[linux-2.6-block.git] / net / core / drop_monitor.c

// SPDX-License-Identifier: GPL-2.0-only
/*
 * 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,
		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
		.doit = net_dm_cmd_config,
	},
	{
		.cmd = NET_DM_CMD_START,
		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
		.doit = net_dm_cmd_trace,
	},
	{
		.cmd = NET_DM_CMD_STOP,
		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
		.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
09c434b8	1	// SPDX-License-Identifier: GPL-2.0-only
9a8afc8d NH	2	/*
	3	* Monitoring code for network dropped packet alerts
	4	*
	5	* Copyright (C) 2009 Neil Horman <nhorman@tuxdriver.com>
	6	*/
	7
e005d193 JP	8	#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
e005d193 JP	9
9a8afc8d NH	10	#include <linux/netdevice.h>
	11	#include <linux/etherdevice.h>
	12	#include <linux/string.h>
	13	#include <linux/if_arp.h>
	14	#include <linux/inetdevice.h>
	15	#include <linux/inet.h>
	16	#include <linux/interrupt.h>
	17	#include <linux/netpoll.h>
	18	#include <linux/sched.h>
	19	#include <linux/delay.h>
	20	#include <linux/types.h>
	21	#include <linux/workqueue.h>
	22	#include <linux/netlink.h>
	23	#include <linux/net_dropmon.h>
	24	#include <linux/percpu.h>
	25	#include <linux/timer.h>
	26	#include <linux/bitops.h>
5a0e3ad6	27	#include <linux/slab.h>
cad456d5	28	#include <linux/module.h>
9a8afc8d	29	#include <net/genetlink.h>
4ea7e386	30	#include <net/netevent.h>
9a8afc8d	31
ad8d75ff	32	#include <trace/events/skb.h>
9cbc1cb8	33	#include <trace/events/napi.h>
9a8afc8d NH	34
	35	#include <asm/unaligned.h>
	36
	37	#define TRACE_ON 1
	38	#define TRACE_OFF 0
	39
9a8afc8d NH	40	/*
	41	* Globals, our netlink socket pointer
	42	* and the work handle that will send up
	43	* netlink alerts
	44	*/
4ea7e386	45	static int trace_state = TRACE_OFF;
cde2e9a6	46	static DEFINE_MUTEX(trace_state_mutex);
9a8afc8d NH	47
9a8afc8d NH	48	struct per_cpu_dm_data {
bec4596b ED	49	spinlock_t lock;
	50	struct sk_buff *skb;
	51	struct work_struct dm_alert_work;
	52	struct timer_list send_timer;
9a8afc8d NH	53	};
9a8afc8d NH	54
4ea7e386 NH	55	struct dm_hw_stat_delta {
4ea7e386 NH	56	struct net_device *dev;
5848cc09	57	unsigned long last_rx;
4ea7e386 NH	58	struct list_head list;
	59	struct rcu_head rcu;
	60	unsigned long last_drop_val;
	61	};
	62
489111e5	63	static struct genl_family net_drop_monitor_family;
9a8afc8d NH	64
	65	static DEFINE_PER_CPU(struct per_cpu_dm_data, dm_cpu_data);
	66
	67	static int dm_hit_limit = 64;
	68	static int dm_delay = 1;
4ea7e386 NH	69	static unsigned long dm_hw_check_delta = 2*HZ;
4ea7e386 NH	70	static LIST_HEAD(hw_stats_list);
9a8afc8d	71
bec4596b	72	static struct sk_buff reset_per_cpu_data(struct per_cpu_dm_data data)
9a8afc8d NH	73	{
	74	size_t al;
	75	struct net_dm_alert_msg *msg;
683703a2	76	struct nlattr *nla;
3885ca78	77	struct sk_buff *skb;
bec4596b	78	unsigned long flags;
4200462d	79	void *msg_header;
9a8afc8d NH	80
	81	al = sizeof(struct net_dm_alert_msg);
	82	al += dm_hit_limit * sizeof(struct net_dm_drop_point);
683703a2 NH	83	al += sizeof(struct nlattr);
683703a2 NH	84
3885ca78 NH	85	skb = genlmsg_new(al, GFP_KERNEL);
3885ca78 NH	86
4200462d RW	87	if (!skb)
4200462d RW	88	goto err;
3885ca78	89
4200462d RW	90	msg_header = genlmsg_put(skb, 0, 0, &net_drop_monitor_family,
	91	0, NET_DM_CMD_ALERT);
	92	if (!msg_header) {
	93	nlmsg_free(skb);
	94	skb = NULL;
	95	goto err;
	96	}
	97	nla = nla_reserve(skb, NLA_UNSPEC,
	98	sizeof(struct net_dm_alert_msg));
	99	if (!nla) {
	100	nlmsg_free(skb);
	101	skb = NULL;
	102	goto err;
	103	}
	104	msg = nla_data(nla);
	105	memset(msg, 0, al);
4200462d RW	106	goto out;
	107
	108	err:
	109	mod_timer(&data->send_timer, jiffies + HZ / 10);
	110	out:
bec4596b ED	111	spin_lock_irqsave(&data->lock, flags);
	112	swap(data->skb, skb);
	113	spin_unlock_irqrestore(&data->lock, flags);
	114
3b48ab22 RW	115	if (skb) {
	116	struct nlmsghdr nlh = (struct nlmsghdr )skb->data;
	117	struct genlmsghdr gnlh = (struct genlmsghdr )nlmsg_data(nlh);
	118
	119	genlmsg_end(skb, genlmsg_data(gnlh));
	120	}
	121
bec4596b	122	return skb;
9a8afc8d NH	123	}
9a8afc8d NH	124
85bae4bd	125	static const struct genl_multicast_group dropmon_mcgrps[] = {
2a94fe48	126	{ .name = "events", },
e5dcecba JB	127	};
e5dcecba JB	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	138	if (skb)
68eb5503	139	genlmsg_multicast(&net_drop_monitor_family, skb, 0,
2a94fe48	140	0, GFP_KERNEL);
9a8afc8d NH	141	}
	142
	143	/*
	144	* This is the timer function to delay the sending of an alert
	145	* in the event that more drops will arrive during the
bec4596b	146	* hysteresis period.
9a8afc8d	147	*/
e99e88a9	148	static void sched_send_work(struct timer_list *t)
9a8afc8d	149	{
e99e88a9	150	struct per_cpu_dm_data *data = from_timer(data, t, send_timer);
3885ca78	151
bec4596b	152	schedule_work(&data->dm_alert_work);
9a8afc8d NH	153	}
9a8afc8d NH	154
4ea7e386	155	static void trace_drop_common(struct sk_buff skb, void location)
9a8afc8d NH	156	{
	157	struct net_dm_alert_msg *msg;
	158	struct nlmsghdr *nlh;
683703a2	159	struct nlattr *nla;
9a8afc8d	160	int i;
3885ca78	161	struct sk_buff *dskb;
bec4596b ED	162	struct per_cpu_dm_data *data;
bec4596b ED	163	unsigned long flags;
9a8afc8d	164
bec4596b	165	local_irq_save(flags);
903ceff7	166	data = this_cpu_ptr(&dm_cpu_data);
bec4596b ED	167	spin_lock(&data->lock);
bec4596b ED	168	dskb = data->skb;
3885ca78 NH	169
	170	if (!dskb)
	171	goto out;
	172
3885ca78	173	nlh = (struct nlmsghdr *)dskb->data;
683703a2 NH	174	nla = genlmsg_data(nlmsg_data(nlh));
683703a2 NH	175	msg = nla_data(nla);
9a8afc8d NH	176	for (i = 0; i < msg->entries; i++) {
	177	if (!memcmp(&location, msg->points[i].pc, sizeof(void *))) {
	178	msg->points[i].count++;
	179	goto out;
	180	}
	181	}
bec4596b ED	182	if (msg->entries == dm_hit_limit)
bec4596b ED	183	goto out;
9a8afc8d NH	184	/*
	185	* We need to create a new entry
	186	*/
3885ca78	187	__nla_reserve_nohdr(dskb, sizeof(struct net_dm_drop_point));
683703a2	188	nla->nla_len += NLA_ALIGN(sizeof(struct net_dm_drop_point));
9a8afc8d NH	189	memcpy(msg->points[msg->entries].pc, &location, sizeof(void *));
	190	msg->points[msg->entries].count = 1;
	191	msg->entries++;
	192
	193	if (!timer_pending(&data->send_timer)) {
	194	data->send_timer.expires = jiffies + dm_delay * HZ;
bec4596b	195	add_timer(&data->send_timer);
9a8afc8d NH	196	}
	197
	198	out:
bec4596b	199	spin_unlock_irqrestore(&data->lock, flags);
9a8afc8d NH	200	}
9a8afc8d NH	201
38516ab5	202	static void trace_kfree_skb_hit(void ignore, struct sk_buff skb, void *location)
4ea7e386 NH	203	{
	204	trace_drop_common(skb, location);
	205	}
	206
1db19db7 JDB	207	static void trace_napi_poll_hit(void ignore, struct napi_struct napi,
1db19db7 JDB	208	int work, int budget)
4ea7e386 NH	209	{
	210	struct dm_hw_stat_delta *new_stat;
	211
	212	/*
5848cc09	213	* Don't check napi structures with no associated device
4ea7e386	214	*/
5848cc09	215	if (!napi->dev)
4ea7e386 NH	216	return;
	217
	218	rcu_read_lock();
	219	list_for_each_entry_rcu(new_stat, &hw_stats_list, list) {
5848cc09 NH	220	/*
	221	* only add a note to our monitor buffer if:
	222	* 1) this is the dev we received on
	223	* 2) its after the last_rx delta
	224	* 3) our rx_dropped count has gone up
	225	*/
4ea7e386	226	if ((new_stat->dev == napi->dev) &&
5848cc09	227	(time_after(jiffies, new_stat->last_rx + dm_hw_check_delta)) &&
4ea7e386 NH	228	(napi->dev->stats.rx_dropped != new_stat->last_drop_val)) {
	229	trace_drop_common(NULL, NULL);
	230	new_stat->last_drop_val = napi->dev->stats.rx_dropped;
5848cc09	231	new_stat->last_rx = jiffies;
4ea7e386 NH	232	break;
	233	}
	234	}
	235	rcu_read_unlock();
	236	}
	237
9a8afc8d NH	238	static int set_all_monitor_traces(int state)
	239	{
	240	int rc = 0;
4ea7e386 NH	241	struct dm_hw_stat_delta *new_stat = NULL;
	242	struct dm_hw_stat_delta *temp;
	243
cde2e9a6	244	mutex_lock(&trace_state_mutex);
9a8afc8d	245
4b706372 NH	246	if (state == trace_state) {
	247	rc = -EAGAIN;
	248	goto out_unlock;
	249	}
	250
9a8afc8d NH	251	switch (state) {
9a8afc8d NH	252	case TRACE_ON:
cad456d5 NH	253	if (!try_module_get(THIS_MODULE)) {
	254	rc = -ENODEV;
	255	break;
	256	}
	257
38516ab5 SR	258	rc \|= register_trace_kfree_skb(trace_kfree_skb_hit, NULL);
38516ab5 SR	259	rc \|= register_trace_napi_poll(trace_napi_poll_hit, NULL);
9a8afc8d	260	break;
cad456d5	261
9a8afc8d	262	case TRACE_OFF:
38516ab5 SR	263	rc \|= unregister_trace_kfree_skb(trace_kfree_skb_hit, NULL);
38516ab5 SR	264	rc \|= unregister_trace_napi_poll(trace_napi_poll_hit, NULL);
9a8afc8d NH	265
9a8afc8d NH	266	tracepoint_synchronize_unregister();
4ea7e386 NH	267
	268	/*
	269	* Clean the device list
	270	*/
	271	list_for_each_entry_safe(new_stat, temp, &hw_stats_list, list) {
	272	if (new_stat->dev == NULL) {
	273	list_del_rcu(&new_stat->list);
fa81c0e1	274	kfree_rcu(new_stat, rcu);
4ea7e386 NH	275	}
4ea7e386 NH	276	}
cad456d5 NH	277
	278	module_put(THIS_MODULE);
	279
9a8afc8d NH	280	break;
	281	default:
	282	rc = 1;
	283	break;
	284	}
	285
4ea7e386 NH	286	if (!rc)
4ea7e386 NH	287	trace_state = state;
4b706372 NH	288	else
4b706372 NH	289	rc = -EINPROGRESS;
4ea7e386	290
4b706372	291	out_unlock:
cde2e9a6	292	mutex_unlock(&trace_state_mutex);
4ea7e386	293
9a8afc8d NH	294	return rc;
	295	}
	296
	297
	298	static int net_dm_cmd_config(struct sk_buff *skb,
	299	struct genl_info *info)
	300	{
	301	return -ENOTSUPP;
	302	}
	303
	304	static int net_dm_cmd_trace(struct sk_buff *skb,
	305	struct genl_info *info)
	306	{
	307	switch (info->genlhdr->cmd) {
	308	case NET_DM_CMD_START:
	309	return set_all_monitor_traces(TRACE_ON);
9a8afc8d NH	310	case NET_DM_CMD_STOP:
9a8afc8d NH	311	return set_all_monitor_traces(TRACE_OFF);
9a8afc8d NH	312	}
	313
	314	return -ENOTSUPP;
	315	}
	316
4ea7e386	317	static int dropmon_net_event(struct notifier_block *ev_block,
351638e7	318	unsigned long event, void *ptr)
4ea7e386	319	{
351638e7	320	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4ea7e386 NH	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	355
4534de83	356	static const struct genl_ops dropmon_ops[] = {
9a8afc8d NH	357	{
9a8afc8d NH	358	.cmd = NET_DM_CMD_CONFIG,
ef6243ac	359	.validate = GENL_DONT_VALIDATE_STRICT \| GENL_DONT_VALIDATE_DUMP,
9a8afc8d NH	360	.doit = net_dm_cmd_config,
	361	},
	362	{
	363	.cmd = NET_DM_CMD_START,
ef6243ac	364	.validate = GENL_DONT_VALIDATE_STRICT \| GENL_DONT_VALIDATE_DUMP,
9a8afc8d NH	365	.doit = net_dm_cmd_trace,
	366	},
	367	{
	368	.cmd = NET_DM_CMD_STOP,
ef6243ac	369	.validate = GENL_DONT_VALIDATE_STRICT \| GENL_DONT_VALIDATE_DUMP,
9a8afc8d NH	370	.doit = net_dm_cmd_trace,
	371	},
	372	};
	373
56989f6d	374	static struct genl_family net_drop_monitor_family __ro_after_init = {
489111e5 JB	375	.hdrsize = 0,
	376	.name = "NET_DM",
	377	.version = 2,
	378	.module = THIS_MODULE,
	379	.ops = dropmon_ops,
	380	.n_ops = ARRAY_SIZE(dropmon_ops),
	381	.mcgrps = dropmon_mcgrps,
	382	.n_mcgrps = ARRAY_SIZE(dropmon_mcgrps),
	383	};
	384
4ea7e386 NH	385	static struct notifier_block dropmon_net_notifier = {
	386	.notifier_call = dropmon_net_event
	387	};
	388
9a8afc8d NH	389	static int __init init_net_drop_monitor(void)
9a8afc8d NH	390	{
9a8afc8d	391	struct per_cpu_dm_data *data;
a256be70 CG	392	int cpu, rc;
a256be70 CG	393
e005d193	394	pr_info("Initializing network drop monitor service\n");
9a8afc8d NH	395
9a8afc8d NH	396	if (sizeof(void *) > 8) {
e005d193	397	pr_err("Unable to store program counters on this arch, Drop monitor failed\n");
9a8afc8d NH	398	return -ENOSPC;
	399	}
	400
489111e5	401	rc = genl_register_family(&net_drop_monitor_family);
a256be70	402	if (rc) {
e005d193	403	pr_err("Could not create drop monitor netlink family\n");
a256be70	404	return rc;
9a8afc8d	405	}
2a94fe48	406	WARN_ON(net_drop_monitor_family.mcgrp_offset != NET_DM_GRP_ALERT);
e5dcecba	407
4ea7e386 NH	408	rc = register_netdevice_notifier(&dropmon_net_notifier);
4ea7e386 NH	409	if (rc < 0) {
e005d193	410	pr_crit("Failed to register netdevice notifier\n");
4ea7e386 NH	411	goto out_unreg;
	412	}
	413
9a8afc8d NH	414	rc = 0;
9a8afc8d NH	415
cad456d5	416	for_each_possible_cpu(cpu) {
9a8afc8d	417	data = &per_cpu(dm_cpu_data, cpu);
9a8afc8d	418	INIT_WORK(&data->dm_alert_work, send_dm_alert);
e99e88a9	419	timer_setup(&data->send_timer, sched_send_work, 0);
bec4596b	420	spin_lock_init(&data->lock);
4fdcfa12	421	reset_per_cpu_data(data);
9a8afc8d	422	}
4ea7e386	423
3885ca78	424
9a8afc8d NH	425	goto out;
	426
	427	out_unreg:
	428	genl_unregister_family(&net_drop_monitor_family);
	429	out:
	430	return rc;
	431	}
	432
cad456d5 NH	433	static void exit_net_drop_monitor(void)
	434	{
	435	struct per_cpu_dm_data *data;
	436	int cpu;
	437
	438	BUG_ON(unregister_netdevice_notifier(&dropmon_net_notifier));
	439
	440	/*
	441	* Because of the module_get/put we do in the trace state change path
	442	* we are guarnateed not to have any current users when we get here
	443	* all we need to do is make sure that we don't have any running timers
	444	* or pending schedule calls
	445	*/
	446
	447	for_each_possible_cpu(cpu) {
	448	data = &per_cpu(dm_cpu_data, cpu);
	449	del_timer_sync(&data->send_timer);
	450	cancel_work_sync(&data->dm_alert_work);
	451	/*
	452	* At this point, we should have exclusive access
	453	* to this struct and can free the skb inside it
	454	*/
	455	kfree_skb(data->skb);
	456	}
	457
	458	BUG_ON(genl_unregister_family(&net_drop_monitor_family));
	459	}
	460
	461	module_init(init_net_drop_monitor);
	462	module_exit(exit_net_drop_monitor);
	463
	464	MODULE_LICENSE("GPL v2");
	465	MODULE_AUTHOR("Neil Horman <nhorman@tuxdriver.com>");
3fdcbd45	466	MODULE_ALIAS_GENL_FAMILY("NET_DM");