[9291] in linux-scsi channel archive
Re: concurrent commands to same LUN ?
daemon@ATHENA.MIT.EDU (Martin Peschke)
Thu Aug 3 07:06:24 2000
Date: Thu, 3 Aug 2000 12:49:17 +0200 (MET DST)
From: Martin Peschke <peschke@fh-brandenburg.de>
To: Kurt Garloff <garloff@suse.de>
cc: Linux SCSI list <linux-scsi@vger.rutgers.edu>
In-Reply-To: <20000726195056.S32091@garloff.etpnet.phys.tue.nl>
Message-ID: <Pine.GSO.4.10.10008031143300.20485-100000@zeus.fh-brandenburg.de>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
On Wed, 26 Jul 2000, Kurt Garloff wrote:
> On Mon, Jul 24, 2000 at 06:07:39PM +0200, Martin Peschke wrote:
> > I am just thinking about multiple paths in Fibre Channel environments
> > because of several FC adapters within a single host connected to the same
> > switched fabric.
>
> Be careful. As Linux does not know that the devices it sees multiple times
> (on each host adapter once) are actually the same, it will not ense data
> integrity. If you change the contents of a device (filesystem) on one
> adapter and read back from the other one, your buffer cache and your FS will
> look inconsistent.
I'm currently looking for ways to implement multipathing above the SCSI
stack or alternately within a HBA driver. In my opinion, such
solutions can only be workarounds. Multipathing
should be a midlayer facility, since it is related to different things of
the current midlayer, i.e. device scanning and error recovery. Currently I
see it from a Fibre Channel point of view. I do not know which
requirements will be raised by other future attachments, like
Firewire. But I think there is a need to discuss a stack that considers
different physical interfaces, their protocols and characteristics. The
current SCSI stack is designed for Parallel SCSI, even if someone wrote a
FC-4 backend.
Regards
Martin Peschke
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.rutgers.edu
