[PATCH] Revert tcp_skb_cb to it's original size using 'struct tcp_skb_cb_ext'
by Rao Shoaib
This patch allocates a structure 'struct tcp_skb_cb_ext' to store values
that do not fit in tcp_skb_cb. On the Tx side the structure is allocated
as part of the skb carrying the payload. It is freed when the data is acked.
The structure is not passed down below TCP. On the Rxside the structure is
allocated when the packet is received by TCP, it is freed when the skb
is freed.
No changes are required to any of the functions implementing various skb
operations.
There might be some performance degradation due to extra allocation.
Maintaining a cache of 'struct tcp_skb_cb_ext' might help.
This patch reverses the size of tcp_skb_cb back to 48.
Signed-off-by: Rao Shoaib <rao.shoaib(a)oracle.com>
---
include/linux/skbuff.h | 2 +-
include/net/mptcp.h | 6 ++-
include/net/tcp.h | 105 ++++++++++++++++++++++++++++++++++++-----------
net/ipv4/syncookies.c | 2 +-
net/ipv4/tcp.c | 18 ++++++--
net/ipv4/tcp_input.c | 16 ++++++--
net/ipv4/tcp_ipv4.c | 22 ++++++----
net/ipv4/tcp_output.c | 25 +++++++----
net/ipv6/syncookies.c | 2 +-
net/ipv6/tcp_ipv6.c | 36 +++++++++++-----
net/mptcp/mptcp_ctrl.c | 2 +-
net/mptcp/mptcp_input.c | 37 +++++++++++++++--
net/mptcp/mptcp_ipv4.c | 6 ++-
net/mptcp/mptcp_ipv6.c | 3 +-
net/mptcp/mptcp_output.c | 29 +++++++------
15 files changed, 230 insertions(+), 81 deletions(-)
diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index f66cd5e..ca2e26a 100644
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
@@ -540,7 +540,7 @@ struct sk_buff {
* want to keep them across layers you have to do a skb_clone()
* first. This is owned by whoever has the skb queued ATM.
*/
- char cb[56] __aligned(8);
+ char cb[48] __aligned(8);
unsigned long _skb_refdst;
void (*destructor)(struct sk_buff *skb);
diff --git a/include/net/mptcp.h b/include/net/mptcp.h
index 876f1e6..ff0df36 100644
--- a/include/net/mptcp.h
+++ b/include/net/mptcp.h
@@ -1000,12 +1000,14 @@ static inline void mptcp_sub_force_close_all(struct mptcp_cb *mpcb,
static inline bool mptcp_is_data_seq(const struct sk_buff *skb)
{
- return TCP_SKB_CB(skb)->mptcp_flags & MPTCPHDR_SEQ;
+ return ((TCP_SKB_CB(skb)->tcp_skb_ext == NULL) ? 0 :
+ TCP_SKB_CB(skb)->mptcp_flags & MPTCPHDR_SEQ);
}
static inline bool mptcp_is_data_fin(const struct sk_buff *skb)
{
- return TCP_SKB_CB(skb)->mptcp_flags & MPTCPHDR_FIN;
+ return ((TCP_SKB_CB(skb)->tcp_skb_ext == NULL) ? 0 :
+ TCP_SKB_CB(skb)->mptcp_flags & MPTCPHDR_FIN);
}
/* Is it a data-fin while in infinite mapping mode?
diff --git a/include/net/tcp.h b/include/net/tcp.h
index 655ecd4..abe3641 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -822,6 +822,30 @@ static inline u32 tcp_skb_timestamp(const struct sk_buff *skb)
#define TCPHDR_ECE 0x40
#define TCPHDR_CWR 0x80
+struct tcp_skb_cb_ext {
+ union {
+ struct inet_skb_parm h4;
+#if IS_ENABLED(CONFIG_IPV6)
+ struct inet6_skb_parm h6;
+#endif
+ } header;
+
+ union { /* For MPTCP outgoing frames */
+ __u32 path_mask; /* paths that tried to send this skb */
+ __u32 dss[6]; /* DSS options */
+ };
+
+ __u8 mptcp_flags; /* flags for the MPTCP layer */
+ __u8 dss_off; /* Number of 4-byte words until */
+};
+
+#define tcp_ipcb_v4 tcp_skb_ext->header.h4
+#define tcp_ipcb_v6 tcp_skb_ext->header.h6
+#define mptcp_flags tcp_skb_ext->mptcp_flags
+#define path_mask tcp_skb_ext->path_mask
+#define dss_off tcp_skb_ext->dss_off
+#define dss tcp_skb_ext->dss
+
/* This is what the send packet queuing engine uses to pass
* TCP per-packet control information to the transmission code.
* We also store the host-order sequence numbers in here too.
@@ -829,7 +853,7 @@ static inline u32 tcp_skb_timestamp(const struct sk_buff *skb)
* If this grows please adjust skbuff.h:skbuff->cb[xxx] size appropriately.
*/
struct tcp_skb_cb {
- __u32 seq; /* Starting sequence number */
+ __u32 seq; /* Starting sequence number */
__u32 end_seq; /* SEQ + FIN + SYN + datalen */
union {
/* Note : tcp_tw_isn is used in input path only
@@ -842,11 +866,6 @@ struct tcp_skb_cb {
__u32 tcp_gso_segs;
};
-#ifdef CONFIG_MPTCP
- __u8 mptcp_flags; /* flags for the MPTCP layer */
- __u8 dss_off; /* Number of 4-byte words until
- * seq-number */
-#endif
__u8 tcp_flags; /* TCP header flags. (tcp[13]) */
__u8 sacked; /* State flags for SACK/FACK. */
@@ -861,25 +880,65 @@ struct tcp_skb_cb {
__u8 ip_dsfield; /* IPv4 tos or IPv6 dsfield */
/* 1 byte hole */
+ struct tcp_skb_cb_ext *tcp_skb_ext;
__u32 ack_seq; /* Sequence number ACK'd */
- union {
- union {
- struct inet_skb_parm h4;
-#if IS_ENABLED(CONFIG_IPV6)
- struct inet6_skb_parm h6;
-#endif
- } header; /* For incoming frames */
-#ifdef CONFIG_MPTCP
- union { /* For MPTCP outgoing frames */
- __u32 path_mask; /* paths that tried to send this skb */
- __u32 dss[6]; /* DSS options */
- };
-#endif
- };
};
#define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0]))
+static inline struct tcp_skb_cb_ext *tcp_skb_cb_ext_alloc(struct sk_buff *skb)
+{
+ int extra;
+
+ if (TCP_SKB_CB(skb)->tcp_skb_ext != NULL) {
+ return (TCP_SKB_CB(skb)->tcp_skb_ext);
+ }
+
+ extra = SKB_DATA_ALIGN(sizeof(struct tcp_skb_cb_ext));
+
+ TCP_SKB_CB(skb)->tcp_skb_ext = kzalloc(extra, GFP_KERNEL);
+ if (TCP_SKB_CB(skb)->tcp_skb_ext != NULL) {
+ skb->truesize += extra;
+ return (TCP_SKB_CB(skb)->tcp_skb_ext);
+ }
+ return (NULL);
+}
+
+static inline void tcp_sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
+{
+ if (TCP_SKB_CB(skb)->tcp_skb_ext != NULL) {
+ kfree (TCP_SKB_CB(skb)->tcp_skb_ext);
+ TCP_SKB_CB(skb)->tcp_skb_ext = NULL;
+ }
+ sk_wmem_free_skb(sk, skb);
+}
+
+typedef enum {
+ SKB_DEFAULT,
+ SKB_FCLONE
+} tcp_skb_flag_t;
+
+static inline struct sk_buff *tcp_alloc_skb(unsigned int size,
+ gfp_t priority, tcp_skb_flag_t tcp_skb_flag) {
+ struct sk_buff *skb;
+
+ switch (tcp_skb_flag) {
+ case (SKB_FCLONE) :
+ skb = alloc_skb_fclone(size, priority);
+ break;
+ case (SKB_DEFAULT):
+ default:
+ skb = alloc_skb(size, priority);
+ break;
+ }
+
+ if (skb && (tcp_skb_cb_ext_alloc(skb) == NULL)) {
+ kfree_skb(skb);
+ return (NULL);
+ }
+
+ return (skb);
+}
#if IS_ENABLED(CONFIG_IPV6)
/* This is the variant of inet6_iif() that must be used by TCP,
@@ -887,7 +946,7 @@ struct tcp_skb_cb {
*/
static inline int tcp_v6_iif(const struct sk_buff *skb)
{
- return TCP_SKB_CB(skb)->header.h6.iif;
+ return TCP_SKB_CB(skb)->tcp_ipcb_v6.iif;
}
#endif
@@ -1533,7 +1592,7 @@ static inline void tcp_write_queue_purge(struct sock *sk)
struct sk_buff *skb;
while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
- sk_wmem_free_skb(sk, skb);
+ tcp_sk_wmem_free_skb(sk, skb);
sk_mem_reclaim(sk);
tcp_clear_all_retrans_hints(tcp_sk(sk));
}
@@ -1869,7 +1928,7 @@ void tcp_init(void);
*/
static inline struct ip_options_rcu *tcp_v4_save_options(struct sk_buff *skb)
{
- const struct ip_options *opt = &TCP_SKB_CB(skb)->header.h4.opt;
+ const struct ip_options *opt = &TCP_SKB_CB(skb)->tcp_ipcb_v4.opt;
struct ip_options_rcu *dopt = NULL;
if (opt->optlen) {
diff --git a/net/ipv4/syncookies.c b/net/ipv4/syncookies.c
index 5ce0cff..2da0282 100644
--- a/net/ipv4/syncookies.c
+++ b/net/ipv4/syncookies.c
@@ -310,7 +310,7 @@ EXPORT_SYMBOL(cookie_ecn_ok);
struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb)
{
- struct ip_options *opt = &TCP_SKB_CB(skb)->header.h4.opt;
+ struct ip_options *opt = &TCP_SKB_CB(skb)->tcp_ipcb_v4.opt;
struct tcp_options_received tcp_opt;
struct mptcp_options_received mopt;
struct inet_request_sock *ireq;
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 63c107c..f3cbaf2 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -852,7 +852,12 @@ struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp)
skb = alloc_skb_fclone(size + sk->sk_prot->max_header, gfp);
if (skb) {
- if (sk_wmem_schedule(sk, skb->truesize)) {
+ /*
+ * allocate tcp_skb_cb_ext
+ */
+ (void)tcp_skb_cb_ext_alloc(skb);
+ if ((TCP_SKB_CB(skb)->tcp_skb_ext != NULL) &&
+ (sk_wmem_schedule(sk, skb->truesize))) {
skb_reserve(skb, sk->sk_prot->max_header);
/*
* Make sure that we have exactly size bytes
@@ -861,11 +866,14 @@ struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp)
skb->reserved_tailroom = skb->end - skb->tail - size;
return skb;
}
+ if (TCP_SKB_CB(skb)->tcp_skb_ext != NULL)
+ kfree (TCP_SKB_CB(skb)->tcp_skb_ext);
__kfree_skb(skb);
} else {
sk->sk_prot->enter_memory_pressure(sk);
sk_stream_moderate_sndbuf(sk);
}
+
return NULL;
}
@@ -1344,7 +1352,7 @@ do_fault:
* reset, where we can be unlinking the send_head.
*/
tcp_check_send_head(sk, skb);
- sk_wmem_free_skb(sk, skb);
+ tcp_sk_wmem_free_skb(sk, skb);
}
do_error:
@@ -3160,7 +3168,11 @@ void __init tcp_init(void)
int max_rshare, max_wshare, cnt;
unsigned int i;
- sock_skb_cb_check_size(sizeof(struct tcp_skb_cb));
+ /*
+ * copied from the current upstream tree
+ */
+ BUILD_BUG_ON(sizeof(struct tcp_skb_cb) >
+ FIELD_SIZEOF(struct sk_buff, cb));
percpu_counter_init(&tcp_sockets_allocated, 0, GFP_KERNEL);
percpu_counter_init(&tcp_orphan_count, 0, GFP_KERNEL);
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index 16a7256..cb95a7e 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -1354,7 +1354,7 @@ static bool tcp_shifted_skb(struct sock *sk, struct sk_buff *skb,
tcp_advance_highest_sack(sk, skb);
tcp_unlink_write_queue(skb, sk);
- sk_wmem_free_skb(sk, skb);
+ tcp_sk_wmem_free_skb(sk, skb);
NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_SACKMERGED);
@@ -3144,7 +3144,7 @@ static int tcp_clean_rtx_queue(struct sock *sk, int prior_fackets,
break;
tcp_unlink_write_queue(skb, sk);
- sk_wmem_free_skb(sk, skb);
+ tcp_sk_wmem_free_skb(sk, skb);
if (unlikely(skb == tp->retransmit_skb_hint))
tp->retransmit_skb_hint = NULL;
if (unlikely(skb == tp->lost_skb_hint))
@@ -4639,6 +4639,8 @@ out_of_window:
tcp_enter_quickack_mode(sk);
inet_csk_schedule_ack(sk);
drop:
+ if (TCP_SKB_CB(skb)->tcp_skb_ext != NULL)
+ kfree (TCP_SKB_CB(skb)->tcp_skb_ext);
__kfree_skb(skb);
return;
}
@@ -4749,11 +4751,14 @@ restart:
int copy = min_t(int, SKB_MAX_ORDER(0, 0), end - start);
struct sk_buff *nskb;
- nskb = alloc_skb(copy, GFP_ATOMIC);
+ nskb = tcp_alloc_skb(copy, GFP_ATOMIC, SKB_DEFAULT);
if (!nskb)
return;
memcpy(nskb->cb, skb->cb, sizeof(skb->cb));
+ memcpy(TCP_SKB_CB(nskb)->tcp_skb_ext,
+ TCP_SKB_CB(skb)->tcp_skb_ext,
+ sizeof(struct tcp_skb_cb_ext));
TCP_SKB_CB(nskb)->seq = TCP_SKB_CB(nskb)->end_seq = start;
__skb_queue_before(list, skb, nskb);
skb_set_owner_r(nskb, sk);
@@ -5251,6 +5256,8 @@ syn_challenge:
discard:
if (mptcp(tp))
mptcp_reset_mopt(tp);
+ if (TCP_SKB_CB(skb)->tcp_skb_ext != NULL)
+ kfree (TCP_SKB_CB(skb)->tcp_skb_ext);
__kfree_skb(skb);
return false;
}
@@ -5473,8 +5480,11 @@ csum_error:
TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS);
discard:
+ if (TCP_SKB_CB(skb)->tcp_skb_ext != NULL)
+ kfree (TCP_SKB_CB(skb)->tcp_skb_ext);
__kfree_skb(skb);
}
+
EXPORT_SYMBOL(tcp_rcv_established);
void tcp_finish_connect(struct sock *sk, struct sk_buff *skb)
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index 6967a86..c035271 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -694,7 +694,7 @@ void tcp_v4_send_reset(struct sock *sk, struct sk_buff *skb)
arg.tos = ip_hdr(skb)->tos;
ip_send_unicast_reply(*this_cpu_ptr(net->ipv4.tcp_sk),
- skb, &TCP_SKB_CB(skb)->header.h4.opt,
+ skb, &TCP_SKB_CB(skb)->tcp_ipcb_v4.opt,
ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
&arg, arg.iov[0].iov_len);
@@ -798,7 +798,7 @@ static void tcp_v4_send_ack(struct net *net,
arg.bound_dev_if = oif;
arg.tos = tos;
ip_send_unicast_reply(*this_cpu_ptr(net->ipv4.tcp_sk),
- skb, &TCP_SKB_CB(skb)->header.h4.opt,
+ skb, &TCP_SKB_CB(skb)->tcp_ipcb_v4.opt,
ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
&arg, arg.iov[0].iov_len);
@@ -1599,6 +1599,8 @@ int tcp_v4_rcv(struct sk_buff *skb)
int ret;
struct net *net = dev_net(skb->dev);
+ TCP_SKB_CB(skb)->tcp_skb_ext = NULL;
+
if (skb->pkt_type != PACKET_HOST)
goto discard_it;
@@ -1612,6 +1614,7 @@ int tcp_v4_rcv(struct sk_buff *skb)
if (th->doff < sizeof(struct tcphdr) / 4)
goto bad_packet;
+
if (!pskb_may_pull(skb, th->doff * 4))
goto discard_it;
@@ -1625,10 +1628,14 @@ int tcp_v4_rcv(struct sk_buff *skb)
th = tcp_hdr(skb);
iph = ip_hdr(skb);
+
+ if (tcp_skb_cb_ext_alloc(skb) == NULL)
+ goto discard_it;
+
/* This is tricky : We move IPCB at its correct location into TCP_SKB_CB()
* barrier() makes sure compiler wont play fool^Waliasing games.
*/
- memmove(&TCP_SKB_CB(skb)->header.h4, IPCB(skb),
+ memmove(&TCP_SKB_CB(skb)->tcp_ipcb_v4, IPCB(skb),
sizeof(struct inet_skb_parm));
barrier();
@@ -1636,10 +1643,7 @@ int tcp_v4_rcv(struct sk_buff *skb)
TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin +
skb->len - th->doff * 4);
TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
-#ifdef CONFIG_MPTCP
- TCP_SKB_CB(skb)->mptcp_flags = 0;
- TCP_SKB_CB(skb)->dss_off = 0;
-#endif
+
TCP_SKB_CB(skb)->tcp_flags = tcp_flag_byte(th);
TCP_SKB_CB(skb)->tcp_tw_isn = 0;
TCP_SKB_CB(skb)->ip_dsfield = ipv4_get_dsfield(iph);
@@ -1752,6 +1756,10 @@ bad_packet:
discard_it:
/* Discard frame. */
+ if (TCP_SKB_CB(skb)->tcp_skb_ext != NULL) {
+ kfree(TCP_SKB_CB(skb)->tcp_skb_ext);
+ TCP_SKB_CB(skb)->tcp_skb_ext = NULL;
+ }
kfree_skb(skb);
return 0;
diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index a635483..25f8d95 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -540,6 +540,10 @@ static void tcp_options_write(__be32 *ptr, struct tcp_sock *tp,
if (unlikely(OPTION_MPTCP & opts->options))
mptcp_options_write(ptr, tp, opts, skb);
+ /*
+ * Now that we have written the options, NULL out the pointer
+ */
+ TCP_SKB_CB(skb)->tcp_skb_ext = NULL;
}
/* Compute TCP options for SYN packets. This is not the final
@@ -2008,7 +2012,7 @@ static int tcp_mtu_probe(struct sock *sk)
* Throw it away. */
TCP_SKB_CB(nskb)->tcp_flags |= TCP_SKB_CB(skb)->tcp_flags;
tcp_unlink_write_queue(skb, sk);
- sk_wmem_free_skb(sk, skb);
+ tcp_sk_wmem_free_skb(sk, skb);
} else {
TCP_SKB_CB(nskb)->tcp_flags |= TCP_SKB_CB(skb)->tcp_flags &
~(TCPHDR_FIN|TCPHDR_PSH);
@@ -2533,7 +2537,7 @@ static void tcp_collapse_retrans(struct sock *sk, struct sk_buff *skb)
tcp_adjust_pcount(sk, next_skb, tcp_skb_pcount(next_skb));
- sk_wmem_free_skb(sk, next_skb);
+ tcp_sk_wmem_free_skb(sk, next_skb);
}
/* Check if coalescing SKBs is legal. */
@@ -2905,7 +2909,8 @@ coalesce:
return;
}
} else {
- skb = alloc_skb_fclone(MAX_TCP_HEADER, sk->sk_allocation);
+ skb = tcp_alloc_skb(MAX_TCP_HEADER, sk->sk_allocation,
+ SKB_FCLONE);
if (unlikely(!skb)) {
if (tskb)
goto coalesce;
@@ -2931,8 +2936,8 @@ void tcp_send_active_reset(struct sock *sk, gfp_t priority)
struct sk_buff *skb;
/* NOTE: No TCP options attached and we never retransmit this. */
- skb = alloc_skb(MAX_TCP_HEADER, priority);
- if (!skb) {
+ skb = tcp_alloc_skb(MAX_TCP_HEADER, priority, SKB_DEFAULT);
+ if (!skb) {
NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPABORTFAILED);
return;
}
@@ -2972,7 +2977,7 @@ int tcp_send_synack(struct sock *sk)
tcp_unlink_write_queue(skb, sk);
__skb_header_release(nskb);
__tcp_add_write_queue_head(sk, nskb);
- sk_wmem_free_skb(sk, skb);
+ tcp_sk_wmem_free_skb(sk, skb);
sk->sk_wmem_queued += nskb->truesize;
sk_mem_charge(sk, nskb->truesize);
skb = nskb;
@@ -3424,7 +3429,8 @@ void tcp_send_ack(struct sock *sk)
* tcp_transmit_skb() will set the ownership to this
* sock.
*/
- buff = alloc_skb(MAX_TCP_HEADER, sk_gfp_atomic(sk, GFP_ATOMIC));
+ buff = tcp_alloc_skb(MAX_TCP_HEADER, sk_gfp_atomic(sk, GFP_ATOMIC),
+ SKB_DEFAULT);
if (!buff) {
inet_csk_schedule_ack(sk);
inet_csk(sk)->icsk_ack.ato = TCP_ATO_MIN;
@@ -3468,8 +3474,9 @@ int tcp_xmit_probe_skb(struct sock *sk, int urgent)
struct sk_buff *skb;
/* We don't queue it, tcp_transmit_skb() sets ownership. */
- skb = alloc_skb(MAX_TCP_HEADER, sk_gfp_atomic(sk, GFP_ATOMIC));
- if (!skb)
+ skb = tcp_alloc_skb(MAX_TCP_HEADER, sk_gfp_atomic(sk, GFP_ATOMIC),
+ SKB_DEFAULT);
+ if (!skb)
return -1;
/* Reserve space for headers and set control bits. */
diff --git a/net/ipv6/syncookies.c b/net/ipv6/syncookies.c
index faea999..c349380 100644
--- a/net/ipv6/syncookies.c
+++ b/net/ipv6/syncookies.c
@@ -242,7 +242,7 @@ struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb)
ireq->ir_num = ntohs(th->dest);
ireq->ir_v6_rmt_addr = ipv6_hdr(skb)->saddr;
ireq->ir_v6_loc_addr = ipv6_hdr(skb)->daddr;
- if (ipv6_opt_accepted(sk, skb, &TCP_SKB_CB(skb)->header.h6) ||
+ if (ipv6_opt_accepted(sk, skb, &TCP_SKB_CB(skb)->tcp_ipcb_v6) ||
np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
atomic_inc(&skb->users);
diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
index eba2436..0218d6a 100644
--- a/net/ipv6/tcp_ipv6.c
+++ b/net/ipv6/tcp_ipv6.c
@@ -699,7 +699,7 @@ static int tcp_v6_init_req(struct request_sock *req, struct sock *sk,
ireq->ir_iif = tcp_v6_iif(skb);
if (!TCP_SKB_CB(skb)->tcp_tw_isn &&
- (ipv6_opt_accepted(sk, skb, &TCP_SKB_CB(skb)->header.h6) ||
+ (ipv6_opt_accepted(sk, skb, &TCP_SKB_CB(skb)->tcp_ipcb_v6) ||
np->rxopt.bits.rxinfo ||
np->rxopt.bits.rxoinfo || np->rxopt.bits.rxhlim ||
np->rxopt.bits.rxohlim || np->repflow)) {
@@ -771,8 +771,8 @@ static void tcp_v6_send_response(struct sock *sk, struct sk_buff *skb, u32 seq,
if (mptcp)
tot_len += MPTCP_SUB_LEN_DSS + MPTCP_SUB_LEN_ACK;
#endif
- buff = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) + tot_len,
- GFP_ATOMIC);
+ buff = tcp_alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) + tot_len,
+ GFP_ATOMIC, SKB_DEFAULT);
if (!buff)
return;
@@ -1356,6 +1356,11 @@ reset:
discard:
if (opt_skb)
__kfree_skb(opt_skb);
+ if (TCP_SKB_CB(skb)->tcp_skb_ext != NULL) {
+ kfree (TCP_SKB_CB(skb)->tcp_skb_ext);
+ TCP_SKB_CB(skb)->tcp_skb_ext = NULL;
+ }
+
kfree_skb(skb);
return 0;
csum_err:
@@ -1383,7 +1388,7 @@ ipv6_pktoptions:
np->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(opt_skb));
if (np->repflow)
np->flow_label = ip6_flowlabel(ipv6_hdr(opt_skb));
- if (ipv6_opt_accepted(sk, opt_skb, &TCP_SKB_CB(opt_skb)->header.h6)) {
+ if (ipv6_opt_accepted(sk, opt_skb, &TCP_SKB_CB(opt_skb)->tcp_ipcb_v6)) {
skb_set_owner_r(opt_skb, sk);
opt_skb = xchg(&np->pktoptions, opt_skb);
} else {
@@ -1404,7 +1409,7 @@ static void tcp_v6_fill_cb(struct sk_buff *skb, const struct ipv6hdr *hdr,
* _decode_session6() uses IP6CB().
* barrier() makes sure compiler won't play aliasing games.
*/
- memmove(&TCP_SKB_CB(skb)->header.h6, IP6CB(skb),
+ memmove(&TCP_SKB_CB(skb)->tcp_ipcb_v6, IP6CB(skb),
sizeof(struct inet6_skb_parm));
barrier();
@@ -1412,10 +1417,7 @@ static void tcp_v6_fill_cb(struct sk_buff *skb, const struct ipv6hdr *hdr,
TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin +
skb->len - th->doff*4);
TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
-#ifdef CONFIG_MPTCP
- TCP_SKB_CB(skb)->mptcp_flags = 0;
- TCP_SKB_CB(skb)->dss_off = 0;
-#endif
+
TCP_SKB_CB(skb)->tcp_flags = tcp_flag_byte(th);
TCP_SKB_CB(skb)->tcp_tw_isn = 0;
TCP_SKB_CB(skb)->ip_dsfield = ipv6_get_dsfield(hdr);
@@ -1427,7 +1429,7 @@ static void tcp_v6_restore_cb(struct sk_buff *skb)
/* We need to move header back to the beginning if xfrm6_policy_check()
* and tcp_v6_fill_cb() are going to be called again.
*/
- memmove(IP6CB(skb), &TCP_SKB_CB(skb)->header.h6,
+ memmove(IP6CB(skb), &TCP_SKB_CB(skb)->tcp_ipcb_v6,
sizeof(struct inet6_skb_parm));
}
@@ -1439,6 +1441,8 @@ static int tcp_v6_rcv(struct sk_buff *skb)
int ret;
struct net *net = dev_net(skb->dev);
+ TCP_SKB_CB(skb)->tcp_skb_ext = NULL;
+
if (skb->pkt_type != PACKET_HOST)
goto discard_it;
@@ -1454,6 +1458,7 @@ static int tcp_v6_rcv(struct sk_buff *skb)
if (th->doff < sizeof(struct tcphdr)/4)
goto bad_packet;
+
if (!pskb_may_pull(skb, th->doff*4))
goto discard_it;
@@ -1467,7 +1472,6 @@ static int tcp_v6_rcv(struct sk_buff *skb)
inet6_iif(skb));
if (!sk)
goto no_tcp_socket;
-
process:
if (sk && sk->sk_state == TCP_TIME_WAIT)
goto do_time_wait;
@@ -1483,6 +1487,9 @@ process:
if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
goto discard_and_relse;
+ if (tcp_skb_cb_ext_alloc(skb) == NULL)
+ goto discard_and_relse;
+
tcp_v6_fill_cb(skb, hdr, th);
#ifdef CONFIG_MPTCP
@@ -1536,6 +1543,9 @@ no_tcp_socket:
if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
goto discard_it;
+ if (tcp_skb_cb_ext_alloc(skb) == NULL)
+ goto discard_it;
+
tcp_v6_fill_cb(skb, hdr, th);
#ifdef CONFIG_MPTCP
@@ -1568,6 +1578,10 @@ bad_packet:
}
discard_it:
+ if (TCP_SKB_CB(skb)->tcp_skb_ext != NULL) {
+ kfree(TCP_SKB_CB(skb)->tcp_skb_ext);
+ TCP_SKB_CB(skb)->tcp_skb_ext = NULL;
+ }
kfree_skb(skb);
return 0;
diff --git a/net/mptcp/mptcp_ctrl.c b/net/mptcp/mptcp_ctrl.c
index 28f348f..0730c26 100644
--- a/net/mptcp/mptcp_ctrl.c
+++ b/net/mptcp/mptcp_ctrl.c
@@ -1459,7 +1459,7 @@ static int mptcp_sub_send_fin(struct sock *sk)
TCP_SKB_CB(skb)->end_seq++;
tp->write_seq++;
} else {
- skb = alloc_skb_fclone(MAX_TCP_HEADER, GFP_ATOMIC);
+ skb = tcp_alloc_skb(MAX_TCP_HEADER, GFP_ATOMIC, SKB_FCLONE);
if (!skb)
return 1;
diff --git a/net/mptcp/mptcp_input.c b/net/mptcp/mptcp_input.c
index 51cbb06..60aa548 100644
--- a/net/mptcp/mptcp_input.c
+++ b/net/mptcp/mptcp_input.c
@@ -146,7 +146,7 @@ static void mptcp_clean_rtx_queue(struct sock *meta_sk, u32 prior_snd_una)
mptcp_sub_close(sk_it, delay);
}
}
- sk_wmem_free_skb(meta_sk, skb);
+ tcp_sk_wmem_free_skb(meta_sk, skb);
}
/* Remove acknowledged data from the reinject queue */
skb_queue_walk_safe(&mpcb->reinject_queue, skb, tmp) {
@@ -494,7 +494,7 @@ static int mptcp_skb_split_tail(struct sk_buff *skb, struct sock *sk, u32 seq)
nsize = 0;
/* Get a new skb... force flag on. */
- buff = alloc_skb(nsize, GFP_ATOMIC);
+ buff = tcp_alloc_skb(nsize, GFP_ATOMIC, SKB_DEFAULT);
if (buff == NULL)
return -ENOMEM;
@@ -542,6 +542,8 @@ static int mptcp_prevalidate_skb(struct sock *sk, struct sk_buff *skb)
*/
tp->copied_seq = TCP_SKB_CB(skb)->end_seq;
__skb_unlink(skb, &sk->sk_receive_queue);
+ if (TCP_SKB_CB(skb)->tcp_skb_ext != NULL)
+ kfree (TCP_SKB_CB(skb)->tcp_skb_ext);
__kfree_skb(skb);
return -1;
}
@@ -706,6 +708,8 @@ static int mptcp_detect_mapping(struct sock *sk, struct sk_buff *skb)
if (tp->mptcp->send_mp_fail) {
tp->copied_seq = TCP_SKB_CB(skb)->end_seq;
__skb_unlink(skb, &sk->sk_receive_queue);
+ if (TCP_SKB_CB(skb)->tcp_skb_ext != NULL)
+ kfree (TCP_SKB_CB(skb)->tcp_skb_ext);
__kfree_skb(skb);
return -1;
}
@@ -844,6 +848,8 @@ static int mptcp_validate_mapping(struct sock *sk, struct sk_buff *skb)
*/
tp->copied_seq = TCP_SKB_CB(skb)->end_seq;
__skb_unlink(skb, &sk->sk_receive_queue);
+ if (TCP_SKB_CB(skb)->tcp_skb_ext != NULL)
+ kfree (TCP_SKB_CB(skb)->tcp_skb_ext);
__kfree_skb(skb);
return -1;
}
@@ -865,6 +871,8 @@ static int mptcp_validate_mapping(struct sock *sk, struct sk_buff *skb)
/* Impossible that we could free skb here, because his
* mapping is known to be valid from previous checks
*/
+ if (TCP_SKB_CB(skb)->tcp_skb_ext != NULL)
+ kfree (TCP_SKB_CB(skb)->tcp_skb_ext);
__kfree_skb(tmp1);
}
}
@@ -1026,6 +1034,7 @@ void mptcp_data_ready(struct sock *sk)
struct sk_buff *skb, *tmp;
int queued = 0;
+
/* restart before the check, because mptcp_fin might have changed the
* state.
*/
@@ -1045,6 +1054,7 @@ restart:
*/
skb_queue_walk_safe(&sk->sk_receive_queue, skb, tmp) {
int ret;
+
/* Pre-validation - e.g., early fallback */
ret = mptcp_prevalidate_skb(sk, skb);
if (ret < 0)
@@ -1108,6 +1118,11 @@ int mptcp_check_req(struct sk_buff *skb, struct net *net)
if (!meta_sk)
return 0;
+ if (tcp_skb_cb_ext_alloc(skb) == NULL) {
+ kfree_skb(skb);
+ return 1;
+ }
+
TCP_SKB_CB(skb)->mptcp_flags |= MPTCPHDR_JOIN;
bh_lock_sock_nested(meta_sk);
@@ -1222,6 +1237,9 @@ int mptcp_lookup_join(struct sk_buff *skb, struct inet_timewait_sock *tw)
inet_twsk_put(tw);
}
+ if (tcp_skb_cb_ext_alloc(skb) == NULL)
+ return (-1);
+
TCP_SKB_CB(skb)->mptcp_flags |= MPTCPHDR_JOIN;
/* OK, this is a new syn/join, let's create a new open request and
* send syn+ack
@@ -1481,7 +1499,9 @@ static void mptcp_data_ack(struct sock *sk, const struct sk_buff *skb)
/* If we are in infinite mapping mode, rx_opt.data_ack has been
* set by mptcp_clean_rtx_infinite.
*/
- if (!(tcb->mptcp_flags & MPTCPHDR_ACK) && !tp->mpcb->infinite_mapping_snd)
+ if ((TCP_SKB_CB(skb)->tcp_skb_ext != NULL) &&
+ !(tcb->mptcp_flags & MPTCPHDR_ACK) &&
+ !tp->mpcb->infinite_mapping_snd)
goto exit;
data_ack = tp->mptcp->rx_opt.data_ack;
@@ -1652,6 +1672,14 @@ void mptcp_parse_options(const uint8_t *ptr, int opsize,
{
const struct mptcp_option *mp_opt = (struct mptcp_option *)ptr;
+ /*
+ * sometimes it may not be necessary to allocate tcp_skb_cb_ext
+ * fix mptcp code to dela with that
+ */
+ if (tcp_skb_cb_ext_alloc((struct sk_buff *)skb) == NULL) {
+ return;
+ }
+
/* If the socket is mp-capable we would have a mopt. */
if (!mopt)
return;
@@ -1754,6 +1782,7 @@ void mptcp_parse_options(const uint8_t *ptr, int opsize,
break;
}
+
ptr += 4;
if (mdss->A) {
@@ -1908,7 +1937,7 @@ void tcp_parse_mptcp_options(const struct sk_buff *skb,
if (opsize > length)
return; /* don't parse partial options */
if (opcode == TCPOPT_MPTCP)
- mptcp_parse_options(ptr - 2, opsize, mopt, skb, NULL);
+ mptcp_parse_options(ptr - 2, opsize, mopt, (struct sk_buff *) skb, NULL);
}
ptr += opsize - 2;
length -= opsize;
diff --git a/net/mptcp/mptcp_ipv4.c b/net/mptcp/mptcp_ipv4.c
index a147b20..b96e5a0 100644
--- a/net/mptcp/mptcp_ipv4.c
+++ b/net/mptcp/mptcp_ipv4.c
@@ -186,7 +186,11 @@ int mptcp_v4_do_rcv(struct sock *meta_sk, struct sk_buff *skb)
struct sock *child, *rsk = NULL;
int ret;
- if (!(TCP_SKB_CB(skb)->mptcp_flags & MPTCPHDR_JOIN)) {
+ /*
+ * If tcp_skb_ext == NULL this is not a JOIN request
+ */
+ if ((TCP_SKB_CB(skb)->tcp_skb_ext == NULL) ||
+ !(TCP_SKB_CB(skb)->mptcp_flags & MPTCPHDR_JOIN)) {
struct tcphdr *th = tcp_hdr(skb);
const struct iphdr *iph = ip_hdr(skb);
struct sock *sk;
diff --git a/net/mptcp/mptcp_ipv6.c b/net/mptcp/mptcp_ipv6.c
index 0de953d..475c6cb 100644
--- a/net/mptcp/mptcp_ipv6.c
+++ b/net/mptcp/mptcp_ipv6.c
@@ -199,7 +199,8 @@ int mptcp_v6_do_rcv(struct sock *meta_sk, struct sk_buff *skb)
struct sock *child, *rsk = NULL;
int ret;
- if (!(TCP_SKB_CB(skb)->mptcp_flags & MPTCPHDR_JOIN)) {
+ if ((TCP_SKB_CB(skb)->tcp_skb_ext == NULL) ||
+ !(TCP_SKB_CB(skb)->mptcp_flags & MPTCPHDR_JOIN)) {
struct tcphdr *th = tcp_hdr(skb);
const struct ipv6hdr *ip6h = ipv6_hdr(skb);
struct sock *sk;
diff --git a/net/mptcp/mptcp_output.c b/net/mptcp/mptcp_output.c
index 691ef6f..de9ae81 100644
--- a/net/mptcp/mptcp_output.c
+++ b/net/mptcp/mptcp_output.c
@@ -413,15 +413,8 @@ static int mptcp_write_dss_data_seq(const struct tcp_sock *tp, struct sk_buff *s
__be32 *start = ptr;
memcpy(ptr, TCP_SKB_CB(skb)->dss, mptcp_dss_len);
-
/* update the data_ack */
start[1] = htonl(mptcp_meta_tp(tp)->rcv_nxt);
-
- /* dss is in a union with inet_skb_parm and
- * the IP layer expects zeroed IPCB fields.
- */
- memset(TCP_SKB_CB(skb)->dss, 0 , mptcp_dss_len);
-
return mptcp_dss_len/sizeof(*ptr);
}
@@ -441,6 +434,14 @@ static bool mptcp_skb_entail(struct sock *sk, struct sk_buff *skb, int reinject)
if (!subskb)
return false;
+ TCP_SKB_CB(subskb)->tcp_skb_ext = NULL;
+ if (tcp_skb_cb_ext_alloc(subskb) == NULL) {
+ kfree_skb(subskb);
+ return false;
+ }
+
+ *(TCP_SKB_CB(subskb)->tcp_skb_ext) = *(TCP_SKB_CB(skb)->tcp_skb_ext);
+
/* At the subflow-level we need to call again tcp_init_tso_segs. We
* force this, by setting pcount to 0. It has been set to 1 prior to
* the call to mptcp_skb_entail.
@@ -495,6 +496,7 @@ static bool mptcp_skb_entail(struct sock *sk, struct sk_buff *skb, int reinject)
err = tcp_transmit_skb(sk, subskb, 1, GFP_ATOMIC);
/* It has not been queued, we can free it now. */
+ kfree (TCP_SKB_CB(subskb)->tcp_skb_ext);
kfree_skb(subskb);
if (err)
@@ -660,7 +662,7 @@ bool mptcp_write_xmit(struct sock *meta_sk, unsigned int mss_now, int nonagle,
struct sk_buff *skb;
int reinject = 0;
unsigned int sublimit;
- __u32 path_mask = 0;
+ __u32 __path_mask = 0;
while ((skb = mpcb->sched_ops->next_segment(meta_sk, &reinject, &subsk,
&sublimit))) {
@@ -746,7 +748,7 @@ bool mptcp_write_xmit(struct sock *meta_sk, unsigned int mss_now, int nonagle,
__tcp_push_pending_frames(subsk, mss_now, TCP_NAGLE_PUSH);
meta_tp->lsndtime = tcp_time_stamp;
- path_mask |= mptcp_pi_to_flag(subtp->mptcp->path_index);
+ __path_mask |= mptcp_pi_to_flag(subtp->mptcp->path_index);
skb_mstamp_get(&skb->skb_mstamp);
if (!reinject) {
@@ -770,7 +772,7 @@ bool mptcp_write_xmit(struct sock *meta_sk, unsigned int mss_now, int nonagle,
mptcp_for_each_sk(mpcb, subsk) {
subtp = tcp_sk(subsk);
- if (!(path_mask & mptcp_pi_to_flag(subtp->mptcp->path_index)))
+ if (!(__path_mask & mptcp_pi_to_flag(subtp->mptcp->path_index)))
continue;
/* We have pushed data on this subflow. We ignore the call to
@@ -1239,8 +1241,9 @@ void mptcp_send_fin(struct sock *meta_sk)
} else {
/* Socket is locked, keep trying until memory is available. */
for (;;) {
- skb = alloc_skb_fclone(MAX_TCP_HEADER,
- meta_sk->sk_allocation);
+ skb = tcp_alloc_skb(MAX_TCP_HEADER,
+ meta_sk->sk_allocation,
+ SKB_FCLONE);
if (skb)
break;
yield();
@@ -1331,7 +1334,7 @@ static void mptcp_ack_retransmit_timer(struct sock *sk)
goto out;
}
- skb = alloc_skb(MAX_TCP_HEADER, GFP_ATOMIC);
+ skb = tcp_alloc_skb(MAX_TCP_HEADER, GFP_ATOMIC, SKB_DEFAULT);
if (skb == NULL) {
sk_reset_timer(sk, &tp->mptcp->mptcp_ack_timer,
jiffies + icsk->icsk_rto);
--
2.7.4
3 years, 10 months
- 2
- 5
Separating MPTCP packet processing from TCP
by Rao Shoaib
Hi,
I am thinking the cleanest implementation would be to separate MPTCP
processing from TCP. On the receive side, when a packet arrives TCP only
does TCP processing (That leaves the code untouched), the packet is than
passed upto MPTCP which does MPTCP processing and can either process the
packet, drop it or send a reset.
The current implementation is doing receive processing in TCP because it
wants to validate the packet and accept it in TCP or not -- but why ?
that seems to be an implementation choice.
In the case where the receiver drops the packet in MPTCP, no data ack
will be sent and MPTCP will re-transmit, It can retransmit even on the
same flow. To achieve this the code requires some change as the DSS
option has to be saved. I think this is doable and is a much cleaner
solution.
Similarly we need to think about the Tx side -- That is not so straight
forward but we need to think harder.
I can work on making the Rx changes but want to discuss it in case I am
missing potential issues and if this is not a got option.
Comments ?
Rao
3 years, 11 months
- 3
- 5
Change to remove MPTCP members from tcp_skb_cb
by Rao Shoaib
Hi,
Here is an attempt to not use tcp_skb_cb with minimal change to MPTCP
code and skb code. I have tested the change by using wget from
multipath-tcp.org. pcap file is attached.
The approach that I have taken is very straight forward. On Tx when an
skb is allocated and the underlying socket is MPTCP , 32 bytes extra are
allocated[1] and reserved as the head room. This space is used for
passing what was being passed in tcp_skb_cb. An used bit in sk_buff is
used to indicate this special case and when skb is copied the extra data
is also copied. clone is not an issue as the data is shared. This work
for retransmission also because tcp_transmit_skb clones the original skb.
On the Rx side, options are parsed in tcp at that point we can reuse the
space that was used for ether and IP headers and no extra space is
required.
The space used is always right below the skb head pointer because the
location of head never changes.
Please take a look and let me know what I have missed.
Regards,
Rao
3 years, 11 months
- 3
- 11
[PATCH] Revert tcp_skb_cb to it's original size
by Rao Shoaib
This is an attempt to address the issue without allocating any space and overloading existing data structures.
Shoaib.
Signed-off-by: Rao Shoaib <rao.shoaib(a)oracle.com>
---
include/linux/skbuff.h | 8 ++-
include/net/mptcp.h | 17 +++---
include/net/tcp.h | 19 +++---
net/ipv4/tcp_ipv4.c | 4 +-
net/ipv4/tcp_output.c | 13 ++++-
net/ipv6/tcp_ipv6.c | 4 +-
net/mptcp/mptcp_input.c | 37 ++++++------
net/mptcp/mptcp_ipv4.c | 4 +-
net/mptcp/mptcp_ipv6.c | 4 +-
net/mptcp/mptcp_output.c | 146 +++++++++++++++++------------------------------
10 files changed, 113 insertions(+), 143 deletions(-)
diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index f66cd5e..e2cdc21 100644
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
@@ -532,7 +532,13 @@ struct sk_buff {
struct rb_node rbnode; /* used in netem & tcp stack */
};
struct sock *sk;
- struct net_device *dev;
+ union {
+ struct net_device *dev;
+ struct {
+ __u8 mptcp_flags;
+ __u8 dss_offset;
+ };
+ };
/*
* This is the control buffer. It is free to use for every
diff --git a/include/net/mptcp.h b/include/net/mptcp.h
index 876f1e6..d9ff2e7 100644
--- a/include/net/mptcp.h
+++ b/include/net/mptcp.h
@@ -807,8 +807,7 @@ void tcp_parse_mptcp_options(const struct sk_buff *skb,
struct mptcp_options_received *mopt);
void mptcp_parse_options(const uint8_t *ptr, int opsize,
struct mptcp_options_received *mopt,
- const struct sk_buff *skb,
- struct tcp_sock *tp);
+ struct sk_buff *skb, struct tcp_sock *tp);
void mptcp_syn_options(const struct sock *sk, struct tcp_out_options *opts,
unsigned *remaining);
void mptcp_synack_options(struct request_sock *req,
@@ -1000,12 +999,12 @@ static inline void mptcp_sub_force_close_all(struct mptcp_cb *mpcb,
static inline bool mptcp_is_data_seq(const struct sk_buff *skb)
{
- return TCP_SKB_CB(skb)->mptcp_flags & MPTCPHDR_SEQ;
+ return skb->mptcp_flags & MPTCPHDR_SEQ;
}
static inline bool mptcp_is_data_fin(const struct sk_buff *skb)
{
- return TCP_SKB_CB(skb)->mptcp_flags & MPTCPHDR_FIN;
+ return skb->mptcp_flags & MPTCPHDR_FIN;
}
/* Is it a data-fin while in infinite mapping mode?
@@ -1034,17 +1033,17 @@ static inline u8 mptcp_get_64_bit(u64 data_seq, struct mptcp_cb *mpcb)
/* Sets the data_seq and returns pointer to the in-skb field of the data_seq.
* If the packet has a 64-bit dseq, the pointer points to the last 32 bits.
*/
-static inline __u32 *mptcp_skb_set_data_seq(const struct sk_buff *skb,
+static inline __u32 *mptcp_skb_set_data_seq(struct sk_buff *skb,
u32 *data_seq,
struct mptcp_cb *mpcb)
{
- __u32 *ptr = (__u32 *)(skb_transport_header(skb) + TCP_SKB_CB(skb)->dss_off);
+ __u32 *ptr = (__u32 *)(skb_transport_header(skb) + skb->dss_offset);
- if (TCP_SKB_CB(skb)->mptcp_flags & MPTCPHDR_SEQ64_SET) {
+ if (skb->mptcp_flags & MPTCPHDR_SEQ64_SET) {
u64 data_seq64 = get_unaligned_be64(ptr);
if (mpcb)
- TCP_SKB_CB(skb)->mptcp_flags |= mptcp_get_64_bit(data_seq64, mpcb);
+ skb->mptcp_flags |= mptcp_get_64_bit(data_seq64, mpcb);
*data_seq = (u32)data_seq64;
ptr++;
@@ -1142,7 +1141,7 @@ static inline void mptcp_reset_mopt(struct tcp_sock *tp)
static inline __be32 mptcp_get_highorder_sndbits(const struct sk_buff *skb,
const struct mptcp_cb *mpcb)
{
- return htonl(mpcb->snd_high_order[(TCP_SKB_CB(skb)->mptcp_flags &
+ return htonl(mpcb->snd_high_order[(skb->mptcp_flags &
MPTCPHDR_SEQ64_INDEX) ? 1 : 0]);
}
diff --git a/include/net/tcp.h b/include/net/tcp.h
index 655ecd4..3258721 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -842,11 +842,6 @@ struct tcp_skb_cb {
__u32 tcp_gso_segs;
};
-#ifdef CONFIG_MPTCP
- __u8 mptcp_flags; /* flags for the MPTCP layer */
- __u8 dss_off; /* Number of 4-byte words until
- * seq-number */
-#endif
__u8 tcp_flags; /* TCP header flags. (tcp[13]) */
__u8 sacked; /* State flags for SACK/FACK. */
@@ -861,7 +856,13 @@ struct tcp_skb_cb {
__u8 ip_dsfield; /* IPv4 tos or IPv6 dsfield */
/* 1 byte hole */
- __u32 ack_seq; /* Sequence number ACK'd */
+ union {
+ __u32 ack_seq; /* Sequence number ACK'd */
+ union {
+ __u32 mptcp_data_seq;
+ __u32 path_mask;
+ };
+ };
union {
union {
struct inet_skb_parm h4;
@@ -869,12 +870,6 @@ struct tcp_skb_cb {
struct inet6_skb_parm h6;
#endif
} header; /* For incoming frames */
-#ifdef CONFIG_MPTCP
- union { /* For MPTCP outgoing frames */
- __u32 path_mask; /* paths that tried to send this skb */
- __u32 dss[6]; /* DSS options */
- };
-#endif
};
};
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index 6967a86..4e8a3e3 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -1637,8 +1637,8 @@ int tcp_v4_rcv(struct sk_buff *skb)
skb->len - th->doff * 4);
TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
#ifdef CONFIG_MPTCP
- TCP_SKB_CB(skb)->mptcp_flags = 0;
- TCP_SKB_CB(skb)->dss_off = 0;
+ skb->mptcp_flags = 0;
+ skb->dss_offset = 0;
#endif
TCP_SKB_CB(skb)->tcp_flags = tcp_flag_byte(th);
TCP_SKB_CB(skb)->tcp_tw_isn = 0;
diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index a635483..9d49125 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -999,7 +999,6 @@ int tcp_transmit_skb(struct sock *sk, struct sk_buff *skb, int clone_it,
th->source = inet->inet_sport;
th->dest = inet->inet_dport;
th->seq = htonl(tcb->seq);
- th->ack_seq = htonl(tp->rcv_nxt);
*(((__be16 *)th) + 6) = htons(((tcp_header_size >> 2) << 12) |
tcb->tcp_flags);
@@ -1029,6 +1028,12 @@ int tcp_transmit_skb(struct sock *sk, struct sk_buff *skb, int clone_it,
if (likely((tcb->tcp_flags & TCPHDR_SYN) == 0))
tcp_ecn_send(sk, skb, tcp_header_size);
+ /*
+ * It is important that we initialize ack_seq after options
+ * have been added as ack_seq is used to pass options value to MPTCP
+ */
+ th->ack_seq = htonl(tp->rcv_nxt);
+
#ifdef CONFIG_TCP_MD5SIG
/* Calculate the MD5 hash, as we have all we need now */
if (md5) {
@@ -1216,6 +1221,12 @@ int tcp_fragment(struct sock *sk, struct sk_buff *skb, u32 len,
TCP_SKB_CB(skb)->tcp_flags = flags & ~(TCPHDR_FIN | TCPHDR_PSH);
TCP_SKB_CB(buff)->tcp_flags = flags;
TCP_SKB_CB(buff)->sacked = TCP_SKB_CB(skb)->sacked;
+ if (mptcp(tp)) {
+ /*
+ * update the mapping
+ */
+ TCP_SKB_CB(buff)->mptcp_data_seq += len;
+ }
if (!skb_shinfo(skb)->nr_frags && skb->ip_summed != CHECKSUM_PARTIAL) {
/* Copy and checksum data tail into the new buffer. */
diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
index eba2436..e0c6132 100644
--- a/net/ipv6/tcp_ipv6.c
+++ b/net/ipv6/tcp_ipv6.c
@@ -1413,8 +1413,8 @@ static void tcp_v6_fill_cb(struct sk_buff *skb, const struct ipv6hdr *hdr,
skb->len - th->doff*4);
TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
#ifdef CONFIG_MPTCP
- TCP_SKB_CB(skb)->mptcp_flags = 0;
- TCP_SKB_CB(skb)->dss_off = 0;
+ skb->mptcp_flags = 0;
+ skb->dss_offset = 0;
#endif
TCP_SKB_CB(skb)->tcp_flags = tcp_flag_byte(th);
TCP_SKB_CB(skb)->tcp_tw_isn = 0;
diff --git a/net/mptcp/mptcp_input.c b/net/mptcp/mptcp_input.c
index 51cbb06..3c267ad 100644
--- a/net/mptcp/mptcp_input.c
+++ b/net/mptcp/mptcp_input.c
@@ -324,8 +324,8 @@ static int mptcp_verif_dss_csum(struct sock *sk)
* in the final csum_partial-call.
*/
u32 offset = skb_transport_offset(tmp) +
- TCP_SKB_CB(tmp)->dss_off;
- if (TCP_SKB_CB(tmp)->mptcp_flags & MPTCPHDR_SEQ64_SET)
+ tmp->dss_offset;
+ if (tmp->mptcp_flags & MPTCPHDR_SEQ64_SET)
offset += 4;
csum_tcp = skb_checksum(tmp, offset,
@@ -758,7 +758,7 @@ static int mptcp_detect_mapping(struct sock *sk, struct sk_buff *skb)
}
/* Does the DSS had 64-bit seqnum's ? */
- if (!(tcb->mptcp_flags & MPTCPHDR_SEQ64_SET)) {
+ if (!(skb->mptcp_flags & MPTCPHDR_SEQ64_SET)) {
/* Wrapped around? */
if (unlikely(after(data_seq, meta_tp->rcv_nxt) && data_seq < meta_tp->rcv_nxt)) {
tp->mptcp->map_data_seq = mptcp_get_data_seq_64(mpcb, !mpcb->rcv_hiseq_index, data_seq);
@@ -767,9 +767,9 @@ static int mptcp_detect_mapping(struct sock *sk, struct sk_buff *skb)
tp->mptcp->map_data_seq = mptcp_get_data_seq_64(mpcb, mpcb->rcv_hiseq_index, data_seq);
}
} else {
- tp->mptcp->map_data_seq = mptcp_get_data_seq_64(mpcb, (tcb->mptcp_flags & MPTCPHDR_SEQ64_INDEX) ? 1 : 0, data_seq);
+ tp->mptcp->map_data_seq = mptcp_get_data_seq_64(mpcb, (skb->mptcp_flags & MPTCPHDR_SEQ64_INDEX) ? 1 : 0, data_seq);
- if (unlikely(tcb->mptcp_flags & MPTCPHDR_SEQ64_OFO)) {
+ if (unlikely(skb->mptcp_flags & MPTCPHDR_SEQ64_OFO)) {
/* We make sure that the data_seq is invalid.
* It will be dropped later.
*/
@@ -1108,7 +1108,7 @@ int mptcp_check_req(struct sk_buff *skb, struct net *net)
if (!meta_sk)
return 0;
- TCP_SKB_CB(skb)->mptcp_flags |= MPTCPHDR_JOIN;
+ skb->mptcp_flags |= MPTCPHDR_JOIN;
bh_lock_sock_nested(meta_sk);
if (sock_owned_by_user(meta_sk)) {
@@ -1222,7 +1222,7 @@ int mptcp_lookup_join(struct sk_buff *skb, struct inet_timewait_sock *tw)
inet_twsk_put(tw);
}
- TCP_SKB_CB(skb)->mptcp_flags |= MPTCPHDR_JOIN;
+ skb->mptcp_flags |= MPTCPHDR_JOIN;
/* OK, this is a new syn/join, let's create a new open request and
* send syn+ack
*/
@@ -1279,7 +1279,7 @@ int mptcp_do_join_short(struct sk_buff *skb,
return -1;
}
- TCP_SKB_CB(skb)->mptcp_flags |= MPTCPHDR_JOIN;
+ skb->mptcp_flags |= MPTCPHDR_JOIN;
/* OK, this is a new syn/join, let's create a new open request and
* send syn+ack
@@ -1461,7 +1461,7 @@ static void mptcp_data_ack(struct sock *sk, const struct sk_buff *skb)
{
struct sock *meta_sk = mptcp_meta_sk(sk);
struct tcp_sock *meta_tp = tcp_sk(meta_sk), *tp = tcp_sk(sk);
- struct tcp_skb_cb *tcb = TCP_SKB_CB(skb);
+ //struct tcp_skb_cb *tcb = TCP_SKB_CB(skb);
u32 prior_snd_una = meta_tp->snd_una;
int prior_packets;
u32 nwin, data_ack, data_seq;
@@ -1481,7 +1481,7 @@ static void mptcp_data_ack(struct sock *sk, const struct sk_buff *skb)
/* If we are in infinite mapping mode, rx_opt.data_ack has been
* set by mptcp_clean_rtx_infinite.
*/
- if (!(tcb->mptcp_flags & MPTCPHDR_ACK) && !tp->mpcb->infinite_mapping_snd)
+ if (!(skb->mptcp_flags & MPTCPHDR_ACK) && !tp->mpcb->infinite_mapping_snd)
goto exit;
data_ack = tp->mptcp->rx_opt.data_ack;
@@ -1647,7 +1647,7 @@ static inline bool is_valid_addropt_opsize(u8 mptcp_ver,
void mptcp_parse_options(const uint8_t *ptr, int opsize,
struct mptcp_options_received *mopt,
- const struct sk_buff *skb,
+ struct sk_buff *skb,
struct tcp_sock *tp)
{
const struct mptcp_option *mp_opt = (struct mptcp_option *)ptr;
@@ -1757,7 +1757,7 @@ void mptcp_parse_options(const uint8_t *ptr, int opsize,
ptr += 4;
if (mdss->A) {
- tcb->mptcp_flags |= MPTCPHDR_ACK;
+ skb->mptcp_flags |= MPTCPHDR_ACK;
if (mdss->a) {
mopt->data_ack = (u32) get_unaligned_be64(ptr);
@@ -1768,13 +1768,14 @@ void mptcp_parse_options(const uint8_t *ptr, int opsize,
}
}
- tcb->dss_off = (ptr - skb_transport_header(skb));
+ //tcb->dss_off = (ptr - skb_transport_header(skb));
+ skb->dss_offset = (ptr - skb_transport_header(skb));
if (mdss->M) {
if (mdss->m) {
u64 data_seq64 = get_unaligned_be64(ptr);
- tcb->mptcp_flags |= MPTCPHDR_SEQ64_SET;
+ skb->mptcp_flags |= MPTCPHDR_SEQ64_SET;
mopt->data_seq = (u32) data_seq64;
ptr += 12; /* 64-bit dseq + subseq */
@@ -1784,15 +1785,15 @@ void mptcp_parse_options(const uint8_t *ptr, int opsize,
}
mopt->data_len = get_unaligned_be16(ptr);
- tcb->mptcp_flags |= MPTCPHDR_SEQ;
+ skb->mptcp_flags |= MPTCPHDR_SEQ;
/* Is a check-sum present? */
if (opsize == mptcp_sub_len_dss(mdss, 1))
- tcb->mptcp_flags |= MPTCPHDR_DSS_CSUM;
+ skb->mptcp_flags |= MPTCPHDR_DSS_CSUM;
/* DATA_FIN only possible with DSS-mapping */
if (mdss->F)
- tcb->mptcp_flags |= MPTCPHDR_FIN;
+ skb->mptcp_flags |= MPTCPHDR_FIN;
}
break;
@@ -2185,7 +2186,7 @@ bool mptcp_handle_options(struct sock *sk, const struct tcphdr *th,
* receiver MUST close the subflow with a RST as it is considered broken.
*/
if (mptcp_is_data_seq(skb) && tp->mpcb->dss_csum &&
- !(TCP_SKB_CB(skb)->mptcp_flags & MPTCPHDR_DSS_CSUM)) {
+ !(skb->mptcp_flags & MPTCPHDR_DSS_CSUM)) {
mptcp_send_reset(sk);
return true;
}
diff --git a/net/mptcp/mptcp_ipv4.c b/net/mptcp/mptcp_ipv4.c
index a147b20..1614836 100644
--- a/net/mptcp/mptcp_ipv4.c
+++ b/net/mptcp/mptcp_ipv4.c
@@ -186,7 +186,7 @@ int mptcp_v4_do_rcv(struct sock *meta_sk, struct sk_buff *skb)
struct sock *child, *rsk = NULL;
int ret;
- if (!(TCP_SKB_CB(skb)->mptcp_flags & MPTCPHDR_JOIN)) {
+ if (!(skb->mptcp_flags & MPTCPHDR_JOIN)) {
struct tcphdr *th = tcp_hdr(skb);
const struct iphdr *iph = ip_hdr(skb);
struct sock *sk;
@@ -217,7 +217,7 @@ int mptcp_v4_do_rcv(struct sock *meta_sk, struct sk_buff *skb)
return ret;
}
- TCP_SKB_CB(skb)->mptcp_flags = 0;
+ skb->mptcp_flags = 0;
/* Has been removed from the tk-table. Thus, no new subflows.
*
diff --git a/net/mptcp/mptcp_ipv6.c b/net/mptcp/mptcp_ipv6.c
index 0de953d..1959d01 100644
--- a/net/mptcp/mptcp_ipv6.c
+++ b/net/mptcp/mptcp_ipv6.c
@@ -199,7 +199,7 @@ int mptcp_v6_do_rcv(struct sock *meta_sk, struct sk_buff *skb)
struct sock *child, *rsk = NULL;
int ret;
- if (!(TCP_SKB_CB(skb)->mptcp_flags & MPTCPHDR_JOIN)) {
+ if (!(skb->mptcp_flags & MPTCPHDR_JOIN)) {
struct tcphdr *th = tcp_hdr(skb);
const struct ipv6hdr *ip6h = ipv6_hdr(skb);
struct sock *sk;
@@ -232,7 +232,7 @@ int mptcp_v6_do_rcv(struct sock *meta_sk, struct sk_buff *skb)
return ret;
}
- TCP_SKB_CB(skb)->mptcp_flags = 0;
+ skb->mptcp_flags = 0;
/* Has been removed from the tk-table. Thus, no new subflows.
*
diff --git a/net/mptcp/mptcp_output.c b/net/mptcp/mptcp_output.c
index 691ef6f..e746565 100644
--- a/net/mptcp/mptcp_output.c
+++ b/net/mptcp/mptcp_output.c
@@ -59,39 +59,17 @@ EXPORT_SYMBOL(mptcp_sub_len_remove_addr_align);
*/
static bool mptcp_reconstruct_mapping(struct sk_buff *skb)
{
- const struct mp_dss *mpdss = (struct mp_dss *)TCP_SKB_CB(skb)->dss;
- u32 *p32;
- u16 *p16;
-
if (!mptcp_is_data_seq(skb))
return false;
- if (!mpdss->M)
- return false;
-
- /* Move the pointer to the data-seq */
- p32 = (u32 *)mpdss;
- p32++;
- if (mpdss->A) {
- p32++;
- if (mpdss->a)
- p32++;
- }
-
- TCP_SKB_CB(skb)->seq = ntohl(*p32);
-
- /* Get the data_len to calculate the end_data_seq */
- p32++;
- p32++;
- p16 = (u16 *)p32;
- TCP_SKB_CB(skb)->end_seq = ntohs(*p16) + TCP_SKB_CB(skb)->seq;
+ TCP_SKB_CB(skb)->seq = TCP_SKB_CB(skb)->mptcp_data_seq;
return true;
}
static bool mptcp_is_reinjected(const struct sk_buff *skb)
{
- return TCP_SKB_CB(skb)->mptcp_flags & MPTCP_REINJECT;
+ return skb->mptcp_flags & MPTCP_REINJECT;
}
static void mptcp_find_and_set_pathmask(const struct sock *meta_sk, struct sk_buff *skb)
@@ -182,7 +160,7 @@ static void __mptcp_reinject_data(struct sk_buff *orig_skb, struct sock *meta_sk
/* Segment goes back to the MPTCP-layer. So, we need to zero the
* path_mask/dss.
*/
- memset(TCP_SKB_CB(skb)->dss, 0 , mptcp_dss_len);
+ TCP_SKB_CB(skb)->path_mask = 0;
/* We need to find out the path-mask from the meta-write-queue
* to properly select a subflow.
@@ -272,7 +250,7 @@ void mptcp_reinject_data(struct sock *sk, int clone_it)
if (mptcp_is_reinjected(skb_it))
continue;
- tcb->mptcp_flags |= MPTCP_REINJECT;
+ skb_it->mptcp_flags |= MPTCP_REINJECT;
__mptcp_reinject_data(skb_it, meta_sk, sk, clone_it);
}
@@ -319,25 +297,45 @@ combine:
}
}
-static int mptcp_write_dss_mapping(const struct tcp_sock *tp, const struct sk_buff *skb,
- __be32 *ptr)
+/*
+ * RFC6824 states that once a particular subflow mapping has been sent
+ * out it must never be changed. However, packets may be split while
+ * they are in the retransmission queue (due to SACK or ACKs) and that
+ * arguably means that we would change the mapping (e.g. it splits it,
+ * our sends out a subset of the initial mapping).
+ *
+ * Furthermore, the skb checksum is not always preserved across splits
+ * (e.g. mptcp_fragment) which would mean that we need to recompute
+ * the DSS checksum in this case.
+ *
+ * To avoid this we save the initial DSS mapping which allows us to
+ * send the same DSS mapping even for fragmented retransmits.
+ */
+
+static int mptcp_write_dss_mapping(const struct tcp_sock *tp,
+ const struct sk_buff *skb, __be32 *ptr)
{
const struct tcp_skb_cb *tcb = TCP_SKB_CB(skb);
- __be32 *start = ptr;
__u16 data_len;
- *ptr++ = htonl(tcb->seq); /* data_seq */
+ *ptr++ = htonl(tcb->mptcp_data_seq); /* data_seq */
/* If it's a non-data DATA_FIN, we set subseq to 0 (draft v7) */
if (mptcp_is_data_fin(skb) && skb->len == 0)
*ptr++ = 0; /* subseq */
else
- *ptr++ = htonl(tp->write_seq - tp->mptcp->snt_isn); /* subseq */
+ *ptr++ = htonl(tcb->seq - tp->mptcp->snt_isn); /* subseq */
- if (tcb->mptcp_flags & MPTCPHDR_INF)
+ if (skb->mptcp_flags & MPTCPHDR_INF)
data_len = 0;
- else
+ else {
data_len = tcb->end_seq - tcb->seq;
+ /*
+ * mptcp_entail_skb adds one for FIN
+ */
+ if (tcb->tcp_flags & TCPHDR_FIN)
+ data_len -= 1;
+ }
if (tp->mpcb->dss_csum && data_len) {
__be16 *p16 = (__be16 *)ptr;
@@ -356,11 +354,11 @@ static int mptcp_write_dss_mapping(const struct tcp_sock *tp, const struct sk_bu
(TCPOPT_NOP));
}
- return ptr - start;
+ return mptcp_dss_len/sizeof(*ptr);
}
-static int mptcp_write_dss_data_ack(const struct tcp_sock *tp, const struct sk_buff *skb,
- __be32 *ptr)
+static int mptcp_write_dss_data_ack(const struct tcp_sock *tp,
+ const struct sk_buff *skb, __be32 *ptr)
{
struct mp_dss *mdss = (struct mp_dss *)ptr;
__be32 *start = ptr;
@@ -377,54 +375,12 @@ static int mptcp_write_dss_data_ack(const struct tcp_sock *tp, const struct sk_b
mdss->len = mptcp_sub_len_dss(mdss, tp->mpcb->dss_csum);
ptr++;
+ /* data_ack */
*ptr++ = htonl(mptcp_meta_tp(tp)->rcv_nxt);
return ptr - start;
}
-/* RFC6824 states that once a particular subflow mapping has been sent
- * out it must never be changed. However, packets may be split while
- * they are in the retransmission queue (due to SACK or ACKs) and that
- * arguably means that we would change the mapping (e.g. it splits it,
- * our sends out a subset of the initial mapping).
- *
- * Furthermore, the skb checksum is not always preserved across splits
- * (e.g. mptcp_fragment) which would mean that we need to recompute
- * the DSS checksum in this case.
- *
- * To avoid this we save the initial DSS mapping which allows us to
- * send the same DSS mapping even for fragmented retransmits.
- */
-static void mptcp_save_dss_data_seq(const struct tcp_sock *tp, struct sk_buff *skb)
-{
- struct tcp_skb_cb *tcb = TCP_SKB_CB(skb);
- __be32 *ptr = (__be32 *)tcb->dss;
-
- tcb->mptcp_flags |= MPTCPHDR_SEQ;
-
- ptr += mptcp_write_dss_data_ack(tp, skb, ptr);
- ptr += mptcp_write_dss_mapping(tp, skb, ptr);
-}
-
-/* Write the saved DSS mapping to the header */
-static int mptcp_write_dss_data_seq(const struct tcp_sock *tp, struct sk_buff *skb,
- __be32 *ptr)
-{
- __be32 *start = ptr;
-
- memcpy(ptr, TCP_SKB_CB(skb)->dss, mptcp_dss_len);
-
- /* update the data_ack */
- start[1] = htonl(mptcp_meta_tp(tp)->rcv_nxt);
-
- /* dss is in a union with inet_skb_parm and
- * the IP layer expects zeroed IPCB fields.
- */
- memset(TCP_SKB_CB(skb)->dss, 0 , mptcp_dss_len);
-
- return mptcp_dss_len/sizeof(*ptr);
-}
-
static bool mptcp_skb_entail(struct sock *sk, struct sk_buff *skb, int reinject)
{
struct tcp_sock *tp = tcp_sk(sk);
@@ -434,7 +390,7 @@ static bool mptcp_skb_entail(struct sock *sk, struct sk_buff *skb, int reinject)
struct sk_buff *subskb = NULL;
if (!reinject)
- TCP_SKB_CB(skb)->mptcp_flags |= (mpcb->snd_hiseq_index ?
+ skb->mptcp_flags |= (mpcb->snd_hiseq_index ?
MPTCPHDR_SEQ64_INDEX : 0);
subskb = pskb_copy_for_clone(skb, GFP_ATOMIC);
@@ -463,14 +419,15 @@ static bool mptcp_skb_entail(struct sock *sk, struct sk_buff *skb, int reinject)
tp->mptcp->fully_established = 1;
tp->mpcb->infinite_mapping_snd = 1;
tp->mptcp->infinite_cutoff_seq = tp->write_seq;
- tcb->mptcp_flags |= MPTCPHDR_INF;
+ skb->mptcp_flags |= MPTCPHDR_INF;
}
if (mptcp_is_data_fin(subskb))
mptcp_combine_dfin(subskb, meta_sk, sk);
- mptcp_save_dss_data_seq(tp, subskb);
-
+ skb->mptcp_flags |= MPTCPHDR_SEQ;
+ tcb->mptcp_data_seq = tcb->seq;
+
tcb->seq = tp->write_seq;
/* Take into account seg len */
@@ -536,9 +493,9 @@ static int mptcp_fragment(struct sock *meta_sk, struct sk_buff *skb, u32 len,
buff = skb->next;
- flags = TCP_SKB_CB(skb)->mptcp_flags;
- TCP_SKB_CB(skb)->mptcp_flags = flags & ~(MPTCPHDR_FIN);
- TCP_SKB_CB(buff)->mptcp_flags = flags;
+ flags = skb->mptcp_flags;
+ skb->mptcp_flags = flags & ~(MPTCPHDR_FIN);
+ buff->mptcp_flags = flags;
TCP_SKB_CB(buff)->path_mask = TCP_SKB_CB(skb)->path_mask;
/* If reinject == 1, the buff will be added to the reinject
@@ -554,7 +511,8 @@ static int mptcp_fragment(struct sock *meta_sk, struct sk_buff *skb, u32 len,
tcp_sk(meta_sk)->mpcb->reinject_queue.qlen++;
meta_sk->sk_write_queue.qlen--;
- if (!before(tcp_sk(meta_sk)->snd_nxt, TCP_SKB_CB(buff)->end_seq)) {
+ if (!before(tcp_sk(meta_sk)->snd_nxt,
+ TCP_SKB_CB(buff)->end_seq)) {
undo = old_factor - tcp_skb_pcount(skb) -
tcp_skb_pcount(buff);
if (undo)
@@ -967,7 +925,7 @@ void mptcp_established_options(struct sock *sk, struct sk_buff *skb,
if (unlikely(mpcb->infinite_mapping_snd) &&
((mpcb->send_infinite_mapping && tcb &&
mptcp_is_data_seq(skb) &&
- !(tcb->mptcp_flags & MPTCPHDR_INF) &&
+ !(skb->mptcp_flags & MPTCPHDR_INF) &&
!before(tcb->seq, tp->mptcp->infinite_cutoff_seq)) ||
!mpcb->send_infinite_mapping))
return;
@@ -1197,10 +1155,10 @@ void mptcp_options_write(__be32 *ptr, struct tcp_sock *tp,
}
if (OPTION_DATA_ACK & opts->mptcp_options) {
- if (!mptcp_is_data_seq(skb))
- ptr += mptcp_write_dss_data_ack(tp, skb, ptr);
- else
- ptr += mptcp_write_dss_data_seq(tp, skb, ptr);
+ ptr += mptcp_write_dss_data_ack(tp, skb, ptr);
+ if (mptcp_is_data_seq(skb))
+ ptr += mptcp_write_dss_mapping(tp, skb, ptr);
+ skb->dev = NULL;
}
if (unlikely(OPTION_MP_PRIO & opts->mptcp_options)) {
struct mp_prio *mpprio = (struct mp_prio *)ptr;
@@ -1233,7 +1191,7 @@ void mptcp_send_fin(struct sock *meta_sk)
mss_now = mptcp_current_mss(meta_sk);
if (tcp_send_head(meta_sk) != NULL) {
- TCP_SKB_CB(skb)->mptcp_flags |= MPTCPHDR_FIN;
+ skb->mptcp_flags |= MPTCPHDR_FIN;
TCP_SKB_CB(skb)->end_seq++;
meta_tp->write_seq++;
} else {
@@ -1250,7 +1208,7 @@ void mptcp_send_fin(struct sock *meta_sk)
tcp_init_nondata_skb(skb, meta_tp->write_seq, TCPHDR_ACK);
TCP_SKB_CB(skb)->end_seq++;
- TCP_SKB_CB(skb)->mptcp_flags |= MPTCPHDR_FIN;
+ skb->mptcp_flags |= MPTCPHDR_FIN;
tcp_queue_skb(meta_sk, skb);
}
__tcp_push_pending_frames(meta_sk, mss_now, TCP_NAGLE_OFF);
--
2.7.4
3 years, 11 months
- 1
- 0
- ← Newer
- 1
- Older →