bsd/kern/kern_physio.c

   1 /*
   2  * Copyright (c) 2000-2004 Apple Computer, Inc. All rights reserved.
   3  *
   4  * @APPLE_OSREFERENCE_LICENSE_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 License
  10  * may not be used to create, or enable the creation or redistribution of,
  11  * unlawful or unlicensed copies of an Apple operating system, or to
  12  * circumvent, violate, or enable the circumvention or violation of, any
  13  * terms of an Apple operating system software license agreement.
  14  *
  15  * Please obtain a copy of the License at
  16  * http://www.opensource.apple.com/apsl/ and read it before using this file.
  17  *
  18  * The Original Code and all software distributed under the License are
  19  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  20  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  21  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  22  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
  23  * Please see the License for the specific language governing rights and
  24  * limitations under the License.
  25  *
  26  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  27  */
  28 /*-
  29  * Copyright (c) 1982, 1986, 1990, 1993
  30  *      The Regents of the University of California.  All rights reserved.
  31  * (c) UNIX System Laboratories, Inc.
  32  * All or some portions of this file are derived from material licensed
  33  * to the University of California by American Telephone and Telegraph
  34  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
  35  * the permission of UNIX System Laboratories, Inc.
  36  *
  37  * Redistribution and use in source and binary forms, with or without
  38  * modification, are permitted provided that the following conditions
  39  * are met:
  40  * 1. Redistributions of source code must retain the above copyright
  41  *    notice, this list of conditions and the following disclaimer.
  42  * 2. Redistributions in binary form must reproduce the above copyright
  43  *    notice, this list of conditions and the following disclaimer in the
  44  *    documentation and/or other materials provided with the distribution.
  45  * 3. All advertising materials mentioning features or use of this software
  46  *    must display the following acknowledgement:
  47  *      This product includes software developed by the University of
  48  *      California, Berkeley and its contributors.
  49  * 4. Neither the name of the University nor the names of its contributors
  50  *    may be used to endorse or promote products derived from this software
  51  *    without specific prior written permission.
  52  *
  53  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  54  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  55  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  56  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  57  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  58  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  59  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  61  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  62  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  63  * SUCH DAMAGE.
  64  *
  65  *      from: @(#)kern_physio.c 8.1 (Berkeley) 6/10/93
  66  */
  67 /*
  68  * HISTORY
  69  * 27-July-97  Umesh Vaishampayan  (umeshv@apple.com)
  70  *      Allow physio() to kernel space.
  71  */
  72
  73 #include <sys/param.h>
  74 #include <sys/systm.h>
  75 #include <sys/buf_internal.h>
  76 #include <sys/conf.h>
  77 #include <sys/proc_internal.h>
  78 #include <sys/uio_internal.h>
  79
  80 int
  81 physio(strategy, bp, dev, flags, minphys, uio, blocksize)
  82         void (*strategy)();
  83         buf_t bp;
  84         dev_t dev;
  85         int flags;
  86         u_int (*minphys)();
  87         struct uio *uio;
  88         int blocksize;
  89 {
  90         struct proc *p = current_proc();
  91         int error, i, nobuf, todo, iosize;
  92 #if LP64KERN
  93         int64_t done;
  94 #else
  95         int done;
  96 #endif
  97
  98         error = 0;
  99         flags &= B_READ | B_WRITE;
 100
 101         /*
 102          * [check user read/write access to the data buffer]
 103          *
 104          * Check each iov one by one.  Note that we know if we're reading or
 105          * writing, so we ignore the uio's rw parameter.  Also note that if
 106          * we're doing a read, that's a *write* to user-space.
 107          */
 108         for (i = 0; i < uio->uio_iovcnt; i++) {
 109                 if(UIO_SEG_IS_USER_SPACE(uio->uio_segflg)) {
 110                         if (!useracc(uio_iov_base_at(uio, i),
 111                                         uio_iov_len_at(uio, i),
 112                                 (flags == B_READ) ? B_WRITE : B_READ))
 113                         return (EFAULT);
 114                 }
 115         }
 116         /* Make sure we have a buffer, creating one if necessary. */
 117         if (nobuf = (bp == NULL)) {
 118                         bp = buf_alloc((vnode_t)0);
 119                 }
 120
 121         /* [while the buffer is marked busy] */
 122         while (((error = (int)buf_acquire(bp, 0, 0, 0)) == EAGAIN));
 123
 124         if (error) {
 125                         if (nobuf)
 126                                 buf_free(bp);
 127                 return (error);
 128         }
 129
 130         /* [set up the fixed part of the buffer for a transfer] */
 131         bp->b_dev = dev;
 132         bp->b_proc = p;
 133
 134         buf_seterror(bp, 0);
 135         /*
 136          * [while there is data to transfer and no I/O error]
 137          * Note that I/O errors are handled with a 'goto' at the bottom
 138          * of the 'while' loop.
 139          */
 140         for (i = 0; i < uio->uio_iovcnt; i++) {
 141                 while (uio_iov_len_at(uio, i) > 0) {
 142                         /*
 143                          * [mark the buffer busy for physical I/O]
 144                          * (i.e. set B_PHYS (because it's an I/O to user
 145                          * memory, and B_RAW, because B_RAW is to be
 146                          * "Set by physio for raw transfers.", in addition
 147                          * to the read/write flag.)
 148                          */
 149                         buf_setflags(bp, B_PHYS | B_RAW | flags);
 150
 151                         if ( (iosize = uio_iov_len_at(uio, i)) > MAXPHYSIO_WIRED)
 152                                 iosize = MAXPHYSIO_WIRED;
 153
 154                         /* [set up the buffer for a maximum-sized transfer] */
 155                         buf_setblkno(bp, uio->uio_offset / blocksize);
 156                         buf_setcount(bp, iosize);
 157                         // LP64todo - fix this!
 158                         buf_setdataptr(bp, CAST_DOWN(caddr_t, uio_iov_base_at(uio, i)));
 159
 160                         /*
 161                          * [call minphys to bound the tranfer size]
 162                          * and remember the amount of data to transfer,
 163                          * for later comparison.
 164                          */
 165                         (*minphys)(bp);
 166                         todo = buf_count(bp);
 167
 168                         /*
 169                          * [lock the part of the user address space involved
 170                          *    in the transfer]
 171                          */
 172
 173                         if(UIO_SEG_IS_USER_SPACE(uio->uio_segflg))
 174                                 vslock(CAST_USER_ADDR_T(buf_dataptr(bp)),
 175                                        (user_size_t)todo);
 176
 177                         /* [call strategy to start the transfer] */
 178                         (*strategy)(bp);
 179
 180
 181                         /* [wait for the transfer to complete] */
 182                         error = (int)buf_biowait(bp);
 183
 184                         /*
 185                          * [unlock the part of the address space previously
 186                          *    locked]
 187                          */
 188                         if(UIO_SEG_IS_USER_SPACE(uio->uio_segflg))
 189                                 vsunlock(CAST_USER_ADDR_T(buf_dataptr(bp)),
 190                                          (user_size_t)todo,
 191                                          (flags & B_READ));
 192
 193                         /*
 194                          * [deduct the transfer size from the total number
 195                          *    of data to transfer]
 196                          */
 197                         done = buf_count(bp) - buf_resid(bp);
 198                         uio_iov_len_add_at(uio, -done, i);
 199                         uio_iov_base_add_at(uio, done, i);
 200                         uio->uio_offset += done;
 201                         uio_setresid(uio, (uio_resid(uio) - done));
 202
 203                         /*
 204                          * Now, check for an error.
 205                          * Also, handle weird end-of-disk semantics.
 206                          */
 207                         if (error || done < todo)
 208                                 goto done;
 209                 }
 210         }
 211
 212 done:
 213         /*
 214          * [clean up the state of the buffer]
 215          * Remember if somebody wants it, so we can wake them up below.
 216          * Also, if we had to steal it, give it back.
 217          */
 218
 219         buf_clearflags(bp, B_PHYS | B_RAW);
 220         if (nobuf)
 221                 buf_free(bp);
 222         else
 223                 {
 224                         buf_drop(bp);
 225                 }
 226
 227         return (error);
 228 }
 229
 230 /*
 231  * Leffler, et al., says on p. 231:
 232  * "The minphys() routine is called by physio() to adjust the
 233  * size of each I/O transfer before the latter is passed to
 234  * the strategy routine..."
 235  *
 236  * so, just adjust the buffer's count accounting to MAXPHYS here,
 237  * and return the new count;
 238  */
 239 u_int
 240 minphys(bp)
 241         struct buf *bp;
 242 {
 243
 244         buf_setcount(bp, min(MAXPHYS, buf_count(bp)));
 245         return buf_count(bp);
 246 }
 247
 248 /*
 249  * Do a read on a device for a user process.
 250  */
 251 rawread(dev, uio)
 252         dev_t dev;
 253         struct uio *uio;
 254 {
 255         return (physio(cdevsw[major(dev)].d_strategy, (struct buf *)NULL,
 256             dev, B_READ, minphys, uio, DEV_BSIZE));
 257 }
 258
 259 /*
 260  * Do a write on a device for a user process.
 261  */
 262 rawwrite(dev, uio)
 263         dev_t dev;
 264         struct uio *uio;
 265 {
 266         return (physio(cdevsw[major(dev)].d_strategy, (struct buf *)NULL,
 267             dev, B_WRITE, minphys, uio, DEV_BSIZE));
 268 }