Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1758967Ab2JZSww (ORCPT ); Fri, 26 Oct 2012 14:52:52 -0400 Received: from www.linutronix.de ([62.245.132.108]:35690 "EHLO Galois.linutronix.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754026Ab2JZSwv (ORCPT ); Fri, 26 Oct 2012 14:52:51 -0400 Date: Fri, 26 Oct 2012 20:52:48 +0200 (CEST) From: Thomas Gleixner To: LKML cc: linux-rt-users , Christoph Lameter Subject: [ANNOUNCE] 3.6.3-rt7 Message-ID: User-Agent: Alpine 2.02 (LFD 1266 2009-07-14) MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII X-Linutronix-Spam-Score: -1.0 X-Linutronix-Spam-Level: - X-Linutronix-Spam-Status: No , -1.0 points, 5.0 required, ALL_TRUSTED=-1,SHORTCIRCUIT=-0.0001 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 5731 Lines: 193 Dear RT Folks, I'm pleased to announce the 3.6.3-rt7 release. Changes since 3.6.3-rt6: * Enable SLUB for RT Last time I looked at SLUB for RT (some years ago) it was just way more painful than dealing with SLAB, but Christoph Lameter has done major surgery on the SLUB code since then and it turns out that making SLUB usable for RT has become very simple. Thanks Christoph! slab.c: 172 insertions(+), 58 deletions(-) slub.c: 17 insertions(+), 13 deletions(-) I did some quick comparisons and even a simple hackbench run shows a significant speedup with SLUB vs. SLAB on RT. I'm not too surprised as SLUBs fastpath does not have the RT induced contention problems which we can observe with SLAB. As usual, give it a good testing and report whatever explodes :) The delta patch against 3.6.3-rt6 is appended below and can be found here: http://www.kernel.org/pub/linux/kernel/projects/rt/3.6/incr/patch-3.6.3-rt6-rt7.patch.xz The RT patch against 3.6.3 can be found here: http://www.kernel.org/pub/linux/kernel/projects/rt/3.6/patch-3.6.3-rt7.patch.xz The split quilt queue is available at: http://www.kernel.org/pub/linux/kernel/projects/rt/3.6/patches-3.6.3-rt7.tar.xz Enjoy, tglx ----------> Index: linux-stable/init/Kconfig =================================================================== --- linux-stable.orig/init/Kconfig +++ linux-stable/init/Kconfig @@ -1442,7 +1442,6 @@ config SLAB config SLUB bool "SLUB (Unqueued Allocator)" - depends on !PREEMPT_RT_FULL help SLUB is a slab allocator that minimizes cache line usage instead of managing queues of cached objects (SLAB approach). Index: linux-stable/localversion-rt =================================================================== --- linux-stable.orig/localversion-rt +++ linux-stable/localversion-rt @@ -1 +1 @@ --rt6 +-rt7 Index: linux-stable/mm/slub.c =================================================================== --- linux-stable.orig/mm/slub.c +++ linux-stable/mm/slub.c @@ -31,6 +31,7 @@ #include #include #include +#include #include @@ -225,6 +226,8 @@ static inline void stat(const struct kme #endif } +static DEFINE_LOCAL_IRQ_LOCK(slub_lock); + /******************************************************************** * Core slab cache functions *******************************************************************/ @@ -1278,7 +1281,7 @@ static struct page *allocate_slab(struct flags &= gfp_allowed_mask; if (flags & __GFP_WAIT) - local_irq_enable(); + local_unlock_irq(slub_lock); flags |= s->allocflags; @@ -1318,7 +1321,7 @@ static struct page *allocate_slab(struct } if (flags & __GFP_WAIT) - local_irq_disable(); + local_lock_irq(slub_lock); if (!page) return NULL; @@ -1959,9 +1962,9 @@ int put_cpu_partial(struct kmem_cache *s * partial array is full. Move the existing * set to the per node partial list. */ - local_irq_save(flags); + local_lock_irqsave(slub_lock, flags); unfreeze_partials(s); - local_irq_restore(flags); + local_unlock_irqrestore(slub_lock, flags); pobjects = 0; pages = 0; stat(s, CPU_PARTIAL_DRAIN); @@ -2201,7 +2204,7 @@ static void *__slab_alloc(struct kmem_ca struct page *page; unsigned long flags; - local_irq_save(flags); + local_lock_irqsave(slub_lock, flags); #ifdef CONFIG_PREEMPT /* * We may have been preempted and rescheduled on a different @@ -2262,7 +2265,7 @@ load_freelist: VM_BUG_ON(!c->page->frozen); c->freelist = get_freepointer(s, freelist); c->tid = next_tid(c->tid); - local_irq_restore(flags); + local_unlock_irqrestore(slub_lock, flags); return freelist; new_slab: @@ -2281,7 +2284,7 @@ new_slab: if (!(gfpflags & __GFP_NOWARN) && printk_ratelimit()) slab_out_of_memory(s, gfpflags, node); - local_irq_restore(flags); + local_unlock_irqrestore(slub_lock, flags); return NULL; } @@ -2296,7 +2299,7 @@ new_slab: deactivate_slab(s, page, get_freepointer(s, freelist)); c->page = NULL; c->freelist = NULL; - local_irq_restore(flags); + local_unlock_irqrestore(slub_lock, flags); return freelist; } @@ -2488,7 +2491,8 @@ static void __slab_free(struct kmem_cach * Otherwise the list_lock will synchronize with * other processors updating the list of slabs. */ - spin_lock_irqsave(&n->list_lock, flags); + local_spin_lock_irqsave(slub_lock, + &n->list_lock, flags); } } @@ -2538,7 +2542,7 @@ static void __slab_free(struct kmem_cach stat(s, FREE_ADD_PARTIAL); } } - spin_unlock_irqrestore(&n->list_lock, flags); + local_spin_unlock_irqrestore(slub_lock, &n->list_lock, flags); return; slab_empty: @@ -2552,7 +2556,7 @@ slab_empty: /* Slab must be on the full list */ remove_full(s, page); - spin_unlock_irqrestore(&n->list_lock, flags); + local_spin_unlock_irqrestore(slub_lock, &n->list_lock, flags); stat(s, FREE_SLAB); discard_slab(s, page); } @@ -4002,9 +4006,9 @@ static int __cpuinit slab_cpuup_callback case CPU_DEAD_FROZEN: mutex_lock(&slab_mutex); list_for_each_entry(s, &slab_caches, list) { - local_irq_save(flags); + local_lock_irqsave(slub_lock, flags); __flush_cpu_slab(s, cpu); - local_irq_restore(flags); + local_unlock_irqrestore(slub_lock, flags); } mutex_unlock(&slab_mutex); break; -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/