From mboxrd@z Thu Jan 1 00:00:00 1970 From: Pat LaVarre Subject: Re: aligned /dev/scd$n reads less rare how Date: 24 Oct 2003 08:41:16 -0600 Sender: linux-scsi-owner@vger.kernel.org Message-ID: <1067006475.16062.10.camel@patehci2> References: <1064271322.4797.9.camel@patehci2> <1064847281.24854.2.camel@patehci2> <1064850600.25057.7.camel@patehci2> <1064858121.4262.0.camel@patehci2> Mime-Version: 1.0 Content-Type: text/plain Content-Transfer-Encoding: 7bit Return-path: Received: from email-out1.iomega.com ([147.178.1.82]:45253 "EHLO email.iomega.com") by vger.kernel.org with ESMTP id S262228AbTJXOli (ORCPT ); Fri, 24 Oct 2003 10:41:38 -0400 Received: from royntex01.iomegacorp.com (unknown [147.178.90.120]) by email.iomega.com (Postfix) with ESMTP id 5522920E8 for ; Fri, 24 Oct 2003 08:41:37 -0600 (MDT) In-Reply-To: <1064858121.4262.0.camel@patehci2> List-Id: linux-scsi@vger.kernel.org To: linux-scsi@vger.kernel.org > Anything easy I can do > to make aligned reads more common? Instead we first could/ should ask: How does Linux align write/read of cd-rw/dvd+rw? We know 2.6 vs. 2.4 newly connects cd-rw as writeable, by http://kniggit.net/wwol26.html of l=kernelnewbies. And "everybody knows" cd-rw needs 64 KiB write alignment, dvd+rw needs 32 KiB write alignment, and both prefer write alignment in reads, right? How then does Linux hint at the mounted fs to align write/read to something other than the 0.5 KiB that was optimal for legacy hdd? > I ask because often thru /dev/scd$n > I see seemingly pointless misalignment, > ... > $ sudo blockdev --flushbufs /dev/scd0 > $ time sudo dd if=/dev/scd0 bs=1M > ... > usb-storage: x 28 00 00 00 00 00 00 00 20 00 Initially we see aligned in address and length to x20 blocks = 64 KiB. > usb-storage: x 28 00 00 00 00 20 00 00 24 00 Soon we see aligned in address but, whoops, not in length. > usb-storage: x 28 00 00 00 00 44 00 00 40 00 > usb-storage: x 28 00 00 00 00 84 00 00 40 00 > usb-storage: x 28 00 00 00 00 C4 00 00 40 00 > ... > usb-storage: x 28 00 00 07 FE C4 00 00 40 00 Mostly we see contiguous after a misaligned access, so not aligned in address though yes aligned in length. Ouch. Pat LaVarre