From: Sage Weil <sage@newdream.net>
Date: Thu, 23 Jul 2009 19:50:34 +0000 (-0700)
Subject: kclient: use a mempool for page vector in writepages
X-Git-Tag: v0.12~90
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=aca6d39d8a06b72ab0a20a2d90c075239de4c69e;p=ceph.git

kclient: use a mempool for page vector in writepages
---

diff --git a/src/kernel/addr.c b/src/kernel/addr.c
index 2c1c422894e3..682552961c3b 100644
--- a/src/kernel/addr.c
+++ b/src/kernel/addr.c
@@ -538,10 +538,35 @@ static void writepages_finish(struct ceph_osd_request *req)
 	ceph_put_wrbuffer_cap_refs(ci, req->r_num_pages, snapc);
 
 	ceph_release_pages(req->r_pages, req->r_num_pages);
-	kfree(req->r_pages);
+	if (req->r_pages_from_pool)
+		mempool_free(req->r_pages,
+			     ceph_client(inode->i_sb)->wb_pagevec_pool);
+	else
+		kfree(req->r_pages);
 	ceph_osdc_put_request(req);
 }
 
+/*
+ * allocate a page vec, either directly, or if necessary, via a the
+ * mempool.  we avoid the mempool if we can because req->r_num_pages
+ * may be less than the maximum write size.
+ */
+static int alloc_page_vec(struct ceph_client *client,
+			  struct ceph_osd_request *req)
+{
+	req->r_pages = kmalloc(sizeof(struct page *) * req->r_num_pages,
+			       GFP_NOFS);
+	if (req->r_pages) {
+		req->r_pages_from_pool = 0;
+		return 0;
+	}
+
+	req->r_pages = mempool_alloc(client->wb_pagevec_pool, GFP_NOFS);
+	req->r_pages_from_pool = 1;
+	WARN_ON(!req->r_pages);
+	return -ENOMEM;
+}
+
 /*
  * initiate async writeback
  */
@@ -744,10 +769,8 @@ get_more_pages:
 					    &inode->i_mtime);
 				max_pages = req->r_num_pages;
 
-				rc = -ENOMEM;
-				req->r_pages = kmalloc(sizeof(*req->r_pages) *
-						       max_pages, GFP_NOFS);
-				if (req->r_pages == NULL)
+				rc = alloc_page_vec(client, req);
+				if (rc)
 					goto out;
 				req->r_callback = writepages_finish;
 				req->r_inode = inode;
diff --git a/src/kernel/osd_client.h b/src/kernel/osd_client.h
index 53822ac39eb6..7634b1f8497b 100644
--- a/src/kernel/osd_client.h
+++ b/src/kernel/osd_client.h
@@ -45,6 +45,7 @@ struct ceph_osd_request {
 	struct ceph_snap_context *r_snapc;    /* snap context for writes */
 	unsigned          r_num_pages;        /* size of page array (follows) */
 	struct page     **r_pages;            /* pages for data payload */
+	int               r_pages_from_pool;
 	int               r_own_pages;        /* if true, i own page list */
 };
 
diff --git a/src/kernel/super.c b/src/kernel/super.c
index 16d674caeef7..8439888c6eb0 100644
--- a/src/kernel/super.c
+++ b/src/kernel/super.c
@@ -724,6 +724,8 @@ static void ceph_destroy_client(struct ceph_client *client)
 		destroy_workqueue(client->trunc_wq);
 	if (client->msgr)
 		ceph_messenger_destroy(client->msgr);
+	if (client->wb_pagevec_pool)
+		mempool_destroy(client->wb_pagevec_pool);
 	kfree(client);
 	dout("destroy_client %p done\n", client);
 }
@@ -975,6 +977,13 @@ static int ceph_set_super(struct super_block *s, void *data)
 
 	s->s_time_gran = 1000;  /* 1000 ns == 1 us */
 
+	/* set up mempools */
+	ret = -ENOMEM;
+	client->wb_pagevec_pool = mempool_create_kmalloc_pool(10,
+			      client->mount_args.wsize >> PAGE_CACHE_SHIFT);
+	if (!client->wb_pagevec_pool)
+		goto fail;
+
 	ret = set_anon_super(s, NULL);  /* what is that second arg for? */
 	if (ret != 0)
 		goto fail;
diff --git a/src/kernel/super.h b/src/kernel/super.h
index 3f9e3e59c270..07d6aa41d4fe 100644
--- a/src/kernel/super.h
+++ b/src/kernel/super.h
@@ -2,12 +2,13 @@
 #define _FS_CEPH_SUPER_H
 
 #include <asm/unaligned.h>
-#include <linux/fs.h>
-#include <linux/wait.h>
+#include <linux/backing-dev.h>
 #include <linux/completion.h>
-#include <linux/pagemap.h>
 #include <linux/exportfs.h>
-#include <linux/backing-dev.h>
+#include <linux/fs.h>
+#include <linux/mempool.h>
+#include <linux/pagemap.h>
+#include <linux/wait.h>
 
 #include "types.h"
 #include "ceph_debug.h"
@@ -132,6 +133,7 @@ struct ceph_client {
 	struct ceph_osd_client osdc;
 
 	/* writeback */
+	mempool_t *wb_pagevec_pool;
 	struct workqueue_struct *wb_wq;
 	struct workqueue_struct *pg_inv_wq;
 	struct workqueue_struct *trunc_wq;