[deliverable/linux.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;

	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) {
		genlmsg_put(skb, 0, 0, &net_drop_monitor_family,
				0, NET_DM_CMD_ALERT);
		nla = nla_reserve(skb, NLA_UNSPEC,
				  sizeof(struct net_dm_alert_msg));
		msg = nla_data(nla);
		memset(msg, 0, al);
	} else {
		mod_timer(&data->send_timer, jiffies + HZ / 10);
	}

	spin_lock_irqsave(&data->lock, flags);
	swap(data->skb, skb);
	spin_unlock_irqrestore(&data->lock, flags);

	return skb;
}

static const struct genl_multicast_group dropmon_mcgrps[] = {
	{ .name = "events", },
};

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(&net_drop_monitor_family, skb, 0,
				  0, 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 = this_cpu_ptr(&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,
				int work, int budget)
{
	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);
	case NET_DM_CMD_STOP:
		return set_all_monitor_traces(TRACE_OFF);
	}

	return -ENOTSUPP;
}

static int dropmon_net_event(struct notifier_block *ev_block,
			     unsigned long event, void *ptr)
{
	struct net_device *dev = netdev_notifier_info_to_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 const 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_groups(&net_drop_monitor_family,
						  dropmon_ops, dropmon_mcgrps);
	if (rc) {
		pr_err("Could not create drop monitor netlink family\n");
		return rc;
	}
	WARN_ON(net_drop_monitor_family.mcgrp_offset != NET_DM_GRP_ALERT);

	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;
9a8afc8d NH	83
	84	al = sizeof(struct net_dm_alert_msg);
	85	al += dm_hit_limit * sizeof(struct net_dm_drop_point);
683703a2 NH	86	al += sizeof(struct nlattr);
683703a2 NH	87
3885ca78 NH	88	skb = genlmsg_new(al, GFP_KERNEL);
	89
	90	if (skb) {
	91	genlmsg_put(skb, 0, 0, &net_drop_monitor_family,
	92	0, NET_DM_CMD_ALERT);
	93	nla = nla_reserve(skb, NLA_UNSPEC,
	94	sizeof(struct net_dm_alert_msg));
	95	msg = nla_data(nla);
	96	memset(msg, 0, al);
bec4596b ED	97	} else {
bec4596b ED	98	mod_timer(&data->send_timer, jiffies + HZ / 10);
3885ca78 NH	99	}
3885ca78 NH	100
bec4596b ED	101	spin_lock_irqsave(&data->lock, flags);
	102	swap(data->skb, skb);
	103	spin_unlock_irqrestore(&data->lock, flags);
	104
	105	return skb;
9a8afc8d NH	106	}
9a8afc8d NH	107
85bae4bd	108	static const struct genl_multicast_group dropmon_mcgrps[] = {
2a94fe48	109	{ .name = "events", },
e5dcecba JB	110	};
e5dcecba JB	111
bec4596b	112	static void send_dm_alert(struct work_struct *work)
9a8afc8d NH	113	{
9a8afc8d NH	114	struct sk_buff *skb;
bec4596b	115	struct per_cpu_dm_data *data;
9a8afc8d	116
bec4596b	117	data = container_of(work, struct per_cpu_dm_data, dm_alert_work);
4fdcfa12	118
bec4596b	119	skb = reset_per_cpu_data(data);
9a8afc8d	120
3885ca78	121	if (skb)
68eb5503	122	genlmsg_multicast(&net_drop_monitor_family, skb, 0,
2a94fe48	123	0, GFP_KERNEL);
9a8afc8d NH	124	}
	125
	126	/*
	127	* This is the timer function to delay the sending of an alert
	128	* in the event that more drops will arrive during the
bec4596b	129	* hysteresis period.
9a8afc8d	130	*/
bec4596b	131	static void sched_send_work(unsigned long _data)
9a8afc8d	132	{
bec4596b	133	struct per_cpu_dm_data data = (struct per_cpu_dm_data )_data;
3885ca78	134
bec4596b	135	schedule_work(&data->dm_alert_work);
9a8afc8d NH	136	}
9a8afc8d NH	137
4ea7e386	138	static void trace_drop_common(struct sk_buff skb, void location)
9a8afc8d NH	139	{
	140	struct net_dm_alert_msg *msg;
	141	struct nlmsghdr *nlh;
683703a2	142	struct nlattr *nla;
9a8afc8d	143	int i;
3885ca78	144	struct sk_buff *dskb;
bec4596b ED	145	struct per_cpu_dm_data *data;
bec4596b ED	146	unsigned long flags;
9a8afc8d	147
bec4596b	148	local_irq_save(flags);
903ceff7	149	data = this_cpu_ptr(&dm_cpu_data);
bec4596b ED	150	spin_lock(&data->lock);
bec4596b ED	151	dskb = data->skb;
3885ca78 NH	152
	153	if (!dskb)
	154	goto out;
	155
3885ca78	156	nlh = (struct nlmsghdr *)dskb->data;
683703a2 NH	157	nla = genlmsg_data(nlmsg_data(nlh));
683703a2 NH	158	msg = nla_data(nla);
9a8afc8d NH	159	for (i = 0; i < msg->entries; i++) {
	160	if (!memcmp(&location, msg->points[i].pc, sizeof(void *))) {
	161	msg->points[i].count++;
	162	goto out;
	163	}
	164	}
bec4596b ED	165	if (msg->entries == dm_hit_limit)
bec4596b ED	166	goto out;
9a8afc8d NH	167	/*
	168	* We need to create a new entry
	169	*/
3885ca78	170	__nla_reserve_nohdr(dskb, sizeof(struct net_dm_drop_point));
683703a2	171	nla->nla_len += NLA_ALIGN(sizeof(struct net_dm_drop_point));
9a8afc8d NH	172	memcpy(msg->points[msg->entries].pc, &location, sizeof(void *));
	173	msg->points[msg->entries].count = 1;
	174	msg->entries++;
	175
	176	if (!timer_pending(&data->send_timer)) {
	177	data->send_timer.expires = jiffies + dm_delay * HZ;
bec4596b	178	add_timer(&data->send_timer);
9a8afc8d NH	179	}
	180
	181	out:
bec4596b	182	spin_unlock_irqrestore(&data->lock, flags);
9a8afc8d NH	183	}
9a8afc8d NH	184
38516ab5	185	static void trace_kfree_skb_hit(void ignore, struct sk_buff skb, void *location)
4ea7e386 NH	186	{
	187	trace_drop_common(skb, location);
	188	}
	189
1db19db7 JDB	190	static void trace_napi_poll_hit(void ignore, struct napi_struct napi,
1db19db7 JDB	191	int work, int budget)
4ea7e386 NH	192	{
	193	struct dm_hw_stat_delta *new_stat;
	194
	195	/*
5848cc09	196	* Don't check napi structures with no associated device
4ea7e386	197	*/
5848cc09	198	if (!napi->dev)
4ea7e386 NH	199	return;
	200
	201	rcu_read_lock();
	202	list_for_each_entry_rcu(new_stat, &hw_stats_list, list) {
5848cc09 NH	203	/*
	204	* only add a note to our monitor buffer if:
	205	* 1) this is the dev we received on
	206	* 2) its after the last_rx delta
	207	* 3) our rx_dropped count has gone up
	208	*/
4ea7e386	209	if ((new_stat->dev == napi->dev) &&
5848cc09	210	(time_after(jiffies, new_stat->last_rx + dm_hw_check_delta)) &&
4ea7e386 NH	211	(napi->dev->stats.rx_dropped != new_stat->last_drop_val)) {
	212	trace_drop_common(NULL, NULL);
	213	new_stat->last_drop_val = napi->dev->stats.rx_dropped;
5848cc09	214	new_stat->last_rx = jiffies;
4ea7e386 NH	215	break;
	216	}
	217	}
	218	rcu_read_unlock();
	219	}
	220
9a8afc8d NH	221	static int set_all_monitor_traces(int state)
	222	{
	223	int rc = 0;
4ea7e386 NH	224	struct dm_hw_stat_delta *new_stat = NULL;
	225	struct dm_hw_stat_delta *temp;
	226
cde2e9a6	227	mutex_lock(&trace_state_mutex);
9a8afc8d	228
4b706372 NH	229	if (state == trace_state) {
	230	rc = -EAGAIN;
	231	goto out_unlock;
	232	}
	233
9a8afc8d NH	234	switch (state) {
9a8afc8d NH	235	case TRACE_ON:
cad456d5 NH	236	if (!try_module_get(THIS_MODULE)) {
	237	rc = -ENODEV;
	238	break;
	239	}
	240
38516ab5 SR	241	rc \|= register_trace_kfree_skb(trace_kfree_skb_hit, NULL);
38516ab5 SR	242	rc \|= register_trace_napi_poll(trace_napi_poll_hit, NULL);
9a8afc8d	243	break;
cad456d5	244
9a8afc8d	245	case TRACE_OFF:
38516ab5 SR	246	rc \|= unregister_trace_kfree_skb(trace_kfree_skb_hit, NULL);
38516ab5 SR	247	rc \|= unregister_trace_napi_poll(trace_napi_poll_hit, NULL);
9a8afc8d NH	248
9a8afc8d NH	249	tracepoint_synchronize_unregister();
4ea7e386 NH	250
	251	/*
	252	* Clean the device list
	253	*/
	254	list_for_each_entry_safe(new_stat, temp, &hw_stats_list, list) {
	255	if (new_stat->dev == NULL) {
	256	list_del_rcu(&new_stat->list);
fa81c0e1	257	kfree_rcu(new_stat, rcu);
4ea7e386 NH	258	}
4ea7e386 NH	259	}
cad456d5 NH	260
	261	module_put(THIS_MODULE);
	262
9a8afc8d NH	263	break;
	264	default:
	265	rc = 1;
	266	break;
	267	}
	268
4ea7e386 NH	269	if (!rc)
4ea7e386 NH	270	trace_state = state;
4b706372 NH	271	else
4b706372 NH	272	rc = -EINPROGRESS;
4ea7e386	273
4b706372	274	out_unlock:
cde2e9a6	275	mutex_unlock(&trace_state_mutex);
4ea7e386	276
9a8afc8d NH	277	return rc;
	278	}
	279
	280
	281	static int net_dm_cmd_config(struct sk_buff *skb,
	282	struct genl_info *info)
	283	{
	284	return -ENOTSUPP;
	285	}
	286
	287	static int net_dm_cmd_trace(struct sk_buff *skb,
	288	struct genl_info *info)
	289	{
	290	switch (info->genlhdr->cmd) {
	291	case NET_DM_CMD_START:
	292	return set_all_monitor_traces(TRACE_ON);
9a8afc8d NH	293	case NET_DM_CMD_STOP:
9a8afc8d NH	294	return set_all_monitor_traces(TRACE_OFF);
9a8afc8d NH	295	}
	296
	297	return -ENOTSUPP;
	298	}
	299
4ea7e386	300	static int dropmon_net_event(struct notifier_block *ev_block,
351638e7	301	unsigned long event, void *ptr)
4ea7e386	302	{
351638e7	303	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4ea7e386 NH	304	struct dm_hw_stat_delta *new_stat = NULL;
	305	struct dm_hw_stat_delta *tmp;
	306
	307	switch (event) {
	308	case NETDEV_REGISTER:
	309	new_stat = kzalloc(sizeof(struct dm_hw_stat_delta), GFP_KERNEL);
	310
	311	if (!new_stat)
	312	goto out;
	313
	314	new_stat->dev = dev;
5848cc09	315	new_stat->last_rx = jiffies;
cde2e9a6	316	mutex_lock(&trace_state_mutex);
4ea7e386	317	list_add_rcu(&new_stat->list, &hw_stats_list);
cde2e9a6	318	mutex_unlock(&trace_state_mutex);
4ea7e386 NH	319	break;
4ea7e386 NH	320	case NETDEV_UNREGISTER:
cde2e9a6	321	mutex_lock(&trace_state_mutex);
4ea7e386 NH	322	list_for_each_entry_safe(new_stat, tmp, &hw_stats_list, list) {
	323	if (new_stat->dev == dev) {
	324	new_stat->dev = NULL;
	325	if (trace_state == TRACE_OFF) {
	326	list_del_rcu(&new_stat->list);
fa81c0e1	327	kfree_rcu(new_stat, rcu);
4ea7e386 NH	328	break;
	329	}
	330	}
	331	}
cde2e9a6	332	mutex_unlock(&trace_state_mutex);
4ea7e386 NH	333	break;
	334	}
	335	out:
	336	return NOTIFY_DONE;
	337	}
9a8afc8d	338
4534de83	339	static const struct genl_ops dropmon_ops[] = {
9a8afc8d NH	340	{
	341	.cmd = NET_DM_CMD_CONFIG,
	342	.doit = net_dm_cmd_config,
	343	},
	344	{
	345	.cmd = NET_DM_CMD_START,
	346	.doit = net_dm_cmd_trace,
	347	},
	348	{
	349	.cmd = NET_DM_CMD_STOP,
	350	.doit = net_dm_cmd_trace,
	351	},
	352	};
	353
4ea7e386 NH	354	static struct notifier_block dropmon_net_notifier = {
	355	.notifier_call = dropmon_net_event
	356	};
	357
9a8afc8d NH	358	static int __init init_net_drop_monitor(void)
9a8afc8d NH	359	{
9a8afc8d	360	struct per_cpu_dm_data *data;
a256be70 CG	361	int cpu, rc;
a256be70 CG	362
e005d193	363	pr_info("Initializing network drop monitor service\n");
9a8afc8d NH	364
9a8afc8d NH	365	if (sizeof(void *) > 8) {
e005d193	366	pr_err("Unable to store program counters on this arch, Drop monitor failed\n");
9a8afc8d NH	367	return -ENOSPC;
	368	}
	369
2a94fe48 JB	370	rc = genl_register_family_with_ops_groups(&net_drop_monitor_family,
2a94fe48 JB	371	dropmon_ops, dropmon_mcgrps);
a256be70	372	if (rc) {
e005d193	373	pr_err("Could not create drop monitor netlink family\n");
a256be70	374	return rc;
9a8afc8d	375	}
2a94fe48	376	WARN_ON(net_drop_monitor_family.mcgrp_offset != NET_DM_GRP_ALERT);
e5dcecba	377
4ea7e386 NH	378	rc = register_netdevice_notifier(&dropmon_net_notifier);
4ea7e386 NH	379	if (rc < 0) {
e005d193	380	pr_crit("Failed to register netdevice notifier\n");
4ea7e386 NH	381	goto out_unreg;
	382	}
	383
9a8afc8d NH	384	rc = 0;
9a8afc8d NH	385
cad456d5	386	for_each_possible_cpu(cpu) {
9a8afc8d	387	data = &per_cpu(dm_cpu_data, cpu);
9a8afc8d NH	388	INIT_WORK(&data->dm_alert_work, send_dm_alert);
9a8afc8d NH	389	init_timer(&data->send_timer);
bec4596b	390	data->send_timer.data = (unsigned long)data;
9a8afc8d	391	data->send_timer.function = sched_send_work;
bec4596b	392	spin_lock_init(&data->lock);
4fdcfa12	393	reset_per_cpu_data(data);
9a8afc8d	394	}
4ea7e386	395
3885ca78	396
9a8afc8d NH	397	goto out;
	398
	399	out_unreg:
	400	genl_unregister_family(&net_drop_monitor_family);
	401	out:
	402	return rc;
	403	}
	404
cad456d5 NH	405	static void exit_net_drop_monitor(void)
	406	{
	407	struct per_cpu_dm_data *data;
	408	int cpu;
	409
	410	BUG_ON(unregister_netdevice_notifier(&dropmon_net_notifier));
	411
	412	/*
	413	* Because of the module_get/put we do in the trace state change path
	414	* we are guarnateed not to have any current users when we get here
	415	* all we need to do is make sure that we don't have any running timers
	416	* or pending schedule calls
	417	*/
	418
	419	for_each_possible_cpu(cpu) {
	420	data = &per_cpu(dm_cpu_data, cpu);
	421	del_timer_sync(&data->send_timer);
	422	cancel_work_sync(&data->dm_alert_work);
	423	/*
	424	* At this point, we should have exclusive access
	425	* to this struct and can free the skb inside it
	426	*/
	427	kfree_skb(data->skb);
	428	}
	429
	430	BUG_ON(genl_unregister_family(&net_drop_monitor_family));
	431	}
	432
	433	module_init(init_net_drop_monitor);
	434	module_exit(exit_net_drop_monitor);
	435
	436	MODULE_LICENSE("GPL v2");
	437	MODULE_AUTHOR("Neil Horman <nhorman@tuxdriver.com>");
3fdcbd45	438	MODULE_ALIAS_GENL_FAMILY("NET_DM");