cgroups: use a hash table for css_set finding

When we attach a process to a different cgroup, the css_set linked-list will be run through to find a suitable existing css_set to use. This patch implements a hash table for better performance. The following benchmarks have been tested: For N in 1, 5, 10, 50, 100, 500, 1000, create N cgroups with one sleeping task in each, and then move an additional task through each cgroup in turn. Here is a test result: N Loop orig - Time(s) hash - Time(s) ---------------------------------------------- 1 10000 1.201231728 1.196311177 5 2000 1.065743872 1.040566424 10 1000 0.991054735 0.986876440 50 200 0.976554203 0.969608733 100 100 0.998504680 0.969218270 500 20 1.157347764 0.962602963 1000 10 1.619521852 1.085140172 Signed-off-by: Li Zefan <lizf@cn.fujitsu.com> Reviewed-by: Paul Menage <menage@google.com> Cc: Balbir Singh <balbir@linux.vnet.ibm.com> Cc: Pavel Emelyanov <xemul@openvz.org> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cgroups: use a hash table for css_set finding
When we attach a process to a different cgroup, the css_set linked-list will be run through to find a suitable existing css_set to use. This patch implements a hash table for better performance. The following benchmarks have been tested: For N in 1, 5, 10, 50, 100, 500, 1000, create N cgroups with one sleeping task in each, and then move an additional task through each cgroup in turn. Here is a test result: N Loop orig - Time(s) hash - Time(s) ---------------------------------------------- 1 10000 1.201231728 1.196311177 5 2000 1.065743872 1.040566424 10 1000 0.991054735 0.986876440 50 200 0.976554203 0.969608733 100 100 0.998504680 0.969218270 500 20 1.157347764 0.962602963 1000 10 1.619521852 1.085140172 Signed-off-by: Li Zefan <lizf@cn.fujitsu.com> Reviewed-by: Paul Menage <menage@google.com> Cc: Balbir Singh <balbir@linux.vnet.ibm.com> Cc: Pavel Emelyanov <xemul@openvz.org> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Li Zefan · Linus Torvalds
1 parent 08ce5f16ee
Showing 2 changed files with 53 additions and 13 deletions Side-by-side Diff
include/linux/cgroup.h
kernel/cgroup.c
@@ -156,6 +156,12 @@
 	struct list_head list;
  
 	/*
+	 * List running through all cgroup groups in the same hash
+	 * slot. Protected by css_set_lock
+	 */
+	struct hlist_node hlist;
+
+	/*
 	 * List running through all tasks using this cgroup
 	 * group. Protected by css_set_lock
 	 */
@@ -174,7 +180,6 @@
 	 * during subsystem registration (at boot time).
 	 */
 	struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT];
-
 };
  
 /*
@@ -44,6 +44,7 @@
 #include <linux/kmod.h>
 #include <linux/delayacct.h>
 #include <linux/cgroupstats.h>
+#include <linux/hash.h>
  
 #include <asm/atomic.h>
  
@@ -193,6 +194,27 @@
 static DEFINE_RWLOCK(css_set_lock);
 static int css_set_count;
  
+/* hash table for cgroup groups. This improves the performance to
+ * find an existing css_set */
+#define CSS_SET_HASH_BITS	7
+#define CSS_SET_TABLE_SIZE	(1 << CSS_SET_HASH_BITS)
+static struct hlist_head css_set_table[CSS_SET_TABLE_SIZE];
+
+static struct hlist_head *css_set_hash(struct cgroup_subsys_state *css[])
+{
+	int i;
+	int index;
+	unsigned long tmp = 0UL;
+
+	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++)
+		tmp += (unsigned long)css[i];
+	tmp = (tmp >> 16) ^ tmp;
+
+	index = hash_long(tmp, CSS_SET_HASH_BITS);
+
+	return &css_set_table[index];
+}
+
 /* We don't maintain the lists running through each css_set to its
  * task until after the first call to cgroup_iter_start(). This
  * reduces the fork()/exit() overhead for people who have cgroups
@@ -219,6 +241,7 @@
 static void unlink_css_set(struct css_set *cg)
 {
 	write_lock(&css_set_lock);
+	hlist_del(&cg->hlist);
 	list_del(&cg->list);
 	css_set_count--;
 	while (!list_empty(&cg->cg_links)) {
@@ -284,9 +307,7 @@
 /*
  * find_existing_css_set() is a helper for
  * find_css_set(), and checks to see whether an existing
- * css_set is suitable. This currently walks a linked-list for
- * simplicity; a later patch will use a hash table for better
- * performance
+ * css_set is suitable.
  *
  * oldcg: the cgroup group that we're using before the cgroup
  * transition
@@ -303,7 +324,9 @@
 {
 	int i;
 	struct cgroupfs_root *root = cgrp->root;
-	struct list_head *l = &init_css_set.list;
+	struct hlist_head *hhead;
+	struct hlist_node *node;
+	struct css_set *cg;
  
 	/* Built the set of subsystem state objects that we want to
 	 * see in the new css_set */
  
@@ -320,18 +343,13 @@
 		}
 	}
  
-	/* Look through existing cgroup groups to find one to reuse */
-	do {
-		struct css_set *cg =
-			list_entry(l, struct css_set, list);
-
+	hhead = css_set_hash(template);
+	hlist_for_each_entry(cg, node, hhead, hlist) {
 		if (!memcmp(template, cg->subsys, sizeof(cg->subsys))) {
 			/* All subsystems matched */
 			return cg;
 		}
-		/* Try the next cgroup group */
-		l = l->next;
-	} while (l != &init_css_set.list);
+	}
  
 	/* No existing cgroup group matched */
 	return NULL;
@@ -393,6 +411,8 @@
 	struct list_head tmp_cg_links;
 	struct cg_cgroup_link *link;
  
+	struct hlist_head *hhead;
+
 	/* First see if we already have a cgroup group that matches
 	 * the desired set */
 	write_lock(&css_set_lock);
@@ -417,6 +437,7 @@
 	kref_init(&res->ref);
 	INIT_LIST_HEAD(&res->cg_links);
 	INIT_LIST_HEAD(&res->tasks);
+	INIT_HLIST_NODE(&res->hlist);
  
 	/* Copy the set of subsystem state objects generated in
 	 * find_existing_css_set() */
@@ -459,6 +480,11 @@
 	/* Link this cgroup group into the list */
 	list_add(&res->list, &init_css_set.list);
 	css_set_count++;
+
+	/* Add this cgroup group to the hash table */
+	hhead = css_set_hash(res->subsys);
+	hlist_add_head(&res->hlist, hhead);
+
 	write_unlock(&css_set_lock);
  
 	return res;
@@ -2508,6 +2534,7 @@
 	INIT_LIST_HEAD(&init_css_set.list);
 	INIT_LIST_HEAD(&init_css_set.cg_links);
 	INIT_LIST_HEAD(&init_css_set.tasks);
+	INIT_HLIST_NODE(&init_css_set.hlist);
 	css_set_count = 1;
 	init_cgroup_root(&rootnode);
 	list_add(&rootnode.root_list, &roots);
@@ -2520,6 +2547,9 @@
 	list_add(&init_css_set_link.cg_link_list,
 		 &init_css_set.cg_links);
  
+	for (i = 0; i < CSS_SET_TABLE_SIZE; i++)
+		INIT_HLIST_HEAD(&css_set_table[i]);
+
 	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
 		struct cgroup_subsys *ss = subsys[i];
  
@@ -2549,6 +2579,7 @@
 {
 	int err;
 	int i;
+	struct hlist_head *hhead;
  
 	err = bdi_init(&cgroup_backing_dev_info);
 	if (err)
@@ -2559,6 +2590,10 @@
 		if (!ss->early_init)
 			cgroup_init_subsys(ss);
 	}
+
+	/* Add init_css_set to the hash table */
+	hhead = css_set_hash(init_css_set.subsys);
+	hlist_add_head(&init_css_set.hlist, hhead);
  
 	err = register_filesystem(&cgroup_fs_type);
 	if (err < 0)
...	...	@@ -156,6 +156,12 @@
156	156	struct list_head list;
157	157
158	158	/*
	159	+ * List running through all cgroup groups in the same hash
	160	+ * slot. Protected by css_set_lock
	161	+ */
	162	+ struct hlist_node hlist;
	163	+
	164	+ /*
159	165	* List running through all tasks using this cgroup
160	166	* group. Protected by css_set_lock
161	167	*/
...	...	@@ -174,7 +180,6 @@
174	180	* during subsystem registration (at boot time).
175	181	*/
176	182	struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT];
177		-
178	183	};
179	184
180	185	/*
...	...	@@ -44,6 +44,7 @@
44	44	#include <linux/kmod.h>
45	45	#include <linux/delayacct.h>
46	46	#include <linux/cgroupstats.h>
	47	+#include <linux/hash.h>
47	48
48	49	#include <asm/atomic.h>
49	50
...	...	@@ -193,6 +194,27 @@
193	194	static DEFINE_RWLOCK(css_set_lock);
194	195	static int css_set_count;
195	196
	197	+/* hash table for cgroup groups. This improves the performance to
	198	+ * find an existing css_set */
	199	+#define CSS_SET_HASH_BITS 7
	200	+#define CSS_SET_TABLE_SIZE (1 << CSS_SET_HASH_BITS)
	201	+static struct hlist_head css_set_table[CSS_SET_TABLE_SIZE];
	202	+
	203	+static struct hlist_head css_set_hash(struct cgroup_subsys_state css[])
	204	+{
	205	+ int i;
	206	+ int index;
	207	+ unsigned long tmp = 0UL;
	208	+
	209	+ for (i = 0; i < CGROUP_SUBSYS_COUNT; i++)
	210	+ tmp += (unsigned long)css[i];
	211	+ tmp = (tmp >> 16) ^ tmp;
	212	+
	213	+ index = hash_long(tmp, CSS_SET_HASH_BITS);
	214	+
	215	+ return &css_set_table[index];
	216	+}
	217	+
196	218	/* We don't maintain the lists running through each css_set to its
197	219	* task until after the first call to cgroup_iter_start(). This
198	220	* reduces the fork()/exit() overhead for people who have cgroups
...	...	@@ -219,6 +241,7 @@
219	241	static void unlink_css_set(struct css_set *cg)
220	242	{
221	243	write_lock(&css_set_lock);
	244	+ hlist_del(&cg->hlist);
222	245	list_del(&cg->list);
223	246	css_set_count--;
224	247	while (!list_empty(&cg->cg_links)) {
...	...	@@ -284,9 +307,7 @@
284	307	/*
285	308	* find_existing_css_set() is a helper for
286	309	* find_css_set(), and checks to see whether an existing
287		- * css_set is suitable. This currently walks a linked-list for
288		- * simplicity; a later patch will use a hash table for better
289		- * performance
	310	+ * css_set is suitable.
290	311	*
291	312	* oldcg: the cgroup group that we're using before the cgroup
292	313	* transition
...	...	@@ -303,7 +324,9 @@
303	324	{
304	325	int i;
305	326	struct cgroupfs_root *root = cgrp->root;
306		- struct list_head *l = &init_css_set.list;
	327	+ struct hlist_head *hhead;
	328	+ struct hlist_node *node;
	329	+ struct css_set *cg;
307	330
308	331	/* Built the set of subsystem state objects that we want to
309	332	* see in the new css_set */
310	333
...	...	@@ -320,18 +343,13 @@
320	343	}
321	344	}
322	345
323		- /* Look through existing cgroup groups to find one to reuse */
324		- do {
325		- struct css_set *cg =
326		- list_entry(l, struct css_set, list);
327		-
	346	+ hhead = css_set_hash(template);
	347	+ hlist_for_each_entry(cg, node, hhead, hlist) {
328	348	if (!memcmp(template, cg->subsys, sizeof(cg->subsys))) {
329	349	/* All subsystems matched */
330	350	return cg;
331	351	}
332		- /* Try the next cgroup group */
333		- l = l->next;
334		- } while (l != &init_css_set.list);
	352	+ }
335	353
336	354	/* No existing cgroup group matched */
337	355	return NULL;
...	...	@@ -393,6 +411,8 @@
393	411	struct list_head tmp_cg_links;
394	412	struct cg_cgroup_link *link;
395	413
	414	+ struct hlist_head *hhead;
	415	+
396	416	/* First see if we already have a cgroup group that matches
397	417	* the desired set */
398	418	write_lock(&css_set_lock);
...	...	@@ -417,6 +437,7 @@
417	437	kref_init(&res->ref);
418	438	INIT_LIST_HEAD(&res->cg_links);
419	439	INIT_LIST_HEAD(&res->tasks);
	440	+ INIT_HLIST_NODE(&res->hlist);
420	441
421	442	/* Copy the set of subsystem state objects generated in
422	443	* find_existing_css_set() */
...	...	@@ -459,6 +480,11 @@
459	480	/* Link this cgroup group into the list */
460	481	list_add(&res->list, &init_css_set.list);
461	482	css_set_count++;
	483	+
	484	+ /* Add this cgroup group to the hash table */
	485	+ hhead = css_set_hash(res->subsys);
	486	+ hlist_add_head(&res->hlist, hhead);
	487	+
462	488	write_unlock(&css_set_lock);
463	489
464	490	return res;
...	...	@@ -2508,6 +2534,7 @@
2508	2534	INIT_LIST_HEAD(&init_css_set.list);
2509	2535	INIT_LIST_HEAD(&init_css_set.cg_links);
2510	2536	INIT_LIST_HEAD(&init_css_set.tasks);
	2537	+ INIT_HLIST_NODE(&init_css_set.hlist);
2511	2538	css_set_count = 1;
2512	2539	init_cgroup_root(&rootnode);
2513	2540	list_add(&rootnode.root_list, &roots);
...	...	@@ -2520,6 +2547,9 @@
2520	2547	list_add(&init_css_set_link.cg_link_list,
2521	2548	&init_css_set.cg_links);
2522	2549
	2550	+ for (i = 0; i < CSS_SET_TABLE_SIZE; i++)
	2551	+ INIT_HLIST_HEAD(&css_set_table[i]);
	2552	+
2523	2553	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
2524	2554	struct cgroup_subsys *ss = subsys[i];
2525	2555
...	...	@@ -2549,6 +2579,7 @@
2549	2579	{
2550	2580	int err;
2551	2581	int i;
	2582	+ struct hlist_head *hhead;
2552	2583
2553	2584	err = bdi_init(&cgroup_backing_dev_info);
2554	2585	if (err)
...	...	@@ -2559,6 +2590,10 @@
2559	2590	if (!ss->early_init)
2560	2591	cgroup_init_subsys(ss);
2561	2592	}
	2593	+
	2594	+ /* Add init_css_set to the hash table */
	2595	+ hhead = css_set_hash(init_css_set.subsys);
	2596	+ hlist_add_head(&init_css_set.hlist, hhead);
2562	2597
2563	2598	err = register_filesystem(&cgroup_fs_type);
2564	2599	if (err < 0)