From 1110f3a9bcf394c06b81a98206aee9b6860653c8 Mon Sep 17 00:00:00 2001 From: Jakub Kicinski Date: Thu, 11 Jan 2018 20:29:03 -0800 Subject: bpf: add map_alloc_check callback .map_alloc callbacks contain a number of checks validating user- -provided map attributes against constraints of a particular map type. For offloaded maps we will need to check map attributes without actually allocating any memory on the host. Add a new callback for validating attributes before any memory is allocated. This callback can be selectively implemented by map types for sharing code with offloads, or simply to separate the logical steps of validation and allocation. Signed-off-by: Jakub Kicinski Reviewed-by: Quentin Monnet Signed-off-by: Daniel Borkmann --- kernel/bpf/syscall.c | 17 +++++++++++++---- 1 file changed, 13 insertions(+), 4 deletions(-) (limited to 'kernel') diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c index 2bac0dc8baba..c0ac03a04880 100644 --- a/kernel/bpf/syscall.c +++ b/kernel/bpf/syscall.c @@ -96,16 +96,25 @@ static int check_uarg_tail_zero(void __user *uaddr, static struct bpf_map *find_and_alloc_map(union bpf_attr *attr) { + const struct bpf_map_ops *ops; struct bpf_map *map; + int err; - if (attr->map_type >= ARRAY_SIZE(bpf_map_types) || - !bpf_map_types[attr->map_type]) + if (attr->map_type >= ARRAY_SIZE(bpf_map_types)) + return ERR_PTR(-EINVAL); + ops = bpf_map_types[attr->map_type]; + if (!ops) return ERR_PTR(-EINVAL); - map = bpf_map_types[attr->map_type]->map_alloc(attr); + if (ops->map_alloc_check) { + err = ops->map_alloc_check(attr); + if (err) + return ERR_PTR(err); + } + map = ops->map_alloc(attr); if (IS_ERR(map)) return map; - map->ops = bpf_map_types[attr->map_type]; + map->ops = ops; map->map_type = attr->map_type; return map; } -- cgit v1.2.3 From daffc5a2e6f4bf4f99b00e183117920e321b6763 Mon Sep 17 00:00:00 2001 From: Jakub Kicinski Date: Thu, 11 Jan 2018 20:29:04 -0800 Subject: bpf: hashtab: move attribute validation before allocation Number of attribute checks are currently performed after hashtab is already allocated. Move them to be able to split them out to the check function later on. Checks have to now be performed on the attr union directly instead of the members of bpf_map, since bpf_map will be allocated later. No functional changes. Signed-off-by: Jakub Kicinski Reviewed-by: Quentin Monnet Signed-off-by: Daniel Borkmann --- kernel/bpf/hashtab.c | 47 +++++++++++++++++++++++------------------------ 1 file changed, 23 insertions(+), 24 deletions(-) (limited to 'kernel') diff --git a/kernel/bpf/hashtab.c b/kernel/bpf/hashtab.c index 3905d4bc5b80..b80f42adf068 100644 --- a/kernel/bpf/hashtab.c +++ b/kernel/bpf/hashtab.c @@ -269,6 +269,28 @@ static struct bpf_map *htab_map_alloc(union bpf_attr *attr) if (numa_node != NUMA_NO_NODE && (percpu || percpu_lru)) return ERR_PTR(-EINVAL); + /* check sanity of attributes. + * value_size == 0 may be allowed in the future to use map as a set + */ + if (attr->max_entries == 0 || attr->key_size == 0 || + attr->value_size == 0) + return ERR_PTR(-EINVAL); + + if (attr->key_size > MAX_BPF_STACK) + /* eBPF programs initialize keys on stack, so they cannot be + * larger than max stack size + */ + return ERR_PTR(-E2BIG); + + if (attr->value_size >= KMALLOC_MAX_SIZE - + MAX_BPF_STACK - sizeof(struct htab_elem)) + /* if value_size is bigger, the user space won't be able to + * access the elements via bpf syscall. This check also makes + * sure that the elem_size doesn't overflow and it's + * kmalloc-able later in htab_map_update_elem() + */ + return ERR_PTR(-E2BIG); + htab = kzalloc(sizeof(*htab), GFP_USER); if (!htab) return ERR_PTR(-ENOMEM); @@ -281,14 +303,6 @@ static struct bpf_map *htab_map_alloc(union bpf_attr *attr) htab->map.map_flags = attr->map_flags; htab->map.numa_node = numa_node; - /* check sanity of attributes. - * value_size == 0 may be allowed in the future to use map as a set - */ - err = -EINVAL; - if (htab->map.max_entries == 0 || htab->map.key_size == 0 || - htab->map.value_size == 0) - goto free_htab; - if (percpu_lru) { /* ensure each CPU's lru list has >=1 elements. * since we are at it, make each lru list has the same @@ -304,22 +318,6 @@ static struct bpf_map *htab_map_alloc(union bpf_attr *attr) /* hash table size must be power of 2 */ htab->n_buckets = roundup_pow_of_two(htab->map.max_entries); - err = -E2BIG; - if (htab->map.key_size > MAX_BPF_STACK) - /* eBPF programs initialize keys on stack, so they cannot be - * larger than max stack size - */ - goto free_htab; - - if (htab->map.value_size >= KMALLOC_MAX_SIZE - - MAX_BPF_STACK - sizeof(struct htab_elem)) - /* if value_size is bigger, the user space won't be able to - * access the elements via bpf syscall. This check also makes - * sure that the elem_size doesn't overflow and it's - * kmalloc-able later in htab_map_update_elem() - */ - goto free_htab; - htab->elem_size = sizeof(struct htab_elem) + round_up(htab->map.key_size, 8); if (percpu) @@ -327,6 +325,7 @@ static struct bpf_map *htab_map_alloc(union bpf_attr *attr) else htab->elem_size += round_up(htab->map.value_size, 8); + err = -E2BIG; /* prevent zero size kmalloc and check for u32 overflow */ if (htab->n_buckets == 0 || htab->n_buckets > U32_MAX / sizeof(struct bucket)) -- cgit v1.2.3 From 9328e0d1bc09e96bd7dc85374f5c2a1e0e04e539 Mon Sep 17 00:00:00 2001 From: Jakub Kicinski Date: Thu, 11 Jan 2018 20:29:05 -0800 Subject: bpf: hashtab: move checks out of alloc function Use the new callback to perform allocation checks for hash maps. Signed-off-by: Jakub Kicinski Reviewed-by: Quentin Monnet Signed-off-by: Daniel Borkmann --- kernel/bpf/hashtab.c | 55 +++++++++++++++++++++++++++++++++++++--------------- 1 file changed, 39 insertions(+), 16 deletions(-) (limited to 'kernel') diff --git a/kernel/bpf/hashtab.c b/kernel/bpf/hashtab.c index b80f42adf068..7fd6519444d3 100644 --- a/kernel/bpf/hashtab.c +++ b/kernel/bpf/hashtab.c @@ -227,7 +227,7 @@ static int alloc_extra_elems(struct bpf_htab *htab) } /* Called from syscall */ -static struct bpf_map *htab_map_alloc(union bpf_attr *attr) +static int htab_map_alloc_check(union bpf_attr *attr) { bool percpu = (attr->map_type == BPF_MAP_TYPE_PERCPU_HASH || attr->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH); @@ -241,9 +241,6 @@ static struct bpf_map *htab_map_alloc(union bpf_attr *attr) bool percpu_lru = (attr->map_flags & BPF_F_NO_COMMON_LRU); bool prealloc = !(attr->map_flags & BPF_F_NO_PREALLOC); int numa_node = bpf_map_attr_numa_node(attr); - struct bpf_htab *htab; - int err, i; - u64 cost; BUILD_BUG_ON(offsetof(struct htab_elem, htab) != offsetof(struct htab_elem, hash_node.pprev)); @@ -254,33 +251,33 @@ static struct bpf_map *htab_map_alloc(union bpf_attr *attr) /* LRU implementation is much complicated than other * maps. Hence, limit to CAP_SYS_ADMIN for now. */ - return ERR_PTR(-EPERM); + return -EPERM; if (attr->map_flags & ~HTAB_CREATE_FLAG_MASK) /* reserved bits should not be used */ - return ERR_PTR(-EINVAL); + return -EINVAL; if (!lru && percpu_lru) - return ERR_PTR(-EINVAL); + return -EINVAL; if (lru && !prealloc) - return ERR_PTR(-ENOTSUPP); + return -ENOTSUPP; if (numa_node != NUMA_NO_NODE && (percpu || percpu_lru)) - return ERR_PTR(-EINVAL); + return -EINVAL; /* check sanity of attributes. * value_size == 0 may be allowed in the future to use map as a set */ if (attr->max_entries == 0 || attr->key_size == 0 || attr->value_size == 0) - return ERR_PTR(-EINVAL); + return -EINVAL; if (attr->key_size > MAX_BPF_STACK) /* eBPF programs initialize keys on stack, so they cannot be * larger than max stack size */ - return ERR_PTR(-E2BIG); + return -E2BIG; if (attr->value_size >= KMALLOC_MAX_SIZE - MAX_BPF_STACK - sizeof(struct htab_elem)) @@ -289,7 +286,28 @@ static struct bpf_map *htab_map_alloc(union bpf_attr *attr) * sure that the elem_size doesn't overflow and it's * kmalloc-able later in htab_map_update_elem() */ - return ERR_PTR(-E2BIG); + return -E2BIG; + + return 0; +} + +static struct bpf_map *htab_map_alloc(union bpf_attr *attr) +{ + bool percpu = (attr->map_type == BPF_MAP_TYPE_PERCPU_HASH || + attr->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH); + bool lru = (attr->map_type == BPF_MAP_TYPE_LRU_HASH || + attr->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH); + /* percpu_lru means each cpu has its own LRU list. + * it is different from BPF_MAP_TYPE_PERCPU_HASH where + * the map's value itself is percpu. percpu_lru has + * nothing to do with the map's value. + */ + bool percpu_lru = (attr->map_flags & BPF_F_NO_COMMON_LRU); + bool prealloc = !(attr->map_flags & BPF_F_NO_PREALLOC); + int numa_node = bpf_map_attr_numa_node(attr); + struct bpf_htab *htab; + int err, i; + u64 cost; htab = kzalloc(sizeof(*htab), GFP_USER); if (!htab) @@ -1142,6 +1160,7 @@ static void htab_map_free(struct bpf_map *map) } const struct bpf_map_ops htab_map_ops = { + .map_alloc_check = htab_map_alloc_check, .map_alloc = htab_map_alloc, .map_free = htab_map_free, .map_get_next_key = htab_map_get_next_key, @@ -1152,6 +1171,7 @@ const struct bpf_map_ops htab_map_ops = { }; const struct bpf_map_ops htab_lru_map_ops = { + .map_alloc_check = htab_map_alloc_check, .map_alloc = htab_map_alloc, .map_free = htab_map_free, .map_get_next_key = htab_map_get_next_key, @@ -1235,6 +1255,7 @@ int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value, } const struct bpf_map_ops htab_percpu_map_ops = { + .map_alloc_check = htab_map_alloc_check, .map_alloc = htab_map_alloc, .map_free = htab_map_free, .map_get_next_key = htab_map_get_next_key, @@ -1244,6 +1265,7 @@ const struct bpf_map_ops htab_percpu_map_ops = { }; const struct bpf_map_ops htab_lru_percpu_map_ops = { + .map_alloc_check = htab_map_alloc_check, .map_alloc = htab_map_alloc, .map_free = htab_map_free, .map_get_next_key = htab_map_get_next_key, @@ -1252,11 +1274,11 @@ const struct bpf_map_ops htab_lru_percpu_map_ops = { .map_delete_elem = htab_lru_map_delete_elem, }; -static struct bpf_map *fd_htab_map_alloc(union bpf_attr *attr) +static int fd_htab_map_alloc_check(union bpf_attr *attr) { if (attr->value_size != sizeof(u32)) - return ERR_PTR(-EINVAL); - return htab_map_alloc(attr); + return -EINVAL; + return htab_map_alloc_check(attr); } static void fd_htab_map_free(struct bpf_map *map) @@ -1327,7 +1349,7 @@ static struct bpf_map *htab_of_map_alloc(union bpf_attr *attr) if (IS_ERR(inner_map_meta)) return inner_map_meta; - map = fd_htab_map_alloc(attr); + map = htab_map_alloc(attr); if (IS_ERR(map)) { bpf_map_meta_free(inner_map_meta); return map; @@ -1371,6 +1393,7 @@ static void htab_of_map_free(struct bpf_map *map) } const struct bpf_map_ops htab_of_maps_map_ops = { + .map_alloc_check = fd_htab_map_alloc_check, .map_alloc = htab_of_map_alloc, .map_free = htab_of_map_free, .map_get_next_key = htab_map_get_next_key, -- cgit v1.2.3 From bd475643d74e8ed78bfd36d941053b0e45974e8e Mon Sep 17 00:00:00 2001 From: Jakub Kicinski Date: Thu, 11 Jan 2018 20:29:06 -0800 Subject: bpf: add helper for copying attrs to struct bpf_map All map types reimplement the field-by-field copy of union bpf_attr members into struct bpf_map. Add a helper to perform this operation. Signed-off-by: Jakub Kicinski Reviewed-by: Quentin Monnet Acked-by: Alexei Starovoitov Signed-off-by: Daniel Borkmann --- kernel/bpf/cpumap.c | 8 +------- kernel/bpf/devmap.c | 8 +------- kernel/bpf/hashtab.c | 9 +-------- kernel/bpf/lpm_trie.c | 7 +------ kernel/bpf/sockmap.c | 8 +------- kernel/bpf/stackmap.c | 6 +----- kernel/bpf/syscall.c | 10 ++++++++++ 7 files changed, 16 insertions(+), 40 deletions(-) (limited to 'kernel') diff --git a/kernel/bpf/cpumap.c b/kernel/bpf/cpumap.c index ce5b669003b2..192151ec9d12 100644 --- a/kernel/bpf/cpumap.c +++ b/kernel/bpf/cpumap.c @@ -94,13 +94,7 @@ static struct bpf_map *cpu_map_alloc(union bpf_attr *attr) if (!cmap) return ERR_PTR(-ENOMEM); - /* mandatory map attributes */ - cmap->map.map_type = attr->map_type; - cmap->map.key_size = attr->key_size; - cmap->map.value_size = attr->value_size; - cmap->map.max_entries = attr->max_entries; - cmap->map.map_flags = attr->map_flags; - cmap->map.numa_node = bpf_map_attr_numa_node(attr); + bpf_map_init_from_attr(&cmap->map, attr); /* Pre-limit array size based on NR_CPUS, not final CPU check */ if (cmap->map.max_entries > NR_CPUS) { diff --git a/kernel/bpf/devmap.c b/kernel/bpf/devmap.c index ebdef54bf7df..565f9ece9115 100644 --- a/kernel/bpf/devmap.c +++ b/kernel/bpf/devmap.c @@ -93,13 +93,7 @@ static struct bpf_map *dev_map_alloc(union bpf_attr *attr) if (!dtab) return ERR_PTR(-ENOMEM); - /* mandatory map attributes */ - dtab->map.map_type = attr->map_type; - dtab->map.key_size = attr->key_size; - dtab->map.value_size = attr->value_size; - dtab->map.max_entries = attr->max_entries; - dtab->map.map_flags = attr->map_flags; - dtab->map.numa_node = bpf_map_attr_numa_node(attr); + bpf_map_init_from_attr(&dtab->map, attr); /* make sure page count doesn't overflow */ cost = (u64) dtab->map.max_entries * sizeof(struct bpf_dtab_netdev *); diff --git a/kernel/bpf/hashtab.c b/kernel/bpf/hashtab.c index 7fd6519444d3..b76828f23b49 100644 --- a/kernel/bpf/hashtab.c +++ b/kernel/bpf/hashtab.c @@ -304,7 +304,6 @@ static struct bpf_map *htab_map_alloc(union bpf_attr *attr) */ bool percpu_lru = (attr->map_flags & BPF_F_NO_COMMON_LRU); bool prealloc = !(attr->map_flags & BPF_F_NO_PREALLOC); - int numa_node = bpf_map_attr_numa_node(attr); struct bpf_htab *htab; int err, i; u64 cost; @@ -313,13 +312,7 @@ static struct bpf_map *htab_map_alloc(union bpf_attr *attr) if (!htab) return ERR_PTR(-ENOMEM); - /* mandatory map attributes */ - htab->map.map_type = attr->map_type; - htab->map.key_size = attr->key_size; - htab->map.value_size = attr->value_size; - htab->map.max_entries = attr->max_entries; - htab->map.map_flags = attr->map_flags; - htab->map.numa_node = numa_node; + bpf_map_init_from_attr(&htab->map, attr); if (percpu_lru) { /* ensure each CPU's lru list has >=1 elements. diff --git a/kernel/bpf/lpm_trie.c b/kernel/bpf/lpm_trie.c index 885e45479680..584e02227671 100644 --- a/kernel/bpf/lpm_trie.c +++ b/kernel/bpf/lpm_trie.c @@ -522,12 +522,7 @@ static struct bpf_map *trie_alloc(union bpf_attr *attr) return ERR_PTR(-ENOMEM); /* copy mandatory map attributes */ - trie->map.map_type = attr->map_type; - trie->map.key_size = attr->key_size; - trie->map.value_size = attr->value_size; - trie->map.max_entries = attr->max_entries; - trie->map.map_flags = attr->map_flags; - trie->map.numa_node = bpf_map_attr_numa_node(attr); + bpf_map_init_from_attr(&trie->map, attr); trie->data_size = attr->key_size - offsetof(struct bpf_lpm_trie_key, data); trie->max_prefixlen = trie->data_size * 8; diff --git a/kernel/bpf/sockmap.c b/kernel/bpf/sockmap.c index 079968680bc3..0314d1783d77 100644 --- a/kernel/bpf/sockmap.c +++ b/kernel/bpf/sockmap.c @@ -513,13 +513,7 @@ static struct bpf_map *sock_map_alloc(union bpf_attr *attr) if (!stab) return ERR_PTR(-ENOMEM); - /* mandatory map attributes */ - stab->map.map_type = attr->map_type; - stab->map.key_size = attr->key_size; - stab->map.value_size = attr->value_size; - stab->map.max_entries = attr->max_entries; - stab->map.map_flags = attr->map_flags; - stab->map.numa_node = bpf_map_attr_numa_node(attr); + bpf_map_init_from_attr(&stab->map, attr); /* make sure page count doesn't overflow */ cost = (u64) stab->map.max_entries * sizeof(struct sock *); diff --git a/kernel/bpf/stackmap.c b/kernel/bpf/stackmap.c index 6c63c2222ea8..b0ecf43f5894 100644 --- a/kernel/bpf/stackmap.c +++ b/kernel/bpf/stackmap.c @@ -88,14 +88,10 @@ static struct bpf_map *stack_map_alloc(union bpf_attr *attr) if (cost >= U32_MAX - PAGE_SIZE) goto free_smap; - smap->map.map_type = attr->map_type; - smap->map.key_size = attr->key_size; + bpf_map_init_from_attr(&smap->map, attr); smap->map.value_size = value_size; - smap->map.max_entries = attr->max_entries; - smap->map.map_flags = attr->map_flags; smap->n_buckets = n_buckets; smap->map.pages = round_up(cost, PAGE_SIZE) >> PAGE_SHIFT; - smap->map.numa_node = bpf_map_attr_numa_node(attr); err = bpf_map_precharge_memlock(smap->map.pages); if (err) diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c index c0ac03a04880..a3f726bb42ea 100644 --- a/kernel/bpf/syscall.c +++ b/kernel/bpf/syscall.c @@ -143,6 +143,16 @@ void bpf_map_area_free(void *area) kvfree(area); } +void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr) +{ + map->map_type = attr->map_type; + map->key_size = attr->key_size; + map->value_size = attr->value_size; + map->max_entries = attr->max_entries; + map->map_flags = attr->map_flags; + map->numa_node = bpf_map_attr_numa_node(attr); +} + int bpf_map_precharge_memlock(u32 pages) { struct user_struct *user = get_current_user(); -- cgit v1.2.3 From 0a9c1991f285f829fd786fa2a9c824c2a3f87bc6 Mon Sep 17 00:00:00 2001 From: Jakub Kicinski Date: Thu, 11 Jan 2018 20:29:07 -0800 Subject: bpf: rename bpf_dev_offload -> bpf_prog_offload With map offload coming, we need to call program offload structure something less ambiguous. Pure rename, no functional changes. Signed-off-by: Jakub Kicinski Reviewed-by: Quentin Monnet Signed-off-by: Daniel Borkmann --- kernel/bpf/offload.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'kernel') diff --git a/kernel/bpf/offload.c b/kernel/bpf/offload.c index 040d4e0edf3f..001ddfde7874 100644 --- a/kernel/bpf/offload.c +++ b/kernel/bpf/offload.c @@ -32,7 +32,7 @@ static LIST_HEAD(bpf_prog_offload_devs); int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr) { - struct bpf_dev_offload *offload; + struct bpf_prog_offload *offload; if (attr->prog_type != BPF_PROG_TYPE_SCHED_CLS && attr->prog_type != BPF_PROG_TYPE_XDP) @@ -72,7 +72,7 @@ err_free: static int __bpf_offload_ndo(struct bpf_prog *prog, enum bpf_netdev_command cmd, struct netdev_bpf *data) { - struct bpf_dev_offload *offload = prog->aux->offload; + struct bpf_prog_offload *offload = prog->aux->offload; struct net_device *netdev; ASSERT_RTNL(); @@ -110,7 +110,7 @@ exit_unlock: int bpf_prog_offload_verify_insn(struct bpf_verifier_env *env, int insn_idx, int prev_insn_idx) { - struct bpf_dev_offload *offload; + struct bpf_prog_offload *offload; int ret = -ENODEV; down_read(&bpf_devs_lock); @@ -124,7 +124,7 @@ int bpf_prog_offload_verify_insn(struct bpf_verifier_env *env, static void __bpf_prog_offload_destroy(struct bpf_prog *prog) { - struct bpf_dev_offload *offload = prog->aux->offload; + struct bpf_prog_offload *offload = prog->aux->offload; struct netdev_bpf data = {}; data.offload.prog = prog; @@ -242,7 +242,7 @@ static int bpf_offload_notification(struct notifier_block *notifier, ulong event, void *ptr) { struct net_device *netdev = netdev_notifier_info_to_dev(ptr); - struct bpf_dev_offload *offload, *tmp; + struct bpf_prog_offload *offload, *tmp; ASSERT_RTNL(); -- cgit v1.2.3 From 5bc2d55c6a69ef9efd11740359974b08ea11f1d7 Mon Sep 17 00:00:00 2001 From: Jakub Kicinski Date: Thu, 11 Jan 2018 20:29:08 -0800 Subject: bpf: offload: factor out netdev checking at allocation time Add a helper to check if netdev could be found and whether it has .ndo_bpf callback. There is no need to check the callback every time it's invoked, ndos can't reasonably be swapped for a set without .ndp_bpf while program is loaded. bpf_dev_offload_check() will also be used by map offload. Signed-off-by: Jakub Kicinski Reviewed-by: Quentin Monnet Signed-off-by: Daniel Borkmann --- kernel/bpf/offload.c | 28 ++++++++++++++++++++-------- 1 file changed, 20 insertions(+), 8 deletions(-) (limited to 'kernel') diff --git a/kernel/bpf/offload.c b/kernel/bpf/offload.c index 001ddfde7874..cdd1e19a668b 100644 --- a/kernel/bpf/offload.c +++ b/kernel/bpf/offload.c @@ -30,9 +30,19 @@ static DECLARE_RWSEM(bpf_devs_lock); static LIST_HEAD(bpf_prog_offload_devs); +static int bpf_dev_offload_check(struct net_device *netdev) +{ + if (!netdev) + return -EINVAL; + if (!netdev->netdev_ops->ndo_bpf) + return -EOPNOTSUPP; + return 0; +} + int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr) { struct bpf_prog_offload *offload; + int err; if (attr->prog_type != BPF_PROG_TYPE_SCHED_CLS && attr->prog_type != BPF_PROG_TYPE_XDP) @@ -49,12 +59,15 @@ int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr) offload->netdev = dev_get_by_index(current->nsproxy->net_ns, attr->prog_ifindex); - if (!offload->netdev) - goto err_free; + err = bpf_dev_offload_check(offload->netdev); + if (err) + goto err_maybe_put; down_write(&bpf_devs_lock); - if (offload->netdev->reg_state != NETREG_REGISTERED) + if (offload->netdev->reg_state != NETREG_REGISTERED) { + err = -EINVAL; goto err_unlock; + } prog->aux->offload = offload; list_add_tail(&offload->offloads, &bpf_prog_offload_devs); dev_put(offload->netdev); @@ -63,10 +76,11 @@ int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr) return 0; err_unlock: up_write(&bpf_devs_lock); - dev_put(offload->netdev); -err_free: +err_maybe_put: + if (offload->netdev) + dev_put(offload->netdev); kfree(offload); - return -EINVAL; + return err; } static int __bpf_offload_ndo(struct bpf_prog *prog, enum bpf_netdev_command cmd, @@ -80,8 +94,6 @@ static int __bpf_offload_ndo(struct bpf_prog *prog, enum bpf_netdev_command cmd, if (!offload) return -ENODEV; netdev = offload->netdev; - if (!netdev->netdev_ops->ndo_bpf) - return -EOPNOTSUPP; data->command = cmd; -- cgit v1.2.3 From a38845729ea3985db5d2544ec3ef3dc8f6313a27 Mon Sep 17 00:00:00 2001 From: Jakub Kicinski Date: Thu, 11 Jan 2018 20:29:09 -0800 Subject: bpf: offload: add map offload infrastructure BPF map offload follow similar path to program offload. At creation time users may specify ifindex of the device on which they want to create the map. Map will be validated by the kernel's .map_alloc_check callback and device driver will be called for the actual allocation. Map will have an empty set of operations associated with it (save for alloc and free callbacks). The real device callbacks are kept in map->offload->dev_ops because they have slightly different signatures. Map operations are called in process context so the driver may communicate with HW freely, msleep(), wait() etc. Map alloc and free callbacks are muxed via existing .ndo_bpf, and are always called with rtnl lock held. Maps and programs are guaranteed to be destroyed before .ndo_uninit (i.e. before unregister_netdev() returns). Map callbacks are invoked with bpf_devs_lock *read* locked, drivers must take care of exclusive locking if necessary. All offload-specific branches are marked with unlikely() (through bpf_map_is_dev_bound()), given that branch penalty will be negligible compared to IO anyway, and we don't want to penalize SW path unnecessarily. Signed-off-by: Jakub Kicinski Reviewed-by: Quentin Monnet Signed-off-by: Daniel Borkmann --- kernel/bpf/offload.c | 188 ++++++++++++++++++++++++++++++++++++++++++++++++-- kernel/bpf/syscall.c | 44 ++++++++++-- kernel/bpf/verifier.c | 7 ++ 3 files changed, 226 insertions(+), 13 deletions(-) (limited to 'kernel') diff --git a/kernel/bpf/offload.c b/kernel/bpf/offload.c index cdd1e19a668b..453785fa1881 100644 --- a/kernel/bpf/offload.c +++ b/kernel/bpf/offload.c @@ -24,11 +24,13 @@ #include #include -/* Protects bpf_prog_offload_devs and offload members of all progs. +/* Protects bpf_prog_offload_devs, bpf_map_offload_devs and offload members + * of all progs. * RTNL lock cannot be taken when holding this lock. */ static DECLARE_RWSEM(bpf_devs_lock); static LIST_HEAD(bpf_prog_offload_devs); +static LIST_HEAD(bpf_map_offload_devs); static int bpf_dev_offload_check(struct net_device *netdev) { @@ -250,11 +252,186 @@ int bpf_prog_offload_info_fill(struct bpf_prog_info *info, const struct bpf_prog_ops bpf_offload_prog_ops = { }; +static int bpf_map_offload_ndo(struct bpf_offloaded_map *offmap, + enum bpf_netdev_command cmd) +{ + struct netdev_bpf data = {}; + struct net_device *netdev; + + ASSERT_RTNL(); + + data.command = cmd; + data.offmap = offmap; + /* Caller must make sure netdev is valid */ + netdev = offmap->netdev; + + return netdev->netdev_ops->ndo_bpf(netdev, &data); +} + +struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr) +{ + struct net *net = current->nsproxy->net_ns; + struct bpf_offloaded_map *offmap; + int err; + + if (!capable(CAP_SYS_ADMIN)) + return ERR_PTR(-EPERM); + if (attr->map_type != BPF_MAP_TYPE_HASH) + return ERR_PTR(-EINVAL); + + offmap = kzalloc(sizeof(*offmap), GFP_USER); + if (!offmap) + return ERR_PTR(-ENOMEM); + + bpf_map_init_from_attr(&offmap->map, attr); + + rtnl_lock(); + down_write(&bpf_devs_lock); + offmap->netdev = __dev_get_by_index(net, attr->map_ifindex); + err = bpf_dev_offload_check(offmap->netdev); + if (err) + goto err_unlock; + + err = bpf_map_offload_ndo(offmap, BPF_OFFLOAD_MAP_ALLOC); + if (err) + goto err_unlock; + + list_add_tail(&offmap->offloads, &bpf_map_offload_devs); + up_write(&bpf_devs_lock); + rtnl_unlock(); + + return &offmap->map; + +err_unlock: + up_write(&bpf_devs_lock); + rtnl_unlock(); + kfree(offmap); + return ERR_PTR(err); +} + +static void __bpf_map_offload_destroy(struct bpf_offloaded_map *offmap) +{ + WARN_ON(bpf_map_offload_ndo(offmap, BPF_OFFLOAD_MAP_FREE)); + /* Make sure BPF_MAP_GET_NEXT_ID can't find this dead map */ + bpf_map_free_id(&offmap->map, true); + list_del_init(&offmap->offloads); + offmap->netdev = NULL; +} + +void bpf_map_offload_map_free(struct bpf_map *map) +{ + struct bpf_offloaded_map *offmap = map_to_offmap(map); + + rtnl_lock(); + down_write(&bpf_devs_lock); + if (offmap->netdev) + __bpf_map_offload_destroy(offmap); + up_write(&bpf_devs_lock); + rtnl_unlock(); + + kfree(offmap); +} + +int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value) +{ + struct bpf_offloaded_map *offmap = map_to_offmap(map); + int ret = -ENODEV; + + down_read(&bpf_devs_lock); + if (offmap->netdev) + ret = offmap->dev_ops->map_lookup_elem(offmap, key, value); + up_read(&bpf_devs_lock); + + return ret; +} + +int bpf_map_offload_update_elem(struct bpf_map *map, + void *key, void *value, u64 flags) +{ + struct bpf_offloaded_map *offmap = map_to_offmap(map); + int ret = -ENODEV; + + if (unlikely(flags > BPF_EXIST)) + return -EINVAL; + + down_read(&bpf_devs_lock); + if (offmap->netdev) + ret = offmap->dev_ops->map_update_elem(offmap, key, value, + flags); + up_read(&bpf_devs_lock); + + return ret; +} + +int bpf_map_offload_delete_elem(struct bpf_map *map, void *key) +{ + struct bpf_offloaded_map *offmap = map_to_offmap(map); + int ret = -ENODEV; + + down_read(&bpf_devs_lock); + if (offmap->netdev) + ret = offmap->dev_ops->map_delete_elem(offmap, key); + up_read(&bpf_devs_lock); + + return ret; +} + +int bpf_map_offload_get_next_key(struct bpf_map *map, void *key, void *next_key) +{ + struct bpf_offloaded_map *offmap = map_to_offmap(map); + int ret = -ENODEV; + + down_read(&bpf_devs_lock); + if (offmap->netdev) + ret = offmap->dev_ops->map_get_next_key(offmap, key, next_key); + up_read(&bpf_devs_lock); + + return ret; +} + +bool bpf_offload_dev_match(struct bpf_prog *prog, struct bpf_map *map) +{ + struct bpf_offloaded_map *offmap; + struct bpf_prog_offload *offload; + bool ret; + + if (!!bpf_prog_is_dev_bound(prog->aux) != !!bpf_map_is_dev_bound(map)) + return false; + if (!bpf_prog_is_dev_bound(prog->aux)) + return true; + + down_read(&bpf_devs_lock); + offload = prog->aux->offload; + offmap = map_to_offmap(map); + + ret = offload && offload->netdev == offmap->netdev; + up_read(&bpf_devs_lock); + + return ret; +} + +static void bpf_offload_orphan_all_progs(struct net_device *netdev) +{ + struct bpf_prog_offload *offload, *tmp; + + list_for_each_entry_safe(offload, tmp, &bpf_prog_offload_devs, offloads) + if (offload->netdev == netdev) + __bpf_prog_offload_destroy(offload->prog); +} + +static void bpf_offload_orphan_all_maps(struct net_device *netdev) +{ + struct bpf_offloaded_map *offmap, *tmp; + + list_for_each_entry_safe(offmap, tmp, &bpf_map_offload_devs, offloads) + if (offmap->netdev == netdev) + __bpf_map_offload_destroy(offmap); +} + static int bpf_offload_notification(struct notifier_block *notifier, ulong event, void *ptr) { struct net_device *netdev = netdev_notifier_info_to_dev(ptr); - struct bpf_prog_offload *offload, *tmp; ASSERT_RTNL(); @@ -265,11 +442,8 @@ static int bpf_offload_notification(struct notifier_block *notifier, break; down_write(&bpf_devs_lock); - list_for_each_entry_safe(offload, tmp, &bpf_prog_offload_devs, - offloads) { - if (offload->netdev == netdev) - __bpf_prog_offload_destroy(offload->prog); - } + bpf_offload_orphan_all_progs(netdev); + bpf_offload_orphan_all_maps(netdev); up_write(&bpf_devs_lock); break; default: diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c index a3f726bb42ea..c691b9e972e3 100644 --- a/kernel/bpf/syscall.c +++ b/kernel/bpf/syscall.c @@ -94,6 +94,11 @@ static int check_uarg_tail_zero(void __user *uaddr, return 0; } +const struct bpf_map_ops bpf_map_offload_ops = { + .map_alloc = bpf_map_offload_map_alloc, + .map_free = bpf_map_offload_map_free, +}; + static struct bpf_map *find_and_alloc_map(union bpf_attr *attr) { const struct bpf_map_ops *ops; @@ -111,6 +116,8 @@ static struct bpf_map *find_and_alloc_map(union bpf_attr *attr) if (err) return ERR_PTR(err); } + if (attr->map_ifindex) + ops = &bpf_map_offload_ops; map = ops->map_alloc(attr); if (IS_ERR(map)) return map; @@ -208,16 +215,25 @@ static int bpf_map_alloc_id(struct bpf_map *map) return id > 0 ? 0 : id; } -static void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock) +void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock) { unsigned long flags; + /* Offloaded maps are removed from the IDR store when their device + * disappears - even if someone holds an fd to them they are unusable, + * the memory is gone, all ops will fail; they are simply waiting for + * refcnt to drop to be freed. + */ + if (!map->id) + return; + if (do_idr_lock) spin_lock_irqsave(&map_idr_lock, flags); else __acquire(&map_idr_lock); idr_remove(&map_idr, map->id); + map->id = 0; if (do_idr_lock) spin_unlock_irqrestore(&map_idr_lock, flags); @@ -397,7 +413,7 @@ static int bpf_obj_name_cpy(char *dst, const char *src) return 0; } -#define BPF_MAP_CREATE_LAST_FIELD map_name +#define BPF_MAP_CREATE_LAST_FIELD map_ifindex /* called via syscall */ static int map_create(union bpf_attr *attr) { @@ -585,8 +601,10 @@ static int map_lookup_elem(union bpf_attr *attr) if (!value) goto free_key; - if (map->map_type == BPF_MAP_TYPE_PERCPU_HASH || - map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH) { + if (bpf_map_is_dev_bound(map)) { + err = bpf_map_offload_lookup_elem(map, key, value); + } else if (map->map_type == BPF_MAP_TYPE_PERCPU_HASH || + map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH) { err = bpf_percpu_hash_copy(map, key, value); } else if (map->map_type == BPF_MAP_TYPE_PERCPU_ARRAY) { err = bpf_percpu_array_copy(map, key, value); @@ -673,7 +691,10 @@ static int map_update_elem(union bpf_attr *attr) goto free_value; /* Need to create a kthread, thus must support schedule */ - if (map->map_type == BPF_MAP_TYPE_CPUMAP) { + if (bpf_map_is_dev_bound(map)) { + err = bpf_map_offload_update_elem(map, key, value, attr->flags); + goto out; + } else if (map->map_type == BPF_MAP_TYPE_CPUMAP) { err = map->ops->map_update_elem(map, key, value, attr->flags); goto out; } @@ -750,6 +771,11 @@ static int map_delete_elem(union bpf_attr *attr) goto err_put; } + if (bpf_map_is_dev_bound(map)) { + err = bpf_map_offload_delete_elem(map, key); + goto out; + } + preempt_disable(); __this_cpu_inc(bpf_prog_active); rcu_read_lock(); @@ -757,7 +783,7 @@ static int map_delete_elem(union bpf_attr *attr) rcu_read_unlock(); __this_cpu_dec(bpf_prog_active); preempt_enable(); - +out: if (!err) trace_bpf_map_delete_elem(map, ufd, key); kfree(key); @@ -807,9 +833,15 @@ static int map_get_next_key(union bpf_attr *attr) if (!next_key) goto free_key; + if (bpf_map_is_dev_bound(map)) { + err = bpf_map_offload_get_next_key(map, key, next_key); + goto out; + } + rcu_read_lock(); err = map->ops->map_get_next_key(map, key, next_key); rcu_read_unlock(); +out: if (err) goto free_next_key; diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c index 48b61caa94cb..ceabb394d2dc 100644 --- a/kernel/bpf/verifier.c +++ b/kernel/bpf/verifier.c @@ -4816,6 +4816,13 @@ static int check_map_prog_compatibility(struct bpf_verifier_env *env, return -EINVAL; } } + + if ((bpf_prog_is_dev_bound(prog->aux) || bpf_map_is_dev_bound(map)) && + !bpf_offload_dev_match(prog, map)) { + verbose(env, "offload device mismatch between prog and map\n"); + return -EINVAL; + } + return 0; } -- cgit v1.2.3