From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S934698Ab1JENMU (ORCPT ); Wed, 5 Oct 2011 09:12:20 -0400 Received: from merlin.infradead.org ([205.233.59.134]:54171 "EHLO merlin.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S934332Ab1JENMT convert rfc822-to-8bit (ORCPT ); Wed, 5 Oct 2011 09:12:19 -0400 Subject: Re: [PATCH] mm, arch: Complete pagefault_disable abstraction From: Peter Zijlstra To: Andrew Morton Cc: "linux-mm@kvack.org" , linux-arch , linux-kernel , Thomas Gleixner Date: Wed, 05 Oct 2011 15:12:14 +0200 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 8BIT X-Mailer: Evolution 3.0.3- Message-ID: <1317820334.6766.23.camel@twins> Mime-Version: 1.0 X-Bad-Reply: 'Re:' in Subject but no References or In-Reply-To headers Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org For those with a strong stomach, here's what we do with kmap_atomic: --- Subject: mm, rt: kmap_atomic scheduling From: Peter Zijlstra Date: Thu, 28 Jul 2011 10:43:51 +0200 In fact, with migrate_disable() existing one could play games with kmap_atomic. You could save/restore the kmap_atomic slots on context switch (if there are any in use of course), this should be esp easy now that we have a kmap_atomic stack. Signed-off-by: Peter Zijlstra [dvhart@linux.intel.com: build fix] Link: http://lkml.kernel.org/r/1311842631.5890.208.camel@twins --- arch/x86/kernel/process_32.c | 36 ++++++++++++++++++++++++++++++++++++ include/linux/sched.h | 5 +++++ mm/memory.c | 2 ++ 3 files changed, 43 insertions(+) Index: linux-2.6/arch/x86/kernel/process_32.c =================================================================== --- linux-2.6.orig/arch/x86/kernel/process_32.c +++ linux-2.6/arch/x86/kernel/process_32.c @@ -38,6 +38,7 @@ #include #include #include +#include #include #include @@ -346,6 +347,41 @@ __switch_to(struct task_struct *prev_p, task_thread_info(next_p)->flags & _TIF_WORK_CTXSW_NEXT)) __switch_to_xtra(prev_p, next_p, tss); +#if defined CONFIG_PREEMPT_RT_FULL && defined CONFIG_HIGHMEM + /* + * Save @prev's kmap_atomic stack + */ + prev_p->kmap_idx = __this_cpu_read(__kmap_atomic_idx); + if (unlikely(prev_p->kmap_idx)) { + int i; + + for (i = 0; i < prev_p->kmap_idx; i++) { + int idx = i + KM_TYPE_NR * smp_processor_id(); + + pte_t *ptep = kmap_pte - idx; + prev_p->kmap_pte[i] = *ptep; + kpte_clear_flush(ptep, __fix_to_virt(FIX_KMAP_BEGIN + idx)); + } + + __this_cpu_write(__kmap_atomic_idx, 0); + } + + /* + * Restore @next_p's kmap_atomic stack + */ + if (unlikely(next_p->kmap_idx)) { + int i; + + __this_cpu_write(__kmap_atomic_idx, next_p->kmap_idx); + + for (i = 0; i < next_p->kmap_idx; i++) { + int idx = i + KM_TYPE_NR * smp_processor_id(); + + set_pte(kmap_pte - idx, next_p->kmap_pte[i]); + } + } +#endif + /* If we're going to preload the fpu context, make sure clts is run while we're batching the cpu state updates. */ if (preload_fpu) Index: linux-2.6/include/linux/sched.h =================================================================== --- linux-2.6.orig/include/linux/sched.h +++ linux-2.6/include/linux/sched.h @@ -63,6 +63,7 @@ struct sched_param { #include #include +#include #include #include #include @@ -1594,6 +1595,10 @@ struct task_struct { struct rcu_head put_rcu; int softirq_nestcnt; #endif +#if defined CONFIG_PREEMPT_RT_FULL && defined CONFIG_HIGHMEM + int kmap_idx; + pte_t kmap_pte[KM_TYPE_NR]; +#endif }; #ifdef CONFIG_PREEMPT_RT_FULL Index: linux-2.6/mm/memory.c =================================================================== --- linux-2.6.orig/mm/memory.c +++ linux-2.6/mm/memory.c @@ -3431,6 +3431,7 @@ unlock: #ifdef CONFIG_PREEMPT_RT_FULL void pagefault_disable(void) { + migrate_disable(); current->pagefault_disabled++; /* * make sure to have issued the store before a pagefault @@ -3448,6 +3449,7 @@ void pagefault_enable(void) */ barrier(); current->pagefault_disabled--; + migrate_enable(); } EXPORT_SYMBOL_GPL(pagefault_enable); #endif