tcp: Register handlers for extra TCP options
Subsystems (5)
Mat Martineau
PATCH v2 1/1 tcp: Register handlers for extra TCP options
Allow additional TCP options to be handled by registered hook functions.
Registered options have a priority that determines the order in which options are prepared and written. Lower priority numbers are handled first.
Option parsing will call the provided 'parse' function when a TCP option number is not recognized by the normal option parsing code.
The 'prepare' function determines the required space for registered options and store associated data. 'write' adds the option to the TCP header.
A static key and RCU synchronization are used to minimize the performance impact of these extensible TCP features.
---
Changes from v1: One 'prepare' callback (no more special callback for request_sock), and add a few missing callback sites (like ipv6).
drivers/infiniband/hw/cxgb4/cm.c | 2 +- include/linux/tcp.h | 22 +++++++ include/net/tcp.h | 40 +++++++++++- net/ipv4/syncookies.c | 2 +- net/ipv4/tcp.c | 133 +++++++++++++++++++++++++++++++++++++++ net/ipv4/tcp_input.c | 16 +++-- net/ipv4/tcp_ipv4.c | 80 ++++++++++++++++++----- net/ipv4/tcp_minisocks.c | 4 +- net/ipv4/tcp_output.c | 43 +++++++------ net/ipv6/syncookies.c | 2 +- net/ipv6/tcp_ipv6.c | 28 ++++++++- 11 files changed, 323 insertions(+), 49 deletions(-)
drivers/infiniband/hw/cxgb4/cm.c +1 -1
@@ -3752,7 +3752,7 @@ static void build_cpl_pass_accept_req(struct sk_buff *skb, int stid , u8 tos)*/ memset (&tmp_opt,0 ,sizeof (tmp_opt));tcp_clear_options (&tmp_opt); -tcp_parse_options (&init_net, skb, &tmp_opt,0 ,NULL ); +tcp_parse_options (&init_net, skb, &tmp_opt,0 ,NULL ,NULL ); req =__skb_push (skb,sizeof (*req));memset (req,0 ,sizeof (*req));
include/linux/tcp.h +22 -0
@@ -112,6 +112,23 @@ static inline void tcp_clear_options(struct tcp_options_received *rx_opt) rx_opt->wscale_ok = rx_opt->snd_wscale =0 ; } +#define OPTION_SACK_ADVERTISE (1 << 0) +#define OPTION_TS (1 << 1) +#define OPTION_MD5 (1 << 2) +#define OPTION_WSCALE (1 << 3) +#define OPTION_FAST_OPEN_COOKIE (1 << 8) + +struct tcp_out_options { +u16 options;/* bit field of OPTION_* */ +u16 mss;/* 0 to disable */ +u8 ws;/* window scale, 0 to disable */ +u8 num_sack_blocks;/* number of SACK blocks to include */ +u8 hash_size;/* bytes in hash_location */ +__u8 *hash_location;/* temporary pointer, overloaded */ +__u32 tsval, tsecr;/* need to include OPTION_TS */ +struct tcp_fastopen_cookie *fastopen_cookie;/* Fast open cookie */ +}; +/* This is the max number of SACKS that we'll generate and process. It's safe * to increase this, although since: * size = TCPOLEN_SACK_BASE_ALIGNED (4) + n * TCPOLEN_SACK_PERBLOCK (8)
@@ -389,6 +406,11 @@ static inline struct tcp_sock *tcp_sk(const struct sock *sk)return (struct tcp_sock *)sk; } +static inline struct sock *tcp_to_sk (const struct tcp_sock *tp) +{ +return (struct sock *)tp; +} +struct tcp_timewait_sock {struct inet_timewait_sock tw_sk;#define tw_rcv_nxt tw_sk.__tw_common.skc_tw_rcv_nxt
include/net/tcp.h +39 -1
@@ -433,7 +433,8 @@ int tcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int nonblock, int flags, int *addr_len);void tcp_parse_options (const struct net *net,const struct sk_buff *skb,struct tcp_options_received *opt_rx, -int estab,struct tcp_fastopen_cookie *foc); +int estab,struct tcp_fastopen_cookie *foc, +struct tcp_sock *tp);const u8 *tcp_parse_md5sig_option (const struct tcphdr *th);/*
@@ -2109,4 +2110,41 @@ static inline bool tcp_bpf_ca_needs_ecn(struct sock *sk) {return (tcp_call_bpf (sk,BPF_SOCK_OPS_NEEDS_ECN ) ==1 ); } + +extern struct static_key_false tcp_extra_options_enabled; + +struct tcp_extra_option_ops { +struct list_head list; +unsigned char option_kind; +unsigned char priority; +void (*parse)(int opsize,const unsigned char *opptr, +const struct sk_buff *skb, +struct tcp_options_received *opt_rx, +struct sock *sk); +/* Return the number of bytes consumed */ +unsigned int (*prepare)(struct sk_buff *skb,u8 flags, +unsigned int remaining, +struct tcp_out_options *opts, +const struct sock *sk); +void (*write)(__be32 *ptr,struct tcp_out_options *opts, +const struct sock *sk); +struct module *owner; +}; + +void tcp_extra_options_parse (int opcode,int opsize,const unsigned char *opptr, +const struct sk_buff *skb, +struct tcp_options_received *opt_rx, +struct sock *sk); + +unsigned int tcp_extra_options_prepare (struct sk_buff *skb,u8 flags, +unsigned int remaining, +struct tcp_out_options *opts, +const struct sock *sk); + +void tcp_extra_options_write (__be32 *ptr,struct tcp_out_options *opts, +const struct sock *sk); + +int tcp_register_extra_option (struct tcp_extra_option_ops *ops); +void tcp_unregister_extra_option (struct tcp_extra_option_ops *ops); +#endif /* _TCP_H */
net/ipv4/syncookies.c +1 -1
@@ -313,7 +313,7 @@ struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb)/* check for timestamp cookie support */ memset (&tcp_opt,0 ,sizeof (tcp_opt)); -tcp_parse_options (sock_net (sk), skb, &tcp_opt,0 ,NULL ); +tcp_parse_options (sock_net (sk), skb, &tcp_opt,0 ,NULL , tp);if (tcp_opt.saw_tstamp && tcp_opt.rcv_tsecr) { tsoff = secure_tcp_ts_off(sock_net(sk),
net/ipv4/tcp.c +133 -0
157 lines not on this page · raw message
net/ipv4/tcp_input.c +11 -5
@@ -3728,7 +3728,7 @@ static void tcp_parse_fastopen_option(int len, const unsigned char *cookie,void tcp_parse_options (const struct net *net,const struct sk_buff *skb,struct tcp_options_received *opt_rx,int estab, -struct tcp_fastopen_cookie *foc) +struct tcp_fastopen_cookie *foc,struct tcp_sock *tp) {const unsigned char *ptr;const struct tcphdr *th =tcp_hdr (skb);
@@ -3830,6 +3830,12 @@ void tcp_parse_options(const struct net *net, ptr + 2, th->syn, foc, true);break ; +default : +tcp_extra_options_parse (opcode, opsize, ptr, + skb, opt_rx, +tcp_to_sk (tp)); +break ; + } ptr += opsize-2 ; length -= opsize;
@@ -3876,7 +3882,7 @@ static bool tcp_fast_parse_options(const struct net *net,return true; } -tcp_parse_options (net, skb, &tp->rx_opt,1 ,NULL ); +tcp_parse_options (net, skb, &tp->rx_opt,1 ,NULL , tp);if (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr) tp->rx_opt.rcv_tsecr -= tp->tsoffset;
@@ -5569,7 +5575,7 @@ static bool tcp_rcv_fastopen_synack(struct sock *sk, struct sk_buff *synack,/* Get original SYNACK MSS value if user MSS sets mss_clamp */ tcp_clear_options (&opt); opt.user_mss = opt.mss_clamp =0 ; -tcp_parse_options (sock_net (sk), synack, &opt,0 ,NULL ); +tcp_parse_options (sock_net (sk), synack, &opt,0 ,NULL , tp); mss = opt.mss_clamp; }
@@ -5623,7 +5629,7 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb,int saved_clamp = tp->rx_opt.mss_clamp;bool fastopen_fail; -tcp_parse_options (sock_net (sk), skb, &tp->rx_opt,0 , &foc); +tcp_parse_options (sock_net (sk), skb, &tp->rx_opt,0 , &foc, tp);if (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr) tp->rx_opt.rcv_tsecr -= tp->tsoffset;
@@ -6299,7 +6305,7 @@ int tcp_conn_request(struct request_sock_ops *rsk_ops, tmp_opt.mss_clamp = af_ops->mss_clamp; tmp_opt.user_mss = tp->rx_opt.user_mss;tcp_parse_options (sock_net (sk), skb, &tmp_opt,0 , - want_cookie ?NULL : &foc); + want_cookie ?NULL : &foc, tp);if (want_cookie && !tmp_opt.saw_tstamp)tcp_clear_options (&tmp_opt);
net/ipv4/tcp_ipv4.c +63 -17
142 lines not on this page · raw message
net/ipv4/tcp_minisocks.c +2 -2
@@ -98,7 +98,7 @@ tcp_timewait_state_process(struct inet_timewait_sock *tw, struct sk_buff *skb, tmp_opt.saw_tstamp =0 ;if (th->doff > (sizeof (*th) >>2 ) && tcptw->tw_ts_recent_stamp) { -tcp_parse_options (twsk_net (tw), skb, &tmp_opt,0 ,NULL ); +tcp_parse_options (twsk_net (tw), skb, &tmp_opt,0 ,NULL ,NULL );if (tmp_opt.saw_tstamp) {if (tmp_opt.rcv_tsecr)
@@ -565,7 +565,7 @@ struct sock *tcp_check_req(struct sock *sk, struct sk_buff *skb, tmp_opt.saw_tstamp =0 ;if (th->doff > (sizeof (struct tcphdr )>>2 )) { -tcp_parse_options (sock_net (sk), skb, &tmp_opt,0 ,NULL ); +tcp_parse_options (sock_net (sk), skb, &tmp_opt,0 ,NULL ,NULL );if (tmp_opt.saw_tstamp) { tmp_opt.ts_recent = req->ts_recent;
net/ipv4/tcp_output.c +24 -19
105 lines not on this page · raw message
net/ipv6/syncookies.c +1 -1
@@ -162,7 +162,7 @@ struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb)/* check for timestamp cookie support */ memset (&tcp_opt,0 ,sizeof (tcp_opt)); -tcp_parse_options (sock_net (sk), skb, &tcp_opt,0 ,NULL ); +tcp_parse_options (sock_net (sk), skb, &tcp_opt,0 ,NULL , tp);if (tcp_opt.saw_tstamp && tcp_opt.rcv_tsecr) { tsoff = secure_tcpv6_ts_off(sock_net(sk),
net/ipv6/tcp_ipv6.c +26 -2
@@ -784,9 +784,10 @@ static void tcp_v6_send_response(const struct sock *sk, struct sk_buff *skb, u32struct flowi6 fl6;struct net *net = sk ?sock_net (sk) :dev_net (skb_dst (skb)->dev);struct sock *ctl_sk = net->ipv6.tcp_sk; -unsigned int tot_len =sizeof (struct tcphdr ); +unsigned int tot_len =0 ;struct dst_entry *dst;__be32 *topt; +struct tcp_out_options extraopts;if (tsecr) tot_len +=TCPOLEN_TSTAMP_ALIGNED ;
@@ -795,10 +796,28 @@ static void tcp_v6_send_response(const struct sock *sk, struct sk_buff *skb, u32 tot_len +=TCPOLEN_MD5SIG_ALIGNED ;#endif +rcu_read_lock (); +if (static_branch_unlikely (&tcp_extra_options_enabled)) { +unsigned int remaining =MAX_TCP_OPTION_SPACE - tot_len; +u8 extraflags = rst ?TCPHDR_RST :0 ; + +if (!rst || !th->ack) + extraflags |=TCPHDR_ACK ; + +memset (&extraopts,0 ,sizeof (extraopts)); + + tot_len +=tcp_extra_options_prepare (skb, extraflags, remaining, + &extraopts, sk); + } + + tot_len +=sizeof (struct tcphdr ); + buff =alloc_skb (MAX_HEADER +sizeof (struct ipv6hdr ) + tot_len,GFP_ATOMIC ); -if (!buff) +if (!buff) { +rcu_read_unlock ();return ; + }skb_reserve (buff,MAX_HEADER +sizeof (struct ipv6hdr ) + tot_len);
@@ -835,6 +854,11 @@ static void tcp_v6_send_response(const struct sock *sk, struct sk_buff *skb, u32 }#endif +if (static_branch_unlikely (&tcp_extra_options_enabled)) +tcp_extra_options_write (topt, &extraopts, sk); + +rcu_read_unlock (); +memset (&fl6,0 ,sizeof (fl6)); fl6.daddr =ipv6_hdr (skb)->saddr; fl6.saddr =ipv6_hdr (skb)->daddr;
Signature · 1 line
2.14.2
Christoph Paasch
Hello Mat,
Quoted 14 lines · Mat Martineau
Allow additional TCP options to be handled by registered hook functions.
Registered options have a priority that determines the order in which options are prepared and written. Lower priority numbers are handled first.
Option parsing will call the provided 'parse' function when a TCP option number is not recognized by the normal option parsing code.
The 'prepare' function determines the required space for registered options and store associated data. 'write' adds the option to the TCP header.
Quoted 8 lines
A static key and RCU synchronization are used to minimize the performance impact of these extensible TCP features.
Signed-off-by Mat Martineau <mathew.j.martineau(a)linux.intel.com>---
Changes from v1: One 'prepare' callback (no more special callback for
request_sock), and add a few missing callback sites (like ipv6).
great, I like that we now have only one 'prepare' callback.
Quoted 56 lines
drivers/infiniband/hw/cxgb4/cm.c | 2 +- include/linux/tcp.h | 22 +++++++ include/net/tcp.h | 40 +++++++++++- net/ipv4/syncookies.c | 2 +- net/ipv4/tcp.c | 133 +++++++++++++++++++++++++++++++++++++++ net/ipv4/tcp_input.c | 16 +++-- net/ipv4/tcp_ipv4.c | 80 ++++++++++++++++++----- net/ipv4/tcp_minisocks.c | 4 +- net/ipv4/tcp_output.c | 43 +++++++------ net/ipv6/syncookies.c | 2 +- net/ipv6/tcp_ipv6.c | 28 ++++++++- 11 files changed, 323 insertions(+), 49 deletions(-)drivers/infiniband/hw/cxgb4/cm.c +1
-1@@ -3752,7 +3752,7 @@ static void build_cpl_pass_accept_req(struct sk_buff *skb, int stid , u8 tos)*/ memset (&tmp_opt,0 ,sizeof (tmp_opt));tcp_clear_options (&tmp_opt); -tcp_parse_options (&init_net, skb, &tmp_opt,0 ,NULL ); +tcp_parse_options (&init_net, skb, &tmp_opt,0 ,NULL ,NULL ); req =__skb_push (skb,sizeof (*req));memset (req,0 ,sizeof (*req));include/linux/tcp.h +17
-0@@ -112,6 +112,23 @@ static inline void tcp_clear_options(struct tcp_options_received *rx_opt) rx_opt->wscale_ok = rx_opt->snd_wscale =0 ; } +#define OPTION_SACK_ADVERTISE (1 << 0) +#define OPTION_TS (1 << 1) +#define OPTION_MD5 (1 << 2) +#define OPTION_WSCALE (1 << 3) +#define OPTION_FAST_OPEN_COOKIE (1 << 8) + +struct tcp_out_options { +u16 options;/* bit field of OPTION_* */ +u16 mss;/* 0 to disable */ +u8 ws;/* window scale, 0 to disable */ +u8 num_sack_blocks;/* number of SACK blocks to include */ +u8 hash_size;/* bytes in hash_location */ +__u8 *hash_location;/* temporary pointer, overloaded */ +__u32 tsval, tsecr;/* need to include OPTION_TS */ +struct tcp_fastopen_cookie *fastopen_cookie;/* Fast open cookie */ +}; +/* This is the max number of SACKS that we'll generate and process. It's safe * to increase this, although since: * size = TCPOLEN_SACK_BASE_ALIGNED (4) + n * TCPOLEN_SACK_PERBLOCK (8)
include/linux/tcp.h static inline struct tcp_sock *tcp_sk(const struct sock *sk) · line 412
Quoted 7 lines
include/linux/tcp.h +4
-0@@ -389,6 +406,11 @@ static inline struct tcp_sock *tcp_sk(const struct sock *sk)return (struct tcp_sock *)sk; } +static inline struct sock *tcp_to_sk (const struct tcp_sock *tp) +{ +return (struct sock *)tp; +}
Nice little function :)
Quoted 128 lines
include/linux/tcp.h +1
-0+struct tcp_timewait_sock {struct inet_timewait_sock tw_sk;#define tw_rcv_nxt tw_sk.__tw_common.skc_tw_rcv_nxtinclude/net/tcp.h +39
-1@@ -433,7 +433,8 @@ int tcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int nonblock, int flags, int *addr_len);void tcp_parse_options (const struct net *net,const struct sk_buff *skb,struct tcp_options_received *opt_rx, -int estab,struct tcp_fastopen_cookie *foc); +int estab,struct tcp_fastopen_cookie *foc, +struct tcp_sock *tp);const u8 *tcp_parse_md5sig_option (const struct tcphdr *th);/* @@ -2109,4 +2110,41 @@ static inline bool tcp_bpf_ca_needs_ecn(struct sock *sk) {return (tcp_call_bpf (sk,BPF_SOCK_OPS_NEEDS_ECN ) ==1 ); } + +extern struct static_key_false tcp_extra_options_enabled; + +struct tcp_extra_option_ops { +struct list_head list; +unsigned char option_kind; +unsigned char priority; +void (*parse)(int opsize,const unsigned char *opptr, +const struct sk_buff *skb, +struct tcp_options_received *opt_rx, +struct sock *sk); +/* Return the number of bytes consumed */ +unsigned int (*prepare)(struct sk_buff *skb,u8 flags, +unsigned int remaining, +struct tcp_out_options *opts, +const struct sock *sk); +void (*write)(__be32 *ptr,struct tcp_out_options *opts, +const struct sock *sk); +struct module *owner; +}; + +void tcp_extra_options_parse (int opcode,int opsize,const unsigned char *opptr, +const struct sk_buff *skb, +struct tcp_options_received *opt_rx, +struct sock *sk); + +unsigned int tcp_extra_options_prepare (struct sk_buff *skb,u8 flags, +unsigned int remaining, +struct tcp_out_options *opts, +const struct sock *sk); + +void tcp_extra_options_write (__be32 *ptr,struct tcp_out_options *opts, +const struct sock *sk); + +int tcp_register_extra_option (struct tcp_extra_option_ops *ops); +void tcp_unregister_extra_option (struct tcp_extra_option_ops *ops); +#endif /* _TCP_H */ net/ipv4/syncookies.c +1
-1@@ -313,7 +313,7 @@ struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb)/* check for timestamp cookie support */ memset (&tcp_opt,0 ,sizeof (tcp_opt)); -tcp_parse_options (sock_net (sk), skb, &tcp_opt,0 ,NULL ); +tcp_parse_options (sock_net (sk), skb, &tcp_opt,0 ,NULL , tp);if (tcp_opt.saw_tstamp && tcp_opt.rcv_tsecr) { tsoff = secure_tcp_ts_off(sock_net(sk),net/ipv4/tcp.c +33
-0@@ -270,6 +270,7 @@#include <linux/time.h> #include <linux/slab.h> #include <linux/errqueue.h> +#include <linux/static_key.h> #include <net/icmp.h> #include <net/inet_common.h> @@ -306,6 +307,13 @@ EXPORT_SYMBOL(tcp_memory_allocated);struct percpu_counter tcp_sockets_allocated;EXPORT_SYMBOL (tcp_sockets_allocated); +/* +* Optional TCP option handlers +*/ +static DEFINE_SPINLOCK (tcp_option_list_lock); +static LIST_HEAD (tcp_option_list); +DEFINE_STATIC_KEY_FALSE (tcp_extra_options_enabled); +/* * TCP splice context */ @@ -3375,6 +3383,130 @@ EXPORT_SYMBOL(tcp_md5_hash_key);#endif +/* Linear search, few entries are expected. The RCU read lock must +* be held before calling. +*/ +static struct tcp_extra_option_ops *tcp_extra_options_find_kind (unsigned char kind) +{ +struct tcp_extra_option_ops *entry; + +list_for_each_entry_rcu (entry, &tcp_option_list, list) { +if (entry->option_kind == kind) +return entry; + } + +return NULL ; +} + +void tcp_extra_options_parse (int opcode,int opsize,const unsigned char *opptr, +const struct sk_buff *skb, +struct tcp_options_received *opt_rx, +struct sock *sk) +{ +struct tcp_extra_option_ops *entry; + +rcu_read_lock (); + entry =tcp_extra_options_find_kind (opcode); +if (entry && entry->parse)
net/ipv4/tcp.c EXPORT_SYMBOL(tcp_md5_hash_key); · line 3420
Quoted 10 lines
net/ipv4/tcp.c +10
-0+ entry->parse (opsize, opptr, skb, opt_rx, sk); +rcu_read_unlock (); +} +EXPORT_SYMBOL_GPL (tcp_extra_options_parse); + +/* The RCU read lock must be held before calling, and should span both +* the call to this function and tcp_extra_options_write to ensure that +* tcp_option_list does not change between the two calls. To preserve +* expected option alignment, always returns a multiple of 4 bytes. +*/
The RCU read lock won't be able to protect the list from being altered. All it will take care of is that the elements on the list won't get free'd while traversing it and that the next-pointer won't get changed. That way the list-traversal is still fine.
If we move the extra-options to a mode where the list is on a per-TCP socket basis, we can avoid handling this. Because we can limit adding/removing TCP-options to happen only when the socket is in TCP_CLOSE. So, once the connection started, the list will always remain the same.
Quoted 191 lines
net/ipv4/tcp.c +90
-0+unsigned int tcp_extra_options_prepare (struct sk_buff *skb,u8 flags, +unsigned int remaining, +struct tcp_out_options *opts, +const struct sock *sk) +{ +struct tcp_extra_option_ops *entry; +unsigned int used =0 ; + +list_for_each_entry_rcu (entry, &tcp_option_list, list) { +if (unlikely (!entry->prepare)) +continue ; + + used += entry->prepare (skb, flags, remaining - used, opts, sk); + } + +return roundup (used,4 ); +} +EXPORT_SYMBOL_GPL (tcp_extra_options_prepare); + +/* The RCU read lock must be held before calling, and should span both +* the call to tcp_extra_options_write and this function to ensure that +* tcp_option_list does not change between the two calls. +*/ +void tcp_extra_options_write (__be32 *ptr,struct tcp_out_options *opts, +const struct sock *sk) +{ +struct tcp_extra_option_ops *entry; + +list_for_each_entry_rcu (entry, &tcp_option_list, list) { +if (unlikely (!entry->write)) +continue ; + + entry->write (ptr, opts, sk); + } +} +EXPORT_SYMBOL_GPL (tcp_extra_options_write); + +int tcp_register_extra_option (struct tcp_extra_option_ops *ops) +{ +struct tcp_extra_option_ops *entry; +struct list_head * add_before = &tcp_option_list; +int ret =0 ; + +if (!ops->option_kind) +return -EINVAL ; + +if (!try_module_get (ops->owner)) +return -ENOENT ; + +spin_lock (&tcp_option_list_lock); + +list_for_each_entry_rcu (entry, &tcp_option_list, list) { +if (entry->option_kind == ops->option_kind) { +pr_notice ("Option kind %u already registered\n" , + ops->option_kind); +spin_unlock (&tcp_option_list_lock); +module_put (ops->owner); +return -EEXIST ; + } + +if (entry->priority <= ops->priority) + add_before = &entry->list; + } + +list_add_tail_rcu (&ops->list, add_before); +pr_debug ("Option kind %u registered\n" , ops->option_kind); + +spin_unlock (&tcp_option_list_lock); + +static_branch_inc (&tcp_extra_options_enabled); + +return ret; +} +EXPORT_SYMBOL_GPL (tcp_register_extra_option); + +void tcp_unregister_extra_option (struct tcp_extra_option_ops *ops) +{ +spin_lock (&tcp_option_list_lock); +list_del_rcu (&ops->list); +spin_unlock (&tcp_option_list_lock); + +synchronize_net (); + +static_branch_dec (&tcp_extra_options_enabled); + +module_put (ops->owner); +} +EXPORT_SYMBOL_GPL (tcp_unregister_extra_option); +void tcp_done (struct sock *sk) {struct request_sock *req =tcp_sk (sk)->fastopen_rsk;@@ -3521,6 +3653,7 @@ void __init tcp_init(void)INIT_HLIST_HEAD (&tcp_hashinfo.bhash[i].chain); } +INIT_LIST_HEAD (&tcp_option_list); cnt = tcp_hashinfo.ehash_mask +1 ; sysctl_tcp_max_orphans = cnt /2 ;net/ipv4/tcp_input.c +11
-5@@ -3728,7 +3728,7 @@ static void tcp_parse_fastopen_option(int len, const unsigned char *cookie,void tcp_parse_options (const struct net *net,const struct sk_buff *skb,struct tcp_options_received *opt_rx,int estab, -struct tcp_fastopen_cookie *foc) +struct tcp_fastopen_cookie *foc,struct tcp_sock *tp) {const unsigned char *ptr;const struct tcphdr *th =tcp_hdr (skb);@@ -3830,6 +3830,12 @@ void tcp_parse_options(const struct net *net, ptr + 2, th->syn, foc, true);break ; +default : +tcp_extra_options_parse (opcode, opsize, ptr, + skb, opt_rx, +tcp_to_sk (tp)); +break ; + } ptr += opsize-2 ; length -= opsize;@@ -3876,7 +3882,7 @@ static bool tcp_fast_parse_options(const struct net *net,return true; } -tcp_parse_options (net, skb, &tp->rx_opt,1 ,NULL ); +tcp_parse_options (net, skb, &tp->rx_opt,1 ,NULL , tp);if (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr) tp->rx_opt.rcv_tsecr -= tp->tsoffset;@@ -5569,7 +5575,7 @@ static bool tcp_rcv_fastopen_synack(struct sock *sk, struct sk_buff *synack,/* Get original SYNACK MSS value if user MSS sets mss_clamp */ tcp_clear_options (&opt); opt.user_mss = opt.mss_clamp =0 ; -tcp_parse_options (sock_net (sk), synack, &opt,0 ,NULL ); +tcp_parse_options (sock_net (sk), synack, &opt,0 ,NULL , tp); mss = opt.mss_clamp; }@@ -5623,7 +5629,7 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb,int saved_clamp = tp->rx_opt.mss_clamp;bool fastopen_fail; -tcp_parse_options (sock_net (sk), skb, &tp->rx_opt,0 , &foc); +tcp_parse_options (sock_net (sk), skb, &tp->rx_opt,0 , &foc, tp);if (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr) tp->rx_opt.rcv_tsecr -= tp->tsoffset;@@ -6299,7 +6305,7 @@ int tcp_conn_request(struct request_sock_ops *rsk_ops, tmp_opt.mss_clamp = af_ops->mss_clamp; tmp_opt.user_mss = tp->rx_opt.user_mss;tcp_parse_options (sock_net (sk), skb, &tmp_opt,0 , - want_cookie ?NULL : &foc); + want_cookie ?NULL : &foc, tp);if (want_cookie && !tmp_opt.saw_tstamp)tcp_clear_options (&tmp_opt);net/ipv4/tcp_ipv4.c +4
-3@@ -598,9 +598,7 @@ static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb)const struct tcphdr *th =tcp_hdr (skb);struct {struct tcphdr th; -#ifdef CONFIG_TCP_MD5SIG -__be32 opt[(TCPOLEN_MD5SIG_ALIGNED >>2 )]; -#endif +__be32 opt[(MAX_TCP_OPTION_SPACE >>2 )]; } rep;struct ip_reply_arg arg;#ifdef CONFIG_TCP_MD5SIG @@ -611,6 +609,7 @@ static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb)struct sock *sk1 =NULL ;#endif struct net *net; +int offset =0 ;/* Never send a reset in response to a reset. */ if (th->rst)@@ -676,17 +675,44 @@ static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb) goto out; } +#endif +
net/ipv4/tcp_ipv4.c static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb) · line 689
Quoted 10 lines
net/ipv4/tcp_ipv4.c +10
-0+if (static_branch_unlikely (&tcp_extra_options_enabled)) { +unsigned int remaining; +unsigned int used; +struct tcp_out_options opts; + + remaining =sizeof (rep.opt); +#ifdef CONFIG_TCP_MD5SIG +if (key) + remaining -=TCPOLEN_MD5SIG_ALIGNED ; +#endif
We will break TCP_MD5 here with this patch if we move it inside the static branch.
Only after the patch that makes TCP_MD5 adopt the framework we can move this code to here.
Cheers,
Christoph
Quoted 331 lines
net/ipv4/tcp_ipv4.c +49
-14+ +memset (&opts,0 ,sizeof (opts)); + +rcu_read_lock (); + used =tcp_extra_options_prepare (NULL ,TCPHDR_RST , remaining, + &opts, sk); + +tcp_extra_options_write (&rep.opt[0 ], &opts, sk); +rcu_read_unlock (); + + arg.iov[0 ].iov_len += used; + offset += used /4 ; + rep.th.doff = arg.iov[0 ].iov_len /4 ; + } +#ifdef CONFIG_TCP_MD5SIG if (key) { - rep.opt[0 ] =htonl ((TCPOPT_NOP <<24 ) | - (TCPOPT_NOP <<16 ) | - (TCPOPT_MD5SIG <<8 ) | -TCPOLEN_MD5SIG ); + rep.opt[offset++] =htonl ((TCPOPT_NOP <<24 ) | + (TCPOPT_NOP <<16 ) | + (TCPOPT_MD5SIG <<8 ) | +TCPOLEN_MD5SIG );/* Update length and the length the header thinks exists */ arg.iov[0 ].iov_len +=TCPOLEN_MD5SIG_ALIGNED ; rep.th.doff = arg.iov[0 ].iov_len /4 ; -tcp_v4_md5_hash_hdr ((__u8 *) &rep.opt[1 ], +tcp_v4_md5_hash_hdr ((__u8 *) &rep.opt[offset], key,ip_hdr (skb)->saddr,ip_hdr (skb)->daddr, &rep.th); }@@ -738,14 +764,11 @@ static void tcp_v4_send_ack(const struct sock *sk,const struct tcphdr *th =tcp_hdr (skb);struct {struct tcphdr th; -__be32 opt[(TCPOLEN_TSTAMP_ALIGNED >>2 ) -#ifdef CONFIG_TCP_MD5SIG - + (TCPOLEN_MD5SIG_ALIGNED >>2 ) -#endif - ]; +__be32 opt[(MAX_TCP_OPTION_SPACE >>2 )]; } rep;struct net *net =sock_net (sk);struct ip_reply_arg arg; +int offset =0 ;memset (&rep.th,0 ,sizeof (struct tcphdr ));memset (&arg,0 ,sizeof (arg));@@ -759,33 +782,56 @@ static void tcp_v4_send_ack(const struct sock *sk, rep.opt[1 ] =htonl (tsval); rep.opt[2 ] =htonl (tsecr); arg.iov[0 ].iov_len +=TCPOLEN_TSTAMP_ALIGNED ; + offset +=3 ; }/* Swap the send and the receive. */ rep.th.dest = th->source; rep.th.source = th->dest; - rep.th.doff = arg.iov[0 ].iov_len /4 ; rep.th.seq =htonl (seq); rep.th.ack_seq =htonl (ack); rep.th.ack =1 ; rep.th.window =htons (win); +if (static_branch_unlikely (&tcp_extra_options_enabled)) { +unsigned int remaining; +unsigned int used; +struct tcp_out_options opts; + + remaining =sizeof (rep.th) +sizeof (rep.opt) - arg.iov[0 ].iov_len;#ifdef CONFIG_TCP_MD5SIG -if (key) { -int offset = (tsecr) ?3 :0 ; +if (key) + remaining -=TCPOLEN_MD5SIG_ALIGNED ; +#endif +memset (&opts,0 ,sizeof (opts)); +rcu_read_lock (); + used =tcp_extra_options_prepare (NULL ,TCPHDR_ACK , remaining, + &opts, sk); + +tcp_extra_options_write (&rep.opt[offset], &opts, sk); +rcu_read_unlock (); + + arg.iov[0 ].iov_len += used; + offset += used /4 ; + } + + rep.th.doff = arg.iov[0 ].iov_len /4 ; + +#ifdef CONFIG_TCP_MD5SIG +if (key) { rep.opt[offset++] =htonl ((TCPOPT_NOP <<24 ) | (TCPOPT_NOP <<16 ) | (TCPOPT_MD5SIG <<8 ) |TCPOLEN_MD5SIG ); arg.iov[0 ].iov_len +=TCPOLEN_MD5SIG_ALIGNED ; - rep.th.doff = arg.iov[0 ].iov_len/4 ;tcp_v4_md5_hash_hdr ((__u8 *) &rep.opt[offset], key,ip_hdr (skb)->saddr,ip_hdr (skb)->daddr, &rep.th); }#endif + arg.flags = reply_flags; arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr, ip_hdr(skb)->saddr, /* XXX */net/ipv4/tcp_minisocks.c +2
-2@@ -98,7 +98,7 @@ tcp_timewait_state_process(struct inet_timewait_sock *tw, struct sk_buff *skb, tmp_opt.saw_tstamp =0 ;if (th->doff > (sizeof (*th) >>2 ) && tcptw->tw_ts_recent_stamp) { -tcp_parse_options (twsk_net (tw), skb, &tmp_opt,0 ,NULL ); +tcp_parse_options (twsk_net (tw), skb, &tmp_opt,0 ,NULL ,NULL );if (tmp_opt.saw_tstamp) {if (tmp_opt.rcv_tsecr)@@ -565,7 +565,7 @@ struct sock *tcp_check_req(struct sock *sk, struct sk_buff *skb, tmp_opt.saw_tstamp =0 ;if (th->doff > (sizeof (struct tcphdr )>>2 )) { -tcp_parse_options (sock_net (sk), skb, &tmp_opt,0 ,NULL ); +tcp_parse_options (sock_net (sk), skb, &tmp_opt,0 ,NULL ,NULL );if (tmp_opt.saw_tstamp) { tmp_opt.ts_recent = req->ts_recent;net/ipv4/tcp_output.c +24
-19@@ -41,6 +41,7 @@#include <linux/compiler.h> #include <linux/gfp.h> #include <linux/module.h> +#include <linux/static_key.h> /* People can turn this off for buggy TCP's found in printers etc. */ int sysctl_tcp_retrans_collapse__read_mostly =1 ;@@ -413,23 +414,6 @@ static inline bool tcp_urg_mode(const struct tcp_sock *tp)return tp->snd_una != tp->snd_up; } -#define OPTION_SACK_ADVERTISE (1 << 0) -#define OPTION_TS (1 << 1) -#define OPTION_MD5 (1 << 2) -#define OPTION_WSCALE (1 << 3) -#define OPTION_FAST_OPEN_COOKIE (1 << 8) - -struct tcp_out_options { -u16 options;/* bit field of OPTION_* */ -u16 mss;/* 0 to disable */ -u8 ws;/* window scale, 0 to disable */ -u8 num_sack_blocks;/* number of SACK blocks to include */ -u8 hash_size;/* bytes in hash_location */ -__u8 *hash_location;/* temporary pointer, overloaded */ -__u32 tsval, tsecr;/* need to include OPTION_TS */ -struct tcp_fastopen_cookie *fastopen_cookie;/* Fast open cookie */ -}; -/* Write previously computed TCP options to the packet. * * Beware: Something in the Internet is very sensitive to the ordering of @@ -536,6 +520,9 @@ static void tcp_options_write(__be32 *ptr, struct tcp_sock *tp, } ptr += (len +3 ) >>2 ; } + +if (static_branch_unlikely (&tcp_extra_options_enabled)) +tcp_extra_options_write (ptr, opts,tcp_to_sk (tp)); }/* Compute TCP options for SYN packets. This is not the final @@ -603,6 +590,11 @@ static unsigned int tcp_syn_options(struct sock *sk, struct sk_buff *skb, } } +if (static_branch_unlikely (&tcp_extra_options_enabled)) + remaining -=tcp_extra_options_prepare (skb,TCPHDR_SYN , + remaining, opts, +tcp_to_sk (tp)); +return MAX_TCP_OPTION_SPACE - remaining; }@@ -663,6 +655,12 @@ static unsigned int tcp_synack_options(struct request_sock *req, } } +if (static_branch_unlikely (&tcp_extra_options_enabled)) + remaining -=tcp_extra_options_prepare (skb, +TCPHDR_SYN |TCPHDR_ACK , + remaining, opts, +req_to_sk (req)); +return MAX_TCP_OPTION_SPACE - remaining; }@@ -696,6 +694,11 @@ static unsigned int tcp_established_options(struct sock *sk, struct sk_buff *skb size +=TCPOLEN_TSTAMP_ALIGNED ; } +if (static_branch_unlikely (&tcp_extra_options_enabled)) + size +=tcp_extra_options_prepare (skb,0 , +MAX_TCP_OPTION_SPACE - size, + opts,tcp_to_sk (tp)); + eff_sacks = tp->rx_opt.num_sacks + tp->rx_opt.dsack;if (unlikely (eff_sacks)) {const unsigned int remaining =MAX_TCP_OPTION_SPACE - size;@@ -1016,6 +1019,7 @@ static int tcp_transmit_skb(struct sock *sk, struct sk_buff *skb, int clone_it, tcb =TCP_SKB_CB (skb);memset (&opts,0 ,sizeof (opts)); +rcu_read_lock ();if (unlikely (tcb->tcp_flags &TCPHDR_SYN )) tcp_options_size =tcp_syn_options (sk, skb, &opts, &md5);else @@ -1092,6 +1096,7 @@ static int tcp_transmit_skb(struct sock *sk, struct sk_buff *skb, int clone_it, md5, sk, skb); }#endif +rcu_read_unlock (); icsk->icsk_af_ops->send_check (sk, skb);@@ -3156,8 +3161,8 @@ struct sk_buff *tcp_make_synack(const struct sock *sk, struct dst_entry *dst,#endif skb->skb_mstamp =tcp_clock_us (); -#ifdef CONFIG_TCP_MD5SIG rcu_read_lock (); +#ifdef CONFIG_TCP_MD5SIG md5 =tcp_rsk (req)->af_specific->req_md5_lookup (sk,req_to_sk (req));#endif skb_set_hash (skb,tcp_rsk (req)->txhash,PKT_HASH_TYPE_L4 );@@ -3196,8 +3201,8 @@ struct sk_buff *tcp_make_synack(const struct sock *sk, struct dst_entry *dst,if (md5)tcp_rsk (req)->af_specific->calc_md5_hash (opts.hash_location, md5,req_to_sk (req), skb); -rcu_read_unlock ();#endif +rcu_read_unlock ();/* Do not fool tcpdump (if any), clean our debris */ skb->tstamp =0 ;net/ipv6/syncookies.c +1
-1@@ -162,7 +162,7 @@ struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb)/* check for timestamp cookie support */ memset (&tcp_opt,0 ,sizeof (tcp_opt)); -tcp_parse_options (sock_net (sk), skb, &tcp_opt,0 ,NULL ); +tcp_parse_options (sock_net (sk), skb, &tcp_opt,0 ,NULL , tp);if (tcp_opt.saw_tstamp && tcp_opt.rcv_tsecr) { tsoff = secure_tcpv6_ts_off(sock_net(sk),net/ipv6/tcp_ipv6.c +26
-2@@ -784,9 +784,10 @@ static void tcp_v6_send_response(const struct sock *sk, struct sk_buff *skb, u32struct flowi6 fl6;struct net *net = sk ?sock_net (sk) :dev_net (skb_dst (skb)->dev);struct sock *ctl_sk = net->ipv6.tcp_sk; -unsigned int tot_len =sizeof (struct tcphdr ); +unsigned int tot_len =0 ;struct dst_entry *dst;__be32 *topt; +struct tcp_out_options extraopts;if (tsecr) tot_len +=TCPOLEN_TSTAMP_ALIGNED ;@@ -795,10 +796,28 @@ static void tcp_v6_send_response(const struct sock *sk, struct sk_buff *skb, u32 tot_len +=TCPOLEN_MD5SIG_ALIGNED ;#endif +rcu_read_lock (); +if (static_branch_unlikely (&tcp_extra_options_enabled)) { +unsigned int remaining =MAX_TCP_OPTION_SPACE - tot_len; +u8 extraflags = rst ?TCPHDR_RST :0 ; + +if (!rst || !th->ack) + extraflags |=TCPHDR_ACK ; + +memset (&extraopts,0 ,sizeof (extraopts)); + + tot_len +=tcp_extra_options_prepare (skb, extraflags, remaining, + &extraopts, sk); + } + + tot_len +=sizeof (struct tcphdr ); + buff =alloc_skb (MAX_HEADER +sizeof (struct ipv6hdr ) + tot_len,GFP_ATOMIC ); -if (!buff) +if (!buff) { +rcu_read_unlock ();return ; + }skb_reserve (buff,MAX_HEADER +sizeof (struct ipv6hdr ) + tot_len);@@ -835,6 +854,11 @@ static void tcp_v6_send_response(const struct sock *sk, struct sk_buff *skb, u32 }#endif +if (static_branch_unlikely (&tcp_extra_options_enabled)) +tcp_extra_options_write (topt, &extraopts, sk); + +rcu_read_unlock (); +memset (&fl6,0 ,sizeof (fl6)); fl6.daddr =ipv6_hdr (skb)->saddr; fl6.saddr =ipv6_hdr (skb)->daddr;Signature · 6 lines
2.14.2 _______________________________________________ mptcp mailing list mptcp(a)lists.01.org https://lists.01.org/mailman/listinfo/mptcp
Mat Martineau
Hi Christoph,
Quoted 229 lines · Christoph Paasch
Hello Mat,
Quoted 14 lines · Mat Martineau
Allow additional TCP options to be handled by registered hook functions.
Registered options have a priority that determines the order in which options are prepared and written. Lower priority numbers are handled first.
Option parsing will call the provided 'parse' function when a TCP option number is not recognized by the normal option parsing code.
The 'prepare' function determines the required space for registered options and store associated data. 'write' adds the option to the TCP header.
Quoted 8 lines
A static key and RCU synchronization are used to minimize the performance impact of these extensible TCP features.
Signed-off-by Mat Martineau <mathew.j.martineau(a)linux.intel.com>---
Changes from v1: One 'prepare' callback (no more special callback for
request_sock), and add a few missing callback sites (like ipv6).great, I like that we now have only one 'prepare' callback.
Quoted 56 lines
drivers/infiniband/hw/cxgb4/cm.c | 2 +- include/linux/tcp.h | 22 +++++++ include/net/tcp.h | 40 +++++++++++- net/ipv4/syncookies.c | 2 +- net/ipv4/tcp.c | 133 +++++++++++++++++++++++++++++++++++++++ net/ipv4/tcp_input.c | 16 +++-- net/ipv4/tcp_ipv4.c | 80 ++++++++++++++++++----- net/ipv4/tcp_minisocks.c | 4 +- net/ipv4/tcp_output.c | 43 +++++++------ net/ipv6/syncookies.c | 2 +- net/ipv6/tcp_ipv6.c | 28 ++++++++- 11 files changed, 323 insertions(+), 49 deletions(-)drivers/infiniband/hw/cxgb4/cm.c +1
-1@@ -3752,7 +3752,7 @@ static void build_cpl_pass_accept_req(struct sk_buff *skb, int stid , u8 tos)*/ memset (&tmp_opt,0 ,sizeof (tmp_opt));tcp_clear_options (&tmp_opt); -tcp_parse_options (&init_net, skb, &tmp_opt,0 ,NULL ); +tcp_parse_options (&init_net, skb, &tmp_opt,0 ,NULL ,NULL ); req =__skb_push (skb,sizeof (*req));memset (req,0 ,sizeof (*req));include/linux/tcp.h +17
-0@@ -112,6 +112,23 @@ static inline void tcp_clear_options(struct tcp_options_received *rx_opt) rx_opt->wscale_ok = rx_opt->snd_wscale =0 ; } +#define OPTION_SACK_ADVERTISE (1 << 0) +#define OPTION_TS (1 << 1) +#define OPTION_MD5 (1 << 2) +#define OPTION_WSCALE (1 << 3) +#define OPTION_FAST_OPEN_COOKIE (1 << 8) + +struct tcp_out_options { +u16 options;/* bit field of OPTION_* */ +u16 mss;/* 0 to disable */ +u8 ws;/* window scale, 0 to disable */ +u8 num_sack_blocks;/* number of SACK blocks to include */ +u8 hash_size;/* bytes in hash_location */ +__u8 *hash_location;/* temporary pointer, overloaded */ +__u32 tsval, tsecr;/* need to include OPTION_TS */ +struct tcp_fastopen_cookie *fastopen_cookie;/* Fast open cookie */ +}; +/* This is the max number of SACKS that we'll generate and process. It's safe * to increase this, although since: * size = TCPOLEN_SACK_BASE_ALIGNED (4) + n * TCPOLEN_SACK_PERBLOCK (8) include/linux/tcp.h static inline struct tcp_sock *tcp_sk(const struct sock *sk) · line 412
Quoted 7 lines
include/linux/tcp.h +4
-0@@ -389,6 +406,11 @@ static inline struct tcp_sock *tcp_sk(const struct sock *sk)return (struct tcp_sock *)sk; } +static inline struct sock *tcp_to_sk (const struct tcp_sock *tp) +{ +return (struct sock *)tp; +}Nice little function :)
Quoted 135 lines
include/linux/tcp.h +1
-0+struct tcp_timewait_sock {struct inet_timewait_sock tw_sk;#define tw_rcv_nxt tw_sk.__tw_common.skc_tw_rcv_nxtinclude/net/tcp.h +39
-1@@ -433,7 +433,8 @@ int tcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int nonblock, int flags, int *addr_len);void tcp_parse_options (const struct net *net,const struct sk_buff *skb,struct tcp_options_received *opt_rx, -int estab,struct tcp_fastopen_cookie *foc); +int estab,struct tcp_fastopen_cookie *foc, +struct tcp_sock *tp);const u8 *tcp_parse_md5sig_option (const struct tcphdr *th);/* @@ -2109,4 +2110,41 @@ static inline bool tcp_bpf_ca_needs_ecn(struct sock *sk) {return (tcp_call_bpf (sk,BPF_SOCK_OPS_NEEDS_ECN ) ==1 ); } + +extern struct static_key_false tcp_extra_options_enabled; + +struct tcp_extra_option_ops { +struct list_head list; +unsigned char option_kind; +unsigned char priority; +void (*parse)(int opsize,const unsigned char *opptr, +const struct sk_buff *skb, +struct tcp_options_received *opt_rx, +struct sock *sk); +/* Return the number of bytes consumed */ +unsigned int (*prepare)(struct sk_buff *skb,u8 flags, +unsigned int remaining, +struct tcp_out_options *opts, +const struct sock *sk); +void (*write)(__be32 *ptr,struct tcp_out_options *opts, +const struct sock *sk); +struct module *owner; +}; + +void tcp_extra_options_parse (int opcode,int opsize,const unsigned char *opptr, +const struct sk_buff *skb, +struct tcp_options_received *opt_rx, +struct sock *sk); + +unsigned int tcp_extra_options_prepare (struct sk_buff *skb,u8 flags, +unsigned int remaining, +struct tcp_out_options *opts, +const struct sock *sk); + +void tcp_extra_options_write (__be32 *ptr,struct tcp_out_options *opts, +const struct sock *sk); + +int tcp_register_extra_option (struct tcp_extra_option_ops *ops); +void tcp_unregister_extra_option (struct tcp_extra_option_ops *ops); +#endif /* _TCP_H */ net/ipv4/syncookies.c +1
-1@@ -313,7 +313,7 @@ struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb)/* check for timestamp cookie support */ memset (&tcp_opt,0 ,sizeof (tcp_opt)); -tcp_parse_options (sock_net (sk), skb, &tcp_opt,0 ,NULL ); +tcp_parse_options (sock_net (sk), skb, &tcp_opt,0 ,NULL , tp);if (tcp_opt.saw_tstamp && tcp_opt.rcv_tsecr) { tsoff = secure_tcp_ts_off(sock_net(sk),net/ipv4/tcp.c +40
-0@@ -270,6 +270,7 @@#include <linux/time.h> #include <linux/slab.h> #include <linux/errqueue.h> +#include <linux/static_key.h> #include <net/icmp.h> #include <net/inet_common.h> @@ -306,6 +307,13 @@ EXPORT_SYMBOL(tcp_memory_allocated);struct percpu_counter tcp_sockets_allocated;EXPORT_SYMBOL (tcp_sockets_allocated); +/* +* Optional TCP option handlers +*/ +static DEFINE_SPINLOCK (tcp_option_list_lock); +static LIST_HEAD (tcp_option_list); +DEFINE_STATIC_KEY_FALSE (tcp_extra_options_enabled); +/* * TCP splice context */ @@ -3375,6 +3383,130 @@ EXPORT_SYMBOL(tcp_md5_hash_key);#endif +/* Linear search, few entries are expected. The RCU read lock must +* be held before calling. +*/ +static struct tcp_extra_option_ops *tcp_extra_options_find_kind (unsigned char kind) +{ +struct tcp_extra_option_ops *entry; + +list_for_each_entry_rcu (entry, &tcp_option_list, list) { +if (entry->option_kind == kind) +return entry; + } + +return NULL ; +} + +void tcp_extra_options_parse (int opcode,int opsize,const unsigned char *opptr, +const struct sk_buff *skb, +struct tcp_options_received *opt_rx, +struct sock *sk) +{ +struct tcp_extra_option_ops *entry; + +rcu_read_lock (); + entry =tcp_extra_options_find_kind (opcode); +if (entry && entry->parse) + entry->parse (opsize, opptr, skb, opt_rx, sk); +rcu_read_unlock (); +} +EXPORT_SYMBOL_GPL (tcp_extra_options_parse); + +/* The RCU read lock must be held before calling, and should span both +* the call to this function and tcp_extra_options_write to ensure that
Quoted 8 lines
net/ipv4/tcp.c EXPORT_SYMBOL(tcp_md5_hash_key); · line 3420
Quoted 3 lines
net/ipv4/tcp.c +3
-0+* tcp_option_list does not change between the two calls. To preserve +* expected option alignment, always returns a multiple of 4 bytes. +*/ The RCU read lock won't be able to protect the list from being altered. All it will take care of is that the elements on the list won't get free'd while traversing it and that the next-pointer won't get changed. That way the list-traversal is still fine.
Thanks for the description. If we switch to per-socket lists then RCU won't be necessary, but I think it's possible to coax RCU in to the behavior I want if:
* the list is copied/rebuilt every time an option is registered or unregistered
and
* tcp_extra_options_prepare returns a pointer to the version of the list for tcp_extra_options_write to use
But then it might as well be an array. The list should be short enough and rarely modified, so it wouldn't be a lot of overhead.
Quoted 4 lines
If we move the extra-options to a mode where the list is on a per-TCP socket basis, we can avoid handling this. Because we can limit adding/removing TCP-options to happen only when the socket is in
TCP_CLOSE. So, once the connection started, the list will always remain the same.
I was thinking a global list might still be useful to avoid complicating request_sock, but now realize that incoming connections would use the extra options list from the listening socket.
Quoted 198 lines
Quoted 198 lines
net/ipv4/tcp.c +90
-0+unsigned int tcp_extra_options_prepare (struct sk_buff *skb,u8 flags, +unsigned int remaining, +struct tcp_out_options *opts, +const struct sock *sk) +{ +struct tcp_extra_option_ops *entry; +unsigned int used =0 ; + +list_for_each_entry_rcu (entry, &tcp_option_list, list) { +if (unlikely (!entry->prepare)) +continue ; + + used += entry->prepare (skb, flags, remaining - used, opts, sk); + } + +return roundup (used,4 ); +} +EXPORT_SYMBOL_GPL (tcp_extra_options_prepare); + +/* The RCU read lock must be held before calling, and should span both +* the call to tcp_extra_options_write and this function to ensure that +* tcp_option_list does not change between the two calls. +*/ +void tcp_extra_options_write (__be32 *ptr,struct tcp_out_options *opts, +const struct sock *sk) +{ +struct tcp_extra_option_ops *entry; + +list_for_each_entry_rcu (entry, &tcp_option_list, list) { +if (unlikely (!entry->write)) +continue ; + + entry->write (ptr, opts, sk); + } +} +EXPORT_SYMBOL_GPL (tcp_extra_options_write); + +int tcp_register_extra_option (struct tcp_extra_option_ops *ops) +{ +struct tcp_extra_option_ops *entry; +struct list_head * add_before = &tcp_option_list; +int ret =0 ; + +if (!ops->option_kind) +return -EINVAL ; + +if (!try_module_get (ops->owner)) +return -ENOENT ; + +spin_lock (&tcp_option_list_lock); + +list_for_each_entry_rcu (entry, &tcp_option_list, list) { +if (entry->option_kind == ops->option_kind) { +pr_notice ("Option kind %u already registered\n" , + ops->option_kind); +spin_unlock (&tcp_option_list_lock); +module_put (ops->owner); +return -EEXIST ; + } + +if (entry->priority <= ops->priority) + add_before = &entry->list; + } + +list_add_tail_rcu (&ops->list, add_before); +pr_debug ("Option kind %u registered\n" , ops->option_kind); + +spin_unlock (&tcp_option_list_lock); + +static_branch_inc (&tcp_extra_options_enabled); + +return ret; +} +EXPORT_SYMBOL_GPL (tcp_register_extra_option); + +void tcp_unregister_extra_option (struct tcp_extra_option_ops *ops) +{ +spin_lock (&tcp_option_list_lock); +list_del_rcu (&ops->list); +spin_unlock (&tcp_option_list_lock); + +synchronize_net (); + +static_branch_dec (&tcp_extra_options_enabled); + +module_put (ops->owner); +} +EXPORT_SYMBOL_GPL (tcp_unregister_extra_option); +void tcp_done (struct sock *sk) {struct request_sock *req =tcp_sk (sk)->fastopen_rsk;@@ -3521,6 +3653,7 @@ void __init tcp_init(void)INIT_HLIST_HEAD (&tcp_hashinfo.bhash[i].chain); } +INIT_LIST_HEAD (&tcp_option_list); cnt = tcp_hashinfo.ehash_mask +1 ; sysctl_tcp_max_orphans = cnt /2 ;net/ipv4/tcp_input.c +11
-5@@ -3728,7 +3728,7 @@ static void tcp_parse_fastopen_option(int len, const unsigned char *cookie,void tcp_parse_options (const struct net *net,const struct sk_buff *skb,struct tcp_options_received *opt_rx,int estab, -struct tcp_fastopen_cookie *foc) +struct tcp_fastopen_cookie *foc,struct tcp_sock *tp) {const unsigned char *ptr;const struct tcphdr *th =tcp_hdr (skb);@@ -3830,6 +3830,12 @@ void tcp_parse_options(const struct net *net, ptr + 2, th->syn, foc, true);break ; +default : +tcp_extra_options_parse (opcode, opsize, ptr, + skb, opt_rx, +tcp_to_sk (tp)); +break ; + } ptr += opsize-2 ; length -= opsize;@@ -3876,7 +3882,7 @@ static bool tcp_fast_parse_options(const struct net *net,return true; } -tcp_parse_options (net, skb, &tp->rx_opt,1 ,NULL ); +tcp_parse_options (net, skb, &tp->rx_opt,1 ,NULL , tp);if (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr) tp->rx_opt.rcv_tsecr -= tp->tsoffset;@@ -5569,7 +5575,7 @@ static bool tcp_rcv_fastopen_synack(struct sock *sk, struct sk_buff *synack,/* Get original SYNACK MSS value if user MSS sets mss_clamp */ tcp_clear_options (&opt); opt.user_mss = opt.mss_clamp =0 ; -tcp_parse_options (sock_net (sk), synack, &opt,0 ,NULL ); +tcp_parse_options (sock_net (sk), synack, &opt,0 ,NULL , tp); mss = opt.mss_clamp; }@@ -5623,7 +5629,7 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb,int saved_clamp = tp->rx_opt.mss_clamp;bool fastopen_fail; -tcp_parse_options (sock_net (sk), skb, &tp->rx_opt,0 , &foc); +tcp_parse_options (sock_net (sk), skb, &tp->rx_opt,0 , &foc, tp);if (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr) tp->rx_opt.rcv_tsecr -= tp->tsoffset;@@ -6299,7 +6305,7 @@ int tcp_conn_request(struct request_sock_ops *rsk_ops, tmp_opt.mss_clamp = af_ops->mss_clamp; tmp_opt.user_mss = tp->rx_opt.user_mss;tcp_parse_options (sock_net (sk), skb, &tmp_opt,0 , - want_cookie ?NULL : &foc); + want_cookie ?NULL : &foc, tp);if (want_cookie && !tmp_opt.saw_tstamp)tcp_clear_options (&tmp_opt);net/ipv4/tcp_ipv4.c +11
-3@@ -598,9 +598,7 @@ static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb)const struct tcphdr *th =tcp_hdr (skb);struct {struct tcphdr th; -#ifdef CONFIG_TCP_MD5SIG -__be32 opt[(TCPOLEN_MD5SIG_ALIGNED >>2 )]; -#endif +__be32 opt[(MAX_TCP_OPTION_SPACE >>2 )]; } rep;struct ip_reply_arg arg;#ifdef CONFIG_TCP_MD5SIG @@ -611,6 +609,7 @@ static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb)struct sock *sk1 =NULL ;#endif struct net *net; +int offset =0 ;/* Never send a reset in response to a reset. */ if (th->rst)@@ -676,17 +675,44 @@ static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb) goto out; } +#endif + +if (static_branch_unlikely (&tcp_extra_options_enabled)) { +unsigned int remaining; +unsigned int used; +struct tcp_out_options opts; + + remaining =sizeof (rep.opt); +#ifdef CONFIG_TCP_MD5SIG
Quoted 8 lines
net/ipv4/tcp_ipv4.c static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb) · line 689
Quoted 3 lines
net/ipv4/tcp_ipv4.c +3
-0+if (key) + remaining -=TCPOLEN_MD5SIG_ALIGNED ; +#endif We will break
TCP_MD5here with this patch if we move it inside the static branch.
Only after the patch that makesTCP_MD5adopt the framework we can move this code to here.
The MD5 option is handled outside the static branch (below), this bit of conditional code adjusts the space available to the extra options. The TCP options are not part of the MD5 digest, so it should be ok to write the extra options early.
Quoted 4 lines
Quoted 4 lines
net/ipv4/tcp_ipv4.c +4
-0+ +memset (&opts,0 ,sizeof (opts)); + +rcu_read_lock ();
Quoted 8 lines
net/ipv4/tcp_ipv4.c static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb) · line 701
Quoted 8 lines
net/ipv4/tcp_ipv4.c +8
-0+ used =tcp_extra_options_prepare (NULL ,TCPHDR_RST , remaining, + &opts, sk); + +tcp_extra_options_write (&rep.opt[0 ], &opts, sk); +rcu_read_unlock (); + + arg.iov[0 ].iov_len += used; + offset += used /4 ;
If any extra options were written, the MD5 option offset is adjusted here.
Quoted 2 lines
net/ipv4/tcp_ipv4.c static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb) · line 703
Quoted 2 lines
net/ipv4/tcp_ipv4.c +2
-0+ rep.th.doff = arg.iov[0 ].iov_len /4 ; + }
(End of static branch)
Quoted 7 lines
net/ipv4/tcp_ipv4.c static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb) · line 707
Quoted 7 lines
net/ipv4/tcp_ipv4.c +2
-4+#ifdef CONFIG_TCP_MD5SIG if (key) { - rep.opt[0 ] =htonl ((TCPOPT_NOP <<24 ) | - (TCPOPT_NOP <<16 ) | - (TCPOPT_MD5SIG <<8 ) | -TCPOLEN_MD5SIG ); + rep.opt[offset++] = htonl((TCPOPT_NOP << 24) |
The MD5 option will be written at the appropriate offset with or without the static branch above. Does that option need to be first? That could be accomodated.
Mat
Quoted 305 lines
Quoted 305 lines
net/ipv4/tcp_ipv4.c +33
-10+ (TCPOPT_NOP <<16 ) | + (TCPOPT_MD5SIG <<8 ) | + TCPOLEN_MD5SIG);/* Update length and the length the header thinks exists */ arg.iov[0 ].iov_len +=TCPOLEN_MD5SIG_ALIGNED ; rep.th.doff = arg.iov[0 ].iov_len /4 ; -tcp_v4_md5_hash_hdr ((__u8 *) &rep.opt[1 ], +tcp_v4_md5_hash_hdr ((__u8 *) &rep.opt[offset], key,ip_hdr (skb)->saddr,ip_hdr (skb)->daddr, &rep.th); }@@ -738,14 +764,11 @@ static void tcp_v4_send_ack(const struct sock *sk,const struct tcphdr *th =tcp_hdr (skb);struct {struct tcphdr th; -__be32 opt[(TCPOLEN_TSTAMP_ALIGNED >>2 ) -#ifdef CONFIG_TCP_MD5SIG - + (TCPOLEN_MD5SIG_ALIGNED >>2 ) -#endif - ]; +__be32 opt[(MAX_TCP_OPTION_SPACE >>2 )]; } rep;struct net *net =sock_net (sk);struct ip_reply_arg arg; +int offset =0 ;memset (&rep.th,0 ,sizeof (struct tcphdr ));memset (&arg,0 ,sizeof (arg));@@ -759,33 +782,56 @@ static void tcp_v4_send_ack(const struct sock *sk, rep.opt[1 ] =htonl (tsval); rep.opt[2 ] =htonl (tsecr); arg.iov[0 ].iov_len +=TCPOLEN_TSTAMP_ALIGNED ; + offset +=3 ; }/* Swap the send and the receive. */ rep.th.dest = th->source; rep.th.source = th->dest; - rep.th.doff = arg.iov[0 ].iov_len /4 ; rep.th.seq =htonl (seq); rep.th.ack_seq =htonl (ack); rep.th.ack =1 ; rep.th.window =htons (win); +if (static_branch_unlikely (&tcp_extra_options_enabled)) { +unsigned int remaining; +unsigned int used; +struct tcp_out_options opts; + + remaining =sizeof (rep.th) +sizeof (rep.opt) - arg.iov[0 ].iov_len;#ifdef CONFIG_TCP_MD5SIG -if (key) { -int offset = (tsecr) ?3 :0 ; +if (key) + remaining -=TCPOLEN_MD5SIG_ALIGNED ; +#endif +memset (&opts,0 ,sizeof (opts)); +rcu_read_lock (); + used =tcp_extra_options_prepare (NULL ,TCPHDR_ACK , remaining, + &opts, sk); + +tcp_extra_options_write (&rep.opt[offset], &opts, sk); +rcu_read_unlock (); + + arg.iov[0 ].iov_len += used; + offset += used /4 ; + } + + rep.th.doff = arg.iov[0 ].iov_len /4 ; + +#ifdef CONFIG_TCP_MD5SIG +if (key) { rep.opt[offset++] =htonl ((TCPOPT_NOP <<24 ) | (TCPOPT_NOP <<16 ) | (TCPOPT_MD5SIG <<8 ) |TCPOLEN_MD5SIG ); arg.iov[0 ].iov_len +=TCPOLEN_MD5SIG_ALIGNED ; - rep.th.doff = arg.iov[0 ].iov_len/4 ;tcp_v4_md5_hash_hdr ((__u8 *) &rep.opt[offset], key,ip_hdr (skb)->saddr,ip_hdr (skb)->daddr, &rep.th); }#endif + arg.flags = reply_flags; arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr, ip_hdr(skb)->saddr, /* XXX */net/ipv4/tcp_minisocks.c +2
-2@@ -98,7 +98,7 @@ tcp_timewait_state_process(struct inet_timewait_sock *tw, struct sk_buff *skb, tmp_opt.saw_tstamp =0 ;if (th->doff > (sizeof (*th) >>2 ) && tcptw->tw_ts_recent_stamp) { -tcp_parse_options (twsk_net (tw), skb, &tmp_opt,0 ,NULL ); +tcp_parse_options (twsk_net (tw), skb, &tmp_opt,0 ,NULL ,NULL );if (tmp_opt.saw_tstamp) {if (tmp_opt.rcv_tsecr)@@ -565,7 +565,7 @@ struct sock *tcp_check_req(struct sock *sk, struct sk_buff *skb, tmp_opt.saw_tstamp =0 ;if (th->doff > (sizeof (struct tcphdr )>>2 )) { -tcp_parse_options (sock_net (sk), skb, &tmp_opt,0 ,NULL ); +tcp_parse_options (sock_net (sk), skb, &tmp_opt,0 ,NULL ,NULL );if (tmp_opt.saw_tstamp) { tmp_opt.ts_recent = req->ts_recent;net/ipv4/tcp_output.c +24
-19@@ -41,6 +41,7 @@#include <linux/compiler.h> #include <linux/gfp.h> #include <linux/module.h> +#include <linux/static_key.h> /* People can turn this off for buggy TCP's found in printers etc. */ int sysctl_tcp_retrans_collapse__read_mostly =1 ;@@ -413,23 +414,6 @@ static inline bool tcp_urg_mode(const struct tcp_sock *tp)return tp->snd_una != tp->snd_up; } -#define OPTION_SACK_ADVERTISE (1 << 0) -#define OPTION_TS (1 << 1) -#define OPTION_MD5 (1 << 2) -#define OPTION_WSCALE (1 << 3) -#define OPTION_FAST_OPEN_COOKIE (1 << 8) - -struct tcp_out_options { -u16 options;/* bit field of OPTION_* */ -u16 mss;/* 0 to disable */ -u8 ws;/* window scale, 0 to disable */ -u8 num_sack_blocks;/* number of SACK blocks to include */ -u8 hash_size;/* bytes in hash_location */ -__u8 *hash_location;/* temporary pointer, overloaded */ -__u32 tsval, tsecr;/* need to include OPTION_TS */ -struct tcp_fastopen_cookie *fastopen_cookie;/* Fast open cookie */ -}; -/* Write previously computed TCP options to the packet. * * Beware: Something in the Internet is very sensitive to the ordering of @@ -536,6 +520,9 @@ static void tcp_options_write(__be32 *ptr, struct tcp_sock *tp, } ptr += (len +3 ) >>2 ; } + +if (static_branch_unlikely (&tcp_extra_options_enabled)) +tcp_extra_options_write (ptr, opts,tcp_to_sk (tp)); }/* Compute TCP options for SYN packets. This is not the final @@ -603,6 +590,11 @@ static unsigned int tcp_syn_options(struct sock *sk, struct sk_buff *skb, } } +if (static_branch_unlikely (&tcp_extra_options_enabled)) + remaining -=tcp_extra_options_prepare (skb,TCPHDR_SYN , + remaining, opts, +tcp_to_sk (tp)); +return MAX_TCP_OPTION_SPACE - remaining; }@@ -663,6 +655,12 @@ static unsigned int tcp_synack_options(struct request_sock *req, } } +if (static_branch_unlikely (&tcp_extra_options_enabled)) + remaining -=tcp_extra_options_prepare (skb, +TCPHDR_SYN |TCPHDR_ACK , + remaining, opts, +req_to_sk (req)); +return MAX_TCP_OPTION_SPACE - remaining; }@@ -696,6 +694,11 @@ static unsigned int tcp_established_options(struct sock *sk, struct sk_buff *skb size +=TCPOLEN_TSTAMP_ALIGNED ; } +if (static_branch_unlikely (&tcp_extra_options_enabled)) + size +=tcp_extra_options_prepare (skb,0 , +MAX_TCP_OPTION_SPACE - size, + opts,tcp_to_sk (tp)); + eff_sacks = tp->rx_opt.num_sacks + tp->rx_opt.dsack;if (unlikely (eff_sacks)) {const unsigned int remaining =MAX_TCP_OPTION_SPACE - size;@@ -1016,6 +1019,7 @@ static int tcp_transmit_skb(struct sock *sk, struct sk_buff *skb, int clone_it, tcb =TCP_SKB_CB (skb);memset (&opts,0 ,sizeof (opts)); +rcu_read_lock ();if (unlikely (tcb->tcp_flags &TCPHDR_SYN )) tcp_options_size =tcp_syn_options (sk, skb, &opts, &md5);else @@ -1092,6 +1096,7 @@ static int tcp_transmit_skb(struct sock *sk, struct sk_buff *skb, int clone_it, md5, sk, skb); }#endif +rcu_read_unlock (); icsk->icsk_af_ops->send_check (sk, skb);@@ -3156,8 +3161,8 @@ struct sk_buff *tcp_make_synack(const struct sock *sk, struct dst_entry *dst,#endif skb->skb_mstamp =tcp_clock_us (); -#ifdef CONFIG_TCP_MD5SIG rcu_read_lock (); +#ifdef CONFIG_TCP_MD5SIG md5 =tcp_rsk (req)->af_specific->req_md5_lookup (sk,req_to_sk (req));#endif skb_set_hash (skb,tcp_rsk (req)->txhash,PKT_HASH_TYPE_L4 );@@ -3196,8 +3201,8 @@ struct sk_buff *tcp_make_synack(const struct sock *sk, struct dst_entry *dst,if (md5)tcp_rsk (req)->af_specific->calc_md5_hash (opts.hash_location, md5,req_to_sk (req), skb); -rcu_read_unlock ();#endif +rcu_read_unlock ();/* Do not fool tcpdump (if any), clean our debris */ skb->tstamp =0 ;net/ipv6/syncookies.c +1
-1@@ -162,7 +162,7 @@ struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb)/* check for timestamp cookie support */ memset (&tcp_opt,0 ,sizeof (tcp_opt)); -tcp_parse_options (sock_net (sk), skb, &tcp_opt,0 ,NULL ); +tcp_parse_options (sock_net (sk), skb, &tcp_opt,0 ,NULL , tp);if (tcp_opt.saw_tstamp && tcp_opt.rcv_tsecr) { tsoff = secure_tcpv6_ts_off(sock_net(sk),net/ipv6/tcp_ipv6.c +26
-3@@ -784,9 +784,10 @@ static void tcp_v6_send_response(const struct sock *sk, struct sk_buff *skb, u32struct flowi6 fl6;struct net *net = sk ?sock_net (sk) :dev_net (skb_dst (skb)->dev);struct sock *ctl_sk = net->ipv6.tcp_sk; -unsigned int tot_len =sizeof (struct tcphdr ); +unsigned int tot_len =0 ;struct dst_entry *dst;__be32 *topt; +struct tcp_out_options extraopts;if (tsecr) tot_len +=TCPOLEN_TSTAMP_ALIGNED ;@@ -795,10 +796,28 @@ static void tcp_v6_send_response(const struct sock *sk, struct sk_buff *skb, u32 tot_len +=TCPOLEN_MD5SIG_ALIGNED ;#endif +rcu_read_lock (); +if (static_branch_unlikely (&tcp_extra_options_enabled)) { +unsigned int remaining =MAX_TCP_OPTION_SPACE - tot_len; +u8 extraflags = rst ?TCPHDR_RST :0 ; + +if (!rst || !th->ack) + extraflags |=TCPHDR_ACK ; + +memset (&extraopts,0 ,sizeof (extraopts)); + + tot_len +=tcp_extra_options_prepare (skb, extraflags, remaining, + &extraopts, sk); + } + + tot_len +=sizeof (struct tcphdr ); + buff =alloc_skb (MAX_HEADER +sizeof (struct ipv6hdr ) + tot_len,GFP_ATOMIC ); -if (!buff) +if (!buff) { +rcu_read_unlock ();return ; + }skb_reserve (buff,MAX_HEADER +sizeof (struct ipv6hdr ) + tot_len);@@ -835,6 +854,11 @@ static void tcp_v6_send_response(const struct sock *sk, struct sk_buff *skb, u32 }#endif +if (static_branch_unlikely (&tcp_extra_options_enabled)) +tcp_extra_options_write (topt, &extraopts, sk); + +rcu_read_unlock (); +memset (&fl6,0 ,sizeof (fl6)); fl6.daddr =ipv6_hdr (skb)->saddr; fl6.saddr =ipv6_hdr (skb)->daddr; --2.14.2
Quoted 5 lines
Quoted 4 lines
_______________________________________________
mptcp mailing list
mptcp(a)lists.01.org
https://lists.01.org/mailman/listinfo/mptcp
--
Mat Martineau
Intel OTC
Christoph Paasch
Hello,
Quoted 252 lines · Mat Martineau
Quoted 227 lines · Christoph Paasch
Quoted 14 lines · Mat Martineau
Allow additional TCP options to be handled by registered hook functions.
Registered options have a priority that determines the order in which options are prepared and written. Lower priority numbers are handled first.
Option parsing will call the provided 'parse' function when a TCP option number is not recognized by the normal option parsing code.
The 'prepare' function determines the required space for registered options and store associated data. 'write' adds the option to the TCP header.
Quoted 8 lines
A static key and RCU synchronization are used to minimize the performance impact of these extensible TCP features.
Signed-off-by Mat Martineau <mathew.j.martineau(a)linux.intel.com>---
Changes from v1: One 'prepare' callback (no more special callback for
request_sock), and add a few missing callback sites (like ipv6).great, I like that we now have only one 'prepare' callback.
Quoted 56 lines
drivers/infiniband/hw/cxgb4/cm.c | 2 +- include/linux/tcp.h | 22 +++++++ include/net/tcp.h | 40 +++++++++++- net/ipv4/syncookies.c | 2 +- net/ipv4/tcp.c | 133 +++++++++++++++++++++++++++++++++++++++ net/ipv4/tcp_input.c | 16 +++-- net/ipv4/tcp_ipv4.c | 80 ++++++++++++++++++----- net/ipv4/tcp_minisocks.c | 4 +- net/ipv4/tcp_output.c | 43 +++++++------ net/ipv6/syncookies.c | 2 +- net/ipv6/tcp_ipv6.c | 28 ++++++++- 11 files changed, 323 insertions(+), 49 deletions(-)drivers/infiniband/hw/cxgb4/cm.c +1
-1@@ -3752,7 +3752,7 @@ static void build_cpl_pass_accept_req(struct sk_buff *skb, int stid , u8 tos)*/ memset (&tmp_opt,0 ,sizeof (tmp_opt));tcp_clear_options (&tmp_opt); -tcp_parse_options (&init_net, skb, &tmp_opt,0 ,NULL ); +tcp_parse_options (&init_net, skb, &tmp_opt,0 ,NULL ,NULL ); req =__skb_push (skb,sizeof (*req));memset (req,0 ,sizeof (*req));include/linux/tcp.h +17
-0@@ -112,6 +112,23 @@ static inline void tcp_clear_options(struct tcp_options_received *rx_opt) rx_opt->wscale_ok = rx_opt->snd_wscale =0 ; } +#define OPTION_SACK_ADVERTISE (1 << 0) +#define OPTION_TS (1 << 1) +#define OPTION_MD5 (1 << 2) +#define OPTION_WSCALE (1 << 3) +#define OPTION_FAST_OPEN_COOKIE (1 << 8) + +struct tcp_out_options { +u16 options;/* bit field of OPTION_* */ +u16 mss;/* 0 to disable */ +u8 ws;/* window scale, 0 to disable */ +u8 num_sack_blocks;/* number of SACK blocks to include */ +u8 hash_size;/* bytes in hash_location */ +__u8 *hash_location;/* temporary pointer, overloaded */ +__u32 tsval, tsecr;/* need to include OPTION_TS */ +struct tcp_fastopen_cookie *fastopen_cookie;/* Fast open cookie */ +}; +/* This is the max number of SACKS that we'll generate and process. It's safe * to increase this, although since: * size = TCPOLEN_SACK_BASE_ALIGNED (4) + n * TCPOLEN_SACK_PERBLOCK (8) include/linux/tcp.h static inline struct tcp_sock *tcp_sk(const struct sock *sk) · line 412
Quoted 7 lines
include/linux/tcp.h +4
-0@@ -389,6 +406,11 @@ static inline struct tcp_sock *tcp_sk(const struct sock *sk)return (struct tcp_sock *)sk; } +static inline struct sock *tcp_to_sk (const struct tcp_sock *tp) +{ +return (struct sock *)tp; +}Nice little function :)
Quoted 135 lines
include/linux/tcp.h +1
-0+struct tcp_timewait_sock {struct inet_timewait_sock tw_sk;#define tw_rcv_nxt tw_sk.__tw_common.skc_tw_rcv_nxtinclude/net/tcp.h +39
-1@@ -433,7 +433,8 @@ int tcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int nonblock, int flags, int *addr_len);void tcp_parse_options (const struct net *net,const struct sk_buff *skb,struct tcp_options_received *opt_rx, -int estab,struct tcp_fastopen_cookie *foc); +int estab,struct tcp_fastopen_cookie *foc, +struct tcp_sock *tp);const u8 *tcp_parse_md5sig_option (const struct tcphdr *th);/* @@ -2109,4 +2110,41 @@ static inline bool tcp_bpf_ca_needs_ecn(struct sock *sk) {return (tcp_call_bpf (sk,BPF_SOCK_OPS_NEEDS_ECN ) ==1 ); } + +extern struct static_key_false tcp_extra_options_enabled; + +struct tcp_extra_option_ops { +struct list_head list; +unsigned char option_kind; +unsigned char priority; +void (*parse)(int opsize,const unsigned char *opptr, +const struct sk_buff *skb, +struct tcp_options_received *opt_rx, +struct sock *sk); +/* Return the number of bytes consumed */ +unsigned int (*prepare)(struct sk_buff *skb,u8 flags, +unsigned int remaining, +struct tcp_out_options *opts, +const struct sock *sk); +void (*write)(__be32 *ptr,struct tcp_out_options *opts, +const struct sock *sk); +struct module *owner; +}; + +void tcp_extra_options_parse (int opcode,int opsize,const unsigned char *opptr, +const struct sk_buff *skb, +struct tcp_options_received *opt_rx, +struct sock *sk); + +unsigned int tcp_extra_options_prepare (struct sk_buff *skb,u8 flags, +unsigned int remaining, +struct tcp_out_options *opts, +const struct sock *sk); + +void tcp_extra_options_write (__be32 *ptr,struct tcp_out_options *opts, +const struct sock *sk); + +int tcp_register_extra_option (struct tcp_extra_option_ops *ops); +void tcp_unregister_extra_option (struct tcp_extra_option_ops *ops); +#endif /* _TCP_H */ net/ipv4/syncookies.c +1
-1@@ -313,7 +313,7 @@ struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb)/* check for timestamp cookie support */ memset (&tcp_opt,0 ,sizeof (tcp_opt)); -tcp_parse_options (sock_net (sk), skb, &tcp_opt,0 ,NULL ); +tcp_parse_options (sock_net (sk), skb, &tcp_opt,0 ,NULL , tp);if (tcp_opt.saw_tstamp && tcp_opt.rcv_tsecr) { tsoff = secure_tcp_ts_off(sock_net(sk),net/ipv4/tcp.c +40
-0@@ -270,6 +270,7 @@#include <linux/time.h> #include <linux/slab.h> #include <linux/errqueue.h> +#include <linux/static_key.h> #include <net/icmp.h> #include <net/inet_common.h> @@ -306,6 +307,13 @@ EXPORT_SYMBOL(tcp_memory_allocated);struct percpu_counter tcp_sockets_allocated;EXPORT_SYMBOL (tcp_sockets_allocated); +/* +* Optional TCP option handlers +*/ +static DEFINE_SPINLOCK (tcp_option_list_lock); +static LIST_HEAD (tcp_option_list); +DEFINE_STATIC_KEY_FALSE (tcp_extra_options_enabled); +/* * TCP splice context */ @@ -3375,6 +3383,130 @@ EXPORT_SYMBOL(tcp_md5_hash_key);#endif +/* Linear search, few entries are expected. The RCU read lock must +* be held before calling. +*/ +static struct tcp_extra_option_ops *tcp_extra_options_find_kind (unsigned char kind) +{ +struct tcp_extra_option_ops *entry; + +list_for_each_entry_rcu (entry, &tcp_option_list, list) { +if (entry->option_kind == kind) +return entry; + } + +return NULL ; +} + +void tcp_extra_options_parse (int opcode,int opsize,const unsigned char *opptr, +const struct sk_buff *skb, +struct tcp_options_received *opt_rx, +struct sock *sk) +{ +struct tcp_extra_option_ops *entry; + +rcu_read_lock (); + entry =tcp_extra_options_find_kind (opcode); +if (entry && entry->parse) + entry->parse (opsize, opptr, skb, opt_rx, sk); +rcu_read_unlock (); +} +EXPORT_SYMBOL_GPL (tcp_extra_options_parse); + +/* The RCU read lock must be held before calling, and should span both +* the call to this function and tcp_extra_options_write to ensure that Quoted 8 lines
net/ipv4/tcp.c EXPORT_SYMBOL(tcp_md5_hash_key); · line 3420
Quoted 3 lines
net/ipv4/tcp.c +3
-0+* tcp_option_list does not change between the two calls. To preserve +* expected option alignment, always returns a multiple of 4 bytes. +*/ The RCU read lock won't be able to protect the list from being altered. All it will take care of is that the elements on the list won't get free'd while traversing it and that the next-pointer won't get changed. That way the list-traversal is still fine.
Thanks for the description. If we switch to per-socket lists then RCU won't be necessary, but I think it's possible to coax RCU in to the behavior I want if:
* the list is copied/rebuilt every time an option is registered or unregistered
and
*
tcp_extra_options_preparereturns a pointer to the version of the list fortcp_extra_options_writeto useBut then it might as well be an array. The list should be short enough and rarely modified, so it wouldn't be a lot of overhead.
Quoted 8 lines
Quoted 4 lines
If we move the extra-options to a mode where the list is on a per-TCP socket basis, we can avoid handling this. Because we can limit adding/removing TCP-options to happen only when the socket is in
TCP_CLOSE. So, once the connection started, the list will always remain the same.I was thinking a global list might still be useful to avoid complicating
request_sock, but now realize that incoming connections would use the extra options list from the listening socket.
I actually think we should copy the option-list over to the request-sock when it gets created. For MD5 it's not absolutely necessary (the keys are anyways already stored in the listener for each peer-IP address).
However, for MPTCP it would be useful. Because, MPTCP requires state per request-sock. And if the list is in the request-sock, we can store the state in there (thus avoid allocating a struct mptcp_request_sock
Quoted 207 lines
Quoted 198 lines
Quoted 198 lines
net/ipv4/tcp.c +90
-0+unsigned int tcp_extra_options_prepare (struct sk_buff *skb,u8 flags, +unsigned int remaining, +struct tcp_out_options *opts, +const struct sock *sk) +{ +struct tcp_extra_option_ops *entry; +unsigned int used =0 ; + +list_for_each_entry_rcu (entry, &tcp_option_list, list) { +if (unlikely (!entry->prepare)) +continue ; + + used += entry->prepare (skb, flags, remaining - used, opts, sk); + } + +return roundup (used,4 ); +} +EXPORT_SYMBOL_GPL (tcp_extra_options_prepare); + +/* The RCU read lock must be held before calling, and should span both +* the call to tcp_extra_options_write and this function to ensure that +* tcp_option_list does not change between the two calls. +*/ +void tcp_extra_options_write (__be32 *ptr,struct tcp_out_options *opts, +const struct sock *sk) +{ +struct tcp_extra_option_ops *entry; + +list_for_each_entry_rcu (entry, &tcp_option_list, list) { +if (unlikely (!entry->write)) +continue ; + + entry->write (ptr, opts, sk); + } +} +EXPORT_SYMBOL_GPL (tcp_extra_options_write); + +int tcp_register_extra_option (struct tcp_extra_option_ops *ops) +{ +struct tcp_extra_option_ops *entry; +struct list_head * add_before = &tcp_option_list; +int ret =0 ; + +if (!ops->option_kind) +return -EINVAL ; + +if (!try_module_get (ops->owner)) +return -ENOENT ; + +spin_lock (&tcp_option_list_lock); + +list_for_each_entry_rcu (entry, &tcp_option_list, list) { +if (entry->option_kind == ops->option_kind) { +pr_notice ("Option kind %u already registered\n" , + ops->option_kind); +spin_unlock (&tcp_option_list_lock); +module_put (ops->owner); +return -EEXIST ; + } + +if (entry->priority <= ops->priority) + add_before = &entry->list; + } + +list_add_tail_rcu (&ops->list, add_before); +pr_debug ("Option kind %u registered\n" , ops->option_kind); + +spin_unlock (&tcp_option_list_lock); + +static_branch_inc (&tcp_extra_options_enabled); + +return ret; +} +EXPORT_SYMBOL_GPL (tcp_register_extra_option); + +void tcp_unregister_extra_option (struct tcp_extra_option_ops *ops) +{ +spin_lock (&tcp_option_list_lock); +list_del_rcu (&ops->list); +spin_unlock (&tcp_option_list_lock); + +synchronize_net (); + +static_branch_dec (&tcp_extra_options_enabled); + +module_put (ops->owner); +} +EXPORT_SYMBOL_GPL (tcp_unregister_extra_option); +void tcp_done (struct sock *sk) {struct request_sock *req =tcp_sk (sk)->fastopen_rsk;@@ -3521,6 +3653,7 @@ void __init tcp_init(void)INIT_HLIST_HEAD (&tcp_hashinfo.bhash[i].chain); } +INIT_LIST_HEAD (&tcp_option_list); cnt = tcp_hashinfo.ehash_mask +1 ; sysctl_tcp_max_orphans = cnt /2 ;net/ipv4/tcp_input.c +11
-5@@ -3728,7 +3728,7 @@ static void tcp_parse_fastopen_option(int len, const unsigned char *cookie,void tcp_parse_options (const struct net *net,const struct sk_buff *skb,struct tcp_options_received *opt_rx,int estab, -struct tcp_fastopen_cookie *foc) +struct tcp_fastopen_cookie *foc,struct tcp_sock *tp) {const unsigned char *ptr;const struct tcphdr *th =tcp_hdr (skb);@@ -3830,6 +3830,12 @@ void tcp_parse_options(const struct net *net, ptr + 2, th->syn, foc, true);break ; +default : +tcp_extra_options_parse (opcode, opsize, ptr, + skb, opt_rx, +tcp_to_sk (tp)); +break ; + } ptr += opsize-2 ; length -= opsize;@@ -3876,7 +3882,7 @@ static bool tcp_fast_parse_options(const struct net *net,return true; } -tcp_parse_options (net, skb, &tp->rx_opt,1 ,NULL ); +tcp_parse_options (net, skb, &tp->rx_opt,1 ,NULL , tp);if (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr) tp->rx_opt.rcv_tsecr -= tp->tsoffset;@@ -5569,7 +5575,7 @@ static bool tcp_rcv_fastopen_synack(struct sock *sk, struct sk_buff *synack,/* Get original SYNACK MSS value if user MSS sets mss_clamp */ tcp_clear_options (&opt); opt.user_mss = opt.mss_clamp =0 ; -tcp_parse_options (sock_net (sk), synack, &opt,0 ,NULL ); +tcp_parse_options (sock_net (sk), synack, &opt,0 ,NULL , tp); mss = opt.mss_clamp; }@@ -5623,7 +5629,7 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb,int saved_clamp = tp->rx_opt.mss_clamp;bool fastopen_fail; -tcp_parse_options (sock_net (sk), skb, &tp->rx_opt,0 , &foc); +tcp_parse_options (sock_net (sk), skb, &tp->rx_opt,0 , &foc, tp);if (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr) tp->rx_opt.rcv_tsecr -= tp->tsoffset;@@ -6299,7 +6305,7 @@ int tcp_conn_request(struct request_sock_ops *rsk_ops, tmp_opt.mss_clamp = af_ops->mss_clamp; tmp_opt.user_mss = tp->rx_opt.user_mss;tcp_parse_options (sock_net (sk), skb, &tmp_opt,0 , - want_cookie ?NULL : &foc); + want_cookie ?NULL : &foc, tp);if (want_cookie && !tmp_opt.saw_tstamp)tcp_clear_options (&tmp_opt);net/ipv4/tcp_ipv4.c +11
-3@@ -598,9 +598,7 @@ static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb)const struct tcphdr *th =tcp_hdr (skb);struct {struct tcphdr th; -#ifdef CONFIG_TCP_MD5SIG -__be32 opt[(TCPOLEN_MD5SIG_ALIGNED >>2 )]; -#endif +__be32 opt[(MAX_TCP_OPTION_SPACE >>2 )]; } rep;struct ip_reply_arg arg;#ifdef CONFIG_TCP_MD5SIG @@ -611,6 +609,7 @@ static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb)struct sock *sk1 =NULL ;#endif struct net *net; +int offset =0 ;/* Never send a reset in response to a reset. */ if (th->rst)@@ -676,17 +675,44 @@ static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb) goto out; } +#endif + +if (static_branch_unlikely (&tcp_extra_options_enabled)) { +unsigned int remaining; +unsigned int used; +struct tcp_out_options opts; + + remaining =sizeof (rep.opt); +#ifdef CONFIG_TCP_MD5SIG Quoted 8 lines
net/ipv4/tcp_ipv4.c static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb) · line 689
Quoted 3 lines
net/ipv4/tcp_ipv4.c +3
-0+if (key) + remaining -=TCPOLEN_MD5SIG_ALIGNED ; +#endif We will break
TCP_MD5here with this patch if we move it inside the static branch.
Only after the patch that makesTCP_MD5adopt the framework we can move this code to here.
Quoted 4 lines
The MD5 option is handled outside the static branch (below), this bit of conditional code adjusts the space available to the extra options. The TCP options are not part of the MD5 digest, so it should be ok to write the extra options early.
Ok, I see. Now I understand. Yes, it's fine then :)
Christoph
Quoted 351 lines
Quoted 13 lines
net/ipv4/tcp_ipv4.c static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb) · line 701
Quoted 13 lines
net/ipv4/tcp_ipv4.c +12
-0+ +memset (&opts,0 ,sizeof (opts)); + +rcu_read_lock (); + used =tcp_extra_options_prepare (NULL ,TCPHDR_RST , remaining, + &opts, sk); + +tcp_extra_options_write (&rep.opt[0 ], &opts, sk); +rcu_read_unlock (); + + arg.iov[0 ].iov_len += used; + offset += used /4 ;If any extra options were written, the MD5 option offset is adjusted here.
Quoted 2 lines
net/ipv4/tcp_ipv4.c static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb) · line 703
Quoted 2 lines
net/ipv4/tcp_ipv4.c +2
-0+ rep.th.doff = arg.iov[0 ].iov_len /4 ; + }(End of static branch)
Quoted 7 lines
net/ipv4/tcp_ipv4.c static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb) · line 707
Quoted 7 lines
net/ipv4/tcp_ipv4.c +2
-4+#ifdef CONFIG_TCP_MD5SIG if (key) { - rep.opt[0 ] =htonl ((TCPOPT_NOP <<24 ) | - (TCPOPT_NOP <<16 ) | - (TCPOPT_MD5SIG <<8 ) | -TCPOLEN_MD5SIG ); + rep.opt[offset++] = htonl((TCPOPT_NOP << 24) |The MD5 option will be written at the appropriate offset with or without the static branch above. Does that option need to be first? That could be accomodated.
Mat
Quoted 305 lines
Quoted 305 lines
net/ipv4/tcp_ipv4.c +33
-10+ (TCPOPT_NOP <<16 ) | + (TCPOPT_MD5SIG <<8 ) | + TCPOLEN_MD5SIG);/* Update length and the length the header thinks exists */ arg.iov[0 ].iov_len +=TCPOLEN_MD5SIG_ALIGNED ; rep.th.doff = arg.iov[0 ].iov_len /4 ; -tcp_v4_md5_hash_hdr ((__u8 *) &rep.opt[1 ], +tcp_v4_md5_hash_hdr ((__u8 *) &rep.opt[offset], key,ip_hdr (skb)->saddr,ip_hdr (skb)->daddr, &rep.th); }@@ -738,14 +764,11 @@ static void tcp_v4_send_ack(const struct sock *sk,const struct tcphdr *th =tcp_hdr (skb);struct {struct tcphdr th; -__be32 opt[(TCPOLEN_TSTAMP_ALIGNED >>2 ) -#ifdef CONFIG_TCP_MD5SIG - + (TCPOLEN_MD5SIG_ALIGNED >>2 ) -#endif - ]; +__be32 opt[(MAX_TCP_OPTION_SPACE >>2 )]; } rep;struct net *net =sock_net (sk);struct ip_reply_arg arg; +int offset =0 ;memset (&rep.th,0 ,sizeof (struct tcphdr ));memset (&arg,0 ,sizeof (arg));@@ -759,33 +782,56 @@ static void tcp_v4_send_ack(const struct sock *sk, rep.opt[1 ] =htonl (tsval); rep.opt[2 ] =htonl (tsecr); arg.iov[0 ].iov_len +=TCPOLEN_TSTAMP_ALIGNED ; + offset +=3 ; }/* Swap the send and the receive. */ rep.th.dest = th->source; rep.th.source = th->dest; - rep.th.doff = arg.iov[0 ].iov_len /4 ; rep.th.seq =htonl (seq); rep.th.ack_seq =htonl (ack); rep.th.ack =1 ; rep.th.window =htons (win); +if (static_branch_unlikely (&tcp_extra_options_enabled)) { +unsigned int remaining; +unsigned int used; +struct tcp_out_options opts; + + remaining =sizeof (rep.th) +sizeof (rep.opt) - arg.iov[0 ].iov_len;#ifdef CONFIG_TCP_MD5SIG -if (key) { -int offset = (tsecr) ?3 :0 ; +if (key) + remaining -=TCPOLEN_MD5SIG_ALIGNED ; +#endif +memset (&opts,0 ,sizeof (opts)); +rcu_read_lock (); + used =tcp_extra_options_prepare (NULL ,TCPHDR_ACK , remaining, + &opts, sk); + +tcp_extra_options_write (&rep.opt[offset], &opts, sk); +rcu_read_unlock (); + + arg.iov[0 ].iov_len += used; + offset += used /4 ; + } + + rep.th.doff = arg.iov[0 ].iov_len /4 ; + +#ifdef CONFIG_TCP_MD5SIG +if (key) { rep.opt[offset++] =htonl ((TCPOPT_NOP <<24 ) | (TCPOPT_NOP <<16 ) | (TCPOPT_MD5SIG <<8 ) |TCPOLEN_MD5SIG ); arg.iov[0 ].iov_len +=TCPOLEN_MD5SIG_ALIGNED ; - rep.th.doff = arg.iov[0 ].iov_len/4 ;tcp_v4_md5_hash_hdr ((__u8 *) &rep.opt[offset], key,ip_hdr (skb)->saddr,ip_hdr (skb)->daddr, &rep.th); }#endif + arg.flags = reply_flags; arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr, ip_hdr(skb)->saddr, /* XXX */net/ipv4/tcp_minisocks.c +2
-2@@ -98,7 +98,7 @@ tcp_timewait_state_process(struct inet_timewait_sock *tw, struct sk_buff *skb, tmp_opt.saw_tstamp =0 ;if (th->doff > (sizeof (*th) >>2 ) && tcptw->tw_ts_recent_stamp) { -tcp_parse_options (twsk_net (tw), skb, &tmp_opt,0 ,NULL ); +tcp_parse_options (twsk_net (tw), skb, &tmp_opt,0 ,NULL ,NULL );if (tmp_opt.saw_tstamp) {if (tmp_opt.rcv_tsecr)@@ -565,7 +565,7 @@ struct sock *tcp_check_req(struct sock *sk, struct sk_buff *skb, tmp_opt.saw_tstamp =0 ;if (th->doff > (sizeof (struct tcphdr )>>2 )) { -tcp_parse_options (sock_net (sk), skb, &tmp_opt,0 ,NULL ); +tcp_parse_options (sock_net (sk), skb, &tmp_opt,0 ,NULL ,NULL );if (tmp_opt.saw_tstamp) { tmp_opt.ts_recent = req->ts_recent;net/ipv4/tcp_output.c +24
-19@@ -41,6 +41,7 @@#include <linux/compiler.h> #include <linux/gfp.h> #include <linux/module.h> +#include <linux/static_key.h> /* People can turn this off for buggy TCP's found in printers etc. */ int sysctl_tcp_retrans_collapse__read_mostly =1 ;@@ -413,23 +414,6 @@ static inline bool tcp_urg_mode(const struct tcp_sock *tp)return tp->snd_una != tp->snd_up; } -#define OPTION_SACK_ADVERTISE (1 << 0) -#define OPTION_TS (1 << 1) -#define OPTION_MD5 (1 << 2) -#define OPTION_WSCALE (1 << 3) -#define OPTION_FAST_OPEN_COOKIE (1 << 8) - -struct tcp_out_options { -u16 options;/* bit field of OPTION_* */ -u16 mss;/* 0 to disable */ -u8 ws;/* window scale, 0 to disable */ -u8 num_sack_blocks;/* number of SACK blocks to include */ -u8 hash_size;/* bytes in hash_location */ -__u8 *hash_location;/* temporary pointer, overloaded */ -__u32 tsval, tsecr;/* need to include OPTION_TS */ -struct tcp_fastopen_cookie *fastopen_cookie;/* Fast open cookie */ -}; -/* Write previously computed TCP options to the packet. * * Beware: Something in the Internet is very sensitive to the ordering of @@ -536,6 +520,9 @@ static void tcp_options_write(__be32 *ptr, struct tcp_sock *tp, } ptr += (len +3 ) >>2 ; } + +if (static_branch_unlikely (&tcp_extra_options_enabled)) +tcp_extra_options_write (ptr, opts,tcp_to_sk (tp)); }/* Compute TCP options for SYN packets. This is not the final @@ -603,6 +590,11 @@ static unsigned int tcp_syn_options(struct sock *sk, struct sk_buff *skb, } } +if (static_branch_unlikely (&tcp_extra_options_enabled)) + remaining -=tcp_extra_options_prepare (skb,TCPHDR_SYN , + remaining, opts, +tcp_to_sk (tp)); +return MAX_TCP_OPTION_SPACE - remaining; }@@ -663,6 +655,12 @@ static unsigned int tcp_synack_options(struct request_sock *req, } } +if (static_branch_unlikely (&tcp_extra_options_enabled)) + remaining -=tcp_extra_options_prepare (skb, +TCPHDR_SYN |TCPHDR_ACK , + remaining, opts, +req_to_sk (req)); +return MAX_TCP_OPTION_SPACE - remaining; }@@ -696,6 +694,11 @@ static unsigned int tcp_established_options(struct sock *sk, struct sk_buff *skb size +=TCPOLEN_TSTAMP_ALIGNED ; } +if (static_branch_unlikely (&tcp_extra_options_enabled)) + size +=tcp_extra_options_prepare (skb,0 , +MAX_TCP_OPTION_SPACE - size, + opts,tcp_to_sk (tp)); + eff_sacks = tp->rx_opt.num_sacks + tp->rx_opt.dsack;if (unlikely (eff_sacks)) {const unsigned int remaining =MAX_TCP_OPTION_SPACE - size;@@ -1016,6 +1019,7 @@ static int tcp_transmit_skb(struct sock *sk, struct sk_buff *skb, int clone_it, tcb =TCP_SKB_CB (skb);memset (&opts,0 ,sizeof (opts)); +rcu_read_lock ();if (unlikely (tcb->tcp_flags &TCPHDR_SYN )) tcp_options_size =tcp_syn_options (sk, skb, &opts, &md5);else @@ -1092,6 +1096,7 @@ static int tcp_transmit_skb(struct sock *sk, struct sk_buff *skb, int clone_it, md5, sk, skb); }#endif +rcu_read_unlock (); icsk->icsk_af_ops->send_check (sk, skb);@@ -3156,8 +3161,8 @@ struct sk_buff *tcp_make_synack(const struct sock *sk, struct dst_entry *dst,#endif skb->skb_mstamp =tcp_clock_us (); -#ifdef CONFIG_TCP_MD5SIG rcu_read_lock (); +#ifdef CONFIG_TCP_MD5SIG md5 =tcp_rsk (req)->af_specific->req_md5_lookup (sk,req_to_sk (req));#endif skb_set_hash (skb,tcp_rsk (req)->txhash,PKT_HASH_TYPE_L4 );@@ -3196,8 +3201,8 @@ struct sk_buff *tcp_make_synack(const struct sock *sk, struct dst_entry *dst,if (md5)tcp_rsk (req)->af_specific->calc_md5_hash (opts.hash_location, md5,req_to_sk (req), skb); -rcu_read_unlock ();#endif +rcu_read_unlock ();/* Do not fool tcpdump (if any), clean our debris */ skb->tstamp =0 ;net/ipv6/syncookies.c +1
-1@@ -162,7 +162,7 @@ struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb)/* check for timestamp cookie support */ memset (&tcp_opt,0 ,sizeof (tcp_opt)); -tcp_parse_options (sock_net (sk), skb, &tcp_opt,0 ,NULL ); +tcp_parse_options (sock_net (sk), skb, &tcp_opt,0 ,NULL , tp);if (tcp_opt.saw_tstamp && tcp_opt.rcv_tsecr) { tsoff = secure_tcpv6_ts_off(sock_net(sk),net/ipv6/tcp_ipv6.c +26
-3@@ -784,9 +784,10 @@ static void tcp_v6_send_response(const struct sock *sk, struct sk_buff *skb, u32struct flowi6 fl6;struct net *net = sk ?sock_net (sk) :dev_net (skb_dst (skb)->dev);struct sock *ctl_sk = net->ipv6.tcp_sk; -unsigned int tot_len =sizeof (struct tcphdr ); +unsigned int tot_len =0 ;struct dst_entry *dst;__be32 *topt; +struct tcp_out_options extraopts;if (tsecr) tot_len +=TCPOLEN_TSTAMP_ALIGNED ;@@ -795,10 +796,28 @@ static void tcp_v6_send_response(const struct sock *sk, struct sk_buff *skb, u32 tot_len +=TCPOLEN_MD5SIG_ALIGNED ;#endif +rcu_read_lock (); +if (static_branch_unlikely (&tcp_extra_options_enabled)) { +unsigned int remaining =MAX_TCP_OPTION_SPACE - tot_len; +u8 extraflags = rst ?TCPHDR_RST :0 ; + +if (!rst || !th->ack) + extraflags |=TCPHDR_ACK ; + +memset (&extraopts,0 ,sizeof (extraopts)); + + tot_len +=tcp_extra_options_prepare (skb, extraflags, remaining, + &extraopts, sk); + } + + tot_len +=sizeof (struct tcphdr ); + buff =alloc_skb (MAX_HEADER +sizeof (struct ipv6hdr ) + tot_len,GFP_ATOMIC ); -if (!buff) +if (!buff) { +rcu_read_unlock ();return ; + }skb_reserve (buff,MAX_HEADER +sizeof (struct ipv6hdr ) + tot_len);@@ -835,6 +854,11 @@ static void tcp_v6_send_response(const struct sock *sk, struct sk_buff *skb, u32 }#endif +if (static_branch_unlikely (&tcp_extra_options_enabled)) +tcp_extra_options_write (topt, &extraopts, sk); + +rcu_read_unlock (); +memset (&fl6,0 ,sizeof (fl6)); fl6.daddr =ipv6_hdr (skb)->saddr; fl6.saddr =ipv6_hdr (skb)->daddr; --2.14.2
Quoted 5 lines
Quoted 4 lines
_______________________________________________
mptcp mailing list
mptcp(a)lists.01.org
https://lists.01.org/mailman/listinfo/mptcp--
Mat Martineau
Intel OTC
Christoph Paasch
Quoted 737 lines · Mat Martineau
Allow additional TCP options to be handled by registered hook functions.
Registered options have a priority that determines the order in which options are prepared and written. Lower priority numbers are handled first.
Option parsing will call the provided 'parse' function when a TCP option number is not recognized by the normal option parsing code.
The 'prepare' function determines the required space for registered options and store associated data. 'write' adds the option to the TCP header.
A static key and RCU synchronization are used to minimize the performance impact of these extensible TCP features.
Signed-off-by Mat Martineau <mathew.j.martineau(a)linux.intel.com>---
Changes from v1: One 'prepare' callback (no more special callback for
request_sock), and add a few missing callback sites (like ipv6).drivers/infiniband/hw/cxgb4/cm.c | 2 +- include/linux/tcp.h | 22 +++++++ include/net/tcp.h | 40 +++++++++++- net/ipv4/syncookies.c | 2 +- net/ipv4/tcp.c | 133 +++++++++++++++++++++++++++++++++++++++ net/ipv4/tcp_input.c | 16 +++-- net/ipv4/tcp_ipv4.c | 80 ++++++++++++++++++----- net/ipv4/tcp_minisocks.c | 4 +- net/ipv4/tcp_output.c | 43 +++++++------ net/ipv6/syncookies.c | 2 +- net/ipv6/tcp_ipv6.c | 28 ++++++++- 11 files changed, 323 insertions(+), 49 deletions(-)drivers/infiniband/hw/cxgb4/cm.c +1
-1@@ -3752,7 +3752,7 @@ static void build_cpl_pass_accept_req(struct sk_buff *skb, int stid , u8 tos)*/ memset (&tmp_opt,0 ,sizeof (tmp_opt));tcp_clear_options (&tmp_opt); -tcp_parse_options (&init_net, skb, &tmp_opt,0 ,NULL ); +tcp_parse_options (&init_net, skb, &tmp_opt,0 ,NULL ,NULL ); req =__skb_push (skb,sizeof (*req));memset (req,0 ,sizeof (*req));include/linux/tcp.h +22
-0@@ -112,6 +112,23 @@ static inline void tcp_clear_options(struct tcp_options_received *rx_opt) rx_opt->wscale_ok = rx_opt->snd_wscale =0 ; } +#define OPTION_SACK_ADVERTISE (1 << 0) +#define OPTION_TS (1 << 1) +#define OPTION_MD5 (1 << 2) +#define OPTION_WSCALE (1 << 3) +#define OPTION_FAST_OPEN_COOKIE (1 << 8) + +struct tcp_out_options { +u16 options;/* bit field of OPTION_* */ +u16 mss;/* 0 to disable */ +u8 ws;/* window scale, 0 to disable */ +u8 num_sack_blocks;/* number of SACK blocks to include */ +u8 hash_size;/* bytes in hash_location */ +__u8 *hash_location;/* temporary pointer, overloaded */ +__u32 tsval, tsecr;/* need to include OPTION_TS */ +struct tcp_fastopen_cookie *fastopen_cookie;/* Fast open cookie */ +}; +/* This is the max number of SACKS that we'll generate and process. It's safe * to increase this, although since: * size = TCPOLEN_SACK_BASE_ALIGNED (4) + n * TCPOLEN_SACK_PERBLOCK (8) @@ -389,6 +406,11 @@ static inline struct tcp_sock *tcp_sk(const struct sock *sk)return (struct tcp_sock *)sk; } +static inline struct sock *tcp_to_sk (const struct tcp_sock *tp) +{ +return (struct sock *)tp; +} +struct tcp_timewait_sock {struct inet_timewait_sock tw_sk;#define tw_rcv_nxt tw_sk.__tw_common.skc_tw_rcv_nxtinclude/net/tcp.h +39
-1@@ -433,7 +433,8 @@ int tcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int nonblock, int flags, int *addr_len);void tcp_parse_options (const struct net *net,const struct sk_buff *skb,struct tcp_options_received *opt_rx, -int estab,struct tcp_fastopen_cookie *foc); +int estab,struct tcp_fastopen_cookie *foc, +struct tcp_sock *tp);const u8 *tcp_parse_md5sig_option (const struct tcphdr *th);/* @@ -2109,4 +2110,41 @@ static inline bool tcp_bpf_ca_needs_ecn(struct sock *sk) {return (tcp_call_bpf (sk,BPF_SOCK_OPS_NEEDS_ECN ) ==1 ); } + +extern struct static_key_false tcp_extra_options_enabled; + +struct tcp_extra_option_ops { +struct list_head list; +unsigned char option_kind; +unsigned char priority; +void (*parse)(int opsize,const unsigned char *opptr, +const struct sk_buff *skb, +struct tcp_options_received *opt_rx, +struct sock *sk); +/* Return the number of bytes consumed */ +unsigned int (*prepare)(struct sk_buff *skb,u8 flags, +unsigned int remaining, +struct tcp_out_options *opts, +const struct sock *sk); +void (*write)(__be32 *ptr,struct tcp_out_options *opts, +const struct sock *sk); +struct module *owner; +}; + +void tcp_extra_options_parse (int opcode,int opsize,const unsigned char *opptr, +const struct sk_buff *skb, +struct tcp_options_received *opt_rx, +struct sock *sk); + +unsigned int tcp_extra_options_prepare (struct sk_buff *skb,u8 flags, +unsigned int remaining, +struct tcp_out_options *opts, +const struct sock *sk); + +void tcp_extra_options_write (__be32 *ptr,struct tcp_out_options *opts, +const struct sock *sk); + +int tcp_register_extra_option (struct tcp_extra_option_ops *ops); +void tcp_unregister_extra_option (struct tcp_extra_option_ops *ops); +#endif /* _TCP_H */ net/ipv4/syncookies.c +1
-1@@ -313,7 +313,7 @@ struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb)/* check for timestamp cookie support */ memset (&tcp_opt,0 ,sizeof (tcp_opt)); -tcp_parse_options (sock_net (sk), skb, &tcp_opt,0 ,NULL ); +tcp_parse_options (sock_net (sk), skb, &tcp_opt,0 ,NULL , tp);if (tcp_opt.saw_tstamp && tcp_opt.rcv_tsecr) { tsoff = secure_tcp_ts_off(sock_net(sk),net/ipv4/tcp.c +133
-0@@ -270,6 +270,7 @@#include <linux/time.h> #include <linux/slab.h> #include <linux/errqueue.h> +#include <linux/static_key.h> #include <net/icmp.h> #include <net/inet_common.h> @@ -306,6 +307,13 @@ EXPORT_SYMBOL(tcp_memory_allocated);struct percpu_counter tcp_sockets_allocated;EXPORT_SYMBOL (tcp_sockets_allocated); +/* +* Optional TCP option handlers +*/ +static DEFINE_SPINLOCK (tcp_option_list_lock); +static LIST_HEAD (tcp_option_list); +DEFINE_STATIC_KEY_FALSE (tcp_extra_options_enabled); +/* * TCP splice context */ @@ -3375,6 +3383,130 @@ EXPORT_SYMBOL(tcp_md5_hash_key);#endif +/* Linear search, few entries are expected. The RCU read lock must +* be held before calling. +*/ +static struct tcp_extra_option_ops *tcp_extra_options_find_kind (unsigned char kind) +{ +struct tcp_extra_option_ops *entry; + +list_for_each_entry_rcu (entry, &tcp_option_list, list) { +if (entry->option_kind == kind) +return entry; + } + +return NULL ; +} + +void tcp_extra_options_parse (int opcode,int opsize,const unsigned char *opptr, +const struct sk_buff *skb, +struct tcp_options_received *opt_rx, +struct sock *sk) +{ +struct tcp_extra_option_ops *entry; + +rcu_read_lock (); + entry =tcp_extra_options_find_kind (opcode); +if (entry && entry->parse) + entry->parse (opsize, opptr, skb, opt_rx, sk); +rcu_read_unlock (); +} +EXPORT_SYMBOL_GPL (tcp_extra_options_parse); + +/* The RCU read lock must be held before calling, and should span both +* the call to this function and tcp_extra_options_write to ensure that +* tcp_option_list does not change between the two calls. To preserve +* expected option alignment, always returns a multiple of 4 bytes. +*/ +unsigned int tcp_extra_options_prepare (struct sk_buff *skb,u8 flags, +unsigned int remaining, +struct tcp_out_options *opts, +const struct sock *sk) +{ +struct tcp_extra_option_ops *entry; +unsigned int used =0 ; + +list_for_each_entry_rcu (entry, &tcp_option_list, list) { +if (unlikely (!entry->prepare)) +continue ; + + used += entry->prepare (skb, flags, remaining - used, opts, sk); + } + +return roundup (used,4 ); +} +EXPORT_SYMBOL_GPL (tcp_extra_options_prepare); + +/* The RCU read lock must be held before calling, and should span both +* the call to tcp_extra_options_write and this function to ensure that +* tcp_option_list does not change between the two calls. +*/ +void tcp_extra_options_write (__be32 *ptr,struct tcp_out_options *opts, +const struct sock *sk) +{ +struct tcp_extra_option_ops *entry; + +list_for_each_entry_rcu (entry, &tcp_option_list, list) { +if (unlikely (!entry->write)) +continue ; + + entry->write (ptr, opts, sk); + } +} +EXPORT_SYMBOL_GPL (tcp_extra_options_write); + +int tcp_register_extra_option (struct tcp_extra_option_ops *ops) +{ +struct tcp_extra_option_ops *entry; +struct list_head * add_before = &tcp_option_list; +int ret =0 ; + +if (!ops->option_kind) +return -EINVAL ; + +if (!try_module_get (ops->owner)) +return -ENOENT ; + +spin_lock (&tcp_option_list_lock); + +list_for_each_entry_rcu (entry, &tcp_option_list, list) { +if (entry->option_kind == ops->option_kind) { +pr_notice ("Option kind %u already registered\n" , + ops->option_kind); +spin_unlock (&tcp_option_list_lock); +module_put (ops->owner); +return -EEXIST ; + } + +if (entry->priority <= ops->priority) + add_before = &entry->list; + } + +list_add_tail_rcu (&ops->list, add_before); +pr_debug ("Option kind %u registered\n" , ops->option_kind); + +spin_unlock (&tcp_option_list_lock); + +static_branch_inc (&tcp_extra_options_enabled); + +return ret; +} +EXPORT_SYMBOL_GPL (tcp_register_extra_option); + +void tcp_unregister_extra_option (struct tcp_extra_option_ops *ops) +{ +spin_lock (&tcp_option_list_lock); +list_del_rcu (&ops->list); +spin_unlock (&tcp_option_list_lock); + +synchronize_net (); + +static_branch_dec (&tcp_extra_options_enabled); + +module_put (ops->owner); +} +EXPORT_SYMBOL_GPL (tcp_unregister_extra_option); +void tcp_done (struct sock *sk) {struct request_sock *req =tcp_sk (sk)->fastopen_rsk;@@ -3521,6 +3653,7 @@ void __init tcp_init(void)INIT_HLIST_HEAD (&tcp_hashinfo.bhash[i].chain); } +INIT_LIST_HEAD (&tcp_option_list); cnt = tcp_hashinfo.ehash_mask +1 ; sysctl_tcp_max_orphans = cnt /2 ;net/ipv4/tcp_input.c +11
-5@@ -3728,7 +3728,7 @@ static void tcp_parse_fastopen_option(int len, const unsigned char *cookie,void tcp_parse_options (const struct net *net,const struct sk_buff *skb,struct tcp_options_received *opt_rx,int estab, -struct tcp_fastopen_cookie *foc) +struct tcp_fastopen_cookie *foc,struct tcp_sock *tp) {const unsigned char *ptr;const struct tcphdr *th =tcp_hdr (skb);@@ -3830,6 +3830,12 @@ void tcp_parse_options(const struct net *net, ptr + 2, th->syn, foc, true);break ; +default : +tcp_extra_options_parse (opcode, opsize, ptr, + skb, opt_rx, +tcp_to_sk (tp)); +break ; + } ptr += opsize-2 ; length -= opsize;@@ -3876,7 +3882,7 @@ static bool tcp_fast_parse_options(const struct net *net,return true; } -tcp_parse_options (net, skb, &tp->rx_opt,1 ,NULL ); +tcp_parse_options (net, skb, &tp->rx_opt,1 ,NULL , tp);if (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr) tp->rx_opt.rcv_tsecr -= tp->tsoffset;@@ -5569,7 +5575,7 @@ static bool tcp_rcv_fastopen_synack(struct sock *sk, struct sk_buff *synack,/* Get original SYNACK MSS value if user MSS sets mss_clamp */ tcp_clear_options (&opt); opt.user_mss = opt.mss_clamp =0 ; -tcp_parse_options (sock_net (sk), synack, &opt,0 ,NULL ); +tcp_parse_options (sock_net (sk), synack, &opt,0 ,NULL , tp); mss = opt.mss_clamp; }@@ -5623,7 +5629,7 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb,int saved_clamp = tp->rx_opt.mss_clamp;bool fastopen_fail; -tcp_parse_options (sock_net (sk), skb, &tp->rx_opt,0 , &foc); +tcp_parse_options (sock_net (sk), skb, &tp->rx_opt,0 , &foc, tp);if (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr) tp->rx_opt.rcv_tsecr -= tp->tsoffset;@@ -6299,7 +6305,7 @@ int tcp_conn_request(struct request_sock_ops *rsk_ops, tmp_opt.mss_clamp = af_ops->mss_clamp; tmp_opt.user_mss = tp->rx_opt.user_mss;tcp_parse_options (sock_net (sk), skb, &tmp_opt,0 , - want_cookie ?NULL : &foc); + want_cookie ?NULL : &foc, tp);if (want_cookie && !tmp_opt.saw_tstamp)tcp_clear_options (&tmp_opt);net/ipv4/tcp_ipv4.c +63
-17@@ -598,9 +598,7 @@ static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb)const struct tcphdr *th =tcp_hdr (skb);struct {struct tcphdr th; -#ifdef CONFIG_TCP_MD5SIG -__be32 opt[(TCPOLEN_MD5SIG_ALIGNED >>2 )]; -#endif +__be32 opt[(MAX_TCP_OPTION_SPACE >>2 )]; } rep;struct ip_reply_arg arg;#ifdef CONFIG_TCP_MD5SIG @@ -611,6 +609,7 @@ static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb)struct sock *sk1 =NULL ;#endif struct net *net; +int offset =0 ;/* Never send a reset in response to a reset. */ if (th->rst)@@ -676,17 +675,44 @@ static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb) goto out; } +#endif + +if (static_branch_unlikely (&tcp_extra_options_enabled)) { +unsigned int remaining; +unsigned int used; +struct tcp_out_options opts; + + remaining =sizeof (rep.opt); +#ifdef CONFIG_TCP_MD5SIG +if (key) + remaining -=TCPOLEN_MD5SIG_ALIGNED ; +#endif + +memset (&opts,0 ,sizeof (opts)); + +rcu_read_lock (); + used =tcp_extra_options_prepare (NULL ,TCPHDR_RST , remaining, + &opts, sk); + +tcp_extra_options_write (&rep.opt[0 ], &opts, sk); +rcu_read_unlock (); + + arg.iov[0 ].iov_len += used; + offset += used /4 ; + rep.th.doff = arg.iov[0 ].iov_len /4 ; + } +#ifdef CONFIG_TCP_MD5SIG if (key) { - rep.opt[0 ] =htonl ((TCPOPT_NOP <<24 ) | - (TCPOPT_NOP <<16 ) | - (TCPOPT_MD5SIG <<8 ) | -TCPOLEN_MD5SIG ); + rep.opt[offset++] =htonl ((TCPOPT_NOP <<24 ) | + (TCPOPT_NOP <<16 ) | + (TCPOPT_MD5SIG <<8 ) | +TCPOLEN_MD5SIG );/* Update length and the length the header thinks exists */ arg.iov[0 ].iov_len +=TCPOLEN_MD5SIG_ALIGNED ; rep.th.doff = arg.iov[0 ].iov_len /4 ; -tcp_v4_md5_hash_hdr ((__u8 *) &rep.opt[1 ], +tcp_v4_md5_hash_hdr ((__u8 *) &rep.opt[offset], key,ip_hdr (skb)->saddr,ip_hdr (skb)->daddr, &rep.th); }@@ -738,14 +764,11 @@ static void tcp_v4_send_ack(const struct sock *sk,const struct tcphdr *th =tcp_hdr (skb);struct {struct tcphdr th; -__be32 opt[(TCPOLEN_TSTAMP_ALIGNED >>2 ) -#ifdef CONFIG_TCP_MD5SIG - + (TCPOLEN_MD5SIG_ALIGNED >>2 ) -#endif - ]; +__be32 opt[(MAX_TCP_OPTION_SPACE >>2 )]; } rep;struct net *net =sock_net (sk);struct ip_reply_arg arg; +int offset =0 ;memset (&rep.th,0 ,sizeof (struct tcphdr ));memset (&arg,0 ,sizeof (arg));@@ -759,33 +782,56 @@ static void tcp_v4_send_ack(const struct sock *sk, rep.opt[1 ] =htonl (tsval); rep.opt[2 ] =htonl (tsecr); arg.iov[0 ].iov_len +=TCPOLEN_TSTAMP_ALIGNED ; + offset +=3 ; }/* Swap the send and the receive. */ rep.th.dest = th->source; rep.th.source = th->dest; - rep.th.doff = arg.iov[0 ].iov_len /4 ; rep.th.seq =htonl (seq); rep.th.ack_seq =htonl (ack); rep.th.ack =1 ; rep.th.window =htons (win); +if (static_branch_unlikely (&tcp_extra_options_enabled)) { +unsigned int remaining; +unsigned int used; +struct tcp_out_options opts; + + remaining =sizeof (rep.th) +sizeof (rep.opt) - arg.iov[0 ].iov_len;#ifdef CONFIG_TCP_MD5SIG -if (key) { -int offset = (tsecr) ?3 :0 ; +if (key) + remaining -=TCPOLEN_MD5SIG_ALIGNED ; +#endif +memset (&opts,0 ,sizeof (opts)); +rcu_read_lock (); + used =tcp_extra_options_prepare (NULL ,TCPHDR_ACK , remaining, + &opts, sk); + +tcp_extra_options_write (&rep.opt[offset], &opts, sk); +rcu_read_unlock (); + + arg.iov[0 ].iov_len += used; + offset += used /4 ; + } + + rep.th.doff = arg.iov[0 ].iov_len /4 ; + +#ifdef CONFIG_TCP_MD5SIG +if (key) { rep.opt[offset++] =htonl ((TCPOPT_NOP <<24 ) | (TCPOPT_NOP <<16 ) | (TCPOPT_MD5SIG <<8 ) |TCPOLEN_MD5SIG ); arg.iov[0 ].iov_len +=TCPOLEN_MD5SIG_ALIGNED ; - rep.th.doff = arg.iov[0 ].iov_len/4 ;tcp_v4_md5_hash_hdr ((__u8 *) &rep.opt[offset], key,ip_hdr (skb)->saddr,ip_hdr (skb)->daddr, &rep.th); }#endif + arg.flags = reply_flags; arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr, ip_hdr(skb)->saddr, /* XXX */net/ipv4/tcp_minisocks.c +2
-2@@ -98,7 +98,7 @@ tcp_timewait_state_process(struct inet_timewait_sock *tw, struct sk_buff *skb, tmp_opt.saw_tstamp =0 ;if (th->doff > (sizeof (*th) >>2 ) && tcptw->tw_ts_recent_stamp) { -tcp_parse_options (twsk_net (tw), skb, &tmp_opt,0 ,NULL ); +tcp_parse_options (twsk_net (tw), skb, &tmp_opt,0 ,NULL ,NULL );if (tmp_opt.saw_tstamp) {if (tmp_opt.rcv_tsecr)@@ -565,7 +565,7 @@ struct sock *tcp_check_req(struct sock *sk, struct sk_buff *skb, tmp_opt.saw_tstamp =0 ;if (th->doff > (sizeof (struct tcphdr )>>2 )) { -tcp_parse_options (sock_net (sk), skb, &tmp_opt,0 ,NULL ); +tcp_parse_options (sock_net (sk), skb, &tmp_opt,0 ,NULL ,NULL );if (tmp_opt.saw_tstamp) { tmp_opt.ts_recent = req->ts_recent;net/ipv4/tcp_output.c +24
-19@@ -41,6 +41,7 @@#include <linux/compiler.h> #include <linux/gfp.h> #include <linux/module.h> +#include <linux/static_key.h> /* People can turn this off for buggy TCP's found in printers etc. */ int sysctl_tcp_retrans_collapse__read_mostly =1 ;@@ -413,23 +414,6 @@ static inline bool tcp_urg_mode(const struct tcp_sock *tp)return tp->snd_una != tp->snd_up; } -#define OPTION_SACK_ADVERTISE (1 << 0) -#define OPTION_TS (1 << 1) -#define OPTION_MD5 (1 << 2) -#define OPTION_WSCALE (1 << 3) -#define OPTION_FAST_OPEN_COOKIE (1 << 8) - -struct tcp_out_options { -u16 options;/* bit field of OPTION_* */ -u16 mss;/* 0 to disable */ -u8 ws;/* window scale, 0 to disable */ -u8 num_sack_blocks;/* number of SACK blocks to include */ -u8 hash_size;/* bytes in hash_location */ -__u8 *hash_location;/* temporary pointer, overloaded */ -__u32 tsval, tsecr;/* need to include OPTION_TS */ -struct tcp_fastopen_cookie *fastopen_cookie;/* Fast open cookie */ -}; -/* Write previously computed TCP options to the packet. * * Beware: Something in the Internet is very sensitive to the ordering of @@ -536,6 +520,9 @@ static void tcp_options_write(__be32 *ptr, struct tcp_sock *tp, } ptr += (len +3 ) >>2 ; } + +if (static_branch_unlikely (&tcp_extra_options_enabled)) +tcp_extra_options_write (ptr, opts,tcp_to_sk (tp)); }/* Compute TCP options for SYN packets. This is not the final @@ -603,6 +590,11 @@ static unsigned int tcp_syn_options(struct sock *sk, struct sk_buff *skb, } } +if (static_branch_unlikely (&tcp_extra_options_enabled)) + remaining -=tcp_extra_options_prepare (skb,TCPHDR_SYN , + remaining, opts, +tcp_to_sk (tp)); +return MAX_TCP_OPTION_SPACE - remaining; }@@ -663,6 +655,12 @@ static unsigned int tcp_synack_options(struct request_sock *req, } } +if (static_branch_unlikely (&tcp_extra_options_enabled)) + remaining -=tcp_extra_options_prepare (skb, +TCPHDR_SYN |TCPHDR_ACK , + remaining, opts, +req_to_sk (req)); +return MAX_TCP_OPTION_SPACE - remaining; }@@ -696,6 +694,11 @@ static unsigned int tcp_established_options(struct sock *sk, struct sk_buff *skb size +=TCPOLEN_TSTAMP_ALIGNED ; } +if (static_branch_unlikely (&tcp_extra_options_enabled)) + size +=tcp_extra_options_prepare (skb,0 , +MAX_TCP_OPTION_SPACE - size, + opts,tcp_to_sk (tp)); + eff_sacks = tp->rx_opt.num_sacks + tp->rx_opt.dsack;if (unlikely (eff_sacks)) {const unsigned int remaining =MAX_TCP_OPTION_SPACE - size;@@ -1016,6 +1019,7 @@ static int tcp_transmit_skb(struct sock *sk, struct sk_buff *skb, int clone_it, tcb =TCP_SKB_CB (skb);memset (&opts,0 ,sizeof (opts)); +rcu_read_lock ();if (unlikely (tcb->tcp_flags &TCPHDR_SYN )) tcp_options_size =tcp_syn_options (sk, skb, &opts, &md5);else @@ -1092,6 +1096,7 @@ static int tcp_transmit_skb(struct sock *sk, struct sk_buff *skb, int clone_it, md5, sk, skb); }#endif +rcu_read_unlock (); icsk->icsk_af_ops->send_check (sk, skb);@@ -3156,8 +3161,8 @@ struct sk_buff *tcp_make_synack(const struct sock *sk, struct dst_entry *dst,#endif skb->skb_mstamp =tcp_clock_us (); -#ifdef CONFIG_TCP_MD5SIG rcu_read_lock (); +#ifdef CONFIG_TCP_MD5SIG md5 =tcp_rsk (req)->af_specific->req_md5_lookup (sk,req_to_sk (req));#endif skb_set_hash (skb,tcp_rsk (req)->txhash,PKT_HASH_TYPE_L4 );@@ -3196,8 +3201,8 @@ struct sk_buff *tcp_make_synack(const struct sock *sk, struct dst_entry *dst,if (md5)tcp_rsk (req)->af_specific->calc_md5_hash (opts.hash_location, md5,req_to_sk (req), skb); -rcu_read_unlock ();#endif +rcu_read_unlock ();/* Do not fool tcpdump (if any), clean our debris */ skb->tstamp =0 ;net/ipv6/syncookies.c +1
-1@@ -162,7 +162,7 @@ struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb)/* check for timestamp cookie support */ memset (&tcp_opt,0 ,sizeof (tcp_opt)); -tcp_parse_options (sock_net (sk), skb, &tcp_opt,0 ,NULL ); +tcp_parse_options (sock_net (sk), skb, &tcp_opt,0 ,NULL , tp);if (tcp_opt.saw_tstamp && tcp_opt.rcv_tsecr) { tsoff = secure_tcpv6_ts_off(sock_net(sk),net/ipv6/tcp_ipv6.c +21
-2@@ -784,9 +784,10 @@ static void tcp_v6_send_response(const struct sock *sk, struct sk_buff *skb, u32struct flowi6 fl6;struct net *net = sk ?sock_net (sk) :dev_net (skb_dst (skb)->dev);struct sock *ctl_sk = net->ipv6.tcp_sk; -unsigned int tot_len =sizeof (struct tcphdr ); +unsigned int tot_len =0 ;struct dst_entry *dst;__be32 *topt; +struct tcp_out_options extraopts;if (tsecr) tot_len +=TCPOLEN_TSTAMP_ALIGNED ;@@ -795,10 +796,28 @@ static void tcp_v6_send_response(const struct sock *sk, struct sk_buff *skb, u32 tot_len +=TCPOLEN_MD5SIG_ALIGNED ;#endif +rcu_read_lock (); +if (static_branch_unlikely (&tcp_extra_options_enabled)) { +unsigned int remaining =MAX_TCP_OPTION_SPACE - tot_len; +u8 extraflags = rst ?TCPHDR_RST :0 ; + +if (!rst || !th->ack) + extraflags |=TCPHDR_ACK ; + +memset (&extraopts,0 ,sizeof (extraopts)); + + tot_len +=tcp_extra_options_prepare (skb, extraflags, remaining, + &extraopts, sk); + } + + tot_len +=sizeof (struct tcphdr ); + buff =alloc_skb (MAX_HEADER +sizeof (struct ipv6hdr ) + tot_len,GFP_ATOMIC ); -if (!buff) +if (!buff) { +rcu_read_unlock ();return ; + }skb_reserve (buff,MAX_HEADER +sizeof (struct ipv6hdr ) + tot_len);
net/ipv6/tcp_ipv6.c static void tcp_v6_send_response(const struct sock *sk, struct sk_buff *skb, u32 · line 858
Quoted 5 lines
net/ipv6/tcp_ipv6.c +2
-0@@ -835,6 +854,11 @@ static void tcp_v6_send_response(const struct sock *sk, struct sk_buff *skb, u32 }#endif +if (static_branch_unlikely (&tcp_extra_options_enabled)) +tcp_extra_options_write (topt, &extraopts, sk);
I'm currently rebasing my changes on top of this patch here. We will have to see how to handle this here correctly.
Because, MD5 needs some more info to generate the hash correctly. Notable, IP-addresses and the TCP-header.
tcp_extra_options_write done in this way here won't suffice.
I will see how to handle this and resubmit with a v2 of my series.
Christoph
Quoted 13 lines
net/ipv6/tcp_ipv6.c +3
-0+ +rcu_read_unlock (); +memset (&fl6,0 ,sizeof (fl6)); fl6.daddr =ipv6_hdr (skb)->saddr; fl6.saddr =ipv6_hdr (skb)->daddr;Signature · 6 lines
2.14.2 _______________________________________________ mptcp mailing list mptcp(a)lists.01.org https://lists.01.org/mailman/listinfo/mptcp