[mirror_ubuntu-bionic-kernel.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;

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 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(struct timer_list *t)
{
	struct per_cpu_dm_data *data = from_timer(data, t, send_timer);

	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 genl_family net_drop_monitor_family __ro_after_init = {
	.hdrsize        = 0,
	.name           = "NET_DM",
	.version        = 2,
	.module		= THIS_MODULE,
	.ops		= dropmon_ops,
	.n_ops		= ARRAY_SIZE(dropmon_ops),
	.mcgrps		= dropmon_mcgrps,
	.n_mcgrps	= ARRAY_SIZE(dropmon_mcgrps),
};

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(&net_drop_monitor_family);
	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);
		timer_setup(&data->send_timer, sched_send_work, 0);
		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
489111e5	62	static struct genl_family net_drop_monitor_family;
9a8afc8d NH	63
	64	static DEFINE_PER_CPU(struct per_cpu_dm_data, dm_cpu_data);
	65
	66	static int dm_hit_limit = 64;
	67	static int dm_delay = 1;
4ea7e386 NH	68	static unsigned long dm_hw_check_delta = 2*HZ;
4ea7e386 NH	69	static LIST_HEAD(hw_stats_list);
9a8afc8d	70
bec4596b	71	static struct sk_buff reset_per_cpu_data(struct per_cpu_dm_data data)
9a8afc8d NH	72	{
	73	size_t al;
	74	struct net_dm_alert_msg *msg;
683703a2	75	struct nlattr *nla;
3885ca78	76	struct sk_buff *skb;
bec4596b	77	unsigned long flags;
4200462d	78	void *msg_header;
9a8afc8d NH	79
	80	al = sizeof(struct net_dm_alert_msg);
	81	al += dm_hit_limit * sizeof(struct net_dm_drop_point);
683703a2 NH	82	al += sizeof(struct nlattr);
683703a2 NH	83
3885ca78 NH	84	skb = genlmsg_new(al, GFP_KERNEL);
3885ca78 NH	85
4200462d RW	86	if (!skb)
4200462d RW	87	goto err;
3885ca78	88
4200462d RW	89	msg_header = genlmsg_put(skb, 0, 0, &net_drop_monitor_family,
	90	0, NET_DM_CMD_ALERT);
	91	if (!msg_header) {
	92	nlmsg_free(skb);
	93	skb = NULL;
	94	goto err;
	95	}
	96	nla = nla_reserve(skb, NLA_UNSPEC,
	97	sizeof(struct net_dm_alert_msg));
	98	if (!nla) {
	99	nlmsg_free(skb);
	100	skb = NULL;
	101	goto err;
	102	}
	103	msg = nla_data(nla);
	104	memset(msg, 0, al);
4200462d RW	105	goto out;
	106
	107	err:
	108	mod_timer(&data->send_timer, jiffies + HZ / 10);
	109	out:
bec4596b ED	110	spin_lock_irqsave(&data->lock, flags);
	111	swap(data->skb, skb);
	112	spin_unlock_irqrestore(&data->lock, flags);
	113
3b48ab22 RW	114	if (skb) {
	115	struct nlmsghdr nlh = (struct nlmsghdr )skb->data;
	116	struct genlmsghdr gnlh = (struct genlmsghdr )nlmsg_data(nlh);
	117
	118	genlmsg_end(skb, genlmsg_data(gnlh));
	119	}
	120
bec4596b	121	return skb;
9a8afc8d NH	122	}
9a8afc8d NH	123
85bae4bd	124	static const struct genl_multicast_group dropmon_mcgrps[] = {
2a94fe48	125	{ .name = "events", },
e5dcecba JB	126	};
e5dcecba JB	127
bec4596b	128	static void send_dm_alert(struct work_struct *work)
9a8afc8d NH	129	{
9a8afc8d NH	130	struct sk_buff *skb;
bec4596b	131	struct per_cpu_dm_data *data;
9a8afc8d	132
bec4596b	133	data = container_of(work, struct per_cpu_dm_data, dm_alert_work);
4fdcfa12	134
bec4596b	135	skb = reset_per_cpu_data(data);
9a8afc8d	136
3885ca78	137	if (skb)
68eb5503	138	genlmsg_multicast(&net_drop_monitor_family, skb, 0,
2a94fe48	139	0, 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	*/
e99e88a9	147	static void sched_send_work(struct timer_list *t)
9a8afc8d	148	{
e99e88a9	149	struct per_cpu_dm_data *data = from_timer(data, t, send_timer);
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	164	local_irq_save(flags);
903ceff7	165	data = this_cpu_ptr(&dm_cpu_data);
bec4596b ED	166	spin_lock(&data->lock);
bec4596b ED	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
1db19db7 JDB	206	static void trace_napi_poll_hit(void ignore, struct napi_struct napi,
1db19db7 JDB	207	int work, int budget)
4ea7e386 NH	208	{
	209	struct dm_hw_stat_delta *new_stat;
	210
	211	/*
5848cc09	212	* Don't check napi structures with no associated device
4ea7e386	213	*/
5848cc09	214	if (!napi->dev)
4ea7e386 NH	215	return;
	216
	217	rcu_read_lock();
	218	list_for_each_entry_rcu(new_stat, &hw_stats_list, list) {
5848cc09 NH	219	/*
	220	* only add a note to our monitor buffer if:
	221	* 1) this is the dev we received on
	222	* 2) its after the last_rx delta
	223	* 3) our rx_dropped count has gone up
	224	*/
4ea7e386	225	if ((new_stat->dev == napi->dev) &&
5848cc09	226	(time_after(jiffies, new_stat->last_rx + dm_hw_check_delta)) &&
4ea7e386 NH	227	(napi->dev->stats.rx_dropped != new_stat->last_drop_val)) {
	228	trace_drop_common(NULL, NULL);
	229	new_stat->last_drop_val = napi->dev->stats.rx_dropped;
5848cc09	230	new_stat->last_rx = jiffies;
4ea7e386 NH	231	break;
	232	}
	233	}
	234	rcu_read_unlock();
	235	}
	236
9a8afc8d NH	237	static int set_all_monitor_traces(int state)
	238	{
	239	int rc = 0;
4ea7e386 NH	240	struct dm_hw_stat_delta *new_stat = NULL;
	241	struct dm_hw_stat_delta *temp;
	242
cde2e9a6	243	mutex_lock(&trace_state_mutex);
9a8afc8d	244
4b706372 NH	245	if (state == trace_state) {
	246	rc = -EAGAIN;
	247	goto out_unlock;
	248	}
	249
9a8afc8d NH	250	switch (state) {
9a8afc8d NH	251	case TRACE_ON:
cad456d5 NH	252	if (!try_module_get(THIS_MODULE)) {
	253	rc = -ENODEV;
	254	break;
	255	}
	256
38516ab5 SR	257	rc \|= register_trace_kfree_skb(trace_kfree_skb_hit, NULL);
38516ab5 SR	258	rc \|= register_trace_napi_poll(trace_napi_poll_hit, NULL);
9a8afc8d	259	break;
cad456d5	260
9a8afc8d	261	case TRACE_OFF:
38516ab5 SR	262	rc \|= unregister_trace_kfree_skb(trace_kfree_skb_hit, NULL);
38516ab5 SR	263	rc \|= unregister_trace_napi_poll(trace_napi_poll_hit, NULL);
9a8afc8d NH	264
9a8afc8d NH	265	tracepoint_synchronize_unregister();
4ea7e386 NH	266
	267	/*
	268	* Clean the device list
	269	*/
	270	list_for_each_entry_safe(new_stat, temp, &hw_stats_list, list) {
	271	if (new_stat->dev == NULL) {
	272	list_del_rcu(&new_stat->list);
fa81c0e1	273	kfree_rcu(new_stat, rcu);
4ea7e386 NH	274	}
4ea7e386 NH	275	}
cad456d5 NH	276
	277	module_put(THIS_MODULE);
	278
9a8afc8d NH	279	break;
	280	default:
	281	rc = 1;
	282	break;
	283	}
	284
4ea7e386 NH	285	if (!rc)
4ea7e386 NH	286	trace_state = state;
4b706372 NH	287	else
4b706372 NH	288	rc = -EINPROGRESS;
4ea7e386	289
4b706372	290	out_unlock:
cde2e9a6	291	mutex_unlock(&trace_state_mutex);
4ea7e386	292
9a8afc8d NH	293	return rc;
	294	}
	295
	296
	297	static int net_dm_cmd_config(struct sk_buff *skb,
	298	struct genl_info *info)
	299	{
	300	return -ENOTSUPP;
	301	}
	302
	303	static int net_dm_cmd_trace(struct sk_buff *skb,
	304	struct genl_info *info)
	305	{
	306	switch (info->genlhdr->cmd) {
	307	case NET_DM_CMD_START:
	308	return set_all_monitor_traces(TRACE_ON);
9a8afc8d NH	309	case NET_DM_CMD_STOP:
9a8afc8d NH	310	return set_all_monitor_traces(TRACE_OFF);
9a8afc8d NH	311	}
	312
	313	return -ENOTSUPP;
	314	}
	315
4ea7e386	316	static int dropmon_net_event(struct notifier_block *ev_block,
351638e7	317	unsigned long event, void *ptr)
4ea7e386	318	{
351638e7	319	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4ea7e386 NH	320	struct dm_hw_stat_delta *new_stat = NULL;
	321	struct dm_hw_stat_delta *tmp;
	322
	323	switch (event) {
	324	case NETDEV_REGISTER:
	325	new_stat = kzalloc(sizeof(struct dm_hw_stat_delta), GFP_KERNEL);
	326
	327	if (!new_stat)
	328	goto out;
	329
	330	new_stat->dev = dev;
5848cc09	331	new_stat->last_rx = jiffies;
cde2e9a6	332	mutex_lock(&trace_state_mutex);
4ea7e386	333	list_add_rcu(&new_stat->list, &hw_stats_list);
cde2e9a6	334	mutex_unlock(&trace_state_mutex);
4ea7e386 NH	335	break;
4ea7e386 NH	336	case NETDEV_UNREGISTER:
cde2e9a6	337	mutex_lock(&trace_state_mutex);
4ea7e386 NH	338	list_for_each_entry_safe(new_stat, tmp, &hw_stats_list, list) {
	339	if (new_stat->dev == dev) {
	340	new_stat->dev = NULL;
	341	if (trace_state == TRACE_OFF) {
	342	list_del_rcu(&new_stat->list);
fa81c0e1	343	kfree_rcu(new_stat, rcu);
4ea7e386 NH	344	break;
	345	}
	346	}
	347	}
cde2e9a6	348	mutex_unlock(&trace_state_mutex);
4ea7e386 NH	349	break;
	350	}
	351	out:
	352	return NOTIFY_DONE;
	353	}
9a8afc8d	354
4534de83	355	static const struct genl_ops dropmon_ops[] = {
9a8afc8d NH	356	{
	357	.cmd = NET_DM_CMD_CONFIG,
	358	.doit = net_dm_cmd_config,
	359	},
	360	{
	361	.cmd = NET_DM_CMD_START,
	362	.doit = net_dm_cmd_trace,
	363	},
	364	{
	365	.cmd = NET_DM_CMD_STOP,
	366	.doit = net_dm_cmd_trace,
	367	},
	368	};
	369
56989f6d	370	static struct genl_family net_drop_monitor_family __ro_after_init = {
489111e5 JB	371	.hdrsize = 0,
	372	.name = "NET_DM",
	373	.version = 2,
	374	.module = THIS_MODULE,
	375	.ops = dropmon_ops,
	376	.n_ops = ARRAY_SIZE(dropmon_ops),
	377	.mcgrps = dropmon_mcgrps,
	378	.n_mcgrps = ARRAY_SIZE(dropmon_mcgrps),
	379	};
	380
4ea7e386 NH	381	static struct notifier_block dropmon_net_notifier = {
	382	.notifier_call = dropmon_net_event
	383	};
	384
9a8afc8d NH	385	static int __init init_net_drop_monitor(void)
9a8afc8d NH	386	{
9a8afc8d	387	struct per_cpu_dm_data *data;
a256be70 CG	388	int cpu, rc;
a256be70 CG	389
e005d193	390	pr_info("Initializing network drop monitor service\n");
9a8afc8d NH	391
9a8afc8d NH	392	if (sizeof(void *) > 8) {
e005d193	393	pr_err("Unable to store program counters on this arch, Drop monitor failed\n");
9a8afc8d NH	394	return -ENOSPC;
	395	}
	396
489111e5	397	rc = genl_register_family(&net_drop_monitor_family);
a256be70	398	if (rc) {
e005d193	399	pr_err("Could not create drop monitor netlink family\n");
a256be70	400	return rc;
9a8afc8d	401	}
2a94fe48	402	WARN_ON(net_drop_monitor_family.mcgrp_offset != NET_DM_GRP_ALERT);
e5dcecba	403
4ea7e386 NH	404	rc = register_netdevice_notifier(&dropmon_net_notifier);
4ea7e386 NH	405	if (rc < 0) {
e005d193	406	pr_crit("Failed to register netdevice notifier\n");
4ea7e386 NH	407	goto out_unreg;
	408	}
	409
9a8afc8d NH	410	rc = 0;
9a8afc8d NH	411
cad456d5	412	for_each_possible_cpu(cpu) {
9a8afc8d	413	data = &per_cpu(dm_cpu_data, cpu);
9a8afc8d	414	INIT_WORK(&data->dm_alert_work, send_dm_alert);
e99e88a9	415	timer_setup(&data->send_timer, sched_send_work, 0);
bec4596b	416	spin_lock_init(&data->lock);
4fdcfa12	417	reset_per_cpu_data(data);
9a8afc8d	418	}
4ea7e386	419
3885ca78	420
9a8afc8d NH	421	goto out;
	422
	423	out_unreg:
	424	genl_unregister_family(&net_drop_monitor_family);
	425	out:
	426	return rc;
	427	}
	428
cad456d5 NH	429	static void exit_net_drop_monitor(void)
	430	{
	431	struct per_cpu_dm_data *data;
	432	int cpu;
	433
	434	BUG_ON(unregister_netdevice_notifier(&dropmon_net_notifier));
	435
	436	/*
	437	* Because of the module_get/put we do in the trace state change path
	438	* we are guarnateed not to have any current users when we get here
	439	* all we need to do is make sure that we don't have any running timers
	440	* or pending schedule calls
	441	*/
	442
	443	for_each_possible_cpu(cpu) {
	444	data = &per_cpu(dm_cpu_data, cpu);
	445	del_timer_sync(&data->send_timer);
	446	cancel_work_sync(&data->dm_alert_work);
	447	/*
	448	* At this point, we should have exclusive access
	449	* to this struct and can free the skb inside it
	450	*/
	451	kfree_skb(data->skb);
	452	}
	453
	454	BUG_ON(genl_unregister_family(&net_drop_monitor_family));
	455	}
	456
	457	module_init(init_net_drop_monitor);
	458	module_exit(exit_net_drop_monitor);
	459
	460	MODULE_LICENSE("GPL v2");
	461	MODULE_AUTHOR("Neil Horman <nhorman@tuxdriver.com>");
3fdcbd45	462	MODULE_ALIAS_GENL_FAMILY("NET_DM");