xref: /freebsd/bin/pax/tables.h (revision 952d112864d8008aa87278a30a539d888a8493cd)
1 /*-
2  * Copyright (c) 1992 Keith Muller.
3  * Copyright (c) 1992, 1993
4  *	The Regents of the University of California.  All rights reserved.
5  *
6  * This code is derived from software contributed to Berkeley by
7  * Keith Muller of the University of California, San Diego.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. All advertising materials mentioning features or use of this software
18  *    must display the following acknowledgement:
19  *	This product includes software developed by the University of
20  *	California, Berkeley and its contributors.
21  * 4. Neither the name of the University nor the names of its contributors
22  *    may be used to endorse or promote products derived from this software
23  *    without specific prior written permission.
24  *
25  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35  * SUCH DAMAGE.
36  *
37  *	@(#)tables.h	8.1 (Berkeley) 5/31/93
38  *	$Id$
39  */
40 
41 /*
42  * data structures and constants used by the different databases kept by pax
43  */
44 
45 /*
46  * Hash Table Sizes MUST BE PRIME, if set too small performance suffers.
47  * Probably safe to expect 500000 inodes per tape. Assuming good key
48  * distribution (inodes) chains of under 50 long (worse case) is ok.
49  */
50 #define L_TAB_SZ	2503		/* hard link hash table size */
51 #define F_TAB_SZ	50503		/* file time hash table size */
52 #define N_TAB_SZ	541		/* interactive rename hash table */
53 #define D_TAB_SZ	317		/* unique device mapping table */
54 #define A_TAB_SZ	317		/* ftree dir access time reset table */
55 #define MAXKEYLEN	64		/* max number of chars for hash */
56 
57 /*
58  * file hard link structure (hashed by dev/ino and chained) used to find the
59  * hard links in a file system or with some archive formats (cpio)
60  */
61 typedef struct hrdlnk {
62 	char		*name;	/* name of first file seen with this ino/dev */
63 	dev_t		dev;	/* files device number */
64 	ino_t		ino;	/* files inode number */
65 	u_long		nlink;	/* expected link count */
66 	struct hrdlnk	*fow;
67 } HRDLNK;
68 
69 /*
70  * Archive write update file time table (the -u, -C flag), hashed by filename.
71  * Filenames are stored in a scratch file at seek offset into the file. The
72  * file time (mod time) and the file name length (for a quick check) are
73  * stored in a hash table node. We were forced to use a scratch file because
74  * with -u, the mtime for every node in the archive must always be available
75  * to compare against (and this data can get REALLY large with big archives).
76  * By being careful to read only when we have a good chance of a match, the
77  * performance loss is not measurable (and the size of the archive we can
78  * handle is greatly increased).
79  */
80 typedef struct ftm {
81 	int		namelen;	/* file name length */
82 	time_t		mtime;		/* files last modification time */
83 	off_t		seek;		/* loacation in scratch file */
84 	struct ftm	*fow;
85 } FTM;
86 
87 /*
88  * Interactive rename table (-i flag), hashed by orig filename.
89  * We assume this will not be a large table as this mapping data can only be
90  * obtained through interactive input by the user. Nobody is going to type in
91  * changes for 500000 files? We use chaining to resolve collisions.
92  */
93 
94 typedef struct namt {
95 	char		*oname;		/* old name */
96 	char		*nname;		/* new name typed in by the user */
97 	struct namt	*fow;
98 } NAMT;
99 
100 /*
101  * Unique device mapping tables. Some protocols (e.g. cpio) require that the
102  * <c_dev,c_ino> pair will uniquely identify a file in an archive unless they
103  * are links to the same file. Appending to archives can break this. For those
104  * protocols that have this requirement we map c_dev to a unique value not seen
105  * in the archive when we append. We also try to handle inode truncation with
106  * this table. (When the inode field in the archive header are too small, we
107  * remap the dev on writes to remove accidental collisions).
108  *
109  * The list is hashed by device number using chain collision resolution. Off of
110  * each DEVT are linked the various remaps for this device based on those bits
111  * in the inode which were truncated. For example if we are just remapping to
112  * avoid a device number during an update append, off the DEVT we would have
113  * only a single DLIST that has a truncation id of 0 (no inode bits were
114  * stripped for this device so far). When we spot inode truncation we create
115  * a new mapping based on the set of bits in the inode which were stripped off.
116  * so if the top four bits of the inode are stripped and they have a pattern of
117  * 0110...... (where . are those bits not truncated) we would have a mapping
118  * assigned for all inodes that has the same 0110.... pattern (with this dev
119  * number of course). This keeps the mapping sparse and should be able to store
120  * close to the limit of files which can be represented by the optimal
121  * combination of dev and inode bits, and without creating a fouled up archive.
122  * Note we also remap truncated devs in the same way (an exercise for the
123  * dedicated reader; always wanted to say that...:)
124  */
125 
126 typedef struct devt {
127 	dev_t		dev;	/* the orig device number we now have to map */
128 	struct devt	*fow;	/* new device map list */
129 	struct dlist	*list;	/* map list based on inode truncation bits */
130 } DEVT;
131 
132 typedef struct dlist {
133 	ino_t trunc_bits;	/* truncation pattern for a specific map */
134 	dev_t dev;		/* the new device id we use */
135 	struct dlist *fow;
136 } DLIST;
137 
138 /*
139  * ftree directory access time reset table. When we are done with with a
140  * subtree we reset the access and mod time of the directory when the tflag is
141  * set. Not really explicitly specified in the pax spec, but easy and fast to
142  * do (and this may have even been intended in the spec, it is not clear).
143  * table is hashed by inode with chaining.
144  */
145 
146 typedef struct atdir {
147 	char *name;	/* name of directory to reset */
148 	dev_t dev;	/* dev and inode for fast lookup */
149 	ino_t ino;
150 	time_t mtime;	/* access and mod time to reset to */
151 	time_t atime;
152 	struct atdir *fow;
153 } ATDIR;
154 
155 /*
156  * created directory time and mode storage entry. After pax is finished during
157  * extraction or copy, we must reset directory access modes and times that
158  * may have been modified after creation (they no longer have the specified
159  * times and/or modes). We must reset time in the reverse order of creation,
160  * because entries are added  from the top of the file tree to the bottom.
161  * We MUST reset times from leaf to root (it will not work the other
162  * direction).  Entries are recorded into a spool file to make reverse
163  * reading faster.
164  */
165 
166 typedef struct dirdata {
167 	int nlen;	/* length of the directory name (includes \0) */
168 	off_t npos;	/* position in file where this dir name starts */
169 	mode_t mode;	/* file mode to restore */
170 	time_t mtime;	/* mtime to set */
171 	time_t atime;	/* atime to set */
172 	int frc_mode;	/* do we force mode settings? */
173 } DIRDATA;
174