netprio_cgroup.c 6.4 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11
/*
 * net/core/netprio_cgroup.c	Priority Control Group
 *
 *		This program is free software; you can redistribute it and/or
 *		modify it under the terms of the GNU General Public License
 *		as published by the Free Software Foundation; either version
 *		2 of the License, or (at your option) any later version.
 *
 * Authors:	Neil Horman <nhorman@tuxdriver.com>
 */

J
Joe Perches 已提交
12 13
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

14 15 16 17 18 19 20 21 22 23 24 25 26 27
#include <linux/module.h>
#include <linux/slab.h>
#include <linux/types.h>
#include <linux/string.h>
#include <linux/errno.h>
#include <linux/skbuff.h>
#include <linux/cgroup.h>
#include <linux/rcupdate.h>
#include <linux/atomic.h>
#include <net/rtnetlink.h>
#include <net/pkt_cls.h>
#include <net/sock.h>
#include <net/netprio_cgroup.h>

28 29
#include <linux/fdtable.h>

30
#define PRIOMAP_MIN_SZ		128
31

32
/*
S
stephen hemminger 已提交
33
 * Extend @dev->priomap so that it's large enough to accommodate
34 35 36 37
 * @target_idx.  @dev->priomap.priomap_len > @target_idx after successful
 * return.  Must be called under rtnl lock.
 */
static int extend_netdev_table(struct net_device *dev, u32 target_idx)
38
{
39 40
	struct netprio_map *old, *new;
	size_t new_sz, new_len;
41

42
	/* is the existing priomap large enough? */
43
	old = rtnl_dereference(dev->priomap);
44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62
	if (old && old->priomap_len > target_idx)
		return 0;

	/*
	 * Determine the new size.  Let's keep it power-of-two.  We start
	 * from PRIOMAP_MIN_SZ and double it until it's large enough to
	 * accommodate @target_idx.
	 */
	new_sz = PRIOMAP_MIN_SZ;
	while (true) {
		new_len = (new_sz - offsetof(struct netprio_map, priomap)) /
			sizeof(new->priomap[0]);
		if (new_len > target_idx)
			break;
		new_sz *= 2;
		/* overflowed? */
		if (WARN_ON(new_sz < PRIOMAP_MIN_SZ))
			return -ENOSPC;
	}
63

64 65
	/* allocate & copy */
	new = kzalloc(new_sz, GFP_KERNEL);
66
	if (!new)
67
		return -ENOMEM;
68

69 70 71
	if (old)
		memcpy(new->priomap, old->priomap,
		       old->priomap_len * sizeof(old->priomap[0]));
72

73
	new->priomap_len = new_len;
74

75
	/* install the new priomap */
76 77 78
	rcu_assign_pointer(dev->priomap, new);
	if (old)
		kfree_rcu(old, rcu);
79 80 81
	return 0;
}

82 83
/**
 * netprio_prio - return the effective netprio of a cgroup-net_device pair
84
 * @css: css part of the target pair
85 86 87 88
 * @dev: net_device part of the target pair
 *
 * Should be called under RCU read or rtnl lock.
 */
89
static u32 netprio_prio(struct cgroup_subsys_state *css, struct net_device *dev)
90 91
{
	struct netprio_map *map = rcu_dereference_rtnl(dev->priomap);
92
	int id = css->cgroup->id;
93

94 95
	if (map && id < map->priomap_len)
		return map->priomap[id];
96 97 98 99 100
	return 0;
}

/**
 * netprio_set_prio - set netprio on a cgroup-net_device pair
101
 * @css: css part of the target pair
102 103 104
 * @dev: net_device part of the target pair
 * @prio: prio to set
 *
105
 * Set netprio to @prio on @css-@dev pair.  Should be called under rtnl
106 107
 * lock and may fail under memory pressure for non-zero @prio.
 */
108 109
static int netprio_set_prio(struct cgroup_subsys_state *css,
			    struct net_device *dev, u32 prio)
110 111
{
	struct netprio_map *map;
112
	int id = css->cgroup->id;
113 114 115 116
	int ret;

	/* avoid extending priomap for zero writes */
	map = rtnl_dereference(dev->priomap);
117
	if (!prio && (!map || map->priomap_len <= id))
118 119
		return 0;

120
	ret = extend_netdev_table(dev, id);
121 122 123 124
	if (ret)
		return ret;

	map = rtnl_dereference(dev->priomap);
125
	map->priomap[id] = prio;
126 127 128
	return 0;
}

129 130
static struct cgroup_subsys_state *
cgrp_css_alloc(struct cgroup_subsys_state *parent_css)
131
{
132
	struct cgroup_subsys_state *css;
133

134 135
	css = kzalloc(sizeof(*css), GFP_KERNEL);
	if (!css)
136 137
		return ERR_PTR(-ENOMEM);

138
	return css;
139 140
}

141
static int cgrp_css_online(struct cgroup_subsys_state *css)
142
{
143
	struct cgroup_subsys_state *parent_css = css_parent(css);
144
	struct net_device *dev;
145 146
	int ret = 0;

147
	if (!parent_css)
148
		return 0;
149 150

	rtnl_lock();
151 152 153 154 155
	/*
	 * Inherit prios from the parent.  As all prios are set during
	 * onlining, there is no need to clear them on offline.
	 */
	for_each_netdev(&init_net, dev) {
156
		u32 prio = netprio_prio(parent_css, dev);
157

158
		ret = netprio_set_prio(css, dev, prio);
159 160 161
		if (ret)
			break;
	}
162
	rtnl_unlock();
163 164 165
	return ret;
}

166
static void cgrp_css_free(struct cgroup_subsys_state *css)
167
{
168
	kfree(css);
169 170
}

171
static u64 read_prioidx(struct cgroup_subsys_state *css, struct cftype *cft)
172
{
173
	return css->cgroup->id;
174 175
}

176
static int read_priomap(struct seq_file *sf, void *v)
177 178 179 180
{
	struct net_device *dev;

	rcu_read_lock();
181
	for_each_netdev_rcu(&init_net, dev)
182 183
		seq_printf(sf, "%s %u\n", dev->name,
			   netprio_prio(seq_css(sf), dev));
184 185 186 187
	rcu_read_unlock();
	return 0;
}

188
static int write_priomap(struct cgroup_subsys_state *css, struct cftype *cft,
189
			 char *buffer)
190
{
191
	char devname[IFNAMSIZ + 1];
192
	struct net_device *dev;
193 194
	u32 prio;
	int ret;
195

196 197
	if (sscanf(buffer, "%"__stringify(IFNAMSIZ)"s %u", devname, &prio) != 2)
		return -EINVAL;
198 199 200

	dev = dev_get_by_name(&init_net, devname);
	if (!dev)
201
		return -ENODEV;
202

203
	rtnl_lock();
204

205
	ret = netprio_set_prio(css, dev, prio);
206

207
	rtnl_unlock();
208 209 210 211
	dev_put(dev);
	return ret;
}

A
Al Viro 已提交
212 213 214 215 216 217 218 219 220
static int update_netprio(const void *v, struct file *file, unsigned n)
{
	int err;
	struct socket *sock = sock_from_file(file, &err);
	if (sock)
		sock->sk->sk_cgrp_prioidx = (u32)(unsigned long)v;
	return 0;
}

221 222
static void net_prio_attach(struct cgroup_subsys_state *css,
			    struct cgroup_taskset *tset)
223 224
{
	struct task_struct *p;
225
	void *v = (void *)(unsigned long)css->cgroup->id;
226

227
	cgroup_taskset_for_each(p, tset) {
228
		task_lock(p);
A
Al Viro 已提交
229
		iterate_fd(p->files, 0, update_netprio, v);
230 231 232 233
		task_unlock(p);
	}
}

234 235 236 237 238 239 240
static struct cftype ss_files[] = {
	{
		.name = "prioidx",
		.read_u64 = read_prioidx,
	},
	{
		.name = "ifpriomap",
241
		.seq_show = read_priomap,
242 243
		.write_string = write_priomap,
	},
244
	{ }	/* terminate */
245 246
};

247
struct cgroup_subsys net_prio_cgrp_subsys = {
248
	.css_alloc	= cgrp_css_alloc,
249
	.css_online	= cgrp_css_online,
250
	.css_free	= cgrp_css_free,
251
	.attach		= net_prio_attach,
252
	.base_cftypes	= ss_files,
253
};
254 255 256 257

static int netprio_device_event(struct notifier_block *unused,
				unsigned long event, void *ptr)
{
258
	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
259 260 261 262 263 264 265 266 267 268
	struct netprio_map *old;

	/*
	 * Note this is called with rtnl_lock held so we have update side
	 * protection on our rcu assignments
	 */

	switch (event) {
	case NETDEV_UNREGISTER:
		old = rtnl_dereference(dev->priomap);
269
		RCU_INIT_POINTER(dev->priomap, NULL);
270 271 272 273 274 275 276 277 278 279 280 281 282 283
		if (old)
			kfree_rcu(old, rcu);
		break;
	}
	return NOTIFY_DONE;
}

static struct notifier_block netprio_device_notifier = {
	.notifier_call = netprio_device_event
};

static int __init init_cgroup_netprio(void)
{
	register_netdevice_notifier(&netprio_device_notifier);
284
	return 0;
285 286
}

287
subsys_initcall(init_cgroup_netprio);
288
MODULE_LICENSE("GPL v2");