From: Christoph Lameter <cl@linux-foundation.org> To: David Rientjes <rientjes@google.com> Cc: Pekka Enberg <penberg@cs.helsinki.fi>, Andi Kleen <andi@firstfloor.org>, Nick Piggin <npiggin@suse.de>, linux-kernel@vger.kernel.org, linux-mm@kvack.org, haicheng.li@intel.com, KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Subject: Re: [PATCH] [4/4] SLAB: Fix node add timer race in cache_reap Date: Thu, 25 Feb 2010 16:31:00 -0600 (CST) [thread overview] Message-ID: <alpine.DEB.2.00.1002251627040.18861@router.home> (raw) In-Reply-To: <alpine.DEB.2.00.1002251315010.3501@chino.kir.corp.google.com> On Thu, 25 Feb 2010, David Rientjes wrote: > On Thu, 25 Feb 2010, Christoph Lameter wrote: > > > > I don't see how memory hotadd with a new node being onlined could have > > > worked fine before since slab lacked any memory hotplug notifier until > > > Andi just added it. > > > > AFAICR The cpu notifier took on that role in the past. > > > > The cpu notifier isn't involved if the firmware notifies the kernel that a > new ACPI memory device has been added or you write a start address to > /sys/devices/system/memory/probe. Hot-added memory devices can include > ACPI_SRAT_MEM_HOT_PLUGGABLE entries in the SRAT for x86 that assign them > non-online node ids (although all such entries get their bits set in > node_possible_map at boot), so a new pgdat may be allocated for the node's > registered range. Yes Andi's work makes it explicit but there is already code in the cpu notifier (see cpuup_prepare) that seems to have been intended to initialize the node structures. Wonder why the hotplug people never addressed that issue? Kame? list_for_each_entry(cachep, &cache_chain, next) { /* * Set up the size64 kmemlist for cpu before we can * begin anything. Make sure some other cpu on this * node has not already allocated this */ if (!cachep->nodelists[node]) { l3 = kmalloc_node(memsize, GFP_KERNEL, node); if (!l3) goto bad; kmem_list3_init(l3); l3->next_reap = jiffies + REAPTIMEOUT_LIST3 + ((unsigned long)cachep) % REAPTIMEOUT_LIST3; /* * The l3s don't come and go as CPUs come and * go. cache_chain_mutex is sufficient * protection here. */ cachep->nodelists[node] = l3; } spin_lock_irq(&cachep->nodelists[node]->list_lock); cachep->nodelists[node]->free_limit = (1 + nr_cpus_node(node)) * cachep->batchcount + cachep->num; spin_unlock_irq(&cachep->nodelists[node]->list_lock); } > kmalloc_node() in generic kernel code. All that is done under > MEM_GOING_ONLINE and not MEM_ONLINE, which is why I suggest the first and > fourth patch in this series may not be necessary if we prevent setting the > bit in the nodemask or building the zonelists until the slab nodelists are > ready. That sounds good.
WARNING: multiple messages have this Message-ID (diff)
From: Christoph Lameter <cl@linux-foundation.org> To: David Rientjes <rientjes@google.com> Cc: Pekka Enberg <penberg@cs.helsinki.fi>, Andi Kleen <andi@firstfloor.org>, Nick Piggin <npiggin@suse.de>, linux-kernel@vger.kernel.org, linux-mm@kvack.org, haicheng.li@intel.com, KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Subject: Re: [PATCH] [4/4] SLAB: Fix node add timer race in cache_reap Date: Thu, 25 Feb 2010 16:31:00 -0600 (CST) [thread overview] Message-ID: <alpine.DEB.2.00.1002251627040.18861@router.home> (raw) In-Reply-To: <alpine.DEB.2.00.1002251315010.3501@chino.kir.corp.google.com> On Thu, 25 Feb 2010, David Rientjes wrote: > On Thu, 25 Feb 2010, Christoph Lameter wrote: > > > > I don't see how memory hotadd with a new node being onlined could have > > > worked fine before since slab lacked any memory hotplug notifier until > > > Andi just added it. > > > > AFAICR The cpu notifier took on that role in the past. > > > > The cpu notifier isn't involved if the firmware notifies the kernel that a > new ACPI memory device has been added or you write a start address to > /sys/devices/system/memory/probe. Hot-added memory devices can include > ACPI_SRAT_MEM_HOT_PLUGGABLE entries in the SRAT for x86 that assign them > non-online node ids (although all such entries get their bits set in > node_possible_map at boot), so a new pgdat may be allocated for the node's > registered range. Yes Andi's work makes it explicit but there is already code in the cpu notifier (see cpuup_prepare) that seems to have been intended to initialize the node structures. Wonder why the hotplug people never addressed that issue? Kame? list_for_each_entry(cachep, &cache_chain, next) { /* * Set up the size64 kmemlist for cpu before we can * begin anything. Make sure some other cpu on this * node has not already allocated this */ if (!cachep->nodelists[node]) { l3 = kmalloc_node(memsize, GFP_KERNEL, node); if (!l3) goto bad; kmem_list3_init(l3); l3->next_reap = jiffies + REAPTIMEOUT_LIST3 + ((unsigned long)cachep) % REAPTIMEOUT_LIST3; /* * The l3s don't come and go as CPUs come and * go. cache_chain_mutex is sufficient * protection here. */ cachep->nodelists[node] = l3; } spin_lock_irq(&cachep->nodelists[node]->list_lock); cachep->nodelists[node]->free_limit = (1 + nr_cpus_node(node)) * cachep->batchcount + cachep->num; spin_unlock_irq(&cachep->nodelists[node]->list_lock); } > kmalloc_node() in generic kernel code. All that is done under > MEM_GOING_ONLINE and not MEM_ONLINE, which is why I suggest the first and > fourth patch in this series may not be necessary if we prevent setting the > bit in the nodemask or building the zonelists until the slab nodelists are > ready. That sounds good. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
next prev parent reply other threads:[~2010-02-25 22:31 UTC|newest] Thread overview: 174+ messages / expand[flat|nested] mbox.gz Atom feed top 2010-02-11 20:53 [PATCH] [0/4] Update slab memory hotplug series Andi Kleen 2010-02-11 20:53 ` Andi Kleen 2010-02-11 20:54 ` [PATCH] [1/4] SLAB: Handle node-not-up case in fallback_alloc() v2 Andi Kleen 2010-02-11 20:54 ` Andi Kleen 2010-02-11 21:41 ` David Rientjes 2010-02-11 21:41 ` David Rientjes 2010-02-11 21:55 ` Andi Kleen 2010-02-11 21:55 ` Andi Kleen 2010-02-15 6:04 ` Nick Piggin 2010-02-15 6:04 ` Nick Piggin 2010-02-15 10:07 ` Andi Kleen 2010-02-15 10:07 ` Andi Kleen 2010-02-15 10:22 ` Nick Piggin 2010-02-15 10:22 ` Nick Piggin 2010-02-11 20:54 ` [PATCH] [2/4] SLAB: Separate node initialization into separate function Andi Kleen 2010-02-11 20:54 ` Andi Kleen 2010-02-11 21:44 ` David Rientjes 2010-02-11 21:44 ` David Rientjes 2010-02-11 20:54 ` [PATCH] [3/4] SLAB: Set up the l3 lists for the memory of freshly added memory v2 Andi Kleen 2010-02-11 20:54 ` Andi Kleen 2010-02-11 21:45 ` David Rientjes 2010-02-11 21:45 ` David Rientjes 2010-02-15 6:06 ` Nick Piggin 2010-02-15 6:06 ` Nick Piggin 2010-02-15 21:47 ` David Rientjes 2010-02-15 21:47 ` David Rientjes 2010-02-16 14:04 ` Nick Piggin 2010-02-16 14:04 ` Nick Piggin 2010-02-16 20:45 ` Pekka Enberg 2010-02-16 20:45 ` Pekka Enberg 2010-02-11 20:54 ` [PATCH] [4/4] SLAB: Fix node add timer race in cache_reap Andi Kleen 2010-02-11 20:54 ` Andi Kleen 2010-02-11 21:45 ` David Rientjes 2010-02-11 21:45 ` David Rientjes 2010-02-15 6:15 ` Nick Piggin 2010-02-15 6:15 ` Nick Piggin 2010-02-15 10:32 ` Andi Kleen 2010-02-15 10:32 ` Andi Kleen 2010-02-15 10:41 ` Nick Piggin 2010-02-15 10:41 ` Nick Piggin 2010-02-15 10:52 ` Andi Kleen 2010-02-15 10:52 ` Andi Kleen 2010-02-15 11:01 ` Nick Piggin 2010-02-15 11:01 ` Nick Piggin 2010-02-15 15:30 ` Andi Kleen 2010-02-15 15:30 ` Andi Kleen 2010-02-19 18:22 ` Christoph Lameter 2010-02-19 18:22 ` Christoph Lameter 2010-02-20 9:01 ` Andi Kleen 2010-02-20 9:01 ` Andi Kleen 2010-02-22 10:53 ` Pekka Enberg 2010-02-22 10:53 ` Pekka Enberg 2010-02-22 14:31 ` Andi Kleen 2010-02-22 14:31 ` Andi Kleen 2010-02-22 16:11 ` Pekka Enberg 2010-02-22 16:11 ` Pekka Enberg 2010-02-22 20:20 ` Andi Kleen 2010-02-22 20:20 ` Andi Kleen 2010-02-24 15:49 ` Christoph Lameter 2010-02-24 15:49 ` Christoph Lameter 2010-02-25 7:26 ` Pekka Enberg 2010-02-25 7:26 ` Pekka Enberg 2010-02-25 8:01 ` David Rientjes 2010-02-25 8:01 ` David Rientjes 2010-02-25 18:30 ` Christoph Lameter 2010-02-25 18:30 ` Christoph Lameter 2010-02-25 21:45 ` David Rientjes 2010-02-25 21:45 ` David Rientjes 2010-02-25 22:31 ` Christoph Lameter [this message] 2010-02-25 22:31 ` Christoph Lameter 2010-02-26 10:45 ` Pekka Enberg 2010-02-26 10:45 ` Pekka Enberg 2010-02-26 11:43 ` Andi Kleen 2010-02-26 11:43 ` Andi Kleen 2010-02-26 12:35 ` Pekka Enberg 2010-02-26 12:35 ` Pekka Enberg 2010-02-26 14:08 ` Andi Kleen 2010-02-26 14:08 ` Andi Kleen 2010-02-26 1:09 ` KAMEZAWA Hiroyuki 2010-02-26 1:09 ` KAMEZAWA Hiroyuki 2010-02-26 11:41 ` Andi Kleen 2010-02-26 11:41 ` Andi Kleen 2010-02-26 15:04 ` Christoph Lameter 2010-02-26 15:04 ` Christoph Lameter 2010-02-26 15:05 ` Christoph Lameter 2010-02-26 15:05 ` Christoph Lameter 2010-02-26 15:59 ` Andi Kleen 2010-02-26 15:59 ` Andi Kleen 2010-02-26 15:57 ` Andi Kleen 2010-02-26 15:57 ` Andi Kleen 2010-02-26 17:24 ` Christoph Lameter 2010-02-26 17:24 ` Christoph Lameter 2010-02-26 17:31 ` Andi Kleen 2010-02-26 17:31 ` Andi Kleen 2010-03-01 1:59 ` KAMEZAWA Hiroyuki 2010-03-01 1:59 ` KAMEZAWA Hiroyuki 2010-03-01 10:27 ` David Rientjes 2010-03-01 10:27 ` David Rientjes 2010-02-27 0:01 ` David Rientjes 2010-02-27 0:01 ` David Rientjes 2010-03-01 10:24 ` [patch] slab: add memory hotplug support David Rientjes 2010-03-01 10:24 ` David Rientjes 2010-03-02 5:53 ` Pekka Enberg 2010-03-02 5:53 ` Pekka Enberg 2010-03-02 20:20 ` Christoph Lameter 2010-03-02 20:20 ` Christoph Lameter 2010-03-02 21:03 ` David Rientjes 2010-03-02 21:03 ` David Rientjes 2010-03-03 1:28 ` KAMEZAWA Hiroyuki 2010-03-03 1:28 ` KAMEZAWA Hiroyuki 2010-03-03 2:39 ` David Rientjes 2010-03-03 2:39 ` David Rientjes 2010-03-03 2:51 ` KAMEZAWA Hiroyuki 2010-03-03 2:51 ` KAMEZAWA Hiroyuki 2010-03-02 12:53 ` Andi Kleen 2010-03-02 12:53 ` Andi Kleen 2010-03-02 15:04 ` Pekka Enberg 2010-03-02 15:04 ` Pekka Enberg 2010-03-03 14:34 ` Andi Kleen 2010-03-03 14:34 ` Andi Kleen 2010-03-03 15:46 ` Christoph Lameter 2010-03-03 15:46 ` Christoph Lameter 2010-03-02 21:17 ` David Rientjes 2010-03-02 21:17 ` David Rientjes 2010-03-05 6:20 ` Nick Piggin 2010-03-05 6:20 ` Nick Piggin 2010-03-05 12:47 ` Anca Emanuel 2010-03-05 12:47 ` Anca Emanuel 2010-03-05 13:58 ` Anca Emanuel 2010-03-05 13:58 ` Anca Emanuel 2010-03-05 14:11 ` Christoph Lameter 2010-03-05 14:11 ` Christoph Lameter 2010-03-08 3:06 ` Andi Kleen 2010-03-08 3:06 ` Andi Kleen 2010-03-08 2:58 ` Andi Kleen 2010-03-08 2:58 ` Andi Kleen 2010-03-08 23:19 ` David Rientjes 2010-03-08 23:19 ` David Rientjes 2010-03-09 13:46 ` Nick Piggin 2010-03-09 13:46 ` Nick Piggin 2010-03-22 17:28 ` Pekka Enberg 2010-03-22 17:28 ` Pekka Enberg 2010-03-22 21:12 ` Nick Piggin 2010-03-22 21:12 ` Nick Piggin 2010-03-28 2:13 ` David Rientjes 2010-03-28 2:13 ` David Rientjes 2010-03-28 2:40 ` [patch v2] " David Rientjes 2010-03-28 2:40 ` David Rientjes 2010-03-30 9:01 ` Pekka Enberg 2010-03-30 9:01 ` Pekka Enberg 2010-03-30 16:43 ` Christoph Lameter 2010-03-30 16:43 ` Christoph Lameter 2010-04-04 20:45 ` David Rientjes 2010-04-04 20:45 ` David Rientjes 2010-04-07 16:29 ` Pekka Enberg 2010-04-07 16:29 ` Pekka Enberg 2010-02-25 18:34 ` [PATCH] [4/4] SLAB: Fix node add timer race in cache_reap Christoph Lameter 2010-02-25 18:34 ` Christoph Lameter 2010-02-25 18:46 ` Pekka Enberg 2010-02-25 18:46 ` Pekka Enberg 2010-02-25 19:19 ` Christoph Lameter 2010-02-25 19:19 ` Christoph Lameter 2010-03-02 12:55 ` Andi Kleen 2010-03-02 12:55 ` Andi Kleen 2010-02-19 18:22 ` Christoph Lameter 2010-02-19 18:22 ` Christoph Lameter 2010-02-22 10:57 ` Pekka Enberg 2010-02-22 10:57 ` Pekka Enberg 2010-02-13 10:24 ` [PATCH] [0/4] Update slab memory hotplug series Pekka Enberg 2010-02-13 10:24 ` Pekka Enberg -- strict thread matches above, loose matches on Subject: below -- 2010-02-03 21:39 [PATCH] [0/4] SLAB: Fix a couple of slab memory hotadd issues Andi Kleen 2010-02-03 21:39 ` [PATCH] [4/4] SLAB: Fix node add timer race in cache_reap Andi Kleen 2010-02-03 21:39 ` Andi Kleen 2010-02-05 19:16 ` Christoph Lameter 2010-02-05 19:16 ` Christoph Lameter
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=alpine.DEB.2.00.1002251627040.18861@router.home \ --to=cl@linux-foundation.org \ --cc=andi@firstfloor.org \ --cc=haicheng.li@intel.com \ --cc=kamezawa.hiroyu@jp.fujitsu.com \ --cc=linux-kernel@vger.kernel.org \ --cc=linux-mm@kvack.org \ --cc=npiggin@suse.de \ --cc=penberg@cs.helsinki.fi \ --cc=rientjes@google.com \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.