home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.hp
- Path: sparky!uunet!spool.mu.edu!sdd.hp.com!scd.hp.com!hpuerca.atl.hp.com!der
- From: der@hpuerca.atl.hp.com (Dave Ritchie)
- Subject: Re: Disk Array & 2GB HP-UX limit
- Message-ID: <BunK3M.GKz@hpuerca.atl.hp.com>
- Organization: Hewlett-Packard NARC Atlanta
- References: <51980002@hpsgm2.sgp.hp.com> <BuBwMI.MFC@cup.hp.com> <1992Sep12.005322.3016@polari>
- Date: Wed, 16 Sep 1992 03:52:33 GMT
- Lines: 29
-
- In <1992Sep12.005322.3016@polari> lampi@polari.online.com writes:
-
- >Why would it take 64 bit pointers to address >4GB file systems? After all,
- >we're not necessarily talking about >4GB *files* (which would require
- >such pointers), we're talking about >4GB *volumes*. It's really a stupid
- >limitation that should have nothing to do with the 32 bit limit of
- >pointers. After all, we're not talking about a pointer based from the
- >beginning of the logical volume when we address the contents of a given file.
-
- >Or are we? :-)
-
- >+-+ Michael Lampi lampi@polari.online.com
-
- We are, actually. Think about the case when performing a seek to a
- particular block of a mounted file system. Since HP-UX and other
- Unix-derived operating systems allow devices to be manipulated using
- the same routines that are used with normal files, it would require a
- change in the semantics of how lseek works for device files vs. regular
- files (since the current semantics of lseek are tied to a 32-bit offset).
-
- There are also problems with drivers (as I recall, there are 22 bits of SCSI
- logical block addressing when using 10 byte commmands, which limits you to
- 4GByte devices when using 1k blocks).
-
- -- Dave Ritchie
- der@hpuerca.atl.hp.com
- HP North American Response Center, Atlanta
-
-
-