Revert "net: Add a second bind table hashed by port and address"

This reverts: commit d5a42de8bdbe ("net: Add a second bind table hashed by port and address") commit 538aaf9b2383 ("selftests: Add test for timing a bind request to a port with a populated bhash entry") Link: https://lore.kernel.org/netdev/20220520001834.2247810-1-kuba@kernel.org/ There are a few things that need to be fixed here: * Updating bhash2 in cases where the socket's rcv saddr changes * Adding bhash2 hashbucket locks Links to syzbot reports: https://lore.kernel.org/netdev/00000000000022208805e0df247a@google.com/ https://lore.kernel.org/netdev/0000000000003f33bc05dfaf44fe@google.com/

Joanne Koong committed 4 years ago

Commit 593d1eb1 parent 219b51a10 files · +83 −611Browse files

Showing the changes to include (3 of 10 files). Show all

Unified · Split

3 files

include/net/inet_connection_sock.h +0 −3
include/net/inet_hashtables.h +1 −67
include/net/sock.h +0 −14

include/net/inet_connection_sock.h+0−3

@@ -25,7 +25,6 @@ · 24 unmodified lines
25	25	#undef INET_CSK_CLEAR_TIMERS
26	26
27	27	struct inet_bind_bucket;
28		−struct inet_bind2_bucket;
29	28	struct tcp_congestion_ops;
30	29
31	30	/*
@@ -58,7 +57,6 @@ struct inet_connection_sock_af_ops { · 26 unmodified lines
58	57	*
59	58	* @icsk_accept_queue: FIFO of established children
60	59	* @icsk_bind_hash: Bind node
61		− * @icsk_bind2_hash: Bind node in the bhash2 table
62	60	* @icsk_timeout: Timeout
63	61	* @icsk_retransmit_timer: Resend (no ack)
64	62	* @icsk_rto: Retransmit timeout
@@ -85,7 +83,6 @@ struct inet_connection_sock { · 20 unmodified lines
85	83	struct inet_sock icsk_inet;
86	84	struct request_sock_queue icsk_accept_queue;
87	85	struct inet_bind_bucket *icsk_bind_hash;
88		− struct inet_bind2_bucket *icsk_bind2_hash;
89	86	unsigned long icsk_timeout;
90	87	struct timer_list icsk_retransmit_timer;
91	88	struct timer_list icsk_delack_timer;

include/net/inet_hashtables.h+1−67

@@ -90,32 +90,11 @@ struct inet_bind_bucket { · 89 unmodified lines
90	90	struct hlist_head owners;
91	91	};
92	92
93		−struct inet_bind2_bucket {
94		− possible_net_t ib_net;
95		− int l3mdev;
96		− unsigned short port;
97		− union {
98		−#if IS_ENABLED(CONFIG_IPV6)
99		− struct in6_addr v6_rcv_saddr;
100		−#endif
101		− __be32 rcv_saddr;
102		− };
103		− /* Node in the inet2_bind_hashbucket chain */
104		− struct hlist_node node;
105		− /* List of sockets hashed to this bucket */
106		− struct hlist_head owners;
107		−};
108		−
109	93	static inline struct net ib_net(struct inet_bind_bucket ib)
110	94	{
111	95	return read_pnet(&ib->ib_net);
112	96	}
113	97
114		−static inline struct net ib2_net(struct inet_bind2_bucket ib)
115		−{
116		− return read_pnet(&ib->ib_net);
117		−}
118		−
119	98	#define inet_bind_bucket_for_each(tb, head) \
120	99	hlist_for_each_entry(tb, head, node)
121	100
@@ -124,15 +103,6 @@ struct inet_bind_hashbucket { · 2 unmodified lines
124	103	struct hlist_head chain;
125	104	};
126	105
127		−/* This is synchronized using the inet_bind_hashbucket's spinlock.
128		− * Instead of having separate spinlocks, the inet_bind2_hashbucket can share
129		− * the inet_bind_hashbucket's given that in every case where the bhash2 table
130		− * is useful, a lookup in the bhash table also occurs.
131		− */
132		−struct inet_bind2_hashbucket {
133		− struct hlist_head chain;
134		−};
135		−
136	106	/* Sockets can be hashed in established or listening table.
137	107	* We must use different 'nulls' end-of-chain value for all hash buckets :
138	108	* A socket might transition from ESTABLISH to LISTEN state without
@@ -164,12 +134,6 @@ struct inet_hashinfo { · 25 unmodified lines
164	134	*/
165	135	struct kmem_cache *bind_bucket_cachep;
166	136	struct inet_bind_hashbucket *bhash;
167		− /* The 2nd binding table hashed by port and address.
168		− * This is used primarily for expediting the resolution of bind
169		− * conflicts.
170		− */
171		− struct kmem_cache *bind2_bucket_cachep;
172		− struct inet_bind2_hashbucket *bhash2;
173	137	unsigned int bhash_size;
174	138
175	139	/* The 2nd listener table hashed by local port and address */
@@ -229,36 +193,6 @@ inet_bind_bucket_create(struct kmem_cache cachep, struct net net, · 53 unmodified lines
229	193	void inet_bind_bucket_destroy(struct kmem_cache *cachep,
230	194	struct inet_bind_bucket *tb);
231	195
232		−static inline bool check_bind_bucket_match(struct inet_bind_bucket *tb,
233		− struct net *net,
234		− const unsigned short port,
235		− int l3mdev)
236		−{
237		− return net_eq(ib_net(tb), net) && tb->port == port &&
238		− tb->l3mdev == l3mdev;
239		−}
240		−
241		−struct inet_bind2_bucket *
242		−inet_bind2_bucket_create(struct kmem_cache cachep, struct net net,
243		− struct inet_bind2_hashbucket *head,
244		− const unsigned short port, int l3mdev,
245		− const struct sock *sk);
246		−
247		−void inet_bind2_bucket_destroy(struct kmem_cache *cachep,
248		− struct inet_bind2_bucket *tb);
249		−
250		−struct inet_bind2_bucket *
251		−inet_bind2_bucket_find(struct inet_hashinfo hinfo, struct net net,
252		− const unsigned short port, int l3mdev,
253		− struct sock *sk,
254		− struct inet_bind2_hashbucket **head);
255		−
256		−bool check_bind2_bucket_match_nulladdr(struct inet_bind2_bucket *tb,
257		− struct net *net,
258		− const unsigned short port,
259		− int l3mdev,
260		− const struct sock *sk);
261		−
262	196	static inline u32 inet_bhashfn(const struct net *net, const __u16 lport,
263	197	const u32 bhash_size)
264	198	{
@@ -266,7 +200,7 @@ static inline u32 inet_bhashfn(const struct net *net, const __u16 lport, · 1 unmodified line
266	200	}
267	201
268	202	void inet_bind_hash(struct sock sk, struct inet_bind_bucket tb,
269		− struct inet_bind2_bucket *tb2, const unsigned short snum);
	203	+ const unsigned short snum);
270	204
271	205	/* Caller must disable local BH processing. */
272	206	int __inet_inherit_port(const struct sock sk, struct sock child);

include/net/sock.h+0−14

@@ -348,7 +348,6 @@ struct sk_filter; · 347 unmodified lines
348	348	* @sk_txtime_report_errors: set report errors mode for SO_TXTIME
349	349	* @sk_txtime_unused: unused txtime flags
350	350	* @ns_tracker: tracker for netns reference
351		− * @sk_bind2_node: bind node in the bhash2 table
352	351	*/
353	352	struct sock {
354	353	/*
@@ -538,7 +537,6 @@ struct sock { · 183 unmodified lines
538	537	#endif
539	538	struct rcu_head sk_rcu;
540	539	netns_tracker ns_tracker;
541		− struct hlist_node sk_bind2_node;
542	540	};
543	541
544	542	enum sk_pacing {
@@ -819,16 +817,6 @@ static inline void sk_add_bind_node(struct sock *sk, · 274 unmodified lines
819	817	hlist_add_head(&sk->sk_bind_node, list);
820	818	}
821	819
822		−static inline void __sk_del_bind2_node(struct sock *sk)
823		−{
824		− __hlist_del(&sk->sk_bind2_node);
825		−}
826		−
827		−static inline void sk_add_bind2_node(struct sock sk, struct hlist_head list)
828		−{
829		− hlist_add_head(&sk->sk_bind2_node, list);
830		−}
831		−
832	820	#define sk_for_each(__sk, list) \
833	821	hlist_for_each_entry(__sk, list, sk_node)
834	822	#define sk_for_each_rcu(__sk, list) \
@@ -846,8 +834,6 @@ static inline void sk_add_bind2_node(struct sock sk, struct hlist_head list) · 11 unmodified lines
846	834	hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
847	835	#define sk_for_each_bound(__sk, list) \
848	836	hlist_for_each_entry(__sk, list, sk_bind_node)
849		−#define sk_for_each_bound_bhash2(__sk, list) \
850		− hlist_for_each_entry(__sk, list, sk_bind2_node)
851	837
852	838	/**
853	839	* sk_for_each_entry_offset_rcu - iterate over a list at a given struct offset