Section: Maintenance Commands (8)
Updated: November 2005
The undelete mode takes the files either matching the regular expression (option -m) or specified by the inode-expressions and recovers as much of the data as possible. It saves the result to another location. Partly for safety, but mostly because NTFS write support isn't finished.
ntfsundelete cannot perform the impossible.
When a file is deleted the MFT Record is marked as not in use and the bitmap representing the disk usage is updated. If the power isn't turned off immediately, the free space, where the file used to live, may become overwritten. Worse, the MFT Record may be reused for another file. If this happens it is impossible to tell where the file was on disk.
In NTFS all the filenames are stored as Unicode. They will be converted into the current locale for display by ntfsundelete. The utility has successfully displayed some Chinese pictogram filenames and then correctly recovered them.
In rare circumstances, a single MFT Record will not be large enough to hold the metadata describing a file (a file would have to be in hundreds of fragments for this to happen). In these cases one MFT record may hold the filename, but another will hold the information about the data. ntfsundelete will not try and piece together such records. It will simply show unnamed files with data.
To recover a file ntfsundelete has to read the file's metadata. Unfortunately, this isn't always intact. When a file is deleted, the metadata can be left in an inconsistent state. e.g. the file size may be zero; the dates of the file may be set to the time it was deleted, or random.
To be safe ntfsundelete will pick the largest file size it finds and write that to disk. It will also try and set the file's date to the last modified date. This date may be the correct last modified date, or something unexpected.
Below is a summary of all the options that ntfsundelete accepts. Nearly all options have two equivalent names. The short name is preceded by and the long name is preceded by -. Any single letter options, that don't take an argument, can be combined into a single command, e.g. -fv is equivalent to -f -v. Long named options can be abbreviated to any unique prefix of their name.
The output of scan will be:
Inode Flags %age Date Time Size Filename 6038 FN.. 93% 2002-07-17 13:42 26629 thesis.doc
<TABLE BORDER> <TR> <TD> <TABLE> <TR VALIGN="top| <TD CLASS="c1|Flag</TD> <TD>Description
</TD> </TR> <TR VALIGN="top| <TD>F/D</TD> <TD>File/Directory
</TD> </TR> <TR VALIGN="top| <TD>N/R</TD> <TD>(Non-)Resident data stream
</TD> </TR> <TR VALIGN="top| <TD>C/E</TD> <TD>Compressed/Encrypted data stream
</TD> </TR> <TR VALIGN="top| <TD>!</TD> <TD>Missing attributes
</TD> </TR> </TABLE> </TD> </TR> </TABLE>
The percentage field shows how much of the file can potentially be recovered.
When the file is recovered it will be given its original name, unless the --output option is used.
Look for deleted files on /dev/hda1.
Look for deleted documents on /dev/hda1.
ntfsundelete /dev/hda1 -s -m '*.doc'
Look for deleted files between 5000 and 6000000 bytes, with at least 90% of the data recoverable, on /dev/hda1.
ntfsundelete /dev/hda1 -S 5k-6m -p 90
Look for deleted files altered in the last two days
ntfsundelete /dev/hda1 -t 2d
Undelete inodes 2, 5 and 100 to 131 of device /dev/sda1
ntfsundelete /dev/sda1 -u -i 2,5,100-131
Undelete inode number 3689, call the file 'work.doc', set it to recovered size and put it in the user's home directory.
ntfsundelete /dev/hda1 -u -T -i 3689 -o work.doc -d ~
Save MFT Records 3689 to 3690 to a file 'debug'
ntfsundelete /dev/hda1 -c 3689-3690 -o debug
There are some small limitations to ntfsundelete, but currently no known bugs. If you find a bug please send an email describing the problem to the development team:
ntfsundelete is part of the ntfs-3g package and is available from:
Tutoriais de Tecnologia Web