152                         buf_setdataptr(bp, CAST_DOWN(caddr_t, uio_iov_base_at(uio, i)));
 153
 154                         /*
 155                          * [call minphys to bound the tranfer size]
 156                          * and remember the amount of data to transfer,
 157                          * for later comparison.
 158                          */
 159                         (*minphys)(bp);
 160                         todo = buf_count(bp);
 161
 162                         /*
 163                          * [lock the part of the user address space involved
 164                          *    in the transfer]
 165                          */
 166
 167                         if(UIO_SEG_IS_USER_SPACE(uio->uio_segflg))
 168                                 vslock(CAST_USER_ADDR_T(buf_dataptr(bp)),
 169                                        (user_size_t)todo);
 170
 171                         /* [call strategy to start the transfer] */
 172                         (*strategy)(bp);
 173
 174
 175                         /* [wait for the transfer to complete] */
 176                         error = (int)buf_biowait(bp);
 177
 178                         /*
 179                          * [unlock the part of the address space previously
 180                          *    locked]
 181                          */
 182                         if(UIO_SEG_IS_USER_SPACE(uio->uio_segflg))
 183                                 vsunlock(CAST_USER_ADDR_T(buf_dataptr(bp)),
 184                                          (user_size_t)todo,
 185                                          (flags & B_READ));
 186
 187                         /*
 188                          * [deduct the transfer size from the total number
 189                          *    of data to transfer]
 190                          */
 191                         done = buf_count(bp) - buf_resid(bp);
 192                         uio_iov_len_add_at(uio, -done, i);
 193                         uio_iov_base_add_at(uio, done, i);
 194                         uio->uio_offset += done;
 195                         uio_setresid(uio, (uio_resid(uio) - done));
 196
 197                         /*
 198                          * Now, check for an error.
 199                          * Also, handle weird end-of-disk semantics.
 200                          */
 201                         if (error || done < todo)
 202                                 goto done;
 203                 }
 204         }
 205
 206 done:
 207         /*
 208          * [clean up the state of the buffer]
 209          * Remember if somebody wants it, so we can wake them up below.
 210          * Also, if we had to steal it, give it back.
 211          */
 212
 213         buf_clearflags(bp, B_PHYS | B_RAW);
 214         if (nobuf)
 215                 buf_free(bp);
 216         else
 217                 {
 218                         buf_drop(bp);
 219                 }
 220
 221         return (error);
 222 }
 223
 224 /*
 225  * Leffler, et al., says on p. 231:
 226  * "The minphys() routine is called by physio() to adjust the
 227  * size of each I/O transfer before the latter is passed to
 228  * the strategy routine..."
 229  *
 230  * so, just adjust the buffer's count accounting to MAXPHYS here,
 231  * and return the new count;
 232  */
 233 u_int
 234 minphys(bp)
 235         struct buf *bp;
 236 {
 237
 238         buf_setcount(bp, min(MAXPHYS, buf_count(bp)));
 239         return buf_count(bp);
 240 }
 241
 242 /*
 243  * Do a read on a device for a user process.
 244  */
 245 rawread(dev, uio)
 246         dev_t dev;
 247         struct uio *uio;
 248 {
 249         return (physio(cdevsw[major(dev)].d_strategy, (struct buf *)NULL,
 250             dev, B_READ, minphys, uio, DEV_BSIZE));
 251 }
 252
 253 /*
 254  * Do a write on a device for a user process.
 255  */
 256 rawwrite(dev, uio)
 257         dev_t dev;
 258         struct uio *uio;
 259 {
 260         return (physio(cdevsw[major(dev)].d_strategy, (struct buf *)NULL,
 261             dev, B_WRITE, minphys, uio, DEV_BSIZE));
 262 }