xref: /openbmc/linux/kernel/bpf/devmap.c (revision 7211ec63)
1 /* Copyright (c) 2017 Covalent IO, Inc. http://covalent.io
2  *
3  * This program is free software; you can redistribute it and/or
4  * modify it under the terms of version 2 of the GNU General Public
5  * License as published by the Free Software Foundation.
6  *
7  * This program is distributed in the hope that it will be useful, but
8  * WITHOUT ANY WARRANTY; without even the implied warranty of
9  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
10  * General Public License for more details.
11  */
12 
13 /* Devmaps primary use is as a backend map for XDP BPF helper call
14  * bpf_redirect_map(). Because XDP is mostly concerned with performance we
15  * spent some effort to ensure the datapath with redirect maps does not use
16  * any locking. This is a quick note on the details.
17  *
18  * We have three possible paths to get into the devmap control plane bpf
19  * syscalls, bpf programs, and driver side xmit/flush operations. A bpf syscall
20  * will invoke an update, delete, or lookup operation. To ensure updates and
21  * deletes appear atomic from the datapath side xchg() is used to modify the
22  * netdev_map array. Then because the datapath does a lookup into the netdev_map
23  * array (read-only) from an RCU critical section we use call_rcu() to wait for
24  * an rcu grace period before free'ing the old data structures. This ensures the
25  * datapath always has a valid copy. However, the datapath does a "flush"
26  * operation that pushes any pending packets in the driver outside the RCU
27  * critical section. Each bpf_dtab_netdev tracks these pending operations using
28  * an atomic per-cpu bitmap. The bpf_dtab_netdev object will not be destroyed
29  * until all bits are cleared indicating outstanding flush operations have
30  * completed.
31  *
32  * BPF syscalls may race with BPF program calls on any of the update, delete
33  * or lookup operations. As noted above the xchg() operation also keep the
34  * netdev_map consistent in this case. From the devmap side BPF programs
35  * calling into these operations are the same as multiple user space threads
36  * making system calls.
37  *
38  * Finally, any of the above may race with a netdev_unregister notifier. The
39  * unregister notifier must search for net devices in the map structure that
40  * contain a reference to the net device and remove them. This is a two step
41  * process (a) dereference the bpf_dtab_netdev object in netdev_map and (b)
42  * check to see if the ifindex is the same as the net_device being removed.
43  * When removing the dev a cmpxchg() is used to ensure the correct dev is
44  * removed, in the case of a concurrent update or delete operation it is
45  * possible that the initially referenced dev is no longer in the map. As the
46  * notifier hook walks the map we know that new dev references can not be
47  * added by the user because core infrastructure ensures dev_get_by_index()
48  * calls will fail at this point.
49  */
50 #include <linux/bpf.h>
51 #include <linux/filter.h>
52 
53 struct bpf_dtab_netdev {
54 	struct net_device *dev;
55 	struct bpf_dtab *dtab;
56 	unsigned int bit;
57 	struct rcu_head rcu;
58 };
59 
60 struct bpf_dtab {
61 	struct bpf_map map;
62 	struct bpf_dtab_netdev **netdev_map;
63 	unsigned long __percpu *flush_needed;
64 	struct list_head list;
65 };
66 
67 static DEFINE_SPINLOCK(dev_map_lock);
68 static LIST_HEAD(dev_map_list);
69 
70 static u64 dev_map_bitmap_size(const union bpf_attr *attr)
71 {
72 	return BITS_TO_LONGS(attr->max_entries) * sizeof(unsigned long);
73 }
74 
75 static struct bpf_map *dev_map_alloc(union bpf_attr *attr)
76 {
77 	struct bpf_dtab *dtab;
78 	int err = -EINVAL;
79 	u64 cost;
80 
81 	/* check sanity of attributes */
82 	if (attr->max_entries == 0 || attr->key_size != 4 ||
83 	    attr->value_size != 4 || attr->map_flags & ~BPF_F_NUMA_NODE)
84 		return ERR_PTR(-EINVAL);
85 
86 	dtab = kzalloc(sizeof(*dtab), GFP_USER);
87 	if (!dtab)
88 		return ERR_PTR(-ENOMEM);
89 
90 	/* mandatory map attributes */
91 	dtab->map.map_type = attr->map_type;
92 	dtab->map.key_size = attr->key_size;
93 	dtab->map.value_size = attr->value_size;
94 	dtab->map.max_entries = attr->max_entries;
95 	dtab->map.map_flags = attr->map_flags;
96 	dtab->map.numa_node = bpf_map_attr_numa_node(attr);
97 
98 	/* make sure page count doesn't overflow */
99 	cost = (u64) dtab->map.max_entries * sizeof(struct bpf_dtab_netdev *);
100 	cost += dev_map_bitmap_size(attr) * num_possible_cpus();
101 	if (cost >= U32_MAX - PAGE_SIZE)
102 		goto free_dtab;
103 
104 	dtab->map.pages = round_up(cost, PAGE_SIZE) >> PAGE_SHIFT;
105 
106 	/* if map size is larger than memlock limit, reject it early */
107 	err = bpf_map_precharge_memlock(dtab->map.pages);
108 	if (err)
109 		goto free_dtab;
110 
111 	err = -ENOMEM;
112 
113 	/* A per cpu bitfield with a bit per possible net device */
114 	dtab->flush_needed = __alloc_percpu(dev_map_bitmap_size(attr),
115 					    __alignof__(unsigned long));
116 	if (!dtab->flush_needed)
117 		goto free_dtab;
118 
119 	dtab->netdev_map = bpf_map_area_alloc(dtab->map.max_entries *
120 					      sizeof(struct bpf_dtab_netdev *),
121 					      dtab->map.numa_node);
122 	if (!dtab->netdev_map)
123 		goto free_dtab;
124 
125 	spin_lock(&dev_map_lock);
126 	list_add_tail_rcu(&dtab->list, &dev_map_list);
127 	spin_unlock(&dev_map_lock);
128 
129 	return &dtab->map;
130 free_dtab:
131 	free_percpu(dtab->flush_needed);
132 	kfree(dtab);
133 	return ERR_PTR(err);
134 }
135 
136 static void dev_map_free(struct bpf_map *map)
137 {
138 	struct bpf_dtab *dtab = container_of(map, struct bpf_dtab, map);
139 	int i, cpu;
140 
141 	/* At this point bpf_prog->aux->refcnt == 0 and this map->refcnt == 0,
142 	 * so the programs (can be more than one that used this map) were
143 	 * disconnected from events. Wait for outstanding critical sections in
144 	 * these programs to complete. The rcu critical section only guarantees
145 	 * no further reads against netdev_map. It does __not__ ensure pending
146 	 * flush operations (if any) are complete.
147 	 */
148 
149 	spin_lock(&dev_map_lock);
150 	list_del_rcu(&dtab->list);
151 	spin_unlock(&dev_map_lock);
152 
153 	synchronize_rcu();
154 
155 	/* To ensure all pending flush operations have completed wait for flush
156 	 * bitmap to indicate all flush_needed bits to be zero on _all_ cpus.
157 	 * Because the above synchronize_rcu() ensures the map is disconnected
158 	 * from the program we can assume no new bits will be set.
159 	 */
160 	for_each_online_cpu(cpu) {
161 		unsigned long *bitmap = per_cpu_ptr(dtab->flush_needed, cpu);
162 
163 		while (!bitmap_empty(bitmap, dtab->map.max_entries))
164 			cond_resched();
165 	}
166 
167 	for (i = 0; i < dtab->map.max_entries; i++) {
168 		struct bpf_dtab_netdev *dev;
169 
170 		dev = dtab->netdev_map[i];
171 		if (!dev)
172 			continue;
173 
174 		dev_put(dev->dev);
175 		kfree(dev);
176 	}
177 
178 	free_percpu(dtab->flush_needed);
179 	bpf_map_area_free(dtab->netdev_map);
180 	kfree(dtab);
181 }
182 
183 static int dev_map_get_next_key(struct bpf_map *map, void *key, void *next_key)
184 {
185 	struct bpf_dtab *dtab = container_of(map, struct bpf_dtab, map);
186 	u32 index = key ? *(u32 *)key : U32_MAX;
187 	u32 *next = next_key;
188 
189 	if (index >= dtab->map.max_entries) {
190 		*next = 0;
191 		return 0;
192 	}
193 
194 	if (index == dtab->map.max_entries - 1)
195 		return -ENOENT;
196 	*next = index + 1;
197 	return 0;
198 }
199 
200 void __dev_map_insert_ctx(struct bpf_map *map, u32 bit)
201 {
202 	struct bpf_dtab *dtab = container_of(map, struct bpf_dtab, map);
203 	unsigned long *bitmap = this_cpu_ptr(dtab->flush_needed);
204 
205 	__set_bit(bit, bitmap);
206 }
207 
208 /* __dev_map_flush is called from xdp_do_flush_map() which _must_ be signaled
209  * from the driver before returning from its napi->poll() routine. The poll()
210  * routine is called either from busy_poll context or net_rx_action signaled
211  * from NET_RX_SOFTIRQ. Either way the poll routine must complete before the
212  * net device can be torn down. On devmap tear down we ensure the ctx bitmap
213  * is zeroed before completing to ensure all flush operations have completed.
214  */
215 void __dev_map_flush(struct bpf_map *map)
216 {
217 	struct bpf_dtab *dtab = container_of(map, struct bpf_dtab, map);
218 	unsigned long *bitmap = this_cpu_ptr(dtab->flush_needed);
219 	u32 bit;
220 
221 	for_each_set_bit(bit, bitmap, map->max_entries) {
222 		struct bpf_dtab_netdev *dev = READ_ONCE(dtab->netdev_map[bit]);
223 		struct net_device *netdev;
224 
225 		/* This is possible if the dev entry is removed by user space
226 		 * between xdp redirect and flush op.
227 		 */
228 		if (unlikely(!dev))
229 			continue;
230 
231 		__clear_bit(bit, bitmap);
232 		netdev = dev->dev;
233 		if (likely(netdev->netdev_ops->ndo_xdp_flush))
234 			netdev->netdev_ops->ndo_xdp_flush(netdev);
235 	}
236 }
237 
238 /* rcu_read_lock (from syscall and BPF contexts) ensures that if a delete and/or
239  * update happens in parallel here a dev_put wont happen until after reading the
240  * ifindex.
241  */
242 struct net_device  *__dev_map_lookup_elem(struct bpf_map *map, u32 key)
243 {
244 	struct bpf_dtab *dtab = container_of(map, struct bpf_dtab, map);
245 	struct bpf_dtab_netdev *dev;
246 
247 	if (key >= map->max_entries)
248 		return NULL;
249 
250 	dev = READ_ONCE(dtab->netdev_map[key]);
251 	return dev ? dev->dev : NULL;
252 }
253 
254 static void *dev_map_lookup_elem(struct bpf_map *map, void *key)
255 {
256 	struct net_device *dev = __dev_map_lookup_elem(map, *(u32 *)key);
257 
258 	return dev ? &dev->ifindex : NULL;
259 }
260 
261 static void dev_map_flush_old(struct bpf_dtab_netdev *dev)
262 {
263 	if (dev->dev->netdev_ops->ndo_xdp_flush) {
264 		struct net_device *fl = dev->dev;
265 		unsigned long *bitmap;
266 		int cpu;
267 
268 		for_each_online_cpu(cpu) {
269 			bitmap = per_cpu_ptr(dev->dtab->flush_needed, cpu);
270 			__clear_bit(dev->bit, bitmap);
271 
272 			fl->netdev_ops->ndo_xdp_flush(dev->dev);
273 		}
274 	}
275 }
276 
277 static void __dev_map_entry_free(struct rcu_head *rcu)
278 {
279 	struct bpf_dtab_netdev *dev;
280 
281 	dev = container_of(rcu, struct bpf_dtab_netdev, rcu);
282 	dev_map_flush_old(dev);
283 	dev_put(dev->dev);
284 	kfree(dev);
285 }
286 
287 static int dev_map_delete_elem(struct bpf_map *map, void *key)
288 {
289 	struct bpf_dtab *dtab = container_of(map, struct bpf_dtab, map);
290 	struct bpf_dtab_netdev *old_dev;
291 	int k = *(u32 *)key;
292 
293 	if (k >= map->max_entries)
294 		return -EINVAL;
295 
296 	/* Use call_rcu() here to ensure any rcu critical sections have
297 	 * completed, but this does not guarantee a flush has happened
298 	 * yet. Because driver side rcu_read_lock/unlock only protects the
299 	 * running XDP program. However, for pending flush operations the
300 	 * dev and ctx are stored in another per cpu map. And additionally,
301 	 * the driver tear down ensures all soft irqs are complete before
302 	 * removing the net device in the case of dev_put equals zero.
303 	 */
304 	old_dev = xchg(&dtab->netdev_map[k], NULL);
305 	if (old_dev)
306 		call_rcu(&old_dev->rcu, __dev_map_entry_free);
307 	return 0;
308 }
309 
310 static int dev_map_update_elem(struct bpf_map *map, void *key, void *value,
311 				u64 map_flags)
312 {
313 	struct bpf_dtab *dtab = container_of(map, struct bpf_dtab, map);
314 	struct net *net = current->nsproxy->net_ns;
315 	struct bpf_dtab_netdev *dev, *old_dev;
316 	u32 i = *(u32 *)key;
317 	u32 ifindex = *(u32 *)value;
318 
319 	if (unlikely(map_flags > BPF_EXIST))
320 		return -EINVAL;
321 	if (unlikely(i >= dtab->map.max_entries))
322 		return -E2BIG;
323 	if (unlikely(map_flags == BPF_NOEXIST))
324 		return -EEXIST;
325 
326 	if (!ifindex) {
327 		dev = NULL;
328 	} else {
329 		dev = kmalloc_node(sizeof(*dev), GFP_ATOMIC | __GFP_NOWARN,
330 				   map->numa_node);
331 		if (!dev)
332 			return -ENOMEM;
333 
334 		dev->dev = dev_get_by_index(net, ifindex);
335 		if (!dev->dev) {
336 			kfree(dev);
337 			return -EINVAL;
338 		}
339 
340 		dev->bit = i;
341 		dev->dtab = dtab;
342 	}
343 
344 	/* Use call_rcu() here to ensure rcu critical sections have completed
345 	 * Remembering the driver side flush operation will happen before the
346 	 * net device is removed.
347 	 */
348 	old_dev = xchg(&dtab->netdev_map[i], dev);
349 	if (old_dev)
350 		call_rcu(&old_dev->rcu, __dev_map_entry_free);
351 
352 	return 0;
353 }
354 
355 const struct bpf_map_ops dev_map_ops = {
356 	.map_alloc = dev_map_alloc,
357 	.map_free = dev_map_free,
358 	.map_get_next_key = dev_map_get_next_key,
359 	.map_lookup_elem = dev_map_lookup_elem,
360 	.map_update_elem = dev_map_update_elem,
361 	.map_delete_elem = dev_map_delete_elem,
362 };
363 
364 static int dev_map_notification(struct notifier_block *notifier,
365 				ulong event, void *ptr)
366 {
367 	struct net_device *netdev = netdev_notifier_info_to_dev(ptr);
368 	struct bpf_dtab *dtab;
369 	int i;
370 
371 	switch (event) {
372 	case NETDEV_UNREGISTER:
373 		/* This rcu_read_lock/unlock pair is needed because
374 		 * dev_map_list is an RCU list AND to ensure a delete
375 		 * operation does not free a netdev_map entry while we
376 		 * are comparing it against the netdev being unregistered.
377 		 */
378 		rcu_read_lock();
379 		list_for_each_entry_rcu(dtab, &dev_map_list, list) {
380 			for (i = 0; i < dtab->map.max_entries; i++) {
381 				struct bpf_dtab_netdev *dev, *odev;
382 
383 				dev = READ_ONCE(dtab->netdev_map[i]);
384 				if (!dev ||
385 				    dev->dev->ifindex != netdev->ifindex)
386 					continue;
387 				odev = cmpxchg(&dtab->netdev_map[i], dev, NULL);
388 				if (dev == odev)
389 					call_rcu(&dev->rcu,
390 						 __dev_map_entry_free);
391 			}
392 		}
393 		rcu_read_unlock();
394 		break;
395 	default:
396 		break;
397 	}
398 	return NOTIFY_OK;
399 }
400 
401 static struct notifier_block dev_map_notifier = {
402 	.notifier_call = dev_map_notification,
403 };
404 
405 static int __init dev_map_init(void)
406 {
407 	register_netdevice_notifier(&dev_map_notifier);
408 	return 0;
409 }
410 
411 subsys_initcall(dev_map_init);
412