From mboxrd@z Thu Jan 1 00:00:00 1970 From: Dave Jones Subject: Re: [PATCH] Poll-based IDE driver Date: Wed, 8 Oct 2003 13:15:23 +0100 Sender: linux-kernel-owner@vger.kernel.org Message-ID: <20031008121523.GH705@redhat.com> References: <20030917144120.A11425@in.ibm.com> <1063806900.12279.47.camel@dhcp23.swansea.linux.org.uk> <20031008151357.A31976@in.ibm.com> <20031008115051.GD705@redhat.com> <20031008174458.A32517@in.ibm.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <20031008174458.A32517@in.ibm.com> To: Srivatsa Vaddagiri Cc: Alan Cox , lkcd-devel@lists.sourceforge.net, Linux Kernel Mailing List , linux-ide@vger.kernel.org List-Id: linux-ide@vger.kernel.org On Wed, Oct 08, 2003 at 05:44:58PM +0530, Srivatsa Vaddagiri wrote: > > Why not just use udelay() ? The above code cannot possibly do > > the right thing on all processors. > > Since my code is supposed to run when system is crashing, I would like > to avoid calling any function in the kernel as far as possible, since > the kernel and its data structures may be in a inconsistent state > and/or corrupted. By the same principle, your dump_udelay() is just as likely to get stomped on as the kernels udelay() function. > I do realize that the above code does not provide accurate > delay and may not work on all platforms. In that direction > I was considering using the loops_per_jiffy variable > which may provide more accurate/platform-independent delay (?) .. If you must reinvent a wheel, use a round one. Copy udelay(). Dave -- Dave Jones http://www.codemonkey.org.uk