UFS filesystem size limit

Matthew Dillon dillon at apollo.backplane.com
Fri Sep 2 10:55:49 PDT 2005


:For reference, I'm using:
:
:# uname -sv
:DragonFly DragonFly 1.3-Preview #0: Fri Jul  8 14:09:49 CEST 2005
: root at xxxxxxxxxxxxxxxxxxxxxx:/usr/obj/usr/src/sys/GENERIC 
:
:I have created a vinum partition that is 1.12 terabytes. As some of
:you may expect, /sbin/newfs is not happy with this size (# sectors in
:partition > max value for a signed 32 bit integer).
:
:Is there any current way to get a filesystem past the terabyte limit? 
:------
:Dave Hayes - Consultant - Altadena CA, USA - dave at xxxxxxxxxxx 
:>>> The opinions expressed above are entirely my own <<<
:
:Only one who is seeking certainty can be uncertain.

    There are a number of limits that effectively make it 1TB due
    to the block representation for device I/O (2^31 x 512).  There
    is also a theoretical UFS1 limit of (2^31 x fs_block_size)
    which implies e.g. (2^31 x 8192) = 16TB (more if you use a
    larger block size), but there are probably still internal 
    overflows in UFS1 that effectively make that limit 1TB too.

    Hiten has been working on the BUF/BIO subsystem.  He and I have
    been going through bunches of patch sets but I really doubt 
    that we will have a 64-bit clean I/O path by the next release.
    At the moment we are simply working to separate the b_lblkno/
    b_blkno/b_pblkno mess and to introduce a layer that we will
    be able to use to implement I/O scheduling.

    FreeBSD's UFS2 can do larger filesystems now, but you would have
    to inquire on the FreeBSD lists as to the stability.

					-Matt
					Matthew Dillon 
					<dillon at xxxxxxxxxxxxx>





More information about the Users mailing list