| 1 | /* |
| 2 | * Copyright (c) 1982, 1986, 1988 Regents of the University of California. |
| 3 | * All rights reserved. |
| 4 | * |
| 5 | * Redistribution and use in source and binary forms are permitted |
| 6 | * provided that the above copyright notice and this paragraph are |
| 7 | * duplicated in all such forms and that any documentation, |
| 8 | * advertising materials, and other materials related to such |
| 9 | * distribution and use acknowledge that the software was developed |
| 10 | * by the University of California, Berkeley. The name of the |
| 11 | * University may not be used to endorse or promote products derived |
| 12 | * from this software without specific prior written permission. |
| 13 | * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR |
| 14 | * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED |
| 15 | * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. |
| 16 | * |
| 17 | * @(#)ufs_disksubr.c 7.12 (Berkeley) %G% |
| 18 | */ |
| 19 | |
| 20 | #include "param.h" |
| 21 | #include "systm.h" |
| 22 | #include "buf.h" |
| 23 | #include "disklabel.h" |
| 24 | #include "syslog.h" |
| 25 | #include "user.h" |
| 26 | |
| 27 | /* |
| 28 | * Seek sort for disks. We depend on the driver |
| 29 | * which calls us using b_resid as the current cylinder number. |
| 30 | * |
| 31 | * The argument dp structure holds a b_actf activity chain pointer |
| 32 | * on which we keep two queues, sorted in ascending cylinder order. |
| 33 | * The first queue holds those requests which are positioned after |
| 34 | * the current cylinder (in the first request); the second holds |
| 35 | * requests which came in after their cylinder number was passed. |
| 36 | * Thus we implement a one way scan, retracting after reaching the |
| 37 | * end of the drive to the first request on the second queue, |
| 38 | * at which time it becomes the first queue. |
| 39 | * |
| 40 | * A one-way scan is natural because of the way UNIX read-ahead |
| 41 | * blocks are allocated. |
| 42 | */ |
| 43 | |
| 44 | #define b_cylin b_resid |
| 45 | |
| 46 | disksort(dp, bp) |
| 47 | register struct buf *dp, *bp; |
| 48 | { |
| 49 | register struct buf *ap; |
| 50 | |
| 51 | /* |
| 52 | * If nothing on the activity queue, then |
| 53 | * we become the only thing. |
| 54 | */ |
| 55 | ap = dp->b_actf; |
| 56 | if(ap == NULL) { |
| 57 | dp->b_actf = bp; |
| 58 | dp->b_actl = bp; |
| 59 | bp->av_forw = NULL; |
| 60 | return; |
| 61 | } |
| 62 | /* |
| 63 | * If we lie after the first (currently active) |
| 64 | * request, then we must locate the second request list |
| 65 | * and add ourselves to it. |
| 66 | */ |
| 67 | if (bp->b_cylin < ap->b_cylin) { |
| 68 | while (ap->av_forw) { |
| 69 | /* |
| 70 | * Check for an ``inversion'' in the |
| 71 | * normally ascending cylinder numbers, |
| 72 | * indicating the start of the second request list. |
| 73 | */ |
| 74 | if (ap->av_forw->b_cylin < ap->b_cylin) { |
| 75 | /* |
| 76 | * Search the second request list |
| 77 | * for the first request at a larger |
| 78 | * cylinder number. We go before that; |
| 79 | * if there is no such request, we go at end. |
| 80 | */ |
| 81 | do { |
| 82 | if (bp->b_cylin < ap->av_forw->b_cylin) |
| 83 | goto insert; |
| 84 | if (bp->b_cylin == ap->av_forw->b_cylin && |
| 85 | bp->b_blkno < ap->av_forw->b_blkno) |
| 86 | goto insert; |
| 87 | ap = ap->av_forw; |
| 88 | } while (ap->av_forw); |
| 89 | goto insert; /* after last */ |
| 90 | } |
| 91 | ap = ap->av_forw; |
| 92 | } |
| 93 | /* |
| 94 | * No inversions... we will go after the last, and |
| 95 | * be the first request in the second request list. |
| 96 | */ |
| 97 | goto insert; |
| 98 | } |
| 99 | /* |
| 100 | * Request is at/after the current request... |
| 101 | * sort in the first request list. |
| 102 | */ |
| 103 | while (ap->av_forw) { |
| 104 | /* |
| 105 | * We want to go after the current request |
| 106 | * if there is an inversion after it (i.e. it is |
| 107 | * the end of the first request list), or if |
| 108 | * the next request is a larger cylinder than our request. |
| 109 | */ |
| 110 | if (ap->av_forw->b_cylin < ap->b_cylin || |
| 111 | bp->b_cylin < ap->av_forw->b_cylin || |
| 112 | (bp->b_cylin == ap->av_forw->b_cylin && |
| 113 | bp->b_blkno < ap->av_forw->b_blkno)) |
| 114 | goto insert; |
| 115 | ap = ap->av_forw; |
| 116 | } |
| 117 | /* |
| 118 | * Neither a second list nor a larger |
| 119 | * request... we go at the end of the first list, |
| 120 | * which is the same as the end of the whole schebang. |
| 121 | */ |
| 122 | insert: |
| 123 | bp->av_forw = ap->av_forw; |
| 124 | ap->av_forw = bp; |
| 125 | if (ap == dp->b_actl) |
| 126 | dp->b_actl = bp; |
| 127 | } |
| 128 | |
| 129 | /* |
| 130 | * Attempt to read a disk label from a device |
| 131 | * using the indicated stategy routine. |
| 132 | * The label must be partly set up before this: |
| 133 | * secpercyl and anything required in the strategy routine |
| 134 | * (e.g., sector size) must be filled in before calling us. |
| 135 | * Returns null on success and an error string on failure. |
| 136 | */ |
| 137 | char * |
| 138 | readdisklabel(dev, strat, lp) |
| 139 | dev_t dev; |
| 140 | int (*strat)(); |
| 141 | register struct disklabel *lp; |
| 142 | { |
| 143 | register struct buf *bp; |
| 144 | struct disklabel *dlp; |
| 145 | char *msg = NULL; |
| 146 | |
| 147 | if (lp->d_secperunit == 0) |
| 148 | lp->d_secperunit = 0x1fffffff; |
| 149 | lp->d_npartitions = 1; |
| 150 | if (lp->d_partitions[0].p_size == 0) |
| 151 | lp->d_partitions[0].p_size = 0x1fffffff; |
| 152 | lp->d_partitions[0].p_offset = 0; |
| 153 | |
| 154 | bp = geteblk((int)lp->d_secsize); |
| 155 | bp->b_dev = dev; |
| 156 | bp->b_blkno = LABELSECTOR; |
| 157 | bp->b_bcount = lp->d_secsize; |
| 158 | bp->b_flags = B_BUSY | B_READ; |
| 159 | bp->b_cylin = LABELSECTOR / lp->d_secpercyl; |
| 160 | (*strat)(bp); |
| 161 | if (biowait(bp)) { |
| 162 | msg = "I/O error"; |
| 163 | } else for (dlp = (struct disklabel *)bp->b_un.b_addr; |
| 164 | dlp <= (struct disklabel *)(bp->b_un.b_addr+DEV_BSIZE-sizeof(*dlp)); |
| 165 | dlp = (struct disklabel *)((char *)dlp + sizeof(long))) { |
| 166 | if (dlp->d_magic != DISKMAGIC || dlp->d_magic2 != DISKMAGIC) { |
| 167 | if (msg == NULL) |
| 168 | msg = "no disk label"; |
| 169 | } else if (dkcksum(dlp) != 0) |
| 170 | msg = "disk label corrupted"; |
| 171 | else { |
| 172 | *lp = *dlp; |
| 173 | msg = NULL; |
| 174 | break; |
| 175 | } |
| 176 | } |
| 177 | if (lp->d_npartitions > MAXPARTITIONS) |
| 178 | lp->d_npartitions = MAXPARTITIONS; |
| 179 | bp->b_flags = B_INVAL | B_AGE; |
| 180 | brelse(bp); |
| 181 | return (msg); |
| 182 | } |
| 183 | |
| 184 | /* |
| 185 | * Check new disk label for sensibility |
| 186 | * before setting it. |
| 187 | */ |
| 188 | setdisklabel(olp, nlp, openmask) |
| 189 | register struct disklabel *olp, *nlp; |
| 190 | u_long openmask; |
| 191 | { |
| 192 | register i; |
| 193 | register struct partition *opp, *npp; |
| 194 | |
| 195 | if (nlp->d_magic != DISKMAGIC || nlp->d_magic2 != DISKMAGIC || |
| 196 | dkcksum(nlp) != 0) |
| 197 | return (EINVAL); |
| 198 | while ((i = ffs((long)openmask)) != 0) { |
| 199 | i--; |
| 200 | openmask &= ~(1 << i); |
| 201 | if (nlp->d_npartitions <= i) |
| 202 | return (EBUSY); |
| 203 | opp = &olp->d_partitions[i]; |
| 204 | npp = &nlp->d_partitions[i]; |
| 205 | if (npp->p_offset != opp->p_offset || npp->p_size < opp->p_size) |
| 206 | return (EBUSY); |
| 207 | /* |
| 208 | * Copy internally-set partition information |
| 209 | * if new label doesn't include it. XXX |
| 210 | */ |
| 211 | if (npp->p_fstype == FS_UNUSED && opp->p_fstype != FS_UNUSED) { |
| 212 | npp->p_fstype = opp->p_fstype; |
| 213 | npp->p_fsize = opp->p_fsize; |
| 214 | npp->p_frag = opp->p_frag; |
| 215 | npp->p_cpg = opp->p_cpg; |
| 216 | } |
| 217 | } |
| 218 | nlp->d_checksum = 0; |
| 219 | nlp->d_checksum = dkcksum(nlp); |
| 220 | *olp = *nlp; |
| 221 | return (0); |
| 222 | } |
| 223 | |
| 224 | /* encoding of disk minor numbers, should be elsewhere... */ |
| 225 | #define dkunit(dev) (minor(dev) >> 3) |
| 226 | #define dkpart(dev) (minor(dev) & 07) |
| 227 | #define dkminor(unit, part) (((unit) << 3) | (part)) |
| 228 | |
| 229 | /* |
| 230 | * Write disk label back to device after modification. |
| 231 | */ |
| 232 | writedisklabel(dev, strat, lp) |
| 233 | dev_t dev; |
| 234 | int (*strat)(); |
| 235 | register struct disklabel *lp; |
| 236 | { |
| 237 | struct buf *bp; |
| 238 | struct disklabel *dlp; |
| 239 | int labelpart; |
| 240 | int error = 0; |
| 241 | |
| 242 | labelpart = dkpart(dev); |
| 243 | if (lp->d_partitions[labelpart].p_offset != 0) { |
| 244 | if (lp->d_partitions[0].p_offset != 0) |
| 245 | return (EXDEV); /* not quite right */ |
| 246 | labelpart = 0; |
| 247 | } |
| 248 | bp = geteblk((int)lp->d_secsize); |
| 249 | bp->b_dev = makedev(major(dev), dkminor(dkunit(dev), labelpart)); |
| 250 | bp->b_blkno = LABELSECTOR; |
| 251 | bp->b_bcount = lp->d_secsize; |
| 252 | bp->b_flags = B_READ; |
| 253 | (*strat)(bp); |
| 254 | if (error = biowait(bp)) |
| 255 | goto done; |
| 256 | for (dlp = (struct disklabel *)bp->b_un.b_addr; |
| 257 | dlp <= (struct disklabel *) |
| 258 | (bp->b_un.b_addr + lp->d_secsize - sizeof(*dlp)); |
| 259 | dlp = (struct disklabel *)((char *)dlp + sizeof(long))) { |
| 260 | if (dlp->d_magic == DISKMAGIC && dlp->d_magic2 == DISKMAGIC && |
| 261 | dkcksum(dlp) == 0) { |
| 262 | *dlp = *lp; |
| 263 | bp->b_flags = B_WRITE; |
| 264 | (*strat)(bp); |
| 265 | error = biowait(bp); |
| 266 | goto done; |
| 267 | } |
| 268 | } |
| 269 | error = ESRCH; |
| 270 | done: |
| 271 | brelse(bp); |
| 272 | return (error); |
| 273 | } |
| 274 | |
| 275 | /* |
| 276 | * Compute checksum for disk label. |
| 277 | */ |
| 278 | dkcksum(lp) |
| 279 | register struct disklabel *lp; |
| 280 | { |
| 281 | register u_short *start, *end; |
| 282 | register u_short sum = 0; |
| 283 | |
| 284 | start = (u_short *)lp; |
| 285 | end = (u_short *)&lp->d_partitions[lp->d_npartitions]; |
| 286 | while (start < end) |
| 287 | sum ^= *start++; |
| 288 | return (sum); |
| 289 | } |
| 290 | |
| 291 | /* |
| 292 | * Disk error is the preface to plaintive error messages |
| 293 | * about failing disk transfers. It prints messages of the form |
| 294 | |
| 295 | hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d) |
| 296 | |
| 297 | * if the offset of the error in the transfer and a disk label |
| 298 | * are both available. blkdone should be -1 if the position of the error |
| 299 | * is unknown; the disklabel pointer may be null from drivers that have not |
| 300 | * been converted to use them. The message is printed with printf |
| 301 | * if pri is LOG_PRINTF, otherwise it uses log at the specified priority. |
| 302 | * The message should be completed (with at least a newline) with printf |
| 303 | * or addlog, respectively. There is no trailing space. |
| 304 | */ |
| 305 | diskerr(bp, dname, what, pri, blkdone, lp) |
| 306 | register struct buf *bp; |
| 307 | char *dname, *what; |
| 308 | int pri, blkdone; |
| 309 | register struct disklabel *lp; |
| 310 | { |
| 311 | int unit = dkunit(bp->b_dev), part = dkpart(bp->b_dev); |
| 312 | register int (*pr)(), sn; |
| 313 | char partname = 'a' + part; |
| 314 | extern printf(), addlog(); |
| 315 | |
| 316 | if (pri != LOG_PRINTF) { |
| 317 | log(pri, ""); |
| 318 | pr = addlog; |
| 319 | } else |
| 320 | pr = printf; |
| 321 | (*pr)("%s%d%c: %s %sing fsbn ", dname, unit, partname, what, |
| 322 | bp->b_flags & B_READ ? "read" : "writ"); |
| 323 | sn = bp->b_blkno; |
| 324 | if (bp->b_bcount <= DEV_BSIZE) |
| 325 | (*pr)("%d", sn); |
| 326 | else { |
| 327 | if (blkdone >= 0) { |
| 328 | sn += blkdone; |
| 329 | (*pr)("%d of ", sn); |
| 330 | } |
| 331 | (*pr)("%d-%d", bp->b_blkno, |
| 332 | bp->b_blkno + (bp->b_bcount - 1) / DEV_BSIZE); |
| 333 | } |
| 334 | if (lp && (blkdone >= 0 || bp->b_bcount <= lp->d_secsize)) { |
| 335 | #ifdef tahoe |
| 336 | sn *= DEV_BSIZE / lp->d_secsize; /* XXX */ |
| 337 | #endif |
| 338 | sn += lp->d_partitions[part].p_offset; |
| 339 | (*pr)(" (%s%d bn %d; cn %d", dname, unit, sn, |
| 340 | sn / lp->d_secpercyl); |
| 341 | sn %= lp->d_secpercyl; |
| 342 | (*pr)(" tn %d sn %d)", sn / lp->d_nsectors, sn % lp->d_nsectors); |
| 343 | } |
| 344 | } |