bin/pax/tables.h

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