13. Reading VMS tapes from Linux

(This section was written by Mike Miller)

13.1. Introduction

From time to time you may want to read tapes made on a VMS machine (or tapes that are made to be readable by VMS and *nix systems). In general, this is quite easy for DECFILES11A tapes.

Although you may be reading this as part of a Linux mini-HOWTO, I believe that the information here is applicable to any *nix system - I've done this on Linux, HP, Sun and DEC *nix systems. The main platform dependences that I know are device names, which can differ on different systems, and the options to mt for specifying the device name (for example, mt -f on Linux and mt -t on HPUX 9).

Caveat - I've only tried this with Exabyte 8mm SCSI tape drives. If you've read other formats (still got those 9-tracks lying around?) let me know and I'll add a note here.

13.2. The Basics

When reading a tape that has been made with the VMS ``copy'' command (or has at least been made to look like it was made with copy) all you need to know is there will be three files on the tape for each actual data file - a header, the data, and a trailer. The header and trailer are interesting in that they contain info on the file as it existed under VMS. The data is, well, the data. Each of these files can be extracted from the tape with the dd command. The tape can be positioned by skipping around with the mt command.

Example: I've got VMS tape with a series of files on it. The first two were originally named ce66-2.evt and ce66-3.evt on a VMS system. The tape label is c66a2.

If I execute these commands:

> dd if=$TAPE bs=16k of=header1
> dd if=$TAPE bs=16k of=data1
> dd if=$TAPE bs=16k of=trailer1
> dd if=$TAPE bs=16k of=header2
> dd if=$TAPE bs=16k of=data2
> dd if=$TAPE bs=16k of=trailer2

I'm left with six files: header1, data1, trailer1, header2, data2 and trailer2. The syntax here is if="input file", bs="block size" and of="output file". TAPE is expected to be a variable containing the device name of your tape drive - for example, /dev/nts0 if you are using the first SCSI tape on Linux.

If you wanted to read the second file, but not the first, you didn't care about the header, and you wanted to use the original file name, do this:

> mt -f $TAPE fsf 4
> dd if=$TAPE bs=16k of=ce66-2.evt
> mt -f $TAPE fsf 1

Note the 4 - skip three files for the first file on the tape and one for the next header. The second mt skips the second file's trailer and positions the tape at the beginning of the next file - the third VMS header. You can also use mt to skip backwards (bsf), rewind (rewind) and rewind and unload the tape (offline, rewoffl).

13.3. Some details

The header and trailer files contain uppercase ASCII data used by VMS to store file information such as block size. They also contain the file name, which can be handy if you want to build scripts that automate read files or search for a particular file. The first header on a tape volume is slightly different than subsequent headers.

For a file that is the first file on the tape, as in header1 of the above example, the first four characters will be "VOL1" followed by the volume name. In the example, header1 starts with "VOL1C66A2". This is followed by a series of spaces terminated with a numeral. After that is the string "HDR1" which indicates that this is a file header. The characters immediately following the HDR1 string are the VMS file name. In in the example, this is "HDR1CE66-2.EVT". The next field is the volume name again.

For files that are not the first file on the tape, the initial VOL1 field is not present. Other than that the header has the same structure as for the initial file. Another useful field is the 7th field, which will end with "DECFILES11A". This must be present on tapes that conform to the DEC Files11A standard.

       field	 initial header		subsequent headers
       =====	 =============		==================
         1	 VOL1 + volume name	HDR1 + file name
         2	 3HDR1 + file name	volume name
         3	 volume name		

         6				...DECFILES11A
	 7	...DECFILES11A

For full details on the header and trailer format, see the DEC FILES11A documentation (on the orange/grey wall - ask your local VMS folks :-).

13.4. Comment on Block Size

In the example, I used a block size of 16k. On a *nix system, there is no block size associated with a file on disk while, under VMS, each file has a specific block size. That means that block size doesn't matter too much on the Linux end... unless it makes it hard to read the tape. If you have difficulty figuring out the block size and reading a tape, you can try setting the hardware block size on your tape drive using `mt -f $TAPE setblk 0'. The exact form of the setblk option (and its availability) may depend on the version of mt, the tape drive hardware interface and on your particular flavor of *nix.

(Thanks to Wojtek Skulski () for pointing out setblk.)