projects/linux·blob v4.17 · refs

412 lines · 12.3 KiB

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#define DEV_CREATE_FLAG_MASK \
54 (BPF_F_NUMA_NODE | BPF_F_RDONLY | BPF_F_WRONLY)
55
56struct bpf_dtab_netdev {
57 struct net_device *dev;
58 struct bpf_dtab *dtab;
59 unsigned int bit;
60 struct rcu_head rcu;
61};
62
63struct bpf_dtab {
64 struct bpf_map map;
65 struct bpf_dtab_netdev **netdev_map;
66 unsigned long __percpu *flush_needed;
67 struct list_head list;
68};
69
70static DEFINE_SPINLOCK(dev_map_lock);
71static LIST_HEAD(dev_map_list);
72
73static u64 dev_map_bitmap_size(const union bpf_attr *attr)
74{
75 return BITS_TO_LONGS((u64) attr->max_entries) * sizeof(unsigned long);
76}
77
78static struct bpf_map *dev_map_alloc(union bpf_attr *attr)
79{
80 struct bpf_dtab *dtab;
81 int err = -EINVAL;
82 u64 cost;
83
84 if (!capable(CAP_NET_ADMIN))
85 return ERR_PTR(-EPERM);
86
87 /* check sanity of attributes */
88 if (attr->max_entries == 0 || attr->key_size != 4 ||
89 attr->value_size != 4 || attr->map_flags & ~DEV_CREATE_FLAG_MASK)
90 return ERR_PTR(-EINVAL);
91
92 dtab = kzalloc(sizeof(*dtab), GFP_USER);
93 if (!dtab)
94 return ERR_PTR(-ENOMEM);
95
96 bpf_map_init_from_attr(&dtab->map, 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_gfp(dev_map_bitmap_size(attr),
115 __alignof__(unsigned long),
116 GFP_KERNEL | __GFP_NOWARN);
117 if (!dtab->flush_needed)
118 goto free_dtab;
119
120 dtab->netdev_map = bpf_map_area_alloc(dtab->map.max_entries *
121 sizeof(struct bpf_dtab_netdev *),
122 dtab->map.numa_node);
123 if (!dtab->netdev_map)
124 goto free_dtab;
125
126 spin_lock(&dev_map_lock);
127 list_add_tail_rcu(&dtab->list, &dev_map_list);
128 spin_unlock(&dev_map_lock);
129
130 return &dtab->map;
131free_dtab:
132 free_percpu(dtab->flush_needed);
133 kfree(dtab);
134 return ERR_PTR(err);
135}
136
137static void dev_map_free(struct bpf_map *map)
138{
139 struct bpf_dtab *dtab = container_of(map, struct bpf_dtab, map);
140 int i, cpu;
141
142 /* At this point bpf_prog->aux->refcnt == 0 and this map->refcnt == 0,
143 * so the programs (can be more than one that used this map) were
144 * disconnected from events. Wait for outstanding critical sections in
145 * these programs to complete. The rcu critical section only guarantees
146 * no further reads against netdev_map. It does __not__ ensure pending
147 * flush operations (if any) are complete.
148 */
149
150 spin_lock(&dev_map_lock);
151 list_del_rcu(&dtab->list);
152 spin_unlock(&dev_map_lock);
153
154 synchronize_rcu();
155
156 /* To ensure all pending flush operations have completed wait for flush
157 * bitmap to indicate all flush_needed bits to be zero on _all_ cpus.
158 * Because the above synchronize_rcu() ensures the map is disconnected
159 * from the program we can assume no new bits will be set.
160 */
161 for_each_online_cpu(cpu) {
162 unsigned long *bitmap = per_cpu_ptr(dtab->flush_needed, cpu);
163
164 while (!bitmap_empty(bitmap, dtab->map.max_entries))
165 cond_resched();
166 }
167
168 for (i = 0; i < dtab->map.max_entries; i++) {
169 struct bpf_dtab_netdev *dev;
170
171 dev = dtab->netdev_map[i];
172 if (!dev)
173 continue;
174
175 dev_put(dev->dev);
176 kfree(dev);
177 }
178
179 free_percpu(dtab->flush_needed);
180 bpf_map_area_free(dtab->netdev_map);
181 kfree(dtab);
182}
183
184static int dev_map_get_next_key(struct bpf_map *map, void *key, void *next_key)
185{
186 struct bpf_dtab *dtab = container_of(map, struct bpf_dtab, map);
187 u32 index = key ? *(u32 *)key : U32_MAX;
188 u32 *next = next_key;
189
190 if (index >= dtab->map.max_entries) {
191 *next = 0;
192 return 0;
193 }
194
195 if (index == dtab->map.max_entries - 1)
196 return -ENOENT;
197 *next = index + 1;
198 return 0;
199}
200
201void __dev_map_insert_ctx(struct bpf_map *map, u32 bit)
202{
203 struct bpf_dtab *dtab = container_of(map, struct bpf_dtab, map);
204 unsigned long *bitmap = this_cpu_ptr(dtab->flush_needed);
205
206 __set_bit(bit, bitmap);
207}
208
209/* __dev_map_flush is called from xdp_do_flush_map() which _must_ be signaled
210 * from the driver before returning from its napi->poll() routine. The poll()
211 * routine is called either from busy_poll context or net_rx_action signaled
212 * from NET_RX_SOFTIRQ. Either way the poll routine must complete before the
213 * net device can be torn down. On devmap tear down we ensure the ctx bitmap
214 * is zeroed before completing to ensure all flush operations have completed.
215 */
216void __dev_map_flush(struct bpf_map *map)
217{
218 struct bpf_dtab *dtab = container_of(map, struct bpf_dtab, map);
219 unsigned long *bitmap = this_cpu_ptr(dtab->flush_needed);
220 u32 bit;
221
222 for_each_set_bit(bit, bitmap, map->max_entries) {
223 struct bpf_dtab_netdev *dev = READ_ONCE(dtab->netdev_map[bit]);
224 struct net_device *netdev;
225
226 /* This is possible if the dev entry is removed by user space
227 * between xdp redirect and flush op.
228 */
229 if (unlikely(!dev))
230 continue;
231
232 __clear_bit(bit, bitmap);
233 netdev = dev->dev;
234 if (likely(netdev->netdev_ops->ndo_xdp_flush))
235 netdev->netdev_ops->ndo_xdp_flush(netdev);
236 }
237}
238
239/* rcu_read_lock (from syscall and BPF contexts) ensures that if a delete and/or
240 * update happens in parallel here a dev_put wont happen until after reading the
241 * ifindex.
242 */
243struct net_device *__dev_map_lookup_elem(struct bpf_map *map, u32 key)
244{
245 struct bpf_dtab *dtab = container_of(map, struct bpf_dtab, map);
246 struct bpf_dtab_netdev *dev;
247
248 if (key >= map->max_entries)
249 return NULL;
250
251 dev = READ_ONCE(dtab->netdev_map[key]);
252 return dev ? dev->dev : NULL;
253}
254
255static void *dev_map_lookup_elem(struct bpf_map *map, void *key)
256{
257 struct net_device *dev = __dev_map_lookup_elem(map, *(u32 *)key);
258
259 return dev ? &dev->ifindex : NULL;
260}
261
262static void dev_map_flush_old(struct bpf_dtab_netdev *dev)
263{
264 if (dev->dev->netdev_ops->ndo_xdp_flush) {
265 struct net_device *fl = dev->dev;
266 unsigned long *bitmap;
267 int cpu;
268
269 for_each_online_cpu(cpu) {
270 bitmap = per_cpu_ptr(dev->dtab->flush_needed, cpu);
271 __clear_bit(dev->bit, bitmap);
272
273 fl->netdev_ops->ndo_xdp_flush(dev->dev);
274 }
275 }
276}
277
278static void __dev_map_entry_free(struct rcu_head *rcu)
279{
280 struct bpf_dtab_netdev *dev;
281
282 dev = container_of(rcu, struct bpf_dtab_netdev, rcu);
283 dev_map_flush_old(dev);
284 dev_put(dev->dev);
285 kfree(dev);
286}
287
288static int dev_map_delete_elem(struct bpf_map *map, void *key)
289{
290 struct bpf_dtab *dtab = container_of(map, struct bpf_dtab, map);
291 struct bpf_dtab_netdev *old_dev;
292 int k = *(u32 *)key;
293
294 if (k >= map->max_entries)
295 return -EINVAL;
296
297 /* Use call_rcu() here to ensure any rcu critical sections have
298 * completed, but this does not guarantee a flush has happened
299 * yet. Because driver side rcu_read_lock/unlock only protects the
300 * running XDP program. However, for pending flush operations the
301 * dev and ctx are stored in another per cpu map. And additionally,
302 * the driver tear down ensures all soft irqs are complete before
303 * removing the net device in the case of dev_put equals zero.
304 */
305 old_dev = xchg(&dtab->netdev_map[k], NULL);
306 if (old_dev)
307 call_rcu(&old_dev->rcu, __dev_map_entry_free);
308 return 0;
309}
310
311static int dev_map_update_elem(struct bpf_map *map, void *key, void *value,
312 u64 map_flags)
313{
314 struct bpf_dtab *dtab = container_of(map, struct bpf_dtab, map);
315 struct net *net = current->nsproxy->net_ns;
316 struct bpf_dtab_netdev *dev, *old_dev;
317 u32 i = *(u32 *)key;
318 u32 ifindex = *(u32 *)value;
319
320 if (unlikely(map_flags > BPF_EXIST))
321 return -EINVAL;
322 if (unlikely(i >= dtab->map.max_entries))
323 return -E2BIG;
324 if (unlikely(map_flags == BPF_NOEXIST))
325 return -EEXIST;
326
327 if (!ifindex) {
328 dev = NULL;
329 } else {
330 dev = kmalloc_node(sizeof(*dev), GFP_ATOMIC | __GFP_NOWARN,
331 map->numa_node);
332 if (!dev)
333 return -ENOMEM;
334
335 dev->dev = dev_get_by_index(net, ifindex);
336 if (!dev->dev) {
337 kfree(dev);
338 return -EINVAL;
339 }
340
341 dev->bit = i;
342 dev->dtab = dtab;
343 }
344
345 /* Use call_rcu() here to ensure rcu critical sections have completed
346 * Remembering the driver side flush operation will happen before the
347 * net device is removed.
348 */
349 old_dev = xchg(&dtab->netdev_map[i], dev);
350 if (old_dev)
351 call_rcu(&old_dev->rcu, __dev_map_entry_free);
352
353 return 0;
354}
355
356const struct bpf_map_ops dev_map_ops = {
357 .map_alloc = dev_map_alloc,
358 .map_free = dev_map_free,
359 .map_get_next_key = dev_map_get_next_key,
360 .map_lookup_elem = dev_map_lookup_elem,
361 .map_update_elem = dev_map_update_elem,
362 .map_delete_elem = dev_map_delete_elem,
363};
364
365static int dev_map_notification(struct notifier_block *notifier,
366 ulong event, void *ptr)
367{
368 struct net_device *netdev = netdev_notifier_info_to_dev(ptr);
369 struct bpf_dtab *dtab;
370 int i;
371
372 switch (event) {
373 case NETDEV_UNREGISTER:
374 /* This rcu_read_lock/unlock pair is needed because
375 * dev_map_list is an RCU list AND to ensure a delete
376 * operation does not free a netdev_map entry while we
377 * are comparing it against the netdev being unregistered.
378 */
379 rcu_read_lock();
380 list_for_each_entry_rcu(dtab, &dev_map_list, list) {
381 for (i = 0; i < dtab->map.max_entries; i++) {
382 struct bpf_dtab_netdev *dev, *odev;
383
384 dev = READ_ONCE(dtab->netdev_map[i]);
385 if (!dev ||
386 dev->dev->ifindex != netdev->ifindex)
387 continue;
388 odev = cmpxchg(&dtab->netdev_map[i], dev, NULL);
389 if (dev == odev)
390 call_rcu(&dev->rcu,
391 __dev_map_entry_free);
392 }
393 }
394 rcu_read_unlock();
395 break;
396 default:
397 break;
398 }
399 return NOTIFY_OK;
400}
401
402static struct notifier_block dev_map_notifier = {
403 .notifier_call = dev_map_notification,
404};
405
406static int __init dev_map_init(void)
407{
408 register_netdevice_notifier(&dev_map_notifier);
409 return 0;
410}
411
412subsys_initcall(dev_map_init);