From fdb5c4531c1e0e50e609df83f736b6f3a02896e2 Mon Sep 17 00:00:00 2001 From: Sean Young Date: Tue, 19 Jun 2018 00:04:24 +0100 Subject: bpf: fix attach type BPF_LIRC_MODE2 dependency wrt CONFIG_CGROUP_BPF If the kernel is compiled with CONFIG_CGROUP_BPF not enabled, it is not possible to attach, detach or query IR BPF programs to /dev/lircN devices, making them impossible to use. For embedded devices, it should be possible to use IR decoding without cgroups or CONFIG_CGROUP_BPF enabled. This change requires some refactoring, since bpf_prog_{attach,detach,query} functions are now always compiled, but their code paths for cgroups need moving out. Rather than a #ifdef CONFIG_CGROUP_BPF in kernel/bpf/syscall.c, moving them to kernel/bpf/cgroup.c and kernel/bpf/sockmap.c does not require #ifdefs since that is already conditionally compiled. Fixes: f4364dcfc86d ("media: rc: introduce BPF_PROG_LIRC_MODE2") Signed-off-by: Sean Young Signed-off-by: Daniel Borkmann --- kernel/bpf/sockmap.c | 18 ++++++++++++++++++ 1 file changed, 18 insertions(+) (limited to 'kernel/bpf/sockmap.c') diff --git a/kernel/bpf/sockmap.c b/kernel/bpf/sockmap.c index 52a91d816c0e..81d0c55a77aa 100644 --- a/kernel/bpf/sockmap.c +++ b/kernel/bpf/sockmap.c @@ -1915,6 +1915,24 @@ int sock_map_prog(struct bpf_map *map, struct bpf_prog *prog, u32 type) return 0; } +int sockmap_get_from_fd(const union bpf_attr *attr, int type, + struct bpf_prog *prog) +{ + int ufd = attr->target_fd; + struct bpf_map *map; + struct fd f; + int err; + + f = fdget(ufd); + map = __bpf_map_get(f); + if (IS_ERR(map)) + return PTR_ERR(map); + + err = sock_map_prog(map, prog, attr->attach_type); + fdput(f); + return err; +} + static void *sock_map_lookup(struct bpf_map *map, void *key) { return NULL; -- cgit v1.2.3 From 9901c5d77e969d8215a8e8d087ef02e6feddc84c Mon Sep 17 00:00:00 2001 From: John Fastabend Date: Sat, 30 Jun 2018 06:17:36 -0700 Subject: bpf: sockmap, fix crash when ipv6 sock is added This fixes a crash where we assign tcp_prot to IPv6 sockets instead of tcpv6_prot. Previously we overwrote the sk->prot field with tcp_prot even in the AF_INET6 case. This patch ensures the correct tcp_prot and tcpv6_prot are used. Tested with 'netserver -6' and 'netperf -H [IPv6]' as well as 'netperf -H [IPv4]'. The ESTABLISHED check resolves the previously crashing case here. Fixes: 174a79ff9515 ("bpf: sockmap with sk redirect support") Reported-by: syzbot+5c063698bdbfac19f363@syzkaller.appspotmail.com Acked-by: Martin KaFai Lau Signed-off-by: John Fastabend Signed-off-by: Wei Wang Signed-off-by: Daniel Borkmann --- kernel/bpf/sockmap.c | 58 +++++++++++++++++++++++++++++++++++++++++++--------- 1 file changed, 48 insertions(+), 10 deletions(-) (limited to 'kernel/bpf/sockmap.c') diff --git a/kernel/bpf/sockmap.c b/kernel/bpf/sockmap.c index 81d0c55a77aa..bfdfbd199c3b 100644 --- a/kernel/bpf/sockmap.c +++ b/kernel/bpf/sockmap.c @@ -140,6 +140,7 @@ static int bpf_tcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, static int bpf_tcp_sendmsg(struct sock *sk, struct msghdr *msg, size_t size); static int bpf_tcp_sendpage(struct sock *sk, struct page *page, int offset, size_t size, int flags); +static void bpf_tcp_close(struct sock *sk, long timeout); static inline struct smap_psock *smap_psock_sk(const struct sock *sk) { @@ -161,7 +162,42 @@ out: return !empty; } -static struct proto tcp_bpf_proto; +enum { + SOCKMAP_IPV4, + SOCKMAP_IPV6, + SOCKMAP_NUM_PROTS, +}; + +enum { + SOCKMAP_BASE, + SOCKMAP_TX, + SOCKMAP_NUM_CONFIGS, +}; + +static struct proto *saved_tcpv6_prot __read_mostly; +static DEFINE_SPINLOCK(tcpv6_prot_lock); +static struct proto bpf_tcp_prots[SOCKMAP_NUM_PROTS][SOCKMAP_NUM_CONFIGS]; +static void build_protos(struct proto prot[SOCKMAP_NUM_CONFIGS], + struct proto *base) +{ + prot[SOCKMAP_BASE] = *base; + prot[SOCKMAP_BASE].close = bpf_tcp_close; + prot[SOCKMAP_BASE].recvmsg = bpf_tcp_recvmsg; + prot[SOCKMAP_BASE].stream_memory_read = bpf_tcp_stream_read; + + prot[SOCKMAP_TX] = prot[SOCKMAP_BASE]; + prot[SOCKMAP_TX].sendmsg = bpf_tcp_sendmsg; + prot[SOCKMAP_TX].sendpage = bpf_tcp_sendpage; +} + +static void update_sk_prot(struct sock *sk, struct smap_psock *psock) +{ + int family = sk->sk_family == AF_INET6 ? SOCKMAP_IPV6 : SOCKMAP_IPV4; + int conf = psock->bpf_tx_msg ? SOCKMAP_TX : SOCKMAP_BASE; + + sk->sk_prot = &bpf_tcp_prots[family][conf]; +} + static int bpf_tcp_init(struct sock *sk) { struct smap_psock *psock; @@ -181,14 +217,17 @@ static int bpf_tcp_init(struct sock *sk) psock->save_close = sk->sk_prot->close; psock->sk_proto = sk->sk_prot; - if (psock->bpf_tx_msg) { - tcp_bpf_proto.sendmsg = bpf_tcp_sendmsg; - tcp_bpf_proto.sendpage = bpf_tcp_sendpage; - tcp_bpf_proto.recvmsg = bpf_tcp_recvmsg; - tcp_bpf_proto.stream_memory_read = bpf_tcp_stream_read; + /* Build IPv6 sockmap whenever the address of tcpv6_prot changes */ + if (sk->sk_family == AF_INET6 && + unlikely(sk->sk_prot != smp_load_acquire(&saved_tcpv6_prot))) { + spin_lock_bh(&tcpv6_prot_lock); + if (likely(sk->sk_prot != saved_tcpv6_prot)) { + build_protos(bpf_tcp_prots[SOCKMAP_IPV6], sk->sk_prot); + smp_store_release(&saved_tcpv6_prot, sk->sk_prot); + } + spin_unlock_bh(&tcpv6_prot_lock); } - - sk->sk_prot = &tcp_bpf_proto; + update_sk_prot(sk, psock); rcu_read_unlock(); return 0; } @@ -1111,8 +1150,7 @@ static void bpf_tcp_msg_add(struct smap_psock *psock, static int bpf_tcp_ulp_register(void) { - tcp_bpf_proto = tcp_prot; - tcp_bpf_proto.close = bpf_tcp_close; + build_protos(bpf_tcp_prots[SOCKMAP_IPV4], &tcp_prot); /* Once BPF TX ULP is registered it is never unregistered. It * will be in the ULP list for the lifetime of the system. Doing * duplicate registers is not a problem. -- cgit v1.2.3 From 54fedb42c6537dcb0102e4a58a88456a6286999d Mon Sep 17 00:00:00 2001 From: John Fastabend Date: Sat, 30 Jun 2018 06:17:41 -0700 Subject: bpf: sockmap, fix smap_list_map_remove when psock is in many maps If a hashmap is free'd with open socks it removes the reference to the hash entry from the psock. If that is the last reference to the psock then it will also be free'd by the reference counting logic. However the current logic that removes the hash reference from the list of references is broken. In smap_list_remove() we first check if the sockmap entry matches and then check if the hashmap entry matches. But, the sockmap entry sill always match because its NULL in this case which causes the first entry to be removed from the list. If this is always the "right" entry (because the user adds/removes entries in order) then everything is OK but otherwise a subsequent bpf_tcp_close() may reference a free'd object. To fix this create two list handlers one for sockmap and one for sockhash. Reported-by: syzbot+0ce137753c78f7b6acc1@syzkaller.appspotmail.com Fixes: 81110384441a ("bpf: sockmap, add hash map support") Acked-by: Martin KaFai Lau Signed-off-by: John Fastabend Signed-off-by: Daniel Borkmann --- kernel/bpf/sockmap.c | 34 ++++++++++++++++++++++------------ 1 file changed, 22 insertions(+), 12 deletions(-) (limited to 'kernel/bpf/sockmap.c') diff --git a/kernel/bpf/sockmap.c b/kernel/bpf/sockmap.c index bfdfbd199c3b..65a937ed5762 100644 --- a/kernel/bpf/sockmap.c +++ b/kernel/bpf/sockmap.c @@ -1602,17 +1602,27 @@ free_stab: return ERR_PTR(err); } -static void smap_list_remove(struct smap_psock *psock, - struct sock **entry, - struct htab_elem *hash_link) +static void smap_list_map_remove(struct smap_psock *psock, + struct sock **entry) { struct smap_psock_map_entry *e, *tmp; list_for_each_entry_safe(e, tmp, &psock->maps, list) { - if (e->entry == entry || e->hash_link == hash_link) { + if (e->entry == entry) + list_del(&e->list); + } +} + +static void smap_list_hash_remove(struct smap_psock *psock, + struct htab_elem *hash_link) +{ + struct smap_psock_map_entry *e, *tmp; + + list_for_each_entry_safe(e, tmp, &psock->maps, list) { + struct htab_elem *c = e->hash_link; + + if (c == hash_link) list_del(&e->list); - break; - } } } @@ -1647,7 +1657,7 @@ static void sock_map_free(struct bpf_map *map) * to be null and queued for garbage collection. */ if (likely(psock)) { - smap_list_remove(psock, &stab->sock_map[i], NULL); + smap_list_map_remove(psock, &stab->sock_map[i]); smap_release_sock(psock, sock); } write_unlock_bh(&sock->sk_callback_lock); @@ -1706,7 +1716,7 @@ static int sock_map_delete_elem(struct bpf_map *map, void *key) if (psock->bpf_parse) smap_stop_sock(psock, sock); - smap_list_remove(psock, &stab->sock_map[k], NULL); + smap_list_map_remove(psock, &stab->sock_map[k]); smap_release_sock(psock, sock); out: write_unlock_bh(&sock->sk_callback_lock); @@ -1908,7 +1918,7 @@ static int sock_map_ctx_update_elem(struct bpf_sock_ops_kern *skops, struct smap_psock *opsock = smap_psock_sk(osock); write_lock_bh(&osock->sk_callback_lock); - smap_list_remove(opsock, &stab->sock_map[i], NULL); + smap_list_map_remove(opsock, &stab->sock_map[i]); smap_release_sock(opsock, osock); write_unlock_bh(&osock->sk_callback_lock); } @@ -2142,7 +2152,7 @@ static void sock_hash_free(struct bpf_map *map) * (psock) to be null and queued for garbage collection. */ if (likely(psock)) { - smap_list_remove(psock, NULL, l); + smap_list_hash_remove(psock, l); smap_release_sock(psock, sock); } write_unlock_bh(&sock->sk_callback_lock); @@ -2322,7 +2332,7 @@ static int sock_hash_ctx_update_elem(struct bpf_sock_ops_kern *skops, psock = smap_psock_sk(l_old->sk); hlist_del_rcu(&l_old->hash_node); - smap_list_remove(psock, NULL, l_old); + smap_list_hash_remove(psock, l_old); smap_release_sock(psock, l_old->sk); free_htab_elem(htab, l_old); } @@ -2390,7 +2400,7 @@ static int sock_hash_delete_elem(struct bpf_map *map, void *key) * to be null and queued for garbage collection. */ if (likely(psock)) { - smap_list_remove(psock, NULL, l); + smap_list_hash_remove(psock, l); smap_release_sock(psock, sock); } write_unlock_bh(&sock->sk_callback_lock); -- cgit v1.2.3 From e9db4ef6bf4ca9894bb324c76e01b8f1a16b2650 Mon Sep 17 00:00:00 2001 From: John Fastabend Date: Sat, 30 Jun 2018 06:17:47 -0700 Subject: bpf: sockhash fix omitted bucket lock in sock_close First the sk_callback_lock() was being used to protect both the sock callback hooks and the psock->maps list. This got overly convoluted after the addition of sockhash (in sockmap it made some sense because masp and callbacks were tightly coupled) so lets split out a specific lock for maps and only use the callback lock for its intended purpose. This fixes a couple cases where we missed using maps lock when it was in fact needed. Also this makes it easier to follow the code because now we can put the locking closer to the actual code its serializing. Next, in sock_hash_delete_elem() the pattern was as follows, sock_hash_delete_elem() [...] spin_lock(bucket_lock) l = lookup_elem_raw() if (l) hlist_del_rcu() write_lock(sk_callback_lock) .... destroy psock ... write_unlock(sk_callback_lock) spin_unlock(bucket_lock) The ordering is necessary because we only know the {p}sock after dereferencing the hash table which we can't do unless we have the bucket lock held. Once we have the bucket lock and the psock element it is deleted from the hashmap to ensure any other path doing a lookup will fail. Finally, the refcnt is decremented and if zero the psock is destroyed. In parallel with the above (or free'ing the map) a tcp close event may trigger tcp_close(). Which at the moment omits the bucket lock altogether (oops!) where the flow looks like this, bpf_tcp_close() [...] write_lock(sk_callback_lock) for each psock->maps // list of maps this sock is part of hlist_del_rcu(ref_hash_node); .... destroy psock ... write_unlock(sk_callback_lock) Obviously, and demonstrated by syzbot, this is broken because we can have multiple threads deleting entries via hlist_del_rcu(). To fix this we might be tempted to wrap the hlist operation in a bucket lock but that would create a lock inversion problem. In summary to follow locking rules the psocks maps list needs the sk_callback_lock (after this patch maps_lock) but we need the bucket lock to do the hlist_del_rcu. To resolve the lock inversion problem pop the head of the maps list repeatedly and remove the reference until no more are left. If a delete happens in parallel from the BPF API that is OK as well because it will do a similar action, lookup the lock in the map/hash, delete it from the map/hash, and dec the refcnt. We check for this case before doing a destroy on the psock to ensure we don't have two threads tearing down a psock. The new logic is as follows, bpf_tcp_close() e = psock_map_pop(psock->maps) // done with map lock bucket_lock() // lock hash list bucket l = lookup_elem_raw(head, hash, key, key_size); if (l) { //only get here if elmnt was not already removed hlist_del_rcu() ... destroy psock... } bucket_unlock() And finally for all the above to work add missing locking around map operations per above. Then add RCU annotations and use rcu_dereference/rcu_assign_pointer to manage values relying on RCU so that the object is not free'd from sock_hash_free() while it is being referenced in bpf_tcp_close(). Reported-by: syzbot+0ce137753c78f7b6acc1@syzkaller.appspotmail.com Fixes: 81110384441a ("bpf: sockmap, add hash map support") Signed-off-by: John Fastabend Signed-off-by: Daniel Borkmann --- kernel/bpf/sockmap.c | 145 ++++++++++++++++++++++++++++++++++----------------- 1 file changed, 96 insertions(+), 49 deletions(-) (limited to 'kernel/bpf/sockmap.c') diff --git a/kernel/bpf/sockmap.c b/kernel/bpf/sockmap.c index 65a937ed5762..ac09b35a9567 100644 --- a/kernel/bpf/sockmap.c +++ b/kernel/bpf/sockmap.c @@ -72,6 +72,7 @@ struct bpf_htab { u32 n_buckets; u32 elem_size; struct bpf_sock_progs progs; + struct rcu_head rcu; }; struct htab_elem { @@ -89,8 +90,8 @@ enum smap_psock_state { struct smap_psock_map_entry { struct list_head list; struct sock **entry; - struct htab_elem *hash_link; - struct bpf_htab *htab; + struct htab_elem __rcu *hash_link; + struct bpf_htab __rcu *htab; }; struct smap_psock { @@ -120,6 +121,7 @@ struct smap_psock { struct bpf_prog *bpf_parse; struct bpf_prog *bpf_verdict; struct list_head maps; + spinlock_t maps_lock; /* Back reference used when sock callback trigger sockmap operations */ struct sock *sock; @@ -258,16 +260,54 @@ out: rcu_read_unlock(); } +static struct htab_elem *lookup_elem_raw(struct hlist_head *head, + u32 hash, void *key, u32 key_size) +{ + struct htab_elem *l; + + hlist_for_each_entry_rcu(l, head, hash_node) { + if (l->hash == hash && !memcmp(&l->key, key, key_size)) + return l; + } + + return NULL; +} + +static inline struct bucket *__select_bucket(struct bpf_htab *htab, u32 hash) +{ + return &htab->buckets[hash & (htab->n_buckets - 1)]; +} + +static inline struct hlist_head *select_bucket(struct bpf_htab *htab, u32 hash) +{ + return &__select_bucket(htab, hash)->head; +} + static void free_htab_elem(struct bpf_htab *htab, struct htab_elem *l) { atomic_dec(&htab->count); kfree_rcu(l, rcu); } +static struct smap_psock_map_entry *psock_map_pop(struct sock *sk, + struct smap_psock *psock) +{ + struct smap_psock_map_entry *e; + + spin_lock_bh(&psock->maps_lock); + e = list_first_entry_or_null(&psock->maps, + struct smap_psock_map_entry, + list); + if (e) + list_del(&e->list); + spin_unlock_bh(&psock->maps_lock); + return e; +} + static void bpf_tcp_close(struct sock *sk, long timeout) { void (*close_fun)(struct sock *sk, long timeout); - struct smap_psock_map_entry *e, *tmp; + struct smap_psock_map_entry *e; struct sk_msg_buff *md, *mtmp; struct smap_psock *psock; struct sock *osk; @@ -286,7 +326,6 @@ static void bpf_tcp_close(struct sock *sk, long timeout) */ close_fun = psock->save_close; - write_lock_bh(&sk->sk_callback_lock); if (psock->cork) { free_start_sg(psock->sock, psock->cork); kfree(psock->cork); @@ -299,20 +338,38 @@ static void bpf_tcp_close(struct sock *sk, long timeout) kfree(md); } - list_for_each_entry_safe(e, tmp, &psock->maps, list) { + e = psock_map_pop(sk, psock); + while (e) { if (e->entry) { osk = cmpxchg(e->entry, sk, NULL); if (osk == sk) { - list_del(&e->list); smap_release_sock(psock, sk); } } else { - hlist_del_rcu(&e->hash_link->hash_node); - smap_release_sock(psock, e->hash_link->sk); - free_htab_elem(e->htab, e->hash_link); + struct htab_elem *link = rcu_dereference(e->hash_link); + struct bpf_htab *htab = rcu_dereference(e->htab); + struct hlist_head *head; + struct htab_elem *l; + struct bucket *b; + + b = __select_bucket(htab, link->hash); + head = &b->head; + raw_spin_lock_bh(&b->lock); + l = lookup_elem_raw(head, + link->hash, link->key, + htab->map.key_size); + /* If another thread deleted this object skip deletion. + * The refcnt on psock may or may not be zero. + */ + if (l) { + hlist_del_rcu(&link->hash_node); + smap_release_sock(psock, link->sk); + free_htab_elem(htab, link); + } + raw_spin_unlock_bh(&b->lock); } + e = psock_map_pop(sk, psock); } - write_unlock_bh(&sk->sk_callback_lock); rcu_read_unlock(); close_fun(sk, timeout); } @@ -1395,7 +1452,9 @@ static void smap_release_sock(struct smap_psock *psock, struct sock *sock) { if (refcount_dec_and_test(&psock->refcnt)) { tcp_cleanup_ulp(sock); + write_lock_bh(&sock->sk_callback_lock); smap_stop_sock(psock, sock); + write_unlock_bh(&sock->sk_callback_lock); clear_bit(SMAP_TX_RUNNING, &psock->state); rcu_assign_sk_user_data(sock, NULL); call_rcu_sched(&psock->rcu, smap_destroy_psock); @@ -1546,6 +1605,7 @@ static struct smap_psock *smap_init_psock(struct sock *sock, int node) INIT_LIST_HEAD(&psock->maps); INIT_LIST_HEAD(&psock->ingress); refcount_set(&psock->refcnt, 1); + spin_lock_init(&psock->maps_lock); rcu_assign_sk_user_data(sock, psock); sock_hold(sock); @@ -1607,10 +1667,12 @@ static void smap_list_map_remove(struct smap_psock *psock, { struct smap_psock_map_entry *e, *tmp; + spin_lock_bh(&psock->maps_lock); list_for_each_entry_safe(e, tmp, &psock->maps, list) { if (e->entry == entry) list_del(&e->list); } + spin_unlock_bh(&psock->maps_lock); } static void smap_list_hash_remove(struct smap_psock *psock, @@ -1618,12 +1680,14 @@ static void smap_list_hash_remove(struct smap_psock *psock, { struct smap_psock_map_entry *e, *tmp; + spin_lock_bh(&psock->maps_lock); list_for_each_entry_safe(e, tmp, &psock->maps, list) { - struct htab_elem *c = e->hash_link; + struct htab_elem *c = rcu_dereference(e->hash_link); if (c == hash_link) list_del(&e->list); } + spin_unlock_bh(&psock->maps_lock); } static void sock_map_free(struct bpf_map *map) @@ -1649,7 +1713,6 @@ static void sock_map_free(struct bpf_map *map) if (!sock) continue; - write_lock_bh(&sock->sk_callback_lock); psock = smap_psock_sk(sock); /* This check handles a racing sock event that can get the * sk_callback_lock before this case but after xchg happens @@ -1660,7 +1723,6 @@ static void sock_map_free(struct bpf_map *map) smap_list_map_remove(psock, &stab->sock_map[i]); smap_release_sock(psock, sock); } - write_unlock_bh(&sock->sk_callback_lock); } rcu_read_unlock(); @@ -1709,7 +1771,6 @@ static int sock_map_delete_elem(struct bpf_map *map, void *key) if (!sock) return -EINVAL; - write_lock_bh(&sock->sk_callback_lock); psock = smap_psock_sk(sock); if (!psock) goto out; @@ -1719,7 +1780,6 @@ static int sock_map_delete_elem(struct bpf_map *map, void *key) smap_list_map_remove(psock, &stab->sock_map[k]); smap_release_sock(psock, sock); out: - write_unlock_bh(&sock->sk_callback_lock); return 0; } @@ -1800,7 +1860,6 @@ static int __sock_map_ctx_update_elem(struct bpf_map *map, } } - write_lock_bh(&sock->sk_callback_lock); psock = smap_psock_sk(sock); /* 2. Do not allow inheriting programs if psock exists and has @@ -1857,7 +1916,9 @@ static int __sock_map_ctx_update_elem(struct bpf_map *map, if (err) goto out_free; smap_init_progs(psock, verdict, parse); + write_lock_bh(&sock->sk_callback_lock); smap_start_sock(psock, sock); + write_unlock_bh(&sock->sk_callback_lock); } /* 4. Place psock in sockmap for use and stop any programs on @@ -1867,9 +1928,10 @@ static int __sock_map_ctx_update_elem(struct bpf_map *map, */ if (map_link) { e->entry = map_link; + spin_lock_bh(&psock->maps_lock); list_add_tail(&e->list, &psock->maps); + spin_unlock_bh(&psock->maps_lock); } - write_unlock_bh(&sock->sk_callback_lock); return err; out_free: smap_release_sock(psock, sock); @@ -1880,7 +1942,6 @@ out_progs: } if (tx_msg) bpf_prog_put(tx_msg); - write_unlock_bh(&sock->sk_callback_lock); kfree(e); return err; } @@ -1917,10 +1978,8 @@ static int sock_map_ctx_update_elem(struct bpf_sock_ops_kern *skops, if (osock) { struct smap_psock *opsock = smap_psock_sk(osock); - write_lock_bh(&osock->sk_callback_lock); smap_list_map_remove(opsock, &stab->sock_map[i]); smap_release_sock(opsock, osock); - write_unlock_bh(&osock->sk_callback_lock); } out: return err; @@ -2109,14 +2168,13 @@ free_htab: return ERR_PTR(err); } -static inline struct bucket *__select_bucket(struct bpf_htab *htab, u32 hash) +static void __bpf_htab_free(struct rcu_head *rcu) { - return &htab->buckets[hash & (htab->n_buckets - 1)]; -} + struct bpf_htab *htab; -static inline struct hlist_head *select_bucket(struct bpf_htab *htab, u32 hash) -{ - return &__select_bucket(htab, hash)->head; + htab = container_of(rcu, struct bpf_htab, rcu); + bpf_map_area_free(htab->buckets); + kfree(htab); } static void sock_hash_free(struct bpf_map *map) @@ -2135,16 +2193,18 @@ static void sock_hash_free(struct bpf_map *map) */ rcu_read_lock(); for (i = 0; i < htab->n_buckets; i++) { - struct hlist_head *head = select_bucket(htab, i); + struct bucket *b = __select_bucket(htab, i); + struct hlist_head *head; struct hlist_node *n; struct htab_elem *l; + raw_spin_lock_bh(&b->lock); + head = &b->head; hlist_for_each_entry_safe(l, n, head, hash_node) { struct sock *sock = l->sk; struct smap_psock *psock; hlist_del_rcu(&l->hash_node); - write_lock_bh(&sock->sk_callback_lock); psock = smap_psock_sk(sock); /* This check handles a racing sock event that can get * the sk_callback_lock before this case but after xchg @@ -2155,13 +2215,12 @@ static void sock_hash_free(struct bpf_map *map) smap_list_hash_remove(psock, l); smap_release_sock(psock, sock); } - write_unlock_bh(&sock->sk_callback_lock); - kfree(l); + free_htab_elem(htab, l); } + raw_spin_unlock_bh(&b->lock); } rcu_read_unlock(); - bpf_map_area_free(htab->buckets); - kfree(htab); + call_rcu(&htab->rcu, __bpf_htab_free); } static struct htab_elem *alloc_sock_hash_elem(struct bpf_htab *htab, @@ -2188,19 +2247,6 @@ static struct htab_elem *alloc_sock_hash_elem(struct bpf_htab *htab, return l_new; } -static struct htab_elem *lookup_elem_raw(struct hlist_head *head, - u32 hash, void *key, u32 key_size) -{ - struct htab_elem *l; - - hlist_for_each_entry_rcu(l, head, hash_node) { - if (l->hash == hash && !memcmp(&l->key, key, key_size)) - return l; - } - - return NULL; -} - static inline u32 htab_map_hash(const void *key, u32 key_len) { return jhash(key, key_len, 0); @@ -2320,9 +2366,12 @@ static int sock_hash_ctx_update_elem(struct bpf_sock_ops_kern *skops, goto bucket_err; } - e->hash_link = l_new; - e->htab = container_of(map, struct bpf_htab, map); + rcu_assign_pointer(e->hash_link, l_new); + rcu_assign_pointer(e->htab, + container_of(map, struct bpf_htab, map)); + spin_lock_bh(&psock->maps_lock); list_add_tail(&e->list, &psock->maps); + spin_unlock_bh(&psock->maps_lock); /* add new element to the head of the list, so that * concurrent search will find it before old elem @@ -2392,7 +2441,6 @@ static int sock_hash_delete_elem(struct bpf_map *map, void *key) struct smap_psock *psock; hlist_del_rcu(&l->hash_node); - write_lock_bh(&sock->sk_callback_lock); psock = smap_psock_sk(sock); /* This check handles a racing sock event that can get the * sk_callback_lock before this case but after xchg happens @@ -2403,7 +2451,6 @@ static int sock_hash_delete_elem(struct bpf_map *map, void *key) smap_list_hash_remove(psock, l); smap_release_sock(psock, sock); } - write_unlock_bh(&sock->sk_callback_lock); free_htab_elem(htab, l); ret = 0; } -- cgit v1.2.3 From caac76a5170eb508529bbff9d9300e9c57126444 Mon Sep 17 00:00:00 2001 From: John Fastabend Date: Sat, 30 Jun 2018 06:17:52 -0700 Subject: bpf: sockhash, add release routine Add map_release_uref pointer to hashmap ops. This was dropped when original sockhash code was ported into bpf-next before initial commit. Fixes: 81110384441a ("bpf: sockmap, add hash map support") Acked-by: Martin KaFai Lau Signed-off-by: John Fastabend Signed-off-by: Daniel Borkmann --- kernel/bpf/sockmap.c | 1 + 1 file changed, 1 insertion(+) (limited to 'kernel/bpf/sockmap.c') diff --git a/kernel/bpf/sockmap.c b/kernel/bpf/sockmap.c index ac09b35a9567..cf7b6a6dbd1f 100644 --- a/kernel/bpf/sockmap.c +++ b/kernel/bpf/sockmap.c @@ -2496,6 +2496,7 @@ const struct bpf_map_ops sock_hash_ops = { .map_get_next_key = sock_hash_get_next_key, .map_update_elem = sock_hash_update_elem, .map_delete_elem = sock_hash_delete_elem, + .map_release_uref = sock_map_release, }; BPF_CALL_4(bpf_sock_map_update, struct bpf_sock_ops_kern *, bpf_sock, -- cgit v1.2.3 From 547b3aa451ae2739585547db9fbdee11a43ff999 Mon Sep 17 00:00:00 2001 From: John Fastabend Date: Thu, 5 Jul 2018 08:05:56 -0700 Subject: bpf: sockmap, error path can not release psock in multi-map case The current code, in the error path of sock_hash_ctx_update_elem, checks if the sock has a psock in the user data and if so decrements the reference count of the psock. However, if the error happens early in the error path we may have never incremented the psock reference count and if the psock exists because the sock is in another map then we may inadvertently decrement the reference count. Fix this by making the error path only call smap_release_sock if the error happens after the increment. Reported-by: syzbot+d464d2c20c717ef5a6a8@syzkaller.appspotmail.com Fixes: 81110384441a ("bpf: sockmap, add hash map support") Signed-off-by: John Fastabend Signed-off-by: Alexei Starovoitov --- kernel/bpf/sockmap.c | 17 ++++++----------- 1 file changed, 6 insertions(+), 11 deletions(-) (limited to 'kernel/bpf/sockmap.c') diff --git a/kernel/bpf/sockmap.c b/kernel/bpf/sockmap.c index cf7b6a6dbd1f..3847a7ce7dae 100644 --- a/kernel/bpf/sockmap.c +++ b/kernel/bpf/sockmap.c @@ -1896,7 +1896,7 @@ static int __sock_map_ctx_update_elem(struct bpf_map *map, e = kzalloc(sizeof(*e), GFP_ATOMIC | __GFP_NOWARN); if (!e) { err = -ENOMEM; - goto out_progs; + goto out_free; } } @@ -2342,7 +2342,10 @@ static int sock_hash_ctx_update_elem(struct bpf_sock_ops_kern *skops, if (err) goto err; - /* bpf_map_update_elem() can be called in_irq() */ + /* psock is valid here because otherwise above *ctx_update_elem would + * have thrown an error. It is safe to skip error check. + */ + psock = smap_psock_sk(sock); raw_spin_lock_bh(&b->lock); l_old = lookup_elem_raw(head, hash, key, key_size); if (l_old && map_flags == BPF_NOEXIST) { @@ -2360,12 +2363,6 @@ static int sock_hash_ctx_update_elem(struct bpf_sock_ops_kern *skops, goto bucket_err; } - psock = smap_psock_sk(sock); - if (unlikely(!psock)) { - err = -EINVAL; - goto bucket_err; - } - rcu_assign_pointer(e->hash_link, l_new); rcu_assign_pointer(e->htab, container_of(map, struct bpf_htab, map)); @@ -2388,12 +2385,10 @@ static int sock_hash_ctx_update_elem(struct bpf_sock_ops_kern *skops, raw_spin_unlock_bh(&b->lock); return 0; bucket_err: + smap_release_sock(psock, sock); raw_spin_unlock_bh(&b->lock); err: kfree(e); - psock = smap_psock_sk(sock); - if (psock) - smap_release_sock(psock, sock); return err; } -- cgit v1.2.3 From 1d1ef005dbc6de673c62cbd2562290ada3090463 Mon Sep 17 00:00:00 2001 From: John Fastabend Date: Thu, 5 Jul 2018 08:06:01 -0700 Subject: bpf: sockmap, hash table is RCU so readers do not need locks This removes locking from readers of RCU hash table. Its not necessary. Fixes: 81110384441a ("bpf: sockmap, add hash map support") Signed-off-by: John Fastabend Signed-off-by: Alexei Starovoitov --- kernel/bpf/sockmap.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'kernel/bpf/sockmap.c') diff --git a/kernel/bpf/sockmap.c b/kernel/bpf/sockmap.c index 3847a7ce7dae..00fb2e328d1b 100644 --- a/kernel/bpf/sockmap.c +++ b/kernel/bpf/sockmap.c @@ -2467,10 +2467,8 @@ struct sock *__sock_hash_lookup_elem(struct bpf_map *map, void *key) b = __select_bucket(htab, hash); head = &b->head; - raw_spin_lock_bh(&b->lock); l = lookup_elem_raw(head, hash, key, key_size); sk = l ? l->sk : NULL; - raw_spin_unlock_bh(&b->lock); return sk; } -- cgit v1.2.3 From 99ba2b5aba24e022683a7db63204f9e306fe7ab9 Mon Sep 17 00:00:00 2001 From: John Fastabend Date: Thu, 5 Jul 2018 08:50:04 -0700 Subject: bpf: sockhash, disallow bpf_tcp_close and update in parallel After latest lock updates there is no longer anything preventing a close and recvmsg call running in parallel. Additionally, we can race update with close if we close a socket and simultaneously update if via the BPF userspace API (note the cgroup ops are already run with sock_lock held). To resolve this take sock_lock in close and update paths. Reported-by: syzbot+b680e42077a0d7c9a0c4@syzkaller.appspotmail.com Fixes: e9db4ef6bf4c ("bpf: sockhash fix omitted bucket lock in sock_close") Signed-off-by: John Fastabend Signed-off-by: Alexei Starovoitov --- kernel/bpf/sockmap.c | 15 +++++++++++++++ 1 file changed, 15 insertions(+) (limited to 'kernel/bpf/sockmap.c') diff --git a/kernel/bpf/sockmap.c b/kernel/bpf/sockmap.c index 00fb2e328d1b..9c67e96fe336 100644 --- a/kernel/bpf/sockmap.c +++ b/kernel/bpf/sockmap.c @@ -312,10 +312,12 @@ static void bpf_tcp_close(struct sock *sk, long timeout) struct smap_psock *psock; struct sock *osk; + lock_sock(sk); rcu_read_lock(); psock = smap_psock_sk(sk); if (unlikely(!psock)) { rcu_read_unlock(); + release_sock(sk); return sk->sk_prot->close(sk, timeout); } @@ -371,6 +373,7 @@ static void bpf_tcp_close(struct sock *sk, long timeout) e = psock_map_pop(sk, psock); } rcu_read_unlock(); + release_sock(sk); close_fun(sk, timeout); } @@ -2069,7 +2072,13 @@ static int sock_map_update_elem(struct bpf_map *map, return -EOPNOTSUPP; } + lock_sock(skops.sk); + preempt_disable(); + rcu_read_lock(); err = sock_map_ctx_update_elem(&skops, map, key, flags); + rcu_read_unlock(); + preempt_enable(); + release_sock(skops.sk); fput(socket->file); return err; } @@ -2410,7 +2419,13 @@ static int sock_hash_update_elem(struct bpf_map *map, return -EINVAL; } + lock_sock(skops.sk); + preempt_disable(); + rcu_read_lock(); err = sock_hash_ctx_update_elem(&skops, map, key, flags); + rcu_read_unlock(); + preempt_enable(); + release_sock(skops.sk); fput(socket->file); return err; } -- cgit v1.2.3 From 7ebc14d507b4b55105da8d1a1eda323381529cc7 Mon Sep 17 00:00:00 2001 From: John Fastabend Date: Thu, 5 Jul 2018 08:50:10 -0700 Subject: bpf: sockmap, consume_skb in close path Currently, when a sock is closed and the bpf_tcp_close() callback is used we remove memory but do not free the skb. Call consume_skb() if the skb is attached to the buffer. Reported-by: syzbot+d464d2c20c717ef5a6a8@syzkaller.appspotmail.com Fixes: 1aa12bdf1bfb ("bpf: sockmap, add sock close() hook to remove socks") Signed-off-by: John Fastabend Signed-off-by: Alexei Starovoitov --- kernel/bpf/sockmap.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'kernel/bpf/sockmap.c') diff --git a/kernel/bpf/sockmap.c b/kernel/bpf/sockmap.c index 9c67e96fe336..dfc8a8a07c1f 100644 --- a/kernel/bpf/sockmap.c +++ b/kernel/bpf/sockmap.c @@ -571,7 +571,8 @@ static int free_sg(struct sock *sk, int start, struct sk_msg_buff *md) while (sg[i].length) { free += sg[i].length; sk_mem_uncharge(sk, sg[i].length); - put_page(sg_page(&sg[i])); + if (!md->skb) + put_page(sg_page(&sg[i])); sg[i].length = 0; sg[i].page_link = 0; sg[i].offset = 0; @@ -580,6 +581,8 @@ static int free_sg(struct sock *sk, int start, struct sk_msg_buff *md) if (i == MAX_SKB_FRAGS) i = 0; } + if (md->skb) + consume_skb(md->skb); return free; } -- cgit v1.2.3 From 0ea488ff8d23c93da383fcf424825c298b13b1fb Mon Sep 17 00:00:00 2001 From: John Fastabend Date: Thu, 5 Jul 2018 08:50:15 -0700 Subject: bpf: sockmap, convert bpf_compute_data_pointers to bpf_*_sk_skb In commit 'bpf: bpf_compute_data uses incorrect cb structure' (8108a7751512) we added the routine bpf_compute_data_end_sk_skb() to compute the correct data_end values, but this has since been lost. In kernel v4.14 this was correct and the above patch was applied in it entirety. Then when v4.14 was merged into v4.15-rc1 net-next tree we lost the piece that renamed bpf_compute_data_pointers to the new function bpf_compute_data_end_sk_skb. This was done here, e1ea2f9856b7 ("Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net") When it conflicted with the following rename patch, 6aaae2b6c433 ("bpf: rename bpf_compute_data_end into bpf_compute_data_pointers") Finally, after a refactor I thought even the function bpf_compute_data_end_sk_skb() was no longer needed and it was erroneously removed. However, we never reverted the sk_skb_convert_ctx_access() usage of tcp_skb_cb which had been committed and survived the merge conflict. Here we fix this by adding back the helper and *_data_end_sk_skb() usage. Using the bpf_skc_data_end mapping is not correct because it expects a qdisc_skb_cb object but at the sock layer this is not the case. Even though it happens to work here because we don't overwrite any data in-use at the socket layer and the cb structure is cleared later this has potential to create some subtle issues. But, even more concretely the filter.c access check uses tcp_skb_cb. And by some act of chance though, struct bpf_skb_data_end { struct qdisc_skb_cb qdisc_cb; /* 0 28 */ /* XXX 4 bytes hole, try to pack */ void * data_meta; /* 32 8 */ void * data_end; /* 40 8 */ /* size: 48, cachelines: 1, members: 3 */ /* sum members: 44, holes: 1, sum holes: 4 */ /* last cacheline: 48 bytes */ }; and then tcp_skb_cb, struct tcp_skb_cb { [...] struct { __u32 flags; /* 24 4 */ struct sock * sk_redir; /* 32 8 */ void * data_end; /* 40 8 */ } bpf; /* 24 */ }; So when we use offset_of() to track down the byte offset we get 40 in either case and everything continues to work. Fix this mess and use correct structures its unclear how long this might actually work for until someone moves the structs around. Reported-by: Martin KaFai Lau Fixes: e1ea2f9856b7 ("Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net") Fixes: 6aaae2b6c433 ("bpf: rename bpf_compute_data_end into bpf_compute_data_pointers") Signed-off-by: John Fastabend Signed-off-by: Alexei Starovoitov --- kernel/bpf/sockmap.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'kernel/bpf/sockmap.c') diff --git a/kernel/bpf/sockmap.c b/kernel/bpf/sockmap.c index dfc8a8a07c1f..98fb7938beea 100644 --- a/kernel/bpf/sockmap.c +++ b/kernel/bpf/sockmap.c @@ -1236,7 +1236,7 @@ static int smap_verdict_func(struct smap_psock *psock, struct sk_buff *skb) */ TCP_SKB_CB(skb)->bpf.sk_redir = NULL; skb->sk = psock->sock; - bpf_compute_data_pointers(skb); + bpf_compute_data_end_sk_skb(skb); preempt_disable(); rc = (*prog->bpf_func)(skb, prog->insnsi); preempt_enable(); @@ -1491,7 +1491,7 @@ static int smap_parse_func_strparser(struct strparser *strp, * any socket yet. */ skb->sk = psock->sock; - bpf_compute_data_pointers(skb); + bpf_compute_data_end_sk_skb(skb); rc = (*prog->bpf_func)(skb, prog->insnsi); skb->sk = NULL; rcu_read_unlock(); -- cgit v1.2.3