From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jeff Garzik Subject: Re: [PATCH] libata-core More robust parsing for multi_count(v5) Date: Sun, 12 Apr 2009 11:31:16 -0400 Message-ID: <49E20944.7070709@pobox.com> References: <49C1047D.4000008@rtr.ca> <49C11A0C.3070502@rtr.ca> <49C11ED4.2030307@rtr.ca> <49C19087.5000307@gmail.com> <49C190F1.7010202@kernel.org> <49C28133.8050401@rtr.ca> <49C281A5.5080404@rtr.ca> <49C2D642.90706@kernel.org> <49C39690.6090802@rtr.ca> <49E20475.1080409@rtr.ca> <20090412161812.78e92948@lxorguk.ukuu.org.uk> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: Received: from srv5.dvmed.net ([207.36.208.214]:42372 "EHLO mail.dvmed.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754115AbZDLPb1 (ORCPT ); Sun, 12 Apr 2009 11:31:27 -0400 In-Reply-To: <20090412161812.78e92948@lxorguk.ukuu.org.uk> Sender: linux-ide-owner@vger.kernel.org List-Id: linux-ide@vger.kernel.org To: Alan Cox , Mark Lord Cc: Tejun Heo , IDE/ATA development list Alan Cox wrote: >> The limit for sata_mv chipsets seems to actually be 7-sectors or less >> for read/write multiple. Which means a max of 4 in practice. >> >> Similarly, we should also be preventing *any* PIO of more than one DRQ >> for sata_mv. But I don't see a sensible way to do that either. > > Ugghhh.. Our infrastructure just flat doesn't deal well with per-xfer-mode sector count limitations. It is quite easy for us to limit per-xfer sector count on a __per device__ basis -- both SCSI and block layer handle that well. But deviating from that implies potentially having to re-merge (or split) block layer requests -- IOW quickly moves into the non-trivial, cross-stack category of problem. Problem is, of course, sector count limits in ATA can be per-command (or per-xfer-mode), not just per-device. Jeff