Home | History | Annotate | Line # | Download | only in kern
subr_disk.c revision 1.78
      1 /*	$NetBSD: subr_disk.c,v 1.78 2006/04/21 13:53:30 yamt Exp $	*/
      2 
      3 /*-
      4  * Copyright (c) 1996, 1997, 1999, 2000 The NetBSD Foundation, Inc.
      5  * All rights reserved.
      6  *
      7  * This code is derived from software contributed to The NetBSD Foundation
      8  * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
      9  * NASA Ames Research Center.
     10  *
     11  * Redistribution and use in source and binary forms, with or without
     12  * modification, are permitted provided that the following conditions
     13  * are met:
     14  * 1. Redistributions of source code must retain the above copyright
     15  *    notice, this list of conditions and the following disclaimer.
     16  * 2. Redistributions in binary form must reproduce the above copyright
     17  *    notice, this list of conditions and the following disclaimer in the
     18  *    documentation and/or other materials provided with the distribution.
     19  * 3. All advertising materials mentioning features or use of this software
     20  *    must display the following acknowledgement:
     21  *	This product includes software developed by the NetBSD
     22  *	Foundation, Inc. and its contributors.
     23  * 4. Neither the name of The NetBSD Foundation nor the names of its
     24  *    contributors may be used to endorse or promote products derived
     25  *    from this software without specific prior written permission.
     26  *
     27  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     28  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     37  * POSSIBILITY OF SUCH DAMAGE.
     38  */
     39 
     40 /*
     41  * Copyright (c) 1982, 1986, 1988, 1993
     42  *	The Regents of the University of California.  All rights reserved.
     43  * (c) UNIX System Laboratories, Inc.
     44  * All or some portions of this file are derived from material licensed
     45  * to the University of California by American Telephone and Telegraph
     46  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
     47  * the permission of UNIX System Laboratories, Inc.
     48  *
     49  * Redistribution and use in source and binary forms, with or without
     50  * modification, are permitted provided that the following conditions
     51  * are met:
     52  * 1. Redistributions of source code must retain the above copyright
     53  *    notice, this list of conditions and the following disclaimer.
     54  * 2. Redistributions in binary form must reproduce the above copyright
     55  *    notice, this list of conditions and the following disclaimer in the
     56  *    documentation and/or other materials provided with the distribution.
     57  * 3. Neither the name of the University nor the names of its contributors
     58  *    may be used to endorse or promote products derived from this software
     59  *    without specific prior written permission.
     60  *
     61  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     62  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     63  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     64  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     65  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     66  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     67  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     68  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     69  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     70  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     71  * SUCH DAMAGE.
     72  *
     73  *	@(#)ufs_disksubr.c	8.5 (Berkeley) 1/21/94
     74  */
     75 
     76 #include <sys/cdefs.h>
     77 __KERNEL_RCSID(0, "$NetBSD: subr_disk.c,v 1.78 2006/04/21 13:53:30 yamt Exp $");
     78 
     79 #include <sys/param.h>
     80 #include <sys/kernel.h>
     81 #include <sys/malloc.h>
     82 #include <sys/buf.h>
     83 #include <sys/syslog.h>
     84 #include <sys/disklabel.h>
     85 #include <sys/disk.h>
     86 #include <sys/sysctl.h>
     87 #include <lib/libkern/libkern.h>
     88 
     89 /*
     90  * Compute checksum for disk label.
     91  */
     92 u_int
     93 dkcksum(struct disklabel *lp)
     94 {
     95 	u_short *start, *end;
     96 	u_short sum = 0;
     97 
     98 	start = (u_short *)lp;
     99 	end = (u_short *)&lp->d_partitions[lp->d_npartitions];
    100 	while (start < end)
    101 		sum ^= *start++;
    102 	return (sum);
    103 }
    104 
    105 /*
    106  * Disk error is the preface to plaintive error messages
    107  * about failing disk transfers.  It prints messages of the form
    108 
    109 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)
    110 
    111  * if the offset of the error in the transfer and a disk label
    112  * are both available.  blkdone should be -1 if the position of the error
    113  * is unknown; the disklabel pointer may be null from drivers that have not
    114  * been converted to use them.  The message is printed with printf
    115  * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
    116  * The message should be completed (with at least a newline) with printf
    117  * or addlog, respectively.  There is no trailing space.
    118  */
    119 #ifndef PRIdaddr
    120 #define PRIdaddr PRId64
    121 #endif
    122 void
    123 diskerr(const struct buf *bp, const char *dname, const char *what, int pri,
    124     int blkdone, const struct disklabel *lp)
    125 {
    126 	int unit = DISKUNIT(bp->b_dev), part = DISKPART(bp->b_dev);
    127 	void (*pr)(const char *, ...);
    128 	char partname = 'a' + part;
    129 	daddr_t sn;
    130 
    131 	if (/*CONSTCOND*/0)
    132 		/* Compiler will error this is the format is wrong... */
    133 		printf("%" PRIdaddr, bp->b_blkno);
    134 
    135 	if (pri != LOG_PRINTF) {
    136 		static const char fmt[] = "";
    137 		log(pri, fmt);
    138 		pr = addlog;
    139 	} else
    140 		pr = printf;
    141 	(*pr)("%s%d%c: %s %sing fsbn ", dname, unit, partname, what,
    142 	    bp->b_flags & B_READ ? "read" : "writ");
    143 	sn = bp->b_blkno;
    144 	if (bp->b_bcount <= DEV_BSIZE)
    145 		(*pr)("%" PRIdaddr, sn);
    146 	else {
    147 		if (blkdone >= 0) {
    148 			sn += blkdone;
    149 			(*pr)("%" PRIdaddr " of ", sn);
    150 		}
    151 		(*pr)("%" PRIdaddr "-%" PRIdaddr "", bp->b_blkno,
    152 		    bp->b_blkno + (bp->b_bcount - 1) / DEV_BSIZE);
    153 	}
    154 	if (lp && (blkdone >= 0 || bp->b_bcount <= lp->d_secsize)) {
    155 		sn += lp->d_partitions[part].p_offset;
    156 		(*pr)(" (%s%d bn %" PRIdaddr "; cn %" PRIdaddr "",
    157 		    dname, unit, sn, sn / lp->d_secpercyl);
    158 		sn %= lp->d_secpercyl;
    159 		(*pr)(" tn %" PRIdaddr " sn %" PRIdaddr ")",
    160 		    sn / lp->d_nsectors, sn % lp->d_nsectors);
    161 	}
    162 }
    163 
    164 /*
    165  * Searches the iostatlist for the disk corresponding to the
    166  * name provided.
    167  */
    168 struct disk *
    169 disk_find(const char *name)
    170 {
    171 	struct io_stats *stat;
    172 
    173 	stat = iostat_find(name);
    174 
    175 	if ((stat != NULL) && (stat->io_type == IOSTAT_DISK))
    176 		return stat->io_parent;
    177 
    178 	return (NULL);
    179 }
    180 
    181 static void
    182 disk_init0(struct disk *diskp)
    183 {
    184 
    185 	/*
    186 	 * Initialize the wedge-related locks and other fields.
    187 	 */
    188 	lockinit(&diskp->dk_rawlock, PRIBIO, "dkrawlk", 0, 0);
    189 	lockinit(&diskp->dk_openlock, PRIBIO, "dkoplk", 0, 0);
    190 	LIST_INIT(&diskp->dk_wedges);
    191 	diskp->dk_nwedges = 0;
    192 }
    193 
    194 static void
    195 disk_attach0(struct disk *diskp)
    196 {
    197 	/*
    198 	 * Allocate and initialize the disklabel structures.  Note that
    199 	 * it's not safe to sleep here, since we're probably going to be
    200 	 * called during autoconfiguration.
    201 	 */
    202 	diskp->dk_label = malloc(sizeof(struct disklabel), M_DEVBUF, M_NOWAIT);
    203 	diskp->dk_cpulabel = malloc(sizeof(struct cpu_disklabel), M_DEVBUF,
    204 	    M_NOWAIT);
    205 	if ((diskp->dk_label == NULL) || (diskp->dk_cpulabel == NULL))
    206 		panic("disk_attach: can't allocate storage for disklabel");
    207 
    208 	memset(diskp->dk_label, 0, sizeof(struct disklabel));
    209 	memset(diskp->dk_cpulabel, 0, sizeof(struct cpu_disklabel));
    210 
    211 	/*
    212 	 * Set up the stats collection.
    213 	 */
    214 	diskp->dk_stats = iostat_alloc(IOSTAT_DISK);
    215 	diskp->dk_stats->io_parent = (void *) diskp;
    216 	diskp->dk_stats->io_name = diskp->dk_name;
    217 }
    218 
    219 static void
    220 disk_detach0(struct disk *diskp)
    221 {
    222 
    223 	/*
    224 	 * Remove from the drivelist.
    225 	 */
    226 	iostat_free(diskp->dk_stats);
    227 
    228 	/*
    229 	 * Free the space used by the disklabel structures.
    230 	 */
    231 	free(diskp->dk_label, M_DEVBUF);
    232 	free(diskp->dk_cpulabel, M_DEVBUF);
    233 }
    234 
    235 /*
    236  * Attach a disk.
    237  */
    238 void
    239 disk_attach(struct disk *diskp)
    240 {
    241 
    242 	disk_init0(diskp);
    243 	disk_attach0(diskp);
    244 }
    245 
    246 /*
    247  * Detach a disk.
    248  */
    249 void
    250 disk_detach(struct disk *diskp)
    251 {
    252 
    253 	(void) lockmgr(&diskp->dk_openlock, LK_DRAIN, NULL);
    254 	disk_detach0(diskp);
    255 }
    256 
    257 /*
    258  * Initialize a pseudo disk.
    259  */
    260 void
    261 pseudo_disk_init(struct disk *diskp)
    262 {
    263 
    264 	disk_init0(diskp);
    265 }
    266 
    267 /*
    268  * Attach a pseudo disk.
    269  */
    270 void
    271 pseudo_disk_attach(struct disk *diskp)
    272 {
    273 
    274 	disk_attach0(diskp);
    275 }
    276 
    277 /*
    278  * Detach a pseudo disk.
    279  */
    280 void
    281 pseudo_disk_detach(struct disk *diskp)
    282 {
    283 
    284 	disk_detach0(diskp);
    285 }
    286 
    287 /*
    288  * Mark the disk as busy for metrics collection.
    289  */
    290 void
    291 disk_busy(struct disk *diskp)
    292 {
    293 
    294 	iostat_busy(diskp->dk_stats);
    295 }
    296 
    297 /*
    298  * Finished disk operations, gather metrics.
    299  */
    300 void
    301 disk_unbusy(struct disk *diskp, long bcount, int read)
    302 {
    303 
    304 	iostat_unbusy(diskp->dk_stats, bcount, read);
    305 }
    306 
    307 /*
    308  * Bounds checking against the media size, used for the raw partition.
    309  * The sector size passed in should currently always be DEV_BSIZE,
    310  * and the media size the size of the device in DEV_BSIZE sectors.
    311  */
    312 int
    313 bounds_check_with_mediasize(struct buf *bp, int secsize, uint64_t mediasize)
    314 {
    315 	int64_t sz;
    316 
    317 	sz = howmany(bp->b_bcount, secsize);
    318 
    319 	if (bp->b_blkno + sz > mediasize) {
    320 		sz = mediasize - bp->b_blkno;
    321 		if (sz == 0) {
    322 			/* If exactly at end of disk, return EOF. */
    323 			bp->b_resid = bp->b_bcount;
    324 			goto done;
    325 		}
    326 		if (sz < 0) {
    327 			/* If past end of disk, return EINVAL. */
    328 			bp->b_error = EINVAL;
    329 			goto bad;
    330 		}
    331 		/* Otherwise, truncate request. */
    332 		bp->b_bcount = sz << DEV_BSHIFT;
    333 	}
    334 
    335 	return 1;
    336 
    337 bad:
    338 	bp->b_flags |= B_ERROR;
    339 done:
    340 	return 0;
    341 }
    342