From mboxrd@z Thu Jan 1 00:00:00 1970 From: Andi Kleen Subject: Re: [PATCH 5/7] Use %gs for per-cpu sections in kernel Date: Sat, 23 Sep 2006 01:52:43 +0200 Message-ID: <200609230152.43713.ak@muc.de> References: <1158925861.26261.3.camel@localhost.localdomain> <20060922123215.GA98728@muc.de> <45146725.4070109@goop.org> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <45146725.4070109@goop.org> Content-Disposition: inline Sender: linux-kernel-owner@vger.kernel.org To: virtualization@lists.osdl.org Cc: Jeremy Fitzhardinge , lkml - Kernel Mailing List List-Id: virtualization@lists.linuxfoundation.org On Saturday 23 September 2006 00:43, Jeremy Fitzhardinge wrote: > Andi Kleen wrote: > > BTW I changed my copy sorry. I redid the early PDA support > > to not be in assembler. > > I went to the trouble of making the PDA completely set up before any C > code ran. Yes, but your patch never applied to anything even remotely looking like the code in my tree. I got so frustrated that I ended up reimplementing it in a cleaner way. Now head.S calls i386_start_kernel() and that calls pda_init() without any additional assembly code or other special cases etc. This is very similar to how x86-64 works. > which means that that they > have to work from the first function prologue. I mainly did it to fix lockdep. I used to do mcount hacks myself, but you typically need a few special annotations for those anyways so I am not too concerned about them. > It also simplifies things to get all that set up ASAP so there's no > bootstrap dependency problem. Yes no argument on that. -Andi