From owner-freebsd-scsi  Thu May  1 16:44:10 1997
Return-Path: <owner-freebsd-scsi>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.5/8.8.5) id QAA26034
          for freebsd-scsi-outgoing; Thu, 1 May 1997 16:44:10 -0700 (PDT)
Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.19])
          by hub.freebsd.org (8.8.5/8.8.5) with ESMTP id QAA25924;
          Thu, 1 May 1997 16:43:12 -0700 (PDT)
Received: (from bde@localhost) by godzilla.zeta.org.au (8.8.5/8.6.9) id JAA32257; Fri, 2 May 1997 09:38:28 +1000
Date: Fri, 2 May 1997 09:38:28 +1000
From: Bruce Evans <bde@zeta.org.au>
Message-Id: <199705012338.JAA32257@godzilla.zeta.org.au>
To: bde@zeta.org.au, gibbs@plutotech.com
Subject: Re: cvs commit: src/sys/scsi sd.c
Cc: freeBSD-arch@FreeBSD.ORG, freeBSD-scsi@FreeBSD.ORG,
        sos@freefall.freebsd.org
Sender: owner-freebsd-scsi@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk

>So dscheck should do the conversion based on the sector size in the
>disklabel?

Yes.  I'm not sure if we can trust the on-disk value, but drivers set
it before calling dsopen() and dsopen() is a good place to convert it
to a more convenient form (take log2(d_secsize) and use this for shifts).

>Does any of this discussion affect the vn driver?  What about being
>able to create an ISOFS on a hard drive?  Do we only support that via
>vn?

vn could always use 512-byte "sectors', but it needs to know about
the device block size to avoid using the VOP_STRATEGY() interface for
non-physical-block requests.  It would be useful for debugging for vn
to support arbitrary "sector" sizes.  I'm not sure if cd9660 on vn works
now.  I think it does because cd9660 talks to drivers in 512-blocks.

>>You also need to constrain file system fragment sizes to a multiple of
>>the sector size for a simple implementation.  I think only sizes of 512,
>>1024, 2048 and 4096 are practical.
>
>So newfs needs to consult the disklabel and ensure that the frag/block size
>is a a power of 2 multiple of the physical sector size in the disklabel.

If we limit it to powers of 2.  It would be nice to have support for
weird sizes as a stream of blocks.  I just don't see putting file systems
on them.

Bruce