1*57193Smuller /*- 2*57193Smuller * Copyright (c) 1992 Keith Muller. 3*57193Smuller * Copyright (c) 1992 The Regents of the University of California. 4*57193Smuller * All rights reserved. 5*57193Smuller * 6*57193Smuller * This code is derived from software contributed to Berkeley by 7*57193Smuller * Keith Muller of the University of California, San Diego. 8*57193Smuller * 9*57193Smuller * %sccs.include.redist.c% 10*57193Smuller */ 11*57193Smuller 12*57193Smuller #ifndef lint 13*57193Smuller static char sccsid[] = "@(#)buf_subs.c 1.1 (Berkeley) 12/17/92"; 14*57193Smuller #endif /* not lint */ 15*57193Smuller 16*57193Smuller #include <sys/types.h> 17*57193Smuller #include <sys/time.h> 18*57193Smuller #include <sys/stat.h> 19*57193Smuller #include <sys/param.h> 20*57193Smuller #include <stdio.h> 21*57193Smuller #include <ctype.h> 22*57193Smuller #include <errno.h> 23*57193Smuller #include <unistd.h> 24*57193Smuller #include <stdlib.h> 25*57193Smuller #include <string.h> 26*57193Smuller #include "pax.h" 27*57193Smuller #include "extern.h" 28*57193Smuller 29*57193Smuller /* 30*57193Smuller * routines which implement archive and file buffering 31*57193Smuller */ 32*57193Smuller 33*57193Smuller #define MAXFBSZ 4096 /* max block size for hole detection */ 34*57193Smuller #define MINFBSZ 512 /* min block size for hole detection */ 35*57193Smuller #define MAXFLT 10 /* default media read error limit */ 36*57193Smuller 37*57193Smuller static char bufmem[MAXBLK+BLKMULT]; /* i/o buffer + pushback id space */ 38*57193Smuller static char *buf; /* normal start of i/o buffer */ 39*57193Smuller static char *bufend; /* end or last char in i/o buffer */ 40*57193Smuller static char *bufpt; /* read/write point in i/o buffer */ 41*57193Smuller int blksz = MAXBLK; /* block input/output size in bytes */ 42*57193Smuller int wrblksz; /* user spec output size in bytes */ 43*57193Smuller int maxflt = MAXFLT; /* MAX consecutive media errors */ 44*57193Smuller int rdblksz; /* first read blksize (tapes only) */ 45*57193Smuller off_t wrlimit; /* # of bytes written per archive vol */ 46*57193Smuller off_t wrcnt; /* # of bytes written on current vol */ 47*57193Smuller off_t rdcnt; /* # of bytes read on current vol */ 48*57193Smuller 49*57193Smuller /* 50*57193Smuller * wr_start() 51*57193Smuller * set up the buffering system to operate in a write mode 52*57193Smuller * Return: 53*57193Smuller * 0 if ok, -1 if the user specified write block size violates pax spec 54*57193Smuller */ 55*57193Smuller 56*57193Smuller #if __STDC__ 57*57193Smuller int 58*57193Smuller wr_start(void) 59*57193Smuller #else 60*57193Smuller int 61*57193Smuller wr_start() 62*57193Smuller #endif 63*57193Smuller { 64*57193Smuller buf = &(bufmem[BLKMULT]); 65*57193Smuller /* 66*57193Smuller * Check to make sure the write block size meets pax specs. If the user 67*57193Smuller * does not specify a blocksize, we use the format default blocksize. 68*57193Smuller * We must be picky on writes, so we do not allow the user to create an 69*57193Smuller * archive that might be hard to read elsewhere. If all ok, we then 70*57193Smuller * open the first archive volume 71*57193Smuller */ 72*57193Smuller if (!wrblksz) 73*57193Smuller wrblksz = frmt->bsz; 74*57193Smuller if (wrblksz > MAXBLK) { 75*57193Smuller warn(1, "Write block size of %d too large, maximium is: %d", 76*57193Smuller wrblksz, MAXBLK); 77*57193Smuller return(-1); 78*57193Smuller } 79*57193Smuller if (wrblksz % BLKMULT) { 80*57193Smuller warn(1, "Write block size of %d is not a %d byte multiple", 81*57193Smuller wrblksz, BLKMULT); 82*57193Smuller return(-1); 83*57193Smuller } 84*57193Smuller 85*57193Smuller /* 86*57193Smuller * we only allow wrblksz to be used with all archive operations 87*57193Smuller */ 88*57193Smuller blksz = rdblksz = wrblksz; 89*57193Smuller if ((ar_open(arcname) < 0) && (ar_next() < 0)) 90*57193Smuller return(-1); 91*57193Smuller wrcnt = 0; 92*57193Smuller bufend = buf + wrblksz; 93*57193Smuller bufpt = buf; 94*57193Smuller return(0); 95*57193Smuller } 96*57193Smuller 97*57193Smuller /* 98*57193Smuller * rd_start() 99*57193Smuller * set up buffering system to read an archive 100*57193Smuller * Return: 101*57193Smuller * 0 if ok, -1 otherwise 102*57193Smuller */ 103*57193Smuller 104*57193Smuller #if __STDC__ 105*57193Smuller int 106*57193Smuller rd_start(void) 107*57193Smuller #else 108*57193Smuller int 109*57193Smuller rd_start() 110*57193Smuller #endif 111*57193Smuller { 112*57193Smuller /* 113*57193Smuller * leave space for the header pushback (see get_arc()). If we are 114*57193Smuller * going to append and user specified a write block size, check it 115*57193Smuller * right away 116*57193Smuller */ 117*57193Smuller buf = &(bufmem[BLKMULT]); 118*57193Smuller if ((act == APPND) && wrblksz) { 119*57193Smuller if (wrblksz > MAXBLK) { 120*57193Smuller warn(1,"Write block size %d too large, maximium is: %d", 121*57193Smuller wrblksz, MAXBLK); 122*57193Smuller return(-1); 123*57193Smuller } 124*57193Smuller if (wrblksz % BLKMULT) { 125*57193Smuller warn(1, "Write block size %d is not a %d byte multiple", 126*57193Smuller wrblksz, BLKMULT); 127*57193Smuller return(-1); 128*57193Smuller } 129*57193Smuller } 130*57193Smuller 131*57193Smuller /* 132*57193Smuller * open the archive 133*57193Smuller */ 134*57193Smuller if ((ar_open(arcname) < 0) && (ar_next() < 0)) 135*57193Smuller return(-1); 136*57193Smuller bufend = buf + rdblksz; 137*57193Smuller bufpt = bufend; 138*57193Smuller rdcnt = 0; 139*57193Smuller return(0); 140*57193Smuller } 141*57193Smuller 142*57193Smuller /* 143*57193Smuller * cp_start() 144*57193Smuller * set up buffer system for copying within the file system 145*57193Smuller */ 146*57193Smuller 147*57193Smuller #if __STDC__ 148*57193Smuller void 149*57193Smuller cp_start(void) 150*57193Smuller #else 151*57193Smuller void 152*57193Smuller cp_start() 153*57193Smuller #endif 154*57193Smuller { 155*57193Smuller buf = &(bufmem[BLKMULT]); 156*57193Smuller rdblksz = blksz = MAXBLK; 157*57193Smuller } 158*57193Smuller 159*57193Smuller /* 160*57193Smuller * appnd_start() 161*57193Smuller * Set up the buffering system to append new members to an archive that 162*57193Smuller * was just read. The last block(s) of an archive may contain a format 163*57193Smuller * specific trailer. To append a new member, this trailer has to be 164*57193Smuller * removed from the archive. The first byte of the trailer is replaced by 165*57193Smuller * the start of the header of the first file added to the archive. The 166*57193Smuller * format specific end read function tells us how many bytes to move 167*57193Smuller * backwards in the archive to be positioned BEFORE the trailer. Two 168*57193Smuller * different postions have to be adjusted, the O.S. file offset (e.g. the 169*57193Smuller * position of the tape head) and the write point within the data we have 170*57193Smuller * stored in the read (soon to become write) buffer. We may have to move 171*57193Smuller * back several records (the number depends on the size of the archive 172*57193Smuller * record and the size of the format trailer) to read up the record where 173*57193Smuller * the first byte of the trailer is recorded. Trailers may span (and 174*57193Smuller * overlap) record boundries. 175*57193Smuller * We first calculate which record has the first byte of the trailer. We 176*57193Smuller * move the OS file offset back to the start of this record and read it 177*57193Smuller * up. We set the buffer write pointer to be at this byte (the byte where 178*57193Smuller * the trailer starts). We then move the OS file pointer back to the 179*57193Smuller * start of this record so a flush of this buffer will replace the record 180*57193Smuller * in the archive. 181*57193Smuller * A major problem is rewriting this last record. For archives stored 182*57193Smuller * on disk files, this is trival. However, many devices are really picky 183*57193Smuller * about the conditions under which they will allow a write to occur. 184*57193Smuller * Often devices restrict the conditions where writes can be made writes, 185*57193Smuller * so it may not be feasable to append archives stored on all types of 186*57193Smuller * devices. 187*57193Smuller * Return: 188*57193Smuller * 0 for success, -1 for failure 189*57193Smuller */ 190*57193Smuller 191*57193Smuller #if __STDC__ 192*57193Smuller int 193*57193Smuller appnd_start(off_t skcnt) 194*57193Smuller #else 195*57193Smuller int 196*57193Smuller appnd_start(skcnt) 197*57193Smuller off_t skcnt; 198*57193Smuller #endif 199*57193Smuller { 200*57193Smuller register int res; 201*57193Smuller off_t cnt; 202*57193Smuller 203*57193Smuller if (exit_val != 0) 204*57193Smuller return(-1); 205*57193Smuller /* 206*57193Smuller * if the user did not specify a write blocksize, inherit the size used 207*57193Smuller * in the last archive volume read. (If a is set we still use rdblksz 208*57193Smuller * until next volume, cannot shift sizes within a single volume). 209*57193Smuller */ 210*57193Smuller if (!wrblksz) 211*57193Smuller wrblksz = blksz = rdblksz; 212*57193Smuller else 213*57193Smuller blksz = rdblksz; 214*57193Smuller 215*57193Smuller /* 216*57193Smuller * make sure that this volume allows appends 217*57193Smuller */ 218*57193Smuller if (ar_app_ok() < 0) 219*57193Smuller return(-1); 220*57193Smuller 221*57193Smuller /* 222*57193Smuller * Calculate bytes to move back and move in front of record where we 223*57193Smuller * need to start writing from. Remember we have to add in any padding 224*57193Smuller * that might be in the buffer after the trailer in the last block. We 225*57193Smuller * travel skcnt + padding ROUNDED UP to blksize. 226*57193Smuller */ 227*57193Smuller skcnt += bufend - bufpt; 228*57193Smuller if ((cnt = (skcnt/blksz) * blksz) < skcnt) 229*57193Smuller cnt += blksz; 230*57193Smuller if (ar_rev((off_t)cnt) < 0) 231*57193Smuller goto out; 232*57193Smuller 233*57193Smuller /* 234*57193Smuller * We may have gone too far if there is valid data in the block we are 235*57193Smuller * now in front of, read up the block and position the pointer after 236*57193Smuller * the valid data. 237*57193Smuller */ 238*57193Smuller if ((cnt -= skcnt) > 0) { 239*57193Smuller /* 240*57193Smuller * watch out for stupid tape drives. ar_rev() will set rdblksz 241*57193Smuller * to be real physical blocksize so we must loop until we get 242*57193Smuller * the old rdblksz (now in blksz). If ar_rev() fouls up the 243*57193Smuller * determination of the physical block size, we will fail. 244*57193Smuller */ 245*57193Smuller bufpt = buf; 246*57193Smuller bufend = buf + blksz; 247*57193Smuller while (bufpt < bufend) { 248*57193Smuller if ((res = ar_read(bufpt, rdblksz)) <= 0) 249*57193Smuller goto out; 250*57193Smuller bufpt += res; 251*57193Smuller } 252*57193Smuller if (ar_rev((off_t)(bufpt - buf)) < 0) 253*57193Smuller goto out; 254*57193Smuller bufpt = buf + cnt; 255*57193Smuller bufend = buf + blksz; 256*57193Smuller } else { 257*57193Smuller /* 258*57193Smuller * buffer is empty 259*57193Smuller */ 260*57193Smuller bufend = buf + blksz; 261*57193Smuller bufpt = buf; 262*57193Smuller } 263*57193Smuller rdblksz = blksz; 264*57193Smuller rdcnt -= skcnt; 265*57193Smuller wrcnt = 0; 266*57193Smuller 267*57193Smuller /* 268*57193Smuller * At this point we are ready to write. If the device requires special 269*57193Smuller * handling to write at a point were previously recorded data resides, 270*57193Smuller * that is handled in ar_set_wr(). From now on we operate under normal 271*57193Smuller * ARCHIVE mode (write) conditions 272*57193Smuller */ 273*57193Smuller if (ar_set_wr() < 0) 274*57193Smuller return(-1); 275*57193Smuller act = ARCHIVE; 276*57193Smuller return(0); 277*57193Smuller 278*57193Smuller out: 279*57193Smuller warn(1, "Unable to position at the end of the archive, append failed"); 280*57193Smuller return(-1); 281*57193Smuller } 282*57193Smuller 283*57193Smuller /* 284*57193Smuller * rd_sync() 285*57193Smuller * A read error occurred on this archive volume. Resync the buffer and 286*57193Smuller * try to reset the device (if possible) so we can continue to read. Keep 287*57193Smuller * trying to do this until we get a valid read, or we reach the limit on 288*57193Smuller * consecutive read faults (at which point we give up). The user can 289*57193Smuller * adjust the read error limit through a command line option. 290*57193Smuller * Returns: 291*57193Smuller * 0 on success, and -1 on failure 292*57193Smuller */ 293*57193Smuller 294*57193Smuller #if __STDC__ 295*57193Smuller int 296*57193Smuller rd_sync(void) 297*57193Smuller #else 298*57193Smuller int 299*57193Smuller rd_sync() 300*57193Smuller #endif 301*57193Smuller { 302*57193Smuller register int errcnt = 0; 303*57193Smuller register int res; 304*57193Smuller 305*57193Smuller /* 306*57193Smuller * if the user says bail out on first fault, we are out of here... 307*57193Smuller */ 308*57193Smuller if (maxflt == 0) 309*57193Smuller return(-1); 310*57193Smuller if (act == APPND) { 311*57193Smuller warn(1, "Unable to append when there are archive read errors."); 312*57193Smuller return(-1); 313*57193Smuller } 314*57193Smuller 315*57193Smuller /* 316*57193Smuller * poke at device and try to get past media error 317*57193Smuller */ 318*57193Smuller if (ar_rdsync() < 0) { 319*57193Smuller if (ar_next() < 0) 320*57193Smuller return(-1); 321*57193Smuller else 322*57193Smuller rdcnt = 0; 323*57193Smuller } 324*57193Smuller 325*57193Smuller for (;;) { 326*57193Smuller if ((res = ar_read(buf, blksz)) > 0) { 327*57193Smuller /* 328*57193Smuller * All right! got some data, fill that buffer 329*57193Smuller */ 330*57193Smuller bufpt = buf; 331*57193Smuller bufend = buf + res; 332*57193Smuller rdcnt += res; 333*57193Smuller return(0); 334*57193Smuller } 335*57193Smuller 336*57193Smuller /* 337*57193Smuller * Oh well, yet another failed read... 338*57193Smuller * if error limit reached, ditch. o.w. poke device to move past 339*57193Smuller * bad media and try again. if media is badly damaged, we ask 340*57193Smuller * the poor (and upset user at this point) for the next archive 341*57193Smuller * volume. remember the goal on reads is to get the most we 342*57193Smuller * can extract out of the archive. 343*57193Smuller */ 344*57193Smuller if ((maxflt > 0) && (++errcnt > maxflt)) 345*57193Smuller warn(1,"Archive read error limit (%d) reached",maxflt); 346*57193Smuller else if (ar_rdsync() == 0) 347*57193Smuller continue; 348*57193Smuller if (ar_next() < 0) 349*57193Smuller break; 350*57193Smuller rdcnt = 0; 351*57193Smuller errcnt = 0; 352*57193Smuller } 353*57193Smuller return(-1); 354*57193Smuller } 355*57193Smuller 356*57193Smuller /* 357*57193Smuller * pback() 358*57193Smuller * push the data used during the archive id phase back into the I/O 359*57193Smuller * buffer. This is required as we cannot be sure that the header does NOT 360*57193Smuller * overlap a block boundry (as in the case we are trying to recover a 361*57193Smuller * flawed archived). This was not designed to be used for any other 362*57193Smuller * purpose. (What software engineering, HA!) 363*57193Smuller * WARNING: do not even THINK of pback greater than BLKMULT, unless the 364*57193Smuller * pback space is increased. 365*57193Smuller */ 366*57193Smuller 367*57193Smuller #if __STDC__ 368*57193Smuller void 369*57193Smuller pback(char *pt, int cnt) 370*57193Smuller #else 371*57193Smuller void 372*57193Smuller pback(pt, cnt) 373*57193Smuller char *pt; 374*57193Smuller int cnt; 375*57193Smuller #endif 376*57193Smuller { 377*57193Smuller bufpt -= cnt; 378*57193Smuller bcopy(pt, bufpt, cnt); 379*57193Smuller return; 380*57193Smuller } 381*57193Smuller 382*57193Smuller /* 383*57193Smuller * rd_skip() 384*57193Smuller * skip foward in the archive during a archive read. Used to get quickly 385*57193Smuller * past file data and padding for files the user did NOT select. 386*57193Smuller * Return: 387*57193Smuller * 0 if ok, -1 failure, and 1 when EOF on the archive volume was detected. 388*57193Smuller */ 389*57193Smuller 390*57193Smuller #if __STDC__ 391*57193Smuller int 392*57193Smuller rd_skip(off_t skcnt) 393*57193Smuller #else 394*57193Smuller int 395*57193Smuller rd_skip(skcnt) 396*57193Smuller off_t skcnt; 397*57193Smuller #endif 398*57193Smuller { 399*57193Smuller off_t res; 400*57193Smuller off_t cnt; 401*57193Smuller off_t skipped = 0; 402*57193Smuller 403*57193Smuller /* 404*57193Smuller * consume what data we have in the buffer. If we have to move foward 405*57193Smuller * whole records, we call the low level skip function to see if we can 406*57193Smuller * move within the archive without doing the expensive reads on data we 407*57193Smuller * do not want. 408*57193Smuller */ 409*57193Smuller if (skcnt == 0) 410*57193Smuller return(0); 411*57193Smuller res = MIN((bufend - bufpt), skcnt); 412*57193Smuller bufpt += res; 413*57193Smuller skcnt -= res; 414*57193Smuller 415*57193Smuller /* 416*57193Smuller * if skcnt is now 0, then no additional i/o is needed 417*57193Smuller */ 418*57193Smuller if (skcnt == 0) 419*57193Smuller return(0); 420*57193Smuller 421*57193Smuller /* 422*57193Smuller * We have to read more, calculate complete and partial record reads 423*57193Smuller * based on rdblksz. we skip over "cnt" complete records 424*57193Smuller */ 425*57193Smuller res = skcnt%rdblksz; 426*57193Smuller cnt = (skcnt/rdblksz) * rdblksz; 427*57193Smuller 428*57193Smuller /* 429*57193Smuller * if the skip fails, we will have to resync. ar_fow will tell us 430*57193Smuller * how much it can skip over. We will have to read the rest. 431*57193Smuller */ 432*57193Smuller if (ar_fow(cnt, &skipped) < 0) 433*57193Smuller return(-1); 434*57193Smuller res += cnt - skipped; 435*57193Smuller rdcnt += skipped; 436*57193Smuller 437*57193Smuller /* 438*57193Smuller * what is left we have to read (which may be the whole thing if 439*57193Smuller * ar_fow() told us the device can only read to skip records); 440*57193Smuller */ 441*57193Smuller while (res > 0L) { 442*57193Smuller cnt = bufend - bufpt; 443*57193Smuller /* 444*57193Smuller * if the read fails, we will have to resync 445*57193Smuller */ 446*57193Smuller if ((cnt <= 0) && ((cnt = buf_fill()) < 0)) 447*57193Smuller return(-1); 448*57193Smuller if (cnt == 0) 449*57193Smuller return(1); 450*57193Smuller cnt = MIN(cnt, res); 451*57193Smuller bufpt += cnt; 452*57193Smuller res -= cnt; 453*57193Smuller } 454*57193Smuller return(0); 455*57193Smuller } 456*57193Smuller 457*57193Smuller /* 458*57193Smuller * wr_fin() 459*57193Smuller * flush out any data (and pad if required) the last block. We always pad 460*57193Smuller * with zero (even though we do not have to). Padding with 0 makes it a 461*57193Smuller * lot easier to recover if the archive is damaged. zero paddding SHOULD 462*57193Smuller * BE a requirement.... 463*57193Smuller */ 464*57193Smuller 465*57193Smuller #if __STDC__ 466*57193Smuller void 467*57193Smuller wr_fin(void) 468*57193Smuller #else 469*57193Smuller void 470*57193Smuller wr_fin() 471*57193Smuller #endif 472*57193Smuller { 473*57193Smuller if (bufpt > buf) { 474*57193Smuller bzero(bufpt, bufend - bufpt); 475*57193Smuller bufpt = bufend; 476*57193Smuller (void)buf_flush(blksz); 477*57193Smuller } 478*57193Smuller } 479*57193Smuller 480*57193Smuller /* 481*57193Smuller * wr_rdbuf() 482*57193Smuller * fill the write buffer from data passed to it in a buffer (usually used 483*57193Smuller * by format specific write routines to pass a file header). On failure we 484*57193Smuller * punt. We do not allow the user to continue to write flawed archives. 485*57193Smuller * We assume these headers are not very large (the memory copy we use is 486*57193Smuller * a bit expensive). 487*57193Smuller * Return: 488*57193Smuller * 0 if buffer was filled ok, -1 o.w. (buffer flush failure) 489*57193Smuller */ 490*57193Smuller 491*57193Smuller #if __STDC__ 492*57193Smuller int 493*57193Smuller wr_rdbuf(register char *out, register int outcnt) 494*57193Smuller #else 495*57193Smuller int 496*57193Smuller wr_rdbuf(out, outcnt) 497*57193Smuller register char *out; 498*57193Smuller register int outcnt; 499*57193Smuller #endif 500*57193Smuller { 501*57193Smuller register int cnt; 502*57193Smuller 503*57193Smuller /* 504*57193Smuller * while there is data to copy copy into the write buffer. when the 505*57193Smuller * write buffer fills, flush it to the archive and continue 506*57193Smuller */ 507*57193Smuller while (outcnt > 0) { 508*57193Smuller cnt = bufend - bufpt; 509*57193Smuller if ((cnt <= 0) && ((cnt = buf_flush(blksz)) < 0)) 510*57193Smuller return(-1); 511*57193Smuller /* 512*57193Smuller * only move what we have space for 513*57193Smuller */ 514*57193Smuller cnt = MIN(cnt, outcnt); 515*57193Smuller bcopy(out, bufpt, cnt); 516*57193Smuller bufpt += cnt; 517*57193Smuller out += cnt; 518*57193Smuller outcnt -= cnt; 519*57193Smuller } 520*57193Smuller return(0); 521*57193Smuller } 522*57193Smuller 523*57193Smuller /* 524*57193Smuller * rd_wrbuf() 525*57193Smuller * copy from the read buffer into a supplied buffer a specified number of 526*57193Smuller * bytes. If the read buffer is empty fill it and continue to copy. 527*57193Smuller * usually used to obtain a file header for processing by a format 528*57193Smuller * specific read routine. 529*57193Smuller * Return 530*57193Smuller * number of bytes copied to the buffer, 0 indicates EOF on archive volume, 531*57193Smuller * -1 is a read error 532*57193Smuller */ 533*57193Smuller 534*57193Smuller #if __STDC__ 535*57193Smuller int 536*57193Smuller rd_wrbuf(register char *in, register int cpcnt) 537*57193Smuller #else 538*57193Smuller int 539*57193Smuller rd_wrbuf(in, cpcnt) 540*57193Smuller register char *in; 541*57193Smuller register int cpcnt; 542*57193Smuller #endif 543*57193Smuller { 544*57193Smuller register int res; 545*57193Smuller register int cnt; 546*57193Smuller register int incnt = cpcnt; 547*57193Smuller 548*57193Smuller /* 549*57193Smuller * loop until we fill the buffer with the requested number of bytes 550*57193Smuller */ 551*57193Smuller while (incnt > 0) { 552*57193Smuller cnt = bufend - bufpt; 553*57193Smuller if ((cnt <= 0) && ((cnt = buf_fill()) <= 0)) { 554*57193Smuller /* 555*57193Smuller * read error, return what we got (or the error if 556*57193Smuller * no data was copied). The caller must know that an 557*57193Smuller * error occured and has the best knowledge what to 558*57193Smuller * do with it 559*57193Smuller */ 560*57193Smuller if ((res = cpcnt - incnt) > 0) 561*57193Smuller return(res); 562*57193Smuller return(cnt); 563*57193Smuller } 564*57193Smuller 565*57193Smuller /* 566*57193Smuller * calculate how much data to copy based on whats left and 567*57193Smuller * state of buffer 568*57193Smuller */ 569*57193Smuller cnt = MIN(cnt, incnt); 570*57193Smuller bcopy(bufpt, in, cnt); 571*57193Smuller bufpt += cnt; 572*57193Smuller incnt -= cnt; 573*57193Smuller in += cnt; 574*57193Smuller } 575*57193Smuller return(cpcnt); 576*57193Smuller } 577*57193Smuller 578*57193Smuller /* 579*57193Smuller * wr_skip() 580*57193Smuller * skip foward during a write. In other words add padding to the file. 581*57193Smuller * we add zero filled padding as it makes flawed archives much easier to 582*57193Smuller * recover from. the caller tells us how many bytes of padding to add 583*57193Smuller * This routine was not designed to add HUGE amount of padding, just small 584*57193Smuller * amounts (a few 512 byte blocks at most) 585*57193Smuller * Return: 586*57193Smuller * 0 if ok, -1 if there was a buf_flush failure 587*57193Smuller */ 588*57193Smuller 589*57193Smuller #if __STDC__ 590*57193Smuller int 591*57193Smuller wr_skip(off_t skcnt) 592*57193Smuller #else 593*57193Smuller int 594*57193Smuller wr_skip(skcnt) 595*57193Smuller off_t skcnt; 596*57193Smuller #endif 597*57193Smuller { 598*57193Smuller register int cnt; 599*57193Smuller 600*57193Smuller /* 601*57193Smuller * loop while there is more padding to add 602*57193Smuller */ 603*57193Smuller while (skcnt > 0L) { 604*57193Smuller cnt = bufend - bufpt; 605*57193Smuller if ((cnt <= 0) && ((cnt = buf_flush(blksz)) < 0)) 606*57193Smuller return(-1); 607*57193Smuller cnt = MIN(cnt, skcnt); 608*57193Smuller bzero(bufpt, cnt); 609*57193Smuller bufpt += cnt; 610*57193Smuller skcnt -= cnt; 611*57193Smuller } 612*57193Smuller return(0); 613*57193Smuller } 614*57193Smuller 615*57193Smuller /* 616*57193Smuller * wr_rdfile() 617*57193Smuller * fill write buffer with the contents of a file. We are passed an open 618*57193Smuller * file descriptor to the file an the archive structure that describes the 619*57193Smuller * file we are storing. The variable "left" is modified to contain the 620*57193Smuller * number of bytes of the file we were NOT able to write to the archive. 621*57193Smuller * it is important that we always write EXACTLY the number of bytes that 622*57193Smuller * the format specific write routine told us to. The file can also get 623*57193Smuller * bigger, so reading to the end of file would create an improper archive, 624*57193Smuller * we just detect this case and warn the user. We never create a bad 625*57193Smuller * archive if we can avoid it. Of course trying to archive files that are 626*57193Smuller * active is asking for trouble. It we fail, we pass back how much we 627*57193Smuller * could NOT copy and let the caller deal with it. 628*57193Smuller * Return: 629*57193Smuller * 0 ok, -1 if archive write failure. a short read of the file returns a 630*57193Smuller * 0, but "left" is set to be greater than zero. 631*57193Smuller */ 632*57193Smuller 633*57193Smuller #if __STDC__ 634*57193Smuller int 635*57193Smuller wr_rdfile(ARCHD *arcn, int ifd, off_t *left) 636*57193Smuller #else 637*57193Smuller int 638*57193Smuller wr_rdfile(arcn, ifd, left) 639*57193Smuller ARCHD *arcn; 640*57193Smuller int ifd; 641*57193Smuller off_t *left; 642*57193Smuller #endif 643*57193Smuller { 644*57193Smuller register int cnt; 645*57193Smuller register int res = 0; 646*57193Smuller register off_t size = arcn->sb.st_size; 647*57193Smuller struct stat sb; 648*57193Smuller 649*57193Smuller /* 650*57193Smuller * while there are more bytes to write 651*57193Smuller */ 652*57193Smuller while (size > 0L) { 653*57193Smuller cnt = bufend - bufpt; 654*57193Smuller if ((cnt <= 0) && ((cnt = buf_flush(blksz)) < 0)) { 655*57193Smuller *left = size; 656*57193Smuller return(-1); 657*57193Smuller } 658*57193Smuller cnt = MIN(cnt, size); 659*57193Smuller if ((res = read(ifd, bufpt, cnt)) <= 0) 660*57193Smuller break; 661*57193Smuller size -= res; 662*57193Smuller bufpt += res; 663*57193Smuller } 664*57193Smuller 665*57193Smuller /* 666*57193Smuller * better check the file did not change during this operation 667*57193Smuller * or the file read failed. 668*57193Smuller */ 669*57193Smuller if (res < 0) 670*57193Smuller syswarn(1, errno, "Read fault on %s", arcn->org_name); 671*57193Smuller else if (size != 0L) 672*57193Smuller warn(1, "File changed size during read %s", arcn->org_name); 673*57193Smuller else if (fstat(ifd, &sb) < 0) 674*57193Smuller syswarn(1, errno, "Failed stat on %s", arcn->org_name); 675*57193Smuller else if (arcn->sb.st_mtime != sb.st_mtime) 676*57193Smuller warn(1, "File %s was modified during copy to archive", 677*57193Smuller arcn->org_name); 678*57193Smuller *left = size; 679*57193Smuller return(0); 680*57193Smuller } 681*57193Smuller 682*57193Smuller /* 683*57193Smuller * rd_wrfile() 684*57193Smuller * extract the contents of a file from the archive. If we are unable to 685*57193Smuller * extract the entire file (due to failure to write the file) we return 686*57193Smuller * the numbers of bytes we did NOT process. This way the caller knows how 687*57193Smuller * many bytes to skip past to find the next archive header. If the failure 688*57193Smuller * was due to an archive read, we will catch that when we try to skip. If 689*57193Smuller * the format supplies a file data crc value, we calculate the actual crc 690*57193Smuller * so that it can be compared to the value stored in the header 691*57193Smuller * NOTE: 692*57193Smuller * We call a special function to write the file. This function attempts to 693*57193Smuller * restore file holes (blocks of zeros) into the file. When files are 694*57193Smuller * sparse this saves space, and is a LOT faster. For non sparse files 695*57193Smuller * the performance hit is small. As of this writing, no archive supports 696*57193Smuller * information on where the file holes are. 697*57193Smuller * Return: 698*57193Smuller * 0 ok, -1 if archive read failure. if we cannot write the entire file, 699*57193Smuller * we return a 0 but "left" is set to be the amount unwritten 700*57193Smuller */ 701*57193Smuller 702*57193Smuller #if __STDC__ 703*57193Smuller int 704*57193Smuller rd_wrfile(ARCHD *arcn, int ofd, off_t *left) 705*57193Smuller #else 706*57193Smuller int 707*57193Smuller rd_wrfile(arcn, ofd, left) 708*57193Smuller ARCHD *arcn; 709*57193Smuller int ofd; 710*57193Smuller off_t *left; 711*57193Smuller #endif 712*57193Smuller { 713*57193Smuller register int cnt = 0; 714*57193Smuller register off_t size = arcn->sb.st_size; 715*57193Smuller register int res = 0; 716*57193Smuller register char *fnm = arcn->name; 717*57193Smuller int isem = 1; 718*57193Smuller int rem; 719*57193Smuller int sz = MINFBSZ; 720*57193Smuller struct stat sb; 721*57193Smuller u_long crc = 0L; 722*57193Smuller 723*57193Smuller /* 724*57193Smuller * pass the blocksize of the file being written to the write routine, 725*57193Smuller * if an odd size, use the default MINFBSZ 726*57193Smuller */ 727*57193Smuller if (fstat(ofd, &sb) == 0) { 728*57193Smuller if ((sb.st_blksize > 0) && (sb.st_blksize <= MAXFBSZ)) 729*57193Smuller sz = (int)sb.st_blksize; 730*57193Smuller } else 731*57193Smuller syswarn(0,errno,"Unable to obtain block size for file %s",fnm); 732*57193Smuller rem = sz; 733*57193Smuller *left = 0L; 734*57193Smuller 735*57193Smuller /* 736*57193Smuller * Copy the archive to the file the number of bytes specified. We have 737*57193Smuller * to assume that we want to recover file holes as none of the archive 738*57193Smuller * formats can record the location of file holes. 739*57193Smuller */ 740*57193Smuller while (size > 0L) { 741*57193Smuller cnt = bufend - bufpt; 742*57193Smuller /* 743*57193Smuller * if we get a read error, we do not want to skip, as we may 744*57193Smuller * miss a header, so we do not set left, but if we get a write 745*57193Smuller * error, we do want to skip over the unprocessed data. 746*57193Smuller */ 747*57193Smuller if ((cnt <= 0) && ((cnt = buf_fill()) <= 0)) 748*57193Smuller break; 749*57193Smuller cnt = MIN(cnt, size); 750*57193Smuller if ((res = file_write(ofd,bufpt,cnt,&rem,&isem,sz,fnm)) <= 0) { 751*57193Smuller *left = size; 752*57193Smuller break; 753*57193Smuller } 754*57193Smuller 755*57193Smuller if (docrc) { 756*57193Smuller /* 757*57193Smuller * update the actual crc value 758*57193Smuller */ 759*57193Smuller cnt = res; 760*57193Smuller while (--cnt >= 0) 761*57193Smuller crc += *bufpt++ & 0xff; 762*57193Smuller } else 763*57193Smuller bufpt += res; 764*57193Smuller size -= res; 765*57193Smuller } 766*57193Smuller 767*57193Smuller /* 768*57193Smuller * if the last block has a file hole (all zero), we must make sure this 769*57193Smuller * gets updated in the file. We force the last block of zeros to be 770*57193Smuller * written. just closing with the file offset moved foward may not put 771*57193Smuller * a hole at the end of the file. 772*57193Smuller */ 773*57193Smuller if (isem && (arcn->sb.st_size > 0L)) 774*57193Smuller file_flush(ofd, fnm, isem); 775*57193Smuller 776*57193Smuller /* 777*57193Smuller * if we failed from archive read, we do not want to skip 778*57193Smuller */ 779*57193Smuller if ((size > 0L) && (*left == 0L)) 780*57193Smuller return(-1); 781*57193Smuller 782*57193Smuller /* 783*57193Smuller * some formats record a crc on file data. If so, then we compare the 784*57193Smuller * calculated crc to the crc stored in the archive 785*57193Smuller */ 786*57193Smuller if (docrc && (size == 0L) && (arcn->crc != crc)) 787*57193Smuller warn(1,"Actual crc does not match expected crc %s",arcn->name); 788*57193Smuller return(0); 789*57193Smuller } 790*57193Smuller 791*57193Smuller /* 792*57193Smuller * cp_file() 793*57193Smuller * copy the contents of one file to another. used during -rw phase of pax 794*57193Smuller * just as in rd_wrfile() we use a special write function to write the 795*57193Smuller * destination file so we can properly copy files with holes. 796*57193Smuller */ 797*57193Smuller 798*57193Smuller #if __STDC__ 799*57193Smuller void 800*57193Smuller cp_file(ARCHD *arcn, int fd1, int fd2) 801*57193Smuller #else 802*57193Smuller void 803*57193Smuller cp_file(arcn, fd1, fd2) 804*57193Smuller ARCHD *arcn; 805*57193Smuller int fd1; 806*57193Smuller int fd2; 807*57193Smuller #endif 808*57193Smuller { 809*57193Smuller register int cnt; 810*57193Smuller register off_t cpcnt = 0L; 811*57193Smuller register int res = 0; 812*57193Smuller register char *fnm = arcn->name; 813*57193Smuller register int no_hole = 0; 814*57193Smuller int isem = 1; 815*57193Smuller int rem; 816*57193Smuller int sz = MINFBSZ; 817*57193Smuller struct stat sb; 818*57193Smuller 819*57193Smuller /* 820*57193Smuller * check for holes in the source file. If none, we will use regular 821*57193Smuller * write instead of file write. 822*57193Smuller */ 823*57193Smuller if (((off_t)(arcn->sb.st_blocks * BLKMULT)) >= arcn->sb.st_size) 824*57193Smuller ++no_hole; 825*57193Smuller 826*57193Smuller /* 827*57193Smuller * pass the blocksize of the file being written to the write routine, 828*57193Smuller * if an odd size, use the default MINFBSZ 829*57193Smuller */ 830*57193Smuller if (fstat(fd2, &sb) == 0) { 831*57193Smuller if ((sb.st_blksize > 0) && (sb.st_blksize <= MAXFBSZ)) 832*57193Smuller sz = sb.st_blksize; 833*57193Smuller } else 834*57193Smuller syswarn(0,errno,"Unable to obtain block size for file %s",fnm); 835*57193Smuller rem = sz; 836*57193Smuller 837*57193Smuller /* 838*57193Smuller * read the source file and copy to destination file until EOF 839*57193Smuller */ 840*57193Smuller for(;;) { 841*57193Smuller if ((cnt = read(fd1, buf, blksz)) <= 0) 842*57193Smuller break; 843*57193Smuller if (no_hole) 844*57193Smuller res = write(fd2, buf, cnt); 845*57193Smuller else 846*57193Smuller res = file_write(fd2, buf, cnt, &rem, &isem, sz, fnm); 847*57193Smuller if (res != cnt) 848*57193Smuller break; 849*57193Smuller cpcnt += cnt; 850*57193Smuller } 851*57193Smuller 852*57193Smuller /* 853*57193Smuller * check to make sure the copy is valid. 854*57193Smuller */ 855*57193Smuller if (res < 0) 856*57193Smuller syswarn(1, errno, "Failed write during copy of %s to %s", 857*57193Smuller arcn->org_name, arcn->name); 858*57193Smuller else if (cpcnt != arcn->sb.st_size) 859*57193Smuller warn(1, "File %s changed size during copy to %s", 860*57193Smuller arcn->org_name, arcn->name); 861*57193Smuller else if (fstat(fd1, &sb) < 0) 862*57193Smuller syswarn(1, errno, "Failed stat of %s", arcn->org_name); 863*57193Smuller else if (arcn->sb.st_mtime != sb.st_mtime) 864*57193Smuller warn(1, "File %s was modified during copy to %s", 865*57193Smuller arcn->org_name, arcn->name); 866*57193Smuller 867*57193Smuller /* 868*57193Smuller * if the last block has a file hole (all zero), we must make sure this 869*57193Smuller * gets updated in the file. We force the last block of zeros to be 870*57193Smuller * written. just closing with the file offset moved foward may not put 871*57193Smuller * a hole at the end of the file. 872*57193Smuller */ 873*57193Smuller if (!no_hole && isem && (arcn->sb.st_size > 0L)) 874*57193Smuller file_flush(fd2, fnm, isem); 875*57193Smuller return; 876*57193Smuller } 877*57193Smuller 878*57193Smuller /* 879*57193Smuller * buf_fill() 880*57193Smuller * fill the read buffer with the next record (or what we can get) from 881*57193Smuller * the archive volume. 882*57193Smuller * Return: 883*57193Smuller * Number of bytes of data in the read buffer, -1 for read error, and 884*57193Smuller * 0 when finished (user specified termination in ar_next()). 885*57193Smuller */ 886*57193Smuller 887*57193Smuller #if __STDC__ 888*57193Smuller int 889*57193Smuller buf_fill(void) 890*57193Smuller #else 891*57193Smuller int 892*57193Smuller buf_fill() 893*57193Smuller #endif 894*57193Smuller { 895*57193Smuller register int cnt; 896*57193Smuller static int fini = 0; 897*57193Smuller 898*57193Smuller if (fini) 899*57193Smuller return(0); 900*57193Smuller 901*57193Smuller for(;;) { 902*57193Smuller /* 903*57193Smuller * try to fill the buffer. on error the next archive volume is 904*57193Smuller * opened and we try again. 905*57193Smuller */ 906*57193Smuller if ((cnt = ar_read(buf, blksz)) > 0) { 907*57193Smuller bufpt = buf; 908*57193Smuller bufend = buf + cnt; 909*57193Smuller rdcnt += cnt; 910*57193Smuller return(cnt); 911*57193Smuller } 912*57193Smuller 913*57193Smuller /* 914*57193Smuller * errors require resync, EOF goes to next archive 915*57193Smuller */ 916*57193Smuller if (cnt < 0) 917*57193Smuller break; 918*57193Smuller if (ar_next() < 0) { 919*57193Smuller fini = 1; 920*57193Smuller return(0); 921*57193Smuller } 922*57193Smuller rdcnt = 0; 923*57193Smuller } 924*57193Smuller exit_val = 1; 925*57193Smuller return(-1); 926*57193Smuller } 927*57193Smuller 928*57193Smuller /* 929*57193Smuller * buf_flush() 930*57193Smuller * force the write buffer to the archive. We are passed the number of 931*57193Smuller * bytes in the buffer at the point of the flush. When we change archives 932*57193Smuller * the record size might change. (either larger or smaller). 933*57193Smuller * Return: 934*57193Smuller * 0 if all is ok, -1 when a write error occurs. 935*57193Smuller */ 936*57193Smuller 937*57193Smuller #if __STDC__ 938*57193Smuller int 939*57193Smuller buf_flush(register int bufcnt) 940*57193Smuller #else 941*57193Smuller int 942*57193Smuller buf_flush(bufcnt) 943*57193Smuller register int bufcnt; 944*57193Smuller #endif 945*57193Smuller { 946*57193Smuller register int cnt; 947*57193Smuller register int push = 0; 948*57193Smuller register int totcnt = 0; 949*57193Smuller 950*57193Smuller /* 951*57193Smuller * if we have reached the user specified byte count for each archive 952*57193Smuller * volume, prompt for the next volume. (The non-standrad -R flag). 953*57193Smuller * NOTE: If the wrlimit is smaller than wrcnt, we will always write 954*57193Smuller * at least one record. We always round limit UP to next blocksize. 955*57193Smuller */ 956*57193Smuller if ((wrlimit > 0) && (wrcnt > wrlimit)) { 957*57193Smuller warn(0, "User specified archive volume byte limit reached."); 958*57193Smuller if (ar_next() < 0) { 959*57193Smuller wrcnt = 0; 960*57193Smuller exit_val = 1; 961*57193Smuller return(-1); 962*57193Smuller } 963*57193Smuller wrcnt = 0; 964*57193Smuller 965*57193Smuller /* 966*57193Smuller * The new archive volume might have changed the size of the 967*57193Smuller * write blocksize. if so we figure out if we need to write 968*57193Smuller * (one or more times), or if there is now free space left in 969*57193Smuller * the buffer (it is no longer full). bufcnt has the number of 970*57193Smuller * bytes in the buffer, (the blocksize, at the point we were 971*57193Smuller * CALLED). Push has the amount of "extra" data in the buffer 972*57193Smuller * if the block size has shrunk from a volume change. 973*57193Smuller */ 974*57193Smuller bufend = buf + blksz; 975*57193Smuller if (blksz > bufcnt) 976*57193Smuller return(0); 977*57193Smuller if (blksz < bufcnt) 978*57193Smuller push = bufcnt - blksz; 979*57193Smuller } 980*57193Smuller 981*57193Smuller /* 982*57193Smuller * We have enough data to write at least one archive block 983*57193Smuller */ 984*57193Smuller for (;;) { 985*57193Smuller /* 986*57193Smuller * write a block and check if it all went out ok 987*57193Smuller */ 988*57193Smuller cnt = ar_write(buf, blksz); 989*57193Smuller if (cnt == blksz) { 990*57193Smuller /* 991*57193Smuller * the write went ok 992*57193Smuller */ 993*57193Smuller wrcnt += cnt; 994*57193Smuller totcnt += cnt; 995*57193Smuller if (push > 0) { 996*57193Smuller /* we have extra data to push to the front. 997*57193Smuller * check for more than 1 block of push, and if 998*57193Smuller * so we loop back to write again 999*57193Smuller */ 1000*57193Smuller bcopy(bufend, buf, push); 1001*57193Smuller bufpt = buf + push; 1002*57193Smuller if (push >= blksz) { 1003*57193Smuller push -= blksz; 1004*57193Smuller continue; 1005*57193Smuller } 1006*57193Smuller } else 1007*57193Smuller bufpt = buf; 1008*57193Smuller return(totcnt); 1009*57193Smuller } else if (cnt > 0) { 1010*57193Smuller /* 1011*57193Smuller * Oh drat we got a partial write! 1012*57193Smuller * if format doesnt care about alignment let it go, 1013*57193Smuller * we warned the user in ar_write().... but this means 1014*57193Smuller * the last record on this volume violates pax spec.... 1015*57193Smuller */ 1016*57193Smuller totcnt += cnt; 1017*57193Smuller wrcnt += cnt; 1018*57193Smuller bufpt = buf + cnt; 1019*57193Smuller cnt = bufcnt - cnt; 1020*57193Smuller bcopy(bufpt, buf, cnt); 1021*57193Smuller bufpt = buf + cnt; 1022*57193Smuller if (!frmt->blkalgn || ((cnt % frmt->blkalgn) == 0)) 1023*57193Smuller return(totcnt); 1024*57193Smuller break; 1025*57193Smuller } 1026*57193Smuller 1027*57193Smuller /* 1028*57193Smuller * All done, go to next archive 1029*57193Smuller */ 1030*57193Smuller wrcnt = 0; 1031*57193Smuller if (ar_next() < 0) 1032*57193Smuller break; 1033*57193Smuller 1034*57193Smuller /* 1035*57193Smuller * The new archive volume might also have changed the block 1036*57193Smuller * size. if so, figure out if we have too much or too little 1037*57193Smuller * data for using the new block size 1038*57193Smuller */ 1039*57193Smuller bufend = buf + blksz; 1040*57193Smuller if (blksz > bufcnt) 1041*57193Smuller return(0); 1042*57193Smuller if (blksz < bufcnt) 1043*57193Smuller push = bufcnt - blksz; 1044*57193Smuller } 1045*57193Smuller 1046*57193Smuller /* 1047*57193Smuller * write failed, stop pax. we must not create a bad archive! 1048*57193Smuller */ 1049*57193Smuller exit_val = 1; 1050*57193Smuller return(-1); 1051*57193Smuller } 1052