From owner-freebsd-scsi  Wed May  2  6: 1:33 2001
Delivered-To: freebsd-scsi@freebsd.org
Received: from arjun.niksun.com (gwnew.niksun.com [63.148.27.34])
	by hub.freebsd.org (Postfix) with ESMTP
	id 8DC1437B424; Wed,  2 May 2001 06:01:30 -0700 (PDT)
	(envelope-from ath@niksun.com)
Received: from stiegl.niksun.com (stiegl.niksun.com [10.0.0.44])
	by arjun.niksun.com (8.9.3/8.9.3) with ESMTP id JAA17799;
	Wed, 2 May 2001 09:01:29 -0400 (EDT)
	(envelope-from ath@stiegl.niksun.com)
Received: (from ath@localhost)
	by stiegl.niksun.com (8.9.2/8.8.7) id JAA14775;
	Wed, 2 May 2001 09:01:29 -0400 (EDT)
	(envelope-from ath)
To: msmith@FreeBSD.ORG
Cc: freebsd-scsi@FreeBSD.ORG
Subject: Re: AMI MegaRAID on 4.2-RELEASE and bus_dmamap_load errors
References: <200105020234.WAA11069@stiegl.niksun.com>
From: Andrew Heybey <ath@niksun.com>
Date: 02 May 2001 09:01:29 -0400
In-Reply-To: Andrew Heybey's message of "Tue, 01 May 2001 22:34:41 -0400"
Message-ID: <85y9sf7vme.fsf@stiegl.niksun.com>
Lines: 39
X-Mailer: Gnus v5.5/XEmacs 20.4 - "Emerald"
Sender: owner-freebsd-scsi@FreeBSD.ORG
Precedence: bulk
X-Loop: FreeBSD.org

To follow up on my own email:

If I change rawio to page-align the buffer it uses for I/O, then the error
messages go away.

I also note that (for example), dev/aic7xxx_frebbsd.h has:

/*
 * The number of dma segments supported.  The sequencer can handle any number
 * of physically contiguous S/G entrys.  To reduce the driver's memory
 * consumption, we limit the number supported to be sufficient to handle
 * the largest mapping supported by the kernel, MAXPHYS.  Assuming the
 * transfer is as fragmented as possible and unaligned, this turns out to
 * be the number of paged sized transfers in MAXPHYS plus an extra element
 * to handle any unaligned residual.  The sequencer fetches SG elements
 * in 128 byte chucks, so make the number per-transaction a nice multiple
 * of 16 (8 byte S/G elements).
 */
/* XXX Worth the space??? */
#define AHC_NSEG (roundup(btoc(MAXPHYS) + 1, 16))

While dev/amr/amrreg.h has:

/*
 * We could actually use all 17 segments, but using only 16 means that
 * each scatter/gather map is 128 bytes in size, and thus we don't have to worry about
 * maps crossing page boundaries.
 *
 * The AMI documentation says that the limit is 26.  Unfortunately, there's no way to
 * cleanly fit more than 16 entries in without a page boundary.  But is this a concern,
 * since we allocate the s/g maps contiguously anyway?
 */
#define AMR_NSEG		16


I did not feel comfortable changing AMR_NSEG without fully
understanding the problem with crossing a page boundary.

andrew

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-scsi" in the body of the message