Return to BSD News archive
Path: sserve!newshost.anu.edu.au!munnari.oz.au!mel.dit.csiro.au!merlin!harbinger.cc.monash.edu.au!simtel!zombie.ncsc.mil!news.mathworks.com!zib-berlin.de!news.tu-chemnitz.de!irz401!narcisa.sax.de!not-for-mail
From: j@narcisa.sax.de (J Wunsch)
Newsgroups: comp.unix.bsd.freebsd.misc
Subject: Re: 2.0.5 tape blocksize???
Date: 16 Jun 1995 20:01:36 +0200
Organization: Private U**x site, Dresden.
Lines: 23
Message-ID: <3rsgu0$j29@bonnie.tcd-dresden.de>
References: <3rja69$9fu@canyon.sr.hp.com> <3rm34s$3d4@marvin.gmd.de> <3rpomn$97r@canyon.sr.hp.com>
Reply-To: joerg_wunsch@uriah.heep.sax.de
NNTP-Posting-Host: 192.109.108.139
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 8bit
Darryl Okahata <darrylo@sr.hp.com> wrote:
> Out of curiousity, why is the tape buffer size in the default
>kernel limited to under 10240-bytes? The default buffer size in tar is
>20 blocks or 10240 bytes, and it's kinda silly to regen the kernel just
>to use default blocking sizes with tar.
This ain't a kernel problem. DAT and QIC >= 525 are variable-length
recorded, i.e. the driver will write the blocks as long as requested
by the write(2) system call. However, this requires a read(2) system
call of exactly the same length. Apparently, tar (10 KB) and cpio (5
KB) disagree in their defaults, and the tapes are being read with
cpio.
You can force the kernel to write fixed-length records however (it's
an option of mt(1)).
(Regen'ing the kernel won't even change anything at all.)
--
cheers, J"org private: joerg_wunsch@uriah.heep.sax.de
http://www.sax.de/~joerg/
Never trust an operating system you don't have sources for. ;-)