Manual browser: vndcompress(1)

VNDCOMPRESS(1) General Commands Manual VNDCOMPRESS(1)


vndcompress, vnduncompresscompress and uncompress disk images in cloop2 format


vndcompress [-c] [-rR] [-b blocksize] [-k checkpoint-blocks] [-l length] [-p partial-offset] [-w winsize] image compressed-image [blocksize]

vnduncompress [-d] [-w winsize] compressed-image image


The vndcompress utility compresses disk images in cloop2 format, which the kernel's vnd(4) device can interpret as read-only disk devices using the -z option to vnconfig(8).

By default, vndcompress compresses an image, and vnduncompress uncompresses an image, but the -c and -d options can control whether either utility compresses or decompresses.

The following options are available for the compression operation:

-b blocksize
Set the compression block size to blocksize, which must be a multiple of 512 and must be no more than 4294966784, or 0xfffffe00. (On 32-bit systems, the limit may be smaller, limited by the available virtual address space.)

For compatibility with the old version of vndcompress, the compression block size may instead be specified at the end of the command line.

-k checkpoint-blocks
Write a checkpoint after every checkpoint-blocks blocks of output. If interrupted, vndcompress can restart at the last checkpoint with the -r option. You can also request a checkpoint at any time by sending SIGUSR2 to the vndcompress process.
-l length
Specify the length of the input, so that the input may be a pipe, socket, or FIFO. Otherwise, the input must know its size, i.e. must have its size reported in st_size by fstat(2).
-p partial-blocks
Stop after writing partial-blocks blocks of output. This option is mainly useful for automatic testing of restarts.
With the -r option, if restarting fails, then abort right now and don't touch the output file.
Try to restart a partial compression from the last checkpoint. If restarting fails, and the -R option is not specified, then truncate the output file and start compression afresh. Restarting may fail for various reasons: if there have been no checkpoints, or if the output file has been corrupted in some easily recognizable ways, or if the input file's size has changed, or if the block size does not match, and so on.

The following option is available for both compression and decompression:

-w winsize
Use an in-memory window of winsize entries into the table of compressed block offsets. If winsize is zero, vndcompress will use memory proportional to the number of blocks in the uncompressed image, namely 64 bits or 8 bytes per block. If winsize is nonzero, vndcompress will use memory proportional to winsize, and independent of the size of the uncompressed image.

A nonzero winsize requires the compressed image to be a seekable file, which compression requires anyway, in order to record the offsets of compressed blocks once they are compressed and written, but which is a limitation for decompression. Thus, decompressing from a pipe is incompatible with a nonzero winsize.

By default, vndcompress uses a fixed window size, unless decompressing with nonseekable input.


The vndcompress utility exits 0 on success, and >0 if an error occurs.


Compress an ISO 9660 CD-ROM image:

# vndcompress cdrom.iso cdrom.izo

Send a 59 GB disk partition over a local network with netcat (don't do this over the internet!):

# nc 12345 < /dev/rcgd1h

Receive it and save it compressed on another host, showing a progress bar interactively, restarting if possible, and taking a checkpoint every 16 MB (i.e., every 256 compression blocks, which are 64 KB by default):

# nc -l 12345 | progress -l 59g \ 
    vndcompress -l 59g -k 256 -r /dev/stdin disk.cloop2

If the process is interrupted and you rewire your network and disks so that the receiving host now has the disk you want to image, you can start up where you left off, using the -R option to keep vndcompress from clobbering the partial transfer if anything goes wrong:

# vndcompress -l 59g -k 256 -rR /dev/rcgd1h disk.cloop2

Mount the disk with vnd(4), assuming your kernel was built with the VND_COMPRESSION option enabled:

# vnconfig -z vnd0 disk.cloop2 
# mount /dev/vnd0d /mnt


vndcompress responds to the following signals:
Report progress to standard error.
Write a checkpoint to disk now.


The cloop2 format consists of a header, an offset table, and a sequence of compressed blocks. The header is described by the following packed structure:

struct cloop2_header { 
	char		cl2h_magic[128]; 
	uint32_t	cl2h_blocksize; 
	uint32_t	cl2h_n_blocks; 
} __packed;

The cl2h_magic field is an arbitrary sequence of 128 bits, the cl2h_blocksize field is a big-endian integer number of bytes per compression block, and the cl2h_n_blocks field is a big-endian integer number of the compression blocks in the image.

The offset table is a sequence of one more than cl2h_n_blocks big-endian 64-bit integers specifying the offset of each compression block relative to the start of the file. The extra offset specifies the end of the last compression block, which may be truncated if the uncompressed image's size is not a multiple of the compression block size.


vnd(4), vnconfig(8)


The vndcompress command first appeared in NetBSD 3.0. It was rewritten to be more robust, to support restarting partial transfers, and to support bounded memory usage in NetBSD 7.0.
January 21, 2014 NetBSD 7.0