NTFS Overview ============= Legato Systems, Inc. (http://www.legato.com) have sponsored Anton Altaparmakov to develop NTFS on Linux since June 2001. To mount an NTFS volume, use the filesystem type 'ntfs'. The driver currently works only in read-only mode, with no fault-tolerance supported. If you enable the dangerous(!) write support, make sure you can recover from a complete loss of data. Also, download the Linux-NTFS project distribution from Sourceforge at http://sourceforge.net/projects/linux-ntfs/ and always run the included ntfsfix utility after performing a write to an NTFS partition from Linux to fix some of the damage done by the Linux NTFS driver and to schedule an automatic chkdsk when Windows reboots. You should run ntfsfix _after_ unmounting the partition in Linux but _before_ rebooting into Windows. During the next reboot into Windows, chkdsk will be run automatically fixing the remaining damage. If no errors are found it is a good indication that the driver + ntfsfix together worked to full satisfaction. (-; Please note that the experimental write support is limited to Windows NT4 and earlier versions at the moment. If you think you have discovered a bug please have look at the "Known bugs" section below to see whether it isn't known already. For ftdisk support, limited success was reported with volume sets on top of the md driver, although mirror and stripe sets should work as well - if the md driver can be talked into using the same layout as Windows NT. However, using the md driver will fail if any of your NTFS partitions have an odd number of sectors. Supported mount options ======================= iocharset=name Character set to use when returning file names. Unlike VFAT, NTFS suppresses names that contain unconvertible characters. Note that most character sets contain insufficient characters to represent all possible Unicode characters that can exist on NTFS. To be sure you are not missing any files, you are advised to use the iocharset=utf8 which should be capable of representing all Unicode characters. utf8= Use UTF-8 for converting file names. - It is preferable to use iocharset=utf8 instead, but if the utf8 NLS is not available, you can use this utf8 option, which enables the driver's builtin utf8 conversion functions. uid= gid= umask= These options work as documented in mount(8). By default, the files are owned by root and not readable by anyone else. posix= If enabled, the file system distinguishes between upper and lower case. The 8.3 alias names are presented as hard links instead of being suppressed. show_sys_files= If enabled, show all system files as normal files. Note that $MFT does not appear unless specifically requested. For example in bash, use: "ls -l \$MFT". Be careful not to write anything to them or you could crash the kernel and/or corrupt your file system! mft_zone_multiplier= Set the MFT zone multiplier for the volume (this setting is not persistent across mounts and can be changed from mount to mount but cannot be changed on remount). Values of 1 to 4 are allowed, 1 being the default. The MFT zone multiplier determines how much space is reserved for the MFT on the volume. If all other space is used up, then the MFT zone will be shrunk dynamically, so this has no impact on the amount of free space. However, it can have an impact on performance by affecting fragmentation of the MFT. In general use the default. If you have a lot of small files then use a higher value. The values have the following meaning: Value MFT zone size (% of volume size) 1 12.5% 2 25% 3 37.5% 4 50% Known bugs and (mis-)features ============================= - Do not use the driver for writing as it corrupts the file system. If you do use it, get the Linux-NTFS tools and use the ntfsfix utility after dismounting a partition you wrote to. - Writing of extension records is not supported properly. Please send bug reports/comments/feed back/abuse to the Linux-NTFS development list at sourceforge: linux-ntfs-dev@lists.sourceforge.net ChangeLog ========= NTFS 1.1.21: - Fixed bug with reading $MFT where we try to read higher mft records before having read the $DATA attribute of $MFT. (Note this is only a partial solution which will only work in the case that the attribute list is resident or non-resident but $DATA is in the first 1024 bytes. But this should be enough in the majority of cases. I am not going to bother fixing the general case until someone finds this to be a problem for them, which I doubt very much will ever happen...) - Fixed bogus BUG() call in readdir(). NTFS 1.1.20: - Fixed two bugs in ntfs_readwrite_attr(). Thanks to Jan Kara for spotting the out of bounds one. - Check return value of set_blocksize() in ntfs_read_super() and make use of get_hardsect_size() to determine the minimum block size. - Fix return values of ntfs_vcn_to_lcn(). This should stop peoples start of partition being overwritten at random. NTFS 1.1.19: - Fixed ntfs_getdir_unsorted(), ntfs_readdir() and ntfs_printcb() to cope with arbitrary cluster sizes. Very important for Win2k+. Also, make them detect directories which are too large and truncate the enumeration pretending end of directory was reached. Detect more error conditions and overflows. All this fixes the problem where the driver could end up in an infinite loop under certain circumstances. - Fixed potential memory leaks in Unicode conversion functions and setup correct NULL return values. NTFS 1.1.18: - Enhanced & bug fixed cluster deallocation (race fixes, etc.) - Complete rewrite of cluster allocation, now race free. - Fixed several bugs in the attribute modification codepaths. - Hopefully fixed bug where the first sectors of some people's partitions would be overwritten by the mft. And in general fixed up mft extension code a bit (still incomplete though). - Introduce splice_runlist() to allow generic splicing of two run lists into one. - MFT zone is now implemented. [Stage 2 of 3; only lack dynamic growing of mft zone but that is AFAIK not even done by Windows, and the overhead would be so large that it is probably not worth doing at all, so Stage 3 might never happen...] - Complete rewrite of $MFT extension and ntfs inode allocation code. - Made the NTFS driver initialization string show the compile options used (i.e. whether read-only or read-write, whether a module, and whether with debug support). - Modify ntfs_fill_mft_header() to set all fields and to accept more arguments. - Get rid of superfluous add_mft_header(). - Get rid of some unused code. - Fixed several bugs in and generally cleaned up ntfs_readdir, ntfs_getdir_unsorted(), and ntfs_printcb. Now they spew out huge amounts of debug output if debugging is enabled. This will be removed once I know that this works for everyone. - ntfs_readdir now shows hidden files. The only files that are now hidden are the first 16 inodes (i.e. the hard coded system files), which is consistent with Windows NT4. Using the show_sys_files mount option, these files are then shown, too. - Fixed the displaying of the "." and ".." directories. We still cannot cope with more than 65536 files in a directory index block which is not a problem and we now cannot cope with more than 32766 directory index blocks which should not be a problem unless you have a directory with an insanely large number of files in it. The exact number depends on the length of the file names of the directory entries and on the size of the dircetory index blocks. - Fixed all problems with the last file in a directory (e.g. the last file should no longer disappear and tab completion should work). If there are still disappearing files or any other problems with the last file in a directory, please report them! Thanks. - Rewrote ntfs_extend_attr() to use the new cluster allocator and the freshly introduced splice_runlists() function. This simplified ntfs_extend_attr() a lot which in turn seems to have removed one or more bugs from it. - Probably other things I have forgotten... (-; - Removed dollar signs from the names in the system file enumeration. Apparently gcc doesn't support dollar signs on PPC architecture. (Andrzej Krzysztofowicz) NTFS 1.1.17: - Fixed system file handling. No longer need to use show_sys_files option for driver to work fine. System files are now always treated the same, but without the option, they are made invisible to directory listings. As a result system files can once again be opened even without the show_sys_files option. This is important for the statfs system call to work properly, for example. - Implemented MFT zone including mount parameter to tune it (just like in Windows via the registry, only we make it per mount rather than global for the whole driver, so we are better but we have no way of storing the value as we don't have a registry so either specify on each mount or put it in /etc/fstab). [Stage 1 of 3, mount parameter handling.] - Fixed fixup functions to handle corruption cases and to return error codes to the caller. - Made fixup functions apply hotfixes where sensible. [Stage 1 of 2+, in memory only.] - Fixed ommission of "NTFS: " string in ntfs_error() output. - Fixed stupid if statement bug in unistr.c. Thanks to Yann E. Morin for spotting it. - Get rid of all uses of max and min macros. This actually allowed for optimizing the code in several places so it was a Good Thing(TM). - Make ntfs use generic_file_open to enforce the O_LARGEFILE flag. - Detect encrypted files and refuse to access them (return EACCES error code to user space). - Fix handling of encrypted & compressed files so that an encrypted file no longer is considered to be compressed (this was causing kernel segmentation faults). NTFS 1.1.16: - Removed non-functional uni_xlate mount options. - Clarified the semantics of the utf8 and iocharset mount options. - Threw out the non-functional mount options for using hard coded character set conversion. Only kept utf8 one. - Fixed handling of mount options and proper handling of faulty mount options on remount. - Cleaned up character conversion which basically became simplified a lot due to the removal of the above mentioned mount options. - Made character conversion to be always consistent. Previously we could output to the VFS file names which we would then not accept back from the VFS so in effect we were generating ghost entries in the directory listings which could not be accessed by any means. - Simplified time conversion functions drastically without sacrificing accuracy. (-8 - Fixed a use of a pointer before the check for the pointer being NULL, reported by the Stanford checker. - Fixed several missing error checks, reported by the Stanford checker and fixed by Rasmus Andersen. NTFS 1.1.15 (changes since kernel 2.4.4's NTFS driver): - New mount option show_sys_files= to show all system files as normal files. - Support for files and in general any attributes up to the full 2TiB size supported by the NTFS filesystem. Note we only support up to 32-bits worth of inodes/clusters at this point. - Support for more than 128kiB sized runlists (using vmalloc_32() instead of kmalloc()). - Fixed races in allocation of clusters and mft records. - Fixed major bugs in attribute handling / searching / collation. - Fixed major bugs in compressing a run list into a mapping pairs array. - Fixed major bugs in inode allocation. Especially file create and mkdir. - Fixed memory leaks. - Fixed major bug in inode layout assignment of sequence numbers. - Lots of other bug fixes I can't think of right now... - Fixed NULL bug found by the Stanford checker in ntfs_dupuni2map(). - Convert large stack variable to dynamically allocated one in ntfs_get_free_cluster_count() (found by Stanford checker). Kernel 2.4.4: - Started ChangeLog.