linux-mm.kvack.org archive mirror
 help / color / mirror / Atom feed
From: Lee Schermerhorn <lee.schermerhorn@hp.com>
To: linux-mm@kvack.org, linux-numa@vger.kernel.org
Cc: akpm@linux-foundation.org, Mel Gorman <mel@csn.ul.ie>,
	Randy Dunlap <randy.dunlap@oracle.com>,
	Nishanth Aravamudan <nacc@us.ibm.com>,
	David Rientjes <rientjes@google.com>, Adam Litke <agl@us.ibm.com>,
	Andy Whitcroft <apw@canonical.com>,
	eric.whitney@hp.com
Subject: [PATCH 11/11] hugetlb:  offload per node attribute registrations
Date: Mon, 05 Oct 2009 23:19:24 -0400	[thread overview]
Message-ID: <20091006031924.22576.35018.sendpatchset@localhost.localdomain> (raw)
In-Reply-To: <20091006031739.22576.5248.sendpatchset@localhost.localdomain>

[PATCH 11/11] hugetlb:  offload [un]registration of sysfs attr to worker thread

Against:  2.6.31-mmotm-090925-1435

New in V6

V7:  + remove redundant check for memory{ful|less} node from 
       node_hugetlb_work().  Rely on [added] return from
       hugetlb_register_node() to differentiate between transitions
       to/from memoryless state.

This patch offloads the registration and unregistration of per node
hstate sysfs attributes to a worker thread rather than attempt the
allocation/attachment or detachment/freeing of the attributes in 
the context of the memory hotplug handler.

N.B.,  Only tested build, boot, libhugetlbfs regression.
       i.e., no memory hotplug testing.

Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>

 drivers/base/node.c  |   51 ++++++++++++++++++++++++++++++++++++++++++---------
 include/linux/node.h |    5 +++++
 2 files changed, 47 insertions(+), 9 deletions(-)

Index: linux-2.6.31-mmotm-090925-1435/include/linux/node.h
===================================================================
--- linux-2.6.31-mmotm-090925-1435.orig/include/linux/node.h	2009-09-30 15:05:20.000000000 -0400
+++ linux-2.6.31-mmotm-090925-1435/include/linux/node.h	2009-09-30 15:05:58.000000000 -0400
@@ -21,9 +21,14 @@
 
 #include <linux/sysdev.h>
 #include <linux/cpumask.h>
+#include <linux/workqueue.h>
 
 struct node {
 	struct sys_device	sysdev;
+
+#if defined(CONFIG_MEMORY_HOTPLUG_SPARSE) && defined(CONFIG_HUGETLBFS)
+	struct work_struct	node_work;
+#endif
 };
 
 struct memory_block;
Index: linux-2.6.31-mmotm-090925-1435/drivers/base/node.c
===================================================================
--- linux-2.6.31-mmotm-090925-1435.orig/drivers/base/node.c	2009-09-30 15:05:57.000000000 -0400
+++ linux-2.6.31-mmotm-090925-1435/drivers/base/node.c	2009-09-30 15:05:58.000000000 -0400
@@ -186,11 +186,14 @@ static SYSDEV_ATTR(distance, S_IRUGO, no
 static node_registration_func_t __hugetlb_register_node;
 static node_registration_func_t __hugetlb_unregister_node;
 
-static inline void hugetlb_register_node(struct node *node)
+static inline bool hugetlb_register_node(struct node *node)
 {
 	if (__hugetlb_register_node &&
-			node_state(node->sysdev.id, N_HIGH_MEMORY))
+			node_state(node->sysdev.id, N_HIGH_MEMORY)) {
 		__hugetlb_register_node(node);
+		return true;
+	}
+	return false;
 }
 
 static inline void hugetlb_unregister_node(struct node *node)
@@ -387,10 +390,31 @@ static int link_mem_sections(int nid)
 	return err;
 }
 
+#ifdef CONFIG_HUGETLBFS
 /*
  * Handle per node hstate attribute [un]registration on transistions
  * to/from memoryless state.
  */
+static void node_hugetlb_work(struct work_struct *work)
+{
+	struct node *node = container_of(work, struct node, node_work);
+
+	/*
+	 * We only get here when a node transitions to/from memoryless state.
+	 * We can detect which transition occurred by examining whether the
+	 * node has memory now.  hugetlb_register_node() already check this
+	 * so we try to register the attributes.  If that fails, then the
+	 * node has transitioned to memoryless, try to unregister the
+	 * attributes.
+	 */
+	if (!hugetlb_register_node(node))
+		hugetlb_unregister_node(node);
+}
+
+static void init_node_hugetlb_work(int nid)
+{
+	INIT_WORK(&node_devices[nid].node_work, node_hugetlb_work);
+}
 
 static int node_memory_callback(struct notifier_block *self,
 				unsigned long action, void *arg)
@@ -399,14 +423,16 @@ static int node_memory_callback(struct n
 	int nid = mnb->status_change_nid;
 
 	switch (action) {
-	case MEM_ONLINE:    /* memory successfully brought online */
+	case MEM_ONLINE:
+	case MEM_OFFLINE:
+		/*
+		 * offload per node hstate [un]registration to a work thread
+		 * when transitioning to/from memoryless state.
+		 */
 		if (nid != NUMA_NO_NODE)
-			hugetlb_register_node(&node_devices[nid]);
-		break;
-	case MEM_OFFLINE:   /* or offline */
-		if (nid != NUMA_NO_NODE)
-			hugetlb_unregister_node(&node_devices[nid]);
+			schedule_work(&node_devices[nid].node_work);
 		break;
+
 	case MEM_GOING_ONLINE:
 	case MEM_GOING_OFFLINE:
 	case MEM_CANCEL_ONLINE:
@@ -417,7 +443,8 @@ static int node_memory_callback(struct n
 
 	return NOTIFY_OK;
 }
-#else
+#endif	/* CONFIG_HUGETLBFS */
+#else	/* !CONFIG_MEMORY_HOTPLUG_SPARSE */
 static int link_mem_sections(int nid) { return 0; }
 
 static inline int node_memory_callback(struct notifier_block *self,
@@ -425,6 +452,9 @@ static inline int node_memory_callback(s
 {
 	return NOTIFY_OK;
 }
+
+static void init_node_hugetlb_work(int nid) { }
+
 #endif /* CONFIG_MEMORY_HOTPLUG_SPARSE */
 
 int register_one_node(int nid)
@@ -449,6 +479,9 @@ int register_one_node(int nid)
 
 		/* link memory sections under this node */
 		error = link_mem_sections(nid);
+
+		/* initialize work queue for memory hot plug */
+		init_node_hugetlb_work(nid);
 	}
 
 	return error;

--
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>

  parent reply	other threads:[~2009-10-06  3:15 UTC|newest]

Thread overview: 32+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2009-10-06  3:17 [PATCH 0/11] hugetlb: V9 numa control of persistent huge pages alloc/free Lee Schermerhorn
2009-10-06  3:17 ` [PATCH 1/11] hugetlb: rework hstate_next_node_* functions Lee Schermerhorn
2009-10-06  3:17 ` [PATCH 2/11] hugetlb: add nodemask arg to huge page alloc, free and surplus adjust fcns Lee Schermerhorn
2009-10-06  9:09   ` David Rientjes
2009-10-07  3:26   ` David Rientjes
2009-10-07 14:13     ` Lee Schermerhorn
2009-10-06  3:17 ` [PATCH 3/11] hugetlb: factor init_nodemask_of_node Lee Schermerhorn
2009-10-07  3:21   ` David Rientjes
2009-10-06  3:18 ` [PATCH 4/11] hugetlb: derive huge pages nodes allowed from task mempolicy Lee Schermerhorn
2009-10-07  3:26   ` David Rientjes
2009-10-07 16:30     ` Lee Schermerhorn
2009-10-07 20:09       ` David Rientjes
2009-10-06  3:18 ` [PATCH 5/11] hugetlb: accomodate reworked NODEMASK_ALLOC Lee Schermerhorn
2009-10-06  3:18 ` [PATCH 6/11] hugetlb: add generic definition of NUMA_NO_NODE Lee Schermerhorn
2009-10-06  9:28   ` David Rientjes
2009-10-06  3:18 ` [PATCH 7/11] hugetlb: add per node hstate attributes Lee Schermerhorn
2009-10-07  4:04   ` David Rientjes
2009-10-06  3:18 ` [PATCH 8/11] hugetlb: update hugetlb documentation for NUMA controls Lee Schermerhorn
2009-10-06  3:18 ` [PATCH 9/11] hugetlb: use only nodes with memory for huge pages Lee Schermerhorn
2009-10-06  3:18 ` [PATCH 10/11] hugetlb: handle memory hot-plug events Lee Schermerhorn
2009-10-07  4:12   ` David Rientjes
2009-10-06  3:19 ` Lee Schermerhorn [this message]
2009-10-06 16:01   ` [PATCH 11/11] hugetlb: offload per node attribute registrations Andi Kleen
2009-10-06 16:28     ` Lee Schermerhorn
2009-10-06 16:46       ` Andi Kleen
2009-10-06 17:57         ` Lee Schermerhorn
2009-10-07  8:24   ` [patch] mm: clear node in N_HIGH_MEMORY and stop kswapd when all memory is offlined David Rientjes
2009-10-07 14:25     ` Christoph Lameter
2009-10-07 16:48     ` Lee Schermerhorn
2009-10-07 19:53       ` David Rientjes
2009-10-06 16:02 ` [PATCH 0/11] hugetlb: V9 numa control of persistent huge pages alloc/free Andi Kleen
  -- strict thread matches above, loose matches on Subject: below --
2009-09-15 20:43 [PATCH 0/11] hugetlb: V7 constrain allocation/free based on task mempolicy Lee Schermerhorn
2009-09-15 20:45 ` [PATCH 11/11] hugetlb: offload per node attribute registrations Lee Schermerhorn

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=20091006031924.22576.35018.sendpatchset@localhost.localdomain \
    --to=lee.schermerhorn@hp.com \
    --cc=agl@us.ibm.com \
    --cc=akpm@linux-foundation.org \
    --cc=apw@canonical.com \
    --cc=eric.whitney@hp.com \
    --cc=linux-mm@kvack.org \
    --cc=linux-numa@vger.kernel.org \
    --cc=mel@csn.ul.ie \
    --cc=nacc@us.ibm.com \
    --cc=randy.dunlap@oracle.com \
    --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: 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).