xref: /freebsd/bin/pax/tables.h (revision 2008043f386721d58158e37e0d7e50df8095942d)
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  *	@(#)tables.h	8.1 (Berkeley) 5/31/93
36  */
37 
38 /*
39  * data structures and constants used by the different databases kept by pax
40  */
41 
42 /*
43  * Hash Table Sizes MUST BE PRIME, if set too small performance suffers.
44  * Probably safe to expect 500000 inodes per tape. Assuming good key
45  * distribution (inodes) chains of under 50 long (worse case) is ok.
46  */
47 #define L_TAB_SZ	2503		/* hard link hash table size */
48 #define F_TAB_SZ	50503		/* file time hash table size */
49 #define N_TAB_SZ	541		/* interactive rename hash table */
50 #define D_TAB_SZ	317		/* unique device mapping table */
51 #define A_TAB_SZ	317		/* ftree dir access time reset table */
52 #define MAXKEYLEN	64		/* max number of chars for hash */
53 
54 /*
55  * file hard link structure (hashed by dev/ino and chained) used to find the
56  * hard links in a file system or with some archive formats (cpio)
57  */
58 typedef struct hrdlnk {
59 	char		*name;	/* name of first file seen with this ino/dev */
60 	dev_t		dev;	/* files device number */
61 	ino_t		ino;	/* files inode number */
62 	u_long		nlink;	/* expected link count */
63 	struct hrdlnk	*fow;
64 } HRDLNK;
65 
66 /*
67  * Archive write update file time table (the -u, -C flag), hashed by filename.
68  * Filenames are stored in a scratch file at seek offset into the file. The
69  * file time (mod time) and the file name length (for a quick check) are
70  * stored in a hash table node. We were forced to use a scratch file because
71  * with -u, the mtime for every node in the archive must always be available
72  * to compare against (and this data can get REALLY large with big archives).
73  * By being careful to read only when we have a good chance of a match, the
74  * performance loss is not measurable (and the size of the archive we can
75  * handle is greatly increased).
76  */
77 typedef struct ftm {
78 	int		namelen;	/* file name length */
79 	time_t		mtime;		/* files last modification time */
80 	off_t		seek;		/* location in scratch file */
81 	struct ftm	*fow;
82 } FTM;
83 
84 /*
85  * Interactive rename table (-i flag), hashed by orig filename.
86  * We assume this will not be a large table as this mapping data can only be
87  * obtained through interactive input by the user. Nobody is going to type in
88  * changes for 500000 files? We use chaining to resolve collisions.
89  */
90 
91 typedef struct namt {
92 	char		*oname;		/* old name */
93 	char		*nname;		/* new name typed in by the user */
94 	struct namt	*fow;
95 } NAMT;
96 
97 /*
98  * Unique device mapping tables. Some protocols (e.g. cpio) require that the
99  * <c_dev,c_ino> pair will uniquely identify a file in an archive unless they
100  * are links to the same file. Appending to archives can break this. For those
101  * protocols that have this requirement we map c_dev to a unique value not seen
102  * in the archive when we append. We also try to handle inode truncation with
103  * this table. (When the inode field in the archive header are too small, we
104  * remap the dev on writes to remove accidental collisions).
105  *
106  * The list is hashed by device number using chain collision resolution. Off of
107  * each DEVT are linked the various remaps for this device based on those bits
108  * in the inode which were truncated. For example if we are just remapping to
109  * avoid a device number during an update append, off the DEVT we would have
110  * only a single DLIST that has a truncation id of 0 (no inode bits were
111  * stripped for this device so far). When we spot inode truncation we create
112  * a new mapping based on the set of bits in the inode which were stripped off.
113  * so if the top four bits of the inode are stripped and they have a pattern of
114  * 0110...... (where . are those bits not truncated) we would have a mapping
115  * assigned for all inodes that has the same 0110.... pattern (with this dev
116  * number of course). This keeps the mapping sparse and should be able to store
117  * close to the limit of files which can be represented by the optimal
118  * combination of dev and inode bits, and without creating a fouled up archive.
119  * Note we also remap truncated devs in the same way (an exercise for the
120  * dedicated reader; always wanted to say that...:)
121  */
122 
123 typedef struct devt {
124 	dev_t		dev;	/* the orig device number we now have to map */
125 	struct devt	*fow;	/* new device map list */
126 	struct dlist	*list;	/* map list based on inode truncation bits */
127 } DEVT;
128 
129 typedef struct dlist {
130 	ino_t trunc_bits;	/* truncation pattern for a specific map */
131 	dev_t dev;		/* the new device id we use */
132 	struct dlist *fow;
133 } DLIST;
134 
135 /*
136  * ftree directory access time reset table. When we are done with with a
137  * subtree we reset the access and mod time of the directory when the tflag is
138  * set. Not really explicitly specified in the pax spec, but easy and fast to
139  * do (and this may have even been intended in the spec, it is not clear).
140  * table is hashed by inode with chaining.
141  */
142 
143 typedef struct atdir {
144 	char *name;	/* name of directory to reset */
145 	dev_t dev;	/* dev and inode for fast lookup */
146 	ino_t ino;
147 	time_t mtime;	/* access and mod time to reset to */
148 	time_t atime;
149 	struct atdir *fow;
150 } ATDIR;
151 
152 /*
153  * created directory time and mode storage entry. After pax is finished during
154  * extraction or copy, we must reset directory access modes and times that
155  * may have been modified after creation (they no longer have the specified
156  * times and/or modes). We must reset time in the reverse order of creation,
157  * because entries are added  from the top of the file tree to the bottom.
158  * We MUST reset times from leaf to root (it will not work the other
159  * direction).  Entries are recorded into a spool file to make reverse
160  * reading faster.
161  */
162 
163 typedef struct dirdata {
164 	int nlen;	/* length of the directory name (includes \0) */
165 	off_t npos;	/* position in file where this dir name starts */
166 	mode_t mode;	/* file mode to restore */
167 	time_t mtime;	/* mtime to set */
168 	time_t atime;	/* atime to set */
169 	int frc_mode;	/* do we force mode settings? */
170 } DIRDATA;
171