path: root/genisoimage/hash.c

/*
 * This file has been modified for the cdrkit suite.
 *
 * The behaviour and appearence of the program code below can differ to a major
 * extent from the version distributed by the original author(s).
 *
 * For details, see Changelog file distributed with the cdrkit package. If you
 * received this file from another source then ask the distributing person for
 * a log of modifications.
 *
 */

/* @(#)hash.c	1.18 04/06/18 joerg */
/* @(#)hash.c	1.23 06/10/04 joerg */
/*
 * File hash.c - generate hash tables for iso9660 filesystem.
 *
 * Written by Eric Youngdale (1993).
 *
 * Copyright 1993 Yggdrasil Computing, Incorporated
 * Copyright (c) 1999,2000-2006 J. Schilling
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2, or (at your option)
 * any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
 */

/* APPLE_HYB James Pearson j.pearson@ge.ucl.ac.uk 23/2/2000 */

/*
 * From jb@danware.dk:
 *
 * Cygwin fakes inodes by hashing file info, actual collisions observed!
 * This is documented in the cygwin source, look at winsup/cygwin/path.cc
 * and search for the word 'Hash'.  On NT, cygwin ORs together the
 * high and low 32 bits of the 64 bit genuine inode, look at fhandler.cc.
 *
 * Note:	Other operating systems which support the FAT filesystem may
 *		have the same problem because FAT does not use the inode
 *		concept.  For NTFS, genuine inode numbers exist, but they are
 *		64 bits and available only through an open file handle.
 *
 * The solution is the new options -no-cache-inodes/-cache-inodes that
 * allow to disable the genisoimage inode cache.
 */

#include <mconfig.h>
#include "genisoimage.h"
#include <schily.h>

#define	NR_HASH	(16*1024)

#define	HASH_FN(DEV, INO)	((DEV + INO  + (INO >> 8) + (INO << 16)) % NR_HASH)

static struct file_hash *hash_table[NR_HASH];

void		add_hash(struct directory_entry *spnt);
struct file_hash *find_hash(dev_t dev, ino_t inode);
void		flush_hash(void);
void		add_directory_hash(dev_t dev, ino_t inode);
struct file_hash *find_directory_hash(dev_t dev, ino_t inode);
static	unsigned int	name_hash(const char *name);
void		add_file_hash(struct directory_entry *de);
struct directory_entry *find_file_hash(char *name);
static	BOOL		isoname_endsok(char *name);
int		delete_file_hash(struct directory_entry *de);
void		flush_file_hash(void);

void
add_hash(struct directory_entry *spnt)
{
	struct file_hash *s_hash;
	unsigned int    hash_number;

	if (spnt->size == 0 || spnt->starting_block == 0)
		if (spnt->size != 0 && spnt->starting_block == 0) {
			comerrno(EX_BAD,
			"Non zero-length file '%s' assigned zero extent.\n",
							spnt->name);
		};

	if (!cache_inodes)
		return;
	if (spnt->dev == UNCACHED_DEVICE &&
	    (spnt->inode == TABLE_INODE || spnt->inode == UNCACHED_INODE)) {
		return;
	}
	hash_number = HASH_FN((unsigned int) spnt->dev,
						(unsigned int) spnt->inode);

#if 0
	if (verbose > 1)
		fprintf(stderr, "%s ", spnt->name);
#endif
	s_hash = (struct file_hash *) e_malloc(sizeof (struct file_hash));
	s_hash->next = hash_table[hash_number];
	s_hash->inode = spnt->inode;
	s_hash->dev = spnt->dev;
	s_hash->nlink = 0;
	s_hash->starting_block = spnt->starting_block;
	s_hash->size = spnt->size;
#ifdef SORTING
	s_hash->de = spnt;
#endif /* SORTING */
	hash_table[hash_number] = s_hash;
}

struct file_hash *
find_hash(dev_t dev, ino_t inode)
{
	unsigned int    hash_number;
	struct file_hash *spnt;

	if (!cache_inodes)
		return (NULL);
	if (dev == UNCACHED_DEVICE &&
	    (inode == TABLE_INODE || inode == UNCACHED_INODE))
		return (NULL);

	hash_number = HASH_FN((unsigned int) dev, (unsigned int) inode);
	spnt = hash_table[hash_number];
	while (spnt) {
		if (spnt->inode == inode && spnt->dev == dev)
			return (spnt);
		spnt = spnt->next;
	};
	return (NULL);
}

/*
 * based on flush_file_hash() below - needed as we want to re-use the
 * file hash table.
 */
void
flush_hash()
{
	struct file_hash	*fh;
	struct file_hash	*fh1;
	int			i;

	for (i = 0; i < NR_HASH; i++) {
		fh = hash_table[i];
		while (fh) {
			fh1 = fh->next;
			free(fh);
			fh = fh1;
		}
		hash_table[i] = NULL;
	}
}

static struct file_hash *directory_hash_table[NR_HASH];

void
add_directory_hash(dev_t dev, ino_t inode)
{
	struct file_hash *s_hash;
	unsigned int    hash_number;

	if (!cache_inodes)
		return;
	if (dev == UNCACHED_DEVICE &&
	    (inode == TABLE_INODE || inode == UNCACHED_INODE))
		return;

	hash_number = HASH_FN((unsigned int) dev, (unsigned int) inode);

	s_hash = (struct file_hash *) e_malloc(sizeof (struct file_hash));
	s_hash->next = directory_hash_table[hash_number];
	s_hash->inode = inode;
	s_hash->dev = dev;
	s_hash->nlink = 0;
	directory_hash_table[hash_number] = s_hash;
}

struct file_hash *
find_directory_hash(dev_t dev, ino_t inode)
{
	unsigned int    hash_number;
	struct file_hash *spnt;

	if (!cache_inodes)
		return (NULL);
	if (dev == UNCACHED_DEVICE &&
	    (inode == TABLE_INODE || inode == UNCACHED_INODE))
		return (NULL);

	hash_number = HASH_FN((unsigned int) dev, (unsigned int) inode);
	spnt = directory_hash_table[hash_number];
	while (spnt) {
		if (spnt->inode == inode && spnt->dev == dev)
			return (spnt);
		spnt = spnt->next;
	};
	return (NULL);
}

struct name_hash {
	struct name_hash *next;
	struct directory_entry *de;
	int	sum;
};

#define	NR_NAME_HASH	(256*1024)

static struct name_hash *name_hash_table[NR_NAME_HASH] = {0, };

/*
 * Find the hash bucket for this name.
 */
static unsigned int
name_hash(const char *name)
{
	unsigned int	hash = 0;
	const char	*p;

	p = name;

	while (*p) {
		/*
		 * Don't hash the  iso9660 version number.
		 * This way we can detect duplicates in cases where we have
		 * directories (i.e. foo) and non-directories (i.e. foo;1).
		 */
		if (*p == ';') {
			break;
		}
		hash = (hash << 15) + (hash << 3) + (hash >> 3) + (*p++ & 0xFF);
	}
	return (hash % NR_NAME_HASH);
}

void
add_file_hash(struct directory_entry *de)
{
	struct name_hash	*new;
	int			hash;
	Uchar			*p;
	int			sum = 0;

	new = (struct name_hash *) e_malloc(sizeof (struct name_hash));
	new->de = de;
	new->next = NULL;
	for (p = (Uchar *)de->isorec.name; *p; p++) {
		if (*p == ';')
			break;
		sum += *p & 0xFF;
	}
	new->sum = sum;
	hash = name_hash(de->isorec.name);

	/* Now insert into the hash table */
	new->next = name_hash_table[hash];
	name_hash_table[hash] = new;
}

struct directory_entry *
find_file_hash(register char *name)
{
	register char			*p1;
	register char			*p2;
	register struct name_hash	*nh;
	register int			sum = 0;

	if (debug > 1)
		fprintf(stderr, "find_hash('%s')\n", name);

	for (p1 = name; *p1; p1++) {
		if (*p1 == ';')
			break;
		sum += *p1 & 0xFF;
	}

	for (nh = name_hash_table[name_hash(name)]; nh; nh = nh->next) {
		if (nh->sum != sum)
			continue;

		p1 = name;
		p2 = nh->de->isorec.name;
		if (debug > 1)
			fprintf(stderr, "Checking name '%s' isorec.name '%s'\n", p1, p2);

		/* Look for end of string, or a mismatch. */
		while (1 == 1) {
			if ((*p1 == '\0' || *p1 == ';') ||
			    (*p2 == '\0' || *p2 == ';') ||
			    (*p1 != *p2)) {
				break;
			}
			p1++;
			p2++;
		}
		if (!isoname_endsok(p1) || !isoname_endsok(p2)) {
			if (debug > 1) {
				if (!isoname_endsok(p1))
					fprintf(stderr, "'%s' does NOT END OK\n", p1);
				if (!isoname_endsok(p2))
					fprintf(stderr, "'%s' does NOT END OK\n", p2);
			}
			/*
			 * If one file does not end with a valid version number
			 * and the other name ends here, we found a miss match.
			 */
			if (*p1 == '\0' || *p2 == '\0')
				continue;

			if (*p1 == ';' && *p2 == ';') {
				p1++;
				p2++;
				continue;
			}
		}

		/*
		 * If we are at the end of both strings, then we have a match.
		 */
		if ((*p1 == '\0' || *p1 == ';') &&
		    (*p2 == '\0' || *p2 == ';')) {
			return (nh->de);
		}
	}
	return (NULL);
}

/*
 * The macro 'eo' is just an idea on how one might speed up isoname_endsok()
 */
#define	eo(p)	(((p)[0] == '\0') || \
		((p)[0] == ';' && (p)[1] == '1' && (p)[2] == '\0') || \
		isoname_endsok(p))

static BOOL
isoname_endsok(char *name)
{
	int	i;
	char	*p;

	if (*name == '\0')
		return (TRUE);
	if (*name != ';')
		return (FALSE);

	for (p = ++name, i = 0; *p && i < 5; p++, i++) {
		if (*p < '0' || *p > '9')
			return (FALSE);
	}
	i = atoi(name);
	if (i < 1 || i > 32767)
		return (FALSE);
	return (TRUE);
}

int
delete_file_hash(struct directory_entry *de)
{
	struct name_hash	*nh;
	struct name_hash	*prev;
	int			hash;

	prev = NULL;
	hash = name_hash(de->isorec.name);
	for (nh = name_hash_table[hash]; nh; nh = nh->next) {
		if (nh->de == de)
			break;
		prev = nh;
	}
	if (!nh)
		return (1);
	if (!prev)
		name_hash_table[hash] = nh->next;
	else
		prev->next = nh->next;
	free(nh);
	return (0);
}

void
flush_file_hash()
{
	struct name_hash	*nh;
	struct name_hash	*nh1;
	int			i;

	for (i = 0; i < NR_NAME_HASH; i++) {
		nh = name_hash_table[i];
		while (nh) {
			nh1 = nh->next;
			free(nh);
			nh = nh1;
		}
		name_hash_table[i] = NULL;

	}
}