1 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
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.
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.
12 #include <linux/bpf.h>
13 #include <linux/err.h>
14 #include <linux/vmalloc.h>
15 #include <linux/slab.h>
17 #include <linux/filter.h>
18 #include <linux/perf_event.h>
20 /* Called from syscall */
21 static struct bpf_map *array_map_alloc(union bpf_attr *attr)
23 struct bpf_array *array;
24 u32 elem_size, array_size;
26 /* check sanity of attributes */
27 if (attr->max_entries == 0 || attr->key_size != 4 ||
28 attr->value_size == 0)
29 return ERR_PTR(-EINVAL);
31 if (attr->value_size >= 1 << (KMALLOC_SHIFT_MAX - 1))
32 /* if value_size is bigger, the user space won't be able to
33 * access the elements.
35 return ERR_PTR(-E2BIG);
37 elem_size = round_up(attr->value_size, 8);
39 /* check round_up into zero and u32 overflow */
41 attr->max_entries > (U32_MAX - PAGE_SIZE - sizeof(*array)) / elem_size)
42 return ERR_PTR(-ENOMEM);
44 array_size = sizeof(*array) + attr->max_entries * elem_size;
46 /* allocate all map elements and zero-initialize them */
47 array = kzalloc(array_size, GFP_USER | __GFP_NOWARN);
49 array = vzalloc(array_size);
51 return ERR_PTR(-ENOMEM);
54 /* copy mandatory map attributes */
55 array->map.key_size = attr->key_size;
56 array->map.value_size = attr->value_size;
57 array->map.max_entries = attr->max_entries;
58 array->map.pages = round_up(array_size, PAGE_SIZE) >> PAGE_SHIFT;
59 array->elem_size = elem_size;
64 /* Called from syscall or from eBPF program */
65 static void *array_map_lookup_elem(struct bpf_map *map, void *key)
67 struct bpf_array *array = container_of(map, struct bpf_array, map);
68 u32 index = *(u32 *)key;
70 if (index >= array->map.max_entries)
73 return array->value + array->elem_size * index;
76 /* Called from syscall */
77 static int array_map_get_next_key(struct bpf_map *map, void *key, void *next_key)
79 struct bpf_array *array = container_of(map, struct bpf_array, map);
80 u32 index = *(u32 *)key;
81 u32 *next = (u32 *)next_key;
83 if (index >= array->map.max_entries) {
88 if (index == array->map.max_entries - 1)
95 /* Called from syscall or from eBPF program */
96 static int array_map_update_elem(struct bpf_map *map, void *key, void *value,
99 struct bpf_array *array = container_of(map, struct bpf_array, map);
100 u32 index = *(u32 *)key;
102 if (map_flags > BPF_EXIST)
106 if (index >= array->map.max_entries)
107 /* all elements were pre-allocated, cannot insert a new one */
110 if (map_flags == BPF_NOEXIST)
111 /* all elements already exist */
114 memcpy(array->value + array->elem_size * index, value, map->value_size);
118 /* Called from syscall or from eBPF program */
119 static int array_map_delete_elem(struct bpf_map *map, void *key)
124 /* Called when map->refcnt goes to zero, either from workqueue or from syscall */
125 static void array_map_free(struct bpf_map *map)
127 struct bpf_array *array = container_of(map, struct bpf_array, map);
129 /* at this point bpf_prog->aux->refcnt == 0 and this map->refcnt == 0,
130 * so the programs (can be more than one that used this map) were
131 * disconnected from events. Wait for outstanding programs to complete
139 static const struct bpf_map_ops array_ops = {
140 .map_alloc = array_map_alloc,
141 .map_free = array_map_free,
142 .map_get_next_key = array_map_get_next_key,
143 .map_lookup_elem = array_map_lookup_elem,
144 .map_update_elem = array_map_update_elem,
145 .map_delete_elem = array_map_delete_elem,
148 static struct bpf_map_type_list array_type __read_mostly = {
150 .type = BPF_MAP_TYPE_ARRAY,
153 static int __init register_array_map(void)
155 bpf_register_map_type(&array_type);
158 late_initcall(register_array_map);
160 static struct bpf_map *fd_array_map_alloc(union bpf_attr *attr)
162 /* only file descriptors can be stored in this type of map */
163 if (attr->value_size != sizeof(u32))
164 return ERR_PTR(-EINVAL);
165 return array_map_alloc(attr);
168 static void fd_array_map_free(struct bpf_map *map)
170 struct bpf_array *array = container_of(map, struct bpf_array, map);
175 /* make sure it's empty */
176 for (i = 0; i < array->map.max_entries; i++)
177 BUG_ON(array->ptrs[i] != NULL);
181 static void *fd_array_map_lookup_elem(struct bpf_map *map, void *key)
186 /* only called from syscall */
187 static int fd_array_map_update_elem(struct bpf_map *map, void *key,
188 void *value, u64 map_flags)
190 struct bpf_array *array = container_of(map, struct bpf_array, map);
191 void *new_ptr, *old_ptr;
192 u32 index = *(u32 *)key, ufd;
194 if (map_flags != BPF_ANY)
197 if (index >= array->map.max_entries)
201 new_ptr = map->ops->map_fd_get_ptr(map, ufd);
203 return PTR_ERR(new_ptr);
205 old_ptr = xchg(array->ptrs + index, new_ptr);
207 map->ops->map_fd_put_ptr(old_ptr);
212 static int fd_array_map_delete_elem(struct bpf_map *map, void *key)
214 struct bpf_array *array = container_of(map, struct bpf_array, map);
216 u32 index = *(u32 *)key;
218 if (index >= array->map.max_entries)
221 old_ptr = xchg(array->ptrs + index, NULL);
223 map->ops->map_fd_put_ptr(old_ptr);
230 static void *prog_fd_array_get_ptr(struct bpf_map *map, int fd)
232 struct bpf_array *array = container_of(map, struct bpf_array, map);
233 struct bpf_prog *prog = bpf_prog_get(fd);
237 if (!bpf_prog_array_compatible(array, prog)) {
239 return ERR_PTR(-EINVAL);
244 static void prog_fd_array_put_ptr(void *ptr)
246 struct bpf_prog *prog = ptr;
248 bpf_prog_put_rcu(prog);
251 /* decrement refcnt of all bpf_progs that are stored in this map */
252 void bpf_fd_array_map_clear(struct bpf_map *map)
254 struct bpf_array *array = container_of(map, struct bpf_array, map);
257 for (i = 0; i < array->map.max_entries; i++)
258 fd_array_map_delete_elem(map, &i);
261 static const struct bpf_map_ops prog_array_ops = {
262 .map_alloc = fd_array_map_alloc,
263 .map_free = fd_array_map_free,
264 .map_get_next_key = array_map_get_next_key,
265 .map_lookup_elem = fd_array_map_lookup_elem,
266 .map_update_elem = fd_array_map_update_elem,
267 .map_delete_elem = fd_array_map_delete_elem,
268 .map_fd_get_ptr = prog_fd_array_get_ptr,
269 .map_fd_put_ptr = prog_fd_array_put_ptr,
272 static struct bpf_map_type_list prog_array_type __read_mostly = {
273 .ops = &prog_array_ops,
274 .type = BPF_MAP_TYPE_PROG_ARRAY,
277 static int __init register_prog_array_map(void)
279 bpf_register_map_type(&prog_array_type);
282 late_initcall(register_prog_array_map);
284 static void perf_event_array_map_free(struct bpf_map *map)
286 bpf_fd_array_map_clear(map);
287 fd_array_map_free(map);
290 static void *perf_event_fd_array_get_ptr(struct bpf_map *map, int fd)
292 struct perf_event *event;
293 const struct perf_event_attr *attr;
296 file = perf_event_get(fd);
300 event = file->private_data;
302 attr = perf_event_attrs(event);
309 if (attr->type == PERF_TYPE_RAW)
312 if (attr->type == PERF_TYPE_HARDWARE)
315 if (attr->type == PERF_TYPE_SOFTWARE &&
316 attr->config == PERF_COUNT_SW_BPF_OUTPUT)
320 return ERR_PTR(-EINVAL);
323 static void perf_event_fd_array_put_ptr(void *ptr)
325 fput((struct file *)ptr);
328 static const struct bpf_map_ops perf_event_array_ops = {
329 .map_alloc = fd_array_map_alloc,
330 .map_free = perf_event_array_map_free,
331 .map_get_next_key = array_map_get_next_key,
332 .map_lookup_elem = fd_array_map_lookup_elem,
333 .map_update_elem = fd_array_map_update_elem,
334 .map_delete_elem = fd_array_map_delete_elem,
335 .map_fd_get_ptr = perf_event_fd_array_get_ptr,
336 .map_fd_put_ptr = perf_event_fd_array_put_ptr,
339 static struct bpf_map_type_list perf_event_array_type __read_mostly = {
340 .ops = &perf_event_array_ops,
341 .type = BPF_MAP_TYPE_PERF_EVENT_ARRAY,
344 static int __init register_perf_event_array_map(void)
346 bpf_register_map_type(&perf_event_array_type);
349 late_initcall(register_perf_event_array_map);