[linux-2.6-block.git] / net / ipv4 / tcp_memcontrol.c

#include <net/tcp.h>
#include <net/tcp_memcontrol.h>
#include <net/sock.h>
#include <net/ip.h>
#include <linux/nsproxy.h>
#include <linux/memcontrol.h>
#include <linux/module.h>

static inline struct tcp_memcontrol *tcp_from_cgproto(struct cg_proto *cg_proto)
{
	return container_of(cg_proto, struct tcp_memcontrol, cg_proto);
}

static void memcg_tcp_enter_memory_pressure(struct sock *sk)
{
	if (sk->sk_cgrp->memory_pressure)
		*sk->sk_cgrp->memory_pressure = 1;
}
EXPORT_SYMBOL(memcg_tcp_enter_memory_pressure);

int tcp_init_cgroup(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
{
	/*
	 * The root cgroup does not use res_counters, but rather,
	 * rely on the data already collected by the network
	 * subsystem
	 */
	struct res_counter *res_parent = NULL;
	struct cg_proto *cg_proto, *parent_cg;
	struct tcp_memcontrol *tcp;
	struct mem_cgroup *parent = parent_mem_cgroup(memcg);

	cg_proto = tcp_prot.proto_cgroup(memcg);
	if (!cg_proto)
		return 0;

	tcp = tcp_from_cgproto(cg_proto);

	tcp->tcp_prot_mem[0] = sysctl_tcp_mem[0];
	tcp->tcp_prot_mem[1] = sysctl_tcp_mem[1];
	tcp->tcp_prot_mem[2] = sysctl_tcp_mem[2];
	tcp->tcp_memory_pressure = 0;

	parent_cg = tcp_prot.proto_cgroup(parent);
	if (parent_cg)
		res_parent = parent_cg->memory_allocated;

	res_counter_init(&tcp->tcp_memory_allocated, res_parent);
	percpu_counter_init(&tcp->tcp_sockets_allocated, 0);

	cg_proto->enter_memory_pressure = memcg_tcp_enter_memory_pressure;
	cg_proto->memory_pressure = &tcp->tcp_memory_pressure;
	cg_proto->sysctl_mem = tcp->tcp_prot_mem;
	cg_proto->memory_allocated = &tcp->tcp_memory_allocated;
	cg_proto->sockets_allocated = &tcp->tcp_sockets_allocated;
	cg_proto->memcg = memcg;

	return 0;
}
EXPORT_SYMBOL(tcp_init_cgroup);

void tcp_destroy_cgroup(struct mem_cgroup *memcg)
{
	struct cg_proto *cg_proto;
	struct tcp_memcontrol *tcp;

	cg_proto = tcp_prot.proto_cgroup(memcg);
	if (!cg_proto)
		return;

	tcp = tcp_from_cgproto(cg_proto);
	percpu_counter_destroy(&tcp->tcp_sockets_allocated);
}
EXPORT_SYMBOL(tcp_destroy_cgroup);

static int tcp_update_limit(struct mem_cgroup *memcg, u64 val)
{
	struct tcp_memcontrol *tcp;
	struct cg_proto *cg_proto;
	u64 old_lim;
	int i;
	int ret;

	cg_proto = tcp_prot.proto_cgroup(memcg);
	if (!cg_proto)
		return -EINVAL;

	if (val > RES_COUNTER_MAX)
		val = RES_COUNTER_MAX;

	tcp = tcp_from_cgproto(cg_proto);

	old_lim = res_counter_read_u64(&tcp->tcp_memory_allocated, RES_LIMIT);
	ret = res_counter_set_limit(&tcp->tcp_memory_allocated, val);
	if (ret)
		return ret;

	for (i = 0; i < 3; i++)
		tcp->tcp_prot_mem[i] = min_t(long, val >> PAGE_SHIFT,
					     sysctl_tcp_mem[i]);

	if (val == RES_COUNTER_MAX)
		clear_bit(MEMCG_SOCK_ACTIVE, &cg_proto->flags);
	else if (val != RES_COUNTER_MAX) {
		/*
		 * The active bit needs to be written after the static_key
		 * update. This is what guarantees that the socket activation
		 * function is the last one to run. See sock_update_memcg() for
		 * details, and note that we don't mark any socket as belonging
		 * to this memcg until that flag is up.
		 *
		 * We need to do this, because static_keys will span multiple
		 * sites, but we can't control their order. If we mark a socket
		 * as accounted, but the accounting functions are not patched in
		 * yet, we'll lose accounting.
		 *
		 * We never race with the readers in sock_update_memcg(),
		 * because when this value change, the code to process it is not
		 * patched in yet.
		 *
		 * The activated bit is used to guarantee that no two writers
		 * will do the update in the same memcg. Without that, we can't
		 * properly shutdown the static key.
		 */
		if (!test_and_set_bit(MEMCG_SOCK_ACTIVATED, &cg_proto->flags))
			static_key_slow_inc(&memcg_socket_limit_enabled);
		set_bit(MEMCG_SOCK_ACTIVE, &cg_proto->flags);
	}

	return 0;
}

static int tcp_cgroup_write(struct cgroup_subsys_state *css, struct cftype *cft,
			    const char *buffer)
{
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
	unsigned long long val;
	int ret = 0;

	switch (cft->private) {
	case RES_LIMIT:
		/* see memcontrol.c */
		ret = res_counter_memparse_write_strategy(buffer, &val);
		if (ret)
			break;
		ret = tcp_update_limit(memcg, val);
		break;
	default:
		ret = -EINVAL;
		break;
	}
	return ret;
}

static u64 tcp_read_stat(struct mem_cgroup *memcg, int type, u64 default_val)
{
	struct tcp_memcontrol *tcp;
	struct cg_proto *cg_proto;

	cg_proto = tcp_prot.proto_cgroup(memcg);
	if (!cg_proto)
		return default_val;

	tcp = tcp_from_cgproto(cg_proto);
	return res_counter_read_u64(&tcp->tcp_memory_allocated, type);
}

static u64 tcp_read_usage(struct mem_cgroup *memcg)
{
	struct tcp_memcontrol *tcp;
	struct cg_proto *cg_proto;

	cg_proto = tcp_prot.proto_cgroup(memcg);
	if (!cg_proto)
		return atomic_long_read(&tcp_memory_allocated) << PAGE_SHIFT;

	tcp = tcp_from_cgproto(cg_proto);
	return res_counter_read_u64(&tcp->tcp_memory_allocated, RES_USAGE);
}

static u64 tcp_cgroup_read(struct cgroup_subsys_state *css, struct cftype *cft)
{
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
	u64 val;

	switch (cft->private) {
	case RES_LIMIT:
		val = tcp_read_stat(memcg, RES_LIMIT, RES_COUNTER_MAX);
		break;
	case RES_USAGE:
		val = tcp_read_usage(memcg);
		break;
	case RES_FAILCNT:
	case RES_MAX_USAGE:
		val = tcp_read_stat(memcg, cft->private, 0);
		break;
	default:
		BUG();
	}
	return val;
}

static int tcp_cgroup_reset(struct cgroup_subsys_state *css, unsigned int event)
{
	struct mem_cgroup *memcg;
	struct tcp_memcontrol *tcp;
	struct cg_proto *cg_proto;

	memcg = mem_cgroup_from_css(css);
	cg_proto = tcp_prot.proto_cgroup(memcg);
	if (!cg_proto)
		return 0;
	tcp = tcp_from_cgproto(cg_proto);

	switch (event) {
	case RES_MAX_USAGE:
		res_counter_reset_max(&tcp->tcp_memory_allocated);
		break;
	case RES_FAILCNT:
		res_counter_reset_failcnt(&tcp->tcp_memory_allocated);
		break;
	}

	return 0;
}

static struct cftype tcp_files[] = {
	{
		.name = "kmem.tcp.limit_in_bytes",
		.write_string = tcp_cgroup_write,
		.read_u64 = tcp_cgroup_read,
		.private = RES_LIMIT,
	},
	{
		.name = "kmem.tcp.usage_in_bytes",
		.read_u64 = tcp_cgroup_read,
		.private = RES_USAGE,
	},
	{
		.name = "kmem.tcp.failcnt",
		.private = RES_FAILCNT,
		.trigger = tcp_cgroup_reset,
		.read_u64 = tcp_cgroup_read,
	},
	{
		.name = "kmem.tcp.max_usage_in_bytes",
		.private = RES_MAX_USAGE,
		.trigger = tcp_cgroup_reset,
		.read_u64 = tcp_cgroup_read,
	},
	{ }	/* terminate */
};

static int __init tcp_memcontrol_init(void)
{
	WARN_ON(cgroup_add_cftypes(&mem_cgroup_subsys, tcp_files));
	return 0;
}
__initcall(tcp_memcontrol_init);
Commit	Line	Data
d1a4c0b3 GC	1	#include <net/tcp.h>
	2	#include <net/tcp_memcontrol.h>
	3	#include <net/sock.h>
3dc43e3e GC	4	#include <net/ip.h>
3dc43e3e GC	5	#include <linux/nsproxy.h>
d1a4c0b3 GC	6	#include <linux/memcontrol.h>
	7	#include <linux/module.h>
	8
	9	static inline struct tcp_memcontrol tcp_from_cgproto(struct cg_proto cg_proto)
	10	{
	11	return container_of(cg_proto, struct tcp_memcontrol, cg_proto);
	12	}
	13
	14	static void memcg_tcp_enter_memory_pressure(struct sock *sk)
	15	{
c48e074c	16	if (sk->sk_cgrp->memory_pressure)
d1a4c0b3 GC	17	*sk->sk_cgrp->memory_pressure = 1;
	18	}
	19	EXPORT_SYMBOL(memcg_tcp_enter_memory_pressure);
	20
1d62e436	21	int tcp_init_cgroup(struct mem_cgroup memcg, struct cgroup_subsys ss)
d1a4c0b3 GC	22	{
	23	/*
	24	* The root cgroup does not use res_counters, but rather,
	25	* rely on the data already collected by the network
	26	* subsystem
	27	*/
	28	struct res_counter *res_parent = NULL;
	29	struct cg_proto cg_proto, parent_cg;
	30	struct tcp_memcontrol *tcp;
d1a4c0b3 GC	31	struct mem_cgroup *parent = parent_mem_cgroup(memcg);
	32
	33	cg_proto = tcp_prot.proto_cgroup(memcg);
	34	if (!cg_proto)
6bc10349	35	return 0;
d1a4c0b3 GC	36
	37	tcp = tcp_from_cgproto(cg_proto);
	38
a4fe34bf EB	39	tcp->tcp_prot_mem[0] = sysctl_tcp_mem[0];
	40	tcp->tcp_prot_mem[1] = sysctl_tcp_mem[1];
	41	tcp->tcp_prot_mem[2] = sysctl_tcp_mem[2];
d1a4c0b3 GC	42	tcp->tcp_memory_pressure = 0;
	43
	44	parent_cg = tcp_prot.proto_cgroup(parent);
	45	if (parent_cg)
	46	res_parent = parent_cg->memory_allocated;
	47
	48	res_counter_init(&tcp->tcp_memory_allocated, res_parent);
	49	percpu_counter_init(&tcp->tcp_sockets_allocated, 0);
	50
	51	cg_proto->enter_memory_pressure = memcg_tcp_enter_memory_pressure;
	52	cg_proto->memory_pressure = &tcp->tcp_memory_pressure;
	53	cg_proto->sysctl_mem = tcp->tcp_prot_mem;
	54	cg_proto->memory_allocated = &tcp->tcp_memory_allocated;
	55	cg_proto->sockets_allocated = &tcp->tcp_sockets_allocated;
	56	cg_proto->memcg = memcg;
	57
6bc10349	58	return 0;
d1a4c0b3 GC	59	}
	60	EXPORT_SYMBOL(tcp_init_cgroup);
	61
1d62e436	62	void tcp_destroy_cgroup(struct mem_cgroup *memcg)
d1a4c0b3	63	{
d1a4c0b3 GC	64	struct cg_proto *cg_proto;
	65	struct tcp_memcontrol *tcp;
	66
	67	cg_proto = tcp_prot.proto_cgroup(memcg);
	68	if (!cg_proto)
	69	return;
	70
	71	tcp = tcp_from_cgproto(cg_proto);
	72	percpu_counter_destroy(&tcp->tcp_sockets_allocated);
	73	}
	74	EXPORT_SYMBOL(tcp_destroy_cgroup);
3aaabe23 GC	75
	76	static int tcp_update_limit(struct mem_cgroup *memcg, u64 val)
	77	{
3aaabe23 GC	78	struct tcp_memcontrol *tcp;
	79	struct cg_proto *cg_proto;
	80	u64 old_lim;
	81	int i;
	82	int ret;
	83
	84	cg_proto = tcp_prot.proto_cgroup(memcg);
	85	if (!cg_proto)
	86	return -EINVAL;
	87
6de5a8bf SZ	88	if (val > RES_COUNTER_MAX)
6de5a8bf SZ	89	val = RES_COUNTER_MAX;
3aaabe23 GC	90
	91	tcp = tcp_from_cgproto(cg_proto);
	92
	93	old_lim = res_counter_read_u64(&tcp->tcp_memory_allocated, RES_LIMIT);
	94	ret = res_counter_set_limit(&tcp->tcp_memory_allocated, val);
	95	if (ret)
	96	return ret;
	97
	98	for (i = 0; i < 3; i++)
	99	tcp->tcp_prot_mem[i] = min_t(long, val >> PAGE_SHIFT,
a4fe34bf	100	sysctl_tcp_mem[i]);
3aaabe23	101
6de5a8bf	102	if (val == RES_COUNTER_MAX)
3f134619	103	clear_bit(MEMCG_SOCK_ACTIVE, &cg_proto->flags);
6de5a8bf	104	else if (val != RES_COUNTER_MAX) {
3f134619 GC	105	/*
	106	* The active bit needs to be written after the static_key
	107	* update. This is what guarantees that the socket activation
	108	* function is the last one to run. See sock_update_memcg() for
	109	* details, and note that we don't mark any socket as belonging
	110	* to this memcg until that flag is up.
	111	*
	112	* We need to do this, because static_keys will span multiple
	113	* sites, but we can't control their order. If we mark a socket
	114	* as accounted, but the accounting functions are not patched in
	115	* yet, we'll lose accounting.
	116	*
	117	* We never race with the readers in sock_update_memcg(),
	118	* because when this value change, the code to process it is not
	119	* patched in yet.
	120	*
	121	* The activated bit is used to guarantee that no two writers
	122	* will do the update in the same memcg. Without that, we can't
	123	* properly shutdown the static key.
	124	*/
	125	if (!test_and_set_bit(MEMCG_SOCK_ACTIVATED, &cg_proto->flags))
	126	static_key_slow_inc(&memcg_socket_limit_enabled);
	127	set_bit(MEMCG_SOCK_ACTIVE, &cg_proto->flags);
	128	}
3aaabe23 GC	129
	130	return 0;
	131	}
	132
182446d0	133	static int tcp_cgroup_write(struct cgroup_subsys_state css, struct cftype cft,
3aaabe23 GC	134	const char *buffer)
3aaabe23 GC	135	{
182446d0	136	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
3aaabe23 GC	137	unsigned long long val;
	138	int ret = 0;
	139
	140	switch (cft->private) {
	141	case RES_LIMIT:
	142	/* see memcontrol.c */
	143	ret = res_counter_memparse_write_strategy(buffer, &val);
	144	if (ret)
	145	break;
	146	ret = tcp_update_limit(memcg, val);
	147	break;
	148	default:
	149	ret = -EINVAL;
	150	break;
	151	}
	152	return ret;
	153	}
	154
	155	static u64 tcp_read_stat(struct mem_cgroup *memcg, int type, u64 default_val)
	156	{
	157	struct tcp_memcontrol *tcp;
	158	struct cg_proto *cg_proto;
	159
	160	cg_proto = tcp_prot.proto_cgroup(memcg);
	161	if (!cg_proto)
	162	return default_val;
	163
	164	tcp = tcp_from_cgproto(cg_proto);
	165	return res_counter_read_u64(&tcp->tcp_memory_allocated, type);
	166	}
	167
5a6dd343 GC	168	static u64 tcp_read_usage(struct mem_cgroup *memcg)
	169	{
	170	struct tcp_memcontrol *tcp;
	171	struct cg_proto *cg_proto;
	172
	173	cg_proto = tcp_prot.proto_cgroup(memcg);
	174	if (!cg_proto)
	175	return atomic_long_read(&tcp_memory_allocated) << PAGE_SHIFT;
	176
	177	tcp = tcp_from_cgproto(cg_proto);
	178	return res_counter_read_u64(&tcp->tcp_memory_allocated, RES_USAGE);
	179	}
	180
182446d0	181	static u64 tcp_cgroup_read(struct cgroup_subsys_state css, struct cftype cft)
3aaabe23	182	{
182446d0	183	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
3aaabe23 GC	184	u64 val;
	185
	186	switch (cft->private) {
	187	case RES_LIMIT:
6de5a8bf	188	val = tcp_read_stat(memcg, RES_LIMIT, RES_COUNTER_MAX);
3aaabe23	189	break;
5a6dd343 GC	190	case RES_USAGE:
	191	val = tcp_read_usage(memcg);
	192	break;
ffea59e5	193	case RES_FAILCNT:
0850f0f5 GC	194	case RES_MAX_USAGE:
0850f0f5 GC	195	val = tcp_read_stat(memcg, cft->private, 0);
ffea59e5	196	break;
3aaabe23 GC	197	default:
	198	BUG();
	199	}
	200	return val;
	201	}
	202
182446d0	203	static int tcp_cgroup_reset(struct cgroup_subsys_state *css, unsigned int event)
ffea59e5 GC	204	{
	205	struct mem_cgroup *memcg;
	206	struct tcp_memcontrol *tcp;
	207	struct cg_proto *cg_proto;
	208
182446d0	209	memcg = mem_cgroup_from_css(css);
ffea59e5 GC	210	cg_proto = tcp_prot.proto_cgroup(memcg);
	211	if (!cg_proto)
	212	return 0;
	213	tcp = tcp_from_cgproto(cg_proto);
	214
	215	switch (event) {
0850f0f5 GC	216	case RES_MAX_USAGE:
	217	res_counter_reset_max(&tcp->tcp_memory_allocated);
	218	break;
ffea59e5 GC	219	case RES_FAILCNT:
	220	res_counter_reset_failcnt(&tcp->tcp_memory_allocated);
	221	break;
	222	}
	223
	224	return 0;
	225	}
	226
676f7c8f TH	227	static struct cftype tcp_files[] = {
	228	{
	229	.name = "kmem.tcp.limit_in_bytes",
	230	.write_string = tcp_cgroup_write,
	231	.read_u64 = tcp_cgroup_read,
	232	.private = RES_LIMIT,
	233	},
	234	{
	235	.name = "kmem.tcp.usage_in_bytes",
	236	.read_u64 = tcp_cgroup_read,
	237	.private = RES_USAGE,
	238	},
	239	{
	240	.name = "kmem.tcp.failcnt",
	241	.private = RES_FAILCNT,
	242	.trigger = tcp_cgroup_reset,
	243	.read_u64 = tcp_cgroup_read,
	244	},
	245	{
	246	.name = "kmem.tcp.max_usage_in_bytes",
	247	.private = RES_MAX_USAGE,
	248	.trigger = tcp_cgroup_reset,
	249	.read_u64 = tcp_cgroup_read,
	250	},
6bc10349	251	{ } /* terminate */
676f7c8f	252	};
6bc10349 TH	253
	254	static int __init tcp_memcontrol_init(void)
	255	{
	256	WARN_ON(cgroup_add_cftypes(&mem_cgroup_subsys, tcp_files));
	257	return 0;
	258	}
	259	__initcall(tcp_memcontrol_init);