From: Lai Jiangshan <laijs@cn.fujitsu.com>
To: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: <linux-kernel@vger.kernel.org>, Tejun Heo <tj@kernel.org>,
Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>,
"Gu, Zheng" <guz.fnst@cn.fujitsu.com>,
tangchen <tangchen@cn.fujitsu.com>
Subject: Re: [PATCH 1/4] workqueue:Fix unbound workqueue's node affinity detection
Date: Tue, 16 Dec 2014 13:30:32 +0800 [thread overview]
Message-ID: <548FC378.6060809@cn.fujitsu.com> (raw)
In-Reply-To: <548EC29A.5080008@jp.fujitsu.com>
On 12/15/2014 07:14 PM, Kamezawa Hiroyuki wrote:
> Unbound wq pool's node attribute is calculated at its allocation.
> But it's now calculated based on possible cpu<->node information
> which can be wrong after cpu hotplug/unplug.
>
> If wrong pool->node is set, following allocation error will happen.
> ==
> SLUB: Unable to allocate memory on node 2 (gfp=0x80d0)
> cache: kmalloc-192, object size: 192, buffer size: 192, default order:
> 1, min order: 0
> node 0: slabs: 6172, objs: 259224, free: 245741
> node 1: slabs: 3261, objs: 136962, free: 127656
> ==
>
> This patch fixes the node detection by making use of online cpu info.
> Unlike cpumask, the best node can be calculated by degree of overlap
> between attr->cpumask and numanode->online_cpumask.
> This change doesn't corrupt original purpose of the old calculation.
>
> Note: it's expected that this function is called as
> pool_detect_best_node
> get_unbound_pool
> alloc_unbound_pwq
> wq_update_unbound_numa
> called at CPU_ONLINE/CPU_DOWN_PREPARE
> and the latest online cpu info can be applied to a new wq pool,
> which replaces old one.
>
> Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
> ---
> kernel/workqueue.c | 38 ++++++++++++++++++++++++++------------
> 1 file changed, 26 insertions(+), 12 deletions(-)
>
> diff --git a/kernel/workqueue.c b/kernel/workqueue.c
> index 09b685d..7809154 100644
> --- a/kernel/workqueue.c
> +++ b/kernel/workqueue.c
> @@ -3440,6 +3440,31 @@ static void put_unbound_pool(struct worker_pool *pool)
> }
>
> /**
> + * pool_detect_best_node - detect a node which contains specified cpumask.
> + * Should be called with wq_pool_mutex held.
> + * Returns a online node where the most of given cpus are tied to.
> + */
> +static int pool_detect_best_node(const struct cpumask *cpumask)
> +{
> + int node, best, match, selected = NUMA_NO_NODE;
> + static struct cpumask andmask; /* under wq_pool_mutex */
> +
> + if (!wq_numa_enabled ||
> + cpumask_subset(cpu_online_mask, cpumask))
> + goto out;
> + best = 0;
> + /* select a node which contains the most number of cpu */
> + for_each_node_state(node, N_ONLINE) {
> + cpumask_and(&andmask, cpumask, cpumask_of_node(node));
> + match = cpumask_weight(&andmask);
> + if (match > best)
> + selected = best;
> + }
> +out:
> + return selected;
> +}
This is a mixture of fix and development. Why not just keep the original calculation?
if the mask cover multiple nodes, NUMA_NO_NODE is the best for pool->node
after the pool was created. The memory allocation will select the best node
for manage_workers(), from which CPU that the worker actually is running on.
> +
> +/**
> * get_unbound_pool - get a worker_pool with the specified attributes
> * @attrs: the attributes of the worker_pool to get
> *
> @@ -3457,7 +3482,6 @@ static struct worker_pool *get_unbound_pool(const struct workqueue_attrs *attrs)
> {
> u32 hash = wqattrs_hash(attrs);
> struct worker_pool *pool;
> - int node;
>
> lockdep_assert_held(&wq_pool_mutex);
>
> @@ -3482,17 +3506,7 @@ static struct worker_pool *get_unbound_pool(const struct workqueue_attrs *attrs)
> * 'struct workqueue_attrs' comments for detail.
> */
> pool->attrs->no_numa = false;
> -
> - /* if cpumask is contained inside a NUMA node, we belong to that node */
> - if (wq_numa_enabled) {
> - for_each_node(node) {
> - if (cpumask_subset(pool->attrs->cpumask,
> - wq_numa_possible_cpumask[node])) {
> - pool->node = node;
> - break;
> - }
> - }
> - }
> + pool->node = pool_detect_best_node(pool->attrs->cpumask);
>
> if (worker_pool_assign_id(pool) < 0)
> goto fail;
next prev parent reply other threads:[~2014-12-16 5:26 UTC|newest]
Thread overview: 55+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-12-12 10:19 [PATCH 0/5] workqueue: fix bug when numa mapping is changed Lai Jiangshan
2014-12-12 10:19 ` [PATCH 1/5] workqueue: fix memory leak in wq_numa_init() Lai Jiangshan
2014-12-12 17:12 ` Tejun Heo
2014-12-15 5:25 ` Lai Jiangshan
2014-12-12 10:19 ` [PATCH 2/5] workqueue: update wq_numa_possible_cpumask Lai Jiangshan
2014-12-12 17:18 ` Tejun Heo
2014-12-15 2:02 ` Lai Jiangshan
2014-12-25 20:16 ` Tejun Heo
2014-12-18 2:22 ` Lai Jiangshan
2014-12-12 10:19 ` [PATCH 3/5] workqueue: fixup existing pool->node Lai Jiangshan
2014-12-12 17:25 ` Tejun Heo
2014-12-15 1:23 ` Lai Jiangshan
2014-12-25 20:14 ` Tejun Heo
2015-01-13 7:08 ` Lai Jiangshan
2015-01-13 15:24 ` Tejun Heo
2014-12-12 10:19 ` [PATCH 4/5] workqueue: update NUMA affinity for the node lost CPU Lai Jiangshan
2014-12-12 17:27 ` Tejun Heo
2014-12-15 1:28 ` Lai Jiangshan
2014-12-25 20:17 ` Tejun Heo
2014-12-12 10:19 ` [PATCH 5/5] workqueue: retry on NUMA_NO_NODE when create_worker() fails Lai Jiangshan
2014-12-12 16:05 ` KOSAKI Motohiro
2014-12-12 17:29 ` KOSAKI Motohiro
2014-12-12 17:29 ` Tejun Heo
2014-12-12 17:13 ` [PATCH 0/5] workqueue: fix bug when numa mapping is changed Yasuaki Ishimatsu
2014-12-15 1:34 ` Lai Jiangshan
2014-12-18 1:50 ` Yasuaki Ishimatsu
2014-12-13 16:27 ` [PATCH 0/4] workqueue: fix bug when numa mapping is changed v2 Kamezawa Hiroyuki
2014-12-13 16:30 ` [PATCH 1/4] workqueue: add a hook for node hotplug Kamezawa Hiroyuki
2014-12-13 16:33 ` [PATCH 2/4] workqueue: add warning if pool->node is offline Kamezawa Hiroyuki
2014-12-13 16:35 ` [PATCH 3/4] workqueue: remove per-node unbound pool when node goes offline Kamezawa Hiroyuki
2014-12-15 2:06 ` Lai Jiangshan
2014-12-15 2:06 ` Kamezawa Hiroyuki
2014-12-13 16:38 ` [PATCH 4/4] workqueue: handle change in cpu-node relationship Kamezawa Hiroyuki
2014-12-15 2:12 ` Lai Jiangshan
2014-12-15 2:20 ` Kamezawa Hiroyuki
2014-12-15 2:48 ` Lai Jiangshan
2014-12-15 2:55 ` Kamezawa Hiroyuki
2014-12-15 3:30 ` Lai Jiangshan
2014-12-15 3:34 ` Lai Jiangshan
2014-12-15 4:04 ` Kamezawa Hiroyuki
2014-12-15 5:19 ` Lai Jiangshan
2014-12-15 5:33 ` Kamezawa Hiroyuki
2014-12-15 11:11 ` [PATCH 0/4] workqueue: fix memory allocation after numa mapping is changed v3 Kamezawa Hiroyuki
2014-12-15 11:14 ` [PATCH 1/4] workqueue:Fix unbound workqueue's node affinity detection Kamezawa Hiroyuki
2014-12-16 5:30 ` Lai Jiangshan [this message]
2014-12-16 7:32 ` Kamezawa Hiroyuki
2014-12-16 7:54 ` Lai Jiangshan
2014-12-15 11:16 ` [PATCH 2/4] workqueue: update per-cpu workqueue's node affinity at,online-offline Kamezawa Hiroyuki
2014-12-16 5:32 ` Lai Jiangshan
2014-12-16 7:25 ` Kamezawa Hiroyuki
2014-12-15 11:18 ` [PATCH 3/4] workqueue: Update workqueue's possible cpumask when a new node, coming up Kamezawa Hiroyuki
2014-12-16 7:49 ` Lai Jiangshan
2014-12-16 8:10 ` Kamezawa Hiroyuki
2014-12-16 8:18 ` Kamezawa Hiroyuki
2014-12-15 11:22 ` [PATCH 4/4] workqueue: Handle cpu-node affinity change at CPU_ONLINE Kamezawa Hiroyuki
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=548FC378.6060809@cn.fujitsu.com \
--to=laijs@cn.fujitsu.com \
--cc=guz.fnst@cn.fujitsu.com \
--cc=isimatu.yasuaki@jp.fujitsu.com \
--cc=kamezawa.hiroyu@jp.fujitsu.com \
--cc=linux-kernel@vger.kernel.org \
--cc=tangchen@cn.fujitsu.com \
--cc=tj@kernel.org \
/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: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).