Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1760530AbZFOKbx (ORCPT ); Mon, 15 Jun 2009 06:31:53 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1756742AbZFOKbp (ORCPT ); Mon, 15 Jun 2009 06:31:45 -0400 Received: from cantor.suse.de ([195.135.220.2]:57809 "EHLO mx1.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755636AbZFOKbo (ORCPT ); Mon, 15 Jun 2009 06:31:44 -0400 Date: Mon, 15 Jun 2009 12:31:45 +0200 From: Nick Piggin To: Pekka Enberg Cc: Heiko Carstens , torvalds@linux-foundation.org, linux-kernel@vger.kernel.org, akpm@linux-foundation.org, cl@linux-foundation.org, kamezawa.hiroyu@jp.fujitsu.com, lizf@cn.fujitsu.com, mingo@elte.hu, yinghai@kernel.org, benh@kernel.crashing.org Subject: Re: [GIT PULL v2] Early SLAB fixes for 2.6.31 Message-ID: <20090615103145.GB20461@wotan.suse.de> References: <20090615081831.GA5411@osiris.boeblingen.de.ibm.com> <20090615082636.GA30964@wotan.suse.de> <1245054779.23207.6.camel@penberg-laptop> <20090615085209.GB30964@wotan.suse.de> <1245056919.23207.7.camel@penberg-laptop> <20090615102004.GB5411@osiris.boeblingen.de.ibm.com> <1245061304.23207.21.camel@penberg-laptop> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1245061304.23207.21.camel@penberg-laptop> User-Agent: Mutt/1.5.9i Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 4242 Lines: 121 On Mon, Jun 15, 2009 at 01:21:44PM +0300, Pekka Enberg wrote: > On Mon, 2009-06-15 at 12:20 +0200, Heiko Carstens wrote: > > > > --- > > > > mm/slub.c | 17 ++++++++++++++--- > > > > 1 file changed, 14 insertions(+), 3 deletions(-) > > > > > > > > Index: linux-2.6/mm/slub.c > > > > =================================================================== > > > > --- linux-2.6.orig/mm/slub.c > > > > +++ linux-2.6/mm/slub.c > > > > @@ -2596,6 +2596,7 @@ static noinline struct kmem_cache *dma_k > > > > struct kmem_cache *s; > > > > char *text; > > > > size_t realsize; > > > > + unsigned long slabflags; > > > > > > > > s = kmalloc_caches_dma[index]; > > > > if (s) > > > > @@ -2617,9 +2618,18 @@ static noinline struct kmem_cache *dma_k > > > > (unsigned int)realsize); > > > > s = kmalloc(kmem_size, flags & ~SLUB_DMA); > > > > > > > > + /* > > > > + * Must defer sysfs creation to a workqueue because we don't know > > > > + * what context we are called from. Before sysfs comes up, we don't > > > > + * need to do anything because slab_sysfs_init will start by > > > > + * adding all existing slabs to sysfs. > > > > + */ > > > > + slabflags = SLAB_CACHE_DMA; > > > > + if (slab_state >= SYSFS) > > > > + slabflags |= __SYSFS_ADD_DEFERRED; > > > > + > > > > if (!s || !text || !kmem_cache_open(s, flags, text, > > > > - realsize, ARCH_KMALLOC_MINALIGN, > > > > - SLAB_CACHE_DMA|__SYSFS_ADD_DEFERRED, NULL)) { > > > > + realsize, ARCH_KMALLOC_MINALIGN, slabflags, NULL)) { > > > > kfree(s); > > > > kfree(text); > > > > goto unlock_out; > > > > @@ -2628,7 +2638,8 @@ static noinline struct kmem_cache *dma_k > > > > list_add(&s->list, &slab_caches); > > > > kmalloc_caches_dma[index] = s; > > > > > > > > - schedule_work(&sysfs_add_work); > > > > + if (slab_state >= SYSFS) > > > > + schedule_work(&sysfs_add_work); > > > > > > > > unlock_out: > > > > up_write(&slub_lock); > > > > > > Looks good to me. Heiko, does it fix your case? > > > > Yes, works fine. Thanks! > > > > Tested-by: Heiko Carstens > > Thanks! Nick, care to send a patch with your sign-off? Does SLQB need > something like this too (I didn't check)? No, it does not defer DMA slab creation. -- Recent change to use slab allocations earlier exposed a bug where SLUB can call schedule_work and try to call sysfs before it is safe to do so. Reported-by: Heiko Carstens Tested-by: Heiko Carstens Signed-off-by: Nick Piggin --- mm/slub.c | 17 ++++++++++++++--- 1 file changed, 14 insertions(+), 3 deletions(-) Index: linux-2.6/mm/slub.c =================================================================== --- linux-2.6.orig/mm/slub.c +++ linux-2.6/mm/slub.c @@ -2596,6 +2596,7 @@ static noinline struct kmem_cache *dma_k struct kmem_cache *s; char *text; size_t realsize; + unsigned long slabflags; s = kmalloc_caches_dma[index]; if (s) @@ -2617,9 +2618,18 @@ static noinline struct kmem_cache *dma_k (unsigned int)realsize); s = kmalloc(kmem_size, flags & ~SLUB_DMA); + /* + * Must defer sysfs creation to a workqueue because we don't know + * what context we are called from. Before sysfs comes up, we don't + * need to do anything because our sysfs initcall will start by + * adding all existing slabs to sysfs. + */ + slabflags = SLAB_CACHE_DMA; + if (slab_state >= SYSFS) + slabflags |= __SYSFS_ADD_DEFERRED; + if (!s || !text || !kmem_cache_open(s, flags, text, - realsize, ARCH_KMALLOC_MINALIGN, - SLAB_CACHE_DMA|__SYSFS_ADD_DEFERRED, NULL)) { + realsize, ARCH_KMALLOC_MINALIGN, slabflags, NULL)) { kfree(s); kfree(text); goto unlock_out; @@ -2628,7 +2638,8 @@ static noinline struct kmem_cache *dma_k list_add(&s->list, &slab_caches); kmalloc_caches_dma[index] = s; - schedule_work(&sysfs_add_work); + if (slab_state >= SYSFS) + schedule_work(&sysfs_add_work); unlock_out: up_write(&slub_lock); -- 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/