usbloadergx/source/libntfs/dir.c

2898 lines
92 KiB
C
Raw Normal View History

2009-12-19 15:06:57 +01:00
/**
* dir.c - Directory handling code. Originated from the Linux-NTFS project.
*
* Copyright (c) 2002-2005 Anton Altaparmakov
* Copyright (c) 2004-2005 Richard Russon
* Copyright (c) 2004-2008 Szabolcs Szakacsits
* Copyright (c) 2005-2007 Yura Pakhuchiy
* Copyright (c) 2008-2010 Jean-Pierre Andre
2009-12-19 15:06:57 +01:00
*
* This program/include file 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 of the License, or
* (at your option) any later version.
*
* This program/include file 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 (in the main directory of the NTFS-3G
* distribution in the file COPYING); if not, write to the Free Software
* Foundation,Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*/
#ifdef HAVE_CONFIG_H
#include "config.h"
#endif
#ifdef HAVE_STDLIB_H
#include <stdlib.h>
#endif
#ifdef HAVE_ERRNO_H
#include <errno.h>
#endif
#ifdef HAVE_STRING_H
#include <string.h>
#endif
#ifdef HAVE_SYS_STAT_H
#include <sys/stat.h>
#endif
#ifdef HAVE_SYS_SYSMACROS_H
#include <sys/sysmacros.h>
#endif
#include "param.h"
2009-12-19 15:06:57 +01:00
#include "types.h"
#include "debug.h"
#include "attrib.h"
#include "inode.h"
#include "dir.h"
#include "volume.h"
#include "mft.h"
#include "index.h"
#include "ntfstime.h"
#include "lcnalloc.h"
#include "logging.h"
#include "cache.h"
2009-12-19 15:06:57 +01:00
#include "misc.h"
#include "security.h"
#include "reparse.h"
#include "object_id.h"
2009-12-19 15:06:57 +01:00
#ifdef HAVE_SETXATTR
#include <sys/xattr.h>
#endif
/*
* The little endian Unicode strings "$I30", "$SII", "$SDH", "$O"
* and "$Q" as global constants.
*/
ntfschar NTFS_INDEX_I30[5] = { const_cpu_to_le16( '$' ), const_cpu_to_le16( 'I' ),
const_cpu_to_le16( '3' ), const_cpu_to_le16( '0' ),
const_cpu_to_le16( '\0' )
};
ntfschar NTFS_INDEX_SII[5] = { const_cpu_to_le16( '$' ), const_cpu_to_le16( 'S' ),
const_cpu_to_le16( 'I' ), const_cpu_to_le16( 'I' ),
const_cpu_to_le16( '\0' )
};
ntfschar NTFS_INDEX_SDH[5] = { const_cpu_to_le16( '$' ), const_cpu_to_le16( 'S' ),
const_cpu_to_le16( 'D' ), const_cpu_to_le16( 'H' ),
const_cpu_to_le16( '\0' )
};
ntfschar NTFS_INDEX_O[3] = { const_cpu_to_le16( '$' ), const_cpu_to_le16( 'O' ),
const_cpu_to_le16( '\0' )
};
ntfschar NTFS_INDEX_Q[3] = { const_cpu_to_le16( '$' ), const_cpu_to_le16( 'Q' ),
const_cpu_to_le16( '\0' )
};
ntfschar NTFS_INDEX_R[3] = { const_cpu_to_le16( '$' ), const_cpu_to_le16( 'R' ),
const_cpu_to_le16( '\0' )
};
2009-12-19 15:06:57 +01:00
#if CACHE_INODE_SIZE
/*
* Pathname hashing
*
* Based on first char and second char (which may be '\0')
*/
int ntfs_dir_inode_hash( const struct CACHED_GENERIC *cached )
{
const char *path;
const unsigned char *name;
path = ( const char* )cached->variable;
if ( !path )
{
ntfs_log_error( "Bad inode cache entry\n" );
return ( -1 );
}
name = ( const unsigned char* )strrchr( path, '/' );
if ( !name )
name = ( const unsigned char* )path;
return ( ( ( name[0] << 1 ) + name[1] + strlen( ( const char* )name ) )
% ( 2*CACHE_INODE_SIZE ) );
}
2009-12-19 15:06:57 +01:00
/*
* Pathname comparing for entering/fetching from cache
2009-12-19 15:06:57 +01:00
*/
static int inode_cache_compare( const struct CACHED_GENERIC *cached,
const struct CACHED_GENERIC *wanted )
{
return ( !cached->variable
|| strcmp( cached->variable, wanted->variable ) );
2009-12-19 15:06:57 +01:00
}
/*
* Pathname comparing for invalidating entries in cache
2009-12-19 15:06:57 +01:00
*
* A partial path is compared in order to invalidate all paths
* related to a renamed directory
* inode numbers are also checked, as deleting a long name may
* imply deleting a short name and conversely
*
* Only use associated with a CACHE_NOHASH flag
2009-12-19 15:06:57 +01:00
*/
static int inode_cache_inv_compare( const struct CACHED_GENERIC *cached,
const struct CACHED_GENERIC *wanted )
{
int len;
BOOL different;
const struct CACHED_INODE *w;
const struct CACHED_INODE *c;
w = ( const struct CACHED_INODE* )wanted;
c = ( const struct CACHED_INODE* )cached;
if ( w->pathname )
{
len = strlen( w->pathname );
different = !cached->variable
|| ( ( w->inum != MREF( c->inum ) )
&& ( strncmp( c->pathname, w->pathname, len )
|| ( ( c->pathname[len] != '\0' )
&& ( c->pathname[len] != '/' ) ) ) );
}
else
different = !c->pathname
|| ( w->inum != MREF( c->inum ) );
return ( different );
}
#endif
#if CACHE_LOOKUP_SIZE
/*
* File name comparing for entering/fetching from lookup cache
*/
static int lookup_cache_compare( const struct CACHED_GENERIC *cached,
const struct CACHED_GENERIC *wanted )
{
const struct CACHED_LOOKUP *c = ( const struct CACHED_LOOKUP* ) cached;
const struct CACHED_LOOKUP *w = ( const struct CACHED_LOOKUP* ) wanted;
return ( !c->name
|| ( c->parent != w->parent )
|| ( c->namesize != w->namesize )
|| memcmp( c->name, w->name, c->namesize ) );
}
/*
* Inode number comparing for invalidating lookup cache
*
* All entries with designated inode number are invalidated
*
* Only use associated with a CACHE_NOHASH flag
*/
static int lookup_cache_inv_compare( const struct CACHED_GENERIC *cached,
const struct CACHED_GENERIC *wanted )
{
const struct CACHED_LOOKUP *c = ( const struct CACHED_LOOKUP* ) cached;
const struct CACHED_LOOKUP *w = ( const struct CACHED_LOOKUP* ) wanted;
return ( !c->name
|| ( c->parent != w->parent )
|| ( MREF( c->inum ) != MREF( w->inum ) ) );
}
/*
* Lookup hashing
*
* Based on first, second and and last char
*/
int ntfs_dir_lookup_hash( const struct CACHED_GENERIC *cached )
{
const unsigned char *name;
int count;
unsigned int val;
name = ( const unsigned char* )cached->variable;
count = cached->varsize;
if ( !name || !count )
{
ntfs_log_error( "Bad lookup cache entry\n" );
return ( -1 );
}
val = ( name[0] << 2 ) + ( name[1] << 1 ) + name[count - 1] + count;
return ( val % ( 2*CACHE_LOOKUP_SIZE ) );
2009-12-19 15:06:57 +01:00
}
#endif
/**
* ntfs_inode_lookup_by_name - find an inode in a directory given its name
* @dir_ni: ntfs inode of the directory in which to search for the name
* @uname: Unicode name for which to search in the directory
* @uname_len: length of the name @uname in Unicode characters
2009-12-19 15:06:57 +01:00
*
* Look for an inode with name @uname in the directory with inode @dir_ni.
* ntfs_inode_lookup_by_name() walks the contents of the directory looking for
* the Unicode name. If the name is found in the directory, the corresponding
* inode number (>= 0) is returned as a mft reference in cpu format, i.e. it
* is a 64-bit number containing the sequence number.
*
* On error, return -1 with errno set to the error code. If the inode is is not
* found errno is ENOENT.
*
* Note, @uname_len does not include the (optional) terminating NULL character.
*
* Note, we look for a case sensitive match first but we also look for a case
* insensitive match at the same time. If we find a case insensitive match, we
* save that for the case that we don't find an exact match, where we return
* the mft reference of the case insensitive match.
*
* If the volume is mounted with the case sensitive flag set, then we only
* allow exact matches.
*/
u64 ntfs_inode_lookup_by_name( ntfs_inode *dir_ni,
const ntfschar *uname, const int uname_len )
{
VCN vcn;
u64 mref = 0;
s64 br;
ntfs_volume *vol = dir_ni->vol;
ntfs_attr_search_ctx *ctx;
INDEX_ROOT *ir;
INDEX_ENTRY *ie;
INDEX_ALLOCATION *ia;
IGNORE_CASE_BOOL case_sensitivity;
u8 *index_end;
ntfs_attr *ia_na;
int eo, rc;
u32 index_block_size, index_vcn_size;
u8 index_vcn_size_bits;
ntfs_log_trace( "Entering\n" );
if ( !dir_ni || !dir_ni->mrec || !uname || uname_len <= 0 )
{
errno = EINVAL;
return -1;
}
ctx = ntfs_attr_get_search_ctx( dir_ni, NULL );
if ( !ctx )
return -1;
/* Find the index root attribute in the mft record. */
if ( ntfs_attr_lookup( AT_INDEX_ROOT, NTFS_INDEX_I30, 4, CASE_SENSITIVE, 0, NULL,
0, ctx ) )
{
ntfs_log_perror( "Index root attribute missing in directory inode "
"%lld", ( unsigned long long )dir_ni->mft_no );
goto put_err_out;
}
case_sensitivity = ( NVolCaseSensitive( vol ) ? CASE_SENSITIVE : IGNORE_CASE );
/* Get to the index root value. */
ir = ( INDEX_ROOT* )( ( u8* )ctx->attr +
le16_to_cpu( ctx->attr->value_offset ) );
index_block_size = le32_to_cpu( ir->index_block_size );
if ( index_block_size < NTFS_BLOCK_SIZE ||
index_block_size & ( index_block_size - 1 ) )
{
ntfs_log_error( "Index block size %u is invalid.\n",
( unsigned )index_block_size );
goto put_err_out;
}
index_end = ( u8* ) & ir->index + le32_to_cpu( ir->index.index_length );
/* The first index entry. */
ie = ( INDEX_ENTRY* )( ( u8* ) & ir->index +
le32_to_cpu( ir->index.entries_offset ) );
/*
* Loop until we exceed valid memory (corruption case) or until we
* reach the last entry.
*/
for ( ;; ie = ( INDEX_ENTRY* )( ( u8* )ie + le16_to_cpu( ie->length ) ) )
{
/* Bounds checks. */
if ( ( u8* )ie < ( u8* )ctx->mrec || ( u8* )ie +
sizeof( INDEX_ENTRY_HEADER ) > index_end ||
( u8* )ie + le16_to_cpu( ie->key_length ) >
index_end )
{
ntfs_log_error( "Index entry out of bounds in inode %lld"
"\n", ( unsigned long long )dir_ni->mft_no );
goto put_err_out;
}
/*
* The last entry cannot contain a name. It can however contain
* a pointer to a child node in the B+tree so we just break out.
*/
if ( ie->ie_flags & INDEX_ENTRY_END )
break;
if ( !le16_to_cpu( ie->length ) )
{
ntfs_log_error( "Zero length index entry in inode %lld"
"\n", ( unsigned long long )dir_ni->mft_no );
goto put_err_out;
}
/*
* Not a perfect match, need to do full blown collation so we
* know which way in the B+tree we have to go.
*/
rc = ntfs_names_full_collate( uname, uname_len,
( ntfschar* ) & ie->key.file_name.file_name,
ie->key.file_name.file_name_length,
case_sensitivity, vol->upcase, vol->upcase_len );
/*
* If uname collates before the name of the current entry, there
* is definitely no such name in this index but we might need to
* descend into the B+tree so we just break out of the loop.
*/
if ( rc == -1 )
break;
/* The names are not equal, continue the search. */
if ( rc )
continue;
/*
* Perfect match, this will never happen as the
* ntfs_are_names_equal() call will have gotten a match but we
* still treat it correctly.
*/
mref = le64_to_cpu( ie->indexed_file );
ntfs_attr_put_search_ctx( ctx );
return mref;
}
/*
* We have finished with this index without success. Check for the
* presence of a child node and if not present return error code
* ENOENT, unless we have got the mft reference of a matching name
* cached in mref in which case return mref.
*/
if ( !( ie->ie_flags & INDEX_ENTRY_NODE ) )
{
ntfs_attr_put_search_ctx( ctx );
if ( mref )
return mref;
ntfs_log_debug( "Entry not found.\n" );
errno = ENOENT;
return -1;
} /* Child node present, descend into it. */
/* Open the index allocation attribute. */
ia_na = ntfs_attr_open( dir_ni, AT_INDEX_ALLOCATION, NTFS_INDEX_I30, 4 );
if ( !ia_na )
{
ntfs_log_perror( "Failed to open index allocation (inode %lld)",
( unsigned long long )dir_ni->mft_no );
goto put_err_out;
}
/* Allocate a buffer for the current index block. */
ia = ntfs_malloc( index_block_size );
if ( !ia )
{
ntfs_attr_close( ia_na );
goto put_err_out;
}
/* Determine the size of a vcn in the directory index. */
if ( vol->cluster_size <= index_block_size )
{
index_vcn_size = vol->cluster_size;
index_vcn_size_bits = vol->cluster_size_bits;
}
else
{
index_vcn_size = vol->sector_size;
index_vcn_size_bits = vol->sector_size_bits;
}
/* Get the starting vcn of the index_block holding the child node. */
vcn = sle64_to_cpup( ( u8* )ie + le16_to_cpu( ie->length ) - 8 );
2009-12-19 15:06:57 +01:00
descend_into_child_node:
/* Read the index block starting at vcn. */
br = ntfs_attr_mst_pread( ia_na, vcn << index_vcn_size_bits, 1,
index_block_size, ia );
if ( br != 1 )
{
if ( br != -1 )
errno = EIO;
ntfs_log_perror( "Failed to read vcn 0x%llx",
( unsigned long long )vcn );
goto close_err_out;
}
if ( sle64_to_cpu( ia->index_block_vcn ) != vcn )
{
ntfs_log_error( "Actual VCN (0x%llx) of index buffer is different "
"from expected VCN (0x%llx).\n",
( long long )sle64_to_cpu( ia->index_block_vcn ),
( long long )vcn );
errno = EIO;
goto close_err_out;
}
if ( le32_to_cpu( ia->index.allocated_size ) + 0x18 != index_block_size )
{
ntfs_log_error( "Index buffer (VCN 0x%llx) of directory inode 0x%llx "
"has a size (%u) differing from the directory "
"specified size (%u).\n", ( long long )vcn,
( unsigned long long )dir_ni->mft_no,
( unsigned ) le32_to_cpu( ia->index.allocated_size ) + 0x18,
( unsigned )index_block_size );
errno = EIO;
goto close_err_out;
}
index_end = ( u8* ) & ia->index + le32_to_cpu( ia->index.index_length );
if ( index_end > ( u8* )ia + index_block_size )
{
ntfs_log_error( "Size of index buffer (VCN 0x%llx) of directory inode "
"0x%llx exceeds maximum size.\n",
( long long )vcn, ( unsigned long long )dir_ni->mft_no );
errno = EIO;
goto close_err_out;
}
/* The first index entry. */
ie = ( INDEX_ENTRY* )( ( u8* ) & ia->index +
le32_to_cpu( ia->index.entries_offset ) );
/*
* Iterate similar to above big loop but applied to index buffer, thus
* loop until we exceed valid memory (corruption case) or until we
* reach the last entry.
*/
for ( ;; ie = ( INDEX_ENTRY* )( ( u8* )ie + le16_to_cpu( ie->length ) ) )
{
/* Bounds check. */
if ( ( u8* )ie < ( u8* )ia || ( u8* )ie +
sizeof( INDEX_ENTRY_HEADER ) > index_end ||
( u8* )ie + le16_to_cpu( ie->key_length ) >
index_end )
{
ntfs_log_error( "Index entry out of bounds in directory "
"inode %lld.\n",
( unsigned long long )dir_ni->mft_no );
errno = EIO;
goto close_err_out;
}
/*
* The last entry cannot contain a name. It can however contain
* a pointer to a child node in the B+tree so we just break out.
*/
if ( ie->ie_flags & INDEX_ENTRY_END )
break;
if ( !le16_to_cpu( ie->length ) )
{
errno = EIO;
ntfs_log_error( "Zero length index entry in inode %lld"
"\n", ( unsigned long long )dir_ni->mft_no );
goto close_err_out;
}
/*
* Not a perfect match, need to do full blown collation so we
* know which way in the B+tree we have to go.
*/
rc = ntfs_names_full_collate( uname, uname_len,
( ntfschar* ) & ie->key.file_name.file_name,
ie->key.file_name.file_name_length,
case_sensitivity, vol->upcase, vol->upcase_len );
/*
* If uname collates before the name of the current entry, there
* is definitely no such name in this index but we might need to
* descend into the B+tree so we just break out of the loop.
*/
if ( rc == -1 )
break;
/* The names are not equal, continue the search. */
if ( rc )
continue;
mref = le64_to_cpu( ie->indexed_file );
free( ia );
ntfs_attr_close( ia_na );
ntfs_attr_put_search_ctx( ctx );
return mref;
}
/*
* We have finished with this index buffer without success. Check for
* the presence of a child node.
*/
if ( ie->ie_flags & INDEX_ENTRY_NODE )
{
if ( ( ia->index.ih_flags & NODE_MASK ) == LEAF_NODE )
{
ntfs_log_error( "Index entry with child node found in a leaf "
"node in directory inode %lld.\n",
( unsigned long long )dir_ni->mft_no );
errno = EIO;
goto close_err_out;
}
/* Child node present, descend into it. */
vcn = sle64_to_cpup( ( u8* )ie + le16_to_cpu( ie->length ) - 8 );
if ( vcn >= 0 )
goto descend_into_child_node;
ntfs_log_error( "Negative child node vcn in directory inode "
"0x%llx.\n", ( unsigned long long )dir_ni->mft_no );
errno = EIO;
goto close_err_out;
}
free( ia );
ntfs_attr_close( ia_na );
ntfs_attr_put_search_ctx( ctx );
/*
* No child node present, return error code ENOENT, unless we have got
* the mft reference of a matching name cached in mref in which case
* return mref.
*/
if ( mref )
return mref;
ntfs_log_debug( "Entry not found.\n" );
errno = ENOENT;
return -1;
2009-12-19 15:06:57 +01:00
put_err_out:
eo = EIO;
ntfs_log_debug( "Corrupt directory. Aborting lookup.\n" );
2009-12-19 15:06:57 +01:00
eo_put_err_out:
ntfs_attr_put_search_ctx( ctx );
errno = eo;
return -1;
2009-12-19 15:06:57 +01:00
close_err_out:
eo = errno;
free( ia );
ntfs_attr_close( ia_na );
goto eo_put_err_out;
2009-12-19 15:06:57 +01:00
}
/*
* Lookup a file in a directory from its UTF-8 name
*
* The name is first fetched from cache if one is defined
*
* Returns the inode number
* or -1 if not possible (errno tells why)
*/
u64 ntfs_inode_lookup_by_mbsname( ntfs_inode *dir_ni, const char *name )
{
int uname_len;
ntfschar *uname = ( ntfschar* )NULL;
u64 inum;
char *cached_name;
const char *const_name;
if ( !NVolCaseSensitive( dir_ni->vol ) )
{
cached_name = ntfs_uppercase_mbs( name,
dir_ni->vol->upcase, dir_ni->vol->upcase_len );
const_name = cached_name;
}
else
{
cached_name = ( char* )NULL;
const_name = name;
}
if ( const_name )
{
#if CACHE_LOOKUP_SIZE
/*
* fetch inode from cache
*/
if ( dir_ni->vol->lookup_cache )
{
struct CACHED_LOOKUP item;
struct CACHED_LOOKUP *cached;
item.name = const_name;
item.namesize = strlen( const_name ) + 1;
item.parent = dir_ni->mft_no;
cached = ( struct CACHED_LOOKUP* )ntfs_fetch_cache(
dir_ni->vol->lookup_cache,
GENERIC( &item ), lookup_cache_compare );
if ( cached )
{
inum = cached->inum;
if ( inum == ( u64 ) - 1 )
errno = ENOENT;
}
else
{
/* Generate unicode name. */
uname_len = ntfs_mbstoucs( name, &uname );
if ( uname_len >= 0 )
{
inum = ntfs_inode_lookup_by_name( dir_ni,
uname, uname_len );
item.inum = inum;
/* enter into cache, even if not found */
ntfs_enter_cache( dir_ni->vol->lookup_cache,
GENERIC( &item ),
lookup_cache_compare );
free( uname );
}
else
inum = ( s64 ) - 1;
}
}
else
#endif
{
/* Generate unicode name. */
uname_len = ntfs_mbstoucs( cached_name, &uname );
if ( uname_len >= 0 )
inum = ntfs_inode_lookup_by_name( dir_ni,
uname, uname_len );
else
inum = ( s64 ) - 1;
}
if ( cached_name )
free( cached_name );
}
else
inum = ( s64 ) - 1;
return ( inum );
}
/*
* Update a cache lookup record when a name has been defined
*
* The UTF-8 name is required
*/
void ntfs_inode_update_mbsname( ntfs_inode *dir_ni, const char *name, u64 inum )
{
#if CACHE_LOOKUP_SIZE
struct CACHED_LOOKUP item;
struct CACHED_LOOKUP *cached;
char *cached_name;
if ( dir_ni->vol->lookup_cache )
{
if ( !NVolCaseSensitive( dir_ni->vol ) )
{
cached_name = ntfs_uppercase_mbs( name,
dir_ni->vol->upcase, dir_ni->vol->upcase_len );
item.name = cached_name;
}
else
{
cached_name = ( char* )NULL;
item.name = name;
}
if ( item.name )
{
item.namesize = strlen( item.name ) + 1;
item.parent = dir_ni->mft_no;
item.inum = inum;
cached = ( struct CACHED_LOOKUP* )ntfs_enter_cache(
dir_ni->vol->lookup_cache,
GENERIC( &item ), lookup_cache_compare );
if ( cached )
cached->inum = inum;
if ( cached_name )
free( cached_name );
}
}
#endif
}
2009-12-19 15:06:57 +01:00
/**
* ntfs_pathname_to_inode - Find the inode which represents the given pathname
* @vol: An ntfs volume obtained from ntfs_mount
* @parent: A directory inode to begin the search (may be NULL)
* @pathname: Pathname to be located
*
* Take an ASCII pathname and find the inode that represents it. The function
* splits the path and then descends the directory tree. If @parent is NULL,
* then the root directory '.' will be used as the base for the search.
*
* Return: inode Success, the pathname was valid
* NULL Error, the pathname was invalid, or some other error occurred
2009-12-19 15:06:57 +01:00
*/
ntfs_inode *ntfs_pathname_to_inode( ntfs_volume *vol, ntfs_inode *parent,
const char *pathname )
{
u64 inum;
int len, err = 0;
char *p, *q;
ntfs_inode *ni;
ntfs_inode *result = NULL;
ntfschar *unicode = NULL;
char *ascii = NULL;
2009-12-19 15:06:57 +01:00
#if CACHE_INODE_SIZE
struct CACHED_INODE item;
struct CACHED_INODE *cached;
char *fullname;
2009-12-19 15:06:57 +01:00
#endif
if ( !vol || !pathname )
{
errno = EINVAL;
return NULL;
}
ntfs_log_trace( "path: '%s'\n", pathname );
ascii = strdup( pathname );
if ( !ascii )
{
ntfs_log_error( "Out of memory.\n" );
err = ENOMEM;
goto out;
}
p = ascii;
/* Remove leading /'s. */
while ( p && *p && *p == PATH_SEP )
p++;
2009-12-19 15:06:57 +01:00
#if CACHE_INODE_SIZE
fullname = p;
if ( p[0] && ( p[strlen( p )-1] == PATH_SEP ) )
ntfs_log_error( "Unnormalized path %s\n", ascii );
2009-12-19 15:06:57 +01:00
#endif
if ( parent )
{
ni = parent;
}
else
{
2009-12-19 15:06:57 +01:00
#if CACHE_INODE_SIZE
/*
* fetch inode for full path from cache
*/
if ( *fullname )
{
item.pathname = fullname;
item.varsize = strlen( fullname ) + 1;
cached = ( struct CACHED_INODE* )ntfs_fetch_cache(
vol->xinode_cache, GENERIC( &item ),
inode_cache_compare );
}
else
cached = ( struct CACHED_INODE* )NULL;
if ( cached )
{
/*
* return opened inode if found in cache
*/
inum = MREF( cached->inum );
ni = ntfs_inode_open( vol, inum );
if ( !ni )
{
ntfs_log_debug( "Cannot open inode %llu: %s.\n",
( unsigned long long )inum, p );
err = EIO;
}
result = ni;
goto out;
}
2009-12-19 15:06:57 +01:00
#endif
ni = ntfs_inode_open( vol, FILE_root );
if ( !ni )
{
ntfs_log_debug( "Couldn't open the inode of the root "
"directory.\n" );
err = EIO;
result = ( ntfs_inode* )NULL;
goto out;
}
}
while ( p && *p )
{
/* Find the end of the first token. */
q = strchr( p, PATH_SEP );
if ( q != NULL )
{
*q = '\0';
}
2009-12-19 15:06:57 +01:00
#if CACHE_INODE_SIZE
/*
* fetch inode for partial path from cache
*/
cached = ( struct CACHED_INODE* )NULL;
if ( !parent )
{
item.pathname = fullname;
item.varsize = strlen( fullname ) + 1;
cached = ( struct CACHED_INODE* )ntfs_fetch_cache(
vol->xinode_cache, GENERIC( &item ),
inode_cache_compare );
if ( cached )
{
inum = cached->inum;
}
}
/*
* if not in cache, translate, search, then
* insert into cache if found
*/
if ( !cached )
{
len = ntfs_mbstoucs( p, &unicode );
if ( len < 0 )
{
ntfs_log_perror( "Could not convert filename to Unicode:"
" '%s'", p );
err = errno;
goto close;
}
else if ( len > NTFS_MAX_NAME_LEN )
{
err = ENAMETOOLONG;
goto close;
}
inum = ntfs_inode_lookup_by_name( ni, unicode, len );
if ( !parent && ( inum != ( u64 ) - 1 ) )
{
item.inum = inum;
ntfs_enter_cache( vol->xinode_cache,
GENERIC( &item ),
inode_cache_compare );
}
}
2009-12-19 15:06:57 +01:00
#else
len = ntfs_mbstoucs( p, &unicode );
if ( len < 0 )
{
ntfs_log_perror( "Could not convert filename to Unicode:"
" '%s'", p );
err = errno;
goto close;
}
else if ( len > NTFS_MAX_NAME_LEN )
{
err = ENAMETOOLONG;
goto close;
}
inum = ntfs_inode_lookup_by_name( ni, unicode, len );
2009-12-19 15:06:57 +01:00
#endif
if ( inum == ( u64 ) - 1 )
{
ntfs_log_debug( "Couldn't find name '%s' in pathname "
"'%s'.\n", p, pathname );
err = ENOENT;
goto close;
}
if ( ni != parent )
if ( ntfs_inode_close( ni ) )
{
err = errno;
goto out;
}
inum = MREF( inum );
ni = ntfs_inode_open( vol, inum );
if ( !ni )
{
ntfs_log_debug( "Cannot open inode %llu: %s.\n",
( unsigned long long )inum, p );
err = EIO;
goto close;
}
free( unicode );
unicode = NULL;
if ( q ) *q++ = PATH_SEP; /* JPA */
p = q;
while ( p && *p && *p == PATH_SEP )
p++;
}
result = ni;
ni = NULL;
2009-12-19 15:06:57 +01:00
close:
if ( ni && ( ni != parent ) )
if ( ntfs_inode_close( ni ) && !err )
err = errno;
2009-12-19 15:06:57 +01:00
out:
free( ascii );
free( unicode );
if ( err )
errno = err;
return result;
2009-12-19 15:06:57 +01:00
}
/*
* The little endian Unicode string ".." for ntfs_readdir().
*/
static const ntfschar dotdot[3] = { const_cpu_to_le16( '.' ),
const_cpu_to_le16( '.' ),
const_cpu_to_le16( '\0' )
};
2009-12-19 15:06:57 +01:00
/*
* union index_union -
* More helpers for ntfs_readdir().
*/
typedef union
{
INDEX_ROOT *ir;
INDEX_ALLOCATION *ia;
} index_union __attribute__( ( __transparent_union__ ) );
2009-12-19 15:06:57 +01:00
/**
* enum INDEX_TYPE -
* More helpers for ntfs_readdir().
*/
typedef enum
{
INDEX_TYPE_ROOT, /* index root */
INDEX_TYPE_ALLOCATION, /* index allocation */
2009-12-19 15:06:57 +01:00
} INDEX_TYPE;
/**
* ntfs_filldir - ntfs specific filldir method
* @dir_ni: ntfs inode of current directory
* @pos: current position in directory
* @ivcn_bits: log(2) of index vcn size
* @index_type: specifies whether @iu is an index root or an index allocation
* @iu: index root or index block to which @ie belongs
* @ie: current index entry
* @dirent: context for filldir callback supplied by the caller
* @filldir: filldir callback supplied by the caller
2009-12-19 15:06:57 +01:00
*
* Pass information specifying the current directory entry @ie to the @filldir
* callback.
*/
static int ntfs_filldir( ntfs_inode *dir_ni, s64 *pos, u8 ivcn_bits,
const INDEX_TYPE index_type, index_union iu, INDEX_ENTRY *ie,
void *dirent, ntfs_filldir_t filldir )
{
FILE_NAME_ATTR *fn = &ie->key.file_name;
unsigned dt_type;
BOOL metadata;
ntfschar *loname;
int res;
MFT_REF mref;
ntfs_log_trace( "Entering.\n" );
/* Advance the position even if going to skip the entry. */
if ( index_type == INDEX_TYPE_ALLOCATION )
*pos = ( u8* )ie - ( u8* )iu.ia + ( sle64_to_cpu(
iu.ia->index_block_vcn ) << ivcn_bits ) +
dir_ni->vol->mft_record_size;
else /* if (index_type == INDEX_TYPE_ROOT) */
*pos = ( u8* )ie - ( u8* )iu.ir;
/* Skip root directory self reference entry. */
if ( MREF_LE( ie->indexed_file ) == FILE_root )
return 0;
if ( ie->key.file_name.file_attributes & FILE_ATTR_I30_INDEX_PRESENT )
dt_type = NTFS_DT_DIR;
else if ( fn->file_attributes & FILE_ATTR_SYSTEM )
dt_type = NTFS_DT_UNKNOWN;
else
dt_type = NTFS_DT_REG;
/* return metadata files and hidden files if requested */
mref = le64_to_cpu( ie->indexed_file );
metadata = ( MREF( mref ) != FILE_root ) && ( MREF( mref ) < FILE_first_user );
if ( ( !metadata && ( NVolShowHidFiles( dir_ni->vol )
|| !( fn->file_attributes & FILE_ATTR_HIDDEN ) ) )
|| ( NVolShowSysFiles( dir_ni->vol ) && ( NVolShowHidFiles( dir_ni->vol )
|| metadata ) ) )
{
if ( NVolCaseSensitive( dir_ni->vol ) )
{
res = filldir( dirent, fn->file_name,
fn->file_name_length,
fn->file_name_type, *pos,
mref, dt_type );
}
else
{
loname = ( ntfschar* )ntfs_malloc( 2 * fn->file_name_length );
if ( loname )
{
memcpy( loname, fn->file_name,
2*fn->file_name_length );
ntfs_name_locase( loname, fn->file_name_length,
dir_ni->vol->locase,
dir_ni->vol->upcase_len );
res = filldir( dirent, loname,
fn->file_name_length,
fn->file_name_type, *pos,
mref, dt_type );
free( loname );
}
else
res = -1;
}
}
else
res = 0;
return ( res );
2009-12-19 15:06:57 +01:00
}
/**
* ntfs_mft_get_parent_ref - find mft reference of parent directory of an inode
* @ni: ntfs inode whose parent directory to find
2009-12-19 15:06:57 +01:00
*
* Find the parent directory of the ntfs inode @ni. To do this, find the first
* file name attribute in the mft record of @ni and return the parent mft
* reference from that.
*
* Note this only makes sense for directories, since files can be hard linked
* from multiple directories and there is no way for us to tell which one is
* being looked for.
*
* Technically directories can have hard links, too, but we consider that as
* illegal as Linux/UNIX do not support directory hard links.
*
* Return the mft reference of the parent directory on success or -1 on error
* with errno set to the error code.
*/
static MFT_REF ntfs_mft_get_parent_ref( ntfs_inode *ni )
{
MFT_REF mref;
ntfs_attr_search_ctx *ctx;
FILE_NAME_ATTR *fn;
int eo;
ntfs_log_trace( "Entering.\n" );
if ( !ni )
{
errno = EINVAL;
return ERR_MREF( -1 );
}
ctx = ntfs_attr_get_search_ctx( ni, NULL );
if ( !ctx )
return ERR_MREF( -1 );
if ( ntfs_attr_lookup( AT_FILE_NAME, AT_UNNAMED, 0, 0, 0, NULL, 0, ctx ) )
{
ntfs_log_error( "No file name found in inode %lld\n",
( unsigned long long )ni->mft_no );
goto err_out;
}
if ( ctx->attr->non_resident )
{
ntfs_log_error( "File name attribute must be resident (inode "
"%lld)\n", ( unsigned long long )ni->mft_no );
goto io_err_out;
}
fn = ( FILE_NAME_ATTR* )( ( u8* )ctx->attr +
le16_to_cpu( ctx->attr->value_offset ) );
if ( ( u8* )fn + le32_to_cpu( ctx->attr->value_length ) >
( u8* )ctx->attr + le32_to_cpu( ctx->attr->length ) )
{
ntfs_log_error( "Corrupt file name attribute in inode %lld.\n",
( unsigned long long )ni->mft_no );
goto io_err_out;
}
mref = le64_to_cpu( fn->parent_directory );
ntfs_attr_put_search_ctx( ctx );
return mref;
2009-12-19 15:06:57 +01:00
io_err_out:
errno = EIO;
2009-12-19 15:06:57 +01:00
err_out:
eo = errno;
ntfs_attr_put_search_ctx( ctx );
errno = eo;
return ERR_MREF( -1 );
2009-12-19 15:06:57 +01:00
}
/**
* ntfs_readdir - read the contents of an ntfs directory
* @dir_ni: ntfs inode of current directory
* @pos: current position in directory
* @dirent: context for filldir callback supplied by the caller
* @filldir: filldir callback supplied by the caller
2009-12-19 15:06:57 +01:00
*
* Parse the index root and the index blocks that are marked in use in the
* index bitmap and hand each found directory entry to the @filldir callback
* supplied by the caller.
*
* Return 0 on success or -1 on error with errno set to the error code.
*
* Note: Index blocks are parsed in ascending vcn order, from which follows
* that the directory entries are not returned sorted.
*/
int ntfs_readdir( ntfs_inode *dir_ni, s64 *pos,
void *dirent, ntfs_filldir_t filldir )
{
s64 i_size, br, ia_pos, bmp_pos, ia_start;
ntfs_volume *vol;
ntfs_attr *ia_na, *bmp_na = NULL;
ntfs_attr_search_ctx *ctx = NULL;
u8 *index_end, *bmp = NULL;
INDEX_ROOT *ir;
INDEX_ENTRY *ie;
INDEX_ALLOCATION *ia = NULL;
int rc, ir_pos, bmp_buf_size, bmp_buf_pos, eo;
u32 index_block_size, index_vcn_size;
u8 index_block_size_bits, index_vcn_size_bits;
ntfs_log_trace( "Entering.\n" );
if ( !dir_ni || !pos || !filldir )
{
errno = EINVAL;
return -1;
}
if ( !( dir_ni->mrec->flags & MFT_RECORD_IS_DIRECTORY ) )
{
errno = ENOTDIR;
return -1;
}
vol = dir_ni->vol;
ntfs_log_trace( "Entering for inode %lld, *pos 0x%llx.\n",
( unsigned long long )dir_ni->mft_no, ( long long )*pos );
/* Open the index allocation attribute. */
ia_na = ntfs_attr_open( dir_ni, AT_INDEX_ALLOCATION, NTFS_INDEX_I30, 4 );
if ( !ia_na )
{
if ( errno != ENOENT )
{
ntfs_log_perror( "Failed to open index allocation attribute. "
"Directory inode %lld is corrupt or bug",
( unsigned long long )dir_ni->mft_no );
return -1;
}
i_size = 0;
}
else
i_size = ia_na->data_size;
rc = 0;
/* Are we at end of dir yet? */
if ( *pos >= i_size + vol->mft_record_size )
goto done;
/* Emulate . and .. for all directories. */
if ( !*pos )
{
rc = filldir( dirent, dotdot, 1, FILE_NAME_POSIX, *pos,
MK_MREF( dir_ni->mft_no,
le16_to_cpu( dir_ni->mrec->sequence_number ) ),
NTFS_DT_DIR );
if ( rc )
goto err_out;
++*pos;
}
if ( *pos == 1 )
{
MFT_REF parent_mref;
parent_mref = ntfs_mft_get_parent_ref( dir_ni );
if ( parent_mref == ERR_MREF( -1 ) )
{
ntfs_log_perror( "Parent directory not found" );
goto dir_err_out;
}
rc = filldir( dirent, dotdot, 2, FILE_NAME_POSIX, *pos,
parent_mref, NTFS_DT_DIR );
if ( rc )
goto err_out;
++*pos;
}
ctx = ntfs_attr_get_search_ctx( dir_ni, NULL );
if ( !ctx )
goto err_out;
/* Get the offset into the index root attribute. */
ir_pos = ( int ) * pos;
/* Find the index root attribute in the mft record. */
if ( ntfs_attr_lookup( AT_INDEX_ROOT, NTFS_INDEX_I30, 4, CASE_SENSITIVE, 0, NULL,
0, ctx ) )
{
ntfs_log_perror( "Index root attribute missing in directory inode "
"%lld", ( unsigned long long )dir_ni->mft_no );
goto dir_err_out;
}
/* Get to the index root value. */
ir = ( INDEX_ROOT* )( ( u8* )ctx->attr +
le16_to_cpu( ctx->attr->value_offset ) );
/* Determine the size of a vcn in the directory index. */
index_block_size = le32_to_cpu( ir->index_block_size );
if ( index_block_size < NTFS_BLOCK_SIZE ||
index_block_size & ( index_block_size - 1 ) )
{
ntfs_log_error( "Index block size %u is invalid.\n",
( unsigned )index_block_size );
goto dir_err_out;
}
index_block_size_bits = ffs( index_block_size ) - 1;
if ( vol->cluster_size <= index_block_size )
{
index_vcn_size = vol->cluster_size;
index_vcn_size_bits = vol->cluster_size_bits;
}
else
{
index_vcn_size = vol->sector_size;
index_vcn_size_bits = vol->sector_size_bits;
}
/* Are we jumping straight into the index allocation attribute? */
if ( *pos >= vol->mft_record_size )
{
ntfs_attr_put_search_ctx( ctx );
ctx = NULL;
goto skip_index_root;
}
index_end = ( u8* ) & ir->index + le32_to_cpu( ir->index.index_length );
/* The first index entry. */
ie = ( INDEX_ENTRY* )( ( u8* ) & ir->index +
le32_to_cpu( ir->index.entries_offset ) );
/*
* Loop until we exceed valid memory (corruption case) or until we
* reach the last entry or until filldir tells us it has had enough
* or signals an error (both covered by the rc test).
*/
for ( ;; ie = ( INDEX_ENTRY* )( ( u8* )ie + le16_to_cpu( ie->length ) ) )
{
ntfs_log_debug( "In index root, offset %d.\n", ( int )( ( u8* )ie - ( u8* )ir ) );
/* Bounds checks. */
if ( ( u8* )ie < ( u8* )ctx->mrec || ( u8* )ie +
sizeof( INDEX_ENTRY_HEADER ) > index_end ||
( u8* )ie + le16_to_cpu( ie->key_length ) >
index_end )
goto dir_err_out;
/* The last entry cannot contain a name. */
if ( ie->ie_flags & INDEX_ENTRY_END )
break;
if ( !le16_to_cpu( ie->length ) )
goto dir_err_out;
/* Skip index root entry if continuing previous readdir. */
if ( ir_pos > ( u8* )ie - ( u8* )ir )
continue;
/*
* Submit the directory entry to ntfs_filldir(), which will
* invoke the filldir() callback as appropriate.
*/
rc = ntfs_filldir( dir_ni, pos, index_vcn_size_bits,
INDEX_TYPE_ROOT, ir, ie, dirent, filldir );
if ( rc )
{
ntfs_attr_put_search_ctx( ctx );
ctx = NULL;
goto err_out;
}
}
ntfs_attr_put_search_ctx( ctx );
ctx = NULL;
/* If there is no index allocation attribute we are finished. */
if ( !ia_na )
goto EOD;
/* Advance *pos to the beginning of the index allocation. */
*pos = vol->mft_record_size;
2009-12-19 15:06:57 +01:00
skip_index_root:
if ( !ia_na )
goto done;
/* Allocate a buffer for the current index block. */
ia = ntfs_malloc( index_block_size );
if ( !ia )
goto err_out;
bmp_na = ntfs_attr_open( dir_ni, AT_BITMAP, NTFS_INDEX_I30, 4 );
if ( !bmp_na )
{
ntfs_log_perror( "Failed to open index bitmap attribute" );
goto dir_err_out;
}
/* Get the offset into the index allocation attribute. */
ia_pos = *pos - vol->mft_record_size;
bmp_pos = ia_pos >> index_block_size_bits;
if ( bmp_pos >> 3 >= bmp_na->data_size )
{
ntfs_log_error( "Current index position exceeds index bitmap "
"size.\n" );
goto dir_err_out;
}
bmp_buf_size = min( bmp_na->data_size - ( bmp_pos >> 3 ), 4096 );
bmp = ntfs_malloc( bmp_buf_size );
if ( !bmp )
goto err_out;
br = ntfs_attr_pread( bmp_na, bmp_pos >> 3, bmp_buf_size, bmp );
if ( br != bmp_buf_size )
{
if ( br != -1 )
errno = EIO;
ntfs_log_perror( "Failed to read from index bitmap attribute" );
goto err_out;
}
bmp_buf_pos = 0;
/* If the index block is not in use find the next one that is. */
while ( !( bmp[bmp_buf_pos >> 3] & ( 1 << ( bmp_buf_pos & 7 ) ) ) )
{
2009-12-19 15:06:57 +01:00
find_next_index_buffer:
bmp_pos++;
bmp_buf_pos++;
/* If we have reached the end of the bitmap, we are done. */
if ( bmp_pos >> 3 >= bmp_na->data_size )
goto EOD;
ia_pos = bmp_pos << index_block_size_bits;
if ( bmp_buf_pos >> 3 < bmp_buf_size )
continue;
/* Read next chunk from the index bitmap. */
bmp_buf_pos = 0;
if ( ( bmp_pos >> 3 ) + bmp_buf_size > bmp_na->data_size )
bmp_buf_size = bmp_na->data_size - ( bmp_pos >> 3 );
br = ntfs_attr_pread( bmp_na, bmp_pos >> 3, bmp_buf_size, bmp );
if ( br != bmp_buf_size )
{
if ( br != -1 )
errno = EIO;
ntfs_log_perror( "Failed to read from index bitmap attribute" );
goto err_out;
}
}
ntfs_log_debug( "Handling index block 0x%llx.\n", ( long long )bmp_pos );
/* Read the index block starting at bmp_pos. */
br = ntfs_attr_mst_pread( ia_na, bmp_pos << index_block_size_bits, 1,
index_block_size, ia );
if ( br != 1 )
{
if ( br != -1 )
errno = EIO;
ntfs_log_perror( "Failed to read index block" );
goto err_out;
}
ia_start = ia_pos & ~( s64 )( index_block_size - 1 );
if ( sle64_to_cpu( ia->index_block_vcn ) != ia_start >>
index_vcn_size_bits )
{
ntfs_log_error( "Actual VCN (0x%llx) of index buffer is different "
"from expected VCN (0x%llx) in inode 0x%llx.\n",
( long long )sle64_to_cpu( ia->index_block_vcn ),
( long long )ia_start >> index_vcn_size_bits,
( unsigned long long )dir_ni->mft_no );
goto dir_err_out;
}
if ( le32_to_cpu( ia->index.allocated_size ) + 0x18 != index_block_size )
{
ntfs_log_error( "Index buffer (VCN 0x%llx) of directory inode %lld "
"has a size (%u) differing from the directory "
"specified size (%u).\n", ( long long )ia_start >>
index_vcn_size_bits,
( unsigned long long )dir_ni->mft_no,
( unsigned ) le32_to_cpu( ia->index.allocated_size )
+ 0x18, ( unsigned )index_block_size );
goto dir_err_out;
}
index_end = ( u8* ) & ia->index + le32_to_cpu( ia->index.index_length );
if ( index_end > ( u8* )ia + index_block_size )
{
ntfs_log_error( "Size of index buffer (VCN 0x%llx) of directory inode "
"%lld exceeds maximum size.\n",
( long long )ia_start >> index_vcn_size_bits,
( unsigned long long )dir_ni->mft_no );
goto dir_err_out;
}
/* The first index entry. */
ie = ( INDEX_ENTRY* )( ( u8* ) & ia->index +
le32_to_cpu( ia->index.entries_offset ) );
/*
* Loop until we exceed valid memory (corruption case) or until we
* reach the last entry or until ntfs_filldir tells us it has had
* enough or signals an error (both covered by the rc test).
*/
for ( ;; ie = ( INDEX_ENTRY* )( ( u8* )ie + le16_to_cpu( ie->length ) ) )
{
ntfs_log_debug( "In index allocation, offset 0x%llx.\n",
( long long )ia_start + ( ( u8* )ie - ( u8* )ia ) );
/* Bounds checks. */
if ( ( u8* )ie < ( u8* )ia || ( u8* )ie +
sizeof( INDEX_ENTRY_HEADER ) > index_end ||
( u8* )ie + le16_to_cpu( ie->key_length ) >
index_end )
{
ntfs_log_error( "Index entry out of bounds in directory inode "
"%lld.\n", ( unsigned long long )dir_ni->mft_no );
goto dir_err_out;
}
/* The last entry cannot contain a name. */
if ( ie->ie_flags & INDEX_ENTRY_END )
break;
if ( !le16_to_cpu( ie->length ) )
goto dir_err_out;
/* Skip index entry if continuing previous readdir. */
if ( ia_pos - ia_start > ( u8* )ie - ( u8* )ia )
continue;
/*
* Submit the directory entry to ntfs_filldir(), which will
* invoke the filldir() callback as appropriate.
*/
rc = ntfs_filldir( dir_ni, pos, index_vcn_size_bits,
INDEX_TYPE_ALLOCATION, ia, ie, dirent, filldir );
if ( rc )
goto err_out;
}
goto find_next_index_buffer;
2009-12-19 15:06:57 +01:00
EOD:
/* We are finished, set *pos to EOD. */
*pos = i_size + vol->mft_record_size;
2009-12-19 15:06:57 +01:00
done:
free( ia );
free( bmp );
if ( bmp_na )
ntfs_attr_close( bmp_na );
if ( ia_na )
ntfs_attr_close( ia_na );
ntfs_log_debug( "EOD, *pos 0x%llx, returning 0.\n", ( long long )*pos );
return 0;
2009-12-19 15:06:57 +01:00
dir_err_out:
errno = EIO;
2009-12-19 15:06:57 +01:00
err_out:
eo = errno;
ntfs_log_trace( "failed.\n" );
if ( ctx )
ntfs_attr_put_search_ctx( ctx );
free( ia );
free( bmp );
if ( bmp_na )
ntfs_attr_close( bmp_na );
if ( ia_na )
ntfs_attr_close( ia_na );
errno = eo;
return -1;
2009-12-19 15:06:57 +01:00
}
/**
* __ntfs_create - create object on ntfs volume
* @dir_ni: ntfs inode for directory in which create new object
* @securid: id of inheritable security descriptor, 0 if none
* @name: unicode name of new object
* @name_len: length of the name in unicode characters
* @type: type of the object to create
* @dev: major and minor device numbers (obtained from makedev())
* @target: target in unicode (only for symlinks)
* @target_len: length of target in unicode characters
2009-12-19 15:06:57 +01:00
*
* Internal, use ntfs_create{,_device,_symlink} wrappers instead.
*
* @type can be:
* S_IFREG to create regular file
* S_IFDIR to create directory
* S_IFBLK to create block device
* S_IFCHR to create character device
* S_IFLNK to create symbolic link
* S_IFIFO to create FIFO
* S_IFSOCK to create socket
2009-12-19 15:06:57 +01:00
* other values are invalid.
*
* @dev is used only if @type is S_IFBLK or S_IFCHR, in other cases its value
* ignored.
*
* @target and @target_len are used only if @type is S_IFLNK, in other cases
* their value ignored.
*
* Return opened ntfs inode that describes created object on success or NULL
* on error with errno set to the error code.
*/
static ntfs_inode *__ntfs_create( ntfs_inode *dir_ni, le32 securid,
ntfschar *name, u8 name_len, mode_t type, dev_t dev,
ntfschar *target, int target_len )
{
ntfs_inode *ni;
int rollback_data = 0, rollback_sd = 0;
FILE_NAME_ATTR *fn = NULL;
STANDARD_INFORMATION *si = NULL;
int err, fn_len, si_len;
ntfs_log_trace( "Entering.\n" );
/* Sanity checks. */
if ( !dir_ni || !name || !name_len )
{
ntfs_log_error( "Invalid arguments.\n" );
errno = EINVAL;
return NULL;
}
if ( dir_ni->flags & FILE_ATTR_REPARSE_POINT )
{
errno = EOPNOTSUPP;
return NULL;
}
ni = ntfs_mft_record_alloc( dir_ni->vol, NULL );
if ( !ni )
return NULL;
#if CACHE_NIDATA_SIZE
ntfs_inode_invalidate( dir_ni->vol, ni->mft_no );
#endif
/*
* Create STANDARD_INFORMATION attribute.
* JPA Depending on available inherited security descriptor,
* Write STANDARD_INFORMATION v1.2 (no inheritance) or v3
*/
if ( securid )
si_len = sizeof( STANDARD_INFORMATION );
else
si_len = offsetof( STANDARD_INFORMATION, v1_end );
si = ntfs_calloc( si_len );
if ( !si )
{
err = errno;
goto err_out;
}
si->creation_time = ni->creation_time;
si->last_data_change_time = ni->last_data_change_time;
si->last_mft_change_time = ni->last_mft_change_time;
si->last_access_time = ni->last_access_time;
if ( securid )
{
set_nino_flag( ni, v3_Extensions );
ni->owner_id = si->owner_id = 0;
ni->security_id = si->security_id = securid;
ni->quota_charged = si->quota_charged = const_cpu_to_le64( 0 );
ni->usn = si->usn = const_cpu_to_le64( 0 );
}
else
clear_nino_flag( ni, v3_Extensions );
if ( !S_ISREG( type ) && !S_ISDIR( type ) )
{
si->file_attributes = FILE_ATTR_SYSTEM;
ni->flags = FILE_ATTR_SYSTEM;
}
ni->flags |= FILE_ATTR_ARCHIVE;
if ( NVolHideDotFiles( dir_ni->vol )
&& ( name_len > 1 )
&& ( name[0] == const_cpu_to_le16( '.' ) )
&& ( name[1] != const_cpu_to_le16( '.' ) ) )
ni->flags |= FILE_ATTR_HIDDEN;
/*
* Set compression flag according to parent directory
* unless NTFS version < 3.0 or cluster size > 4K
* or compression has been disabled
*/
if ( ( dir_ni->flags & FILE_ATTR_COMPRESSED )
&& ( dir_ni->vol->major_ver >= 3 )
&& NVolCompression( dir_ni->vol )
&& ( dir_ni->vol->cluster_size <= MAX_COMPRESSION_CLUSTER_SIZE )
&& ( S_ISREG( type ) || S_ISDIR( type ) ) )
ni->flags |= FILE_ATTR_COMPRESSED;
/* Add STANDARD_INFORMATION to inode. */
if ( ntfs_attr_add( ni, AT_STANDARD_INFORMATION, AT_UNNAMED, 0,
( u8* )si, si_len ) )
{
err = errno;
ntfs_log_error( "Failed to add STANDARD_INFORMATION "
"attribute.\n" );
goto err_out;
}
if ( !securid )
{
if ( ntfs_sd_add_everyone( ni ) )
{
err = errno;
goto err_out;
}
}
rollback_sd = 1;
if ( S_ISDIR( type ) )
{
INDEX_ROOT *ir = NULL;
INDEX_ENTRY *ie;
int ir_len, index_len;
/* Create INDEX_ROOT attribute. */
index_len = sizeof( INDEX_HEADER ) + sizeof( INDEX_ENTRY_HEADER );
ir_len = offsetof( INDEX_ROOT, index ) + index_len;
ir = ntfs_calloc( ir_len );
if ( !ir )
{
err = errno;
goto err_out;
}
ir->type = AT_FILE_NAME;
ir->collation_rule = COLLATION_FILE_NAME;
ir->index_block_size = cpu_to_le32( ni->vol->indx_record_size );
if ( ni->vol->cluster_size <= ni->vol->indx_record_size )
ir->clusters_per_index_block =
ni->vol->indx_record_size >>
ni->vol->cluster_size_bits;
else
ir->clusters_per_index_block =
ni->vol->indx_record_size >>
ni->vol->sector_size_bits;
ir->index.entries_offset = cpu_to_le32( sizeof( INDEX_HEADER ) );
ir->index.index_length = cpu_to_le32( index_len );
ir->index.allocated_size = cpu_to_le32( index_len );
ie = ( INDEX_ENTRY* )( ( u8* )ir + sizeof( INDEX_ROOT ) );
ie->length = cpu_to_le16( sizeof( INDEX_ENTRY_HEADER ) );
ie->key_length = 0;
ie->ie_flags = INDEX_ENTRY_END;
/* Add INDEX_ROOT attribute to inode. */
if ( ntfs_attr_add( ni, AT_INDEX_ROOT, NTFS_INDEX_I30, 4,
( u8* )ir, ir_len ) )
{
err = errno;
free( ir );
ntfs_log_error( "Failed to add INDEX_ROOT attribute.\n" );
goto err_out;
}
free( ir );
}
else
{
INTX_FILE *data;
int data_len;
switch ( type )
{
case S_IFBLK:
case S_IFCHR:
data_len = offsetof( INTX_FILE, device_end );
data = ntfs_malloc( data_len );
if ( !data )
{
err = errno;
goto err_out;
}
data->major = cpu_to_le64( major( dev ) );
data->minor = cpu_to_le64( minor( dev ) );
if ( type == S_IFBLK )
data->magic = INTX_BLOCK_DEVICE;
if ( type == S_IFCHR )
data->magic = INTX_CHARACTER_DEVICE;
break;
case S_IFLNK:
data_len = sizeof( INTX_FILE_TYPES ) +
target_len * sizeof( ntfschar );
data = ntfs_malloc( data_len );
if ( !data )
{
err = errno;
goto err_out;
}
data->magic = INTX_SYMBOLIC_LINK;
memcpy( data->target, target,
target_len * sizeof( ntfschar ) );
break;
case S_IFSOCK:
data = NULL;
data_len = 1;
break;
default: /* FIFO or regular file. */
data = NULL;
data_len = 0;
break;
}
/* Add DATA attribute to inode. */
if ( ntfs_attr_add( ni, AT_DATA, AT_UNNAMED, 0, ( u8* )data,
data_len ) )
{
err = errno;
ntfs_log_error( "Failed to add DATA attribute.\n" );
free( data );
goto err_out;
}
rollback_data = 1;
free( data );
}
/* Create FILE_NAME attribute. */
fn_len = sizeof( FILE_NAME_ATTR ) + name_len * sizeof( ntfschar );
fn = ntfs_calloc( fn_len );
if ( !fn )
{
err = errno;
goto err_out;
}
fn->parent_directory = MK_LE_MREF( dir_ni->mft_no,
le16_to_cpu( dir_ni->mrec->sequence_number ) );
fn->file_name_length = name_len;
fn->file_name_type = FILE_NAME_POSIX;
if ( S_ISDIR( type ) )
fn->file_attributes = FILE_ATTR_I30_INDEX_PRESENT;
if ( !S_ISREG( type ) && !S_ISDIR( type ) )
fn->file_attributes = FILE_ATTR_SYSTEM;
else
fn->file_attributes |= ni->flags & FILE_ATTR_COMPRESSED;
fn->file_attributes |= FILE_ATTR_ARCHIVE;
fn->file_attributes |= ni->flags & FILE_ATTR_HIDDEN;
fn->creation_time = ni->creation_time;
fn->last_data_change_time = ni->last_data_change_time;
fn->last_mft_change_time = ni->last_mft_change_time;
fn->last_access_time = ni->last_access_time;
if ( ni->mrec->flags & MFT_RECORD_IS_DIRECTORY )
fn->data_size = fn->allocated_size = const_cpu_to_le64( 0 );
else
{
fn->data_size = cpu_to_sle64( ni->data_size );
fn->allocated_size = cpu_to_sle64( ni->allocated_size );
}
memcpy( fn->file_name, name, name_len * sizeof( ntfschar ) );
/* Add FILE_NAME attribute to inode. */
if ( ntfs_attr_add( ni, AT_FILE_NAME, AT_UNNAMED, 0, ( u8* )fn, fn_len ) )
{
err = errno;
ntfs_log_error( "Failed to add FILE_NAME attribute.\n" );
goto err_out;
}
/* Add FILE_NAME attribute to index. */
if ( ntfs_index_add_filename( dir_ni, fn, MK_MREF( ni->mft_no,
le16_to_cpu( ni->mrec->sequence_number ) ) ) )
{
err = errno;
ntfs_log_perror( "Failed to add entry to the index" );
goto err_out;
}
/* Set hard links count and directory flag. */
ni->mrec->link_count = cpu_to_le16( 1 );
if ( S_ISDIR( type ) )
ni->mrec->flags |= MFT_RECORD_IS_DIRECTORY;
ntfs_inode_mark_dirty( ni );
/* Done! */
free( fn );
free( si );
ntfs_log_trace( "Done.\n" );
return ni;
2009-12-19 15:06:57 +01:00
err_out:
ntfs_log_trace( "Failed.\n" );
if ( rollback_sd )
ntfs_attr_remove( ni, AT_SECURITY_DESCRIPTOR, AT_UNNAMED, 0 );
if ( rollback_data )
ntfs_attr_remove( ni, AT_DATA, AT_UNNAMED, 0 );
/*
* Free extent MFT records (should not exist any with current
* ntfs_create implementation, but for any case if something will be
* changed in the future).
*/
while ( ni->nr_extents )
if ( ntfs_mft_record_free( ni->vol, *( ni->extent_nis ) ) )
{
err = errno;
ntfs_log_error( "Failed to free extent MFT record. "
"Leaving inconsistent metadata.\n" );
}
if ( ntfs_mft_record_free( ni->vol, ni ) )
ntfs_log_error( "Failed to free MFT record. "
"Leaving inconsistent metadata. Run chkdsk.\n" );
free( fn );
free( si );
errno = err;
return NULL;
2009-12-19 15:06:57 +01:00
}
/**
* Some wrappers around __ntfs_create() ...
*/
ntfs_inode *ntfs_create( ntfs_inode *dir_ni, le32 securid, ntfschar *name,
u8 name_len, mode_t type )
{
if ( type != S_IFREG && type != S_IFDIR && type != S_IFIFO &&
type != S_IFSOCK )
{
ntfs_log_error( "Invalid arguments.\n" );
return NULL;
}
return __ntfs_create( dir_ni, securid, name, name_len, type, 0, NULL, 0 );
2009-12-19 15:06:57 +01:00
}
ntfs_inode *ntfs_create_device( ntfs_inode *dir_ni, le32 securid,
ntfschar *name, u8 name_len, mode_t type, dev_t dev )
{
if ( type != S_IFCHR && type != S_IFBLK )
{
ntfs_log_error( "Invalid arguments.\n" );
return NULL;
}
return __ntfs_create( dir_ni, securid, name, name_len, type, dev, NULL, 0 );
2009-12-19 15:06:57 +01:00
}
ntfs_inode *ntfs_create_symlink( ntfs_inode *dir_ni, le32 securid,
ntfschar *name, u8 name_len, ntfschar *target, int target_len )
{
if ( !target || !target_len )
{
ntfs_log_error( "%s: Invalid argument (%p, %d)\n", __FUNCTION__,
target, target_len );
return NULL;
}
return __ntfs_create( dir_ni, securid, name, name_len, S_IFLNK, 0,
target, target_len );
2009-12-19 15:06:57 +01:00
}
int ntfs_check_empty_dir( ntfs_inode *ni )
{
ntfs_attr *na;
int ret = 0;
if ( !( ni->mrec->flags & MFT_RECORD_IS_DIRECTORY ) )
return 0;
na = ntfs_attr_open( ni, AT_INDEX_ROOT, NTFS_INDEX_I30, 4 );
if ( !na )
{
errno = EIO;
ntfs_log_perror( "Failed to open directory" );
return -1;
}
/* Non-empty directory? */
if ( ( na->data_size != sizeof( INDEX_ROOT ) + sizeof( INDEX_ENTRY_HEADER ) ) )
{
/* Both ENOTEMPTY and EEXIST are ok. We use the more common. */
errno = ENOTEMPTY;
ntfs_log_debug( "Directory is not empty\n" );
ret = -1;
}
ntfs_attr_close( na );
return ret;
2009-12-19 15:06:57 +01:00
}
static int ntfs_check_unlinkable_dir( ntfs_inode *ni, FILE_NAME_ATTR *fn )
{
int link_count = le16_to_cpu( ni->mrec->link_count );
int ret;
ret = ntfs_check_empty_dir( ni );
if ( !ret || errno != ENOTEMPTY )
return ret;
/*
* Directory is non-empty, so we can unlink only if there is more than
* one "real" hard link, i.e. links aren't different DOS and WIN32 names
*/
if ( ( link_count == 1 ) ||
( link_count == 2 && fn->file_name_type == FILE_NAME_DOS ) )
{
errno = ENOTEMPTY;
ntfs_log_debug( "Non-empty directory without hard links\n" );
goto no_hardlink;
}
ret = 0;
no_hardlink:
return ret;
2009-12-19 15:06:57 +01:00
}
/**
* ntfs_delete - delete file or directory from ntfs volume
* @ni: ntfs inode for object to delte
* @dir_ni: ntfs inode for directory in which delete object
* @name: unicode name of the object to delete
* @name_len: length of the name in unicode characters
2009-12-19 15:06:57 +01:00
*
* @ni is always closed after the call to this function (even if it failed),
* user does not need to call ntfs_inode_close himself.
*
* Return 0 on success or -1 on error with errno set to the error code.
*/
int ntfs_delete( ntfs_volume *vol, const char *pathname,
ntfs_inode *ni, ntfs_inode *dir_ni, ntfschar *name, u8 name_len )
{
ntfs_attr_search_ctx *actx = NULL;
FILE_NAME_ATTR *fn = NULL;
BOOL looking_for_dos_name = FALSE, looking_for_win32_name = FALSE;
BOOL case_sensitive_match = TRUE;
int err = 0;
#if CACHE_NIDATA_SIZE
int i;
#endif
2009-12-19 15:06:57 +01:00
#if CACHE_INODE_SIZE
struct CACHED_INODE item;
const char *p;
u64 inum = ( u64 ) - 1;
int count;
2009-12-19 15:06:57 +01:00
#endif
#if CACHE_LOOKUP_SIZE
struct CACHED_LOOKUP lkitem;
#endif
2009-12-19 15:06:57 +01:00
ntfs_log_trace( "Entering.\n" );
if ( !ni || !dir_ni || !name || !name_len )
{
ntfs_log_error( "Invalid arguments.\n" );
errno = EINVAL;
goto err_out;
}
if ( ni->nr_extents == -1 )
ni = ni->base_ni;
if ( dir_ni->nr_extents == -1 )
dir_ni = dir_ni->base_ni;
/*
* Search for FILE_NAME attribute with such name. If it's in POSIX or
* WIN32_AND_DOS namespace, then simply remove it from index and inode.
* If filename in DOS or in WIN32 namespace, then remove DOS name first,
* only then remove WIN32 name.
*/
actx = ntfs_attr_get_search_ctx( ni, NULL );
if ( !actx )
goto err_out;
2009-12-19 15:06:57 +01:00
search:
while ( !ntfs_attr_lookup( AT_FILE_NAME, AT_UNNAMED, 0, CASE_SENSITIVE,
0, NULL, 0, actx ) )
{
char *s;
BOOL case_sensitive = IGNORE_CASE;
errno = 0;
fn = ( FILE_NAME_ATTR* )( ( u8* )actx->attr +
le16_to_cpu( actx->attr->value_offset ) );
s = ntfs_attr_name_get( fn->file_name, fn->file_name_length );
ntfs_log_trace( "name: '%s' type: %d dos: %d win32: %d "
"case: %d\n", s, fn->file_name_type,
looking_for_dos_name, looking_for_win32_name,
case_sensitive_match );
ntfs_attr_name_free( &s );
if ( looking_for_dos_name )
{
if ( fn->file_name_type == FILE_NAME_DOS )
break;
else
continue;
}
if ( looking_for_win32_name )
{
if ( fn->file_name_type == FILE_NAME_WIN32 )
break;
else
continue;
}
/* Ignore hard links from other directories */
if ( dir_ni->mft_no != MREF_LE( fn->parent_directory ) )
{
ntfs_log_debug( "MFT record numbers don't match "
"(%llu != %llu)\n",
( long long unsigned )dir_ni->mft_no,
( long long unsigned )MREF_LE( fn->parent_directory ) );
continue;
}
if ( fn->file_name_type == FILE_NAME_POSIX || case_sensitive_match )
case_sensitive = CASE_SENSITIVE;
if ( ntfs_names_are_equal( fn->file_name, fn->file_name_length,
name, name_len, case_sensitive,
ni->vol->upcase, ni->vol->upcase_len ) )
{
if ( fn->file_name_type == FILE_NAME_WIN32 )
{
looking_for_dos_name = TRUE;
ntfs_attr_reinit_search_ctx( actx );
continue;
}
if ( fn->file_name_type == FILE_NAME_DOS )
looking_for_dos_name = TRUE;
break;
}
}
if ( errno )
{
/*
* If case sensitive search failed, then try once again
* ignoring case.
*/
if ( errno == ENOENT && case_sensitive_match )
{
case_sensitive_match = FALSE;
ntfs_attr_reinit_search_ctx( actx );
goto search;
}
goto err_out;
}
if ( ntfs_check_unlinkable_dir( ni, fn ) < 0 )
goto err_out;
if ( ntfs_index_remove( dir_ni, ni, fn, le32_to_cpu( actx->attr->value_length ) ) )
goto err_out;
if ( ntfs_attr_record_rm( actx ) )
goto err_out;
ni->mrec->link_count = cpu_to_le16( le16_to_cpu(
ni->mrec->link_count ) - 1 );
ntfs_inode_mark_dirty( ni );
if ( looking_for_dos_name )
{
looking_for_dos_name = FALSE;
looking_for_win32_name = TRUE;
ntfs_attr_reinit_search_ctx( actx );
goto search;
}
/* TODO: Update object id, quota and securiry indexes if required. */
/*
* If hard link count is not equal to zero then we are done. In other
* case there are no reference to this inode left, so we should free all
* non-resident attributes and mark all MFT record as not in use.
*/
#if CACHE_LOOKUP_SIZE
/* invalidate entry in lookup cache */
lkitem.name = ( const char* )NULL;
lkitem.namesize = 0;
lkitem.inum = ni->mft_no;
lkitem.parent = dir_ni->mft_no;
ntfs_invalidate_cache( vol->lookup_cache, GENERIC( &lkitem ),
lookup_cache_inv_compare, CACHE_NOHASH );
#endif
2009-12-19 15:06:57 +01:00
#if CACHE_INODE_SIZE
inum = ni->mft_no;
if ( pathname )
{
/* invalide cache entry, even if there was an error */
/* Remove leading /'s. */
p = pathname;
while ( *p == PATH_SEP )
p++;
if ( p[0] && ( p[strlen( p )-1] == PATH_SEP ) )
ntfs_log_error( "Unnormalized path %s\n", pathname );
item.pathname = p;
item.varsize = strlen( p );
}
else
{
item.pathname = ( const char* )NULL;
item.varsize = 0;
}
item.inum = inum;
count = ntfs_invalidate_cache( vol->xinode_cache, GENERIC( &item ),
inode_cache_inv_compare, CACHE_NOHASH );
if ( pathname && !count )
ntfs_log_error( "Could not delete inode cache entry for %s\n",
pathname );
2009-12-19 15:06:57 +01:00
#endif
if ( ni->mrec->link_count )
{
ntfs_inode_update_times( ni, NTFS_UPDATE_CTIME );
goto ok;
}
if ( ntfs_delete_reparse_index( ni ) )
{
/*
* Failed to remove the reparse index : proceed anyway
* This is not a critical error, the entry is useless
* because of sequence_number, and stopping file deletion
* would be much worse as the file is not referenced now.
*/
err = errno;
}
if ( ntfs_delete_object_id_index( ni ) )
{
/*
* Failed to remove the object id index : proceed anyway
* This is not a critical error.
*/
err = errno;
}
ntfs_attr_reinit_search_ctx( actx );
while ( !ntfs_attrs_walk( actx ) )
{
if ( actx->attr->non_resident )
{
runlist *rl;
rl = ntfs_mapping_pairs_decompress( ni->vol, actx->attr,
NULL );
if ( !rl )
{
err = errno;
ntfs_log_error( "Failed to decompress runlist. "
"Leaving inconsistent metadata.\n" );
continue;
}
if ( ntfs_cluster_free_from_rl( ni->vol, rl ) )
{
err = errno;
ntfs_log_error( "Failed to free clusters. "
"Leaving inconsistent metadata.\n" );
continue;
}
free( rl );
}
}
if ( errno != ENOENT )
{
err = errno;
ntfs_log_error( "Attribute enumeration failed. "
"Probably leaving inconsistent metadata.\n" );
}
/* All extents should be attached after attribute walk. */
#if CACHE_NIDATA_SIZE
/*
* Disconnect extents before deleting them, so they are
* not wrongly moved to cache through the chainings
*/
for ( i = ni->nr_extents - 1; i >= 0; i-- )
{
ni->extent_nis[i]->base_ni = ( ntfs_inode* )NULL;
ni->extent_nis[i]->nr_extents = 0;
if ( ntfs_mft_record_free( ni->vol, ni->extent_nis[i] ) )
{
err = errno;
ntfs_log_error( "Failed to free extent MFT record. "
"Leaving inconsistent metadata.\n" );
}
}
free( ni->extent_nis );
ni->nr_extents = 0;
ni->extent_nis = ( ntfs_inode** )NULL;
#else
while ( ni->nr_extents )
if ( ntfs_mft_record_free( ni->vol, *( ni->extent_nis ) ) )
{
err = errno;
ntfs_log_error( "Failed to free extent MFT record. "
"Leaving inconsistent metadata.\n" );
}
#endif
if ( ntfs_mft_record_free( ni->vol, ni ) )
{
err = errno;
ntfs_log_error( "Failed to free base MFT record. "
"Leaving inconsistent metadata.\n" );
}
ni = NULL;
ok:
ntfs_inode_update_times( dir_ni, NTFS_UPDATE_MCTIME );
2009-12-19 15:06:57 +01:00
out:
if ( actx )
ntfs_attr_put_search_ctx( actx );
if ( ntfs_inode_close( dir_ni ) && !err )
err = errno;
if ( ntfs_inode_close( ni ) && !err )
err = errno;
if ( err )
{
errno = err;
ntfs_log_debug( "Could not delete file: %s\n", strerror( errno ) );
return -1;
}
ntfs_log_trace( "Done.\n" );
return 0;
2009-12-19 15:06:57 +01:00
err_out:
err = errno;
goto out;
2009-12-19 15:06:57 +01:00
}
/**
* ntfs_link - create hard link for file or directory
* @ni: ntfs inode for object to create hard link
* @dir_ni: ntfs inode for directory in which new link should be placed
* @name: unicode name of the new link
* @name_len: length of the name in unicode characters
2009-12-19 15:06:57 +01:00
*
* NOTE: At present we allow creating hardlinks to directories, we use them
* in a temporary state during rename. But it's defenitely bad idea to have
* hard links to directories as a result of operation.
* FIXME: Create internal __ntfs_link that allows hard links to a directories
* and external ntfs_link that do not. Write ntfs_rename that uses __ntfs_link.
*
* Return 0 on success or -1 on error with errno set to the error code.
*/
static int ntfs_link_i( ntfs_inode *ni, ntfs_inode *dir_ni, ntfschar *name,
u8 name_len, FILE_NAME_TYPE_FLAGS nametype )
{
FILE_NAME_ATTR *fn = NULL;
int fn_len, err;
ntfs_log_trace( "Entering.\n" );
if ( !ni || !dir_ni || !name || !name_len ||
ni->mft_no == dir_ni->mft_no )
{
err = EINVAL;
ntfs_log_perror( "ntfs_link wrong arguments" );
goto err_out;
}
if ( ( ni->flags & FILE_ATTR_REPARSE_POINT )
&& !ntfs_possible_symlink( ni ) )
{
err = EOPNOTSUPP;
goto err_out;
}
/* Create FILE_NAME attribute. */
fn_len = sizeof( FILE_NAME_ATTR ) + name_len * sizeof( ntfschar );
fn = ntfs_calloc( fn_len );
if ( !fn )
{
err = errno;
goto err_out;
}
fn->parent_directory = MK_LE_MREF( dir_ni->mft_no,
le16_to_cpu( dir_ni->mrec->sequence_number ) );
fn->file_name_length = name_len;
fn->file_name_type = nametype;
fn->file_attributes = ni->flags;
if ( ni->mrec->flags & MFT_RECORD_IS_DIRECTORY )
{
fn->file_attributes |= FILE_ATTR_I30_INDEX_PRESENT;
fn->data_size = fn->allocated_size = const_cpu_to_le64( 0 );
}
else
{
fn->allocated_size = cpu_to_sle64( ni->allocated_size );
fn->data_size = cpu_to_sle64( ni->data_size );
}
fn->creation_time = ni->creation_time;
fn->last_data_change_time = ni->last_data_change_time;
fn->last_mft_change_time = ni->last_mft_change_time;
fn->last_access_time = ni->last_access_time;
memcpy( fn->file_name, name, name_len * sizeof( ntfschar ) );
/* Add FILE_NAME attribute to index. */
if ( ntfs_index_add_filename( dir_ni, fn, MK_MREF( ni->mft_no,
le16_to_cpu( ni->mrec->sequence_number ) ) ) )
{
err = errno;
ntfs_log_perror( "Failed to add filename to the index" );
goto err_out;
}
/* Add FILE_NAME attribute to inode. */
if ( ntfs_attr_add( ni, AT_FILE_NAME, AT_UNNAMED, 0, ( u8* )fn, fn_len ) )
{
ntfs_log_error( "Failed to add FILE_NAME attribute.\n" );
err = errno;
/* Try to remove just added attribute from index. */
if ( ntfs_index_remove( dir_ni, ni, fn, fn_len ) )
goto rollback_failed;
goto err_out;
}
/* Increment hard links count. */
ni->mrec->link_count = cpu_to_le16( le16_to_cpu(
ni->mrec->link_count ) + 1 );
/* Done! */
ntfs_inode_mark_dirty( ni );
free( fn );
ntfs_log_trace( "Done.\n" );
return 0;
2009-12-19 15:06:57 +01:00
rollback_failed:
ntfs_log_error( "Rollback failed. Leaving inconsistent metadata.\n" );
2009-12-19 15:06:57 +01:00
err_out:
free( fn );
errno = err;
return -1;
2009-12-19 15:06:57 +01:00
}
int ntfs_link( ntfs_inode *ni, ntfs_inode *dir_ni, ntfschar *name, u8 name_len )
{
return ( ntfs_link_i( ni, dir_ni, name, name_len, FILE_NAME_POSIX ) );
2009-12-19 15:06:57 +01:00
}
/*
* Get a parent directory from an inode entry
*
* This is only used in situations where the path used to access
* the current file is not known for sure. The result may be different
* from the path when the file is linked in several parent directories.
*
* Currently this is only used for translating ".." in the target
* of a Vista relative symbolic link
*/
ntfs_inode *ntfs_dir_parent_inode( ntfs_inode *ni )
{
ntfs_inode *dir_ni = ( ntfs_inode* )NULL;
u64 inum;
FILE_NAME_ATTR *fn;
ntfs_attr_search_ctx *ctx;
if ( ni->mft_no != FILE_root )
{
/* find the name in the attributes */
ctx = ntfs_attr_get_search_ctx( ni, NULL );
if ( !ctx )
return ( ( ntfs_inode* )NULL );
if ( !ntfs_attr_lookup( AT_FILE_NAME, AT_UNNAMED, 0,
CASE_SENSITIVE, 0, NULL, 0, ctx ) )
{
/* We know this will always be resident. */
fn = ( FILE_NAME_ATTR* )( ( u8* )ctx->attr +
le16_to_cpu( ctx->attr->value_offset ) );
inum = le64_to_cpu( fn->parent_directory );
if ( inum != ( u64 ) - 1 )
{
dir_ni = ntfs_inode_open( ni->vol, MREF( inum ) );
}
}
ntfs_attr_put_search_ctx( ctx );
}
return ( dir_ni );
}
2009-12-19 15:06:57 +01:00
#ifdef HAVE_SETXATTR
#define MAX_DOS_NAME_LENGTH 12
2009-12-19 15:06:57 +01:00
/*
* Get a DOS name for a file in designated directory
2009-12-19 15:06:57 +01:00
*
* Returns size if found
* 0 if not found
* -1 if there was an error (described by errno)
2009-12-19 15:06:57 +01:00
*/
static int get_dos_name( ntfs_inode *ni, u64 dnum, ntfschar *dosname )
{
size_t outsize = 0;
FILE_NAME_ATTR *fn;
ntfs_attr_search_ctx *ctx;
/* find the name in the attributes */
ctx = ntfs_attr_get_search_ctx( ni, NULL );
if ( !ctx )
return -1;
while ( !ntfs_attr_lookup( AT_FILE_NAME, AT_UNNAMED, 0, CASE_SENSITIVE,
0, NULL, 0, ctx ) )
{
/* We know this will always be resident. */
fn = ( FILE_NAME_ATTR* )( ( u8* )ctx->attr +
le16_to_cpu( ctx->attr->value_offset ) );
if ( ( fn->file_name_type & FILE_NAME_DOS )
&& ( MREF_LE( fn->parent_directory ) == dnum ) )
{
/*
* Found a DOS or WIN32+DOS name for the entry
* copy name, after truncation for safety
*/
outsize = fn->file_name_length;
/* TODO : reject if name is too long ? */
if ( outsize > MAX_DOS_NAME_LENGTH )
outsize = MAX_DOS_NAME_LENGTH;
memcpy( dosname, fn->file_name, outsize*sizeof( ntfschar ) );
}
}
ntfs_attr_put_search_ctx( ctx );
return ( outsize );
2009-12-19 15:06:57 +01:00
}
/*
* Get a long name for a file in designated directory
*
* Returns size if found
* 0 if not found
* -1 if there was an error (described by errno)
*/
static int get_long_name( ntfs_inode *ni, u64 dnum, ntfschar *longname )
{
size_t outsize = 0;
FILE_NAME_ATTR *fn;
ntfs_attr_search_ctx *ctx;
/* find the name in the attributes */
ctx = ntfs_attr_get_search_ctx( ni, NULL );
if ( !ctx )
return -1;
/* first search for WIN32 or DOS+WIN32 names */
while ( !ntfs_attr_lookup( AT_FILE_NAME, AT_UNNAMED, 0, CASE_SENSITIVE,
0, NULL, 0, ctx ) )
{
/* We know this will always be resident. */
fn = ( FILE_NAME_ATTR* )( ( u8* )ctx->attr +
le16_to_cpu( ctx->attr->value_offset ) );
if ( ( fn->file_name_type & FILE_NAME_WIN32 )
&& ( MREF_LE( fn->parent_directory ) == dnum ) )
{
/*
* Found a WIN32 or WIN32+DOS name for the entry
* copy name
*/
outsize = fn->file_name_length;
memcpy( longname, fn->file_name, outsize*sizeof( ntfschar ) );
}
}
/* if not found search for POSIX names */
if ( !outsize )
{
ntfs_attr_reinit_search_ctx( ctx );
while ( !ntfs_attr_lookup( AT_FILE_NAME, AT_UNNAMED, 0, CASE_SENSITIVE,
0, NULL, 0, ctx ) )
{
/* We know this will always be resident. */
fn = ( FILE_NAME_ATTR* )( ( u8* )ctx->attr +
le16_to_cpu( ctx->attr->value_offset ) );
if ( ( fn->file_name_type == FILE_NAME_POSIX )
&& ( MREF_LE( fn->parent_directory ) == dnum ) )
{
/*
* Found a POSIX name for the entry
* copy name
*/
outsize = fn->file_name_length;
memcpy( longname, fn->file_name, outsize*sizeof( ntfschar ) );
}
}
}
ntfs_attr_put_search_ctx( ctx );
return ( outsize );
}
2009-12-19 15:06:57 +01:00
/*
* Get the ntfs DOS name into an extended attribute
2009-12-19 15:06:57 +01:00
*/
int ntfs_get_ntfs_dos_name( ntfs_inode *ni, ntfs_inode *dir_ni,
char *value, size_t size )
{
int outsize = 0;
char *outname = ( char* )NULL;
u64 dnum;
int doslen;
ntfschar dosname[MAX_DOS_NAME_LENGTH];
dnum = dir_ni->mft_no;
doslen = get_dos_name( ni, dnum, dosname );
if ( doslen > 0 )
{
/*
* Found a DOS name for the entry, make
* uppercase and encode into the buffer
* if there is enough space
*/
ntfs_name_upcase( dosname, doslen,
ni->vol->upcase, ni->vol->upcase_len );
if ( ntfs_ucstombs( dosname, doslen, &outname, size ) < 0 )
{
ntfs_log_error( "Cannot represent dosname in current locale.\n" );
outsize = -errno;
}
else
{
outsize = strlen( outname );
if ( value && ( outsize <= ( int )size ) )
memcpy( value, outname, outsize );
else if ( size && ( outsize > ( int )size ) )
outsize = -ERANGE;
free( outname );
}
}
else
{
if ( doslen == 0 )
errno = ENODATA;
outsize = -errno;
}
return ( outsize );
2009-12-19 15:06:57 +01:00
}
/*
* Change the name space of an existing file or directory
2009-12-19 15:06:57 +01:00
*
* Returns the old namespace if successful
* -1 if an error occurred (described by errno)
2009-12-19 15:06:57 +01:00
*/
static int set_namespace( ntfs_inode *ni, ntfs_inode *dir_ni,
ntfschar *name, int len,
FILE_NAME_TYPE_FLAGS nametype )
{
ntfs_attr_search_ctx *actx;
ntfs_index_context *icx;
FILE_NAME_ATTR *fnx;
FILE_NAME_ATTR *fn = NULL;
BOOL found;
int lkup;
int ret;
ret = -1;
actx = ntfs_attr_get_search_ctx( ni, NULL );
if ( actx )
{
found = FALSE;
do
{
lkup = ntfs_attr_lookup( AT_FILE_NAME, AT_UNNAMED, 0,
CASE_SENSITIVE, 0, NULL, 0, actx );
if ( !lkup )
{
fn = ( FILE_NAME_ATTR* )( ( u8* )actx->attr +
le16_to_cpu( actx->attr->value_offset ) );
found = ( MREF_LE( fn->parent_directory )
== dir_ni->mft_no )
&& !memcmp( fn->file_name, name,
len * sizeof( ntfschar ) );
}
}
while ( !lkup && !found );
if ( found )
{
icx = ntfs_index_ctx_get( dir_ni, NTFS_INDEX_I30, 4 );
if ( icx )
{
lkup = ntfs_index_lookup( ( char* )fn, len, icx );
if ( !lkup && icx->data && icx->data_len )
{
fnx = ( FILE_NAME_ATTR* )icx->data;
ret = fn->file_name_type;
fn->file_name_type = nametype;
fnx->file_name_type = nametype;
ntfs_inode_mark_dirty( ni );
ntfs_index_entry_mark_dirty( icx );
}
ntfs_index_ctx_put( icx );
}
}
ntfs_attr_put_search_ctx( actx );
}
return ( ret );
2009-12-19 15:06:57 +01:00
}
/*
* Set a DOS name to a file and adjust name spaces
2009-12-19 15:06:57 +01:00
*
* If the new names are collapsible (same uppercased chars) :
2009-12-19 15:06:57 +01:00
*
* - the existing DOS name or DOS+Win32 name is made Posix
* - if it was a real DOS name, the existing long name is made DOS+Win32
* and the existing DOS name is deleted
* - finally the existing long name is made DOS+Win32 unless already done
*
* If the new names are not collapsible :
2009-12-19 15:06:57 +01:00
*
* - insert the short name as a DOS name
* - delete the old long name or existing short name
* - insert the new long name (as a Win32 or DOS+Win32 name)
*
* Deleting the old long name will not delete the file
* provided the old name was in the Posix name space,
* because the alternate name has been set before.
*
* The inodes of file and parent directory are always closed
*
* Returns 0 if successful
* -1 if failed
2009-12-19 15:06:57 +01:00
*/
static int set_dos_name( ntfs_inode *ni, ntfs_inode *dir_ni,
ntfschar *shortname, int shortlen,
ntfschar *longname, int longlen,
ntfschar *deletename, int deletelen, BOOL existed )
{
unsigned int linkcount;
ntfs_volume *vol;
BOOL collapsible;
BOOL deleted;
BOOL done;
FILE_NAME_TYPE_FLAGS oldnametype;
u64 dnum;
u64 fnum;
int res;
res = -1;
vol = ni->vol;
dnum = dir_ni->mft_no;
fnum = ni->mft_no;
/* save initial link count */
linkcount = le16_to_cpu( ni->mrec->link_count );
/* check whether the same name may be used as DOS and WIN32 */
collapsible = ntfs_collapsible_chars( ni->vol, shortname, shortlen,
longname, longlen );
if ( collapsible )
{
deleted = FALSE;
done = FALSE;
if ( existed )
{
oldnametype = set_namespace( ni, dir_ni, deletename,
deletelen, FILE_NAME_POSIX );
if ( oldnametype == FILE_NAME_DOS )
{
if ( set_namespace( ni, dir_ni, longname, longlen,
FILE_NAME_WIN32_AND_DOS ) >= 0 )
{
if ( !ntfs_delete( vol,
( const char* )NULL, ni, dir_ni,
deletename, deletelen ) )
res = 0;
deleted = TRUE;
}
else
done = TRUE;
}
}
if ( !deleted )
{
if ( !done && ( set_namespace( ni, dir_ni,
longname, longlen,
FILE_NAME_WIN32_AND_DOS ) >= 0 ) )
res = 0;
ntfs_inode_update_times( ni, NTFS_UPDATE_CTIME );
ntfs_inode_update_times( dir_ni, NTFS_UPDATE_MCTIME );
if ( ntfs_inode_close_in_dir( ni, dir_ni ) && !res )
res = -1;
if ( ntfs_inode_close( dir_ni ) && !res )
res = -1;
}
}
else
{
if ( !ntfs_link_i( ni, dir_ni, shortname, shortlen,
FILE_NAME_DOS )
/* make sure a new link was recorded */
&& ( le16_to_cpu( ni->mrec->link_count ) > linkcount ) )
{
/* delete the existing long name or short name */
// is it ok to not provide the path ?
if ( !ntfs_delete( vol, ( char* )NULL, ni, dir_ni,
deletename, deletelen ) )
{
/* delete closes the inodes, so have to open again */
dir_ni = ntfs_inode_open( vol, dnum );
if ( dir_ni )
{
ni = ntfs_inode_open( vol, fnum );
if ( ni )
{
if ( !ntfs_link_i( ni, dir_ni,
longname, longlen,
FILE_NAME_WIN32 ) )
res = 0;
if ( ntfs_inode_close_in_dir( ni,
dir_ni )
&& !res )
res = -1;
}
if ( ntfs_inode_close( dir_ni ) && !res )
res = -1;
}
}
}
else
{
ntfs_inode_close_in_dir( ni, dir_ni );
ntfs_inode_close( dir_ni );
}
}
return ( res );
2009-12-19 15:06:57 +01:00
}
/*
* Set the ntfs DOS name into an extended attribute
2009-12-19 15:06:57 +01:00
*
* The DOS name will be added as another file name attribute
* using the existing file name information from the original
* name or overwriting the DOS Name if one exists.
*
* The inode of the file is always closed
2009-12-19 15:06:57 +01:00
*/
int ntfs_set_ntfs_dos_name( ntfs_inode *ni, ntfs_inode *dir_ni,
const char *value, size_t size, int flags )
{
int res = 0;
int longlen = 0;
int shortlen = 0;
char newname[MAX_DOS_NAME_LENGTH + 1];
ntfschar oldname[MAX_DOS_NAME_LENGTH];
int oldlen;
ntfs_volume *vol;
u64 fnum;
u64 dnum;
BOOL closed = FALSE;
ntfschar *shortname = NULL;
ntfschar longname[NTFS_MAX_NAME_LEN];
vol = ni->vol;
fnum = ni->mft_no;
/* convert the string to the NTFS wide chars */
if ( size > MAX_DOS_NAME_LENGTH )
size = MAX_DOS_NAME_LENGTH;
strncpy( newname, value, size );
newname[size] = 0;
shortlen = ntfs_mbstoucs( newname, &shortname );
/* make sure the short name has valid chars */
if ( ( shortlen < 0 ) || ntfs_forbidden_chars( shortname, shortlen ) )
{
ntfs_inode_close_in_dir( ni, dir_ni );
ntfs_inode_close( dir_ni );
res = -errno;
return res;
}
dnum = dir_ni->mft_no;
longlen = get_long_name( ni, dnum, longname );
if ( longlen > 0 )
{
oldlen = get_dos_name( ni, dnum, oldname );
if ( ( oldlen >= 0 )
&& !ntfs_forbidden_chars( longname, longlen ) )
{
if ( oldlen > 0 )
{
if ( flags & XATTR_CREATE )
{
res = -1;
errno = EEXIST;
}
else if ( ( shortlen == oldlen )
&& !memcmp( shortname, oldname,
oldlen*sizeof( ntfschar ) ) )
/* already set, done */
res = 0;
else
{
res = set_dos_name( ni, dir_ni,
shortname, shortlen,
longname, longlen,
oldname, oldlen, TRUE );
closed = TRUE;
}
}
else
{
if ( flags & XATTR_REPLACE )
{
res = -1;
errno = ENODATA;
}
else
{
res = set_dos_name( ni, dir_ni,
shortname, shortlen,
longname, longlen,
longname, longlen, FALSE );
closed = TRUE;
}
}
}
else
res = -1;
}
else
{
res = -1;
errno = ENOENT;
}
free( shortname );
if ( !closed )
{
ntfs_inode_close_in_dir( ni, dir_ni );
ntfs_inode_close( dir_ni );
}
return ( res ? -1 : 0 );
2009-12-19 15:06:57 +01:00
}
/*
* Delete the ntfs DOS name
2009-12-19 15:06:57 +01:00
*/
int ntfs_remove_ntfs_dos_name( ntfs_inode *ni, ntfs_inode *dir_ni )
{
int res;
int oldnametype;
int longlen = 0;
int shortlen;
u64 dnum;
ntfs_volume *vol;
BOOL deleted = FALSE;
ntfschar shortname[MAX_DOS_NAME_LENGTH];
ntfschar longname[NTFS_MAX_NAME_LEN];
res = -1;
vol = ni->vol;
dnum = dir_ni->mft_no;
longlen = get_long_name( ni, dnum, longname );
if ( longlen > 0 )
{
shortlen = get_dos_name( ni, dnum, shortname );
if ( shortlen >= 0 )
{
/* migrate the long name as Posix */
oldnametype = set_namespace( ni, dir_ni, longname, longlen,
FILE_NAME_POSIX );
switch ( oldnametype )
{
case FILE_NAME_WIN32_AND_DOS :
/* name was Win32+DOS : done */
res = 0;
break;
case FILE_NAME_DOS :
/* name was DOS, make it back to DOS */
set_namespace( ni, dir_ni, longname, longlen,
FILE_NAME_DOS );
errno = ENOENT;
break;
case FILE_NAME_WIN32 :
/* name was Win32, make it Posix and delete */
if ( set_namespace( ni, dir_ni, shortname, shortlen,
FILE_NAME_POSIX ) >= 0 )
{
if ( !ntfs_delete( vol,
( const char* )NULL, ni,
dir_ni, shortname,
shortlen ) )
res = 0;
deleted = TRUE;
}
else
{
/*
* DOS name has been found, but cannot
* migrate to Posix : something bad
* has happened
*/
errno = EIO;
ntfs_log_error( "Could not change"
" DOS name of inode %lld to Posix\n",
( long long )ni->mft_no );
}
break;
default :
/* name was Posix or not found : error */
errno = ENOENT;
break;
}
}
}
else
{
errno = ENOENT;
res = -1;
}
if ( !deleted )
{
ntfs_inode_close_in_dir( ni, dir_ni );
ntfs_inode_close( dir_ni );
}
return ( res );
2009-12-19 15:06:57 +01:00
}
#endif