Skip to content
This repository has been archived by the owner on Apr 18, 2024. It is now read-only.

Commit

Permalink
mptcp: Initialize meta_tp after potentially failing instructions
Browse files Browse the repository at this point in the history 
In mptcp_alloc_mpcb, we are changing quite some fields of the meta_tp
and then do a potentially failing hash-insertion attempt. If this one
fails, we leave the meta_tp with incorrect sequence-numbers,...

Thus, we have to do this operation at the beginning. That way, when it
fails we can just continue in mptcp_create_master_sk when
mptcp_alloc_mpcb fails. Because, we really can't destroy the meta here,
because when we are coming from tcp_rcv_synsent_state_process we already
committed to maintain the metask. We rather move the forced closure now
to __mptcp_check_req_master.

Also, removing the sk_wmem_alloc set to 1, because that's already done
in sk_clone_lock().

Fixes: 369a043 ("mptcp: Build-Fix: Support lockless listeners in mptcp_alloc_mpcb")
Signed-off-by: Christoph Paasch <cpaasch@apple.com>
Signed-off-by: Matthieu Baerts <matthieu.baerts@tessares.net>
  • Loading branch information
@cpaasch @matttbe
cpaasch authored and matttbe committed Mar 8, 2019
1 parent dba9d5c commit 4d0b871
Showing 1 changed file with 79 additions and 79 deletions.
158 changes: 79 additions & 79 deletions net/mptcp/mptcp_ctrl.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -1135,7 +1135,7 @@ static int mptcp_alloc_mpcb(struct sock *meta_sk, __u64 remote_key,
struct sock *master_sk;
struct inet_connection_sock *meta_icsk = inet_csk(meta_sk);
struct tcp_sock *master_tp, *meta_tp = tcp_sk(meta_sk);
u64 idsn;
u64 snd_idsn, rcv_idsn;

dst_release(meta_sk->sk_rx_dst);
meta_sk->sk_rx_dst = NULL;
Expand All @@ -1152,14 +1152,79 @@ static int mptcp_alloc_mpcb(struct sock *meta_sk, __u64 remote_key,

mpcb = kmem_cache_zalloc(mptcp_cb_cache, GFP_ATOMIC);
if (!mpcb) {
/* sk_free (and __sk_free) requirese wmem_alloc to be 1.
* All the rest is set to 0 thanks to __GFP_ZERO above.
*/
refcount_set(&master_sk->sk_wmem_alloc, 1);
sk_free(master_sk);
return -ENOBUFS;
}

/* Store the mptcp version agreed on initial handshake */
mpcb->mptcp_ver = mptcp_ver;

/* Store the keys and generate the peer's token */
mpcb->mptcp_loc_key = meta_tp->mptcp_loc_key;
mpcb->mptcp_loc_token = meta_tp->mptcp_loc_token;

/* Generate Initial data-sequence-numbers */
mptcp_key_sha1(mpcb->mptcp_loc_key, NULL, &snd_idsn);
snd_idsn++;
mpcb->snd_high_order[0] = snd_idsn >> 32;
mpcb->snd_high_order[1] = mpcb->snd_high_order[0] - 1;

mpcb->mptcp_rem_key = remote_key;
mptcp_key_sha1(mpcb->mptcp_rem_key, &mpcb->mptcp_rem_token, &rcv_idsn);
rcv_idsn++;
mpcb->rcv_high_order[0] = rcv_idsn >> 32;
mpcb->rcv_high_order[1] = mpcb->rcv_high_order[0] + 1;

mpcb->meta_sk = meta_sk;
mpcb->master_sk = master_sk;

skb_queue_head_init(&mpcb->reinject_queue);
mutex_init(&mpcb->mpcb_mutex);

/* Init time-wait stuff */
INIT_LIST_HEAD(&mpcb->tw_list);

INIT_HLIST_HEAD(&mpcb->callback_list);
INIT_HLIST_HEAD(&mpcb->conn_list);
spin_lock_init(&mpcb->mpcb_list_lock);

mpcb->orig_sk_rcvbuf = meta_sk->sk_rcvbuf;
mpcb->orig_sk_sndbuf = meta_sk->sk_sndbuf;
mpcb->orig_window_clamp = meta_tp->window_clamp;

/* The meta is directly linked - set refcnt to 1 */
refcount_set(&mpcb->mpcb_refcnt, 1);

if (!meta_tp->inside_tk_table) {
/* Adding the meta_tp in the token hashtable - coming from server-side */
rcu_read_lock_bh();
spin_lock(&mptcp_tk_hashlock);

/* With lockless listeners, we might process two ACKs at the
* same time. With TCP, inet_csk_complete_hashdance takes care
* of this. But, for MPTCP this would be too late if we add
* this MPTCP-socket in the token table (new subflows might
* come in and match on this socket here.
* So, we need to check if someone else already added the token
* and revert in that case. The other guy won the race...
*/
if (mptcp_find_token(mpcb->mptcp_loc_token)) {
spin_unlock(&mptcp_tk_hashlock);
rcu_read_unlock_bh();

inet_put_port(master_sk);
kmem_cache_free(mptcp_cb_cache, mpcb);
sk_free(master_sk);

return -ENOBUFS;
}
__mptcp_hash_insert(meta_tp, mpcb->mptcp_loc_token);

spin_unlock(&mptcp_tk_hashlock);
rcu_read_unlock_bh();
}
master_tp->inside_tk_table = 0;

#if IS_ENABLED(CONFIG_IPV6)
if (meta_icsk->icsk_af_ops == &mptcp_v6_mapped) {
struct tcp6_sock *master_tp6 = (struct tcp6_sock *)master_sk;
Expand Down Expand Up @@ -1216,34 +1281,16 @@ static int mptcp_alloc_mpcb(struct sock *meta_sk, __u64 remote_key,

meta_tp->mptcp = NULL;

/* Store the mptcp version agreed on initial handshake */
mpcb->mptcp_ver = mptcp_ver;

/* Store the keys and generate the peer's token */
mpcb->mptcp_loc_key = meta_tp->mptcp_loc_key;
mpcb->mptcp_loc_token = meta_tp->mptcp_loc_token;

/* Generate Initial data-sequence-numbers */
mptcp_key_sha1(mpcb->mptcp_loc_key, NULL, &idsn);
idsn++;
mpcb->snd_high_order[0] = idsn >> 32;
mpcb->snd_high_order[1] = mpcb->snd_high_order[0] - 1;

meta_tp->write_seq = (u32)idsn;
meta_tp->write_seq = (u32)snd_idsn;
meta_tp->snd_sml = meta_tp->write_seq;
meta_tp->snd_una = meta_tp->write_seq;
meta_tp->snd_nxt = meta_tp->write_seq;
meta_tp->pushed_seq = meta_tp->write_seq;
meta_tp->snd_up = meta_tp->write_seq;

mpcb->mptcp_rem_key = remote_key;
mptcp_key_sha1(mpcb->mptcp_rem_key, &mpcb->mptcp_rem_token, &idsn);
idsn++;
mpcb->rcv_high_order[0] = idsn >> 32;
mpcb->rcv_high_order[1] = mpcb->rcv_high_order[0] + 1;
meta_tp->copied_seq = (u32) idsn;
meta_tp->rcv_nxt = (u32) idsn;
meta_tp->rcv_wup = (u32) idsn;
meta_tp->copied_seq = (u32)rcv_idsn;
meta_tp->rcv_nxt = (u32)rcv_idsn;
meta_tp->rcv_wup = (u32)rcv_idsn;

meta_tp->snd_wl1 = meta_tp->rcv_nxt - 1;
meta_tp->snd_wnd = window;
Expand All @@ -1257,13 +1304,10 @@ static int mptcp_alloc_mpcb(struct sock *meta_sk, __u64 remote_key,
master_tp->meta_sk = meta_sk;
meta_tp->mpcb = mpcb;
meta_tp->meta_sk = meta_sk;
mpcb->meta_sk = meta_sk;
mpcb->master_sk = master_sk;

meta_tp->was_meta_sk = 0;

/* Initialize the queues */
skb_queue_head_init(&mpcb->reinject_queue);
master_tp->out_of_order_queue = RB_ROOT;
master_sk->tcp_rtx_queue = RB_ROOT;
INIT_LIST_HEAD(&master_tp->tsq_node);
Expand All @@ -1275,8 +1319,6 @@ static int mptcp_alloc_mpcb(struct sock *meta_sk, __u64 remote_key,
*/
inet_csk(master_sk)->icsk_bind_hash = NULL;

mutex_init(&mpcb->mpcb_mutex);

/* Init the accept_queue structure, we support a queue of 32 pending
* connections, it does not need to be huge, since we only store here
* pending subflow creations.
Expand Down Expand Up @@ -1304,52 +1346,8 @@ static int mptcp_alloc_mpcb(struct sock *meta_sk, __u64 remote_key,
/* Has been set for sending out the SYN */
inet_csk_clear_xmit_timer(meta_sk, ICSK_TIME_RETRANS);

if (!meta_tp->inside_tk_table) {
/* Adding the meta_tp in the token hashtable - coming from server-side */
rcu_read_lock_bh();
spin_lock(&mptcp_tk_hashlock);

/* With lockless listeners, we might process two ACKs at the
* same time. With TCP, inet_csk_complete_hashdance takes care
* of this. But, for MPTCP this would be too late if we add
* this MPTCP-socket in the token table (new subflows might
* come in and match on this socket here.
* So, we need to check if someone else already added the token
* and revert in that case. The other guy won the race...
*/
if (mptcp_find_token(mpcb->mptcp_loc_token)) {
spin_unlock(&mptcp_tk_hashlock);
rcu_read_unlock_bh();

inet_put_port(master_sk);
kmem_cache_free(mptcp_cb_cache, mpcb);
sk_free(master_sk);

return -ENOBUFS;
}
__mptcp_hash_insert(meta_tp, mpcb->mptcp_loc_token);

spin_unlock(&mptcp_tk_hashlock);
rcu_read_unlock_bh();
}
master_tp->inside_tk_table = 0;

/* Init time-wait stuff */
INIT_LIST_HEAD(&mpcb->tw_list);

INIT_HLIST_HEAD(&mpcb->callback_list);
INIT_HLIST_HEAD(&mpcb->conn_list);
spin_lock_init(&mpcb->mpcb_list_lock);

mptcp_mpcb_inherit_sockopts(meta_sk, master_sk);

mpcb->orig_sk_rcvbuf = meta_sk->sk_rcvbuf;
mpcb->orig_sk_sndbuf = meta_sk->sk_sndbuf;
mpcb->orig_window_clamp = meta_tp->window_clamp;

/* The meta is directly linked - set refcnt to 1 */
refcount_set(&mpcb->mpcb_refcnt, 1);

mptcp_init_path_manager(mpcb);
mptcp_init_scheduler(mpcb);

Expand Down Expand Up @@ -2044,8 +2042,6 @@ int mptcp_create_master_sk(struct sock *meta_sk, __u64 remote_key,

inet_csk_prepare_forced_close(master_sk);
tcp_done(master_sk);
inet_csk_prepare_forced_close(meta_sk);
tcp_done(meta_sk);

err_alloc_mpcb:
return -ENOBUFS;
Expand Down Expand Up @@ -2082,8 +2078,12 @@ static int __mptcp_check_req_master(struct sock *child,
child_tp->mptcp_loc_token = mtreq->mptcp_loc_token;

if (mptcp_create_master_sk(meta_sk, mtreq->mptcp_rem_key,
mtreq->mptcp_ver, child_tp->snd_wnd))
mtreq->mptcp_ver, child_tp->snd_wnd)) {
inet_csk_prepare_forced_close(meta_sk);
tcp_done(meta_sk);

return -ENOBUFS;
}

child = tcp_sk(child)->mpcb->master_sk;
child_tp = tcp_sk(child);
Expand Down

0 comments on commit 4d0b871

Please sign in to comment.