From f1a2e44a3aeccb3ff18d3ccc0b0203e70b95bd92 Mon Sep 17 00:00:00 2001 From: Mauricio Vasquez B Date: Thu, 18 Oct 2018 15:16:25 +0200 Subject: bpf: add queue and stack maps Queue/stack maps implement a FIFO/LIFO data storage for ebpf programs. These maps support peek, pop and push operations that are exposed to eBPF programs through the new bpf_map[peek/pop/push] helpers. Those operations are exposed to userspace applications through the already existing syscalls in the following way: BPF_MAP_LOOKUP_ELEM -> peek BPF_MAP_LOOKUP_AND_DELETE_ELEM -> pop BPF_MAP_UPDATE_ELEM -> push Queue/stack maps are implemented using a buffer, tail and head indexes, hence BPF_F_NO_PREALLOC is not supported. As opposite to other maps, queue and stack do not use RCU for protecting maps values, the bpf_map[peek/pop] have a ARG_PTR_TO_UNINIT_MAP_VALUE argument that is a pointer to a memory zone where to save the value of a map. Basically the same as ARG_PTR_TO_UNINIT_MEM, but the size has not be passed as an extra argument. Our main motivation for implementing queue/stack maps was to keep track of a pool of elements, like network ports in a SNAT, however we forsee other use cases, like for exampling saving last N kernel events in a map and then analysing from userspace. Signed-off-by: Mauricio Vasquez B Acked-by: Song Liu Signed-off-by: Alexei Starovoitov --- kernel/bpf/queue_stack_maps.c | 288 ++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 288 insertions(+) create mode 100644 kernel/bpf/queue_stack_maps.c (limited to 'kernel/bpf/queue_stack_maps.c') diff --git a/kernel/bpf/queue_stack_maps.c b/kernel/bpf/queue_stack_maps.c new file mode 100644 index 000000000000..12a93fb37449 --- /dev/null +++ b/kernel/bpf/queue_stack_maps.c @@ -0,0 +1,288 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * queue_stack_maps.c: BPF queue and stack maps + * + * Copyright (c) 2018 Politecnico di Torino + */ +#include +#include +#include +#include "percpu_freelist.h" + +#define QUEUE_STACK_CREATE_FLAG_MASK \ + (BPF_F_NUMA_NODE | BPF_F_RDONLY | BPF_F_WRONLY) + + +struct bpf_queue_stack { + struct bpf_map map; + raw_spinlock_t lock; + u32 head, tail; + u32 size; /* max_entries + 1 */ + + char elements[0] __aligned(8); +}; + +static struct bpf_queue_stack *bpf_queue_stack(struct bpf_map *map) +{ + return container_of(map, struct bpf_queue_stack, map); +} + +static bool queue_stack_map_is_empty(struct bpf_queue_stack *qs) +{ + return qs->head == qs->tail; +} + +static bool queue_stack_map_is_full(struct bpf_queue_stack *qs) +{ + u32 head = qs->head + 1; + + if (unlikely(head >= qs->size)) + head = 0; + + return head == qs->tail; +} + +/* Called from syscall */ +static int queue_stack_map_alloc_check(union bpf_attr *attr) +{ + /* check sanity of attributes */ + if (attr->max_entries == 0 || attr->key_size != 0 || + attr->map_flags & ~QUEUE_STACK_CREATE_FLAG_MASK) + return -EINVAL; + + if (attr->value_size > KMALLOC_MAX_SIZE) + /* if value_size is bigger, the user space won't be able to + * access the elements. + */ + return -E2BIG; + + return 0; +} + +static struct bpf_map *queue_stack_map_alloc(union bpf_attr *attr) +{ + int ret, numa_node = bpf_map_attr_numa_node(attr); + struct bpf_queue_stack *qs; + u32 size, value_size; + u64 queue_size, cost; + + size = attr->max_entries + 1; + value_size = attr->value_size; + + queue_size = sizeof(*qs) + (u64) value_size * size; + + cost = queue_size; + if (cost >= U32_MAX - PAGE_SIZE) + return ERR_PTR(-E2BIG); + + cost = round_up(cost, PAGE_SIZE) >> PAGE_SHIFT; + + ret = bpf_map_precharge_memlock(cost); + if (ret < 0) + return ERR_PTR(ret); + + qs = bpf_map_area_alloc(queue_size, numa_node); + if (!qs) + return ERR_PTR(-ENOMEM); + + memset(qs, 0, sizeof(*qs)); + + bpf_map_init_from_attr(&qs->map, attr); + + qs->map.pages = cost; + qs->size = size; + + raw_spin_lock_init(&qs->lock); + + return &qs->map; +} + +/* Called when map->refcnt goes to zero, either from workqueue or from syscall */ +static void queue_stack_map_free(struct bpf_map *map) +{ + struct bpf_queue_stack *qs = bpf_queue_stack(map); + + /* at this point bpf_prog->aux->refcnt == 0 and this map->refcnt == 0, + * so the programs (can be more than one that used this map) were + * disconnected from events. Wait for outstanding critical sections in + * these programs to complete + */ + synchronize_rcu(); + + bpf_map_area_free(qs); +} + +static int __queue_map_get(struct bpf_map *map, void *value, bool delete) +{ + struct bpf_queue_stack *qs = bpf_queue_stack(map); + unsigned long flags; + int err = 0; + void *ptr; + + raw_spin_lock_irqsave(&qs->lock, flags); + + if (queue_stack_map_is_empty(qs)) { + err = -ENOENT; + goto out; + } + + ptr = &qs->elements[qs->tail * qs->map.value_size]; + memcpy(value, ptr, qs->map.value_size); + + if (delete) { + if (unlikely(++qs->tail >= qs->size)) + qs->tail = 0; + } + +out: + raw_spin_unlock_irqrestore(&qs->lock, flags); + return err; +} + + +static int __stack_map_get(struct bpf_map *map, void *value, bool delete) +{ + struct bpf_queue_stack *qs = bpf_queue_stack(map); + unsigned long flags; + int err = 0; + void *ptr; + u32 index; + + raw_spin_lock_irqsave(&qs->lock, flags); + + if (queue_stack_map_is_empty(qs)) { + err = -ENOENT; + goto out; + } + + index = qs->head - 1; + if (unlikely(index >= qs->size)) + index = qs->size - 1; + + ptr = &qs->elements[index * qs->map.value_size]; + memcpy(value, ptr, qs->map.value_size); + + if (delete) + qs->head = index; + +out: + raw_spin_unlock_irqrestore(&qs->lock, flags); + return err; +} + +/* Called from syscall or from eBPF program */ +static int queue_map_peek_elem(struct bpf_map *map, void *value) +{ + return __queue_map_get(map, value, false); +} + +/* Called from syscall or from eBPF program */ +static int stack_map_peek_elem(struct bpf_map *map, void *value) +{ + return __stack_map_get(map, value, false); +} + +/* Called from syscall or from eBPF program */ +static int queue_map_pop_elem(struct bpf_map *map, void *value) +{ + return __queue_map_get(map, value, true); +} + +/* Called from syscall or from eBPF program */ +static int stack_map_pop_elem(struct bpf_map *map, void *value) +{ + return __stack_map_get(map, value, true); +} + +/* Called from syscall or from eBPF program */ +static int queue_stack_map_push_elem(struct bpf_map *map, void *value, + u64 flags) +{ + struct bpf_queue_stack *qs = bpf_queue_stack(map); + unsigned long irq_flags; + int err = 0; + void *dst; + + /* BPF_EXIST is used to force making room for a new element in case the + * map is full + */ + bool replace = (flags & BPF_EXIST); + + /* Check supported flags for queue and stack maps */ + if (flags & BPF_NOEXIST || flags > BPF_EXIST) + return -EINVAL; + + raw_spin_lock_irqsave(&qs->lock, irq_flags); + + if (queue_stack_map_is_full(qs)) { + if (!replace) { + err = -E2BIG; + goto out; + } + /* advance tail pointer to overwrite oldest element */ + if (unlikely(++qs->tail >= qs->size)) + qs->tail = 0; + } + + dst = &qs->elements[qs->head * qs->map.value_size]; + memcpy(dst, value, qs->map.value_size); + + if (unlikely(++qs->head >= qs->size)) + qs->head = 0; + +out: + raw_spin_unlock_irqrestore(&qs->lock, irq_flags); + return err; +} + +/* Called from syscall or from eBPF program */ +static void *queue_stack_map_lookup_elem(struct bpf_map *map, void *key) +{ + return NULL; +} + +/* Called from syscall or from eBPF program */ +static int queue_stack_map_update_elem(struct bpf_map *map, void *key, + void *value, u64 flags) +{ + return -EINVAL; +} + +/* Called from syscall or from eBPF program */ +static int queue_stack_map_delete_elem(struct bpf_map *map, void *key) +{ + return -EINVAL; +} + +/* Called from syscall */ +static int queue_stack_map_get_next_key(struct bpf_map *map, void *key, + void *next_key) +{ + return -EINVAL; +} + +const struct bpf_map_ops queue_map_ops = { + .map_alloc_check = queue_stack_map_alloc_check, + .map_alloc = queue_stack_map_alloc, + .map_free = queue_stack_map_free, + .map_lookup_elem = queue_stack_map_lookup_elem, + .map_update_elem = queue_stack_map_update_elem, + .map_delete_elem = queue_stack_map_delete_elem, + .map_push_elem = queue_stack_map_push_elem, + .map_pop_elem = queue_map_pop_elem, + .map_peek_elem = queue_map_peek_elem, + .map_get_next_key = queue_stack_map_get_next_key, +}; + +const struct bpf_map_ops stack_map_ops = { + .map_alloc_check = queue_stack_map_alloc_check, + .map_alloc = queue_stack_map_alloc, + .map_free = queue_stack_map_free, + .map_lookup_elem = queue_stack_map_lookup_elem, + .map_update_elem = queue_stack_map_update_elem, + .map_delete_elem = queue_stack_map_delete_elem, + .map_push_elem = queue_stack_map_push_elem, + .map_pop_elem = stack_map_pop_elem, + .map_peek_elem = stack_map_peek_elem, + .map_get_next_key = queue_stack_map_get_next_key, +}; -- cgit v1.2.3 From d3f66e4116aff8dd0d5bd4067295b9ddb5e2c29c Mon Sep 17 00:00:00 2001 From: Daniel Borkmann Date: Wed, 24 Oct 2018 22:05:48 +0200 Subject: bpf: fix leaking uninitialized memory on pop/peek helpers Commit f1a2e44a3aec ("bpf: add queue and stack maps") added helpers with ARG_PTR_TO_UNINIT_MAP_VALUE. Meaning, the helper is supposed to fill the map value buffer with data instead of reading from it like in other helpers such as map update. However, given the buffer is allowed to be uninitialized (since we fill it in the helper anyway), it also means that the helper is obliged to wipe the memory in case of an error in order to not allow for leaking uninitialized memory. Given pop/peek is both handled inside __{stack,queue}_map_get(), lets wipe it there on error case, that is, empty stack/queue. Fixes: f1a2e44a3aec ("bpf: add queue and stack maps") Signed-off-by: Daniel Borkmann Acked-by: Alexei Starovoitov Cc: Mauricio Vasquez B Acked-by: Mauricio Vasquez B Signed-off-by: Alexei Starovoitov --- kernel/bpf/queue_stack_maps.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'kernel/bpf/queue_stack_maps.c') diff --git a/kernel/bpf/queue_stack_maps.c b/kernel/bpf/queue_stack_maps.c index 12a93fb37449..8bbd72d3a121 100644 --- a/kernel/bpf/queue_stack_maps.c +++ b/kernel/bpf/queue_stack_maps.c @@ -122,6 +122,7 @@ static int __queue_map_get(struct bpf_map *map, void *value, bool delete) raw_spin_lock_irqsave(&qs->lock, flags); if (queue_stack_map_is_empty(qs)) { + memset(value, 0, qs->map.value_size); err = -ENOENT; goto out; } @@ -151,6 +152,7 @@ static int __stack_map_get(struct bpf_map *map, void *value, bool delete) raw_spin_lock_irqsave(&qs->lock, flags); if (queue_stack_map_is_empty(qs)) { + memset(value, 0, qs->map.value_size); err = -ENOENT; goto out; } -- cgit v1.2.3