Memory controller: add switch to control what type of pages to limit

Choose if we want cached pages to be accounted or not. By default both are accounted for. A new set of tunables are added. echo -n 1 > mem_control_type switches the accounting to account for only mapped pages echo -n 3 > mem_control_type switches the behaviour back [bunk@kernel.org: mm/memcontrol.c: clenups] [akpm@linux-foundation.org: fix sparc32 build] Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com> Cc: Pavel Emelianov <xemul@openvz.org> Cc: Paul Menage <menage@google.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: "Eric W. Biederman" <ebiederm@xmission.com> Cc: Nick Piggin <nickpiggin@yahoo.com.au> Cc: Kirill Korotaev <dev@sw.ru> Cc: Herbert Poetzl <herbert@13thfloor.at> Cc: David Rientjes <rientjes@google.com> Cc: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> Signed-off-by: Adrian Bunk <bunk@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Memory controller: add switch to control what type of pages to limit
Choose if we want cached pages to be accounted or not. By default both are accounted for. A new set of tunables are added. echo -n 1 > mem_control_type switches the accounting to account for only mapped pages echo -n 3 > mem_control_type switches the behaviour back [bunk@kernel.org: mm/memcontrol.c: clenups] [akpm@linux-foundation.org: fix sparc32 build] Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com> Cc: Pavel Emelianov <xemul@openvz.org> Cc: Paul Menage <menage@google.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: "Eric W. Biederman" <ebiederm@xmission.com> Cc: Nick Piggin <nickpiggin@yahoo.com.au> Cc: Kirill Korotaev <dev@sw.ru> Cc: Herbert Poetzl <herbert@13thfloor.at> Cc: David Rientjes <rientjes@google.com> Cc: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> Signed-off-by: Adrian Bunk <bunk@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Balbir Singh · Linus Torvalds
1 parent c7ba5c9e81
Showing 4 changed files with 106 additions and 5 deletions Side-by-side Diff
include/linux/memcontrol.h
mm/filemap.c
mm/memcontrol.c
mm/swap_state.c
@@ -22,6 +22,8 @@
  
 struct mem_cgroup;
 struct page_cgroup;
+struct page;
+struct mm_struct;
  
 #ifdef CONFIG_CGROUP_MEM_CONT
  
@@ -40,6 +42,7 @@
 					struct mem_cgroup *mem_cont,
 					int active);
 extern void mem_cgroup_out_of_memory(struct mem_cgroup *mem, gfp_t gfp_mask);
+extern int mem_cgroup_cache_charge(struct page *page, struct mm_struct *mm);
  
 static inline void mem_cgroup_uncharge_page(struct page *page)
 {
@@ -82,6 +85,12 @@
 static inline void mem_cgroup_move_lists(struct page_cgroup *pc,
 						bool active)
 {
+}
+
+static inline int mem_cgroup_cache_charge(struct page *page,
+						struct mm_struct *mm)
+{
+	return 0;
 }
  
 #endif /* CONFIG_CGROUP_MEM_CONT */
@@ -464,7 +464,7 @@
  
 	if (error == 0) {
  
-		error = mem_cgroup_charge(page, current->mm);
+		error = mem_cgroup_cache_charge(page, current->mm);
 		if (error)
 			goto out;
  
@@ -29,6 +29,8 @@
 #include <linux/spinlock.h>
 #include <linux/fs.h>
  
+#include <asm/uaccess.h>
+
 struct cgroup_subsys mem_cgroup_subsys;
 static const int MEM_CGROUP_RECLAIM_RETRIES = 5;
  
@@ -60,6 +62,7 @@
 	 * spin_lock to protect the per cgroup LRU
 	 */
 	spinlock_t lru_lock;
+	unsigned long control_type;	/* control RSS or RSS+Pagecache */
 };
  
 /*
  
@@ -82,7 +85,16 @@
 					/* mapped and cached states     */
 };
  
+enum {
+	MEM_CGROUP_TYPE_UNSPEC = 0,
+	MEM_CGROUP_TYPE_MAPPED,
+	MEM_CGROUP_TYPE_CACHED,
+	MEM_CGROUP_TYPE_ALL,
+	MEM_CGROUP_TYPE_MAX,
+};
  
+static struct mem_cgroup init_mem_cgroup;
+
 static inline
 struct mem_cgroup *mem_cgroup_from_cont(struct cgroup *cont)
 {
  
  
@@ -139,18 +151,18 @@
 		(page->page_cgroup & ~PAGE_CGROUP_LOCK);
 }
  
-void __always_inline lock_page_cgroup(struct page *page)
+static void __always_inline lock_page_cgroup(struct page *page)
 {
 	bit_spin_lock(PAGE_CGROUP_LOCK_BIT, &page->page_cgroup);
 	VM_BUG_ON(!page_cgroup_locked(page));
 }
  
-void __always_inline unlock_page_cgroup(struct page *page)
+static void __always_inline unlock_page_cgroup(struct page *page)
 {
 	bit_spin_unlock(PAGE_CGROUP_LOCK_BIT, &page->page_cgroup);
 }
  
-void __mem_cgroup_move_lists(struct page_cgroup *pc, bool active)
+static void __mem_cgroup_move_lists(struct page_cgroup *pc, bool active)
 {
 	if (active)
 		list_move(&pc->lru, &pc->mem_cgroup->active_list);
@@ -366,6 +378,22 @@
 }
  
 /*
+ * See if the cached pages should be charged at all?
+ */
+int mem_cgroup_cache_charge(struct page *page, struct mm_struct *mm)
+{
+	struct mem_cgroup *mem;
+	if (!mm)
+		mm = &init_mm;
+
+	mem = rcu_dereference(mm->mem_cgroup);
+	if (mem->control_type == MEM_CGROUP_TYPE_ALL)
+		return mem_cgroup_charge(page, mm);
+	else
+		return 0;
+}
+
+/*
  * Uncharging is always a welcome operation, we never complain, simply
  * uncharge.
  */
@@ -375,6 +403,10 @@
 	struct page *page;
 	unsigned long flags;
  
+	/*
+	 * This can handle cases when a page is not charged at all and we
+	 * are switching between handling the control_type.
+	 */
 	if (!pc)
 		return;
  
@@ -425,6 +457,60 @@
 				mem_cgroup_write_strategy);
 }
  
+static ssize_t mem_control_type_write(struct cgroup *cont,
+			struct cftype *cft, struct file *file,
+			const char __user *userbuf,
+			size_t nbytes, loff_t *pos)
+{
+	int ret;
+	char *buf, *end;
+	unsigned long tmp;
+	struct mem_cgroup *mem;
+
+	mem = mem_cgroup_from_cont(cont);
+	buf = kmalloc(nbytes + 1, GFP_KERNEL);
+	ret = -ENOMEM;
+	if (buf == NULL)
+		goto out;
+
+	buf[nbytes] = 0;
+	ret = -EFAULT;
+	if (copy_from_user(buf, userbuf, nbytes))
+		goto out_free;
+
+	ret = -EINVAL;
+	tmp = simple_strtoul(buf, &end, 10);
+	if (*end != '\0')
+		goto out_free;
+
+	if (tmp <= MEM_CGROUP_TYPE_UNSPEC || tmp >= MEM_CGROUP_TYPE_MAX)
+		goto out_free;
+
+	mem->control_type = tmp;
+	ret = nbytes;
+out_free:
+	kfree(buf);
+out:
+	return ret;
+}
+
+static ssize_t mem_control_type_read(struct cgroup *cont,
+				struct cftype *cft,
+				struct file *file, char __user *userbuf,
+				size_t nbytes, loff_t *ppos)
+{
+	unsigned long val;
+	char buf[64], *s;
+	struct mem_cgroup *mem;
+
+	mem = mem_cgroup_from_cont(cont);
+	s = buf;
+	val = mem->control_type;
+	s += sprintf(s, "%lu\n", val);
+	return simple_read_from_buffer((void __user *)userbuf, nbytes,
+			ppos, buf, s - buf);
+}
+
 static struct cftype mem_cgroup_files[] = {
 	{
 		.name = "usage_in_bytes",
@@ -442,6 +528,11 @@
 		.private = RES_FAILCNT,
 		.read = mem_cgroup_read,
 	},
+	{
+		.name = "control_type",
+		.write = mem_control_type_write,
+		.read = mem_control_type_read,
+	},
 };
  
 static struct mem_cgroup init_mem_cgroup;
@@ -464,6 +555,7 @@
 	INIT_LIST_HEAD(&mem->active_list);
 	INIT_LIST_HEAD(&mem->inactive_list);
 	spin_lock_init(&mem->lru_lock);
+	mem->control_type = MEM_CGROUP_TYPE_ALL;
 	return &mem->css;
 }
  
@@ -78,7 +78,7 @@
 	error = radix_tree_preload(gfp_mask);
 	if (!error) {
  
-		error = mem_cgroup_charge(page, current->mm);
+		error = mem_cgroup_cache_charge(page, current->mm);
 		if (error)
 			goto out;
...	...	@@ -22,6 +22,8 @@
22	22
23	23	struct mem_cgroup;
24	24	struct page_cgroup;
	25	+struct page;
	26	+struct mm_struct;
25	27
26	28	#ifdef CONFIG_CGROUP_MEM_CONT
27	29
...	...	@@ -40,6 +42,7 @@
40	42	struct mem_cgroup *mem_cont,
41	43	int active);
42	44	extern void mem_cgroup_out_of_memory(struct mem_cgroup *mem, gfp_t gfp_mask);
	45	+extern int mem_cgroup_cache_charge(struct page page, struct mm_struct mm);
43	46
44	47	static inline void mem_cgroup_uncharge_page(struct page *page)
45	48	{
...	...	@@ -82,6 +85,12 @@
82	85	static inline void mem_cgroup_move_lists(struct page_cgroup *pc,
83	86	bool active)
84	87	{
	88	+}
	89	+
	90	+static inline int mem_cgroup_cache_charge(struct page *page,
	91	+ struct mm_struct *mm)
	92	+{
	93	+ return 0;
85	94	}
86	95
87	96	#endif /* CONFIG_CGROUP_MEM_CONT */
...	...	@@ -464,7 +464,7 @@
464	464
465	465	if (error == 0) {
466	466
467		- error = mem_cgroup_charge(page, current->mm);
	467	+ error = mem_cgroup_cache_charge(page, current->mm);
468	468	if (error)
469	469	goto out;
470	470
...	...	@@ -29,6 +29,8 @@
29	29	#include <linux/spinlock.h>
30	30	#include <linux/fs.h>
31	31
	32	+#include <asm/uaccess.h>
	33	+
32	34	struct cgroup_subsys mem_cgroup_subsys;
33	35	static const int MEM_CGROUP_RECLAIM_RETRIES = 5;
34	36
...	...	@@ -60,6 +62,7 @@
60	62	* spin_lock to protect the per cgroup LRU
61	63	*/
62	64	spinlock_t lru_lock;
	65	+ unsigned long control_type; /* control RSS or RSS+Pagecache */
63	66	};
64	67
65	68	/*
66	69
...	...	@@ -82,7 +85,16 @@
82	85	/* mapped and cached states */
83	86	};
84	87
	88	+enum {
	89	+ MEM_CGROUP_TYPE_UNSPEC = 0,
	90	+ MEM_CGROUP_TYPE_MAPPED,
	91	+ MEM_CGROUP_TYPE_CACHED,
	92	+ MEM_CGROUP_TYPE_ALL,
	93	+ MEM_CGROUP_TYPE_MAX,
	94	+};
85	95
	96	+static struct mem_cgroup init_mem_cgroup;
	97	+
86	98	static inline
87	99	struct mem_cgroup mem_cgroup_from_cont(struct cgroup cont)
88	100	{
89	101
90	102
...	...	@@ -139,18 +151,18 @@
139	151	(page->page_cgroup & ~PAGE_CGROUP_LOCK);
140	152	}
141	153
142		-void __always_inline lock_page_cgroup(struct page *page)
	154	+static void __always_inline lock_page_cgroup(struct page *page)
143	155	{
144	156	bit_spin_lock(PAGE_CGROUP_LOCK_BIT, &page->page_cgroup);
145	157	VM_BUG_ON(!page_cgroup_locked(page));
146	158	}
147	159
148		-void __always_inline unlock_page_cgroup(struct page *page)
	160	+static void __always_inline unlock_page_cgroup(struct page *page)
149	161	{
150	162	bit_spin_unlock(PAGE_CGROUP_LOCK_BIT, &page->page_cgroup);
151	163	}
152	164
153		-void __mem_cgroup_move_lists(struct page_cgroup *pc, bool active)
	165	+static void __mem_cgroup_move_lists(struct page_cgroup *pc, bool active)
154	166	{
155	167	if (active)
156	168	list_move(&pc->lru, &pc->mem_cgroup->active_list);
...	...	@@ -366,6 +378,22 @@
366	378	}
367	379
368	380	/*
	381	+ * See if the cached pages should be charged at all?
	382	+ */
	383	+int mem_cgroup_cache_charge(struct page page, struct mm_struct mm)
	384	+{
	385	+ struct mem_cgroup *mem;
	386	+ if (!mm)
	387	+ mm = &init_mm;
	388	+
	389	+ mem = rcu_dereference(mm->mem_cgroup);
	390	+ if (mem->control_type == MEM_CGROUP_TYPE_ALL)
	391	+ return mem_cgroup_charge(page, mm);
	392	+ else
	393	+ return 0;
	394	+}
	395	+
	396	+/*
369	397	* Uncharging is always a welcome operation, we never complain, simply
370	398	* uncharge.
371	399	*/
...	...	@@ -375,6 +403,10 @@
375	403	struct page *page;
376	404	unsigned long flags;
377	405
	406	+ /*
	407	+ * This can handle cases when a page is not charged at all and we
	408	+ * are switching between handling the control_type.
	409	+ */
378	410	if (!pc)
379	411	return;
380	412
...	...	@@ -425,6 +457,60 @@
425	457	mem_cgroup_write_strategy);
426	458	}
427	459
	460	+static ssize_t mem_control_type_write(struct cgroup *cont,
	461	+ struct cftype cft, struct file file,
	462	+ const char __user *userbuf,
	463	+ size_t nbytes, loff_t *pos)
	464	+{
	465	+ int ret;
	466	+ char buf, end;
	467	+ unsigned long tmp;
	468	+ struct mem_cgroup *mem;
	469	+
	470	+ mem = mem_cgroup_from_cont(cont);
	471	+ buf = kmalloc(nbytes + 1, GFP_KERNEL);
	472	+ ret = -ENOMEM;
	473	+ if (buf == NULL)
	474	+ goto out;
	475	+
	476	+ buf[nbytes] = 0;
	477	+ ret = -EFAULT;
	478	+ if (copy_from_user(buf, userbuf, nbytes))
	479	+ goto out_free;
	480	+
	481	+ ret = -EINVAL;
	482	+ tmp = simple_strtoul(buf, &end, 10);
	483	+ if (*end != '\0')
	484	+ goto out_free;
	485	+
	486	+ if (tmp <= MEM_CGROUP_TYPE_UNSPEC \|\| tmp >= MEM_CGROUP_TYPE_MAX)
	487	+ goto out_free;
	488	+
	489	+ mem->control_type = tmp;
	490	+ ret = nbytes;
	491	+out_free:
	492	+ kfree(buf);
	493	+out:
	494	+ return ret;
	495	+}
	496	+
	497	+static ssize_t mem_control_type_read(struct cgroup *cont,
	498	+ struct cftype *cft,
	499	+ struct file file, char __user userbuf,
	500	+ size_t nbytes, loff_t *ppos)
	501	+{
	502	+ unsigned long val;
	503	+ char buf[64], *s;
	504	+ struct mem_cgroup *mem;
	505	+
	506	+ mem = mem_cgroup_from_cont(cont);
	507	+ s = buf;
	508	+ val = mem->control_type;
	509	+ s += sprintf(s, "%lu\n", val);
	510	+ return simple_read_from_buffer((void __user *)userbuf, nbytes,
	511	+ ppos, buf, s - buf);
	512	+}
	513	+
428	514	static struct cftype mem_cgroup_files[] = {
429	515	{
430	516	.name = "usage_in_bytes",
...	...	@@ -442,6 +528,11 @@
442	528	.private = RES_FAILCNT,
443	529	.read = mem_cgroup_read,
444	530	},
	531	+ {
	532	+ .name = "control_type",
	533	+ .write = mem_control_type_write,
	534	+ .read = mem_control_type_read,
	535	+ },
445	536	};
446	537
447	538	static struct mem_cgroup init_mem_cgroup;
...	...	@@ -464,6 +555,7 @@
464	555	INIT_LIST_HEAD(&mem->active_list);
465	556	INIT_LIST_HEAD(&mem->inactive_list);
466	557	spin_lock_init(&mem->lru_lock);
	558	+ mem->control_type = MEM_CGROUP_TYPE_ALL;
467	559	return &mem->css;
468	560	}
469	561
...	...	@@ -78,7 +78,7 @@
78	78	error = radix_tree_preload(gfp_mask);
79	79	if (!error) {
80	80
81		- error = mem_cgroup_charge(page, current->mm);
	81	+ error = mem_cgroup_cache_charge(page, current->mm);
82	82	if (error)
83	83	goto out;
84	84