xref: /openbmc/linux/kernel/bpf/queue_stack_maps.c (revision 8d81cd1a)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * queue_stack_maps.c: BPF queue and stack maps
4  *
5  * Copyright (c) 2018 Politecnico di Torino
6  */
7 #include <linux/bpf.h>
8 #include <linux/list.h>
9 #include <linux/slab.h>
10 #include <linux/btf_ids.h>
11 #include "percpu_freelist.h"
12 
13 #define QUEUE_STACK_CREATE_FLAG_MASK \
14 	(BPF_F_NUMA_NODE | BPF_F_ACCESS_MASK)
15 
16 struct bpf_queue_stack {
17 	struct bpf_map map;
18 	raw_spinlock_t lock;
19 	u32 head, tail;
20 	u32 size; /* max_entries + 1 */
21 
22 	char elements[] __aligned(8);
23 };
24 
25 static struct bpf_queue_stack *bpf_queue_stack(struct bpf_map *map)
26 {
27 	return container_of(map, struct bpf_queue_stack, map);
28 }
29 
30 static bool queue_stack_map_is_empty(struct bpf_queue_stack *qs)
31 {
32 	return qs->head == qs->tail;
33 }
34 
35 static bool queue_stack_map_is_full(struct bpf_queue_stack *qs)
36 {
37 	u32 head = qs->head + 1;
38 
39 	if (unlikely(head >= qs->size))
40 		head = 0;
41 
42 	return head == qs->tail;
43 }
44 
45 /* Called from syscall */
46 static int queue_stack_map_alloc_check(union bpf_attr *attr)
47 {
48 	/* check sanity of attributes */
49 	if (attr->max_entries == 0 || attr->key_size != 0 ||
50 	    attr->value_size == 0 ||
51 	    attr->map_flags & ~QUEUE_STACK_CREATE_FLAG_MASK ||
52 	    !bpf_map_flags_access_ok(attr->map_flags))
53 		return -EINVAL;
54 
55 	if (attr->value_size > KMALLOC_MAX_SIZE)
56 		/* if value_size is bigger, the user space won't be able to
57 		 * access the elements.
58 		 */
59 		return -E2BIG;
60 
61 	return 0;
62 }
63 
64 static struct bpf_map *queue_stack_map_alloc(union bpf_attr *attr)
65 {
66 	int numa_node = bpf_map_attr_numa_node(attr);
67 	struct bpf_queue_stack *qs;
68 	u64 size, queue_size;
69 
70 	size = (u64) attr->max_entries + 1;
71 	queue_size = sizeof(*qs) + size * attr->value_size;
72 
73 	qs = bpf_map_area_alloc(queue_size, numa_node);
74 	if (!qs)
75 		return ERR_PTR(-ENOMEM);
76 
77 	bpf_map_init_from_attr(&qs->map, attr);
78 
79 	qs->size = size;
80 
81 	raw_spin_lock_init(&qs->lock);
82 
83 	return &qs->map;
84 }
85 
86 /* Called when map->refcnt goes to zero, either from workqueue or from syscall */
87 static void queue_stack_map_free(struct bpf_map *map)
88 {
89 	struct bpf_queue_stack *qs = bpf_queue_stack(map);
90 
91 	bpf_map_area_free(qs);
92 }
93 
94 static long __queue_map_get(struct bpf_map *map, void *value, bool delete)
95 {
96 	struct bpf_queue_stack *qs = bpf_queue_stack(map);
97 	unsigned long flags;
98 	int err = 0;
99 	void *ptr;
100 
101 	if (in_nmi()) {
102 		if (!raw_spin_trylock_irqsave(&qs->lock, flags))
103 			return -EBUSY;
104 	} else {
105 		raw_spin_lock_irqsave(&qs->lock, flags);
106 	}
107 
108 	if (queue_stack_map_is_empty(qs)) {
109 		memset(value, 0, qs->map.value_size);
110 		err = -ENOENT;
111 		goto out;
112 	}
113 
114 	ptr = &qs->elements[qs->tail * qs->map.value_size];
115 	memcpy(value, ptr, qs->map.value_size);
116 
117 	if (delete) {
118 		if (unlikely(++qs->tail >= qs->size))
119 			qs->tail = 0;
120 	}
121 
122 out:
123 	raw_spin_unlock_irqrestore(&qs->lock, flags);
124 	return err;
125 }
126 
127 
128 static long __stack_map_get(struct bpf_map *map, void *value, bool delete)
129 {
130 	struct bpf_queue_stack *qs = bpf_queue_stack(map);
131 	unsigned long flags;
132 	int err = 0;
133 	void *ptr;
134 	u32 index;
135 
136 	if (in_nmi()) {
137 		if (!raw_spin_trylock_irqsave(&qs->lock, flags))
138 			return -EBUSY;
139 	} else {
140 		raw_spin_lock_irqsave(&qs->lock, flags);
141 	}
142 
143 	if (queue_stack_map_is_empty(qs)) {
144 		memset(value, 0, qs->map.value_size);
145 		err = -ENOENT;
146 		goto out;
147 	}
148 
149 	index = qs->head - 1;
150 	if (unlikely(index >= qs->size))
151 		index = qs->size - 1;
152 
153 	ptr = &qs->elements[index * qs->map.value_size];
154 	memcpy(value, ptr, qs->map.value_size);
155 
156 	if (delete)
157 		qs->head = index;
158 
159 out:
160 	raw_spin_unlock_irqrestore(&qs->lock, flags);
161 	return err;
162 }
163 
164 /* Called from syscall or from eBPF program */
165 static long queue_map_peek_elem(struct bpf_map *map, void *value)
166 {
167 	return __queue_map_get(map, value, false);
168 }
169 
170 /* Called from syscall or from eBPF program */
171 static long stack_map_peek_elem(struct bpf_map *map, void *value)
172 {
173 	return __stack_map_get(map, value, false);
174 }
175 
176 /* Called from syscall or from eBPF program */
177 static long queue_map_pop_elem(struct bpf_map *map, void *value)
178 {
179 	return __queue_map_get(map, value, true);
180 }
181 
182 /* Called from syscall or from eBPF program */
183 static long stack_map_pop_elem(struct bpf_map *map, void *value)
184 {
185 	return __stack_map_get(map, value, true);
186 }
187 
188 /* Called from syscall or from eBPF program */
189 static long queue_stack_map_push_elem(struct bpf_map *map, void *value,
190 				      u64 flags)
191 {
192 	struct bpf_queue_stack *qs = bpf_queue_stack(map);
193 	unsigned long irq_flags;
194 	int err = 0;
195 	void *dst;
196 
197 	/* BPF_EXIST is used to force making room for a new element in case the
198 	 * map is full
199 	 */
200 	bool replace = (flags & BPF_EXIST);
201 
202 	/* Check supported flags for queue and stack maps */
203 	if (flags & BPF_NOEXIST || flags > BPF_EXIST)
204 		return -EINVAL;
205 
206 	if (in_nmi()) {
207 		if (!raw_spin_trylock_irqsave(&qs->lock, irq_flags))
208 			return -EBUSY;
209 	} else {
210 		raw_spin_lock_irqsave(&qs->lock, irq_flags);
211 	}
212 
213 	if (queue_stack_map_is_full(qs)) {
214 		if (!replace) {
215 			err = -E2BIG;
216 			goto out;
217 		}
218 		/* advance tail pointer to overwrite oldest element */
219 		if (unlikely(++qs->tail >= qs->size))
220 			qs->tail = 0;
221 	}
222 
223 	dst = &qs->elements[qs->head * qs->map.value_size];
224 	memcpy(dst, value, qs->map.value_size);
225 
226 	if (unlikely(++qs->head >= qs->size))
227 		qs->head = 0;
228 
229 out:
230 	raw_spin_unlock_irqrestore(&qs->lock, irq_flags);
231 	return err;
232 }
233 
234 /* Called from syscall or from eBPF program */
235 static void *queue_stack_map_lookup_elem(struct bpf_map *map, void *key)
236 {
237 	return NULL;
238 }
239 
240 /* Called from syscall or from eBPF program */
241 static long queue_stack_map_update_elem(struct bpf_map *map, void *key,
242 					void *value, u64 flags)
243 {
244 	return -EINVAL;
245 }
246 
247 /* Called from syscall or from eBPF program */
248 static long queue_stack_map_delete_elem(struct bpf_map *map, void *key)
249 {
250 	return -EINVAL;
251 }
252 
253 /* Called from syscall */
254 static int queue_stack_map_get_next_key(struct bpf_map *map, void *key,
255 					void *next_key)
256 {
257 	return -EINVAL;
258 }
259 
260 static u64 queue_stack_map_mem_usage(const struct bpf_map *map)
261 {
262 	u64 usage = sizeof(struct bpf_queue_stack);
263 
264 	usage += ((u64)map->max_entries + 1) * map->value_size;
265 	return usage;
266 }
267 
268 BTF_ID_LIST_SINGLE(queue_map_btf_ids, struct, bpf_queue_stack)
269 const struct bpf_map_ops queue_map_ops = {
270 	.map_meta_equal = bpf_map_meta_equal,
271 	.map_alloc_check = queue_stack_map_alloc_check,
272 	.map_alloc = queue_stack_map_alloc,
273 	.map_free = queue_stack_map_free,
274 	.map_lookup_elem = queue_stack_map_lookup_elem,
275 	.map_update_elem = queue_stack_map_update_elem,
276 	.map_delete_elem = queue_stack_map_delete_elem,
277 	.map_push_elem = queue_stack_map_push_elem,
278 	.map_pop_elem = queue_map_pop_elem,
279 	.map_peek_elem = queue_map_peek_elem,
280 	.map_get_next_key = queue_stack_map_get_next_key,
281 	.map_mem_usage = queue_stack_map_mem_usage,
282 	.map_btf_id = &queue_map_btf_ids[0],
283 };
284 
285 const struct bpf_map_ops stack_map_ops = {
286 	.map_meta_equal = bpf_map_meta_equal,
287 	.map_alloc_check = queue_stack_map_alloc_check,
288 	.map_alloc = queue_stack_map_alloc,
289 	.map_free = queue_stack_map_free,
290 	.map_lookup_elem = queue_stack_map_lookup_elem,
291 	.map_update_elem = queue_stack_map_update_elem,
292 	.map_delete_elem = queue_stack_map_delete_elem,
293 	.map_push_elem = queue_stack_map_push_elem,
294 	.map_pop_elem = stack_map_pop_elem,
295 	.map_peek_elem = stack_map_peek_elem,
296 	.map_get_next_key = queue_stack_map_get_next_key,
297 	.map_mem_usage = queue_stack_map_mem_usage,
298 	.map_btf_id = &queue_map_btf_ids[0],
299 };
300