xref: /linux/kernel/bpf/stackmap.c (revision bf070bb0e6c62ba3075db0a666763ba52c677102)
1 /* Copyright (c) 2016 Facebook
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 #include <linux/bpf.h>
8 #include <linux/jhash.h>
9 #include <linux/filter.h>
10 #include <linux/stacktrace.h>
11 #include <linux/perf_event.h>
12 #include "percpu_freelist.h"
13 
14 #define STACK_CREATE_FLAG_MASK \
15 	(BPF_F_NUMA_NODE | BPF_F_RDONLY | BPF_F_WRONLY)
16 
17 struct stack_map_bucket {
18 	struct pcpu_freelist_node fnode;
19 	u32 hash;
20 	u32 nr;
21 	u64 ip[];
22 };
23 
24 struct bpf_stack_map {
25 	struct bpf_map map;
26 	void *elems;
27 	struct pcpu_freelist freelist;
28 	u32 n_buckets;
29 	struct stack_map_bucket *buckets[];
30 };
31 
32 static int prealloc_elems_and_freelist(struct bpf_stack_map *smap)
33 {
34 	u32 elem_size = sizeof(struct stack_map_bucket) + smap->map.value_size;
35 	int err;
36 
37 	smap->elems = bpf_map_area_alloc(elem_size * smap->map.max_entries,
38 					 smap->map.numa_node);
39 	if (!smap->elems)
40 		return -ENOMEM;
41 
42 	err = pcpu_freelist_init(&smap->freelist);
43 	if (err)
44 		goto free_elems;
45 
46 	pcpu_freelist_populate(&smap->freelist, smap->elems, elem_size,
47 			       smap->map.max_entries);
48 	return 0;
49 
50 free_elems:
51 	bpf_map_area_free(smap->elems);
52 	return err;
53 }
54 
55 /* Called from syscall */
56 static struct bpf_map *stack_map_alloc(union bpf_attr *attr)
57 {
58 	u32 value_size = attr->value_size;
59 	struct bpf_stack_map *smap;
60 	u64 cost, n_buckets;
61 	int err;
62 
63 	if (!capable(CAP_SYS_ADMIN))
64 		return ERR_PTR(-EPERM);
65 
66 	if (attr->map_flags & ~STACK_CREATE_FLAG_MASK)
67 		return ERR_PTR(-EINVAL);
68 
69 	/* check sanity of attributes */
70 	if (attr->max_entries == 0 || attr->key_size != 4 ||
71 	    value_size < 8 || value_size % 8 ||
72 	    value_size / 8 > sysctl_perf_event_max_stack)
73 		return ERR_PTR(-EINVAL);
74 
75 	/* hash table size must be power of 2 */
76 	n_buckets = roundup_pow_of_two(attr->max_entries);
77 
78 	cost = n_buckets * sizeof(struct stack_map_bucket *) + sizeof(*smap);
79 	if (cost >= U32_MAX - PAGE_SIZE)
80 		return ERR_PTR(-E2BIG);
81 
82 	smap = bpf_map_area_alloc(cost, bpf_map_attr_numa_node(attr));
83 	if (!smap)
84 		return ERR_PTR(-ENOMEM);
85 
86 	err = -E2BIG;
87 	cost += n_buckets * (value_size + sizeof(struct stack_map_bucket));
88 	if (cost >= U32_MAX - PAGE_SIZE)
89 		goto free_smap;
90 
91 	smap->map.map_type = attr->map_type;
92 	smap->map.key_size = attr->key_size;
93 	smap->map.value_size = value_size;
94 	smap->map.max_entries = attr->max_entries;
95 	smap->map.map_flags = attr->map_flags;
96 	smap->n_buckets = n_buckets;
97 	smap->map.pages = round_up(cost, PAGE_SIZE) >> PAGE_SHIFT;
98 	smap->map.numa_node = bpf_map_attr_numa_node(attr);
99 
100 	err = bpf_map_precharge_memlock(smap->map.pages);
101 	if (err)
102 		goto free_smap;
103 
104 	err = get_callchain_buffers(sysctl_perf_event_max_stack);
105 	if (err)
106 		goto free_smap;
107 
108 	err = prealloc_elems_and_freelist(smap);
109 	if (err)
110 		goto put_buffers;
111 
112 	return &smap->map;
113 
114 put_buffers:
115 	put_callchain_buffers();
116 free_smap:
117 	bpf_map_area_free(smap);
118 	return ERR_PTR(err);
119 }
120 
121 BPF_CALL_3(bpf_get_stackid, struct pt_regs *, regs, struct bpf_map *, map,
122 	   u64, flags)
123 {
124 	struct bpf_stack_map *smap = container_of(map, struct bpf_stack_map, map);
125 	struct perf_callchain_entry *trace;
126 	struct stack_map_bucket *bucket, *new_bucket, *old_bucket;
127 	u32 max_depth = map->value_size / 8;
128 	/* stack_map_alloc() checks that max_depth <= sysctl_perf_event_max_stack */
129 	u32 init_nr = sysctl_perf_event_max_stack - max_depth;
130 	u32 skip = flags & BPF_F_SKIP_FIELD_MASK;
131 	u32 hash, id, trace_nr, trace_len;
132 	bool user = flags & BPF_F_USER_STACK;
133 	bool kernel = !user;
134 	u64 *ips;
135 
136 	if (unlikely(flags & ~(BPF_F_SKIP_FIELD_MASK | BPF_F_USER_STACK |
137 			       BPF_F_FAST_STACK_CMP | BPF_F_REUSE_STACKID)))
138 		return -EINVAL;
139 
140 	trace = get_perf_callchain(regs, init_nr, kernel, user,
141 				   sysctl_perf_event_max_stack, false, false);
142 
143 	if (unlikely(!trace))
144 		/* couldn't fetch the stack trace */
145 		return -EFAULT;
146 
147 	/* get_perf_callchain() guarantees that trace->nr >= init_nr
148 	 * and trace-nr <= sysctl_perf_event_max_stack, so trace_nr <= max_depth
149 	 */
150 	trace_nr = trace->nr - init_nr;
151 
152 	if (trace_nr <= skip)
153 		/* skipping more than usable stack trace */
154 		return -EFAULT;
155 
156 	trace_nr -= skip;
157 	trace_len = trace_nr * sizeof(u64);
158 	ips = trace->ip + skip + init_nr;
159 	hash = jhash2((u32 *)ips, trace_len / sizeof(u32), 0);
160 	id = hash & (smap->n_buckets - 1);
161 	bucket = READ_ONCE(smap->buckets[id]);
162 
163 	if (bucket && bucket->hash == hash) {
164 		if (flags & BPF_F_FAST_STACK_CMP)
165 			return id;
166 		if (bucket->nr == trace_nr &&
167 		    memcmp(bucket->ip, ips, trace_len) == 0)
168 			return id;
169 	}
170 
171 	/* this call stack is not in the map, try to add it */
172 	if (bucket && !(flags & BPF_F_REUSE_STACKID))
173 		return -EEXIST;
174 
175 	new_bucket = (struct stack_map_bucket *)
176 		pcpu_freelist_pop(&smap->freelist);
177 	if (unlikely(!new_bucket))
178 		return -ENOMEM;
179 
180 	memcpy(new_bucket->ip, ips, trace_len);
181 	new_bucket->hash = hash;
182 	new_bucket->nr = trace_nr;
183 
184 	old_bucket = xchg(&smap->buckets[id], new_bucket);
185 	if (old_bucket)
186 		pcpu_freelist_push(&smap->freelist, &old_bucket->fnode);
187 	return id;
188 }
189 
190 const struct bpf_func_proto bpf_get_stackid_proto = {
191 	.func		= bpf_get_stackid,
192 	.gpl_only	= true,
193 	.ret_type	= RET_INTEGER,
194 	.arg1_type	= ARG_PTR_TO_CTX,
195 	.arg2_type	= ARG_CONST_MAP_PTR,
196 	.arg3_type	= ARG_ANYTHING,
197 };
198 
199 /* Called from eBPF program */
200 static void *stack_map_lookup_elem(struct bpf_map *map, void *key)
201 {
202 	return NULL;
203 }
204 
205 /* Called from syscall */
206 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value)
207 {
208 	struct bpf_stack_map *smap = container_of(map, struct bpf_stack_map, map);
209 	struct stack_map_bucket *bucket, *old_bucket;
210 	u32 id = *(u32 *)key, trace_len;
211 
212 	if (unlikely(id >= smap->n_buckets))
213 		return -ENOENT;
214 
215 	bucket = xchg(&smap->buckets[id], NULL);
216 	if (!bucket)
217 		return -ENOENT;
218 
219 	trace_len = bucket->nr * sizeof(u64);
220 	memcpy(value, bucket->ip, trace_len);
221 	memset(value + trace_len, 0, map->value_size - trace_len);
222 
223 	old_bucket = xchg(&smap->buckets[id], bucket);
224 	if (old_bucket)
225 		pcpu_freelist_push(&smap->freelist, &old_bucket->fnode);
226 	return 0;
227 }
228 
229 static int stack_map_get_next_key(struct bpf_map *map, void *key, void *next_key)
230 {
231 	return -EINVAL;
232 }
233 
234 static int stack_map_update_elem(struct bpf_map *map, void *key, void *value,
235 				 u64 map_flags)
236 {
237 	return -EINVAL;
238 }
239 
240 /* Called from syscall or from eBPF program */
241 static int stack_map_delete_elem(struct bpf_map *map, void *key)
242 {
243 	struct bpf_stack_map *smap = container_of(map, struct bpf_stack_map, map);
244 	struct stack_map_bucket *old_bucket;
245 	u32 id = *(u32 *)key;
246 
247 	if (unlikely(id >= smap->n_buckets))
248 		return -E2BIG;
249 
250 	old_bucket = xchg(&smap->buckets[id], NULL);
251 	if (old_bucket) {
252 		pcpu_freelist_push(&smap->freelist, &old_bucket->fnode);
253 		return 0;
254 	} else {
255 		return -ENOENT;
256 	}
257 }
258 
259 /* Called when map->refcnt goes to zero, either from workqueue or from syscall */
260 static void stack_map_free(struct bpf_map *map)
261 {
262 	struct bpf_stack_map *smap = container_of(map, struct bpf_stack_map, map);
263 
264 	/* wait for bpf programs to complete before freeing stack map */
265 	synchronize_rcu();
266 
267 	bpf_map_area_free(smap->elems);
268 	pcpu_freelist_destroy(&smap->freelist);
269 	bpf_map_area_free(smap);
270 	put_callchain_buffers();
271 }
272 
273 const struct bpf_map_ops stack_map_ops = {
274 	.map_alloc = stack_map_alloc,
275 	.map_free = stack_map_free,
276 	.map_get_next_key = stack_map_get_next_key,
277 	.map_lookup_elem = stack_map_lookup_elem,
278 	.map_update_elem = stack_map_update_elem,
279 	.map_delete_elem = stack_map_delete_elem,
280 };
281