From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ingo Molnar Subject: Re: [RFC][PATCH] kmap_atomic_push Date: Thu, 8 Oct 2009 17:53:44 +0200 Message-ID: <20091008155344.GA11727@elte.hu> References: <1255016123.17055.17.camel@laptop> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <1255016123.17055.17.camel@laptop> Sender: linux-kernel-owner@vger.kernel.org To: Peter Zijlstra , Avi Kivity Cc: Linus Torvalds , Andrew Morton , "hugh.dickins" , David Howells , lkml , linux-arch List-Id: linux-arch.vger.kernel.org * Peter Zijlstra wrote: > The below patchlet changes the kmap_atomic interface to a stack based > one that doesn't require the KM_types anymore. > > This significantly simplifies some code (more still than are present > in this patch -- ie. pte_map_nested can go now) > > This obviously requires that push and pop are matched, I fixed a few > cases that were not properly nested, the (x86) code checks for this > and will go BUG when trying to pop a vaddr that isn't the top one so > abusers should be rather visible. Looks great IMO! Last i proposed this i think either Andrew or Avi had second thoughts about the hard-to-calculate worst-case mapping limit - but i dont think that's a big issue. Lets not change the API names though - the rule is that map/unmap must be properly nested. Ingo From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mx2.mail.elte.hu ([157.181.151.9]:42502 "EHLO mx2.mail.elte.hu" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751051AbZJHPy4 (ORCPT ); Thu, 8 Oct 2009 11:54:56 -0400 Date: Thu, 8 Oct 2009 17:53:44 +0200 From: Ingo Molnar Subject: Re: [RFC][PATCH] kmap_atomic_push Message-ID: <20091008155344.GA11727@elte.hu> References: <1255016123.17055.17.camel@laptop> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1255016123.17055.17.camel@laptop> Sender: linux-arch-owner@vger.kernel.org List-ID: To: Peter Zijlstra , Avi Kivity Cc: Linus Torvalds , Andrew Morton , "hugh.dickins" , David Howells , lkml , linux-arch Message-ID: <20091008155344.PSNf_zlXWczZ9ErwVw9vU2yWltYjMoSODjt0NKL70Dc@z> * Peter Zijlstra wrote: > The below patchlet changes the kmap_atomic interface to a stack based > one that doesn't require the KM_types anymore. > > This significantly simplifies some code (more still than are present > in this patch -- ie. pte_map_nested can go now) > > This obviously requires that push and pop are matched, I fixed a few > cases that were not properly nested, the (x86) code checks for this > and will go BUG when trying to pop a vaddr that isn't the top one so > abusers should be rather visible. Looks great IMO! Last i proposed this i think either Andrew or Avi had second thoughts about the hard-to-calculate worst-case mapping limit - but i dont think that's a big issue. Lets not change the API names though - the rule is that map/unmap must be properly nested. Ingo