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/capability.h> 11 #include "percpu_freelist.h" 12 13 #define QUEUE_STACK_CREATE_FLAG_MASK \ 14 (BPF_F_NUMA_NODE | BPF_F_RDONLY | BPF_F_WRONLY) 15 16 17 struct bpf_queue_stack { 18 struct bpf_map map; 19 raw_spinlock_t lock; 20 u32 head, tail; 21 u32 size; /* max_entries + 1 */ 22 23 char elements[0] __aligned(8); 24 }; 25 26 static struct bpf_queue_stack *bpf_queue_stack(struct bpf_map *map) 27 { 28 return container_of(map, struct bpf_queue_stack, map); 29 } 30 31 static bool queue_stack_map_is_empty(struct bpf_queue_stack *qs) 32 { 33 return qs->head == qs->tail; 34 } 35 36 static bool queue_stack_map_is_full(struct bpf_queue_stack *qs) 37 { 38 u32 head = qs->head + 1; 39 40 if (unlikely(head >= qs->size)) 41 head = 0; 42 43 return head == qs->tail; 44 } 45 46 /* Called from syscall */ 47 static int queue_stack_map_alloc_check(union bpf_attr *attr) 48 { 49 if (!capable(CAP_SYS_ADMIN)) 50 return -EPERM; 51 52 /* check sanity of attributes */ 53 if (attr->max_entries == 0 || attr->key_size != 0 || 54 attr->value_size == 0 || 55 attr->map_flags & ~QUEUE_STACK_CREATE_FLAG_MASK) 56 return -EINVAL; 57 58 if (attr->value_size > KMALLOC_MAX_SIZE) 59 /* if value_size is bigger, the user space won't be able to 60 * access the elements. 61 */ 62 return -E2BIG; 63 64 return 0; 65 } 66 67 static struct bpf_map *queue_stack_map_alloc(union bpf_attr *attr) 68 { 69 int ret, numa_node = bpf_map_attr_numa_node(attr); 70 struct bpf_queue_stack *qs; 71 u64 size, queue_size, cost; 72 73 size = (u64) attr->max_entries + 1; 74 cost = queue_size = sizeof(*qs) + size * attr->value_size; 75 if (cost >= U32_MAX - PAGE_SIZE) 76 return ERR_PTR(-E2BIG); 77 78 cost = round_up(cost, PAGE_SIZE) >> PAGE_SHIFT; 79 80 ret = bpf_map_precharge_memlock(cost); 81 if (ret < 0) 82 return ERR_PTR(ret); 83 84 qs = bpf_map_area_alloc(queue_size, numa_node); 85 if (!qs) 86 return ERR_PTR(-ENOMEM); 87 88 memset(qs, 0, sizeof(*qs)); 89 90 bpf_map_init_from_attr(&qs->map, attr); 91 92 qs->map.pages = cost; 93 qs->size = size; 94 95 raw_spin_lock_init(&qs->lock); 96 97 return &qs->map; 98 } 99 100 /* Called when map->refcnt goes to zero, either from workqueue or from syscall */ 101 static void queue_stack_map_free(struct bpf_map *map) 102 { 103 struct bpf_queue_stack *qs = bpf_queue_stack(map); 104 105 /* at this point bpf_prog->aux->refcnt == 0 and this map->refcnt == 0, 106 * so the programs (can be more than one that used this map) were 107 * disconnected from events. Wait for outstanding critical sections in 108 * these programs to complete 109 */ 110 synchronize_rcu(); 111 112 bpf_map_area_free(qs); 113 } 114 115 static int __queue_map_get(struct bpf_map *map, void *value, bool delete) 116 { 117 struct bpf_queue_stack *qs = bpf_queue_stack(map); 118 unsigned long flags; 119 int err = 0; 120 void *ptr; 121 122 raw_spin_lock_irqsave(&qs->lock, flags); 123 124 if (queue_stack_map_is_empty(qs)) { 125 memset(value, 0, qs->map.value_size); 126 err = -ENOENT; 127 goto out; 128 } 129 130 ptr = &qs->elements[qs->tail * qs->map.value_size]; 131 memcpy(value, ptr, qs->map.value_size); 132 133 if (delete) { 134 if (unlikely(++qs->tail >= qs->size)) 135 qs->tail = 0; 136 } 137 138 out: 139 raw_spin_unlock_irqrestore(&qs->lock, flags); 140 return err; 141 } 142 143 144 static int __stack_map_get(struct bpf_map *map, void *value, bool delete) 145 { 146 struct bpf_queue_stack *qs = bpf_queue_stack(map); 147 unsigned long flags; 148 int err = 0; 149 void *ptr; 150 u32 index; 151 152 raw_spin_lock_irqsave(&qs->lock, flags); 153 154 if (queue_stack_map_is_empty(qs)) { 155 memset(value, 0, qs->map.value_size); 156 err = -ENOENT; 157 goto out; 158 } 159 160 index = qs->head - 1; 161 if (unlikely(index >= qs->size)) 162 index = qs->size - 1; 163 164 ptr = &qs->elements[index * qs->map.value_size]; 165 memcpy(value, ptr, qs->map.value_size); 166 167 if (delete) 168 qs->head = index; 169 170 out: 171 raw_spin_unlock_irqrestore(&qs->lock, flags); 172 return err; 173 } 174 175 /* Called from syscall or from eBPF program */ 176 static int queue_map_peek_elem(struct bpf_map *map, void *value) 177 { 178 return __queue_map_get(map, value, false); 179 } 180 181 /* Called from syscall or from eBPF program */ 182 static int stack_map_peek_elem(struct bpf_map *map, void *value) 183 { 184 return __stack_map_get(map, value, false); 185 } 186 187 /* Called from syscall or from eBPF program */ 188 static int queue_map_pop_elem(struct bpf_map *map, void *value) 189 { 190 return __queue_map_get(map, value, true); 191 } 192 193 /* Called from syscall or from eBPF program */ 194 static int stack_map_pop_elem(struct bpf_map *map, void *value) 195 { 196 return __stack_map_get(map, value, true); 197 } 198 199 /* Called from syscall or from eBPF program */ 200 static int queue_stack_map_push_elem(struct bpf_map *map, void *value, 201 u64 flags) 202 { 203 struct bpf_queue_stack *qs = bpf_queue_stack(map); 204 unsigned long irq_flags; 205 int err = 0; 206 void *dst; 207 208 /* BPF_EXIST is used to force making room for a new element in case the 209 * map is full 210 */ 211 bool replace = (flags & BPF_EXIST); 212 213 /* Check supported flags for queue and stack maps */ 214 if (flags & BPF_NOEXIST || flags > BPF_EXIST) 215 return -EINVAL; 216 217 raw_spin_lock_irqsave(&qs->lock, irq_flags); 218 219 if (queue_stack_map_is_full(qs)) { 220 if (!replace) { 221 err = -E2BIG; 222 goto out; 223 } 224 /* advance tail pointer to overwrite oldest element */ 225 if (unlikely(++qs->tail >= qs->size)) 226 qs->tail = 0; 227 } 228 229 dst = &qs->elements[qs->head * qs->map.value_size]; 230 memcpy(dst, value, qs->map.value_size); 231 232 if (unlikely(++qs->head >= qs->size)) 233 qs->head = 0; 234 235 out: 236 raw_spin_unlock_irqrestore(&qs->lock, irq_flags); 237 return err; 238 } 239 240 /* Called from syscall or from eBPF program */ 241 static void *queue_stack_map_lookup_elem(struct bpf_map *map, void *key) 242 { 243 return NULL; 244 } 245 246 /* Called from syscall or from eBPF program */ 247 static int queue_stack_map_update_elem(struct bpf_map *map, void *key, 248 void *value, u64 flags) 249 { 250 return -EINVAL; 251 } 252 253 /* Called from syscall or from eBPF program */ 254 static int queue_stack_map_delete_elem(struct bpf_map *map, void *key) 255 { 256 return -EINVAL; 257 } 258 259 /* Called from syscall */ 260 static int queue_stack_map_get_next_key(struct bpf_map *map, void *key, 261 void *next_key) 262 { 263 return -EINVAL; 264 } 265 266 const struct bpf_map_ops queue_map_ops = { 267 .map_alloc_check = queue_stack_map_alloc_check, 268 .map_alloc = queue_stack_map_alloc, 269 .map_free = queue_stack_map_free, 270 .map_lookup_elem = queue_stack_map_lookup_elem, 271 .map_update_elem = queue_stack_map_update_elem, 272 .map_delete_elem = queue_stack_map_delete_elem, 273 .map_push_elem = queue_stack_map_push_elem, 274 .map_pop_elem = queue_map_pop_elem, 275 .map_peek_elem = queue_map_peek_elem, 276 .map_get_next_key = queue_stack_map_get_next_key, 277 }; 278 279 const struct bpf_map_ops stack_map_ops = { 280 .map_alloc_check = queue_stack_map_alloc_check, 281 .map_alloc = queue_stack_map_alloc, 282 .map_free = queue_stack_map_free, 283 .map_lookup_elem = queue_stack_map_lookup_elem, 284 .map_update_elem = queue_stack_map_update_elem, 285 .map_delete_elem = queue_stack_map_delete_elem, 286 .map_push_elem = queue_stack_map_push_elem, 287 .map_pop_elem = stack_map_pop_elem, 288 .map_peek_elem = stack_map_peek_elem, 289 .map_get_next_key = queue_stack_map_get_next_key, 290 }; 291