bin/pax/tables.h

   1 /*      $OpenBSD: tables.h,v 1.8 2006/08/05 23:05:13 ray Exp $  */
   2 /*      $NetBSD: tables.h,v 1.3 1995/03/21 09:07:47 cgd Exp $   */
   3
   4 /*-
   5  * Copyright (c) 1992 Keith Muller.
   6  * Copyright (c) 1992, 1993
   7  *      The Regents of the University of California.  All rights reserved.
   8  *
   9  * This code is derived from software contributed to Berkeley by
  10  * Keith Muller of the University of California, San Diego.
  11  *
  12  * Redistribution and use in source and binary forms, with or without
  13  * modification, are permitted provided that the following conditions
  14  * are met:
  15  * 1. Redistributions of source code must retain the above copyright
  16  *    notice, this list of conditions and the following disclaimer.
  17  * 2. Redistributions in binary form must reproduce the above copyright
  18  *    notice, this list of conditions and the following disclaimer in the
  19  *    documentation and/or other materials provided with the distribution.
  20  * 3. Neither the name of the University nor the names of its contributors
  21  *    may be used to endorse or promote products derived from this software
  22  *    without specific prior written permission.
  23  *
  24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  34  * SUCH DAMAGE.
  35  *
  36  *      @(#)tables.h    8.1 (Berkeley) 5/31/93
  37  */
  38
  39 /*
  40  * data structures and constants used by the different databases kept by pax
  41  */
  42
  43 /*
  44  * Hash Table Sizes MUST BE PRIME, if set too small performance suffers.
  45  * Probably safe to expect 500000 inodes per tape. Assuming good key
  46  * distribution (inodes) chains of under 50 long (worst case) is ok.
  47  */
  48 #define L_TAB_SZ        2503            /* hard link hash table size */
  49 #define F_TAB_SZ        50503           /* file time hash table size */
  50 #define N_TAB_SZ        541             /* interactive rename hash table */
  51 #define D_TAB_SZ        317             /* unique device mapping table */
  52 #define A_TAB_SZ        317             /* ftree dir access time reset table */
  53 #define MAXKEYLEN       64              /* max number of chars for hash */
  54 #define DIRP_SIZE       64              /* initial size of created dir table */
  55
  56 /*
  57  * file hard link structure (hashed by dev/ino and chained) used to find the
  58  * hard links in a file system or with some archive formats (cpio)
  59  */
  60 typedef struct hrdlnk {
  61         char            *name;  /* name of first file seen with this ino/dev */
  62         dev_t           dev;    /* files device number */
  63         ino_t           ino;    /* files inode number */
  64         u_long          nlink;  /* expected link count */
  65         struct hrdlnk   *fow;
  66 } HRDLNK;
  67
  68 /*
  69  * Archive write update file time table (the -u, -C flag), hashed by filename.
  70  * Filenames are stored in a scratch file at seek offset into the file. The
  71  * file time (mod time) and the file name length (for a quick check) are
  72  * stored in a hash table node. We were forced to use a scratch file because
  73  * with -u, the mtime for every node in the archive must always be available
  74  * to compare against (and this data can get REALLY large with big archives).
  75  * By being careful to read only when we have a good chance of a match, the
  76  * performance loss is not measurable (and the size of the archive we can
  77  * handle is greatly increased).
  78  */
  79 typedef struct ftm {
  80         int             namelen;        /* file name length */
  81         time_t          mtime;          /* files last modification time */
  82         off_t           seek;           /* location in scratch file */
  83         struct ftm      *fow;
  84 } FTM;
  85
  86 /*
  87  * Interactive rename table (-i flag), hashed by orig filename.
  88  * We assume this will not be a large table as this mapping data can only be
  89  * obtained through interactive input by the user. Nobody is going to type in
  90  * changes for 500000 files? We use chaining to resolve collisions.
  91  */
  92
  93 typedef struct namt {
  94         char            *oname;         /* old name */
  95         char            *nname;         /* new name typed in by the user */
  96         struct namt     *fow;
  97 } NAMT;
  98
  99 /*
 100  * Unique device mapping tables. Some protocols (e.g. cpio) require that the
 101  * <c_dev,c_ino> pair will uniquely identify a file in an archive unless they
 102  * are links to the same file. Appending to archives can break this. For those
 103  * protocols that have this requirement we map c_dev to a unique value not seen
 104  * in the archive when we append. We also try to handle inode truncation with
 105  * this table. (When the inode field in the archive header are too small, we
 106  * remap the dev on writes to remove accidental collisions).
 107  *
 108  * The list is hashed by device number using chain collision resolution. Off of
 109  * each DEVT are linked the various remaps for this device based on those bits
 110  * in the inode which were truncated. For example if we are just remapping to
 111  * avoid a device number during an update append, off the DEVT we would have
 112  * only a single DLIST that has a truncation id of 0 (no inode bits were
 113  * stripped for this device so far). When we spot inode truncation we create
 114  * a new mapping based on the set of bits in the inode which were stripped off.
 115  * so if the top four bits of the inode are stripped and they have a pattern of
 116  * 0110...... (where . are those bits not truncated) we would have a mapping
 117  * assigned for all inodes that has the same 0110.... pattern (with this dev
 118  * number of course). This keeps the mapping sparse and should be able to store
 119  * close to the limit of files which can be represented by the optimal
 120  * combination of dev and inode bits, and without creating a fouled up archive.
 121  * Note we also remap truncated devs in the same way (an exercise for the
 122  * dedicated reader; always wanted to say that...:)
 123  */
 124
 125 typedef struct devt {
 126         dev_t           dev;    /* the orig device number we now have to map */
 127         struct devt     *fow;   /* new device map list */
 128         struct dlist    *list;  /* map list based on inode truncation bits */
 129 } DEVT;
 130
 131 typedef struct dlist {
 132         ino_t trunc_bits;       /* truncation pattern for a specific map */
 133         dev_t dev;              /* the new device id we use */
 134         struct dlist *fow;
 135 } DLIST;
 136
 137 /*
 138  * ftree directory access time reset table. When we are done with a
 139  * subtree we reset the access and mod time of the directory when the tflag is
 140  * set. Not really explicitly specified in the pax spec, but easy and fast to
 141  * do (and this may have even been intended in the spec, it is not clear).
 142  * table is hashed by inode with chaining.
 143  */
 144
 145 typedef struct atdir {
 146         char *name;     /* name of directory to reset */
 147         dev_t dev;      /* dev and inode for fast lookup */
 148         ino_t ino;
 149         time_t mtime;   /* access and mod time to reset to */
 150         time_t atime;
 151         struct atdir *fow;
 152 } ATDIR;
 153
 154 /*
 155  * created directory time and mode storage entry. After pax is finished during
 156  * extraction or copy, we must reset directory access modes and times that
 157  * may have been modified after creation (they no longer have the specified
 158  * times and/or modes). We must reset time in the reverse order of creation,
 159  * because entries are added  from the top of the file tree to the bottom.
 160  * We MUST reset times from leaf to root (it will not work the other
 161  * direction).
 162  */
 163
 164 typedef struct dirdata {
 165         char *name;     /* file name */
 166         time_t mtime;   /* mtime to set */
 167         time_t atime;   /* atime to set */
 168         u_int16_t mode; /* file mode to restore */
 169         u_int16_t frc_mode;     /* do we force mode settings? */
 170 } DIRDATA;