[GitHub/mt8127/android_kernel_alcatel_ttab.git] / net / core / drop_monitor.c

/*
 * Monitoring code for network dropped packet alerts
 *
 * Copyright (C) 2009 Neil Horman <nhorman@tuxdriver.com>
 */

#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

#include <linux/netdevice.h>
#include <linux/etherdevice.h>
#include <linux/string.h>
#include <linux/if_arp.h>
#include <linux/inetdevice.h>
#include <linux/inet.h>
#include <linux/interrupt.h>
#include <linux/netpoll.h>
#include <linux/sched.h>
#include <linux/delay.h>
#include <linux/types.h>
#include <linux/workqueue.h>
#include <linux/netlink.h>
#include <linux/net_dropmon.h>
#include <linux/percpu.h>
#include <linux/timer.h>
#include <linux/bitops.h>
#include <linux/slab.h>
#include <linux/module.h>
#include <net/genetlink.h>
#include <net/netevent.h>

#include <trace/events/skb.h>
#include <trace/events/napi.h>

#include <asm/unaligned.h>

#define TRACE_ON 1
#define TRACE_OFF 0

/*
 * Globals, our netlink socket pointer
 * and the work handle that will send up
 * netlink alerts
 */
static int trace_state = TRACE_OFF;
static DEFINE_MUTEX(trace_state_mutex);

struct per_cpu_dm_data {
	spinlock_t		lock;
	struct sk_buff		*skb;
	struct work_struct	dm_alert_work;
	struct timer_list	send_timer;
};

struct dm_hw_stat_delta {
	struct net_device *dev;
	unsigned long last_rx;
	struct list_head list;
	struct rcu_head rcu;
	unsigned long last_drop_val;
};

static struct genl_family net_drop_monitor_family = {
	.id             = GENL_ID_GENERATE,
	.hdrsize        = 0,
	.name           = "NET_DM",
	.version        = 2,
};

static DEFINE_PER_CPU(struct per_cpu_dm_data, dm_cpu_data);

static int dm_hit_limit = 64;
static int dm_delay = 1;
static unsigned long dm_hw_check_delta = 2*HZ;
static LIST_HEAD(hw_stats_list);

static struct sk_buff *reset_per_cpu_data(struct per_cpu_dm_data *data)
{
	size_t al;
	struct net_dm_alert_msg *msg;
	struct nlattr *nla;
	struct sk_buff *skb;
	unsigned long flags;
	void *msg_header;

	al = sizeof(struct net_dm_alert_msg);
	al += dm_hit_limit * sizeof(struct net_dm_drop_point);
	al += sizeof(struct nlattr);

	skb = genlmsg_new(al, GFP_KERNEL);

	if (!skb)
		goto err;

	msg_header = genlmsg_put(skb, 0, 0, &net_drop_monitor_family,
				 0, NET_DM_CMD_ALERT);
	if (!msg_header) {
		nlmsg_free(skb);
		skb = NULL;
		goto err;
	}
	nla = nla_reserve(skb, NLA_UNSPEC,
			  sizeof(struct net_dm_alert_msg));
	if (!nla) {
		nlmsg_free(skb);
		skb = NULL;
		goto err;
	}
	msg = nla_data(nla);
	memset(msg, 0, al);
	goto out;

err:
	mod_timer(&data->send_timer, jiffies + HZ / 10);
out:
	spin_lock_irqsave(&data->lock, flags);
	swap(data->skb, skb);
	spin_unlock_irqrestore(&data->lock, flags);

	if (skb) {
		struct nlmsghdr *nlh = (struct nlmsghdr *)skb->data;
		struct genlmsghdr *gnlh = (struct genlmsghdr *)nlmsg_data(nlh);

		genlmsg_end(skb, genlmsg_data(gnlh));
	}

	return skb;
}

static void send_dm_alert(struct work_struct *work)
{
	struct sk_buff *skb;
	struct per_cpu_dm_data *data;

	data = container_of(work, struct per_cpu_dm_data, dm_alert_work);

	skb = reset_per_cpu_data(data);

	if (skb)
		genlmsg_multicast(skb, 0, NET_DM_GRP_ALERT, GFP_KERNEL);
}

/*
 * This is the timer function to delay the sending of an alert
 * in the event that more drops will arrive during the
 * hysteresis period.
 */
static void sched_send_work(unsigned long _data)
{
	struct per_cpu_dm_data *data = (struct per_cpu_dm_data *)_data;

	schedule_work(&data->dm_alert_work);
}

static void trace_drop_common(struct sk_buff *skb, void *location)
{
	struct net_dm_alert_msg *msg;
	struct nlmsghdr *nlh;
	struct nlattr *nla;
	int i;
	struct sk_buff *dskb;
	struct per_cpu_dm_data *data;
	unsigned long flags;

	local_irq_save(flags);
	data = &__get_cpu_var(dm_cpu_data);
	spin_lock(&data->lock);
	dskb = data->skb;

	if (!dskb)
		goto out;

	nlh = (struct nlmsghdr *)dskb->data;
	nla = genlmsg_data(nlmsg_data(nlh));
	msg = nla_data(nla);
	for (i = 0; i < msg->entries; i++) {
		if (!memcmp(&location, msg->points[i].pc, sizeof(void *))) {
			msg->points[i].count++;
			goto out;
		}
	}
	if (msg->entries == dm_hit_limit)
		goto out;
	/*
	 * We need to create a new entry
	 */
	__nla_reserve_nohdr(dskb, sizeof(struct net_dm_drop_point));
	nla->nla_len += NLA_ALIGN(sizeof(struct net_dm_drop_point));
	memcpy(msg->points[msg->entries].pc, &location, sizeof(void *));
	msg->points[msg->entries].count = 1;
	msg->entries++;

	if (!timer_pending(&data->send_timer)) {
		data->send_timer.expires = jiffies + dm_delay * HZ;
		add_timer(&data->send_timer);
	}

out:
	spin_unlock_irqrestore(&data->lock, flags);
}

static void trace_kfree_skb_hit(void *ignore, struct sk_buff *skb, void *location)
{
	trace_drop_common(skb, location);
}

static void trace_napi_poll_hit(void *ignore, struct napi_struct *napi)
{
	struct dm_hw_stat_delta *new_stat;

	/*
	 * Don't check napi structures with no associated device
	 */
	if (!napi->dev)
		return;

	rcu_read_lock();
	list_for_each_entry_rcu(new_stat, &hw_stats_list, list) {
		/*
		 * only add a note to our monitor buffer if:
		 * 1) this is the dev we received on
		 * 2) its after the last_rx delta
		 * 3) our rx_dropped count has gone up
		 */
		if ((new_stat->dev == napi->dev)  &&
		    (time_after(jiffies, new_stat->last_rx + dm_hw_check_delta)) &&
		    (napi->dev->stats.rx_dropped != new_stat->last_drop_val)) {
			trace_drop_common(NULL, NULL);
			new_stat->last_drop_val = napi->dev->stats.rx_dropped;
			new_stat->last_rx = jiffies;
			break;
		}
	}
	rcu_read_unlock();
}

static int set_all_monitor_traces(int state)
{
	int rc = 0;
	struct dm_hw_stat_delta *new_stat = NULL;
	struct dm_hw_stat_delta *temp;

	mutex_lock(&trace_state_mutex);

	if (state == trace_state) {
		rc = -EAGAIN;
		goto out_unlock;
	}

	switch (state) {
	case TRACE_ON:
		if (!try_module_get(THIS_MODULE)) {
			rc = -ENODEV;
			break;
		}

		rc |= register_trace_kfree_skb(trace_kfree_skb_hit, NULL);
		rc |= register_trace_napi_poll(trace_napi_poll_hit, NULL);
		break;

	case TRACE_OFF:
		rc |= unregister_trace_kfree_skb(trace_kfree_skb_hit, NULL);
		rc |= unregister_trace_napi_poll(trace_napi_poll_hit, NULL);

		tracepoint_synchronize_unregister();

		/*
		 * Clean the device list
		 */
		list_for_each_entry_safe(new_stat, temp, &hw_stats_list, list) {
			if (new_stat->dev == NULL) {
				list_del_rcu(&new_stat->list);
				kfree_rcu(new_stat, rcu);
			}
		}

		module_put(THIS_MODULE);

		break;
	default:
		rc = 1;
		break;
	}

	if (!rc)
		trace_state = state;
	else
		rc = -EINPROGRESS;

out_unlock:
	mutex_unlock(&trace_state_mutex);

	return rc;
}


static int net_dm_cmd_config(struct sk_buff *skb,
			struct genl_info *info)
{
	return -ENOTSUPP;
}

static int net_dm_cmd_trace(struct sk_buff *skb,
			struct genl_info *info)
{
	switch (info->genlhdr->cmd) {
	case NET_DM_CMD_START:
		return set_all_monitor_traces(TRACE_ON);
		break;
	case NET_DM_CMD_STOP:
		return set_all_monitor_traces(TRACE_OFF);
		break;
	}

	return -ENOTSUPP;
}

static int dropmon_net_event(struct notifier_block *ev_block,
			unsigned long event, void *ptr)
{
	struct net_device *dev = ptr;
	struct dm_hw_stat_delta *new_stat = NULL;
	struct dm_hw_stat_delta *tmp;

	switch (event) {
	case NETDEV_REGISTER:
		new_stat = kzalloc(sizeof(struct dm_hw_stat_delta), GFP_KERNEL);

		if (!new_stat)
			goto out;

		new_stat->dev = dev;
		new_stat->last_rx = jiffies;
		mutex_lock(&trace_state_mutex);
		list_add_rcu(&new_stat->list, &hw_stats_list);
		mutex_unlock(&trace_state_mutex);
		break;
	case NETDEV_UNREGISTER:
		mutex_lock(&trace_state_mutex);
		list_for_each_entry_safe(new_stat, tmp, &hw_stats_list, list) {
			if (new_stat->dev == dev) {
				new_stat->dev = NULL;
				if (trace_state == TRACE_OFF) {
					list_del_rcu(&new_stat->list);
					kfree_rcu(new_stat, rcu);
					break;
				}
			}
		}
		mutex_unlock(&trace_state_mutex);
		break;
	}
out:
	return NOTIFY_DONE;
}

static struct genl_ops dropmon_ops[] = {
	{
		.cmd = NET_DM_CMD_CONFIG,
		.doit = net_dm_cmd_config,
	},
	{
		.cmd = NET_DM_CMD_START,
		.doit = net_dm_cmd_trace,
	},
	{
		.cmd = NET_DM_CMD_STOP,
		.doit = net_dm_cmd_trace,
	},
};

static struct notifier_block dropmon_net_notifier = {
	.notifier_call = dropmon_net_event
};

static int __init init_net_drop_monitor(void)
{
	struct per_cpu_dm_data *data;
	int cpu, rc;

	pr_info("Initializing network drop monitor service\n");

	if (sizeof(void *) > 8) {
		pr_err("Unable to store program counters on this arch, Drop monitor failed\n");
		return -ENOSPC;
	}

	rc = genl_register_family_with_ops(&net_drop_monitor_family,
					   dropmon_ops,
					   ARRAY_SIZE(dropmon_ops));
	if (rc) {
		pr_err("Could not create drop monitor netlink family\n");
		return rc;
	}

	rc = register_netdevice_notifier(&dropmon_net_notifier);
	if (rc < 0) {
		pr_crit("Failed to register netdevice notifier\n");
		goto out_unreg;
	}

	rc = 0;

	for_each_possible_cpu(cpu) {
		data = &per_cpu(dm_cpu_data, cpu);
		INIT_WORK(&data->dm_alert_work, send_dm_alert);
		init_timer(&data->send_timer);
		data->send_timer.data = (unsigned long)data;
		data->send_timer.function = sched_send_work;
		spin_lock_init(&data->lock);
		reset_per_cpu_data(data);
	}


	goto out;

out_unreg:
	genl_unregister_family(&net_drop_monitor_family);
out:
	return rc;
}

static void exit_net_drop_monitor(void)
{
	struct per_cpu_dm_data *data;
	int cpu;

	BUG_ON(unregister_netdevice_notifier(&dropmon_net_notifier));

	/*
	 * Because of the module_get/put we do in the trace state change path
	 * we are guarnateed not to have any current users when we get here
	 * all we need to do is make sure that we don't have any running timers
	 * or pending schedule calls
	 */

	for_each_possible_cpu(cpu) {
		data = &per_cpu(dm_cpu_data, cpu);
		del_timer_sync(&data->send_timer);
		cancel_work_sync(&data->dm_alert_work);
		/*
		 * At this point, we should have exclusive access
		 * to this struct and can free the skb inside it
		 */
		kfree_skb(data->skb);
	}

	BUG_ON(genl_unregister_family(&net_drop_monitor_family));
}

module_init(init_net_drop_monitor);
module_exit(exit_net_drop_monitor);

MODULE_LICENSE("GPL v2");
MODULE_AUTHOR("Neil Horman <nhorman@tuxdriver.com>");
MODULE_ALIAS_GENL_FAMILY("NET_DM");
Commit	Line	Data
9a8afc8d NH	1	/*
	2	* Monitoring code for network dropped packet alerts
	3	*
	4	* Copyright (C) 2009 Neil Horman <nhorman@tuxdriver.com>
	5	*/
	6
e005d193 JP	7	#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
e005d193 JP	8
9a8afc8d NH	9	#include <linux/netdevice.h>
	10	#include <linux/etherdevice.h>
	11	#include <linux/string.h>
	12	#include <linux/if_arp.h>
	13	#include <linux/inetdevice.h>
	14	#include <linux/inet.h>
	15	#include <linux/interrupt.h>
	16	#include <linux/netpoll.h>
	17	#include <linux/sched.h>
	18	#include <linux/delay.h>
	19	#include <linux/types.h>
	20	#include <linux/workqueue.h>
	21	#include <linux/netlink.h>
	22	#include <linux/net_dropmon.h>
	23	#include <linux/percpu.h>
	24	#include <linux/timer.h>
	25	#include <linux/bitops.h>
5a0e3ad6	26	#include <linux/slab.h>
cad456d5	27	#include <linux/module.h>
9a8afc8d	28	#include <net/genetlink.h>
4ea7e386	29	#include <net/netevent.h>
9a8afc8d	30
ad8d75ff	31	#include <trace/events/skb.h>
9cbc1cb8	32	#include <trace/events/napi.h>
9a8afc8d NH	33
	34	#include <asm/unaligned.h>
	35
	36	#define TRACE_ON 1
	37	#define TRACE_OFF 0
	38
9a8afc8d NH	39	/*
	40	* Globals, our netlink socket pointer
	41	* and the work handle that will send up
	42	* netlink alerts
	43	*/
4ea7e386	44	static int trace_state = TRACE_OFF;
cde2e9a6	45	static DEFINE_MUTEX(trace_state_mutex);
9a8afc8d NH	46
9a8afc8d NH	47	struct per_cpu_dm_data {
bec4596b ED	48	spinlock_t lock;
	49	struct sk_buff *skb;
	50	struct work_struct dm_alert_work;
	51	struct timer_list send_timer;
9a8afc8d NH	52	};
9a8afc8d NH	53
4ea7e386 NH	54	struct dm_hw_stat_delta {
4ea7e386 NH	55	struct net_device *dev;
5848cc09	56	unsigned long last_rx;
4ea7e386 NH	57	struct list_head list;
	58	struct rcu_head rcu;
	59	unsigned long last_drop_val;
	60	};
	61
9a8afc8d NH	62	static struct genl_family net_drop_monitor_family = {
	63	.id = GENL_ID_GENERATE,
	64	.hdrsize = 0,
	65	.name = "NET_DM",
683703a2	66	.version = 2,
9a8afc8d NH	67	};
	68
	69	static DEFINE_PER_CPU(struct per_cpu_dm_data, dm_cpu_data);
	70
	71	static int dm_hit_limit = 64;
	72	static int dm_delay = 1;
4ea7e386 NH	73	static unsigned long dm_hw_check_delta = 2*HZ;
4ea7e386 NH	74	static LIST_HEAD(hw_stats_list);
9a8afc8d	75
bec4596b	76	static struct sk_buff reset_per_cpu_data(struct per_cpu_dm_data data)
9a8afc8d NH	77	{
	78	size_t al;
	79	struct net_dm_alert_msg *msg;
683703a2	80	struct nlattr *nla;
3885ca78	81	struct sk_buff *skb;
bec4596b	82	unsigned long flags;
0cfc0624	83	void *msg_header;
9a8afc8d NH	84
	85	al = sizeof(struct net_dm_alert_msg);
	86	al += dm_hit_limit * sizeof(struct net_dm_drop_point);
683703a2 NH	87	al += sizeof(struct nlattr);
683703a2 NH	88
3885ca78 NH	89	skb = genlmsg_new(al, GFP_KERNEL);
3885ca78 NH	90
0cfc0624 RW	91	if (!skb)
0cfc0624 RW	92	goto err;
3885ca78	93
0cfc0624 RW	94	msg_header = genlmsg_put(skb, 0, 0, &net_drop_monitor_family,
	95	0, NET_DM_CMD_ALERT);
	96	if (!msg_header) {
	97	nlmsg_free(skb);
	98	skb = NULL;
	99	goto err;
	100	}
	101	nla = nla_reserve(skb, NLA_UNSPEC,
	102	sizeof(struct net_dm_alert_msg));
	103	if (!nla) {
	104	nlmsg_free(skb);
	105	skb = NULL;
	106	goto err;
	107	}
	108	msg = nla_data(nla);
	109	memset(msg, 0, al);
0cfc0624 RW	110	goto out;
	111
	112	err:
	113	mod_timer(&data->send_timer, jiffies + HZ / 10);
	114	out:
bec4596b ED	115	spin_lock_irqsave(&data->lock, flags);
	116	swap(data->skb, skb);
	117	spin_unlock_irqrestore(&data->lock, flags);
	118
e88f37a1 RW	119	if (skb) {
	120	struct nlmsghdr nlh = (struct nlmsghdr )skb->data;
	121	struct genlmsghdr gnlh = (struct genlmsghdr )nlmsg_data(nlh);
	122
	123	genlmsg_end(skb, genlmsg_data(gnlh));
	124	}
	125
bec4596b	126	return skb;
9a8afc8d NH	127	}
9a8afc8d NH	128
bec4596b	129	static void send_dm_alert(struct work_struct *work)
9a8afc8d NH	130	{
9a8afc8d NH	131	struct sk_buff *skb;
bec4596b	132	struct per_cpu_dm_data *data;
9a8afc8d	133
bec4596b	134	data = container_of(work, struct per_cpu_dm_data, dm_alert_work);
4fdcfa12	135
bec4596b	136	skb = reset_per_cpu_data(data);
9a8afc8d	137
3885ca78 NH	138	if (skb)
3885ca78 NH	139	genlmsg_multicast(skb, 0, NET_DM_GRP_ALERT, GFP_KERNEL);
9a8afc8d NH	140	}
	141
	142	/*
	143	* This is the timer function to delay the sending of an alert
	144	* in the event that more drops will arrive during the
bec4596b	145	* hysteresis period.
9a8afc8d	146	*/
bec4596b	147	static void sched_send_work(unsigned long _data)
9a8afc8d	148	{
bec4596b	149	struct per_cpu_dm_data data = (struct per_cpu_dm_data )_data;
3885ca78	150
bec4596b	151	schedule_work(&data->dm_alert_work);
9a8afc8d NH	152	}
9a8afc8d NH	153
4ea7e386	154	static void trace_drop_common(struct sk_buff skb, void location)
9a8afc8d NH	155	{
	156	struct net_dm_alert_msg *msg;
	157	struct nlmsghdr *nlh;
683703a2	158	struct nlattr *nla;
9a8afc8d	159	int i;
3885ca78	160	struct sk_buff *dskb;
bec4596b ED	161	struct per_cpu_dm_data *data;
bec4596b ED	162	unsigned long flags;
9a8afc8d	163
bec4596b ED	164	local_irq_save(flags);
	165	data = &__get_cpu_var(dm_cpu_data);
	166	spin_lock(&data->lock);
	167	dskb = data->skb;
3885ca78 NH	168
	169	if (!dskb)
	170	goto out;
	171
3885ca78	172	nlh = (struct nlmsghdr *)dskb->data;
683703a2 NH	173	nla = genlmsg_data(nlmsg_data(nlh));
683703a2 NH	174	msg = nla_data(nla);
9a8afc8d NH	175	for (i = 0; i < msg->entries; i++) {
	176	if (!memcmp(&location, msg->points[i].pc, sizeof(void *))) {
	177	msg->points[i].count++;
	178	goto out;
	179	}
	180	}
bec4596b ED	181	if (msg->entries == dm_hit_limit)
bec4596b ED	182	goto out;
9a8afc8d NH	183	/*
	184	* We need to create a new entry
	185	*/
3885ca78	186	__nla_reserve_nohdr(dskb, sizeof(struct net_dm_drop_point));
683703a2	187	nla->nla_len += NLA_ALIGN(sizeof(struct net_dm_drop_point));
9a8afc8d NH	188	memcpy(msg->points[msg->entries].pc, &location, sizeof(void *));
	189	msg->points[msg->entries].count = 1;
	190	msg->entries++;
	191
	192	if (!timer_pending(&data->send_timer)) {
	193	data->send_timer.expires = jiffies + dm_delay * HZ;
bec4596b	194	add_timer(&data->send_timer);
9a8afc8d NH	195	}
	196
	197	out:
bec4596b	198	spin_unlock_irqrestore(&data->lock, flags);
9a8afc8d NH	199	}
9a8afc8d NH	200
38516ab5	201	static void trace_kfree_skb_hit(void ignore, struct sk_buff skb, void *location)
4ea7e386 NH	202	{
	203	trace_drop_common(skb, location);
	204	}
	205
38516ab5	206	static void trace_napi_poll_hit(void ignore, struct napi_struct napi)
4ea7e386 NH	207	{
	208	struct dm_hw_stat_delta *new_stat;
	209
	210	/*
5848cc09	211	* Don't check napi structures with no associated device
4ea7e386	212	*/
5848cc09	213	if (!napi->dev)
4ea7e386 NH	214	return;
	215
	216	rcu_read_lock();
	217	list_for_each_entry_rcu(new_stat, &hw_stats_list, list) {
5848cc09 NH	218	/*
	219	* only add a note to our monitor buffer if:
	220	* 1) this is the dev we received on
	221	* 2) its after the last_rx delta
	222	* 3) our rx_dropped count has gone up
	223	*/
4ea7e386	224	if ((new_stat->dev == napi->dev) &&
5848cc09	225	(time_after(jiffies, new_stat->last_rx + dm_hw_check_delta)) &&
4ea7e386 NH	226	(napi->dev->stats.rx_dropped != new_stat->last_drop_val)) {
	227	trace_drop_common(NULL, NULL);
	228	new_stat->last_drop_val = napi->dev->stats.rx_dropped;
5848cc09	229	new_stat->last_rx = jiffies;
4ea7e386 NH	230	break;
	231	}
	232	}
	233	rcu_read_unlock();
	234	}
	235
9a8afc8d NH	236	static int set_all_monitor_traces(int state)
	237	{
	238	int rc = 0;
4ea7e386 NH	239	struct dm_hw_stat_delta *new_stat = NULL;
	240	struct dm_hw_stat_delta *temp;
	241
cde2e9a6	242	mutex_lock(&trace_state_mutex);
9a8afc8d	243
4b706372 NH	244	if (state == trace_state) {
	245	rc = -EAGAIN;
	246	goto out_unlock;
	247	}
	248
9a8afc8d NH	249	switch (state) {
9a8afc8d NH	250	case TRACE_ON:
cad456d5 NH	251	if (!try_module_get(THIS_MODULE)) {
	252	rc = -ENODEV;
	253	break;
	254	}
	255
38516ab5 SR	256	rc \|= register_trace_kfree_skb(trace_kfree_skb_hit, NULL);
38516ab5 SR	257	rc \|= register_trace_napi_poll(trace_napi_poll_hit, NULL);
9a8afc8d	258	break;
cad456d5	259
9a8afc8d	260	case TRACE_OFF:
38516ab5 SR	261	rc \|= unregister_trace_kfree_skb(trace_kfree_skb_hit, NULL);
38516ab5 SR	262	rc \|= unregister_trace_napi_poll(trace_napi_poll_hit, NULL);
9a8afc8d NH	263
9a8afc8d NH	264	tracepoint_synchronize_unregister();
4ea7e386 NH	265
	266	/*
	267	* Clean the device list
	268	*/
	269	list_for_each_entry_safe(new_stat, temp, &hw_stats_list, list) {
	270	if (new_stat->dev == NULL) {
	271	list_del_rcu(&new_stat->list);
fa81c0e1	272	kfree_rcu(new_stat, rcu);
4ea7e386 NH	273	}
4ea7e386 NH	274	}
cad456d5 NH	275
	276	module_put(THIS_MODULE);
	277
9a8afc8d NH	278	break;
	279	default:
	280	rc = 1;
	281	break;
	282	}
	283
4ea7e386 NH	284	if (!rc)
4ea7e386 NH	285	trace_state = state;
4b706372 NH	286	else
4b706372 NH	287	rc = -EINPROGRESS;
4ea7e386	288
4b706372	289	out_unlock:
cde2e9a6	290	mutex_unlock(&trace_state_mutex);
4ea7e386	291
9a8afc8d NH	292	return rc;
	293	}
	294
	295
	296	static int net_dm_cmd_config(struct sk_buff *skb,
	297	struct genl_info *info)
	298	{
	299	return -ENOTSUPP;
	300	}
	301
	302	static int net_dm_cmd_trace(struct sk_buff *skb,
	303	struct genl_info *info)
	304	{
	305	switch (info->genlhdr->cmd) {
	306	case NET_DM_CMD_START:
	307	return set_all_monitor_traces(TRACE_ON);
	308	break;
	309	case NET_DM_CMD_STOP:
	310	return set_all_monitor_traces(TRACE_OFF);
	311	break;
	312	}
	313
	314	return -ENOTSUPP;
	315	}
	316
4ea7e386 NH	317	static int dropmon_net_event(struct notifier_block *ev_block,
	318	unsigned long event, void *ptr)
	319	{
	320	struct net_device *dev = ptr;
	321	struct dm_hw_stat_delta *new_stat = NULL;
	322	struct dm_hw_stat_delta *tmp;
	323
	324	switch (event) {
	325	case NETDEV_REGISTER:
	326	new_stat = kzalloc(sizeof(struct dm_hw_stat_delta), GFP_KERNEL);
	327
	328	if (!new_stat)
	329	goto out;
	330
	331	new_stat->dev = dev;
5848cc09	332	new_stat->last_rx = jiffies;
cde2e9a6	333	mutex_lock(&trace_state_mutex);
4ea7e386	334	list_add_rcu(&new_stat->list, &hw_stats_list);
cde2e9a6	335	mutex_unlock(&trace_state_mutex);
4ea7e386 NH	336	break;
4ea7e386 NH	337	case NETDEV_UNREGISTER:
cde2e9a6	338	mutex_lock(&trace_state_mutex);
4ea7e386 NH	339	list_for_each_entry_safe(new_stat, tmp, &hw_stats_list, list) {
	340	if (new_stat->dev == dev) {
	341	new_stat->dev = NULL;
	342	if (trace_state == TRACE_OFF) {
	343	list_del_rcu(&new_stat->list);
fa81c0e1	344	kfree_rcu(new_stat, rcu);
4ea7e386 NH	345	break;
	346	}
	347	}
	348	}
cde2e9a6	349	mutex_unlock(&trace_state_mutex);
4ea7e386 NH	350	break;
	351	}
	352	out:
	353	return NOTIFY_DONE;
	354	}
9a8afc8d NH	355
	356	static struct genl_ops dropmon_ops[] = {
	357	{
	358	.cmd = NET_DM_CMD_CONFIG,
	359	.doit = net_dm_cmd_config,
	360	},
	361	{
	362	.cmd = NET_DM_CMD_START,
	363	.doit = net_dm_cmd_trace,
	364	},
	365	{
	366	.cmd = NET_DM_CMD_STOP,
	367	.doit = net_dm_cmd_trace,
	368	},
	369	};
	370
4ea7e386 NH	371	static struct notifier_block dropmon_net_notifier = {
	372	.notifier_call = dropmon_net_event
	373	};
	374
9a8afc8d NH	375	static int __init init_net_drop_monitor(void)
9a8afc8d NH	376	{
9a8afc8d	377	struct per_cpu_dm_data *data;
a256be70 CG	378	int cpu, rc;
a256be70 CG	379
e005d193	380	pr_info("Initializing network drop monitor service\n");
9a8afc8d NH	381
9a8afc8d NH	382	if (sizeof(void *) > 8) {
e005d193	383	pr_err("Unable to store program counters on this arch, Drop monitor failed\n");
9a8afc8d NH	384	return -ENOSPC;
	385	}
	386
a256be70 CG	387	rc = genl_register_family_with_ops(&net_drop_monitor_family,
	388	dropmon_ops,
	389	ARRAY_SIZE(dropmon_ops));
	390	if (rc) {
e005d193	391	pr_err("Could not create drop monitor netlink family\n");
a256be70	392	return rc;
9a8afc8d NH	393	}
9a8afc8d NH	394
4ea7e386 NH	395	rc = register_netdevice_notifier(&dropmon_net_notifier);
4ea7e386 NH	396	if (rc < 0) {
e005d193	397	pr_crit("Failed to register netdevice notifier\n");
4ea7e386 NH	398	goto out_unreg;
	399	}
	400
9a8afc8d NH	401	rc = 0;
9a8afc8d NH	402
cad456d5	403	for_each_possible_cpu(cpu) {
9a8afc8d	404	data = &per_cpu(dm_cpu_data, cpu);
9a8afc8d NH	405	INIT_WORK(&data->dm_alert_work, send_dm_alert);
9a8afc8d NH	406	init_timer(&data->send_timer);
bec4596b	407	data->send_timer.data = (unsigned long)data;
9a8afc8d	408	data->send_timer.function = sched_send_work;
bec4596b	409	spin_lock_init(&data->lock);
4fdcfa12	410	reset_per_cpu_data(data);
9a8afc8d	411	}
4ea7e386	412
3885ca78	413
9a8afc8d NH	414	goto out;
	415
	416	out_unreg:
	417	genl_unregister_family(&net_drop_monitor_family);
	418	out:
	419	return rc;
	420	}
	421
cad456d5 NH	422	static void exit_net_drop_monitor(void)
	423	{
	424	struct per_cpu_dm_data *data;
	425	int cpu;
	426
	427	BUG_ON(unregister_netdevice_notifier(&dropmon_net_notifier));
	428
	429	/*
	430	* Because of the module_get/put we do in the trace state change path
	431	* we are guarnateed not to have any current users when we get here
	432	* all we need to do is make sure that we don't have any running timers
	433	* or pending schedule calls
	434	*/
	435
	436	for_each_possible_cpu(cpu) {
	437	data = &per_cpu(dm_cpu_data, cpu);
	438	del_timer_sync(&data->send_timer);
	439	cancel_work_sync(&data->dm_alert_work);
	440	/*
	441	* At this point, we should have exclusive access
	442	* to this struct and can free the skb inside it
	443	*/
	444	kfree_skb(data->skb);
	445	}
	446
	447	BUG_ON(genl_unregister_family(&net_drop_monitor_family));
	448	}
	449
	450	module_init(init_net_drop_monitor);
	451	module_exit(exit_net_drop_monitor);
	452
	453	MODULE_LICENSE("GPL v2");
	454	MODULE_AUTHOR("Neil Horman <nhorman@tuxdriver.com>");
3fdcbd45	455	MODULE_ALIAS_GENL_FAMILY("NET_DM");