From 5e33a23ba4b56c109b732d57a0a76558a37d9ec5 Mon Sep 17 00:00:00 2001 From: David Howells Date: Fri, 5 Oct 2018 14:05:34 +0100 Subject: rxrpc: Fix some missed refs to init_net Fix some refs to init_net that should've been changed to the appropriate network namespace. Fixes: 2baec2c3f854 ("rxrpc: Support network namespacing") Signed-off-by: David Howells Acked-by: Paolo Abeni --- net/rxrpc/ar-internal.h | 10 ++++++---- net/rxrpc/call_accept.c | 2 +- net/rxrpc/call_object.c | 4 ++-- net/rxrpc/conn_client.c | 10 ++++++---- net/rxrpc/input.c | 4 ++-- net/rxrpc/peer_object.c | 28 +++++++++++++++++----------- 6 files changed, 34 insertions(+), 24 deletions(-) (limited to 'net/rxrpc') diff --git a/net/rxrpc/ar-internal.h b/net/rxrpc/ar-internal.h index ef9554131434..63c43b3a2096 100644 --- a/net/rxrpc/ar-internal.h +++ b/net/rxrpc/ar-internal.h @@ -891,8 +891,9 @@ extern unsigned long rxrpc_conn_idle_client_fast_expiry; extern struct idr rxrpc_client_conn_ids; void rxrpc_destroy_client_conn_ids(void); -int rxrpc_connect_call(struct rxrpc_call *, struct rxrpc_conn_parameters *, - struct sockaddr_rxrpc *, gfp_t); +int rxrpc_connect_call(struct rxrpc_sock *, struct rxrpc_call *, + struct rxrpc_conn_parameters *, struct sockaddr_rxrpc *, + gfp_t); void rxrpc_expose_client_call(struct rxrpc_call *); void rxrpc_disconnect_client_call(struct rxrpc_call *); void rxrpc_put_client_conn(struct rxrpc_connection *); @@ -1045,10 +1046,11 @@ void rxrpc_peer_keepalive_worker(struct work_struct *); */ struct rxrpc_peer *rxrpc_lookup_peer_rcu(struct rxrpc_local *, const struct sockaddr_rxrpc *); -struct rxrpc_peer *rxrpc_lookup_peer(struct rxrpc_local *, +struct rxrpc_peer *rxrpc_lookup_peer(struct rxrpc_sock *, struct rxrpc_local *, struct sockaddr_rxrpc *, gfp_t); struct rxrpc_peer *rxrpc_alloc_peer(struct rxrpc_local *, gfp_t); -void rxrpc_new_incoming_peer(struct rxrpc_local *, struct rxrpc_peer *); +void rxrpc_new_incoming_peer(struct rxrpc_sock *, struct rxrpc_local *, + struct rxrpc_peer *); void rxrpc_destroy_all_peers(struct rxrpc_net *); struct rxrpc_peer *rxrpc_get_peer(struct rxrpc_peer *); struct rxrpc_peer *rxrpc_get_peer_maybe(struct rxrpc_peer *); diff --git a/net/rxrpc/call_accept.c b/net/rxrpc/call_accept.c index 9c7f26d06a52..f55f67894465 100644 --- a/net/rxrpc/call_accept.c +++ b/net/rxrpc/call_accept.c @@ -287,7 +287,7 @@ static struct rxrpc_call *rxrpc_alloc_incoming_call(struct rxrpc_sock *rx, (peer_tail + 1) & (RXRPC_BACKLOG_MAX - 1)); - rxrpc_new_incoming_peer(local, peer); + rxrpc_new_incoming_peer(rx, local, peer); } /* Now allocate and set up the connection */ diff --git a/net/rxrpc/call_object.c b/net/rxrpc/call_object.c index 799f75b6900d..0ca2c2dfd196 100644 --- a/net/rxrpc/call_object.c +++ b/net/rxrpc/call_object.c @@ -287,7 +287,7 @@ struct rxrpc_call *rxrpc_new_client_call(struct rxrpc_sock *rx, /* Set up or get a connection record and set the protocol parameters, * including channel number and call ID. */ - ret = rxrpc_connect_call(call, cp, srx, gfp); + ret = rxrpc_connect_call(rx, call, cp, srx, gfp); if (ret < 0) goto error; @@ -339,7 +339,7 @@ int rxrpc_retry_client_call(struct rxrpc_sock *rx, /* Set up or get a connection record and set the protocol parameters, * including channel number and call ID. */ - ret = rxrpc_connect_call(call, cp, srx, gfp); + ret = rxrpc_connect_call(rx, call, cp, srx, gfp); if (ret < 0) goto error; diff --git a/net/rxrpc/conn_client.c b/net/rxrpc/conn_client.c index 8acf74fe24c0..521189f4b666 100644 --- a/net/rxrpc/conn_client.c +++ b/net/rxrpc/conn_client.c @@ -276,7 +276,8 @@ dont_reuse: * If we return with a connection, the call will be on its waiting list. It's * left to the caller to assign a channel and wake up the call. */ -static int rxrpc_get_client_conn(struct rxrpc_call *call, +static int rxrpc_get_client_conn(struct rxrpc_sock *rx, + struct rxrpc_call *call, struct rxrpc_conn_parameters *cp, struct sockaddr_rxrpc *srx, gfp_t gfp) @@ -289,7 +290,7 @@ static int rxrpc_get_client_conn(struct rxrpc_call *call, _enter("{%d,%lx},", call->debug_id, call->user_call_ID); - cp->peer = rxrpc_lookup_peer(cp->local, srx, gfp); + cp->peer = rxrpc_lookup_peer(rx, cp->local, srx, gfp); if (!cp->peer) goto error; @@ -683,7 +684,8 @@ out: * find a connection for a call * - called in process context with IRQs enabled */ -int rxrpc_connect_call(struct rxrpc_call *call, +int rxrpc_connect_call(struct rxrpc_sock *rx, + struct rxrpc_call *call, struct rxrpc_conn_parameters *cp, struct sockaddr_rxrpc *srx, gfp_t gfp) @@ -696,7 +698,7 @@ int rxrpc_connect_call(struct rxrpc_call *call, rxrpc_discard_expired_client_conns(&rxnet->client_conn_reaper); rxrpc_cull_active_client_conns(rxnet); - ret = rxrpc_get_client_conn(call, cp, srx, gfp); + ret = rxrpc_get_client_conn(rx, call, cp, srx, gfp); if (ret < 0) goto out; diff --git a/net/rxrpc/input.c b/net/rxrpc/input.c index 800f5b8a1baa..c5af9955665b 100644 --- a/net/rxrpc/input.c +++ b/net/rxrpc/input.c @@ -1156,12 +1156,12 @@ void rxrpc_data_ready(struct sock *udp_sk) /* we'll probably need to checksum it (didn't call sock_recvmsg) */ if (skb_checksum_complete(skb)) { rxrpc_free_skb(skb, rxrpc_skb_rx_freed); - __UDP_INC_STATS(&init_net, UDP_MIB_INERRORS, 0); + __UDP_INC_STATS(sock_net(udp_sk), UDP_MIB_INERRORS, 0); _leave(" [CSUM failed]"); return; } - __UDP_INC_STATS(&init_net, UDP_MIB_INDATAGRAMS, 0); + __UDP_INC_STATS(sock_net(udp_sk), UDP_MIB_INDATAGRAMS, 0); /* The UDP protocol already released all skb resources; * we are free to add our own data there. diff --git a/net/rxrpc/peer_object.c b/net/rxrpc/peer_object.c index 01a9febfa367..2d39eaf19620 100644 --- a/net/rxrpc/peer_object.c +++ b/net/rxrpc/peer_object.c @@ -153,8 +153,10 @@ struct rxrpc_peer *rxrpc_lookup_peer_rcu(struct rxrpc_local *local, * assess the MTU size for the network interface through which this peer is * reached */ -static void rxrpc_assess_MTU_size(struct rxrpc_peer *peer) +static void rxrpc_assess_MTU_size(struct rxrpc_sock *rx, + struct rxrpc_peer *peer) { + struct net *net = sock_net(&rx->sk); struct dst_entry *dst; struct rtable *rt; struct flowi fl; @@ -169,7 +171,7 @@ static void rxrpc_assess_MTU_size(struct rxrpc_peer *peer) switch (peer->srx.transport.family) { case AF_INET: rt = ip_route_output_ports( - &init_net, fl4, NULL, + net, fl4, NULL, peer->srx.transport.sin.sin_addr.s_addr, 0, htons(7000), htons(7001), IPPROTO_UDP, 0, 0); if (IS_ERR(rt)) { @@ -188,7 +190,7 @@ static void rxrpc_assess_MTU_size(struct rxrpc_peer *peer) sizeof(struct in6_addr)); fl6->fl6_dport = htons(7001); fl6->fl6_sport = htons(7000); - dst = ip6_route_output(&init_net, NULL, fl6); + dst = ip6_route_output(net, NULL, fl6); if (dst->error) { _leave(" [route err %d]", dst->error); return; @@ -240,10 +242,11 @@ struct rxrpc_peer *rxrpc_alloc_peer(struct rxrpc_local *local, gfp_t gfp) /* * Initialise peer record. */ -static void rxrpc_init_peer(struct rxrpc_peer *peer, unsigned long hash_key) +static void rxrpc_init_peer(struct rxrpc_sock *rx, struct rxrpc_peer *peer, + unsigned long hash_key) { peer->hash_key = hash_key; - rxrpc_assess_MTU_size(peer); + rxrpc_assess_MTU_size(rx, peer); peer->mtu = peer->if_mtu; peer->rtt_last_req = ktime_get_real(); @@ -275,7 +278,8 @@ static void rxrpc_init_peer(struct rxrpc_peer *peer, unsigned long hash_key) /* * Set up a new peer. */ -static struct rxrpc_peer *rxrpc_create_peer(struct rxrpc_local *local, +static struct rxrpc_peer *rxrpc_create_peer(struct rxrpc_sock *rx, + struct rxrpc_local *local, struct sockaddr_rxrpc *srx, unsigned long hash_key, gfp_t gfp) @@ -287,7 +291,7 @@ static struct rxrpc_peer *rxrpc_create_peer(struct rxrpc_local *local, peer = rxrpc_alloc_peer(local, gfp); if (peer) { memcpy(&peer->srx, srx, sizeof(*srx)); - rxrpc_init_peer(peer, hash_key); + rxrpc_init_peer(rx, peer, hash_key); } _leave(" = %p", peer); @@ -299,14 +303,15 @@ static struct rxrpc_peer *rxrpc_create_peer(struct rxrpc_local *local, * since we've already done a search in the list from the non-reentrant context * (the data_ready handler) that is the only place we can add new peers. */ -void rxrpc_new_incoming_peer(struct rxrpc_local *local, struct rxrpc_peer *peer) +void rxrpc_new_incoming_peer(struct rxrpc_sock *rx, struct rxrpc_local *local, + struct rxrpc_peer *peer) { struct rxrpc_net *rxnet = local->rxnet; unsigned long hash_key; hash_key = rxrpc_peer_hash_key(local, &peer->srx); peer->local = local; - rxrpc_init_peer(peer, hash_key); + rxrpc_init_peer(rx, peer, hash_key); spin_lock(&rxnet->peer_hash_lock); hash_add_rcu(rxnet->peer_hash, &peer->hash_link, hash_key); @@ -317,7 +322,8 @@ void rxrpc_new_incoming_peer(struct rxrpc_local *local, struct rxrpc_peer *peer) /* * obtain a remote transport endpoint for the specified address */ -struct rxrpc_peer *rxrpc_lookup_peer(struct rxrpc_local *local, +struct rxrpc_peer *rxrpc_lookup_peer(struct rxrpc_sock *rx, + struct rxrpc_local *local, struct sockaddr_rxrpc *srx, gfp_t gfp) { struct rxrpc_peer *peer, *candidate; @@ -337,7 +343,7 @@ struct rxrpc_peer *rxrpc_lookup_peer(struct rxrpc_local *local, /* The peer is not yet present in hash - create a candidate * for a new record and then redo the search. */ - candidate = rxrpc_create_peer(local, srx, hash_key, gfp); + candidate = rxrpc_create_peer(rx, local, srx, hash_key, gfp); if (!candidate) { _leave(" = NULL [nomem]"); return NULL; -- cgit v1.2.3-59-g8ed1b From 2cfa2271604bb26e75b828d38f357ed084464795 Mon Sep 17 00:00:00 2001 From: David Howells Date: Fri, 5 Oct 2018 14:05:35 +0100 Subject: rxrpc: Fix the data_ready handler Fix the rxrpc_data_ready() function to pick up all packets and to not miss any. There are two problems: (1) The sk_data_ready pointer on the UDP socket is set *after* it is bound. This means that it's open for business before we're ready to dequeue packets and there's a tiny window exists in which a packet can sneak onto the receive queue, but we never know about it. Fix this by setting the pointers on the socket prior to binding it. (2) skb_recv_udp() will return an error (such as ENETUNREACH) if there was an error on the transmission side, even though we set the sk_error_report hook. Because rxrpc_data_ready() returns immediately in such a case, it never actually removes its packet from the receive queue. Fix this by abstracting out the UDP dequeuing and checksumming into a separate function that keeps hammering on skb_recv_udp() until it returns -EAGAIN, passing the packets extracted to the remainder of the function. and two potential problems: (3) It might be possible in some circumstances or in the future for packets to be being added to the UDP receive queue whilst rxrpc is running consuming them, so the data_ready() handler might get called less often than once per packet. Allow for this by fully draining the queue on each call as (2). (4) If a packet fails the checksum check, the code currently returns after discarding the packet without checking for more. Allow for this by fully draining the queue on each call as (2). Fixes: 17926a79320a ("[AF_RXRPC]: Provide secure RxRPC sockets for use by userspace and kernel both") Signed-off-by: David Howells Acked-by: Paolo Abeni --- net/rxrpc/input.c | 68 +++++++++++++++++++++++++++--------------------- net/rxrpc/local_object.c | 11 ++++---- 2 files changed, 44 insertions(+), 35 deletions(-) (limited to 'net/rxrpc') diff --git a/net/rxrpc/input.c b/net/rxrpc/input.c index c5af9955665b..c3114fa66c92 100644 --- a/net/rxrpc/input.c +++ b/net/rxrpc/input.c @@ -1121,7 +1121,7 @@ int rxrpc_extract_header(struct rxrpc_skb_priv *sp, struct sk_buff *skb) * shut down and the local endpoint from going away, thus sk_user_data will not * be cleared until this function returns. */ -void rxrpc_data_ready(struct sock *udp_sk) +void rxrpc_input_packet(struct sock *udp_sk, struct sk_buff *skb) { struct rxrpc_connection *conn; struct rxrpc_channel *chan; @@ -1130,39 +1130,11 @@ void rxrpc_data_ready(struct sock *udp_sk) struct rxrpc_local *local = udp_sk->sk_user_data; struct rxrpc_peer *peer = NULL; struct rxrpc_sock *rx = NULL; - struct sk_buff *skb; unsigned int channel; - int ret, skew = 0; + int skew = 0; _enter("%p", udp_sk); - ASSERT(!irqs_disabled()); - - skb = skb_recv_udp(udp_sk, 0, 1, &ret); - if (!skb) { - if (ret == -EAGAIN) - return; - _debug("UDP socket error %d", ret); - return; - } - - if (skb->tstamp == 0) - skb->tstamp = ktime_get_real(); - - rxrpc_new_skb(skb, rxrpc_skb_rx_received); - - _net("recv skb %p", skb); - - /* we'll probably need to checksum it (didn't call sock_recvmsg) */ - if (skb_checksum_complete(skb)) { - rxrpc_free_skb(skb, rxrpc_skb_rx_freed); - __UDP_INC_STATS(sock_net(udp_sk), UDP_MIB_INERRORS, 0); - _leave(" [CSUM failed]"); - return; - } - - __UDP_INC_STATS(sock_net(udp_sk), UDP_MIB_INDATAGRAMS, 0); - /* The UDP protocol already released all skb resources; * we are free to add our own data there. */ @@ -1181,6 +1153,8 @@ void rxrpc_data_ready(struct sock *udp_sk) } } + if (skb->tstamp == 0) + skb->tstamp = ktime_get_real(); trace_rxrpc_rx_packet(sp); switch (sp->hdr.type) { @@ -1398,3 +1372,37 @@ reject_packet: rxrpc_reject_packet(local, skb); _leave(" [badmsg]"); } + +void rxrpc_data_ready(struct sock *udp_sk) +{ + struct sk_buff *skb; + int ret; + + for (;;) { + skb = skb_recv_udp(udp_sk, 0, 1, &ret); + if (!skb) { + if (ret == -EAGAIN) + return; + + /* If there was a transmission failure, we get an error + * here that we need to ignore. + */ + _debug("UDP socket error %d", ret); + continue; + } + + rxrpc_new_skb(skb, rxrpc_skb_rx_received); + + /* we'll probably need to checksum it (didn't call sock_recvmsg) */ + if (skb_checksum_complete(skb)) { + rxrpc_free_skb(skb, rxrpc_skb_rx_freed); + __UDP_INC_STATS(sock_net(udp_sk), UDP_MIB_INERRORS, 0); + _debug("csum failed"); + continue; + } + + __UDP_INC_STATS(sock_net(udp_sk), UDP_MIB_INDATAGRAMS, 0); + + rxrpc_input_packet(udp_sk, skb); + } +} diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c index 94d234e9c685..30862f44c9f1 100644 --- a/net/rxrpc/local_object.c +++ b/net/rxrpc/local_object.c @@ -122,6 +122,12 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net) return ret; } + /* set the socket up */ + sock = local->socket->sk; + sock->sk_user_data = local; + sock->sk_data_ready = rxrpc_data_ready; + sock->sk_error_report = rxrpc_error_report; + /* if a local address was supplied then bind it */ if (local->srx.transport_len > sizeof(sa_family_t)) { _debug("bind"); @@ -191,11 +197,6 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net) BUG(); } - /* set the socket up */ - sock = local->socket->sk; - sock->sk_user_data = local; - sock->sk_data_ready = rxrpc_data_ready; - sock->sk_error_report = rxrpc_error_report; _leave(" = 0"); return 0; -- cgit v1.2.3-59-g8ed1b From 5271953cad31b97dea80f848c16e96ad66401199 Mon Sep 17 00:00:00 2001 From: David Howells Date: Thu, 4 Oct 2018 11:10:51 +0100 Subject: rxrpc: Use the UDP encap_rcv hook Use the UDP encap_rcv hook to cut the bit out of the rxrpc packet reception in which a packet is placed onto the UDP receive queue and then immediately removed again by rxrpc. Going via the queue in this manner seems like it should be unnecessary. This does, however, require the invention of a value to place in encap_type as that's one of the conditions to switch packets out to the encap_rcv hook. Possibly the value doesn't actually matter for anything other than sockopts on the UDP socket, which aren't accessible outside of rxrpc anyway. This seems to cut a bit of time out of the time elapsed between each sk_buff being timestamped and turning up in rxrpc (the final number in the following trace excerpts). I measured this by making the rxrpc_rx_packet trace point print the time elapsed between the skb being timestamped and the current time (in ns), e.g.: ... 424.278721: rxrpc_rx_packet: ... ACK 25026 So doing a 512MiB DIO read from my test server, with an unmodified kernel: N min max sum mean stddev 27605 2626 7581 7.83992e+07 2840.04 181.029 and with the patch applied: N min max sum mean stddev 27547 1895 12165 6.77461e+07 2459.29 255.02 Signed-off-by: David Howells --- include/uapi/linux/udp.h | 1 + net/rxrpc/ar-internal.h | 2 +- net/rxrpc/input.c | 50 ++++++++++++------------------------------------ net/rxrpc/local_object.c | 27 +++++++++++++++++++++----- 4 files changed, 36 insertions(+), 44 deletions(-) (limited to 'net/rxrpc') diff --git a/include/uapi/linux/udp.h b/include/uapi/linux/udp.h index 09d00f8c442b..09502de447f5 100644 --- a/include/uapi/linux/udp.h +++ b/include/uapi/linux/udp.h @@ -40,5 +40,6 @@ struct udphdr { #define UDP_ENCAP_L2TPINUDP 3 /* rfc2661 */ #define UDP_ENCAP_GTP0 4 /* GSM TS 09.60 */ #define UDP_ENCAP_GTP1U 5 /* 3GPP TS 29.060 */ +#define UDP_ENCAP_RXRPC 6 #endif /* _UAPI_LINUX_UDP_H */ diff --git a/net/rxrpc/ar-internal.h b/net/rxrpc/ar-internal.h index 63c43b3a2096..ab60c0313fd4 100644 --- a/net/rxrpc/ar-internal.h +++ b/net/rxrpc/ar-internal.h @@ -966,7 +966,7 @@ void rxrpc_unpublish_service_conn(struct rxrpc_connection *); /* * input.c */ -void rxrpc_data_ready(struct sock *); +int rxrpc_input_packet(struct sock *, struct sk_buff *); /* * insecure.c diff --git a/net/rxrpc/input.c b/net/rxrpc/input.c index c3114fa66c92..1866aeef2284 100644 --- a/net/rxrpc/input.c +++ b/net/rxrpc/input.c @@ -1121,7 +1121,7 @@ int rxrpc_extract_header(struct rxrpc_skb_priv *sp, struct sk_buff *skb) * shut down and the local endpoint from going away, thus sk_user_data will not * be cleared until this function returns. */ -void rxrpc_input_packet(struct sock *udp_sk, struct sk_buff *skb) +int rxrpc_input_packet(struct sock *udp_sk, struct sk_buff *skb) { struct rxrpc_connection *conn; struct rxrpc_channel *chan; @@ -1135,6 +1135,13 @@ void rxrpc_input_packet(struct sock *udp_sk, struct sk_buff *skb) _enter("%p", udp_sk); + if (skb->tstamp == 0) + skb->tstamp = ktime_get_real(); + + rxrpc_new_skb(skb, rxrpc_skb_rx_received); + + skb_pull(skb, sizeof(struct udphdr)); + /* The UDP protocol already released all skb resources; * we are free to add our own data there. */ @@ -1148,8 +1155,8 @@ void rxrpc_input_packet(struct sock *udp_sk, struct sk_buff *skb) static int lose; if ((lose++ & 7) == 7) { trace_rxrpc_rx_lose(sp); - rxrpc_lose_skb(skb, rxrpc_skb_rx_lost); - return; + rxrpc_free_skb(skb, rxrpc_skb_rx_lost); + return 0; } } @@ -1332,7 +1339,7 @@ discard: rxrpc_free_skb(skb, rxrpc_skb_rx_freed); out: trace_rxrpc_rx_done(0, 0); - return; + return 0; out_unlock: rcu_read_unlock(); @@ -1371,38 +1378,5 @@ reject_packet: trace_rxrpc_rx_done(skb->mark, skb->priority); rxrpc_reject_packet(local, skb); _leave(" [badmsg]"); -} - -void rxrpc_data_ready(struct sock *udp_sk) -{ - struct sk_buff *skb; - int ret; - - for (;;) { - skb = skb_recv_udp(udp_sk, 0, 1, &ret); - if (!skb) { - if (ret == -EAGAIN) - return; - - /* If there was a transmission failure, we get an error - * here that we need to ignore. - */ - _debug("UDP socket error %d", ret); - continue; - } - - rxrpc_new_skb(skb, rxrpc_skb_rx_received); - - /* we'll probably need to checksum it (didn't call sock_recvmsg) */ - if (skb_checksum_complete(skb)) { - rxrpc_free_skb(skb, rxrpc_skb_rx_freed); - __UDP_INC_STATS(sock_net(udp_sk), UDP_MIB_INERRORS, 0); - _debug("csum failed"); - continue; - } - - __UDP_INC_STATS(sock_net(udp_sk), UDP_MIB_INDATAGRAMS, 0); - - rxrpc_input_packet(udp_sk, skb); - } + return 0; } diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c index 30862f44c9f1..cad0691c2bb4 100644 --- a/net/rxrpc/local_object.c +++ b/net/rxrpc/local_object.c @@ -19,6 +19,7 @@ #include #include #include +#include #include #include "ar-internal.h" @@ -108,7 +109,7 @@ static struct rxrpc_local *rxrpc_alloc_local(struct rxrpc_net *rxnet, */ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net) { - struct sock *sock; + struct sock *usk; int ret, opt; _enter("%p{%d,%d}", @@ -123,10 +124,26 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net) } /* set the socket up */ - sock = local->socket->sk; - sock->sk_user_data = local; - sock->sk_data_ready = rxrpc_data_ready; - sock->sk_error_report = rxrpc_error_report; + usk = local->socket->sk; + inet_sk(usk)->mc_loop = 0; + + /* Enable CHECKSUM_UNNECESSARY to CHECKSUM_COMPLETE conversion */ + inet_inc_convert_csum(usk); + + rcu_assign_sk_user_data(usk, local); + + udp_sk(usk)->encap_type = UDP_ENCAP_RXRPC; + udp_sk(usk)->encap_rcv = rxrpc_input_packet; + udp_sk(usk)->encap_destroy = NULL; + udp_sk(usk)->gro_receive = NULL; + udp_sk(usk)->gro_complete = NULL; + + udp_encap_enable(); +#if IS_ENABLED(CONFIG_IPV6) + if (local->srx.transport.family == AF_INET6) + udpv6_encap_enable(); +#endif + usk->sk_error_report = rxrpc_error_report; /* if a local address was supplied then bind it */ if (local->srx.transport_len > sizeof(sa_family_t)) { -- cgit v1.2.3-59-g8ed1b From bfd2821117a751763718f1b5e57216c0d9b19a49 Mon Sep 17 00:00:00 2001 From: David Howells Date: Mon, 8 Oct 2018 15:45:56 +0100 Subject: rxrpc: Don't need to take the RCU read lock in the packet receiver We don't need to take the RCU read lock in the rxrpc packet receive function because it's held further up the stack in the IP input routine around the UDP receive routines. Fix this by dropping the RCU read lock calls from rxrpc_input_packet(). This simplifies the code. Fixes: 70790dbe3f66 ("rxrpc: Pass the last Tx packet marker in the annotation buffer") Signed-off-by: David Howells --- net/rxrpc/input.c | 41 +++++++++++++---------------------------- 1 file changed, 13 insertions(+), 28 deletions(-) (limited to 'net/rxrpc') diff --git a/net/rxrpc/input.c b/net/rxrpc/input.c index 1866aeef2284..2dcef482acf2 100644 --- a/net/rxrpc/input.c +++ b/net/rxrpc/input.c @@ -1120,6 +1120,8 @@ int rxrpc_extract_header(struct rxrpc_skb_priv *sp, struct sk_buff *skb) * The socket is locked by the caller and this prevents the socket from being * shut down and the local endpoint from going away, thus sk_user_data will not * be cleared until this function returns. + * + * Called with the RCU read lock held from the IP layer via UDP. */ int rxrpc_input_packet(struct sock *udp_sk, struct sk_buff *skb) { @@ -1215,8 +1217,6 @@ int rxrpc_input_packet(struct sock *udp_sk, struct sk_buff *skb) if (sp->hdr.serviceId == 0) goto bad_message; - rcu_read_lock(); - if (rxrpc_to_server(sp)) { /* Weed out packets to services we're not offering. Packets * that would begin a call are explicitly rejected and the rest @@ -1228,7 +1228,7 @@ int rxrpc_input_packet(struct sock *udp_sk, struct sk_buff *skb) if (sp->hdr.type == RXRPC_PACKET_TYPE_DATA && sp->hdr.seq == 1) goto unsupported_service; - goto discard_unlock; + goto discard; } } @@ -1248,7 +1248,7 @@ int rxrpc_input_packet(struct sock *udp_sk, struct sk_buff *skb) /* Connection-level packet */ _debug("CONN %p {%d}", conn, conn->debug_id); rxrpc_post_packet_to_conn(conn, skb); - goto out_unlock; + goto out; } /* Note the serial number skew here */ @@ -1267,19 +1267,19 @@ int rxrpc_input_packet(struct sock *udp_sk, struct sk_buff *skb) /* Ignore really old calls */ if (sp->hdr.callNumber < chan->last_call) - goto discard_unlock; + goto discard; if (sp->hdr.callNumber == chan->last_call) { if (chan->call || sp->hdr.type == RXRPC_PACKET_TYPE_ABORT) - goto discard_unlock; + goto discard; /* For the previous service call, if completed * successfully, we discard all further packets. */ if (rxrpc_conn_is_service(conn) && chan->last_type == RXRPC_PACKET_TYPE_ACK) - goto discard_unlock; + goto discard; /* But otherwise we need to retransmit the final packet * from data cached in the connection record. @@ -1290,16 +1290,14 @@ int rxrpc_input_packet(struct sock *udp_sk, struct sk_buff *skb) sp->hdr.serial, sp->hdr.flags, 0); rxrpc_post_packet_to_conn(conn, skb); - goto out_unlock; + goto out; } call = rcu_dereference(chan->call); if (sp->hdr.callNumber > chan->call_id) { - if (rxrpc_to_client(sp)) { - rcu_read_unlock(); + if (rxrpc_to_client(sp)) goto reject_packet; - } if (call) rxrpc_input_implicit_end_call(conn, call); call = NULL; @@ -1318,55 +1316,42 @@ int rxrpc_input_packet(struct sock *udp_sk, struct sk_buff *skb) if (!call || atomic_read(&call->usage) == 0) { if (rxrpc_to_client(sp) || sp->hdr.type != RXRPC_PACKET_TYPE_DATA) - goto bad_message_unlock; + goto bad_message; if (sp->hdr.seq != 1) - goto discard_unlock; + goto discard; call = rxrpc_new_incoming_call(local, rx, peer, conn, skb); - if (!call) { - rcu_read_unlock(); + if (!call) goto reject_packet; - } rxrpc_send_ping(call, skb, skew); mutex_unlock(&call->user_mutex); } rxrpc_input_call_packet(call, skb, skew); - goto discard_unlock; + goto discard; -discard_unlock: - rcu_read_unlock(); discard: rxrpc_free_skb(skb, rxrpc_skb_rx_freed); out: trace_rxrpc_rx_done(0, 0); return 0; -out_unlock: - rcu_read_unlock(); - goto out; - wrong_security: - rcu_read_unlock(); trace_rxrpc_abort(0, "SEC", sp->hdr.cid, sp->hdr.callNumber, sp->hdr.seq, RXKADINCONSISTENCY, EBADMSG); skb->priority = RXKADINCONSISTENCY; goto post_abort; unsupported_service: - rcu_read_unlock(); trace_rxrpc_abort(0, "INV", sp->hdr.cid, sp->hdr.callNumber, sp->hdr.seq, RX_INVALID_OPERATION, EOPNOTSUPP); skb->priority = RX_INVALID_OPERATION; goto post_abort; reupgrade: - rcu_read_unlock(); trace_rxrpc_abort(0, "UPG", sp->hdr.cid, sp->hdr.callNumber, sp->hdr.seq, RX_PROTOCOL_ERROR, EBADMSG); goto protocol_error; -bad_message_unlock: - rcu_read_unlock(); bad_message: trace_rxrpc_abort(0, "BAD", sp->hdr.cid, sp->hdr.callNumber, sp->hdr.seq, RX_PROTOCOL_ERROR, EBADMSG); -- cgit v1.2.3-59-g8ed1b From c479d5f2c2e1ce609da08c075054440d97ddff52 Mon Sep 17 00:00:00 2001 From: David Howells Date: Mon, 8 Oct 2018 15:46:01 +0100 Subject: rxrpc: Don't check RXRPC_CALL_TX_LAST after calling rxrpc_rotate_tx_window() We should only call the function to end a call's Tx phase if we rotated the marked-last packet out of the transmission buffer. Make rxrpc_rotate_tx_window() return an indication of whether it just rotated the packet marked as the last out of the transmit buffer, carrying the information out of the locked section in that function. We can then check the return value instead of examining RXRPC_CALL_TX_LAST. Fixes: 70790dbe3f66 ("rxrpc: Pass the last Tx packet marker in the annotation buffer") Signed-off-by: David Howells --- net/rxrpc/input.c | 35 +++++++++++++++++++---------------- 1 file changed, 19 insertions(+), 16 deletions(-) (limited to 'net/rxrpc') diff --git a/net/rxrpc/input.c b/net/rxrpc/input.c index 2dcef482acf2..8834aa627371 100644 --- a/net/rxrpc/input.c +++ b/net/rxrpc/input.c @@ -216,10 +216,11 @@ static void rxrpc_send_ping(struct rxrpc_call *call, struct sk_buff *skb, /* * Apply a hard ACK by advancing the Tx window. */ -static void rxrpc_rotate_tx_window(struct rxrpc_call *call, rxrpc_seq_t to, +static bool rxrpc_rotate_tx_window(struct rxrpc_call *call, rxrpc_seq_t to, struct rxrpc_ack_summary *summary) { struct sk_buff *skb, *list = NULL; + bool rot_last = false; int ix; u8 annotation; @@ -243,15 +244,17 @@ static void rxrpc_rotate_tx_window(struct rxrpc_call *call, rxrpc_seq_t to, skb->next = list; list = skb; - if (annotation & RXRPC_TX_ANNO_LAST) + if (annotation & RXRPC_TX_ANNO_LAST) { set_bit(RXRPC_CALL_TX_LAST, &call->flags); + rot_last = true; + } if ((annotation & RXRPC_TX_ANNO_MASK) != RXRPC_TX_ANNO_ACK) summary->nr_rot_new_acks++; } spin_unlock(&call->lock); - trace_rxrpc_transmit(call, (test_bit(RXRPC_CALL_TX_LAST, &call->flags) ? + trace_rxrpc_transmit(call, (rot_last ? rxrpc_transmit_rotate_last : rxrpc_transmit_rotate)); wake_up(&call->waitq); @@ -262,6 +265,8 @@ static void rxrpc_rotate_tx_window(struct rxrpc_call *call, rxrpc_seq_t to, skb->next = NULL; rxrpc_free_skb(skb, rxrpc_skb_tx_freed); } + + return rot_last; } /* @@ -332,11 +337,11 @@ static bool rxrpc_receiving_reply(struct rxrpc_call *call) trace_rxrpc_timer(call, rxrpc_timer_init_for_reply, now); } - if (!test_bit(RXRPC_CALL_TX_LAST, &call->flags)) - rxrpc_rotate_tx_window(call, top, &summary); if (!test_bit(RXRPC_CALL_TX_LAST, &call->flags)) { - rxrpc_proto_abort("TXL", call, top); - return false; + if (!rxrpc_rotate_tx_window(call, top, &summary)) { + rxrpc_proto_abort("TXL", call, top); + return false; + } } if (!rxrpc_end_tx_phase(call, true, "ETD")) return false; @@ -897,8 +902,12 @@ static void rxrpc_input_ack(struct rxrpc_call *call, struct sk_buff *skb, if (nr_acks > call->tx_top - hard_ack) return rxrpc_proto_abort("AKN", call, 0); - if (after(hard_ack, call->tx_hard_ack)) - rxrpc_rotate_tx_window(call, hard_ack, &summary); + if (after(hard_ack, call->tx_hard_ack)) { + if (rxrpc_rotate_tx_window(call, hard_ack, &summary)) { + rxrpc_end_tx_phase(call, false, "ETA"); + return; + } + } if (nr_acks > 0) { if (skb_copy_bits(skb, offset, buf.acks, nr_acks) < 0) @@ -907,11 +916,6 @@ static void rxrpc_input_ack(struct rxrpc_call *call, struct sk_buff *skb, &summary); } - if (test_bit(RXRPC_CALL_TX_LAST, &call->flags)) { - rxrpc_end_tx_phase(call, false, "ETA"); - return; - } - if (call->rxtx_annotations[call->tx_top & RXRPC_RXTX_BUFF_MASK] & RXRPC_TX_ANNO_LAST && summary.nr_acks == call->tx_top - hard_ack && @@ -933,8 +937,7 @@ static void rxrpc_input_ackall(struct rxrpc_call *call, struct sk_buff *skb) _proto("Rx ACKALL %%%u", sp->hdr.serial); - rxrpc_rotate_tx_window(call, call->tx_top, &summary); - if (test_bit(RXRPC_CALL_TX_LAST, &call->flags)) + if (rxrpc_rotate_tx_window(call, call->tx_top, &summary)) rxrpc_end_tx_phase(call, false, "ETL"); } -- cgit v1.2.3-59-g8ed1b From dfe995224693798e554ab4770f6d8a096afc60cd Mon Sep 17 00:00:00 2001 From: David Howells Date: Mon, 8 Oct 2018 15:46:05 +0100 Subject: rxrpc: Carry call state out of locked section in rxrpc_rotate_tx_window() Carry the call state out of the locked section in rxrpc_rotate_tx_window() rather than sampling it afterwards. This is only used to select tracepoint data, but could have changed by the time we do the tracepoint. Signed-off-by: David Howells --- net/rxrpc/input.c | 14 ++++++++------ 1 file changed, 8 insertions(+), 6 deletions(-) (limited to 'net/rxrpc') diff --git a/net/rxrpc/input.c b/net/rxrpc/input.c index 8834aa627371..af8ce64f4162 100644 --- a/net/rxrpc/input.c +++ b/net/rxrpc/input.c @@ -278,23 +278,26 @@ static bool rxrpc_rotate_tx_window(struct rxrpc_call *call, rxrpc_seq_t to, static bool rxrpc_end_tx_phase(struct rxrpc_call *call, bool reply_begun, const char *abort_why) { + unsigned int state; ASSERT(test_bit(RXRPC_CALL_TX_LAST, &call->flags)); write_lock(&call->state_lock); - switch (call->state) { + state = call->state; + switch (state) { case RXRPC_CALL_CLIENT_SEND_REQUEST: case RXRPC_CALL_CLIENT_AWAIT_REPLY: if (reply_begun) - call->state = RXRPC_CALL_CLIENT_RECV_REPLY; + call->state = state = RXRPC_CALL_CLIENT_RECV_REPLY; else - call->state = RXRPC_CALL_CLIENT_AWAIT_REPLY; + call->state = state = RXRPC_CALL_CLIENT_AWAIT_REPLY; break; case RXRPC_CALL_SERVER_AWAIT_ACK: __rxrpc_call_completed(call); rxrpc_notify_socket(call); + state = call->state; break; default: @@ -302,11 +305,10 @@ static bool rxrpc_end_tx_phase(struct rxrpc_call *call, bool reply_begun, } write_unlock(&call->state_lock); - if (call->state == RXRPC_CALL_CLIENT_AWAIT_REPLY) { + if (state == RXRPC_CALL_CLIENT_AWAIT_REPLY) trace_rxrpc_transmit(call, rxrpc_transmit_await_reply); - } else { + else trace_rxrpc_transmit(call, rxrpc_transmit_end); - } _leave(" = ok"); return true; -- cgit v1.2.3-59-g8ed1b From 298bc15b2079c324e82d0a6fda39c3d762af7282 Mon Sep 17 00:00:00 2001 From: David Howells Date: Mon, 8 Oct 2018 15:46:11 +0100 Subject: rxrpc: Only take the rwind and mtu values from latest ACK Move the out-of-order and duplicate ACK packet check to before the call to rxrpc_input_ackinfo() so that the receive window size and MTU size are only checked in the latest ACK packet and don't regress. Fixes: 248f219cb8bc ("rxrpc: Rewrite the data and ack handling code") Signed-off-by: David Howells --- net/rxrpc/input.c | 19 ++++++++++--------- 1 file changed, 10 insertions(+), 9 deletions(-) (limited to 'net/rxrpc') diff --git a/net/rxrpc/input.c b/net/rxrpc/input.c index af8ce64f4162..04213a65c1ac 100644 --- a/net/rxrpc/input.c +++ b/net/rxrpc/input.c @@ -868,6 +868,16 @@ static void rxrpc_input_ack(struct rxrpc_call *call, struct sk_buff *skb, rxrpc_propose_ack_respond_to_ack); } + /* Discard any out-of-order or duplicate ACKs. */ + if (before_eq(sp->hdr.serial, call->acks_latest)) { + _debug("discard ACK %d <= %d", + sp->hdr.serial, call->acks_latest); + return; + } + call->acks_latest_ts = skb->tstamp; + call->acks_latest = sp->hdr.serial; + + /* Parse rwind and mtu sizes if provided. */ ioffset = offset + nr_acks + 3; if (skb->len >= ioffset + sizeof(buf.info)) { if (skb_copy_bits(skb, ioffset, &buf.info, sizeof(buf.info)) < 0) @@ -889,15 +899,6 @@ static void rxrpc_input_ack(struct rxrpc_call *call, struct sk_buff *skb, return; } - /* Discard any out-of-order or duplicate ACKs. */ - if (before_eq(sp->hdr.serial, call->acks_latest)) { - _debug("discard ACK %d <= %d", - sp->hdr.serial, call->acks_latest); - return; - } - call->acks_latest_ts = skb->tstamp; - call->acks_latest = sp->hdr.serial; - if (before(hard_ack, call->tx_hard_ack) || after(hard_ack, call->tx_top)) return rxrpc_proto_abort("AKW", call, 0); -- cgit v1.2.3-59-g8ed1b From 647530924f47c93db472ee3cf43b7ef1425581b6 Mon Sep 17 00:00:00 2001 From: David Howells Date: Mon, 8 Oct 2018 15:46:17 +0100 Subject: rxrpc: Fix connection-level abort handling Fix connection-level abort handling to cache the abort and error codes properly so that a new incoming call can be properly aborted if it races with the parent connection being aborted by another CPU. The abort_code and error parameters can then be dropped from rxrpc_abort_calls(). Fixes: f5c17aaeb2ae ("rxrpc: Calls should only have one terminal state") Signed-off-by: David Howells --- net/rxrpc/ar-internal.h | 4 ++-- net/rxrpc/call_accept.c | 4 ++-- net/rxrpc/conn_event.c | 26 +++++++++++++++----------- 3 files changed, 19 insertions(+), 15 deletions(-) (limited to 'net/rxrpc') diff --git a/net/rxrpc/ar-internal.h b/net/rxrpc/ar-internal.h index ab60c0313fd4..45307463b7dd 100644 --- a/net/rxrpc/ar-internal.h +++ b/net/rxrpc/ar-internal.h @@ -442,8 +442,7 @@ struct rxrpc_connection { spinlock_t state_lock; /* state-change lock */ enum rxrpc_conn_cache_state cache_state; enum rxrpc_conn_proto_state state; /* current state of connection */ - u32 local_abort; /* local abort code */ - u32 remote_abort; /* remote abort code */ + u32 abort_code; /* Abort code of connection abort */ int debug_id; /* debug ID for printks */ atomic_t serial; /* packet serial number counter */ unsigned int hi_serial; /* highest serial number received */ @@ -453,6 +452,7 @@ struct rxrpc_connection { u8 security_size; /* security header size */ u8 security_ix; /* security type */ u8 out_clientflag; /* RXRPC_CLIENT_INITIATED if we are client */ + short error; /* Local error code */ }; static inline bool rxrpc_to_server(const struct rxrpc_skb_priv *sp) diff --git a/net/rxrpc/call_accept.c b/net/rxrpc/call_accept.c index f55f67894465..1c4ebc0cb25b 100644 --- a/net/rxrpc/call_accept.c +++ b/net/rxrpc/call_accept.c @@ -405,11 +405,11 @@ struct rxrpc_call *rxrpc_new_incoming_call(struct rxrpc_local *local, case RXRPC_CONN_REMOTELY_ABORTED: rxrpc_set_call_completion(call, RXRPC_CALL_REMOTELY_ABORTED, - conn->remote_abort, -ECONNABORTED); + conn->abort_code, conn->error); break; case RXRPC_CONN_LOCALLY_ABORTED: rxrpc_abort_call("CON", call, sp->hdr.seq, - conn->local_abort, -ECONNABORTED); + conn->abort_code, conn->error); break; default: BUG(); diff --git a/net/rxrpc/conn_event.c b/net/rxrpc/conn_event.c index 6df56ce68861..b6fca8ebb117 100644 --- a/net/rxrpc/conn_event.c +++ b/net/rxrpc/conn_event.c @@ -126,7 +126,7 @@ static void rxrpc_conn_retransmit_call(struct rxrpc_connection *conn, switch (chan->last_type) { case RXRPC_PACKET_TYPE_ABORT: - _proto("Tx ABORT %%%u { %d } [re]", serial, conn->local_abort); + _proto("Tx ABORT %%%u { %d } [re]", serial, conn->abort_code); break; case RXRPC_PACKET_TYPE_ACK: trace_rxrpc_tx_ack(chan->call_debug_id, serial, @@ -153,13 +153,12 @@ static void rxrpc_conn_retransmit_call(struct rxrpc_connection *conn, * pass a connection-level abort onto all calls on that connection */ static void rxrpc_abort_calls(struct rxrpc_connection *conn, - enum rxrpc_call_completion compl, - u32 abort_code, int error) + enum rxrpc_call_completion compl) { struct rxrpc_call *call; int i; - _enter("{%d},%x", conn->debug_id, abort_code); + _enter("{%d},%x", conn->debug_id, conn->abort_code); spin_lock(&conn->channel_lock); @@ -172,9 +171,11 @@ static void rxrpc_abort_calls(struct rxrpc_connection *conn, trace_rxrpc_abort(call->debug_id, "CON", call->cid, call->call_id, 0, - abort_code, error); + conn->abort_code, + conn->error); if (rxrpc_set_call_completion(call, compl, - abort_code, error)) + conn->abort_code, + conn->error)) rxrpc_notify_socket(call); } } @@ -207,10 +208,12 @@ static int rxrpc_abort_connection(struct rxrpc_connection *conn, return 0; } + conn->error = error; + conn->abort_code = abort_code; conn->state = RXRPC_CONN_LOCALLY_ABORTED; spin_unlock_bh(&conn->state_lock); - rxrpc_abort_calls(conn, RXRPC_CALL_LOCALLY_ABORTED, abort_code, error); + rxrpc_abort_calls(conn, RXRPC_CALL_LOCALLY_ABORTED); msg.msg_name = &conn->params.peer->srx.transport; msg.msg_namelen = conn->params.peer->srx.transport_len; @@ -229,7 +232,7 @@ static int rxrpc_abort_connection(struct rxrpc_connection *conn, whdr._rsvd = 0; whdr.serviceId = htons(conn->service_id); - word = htonl(conn->local_abort); + word = htonl(conn->abort_code); iov[0].iov_base = &whdr; iov[0].iov_len = sizeof(whdr); @@ -240,7 +243,7 @@ static int rxrpc_abort_connection(struct rxrpc_connection *conn, serial = atomic_inc_return(&conn->serial); whdr.serial = htonl(serial); - _proto("Tx CONN ABORT %%%u { %d }", serial, conn->local_abort); + _proto("Tx CONN ABORT %%%u { %d }", serial, conn->abort_code); ret = kernel_sendmsg(conn->params.local->socket, &msg, iov, 2, len); if (ret < 0) { @@ -315,9 +318,10 @@ static int rxrpc_process_event(struct rxrpc_connection *conn, abort_code = ntohl(wtmp); _proto("Rx ABORT %%%u { ac=%d }", sp->hdr.serial, abort_code); + conn->error = -ECONNABORTED; + conn->abort_code = abort_code; conn->state = RXRPC_CONN_REMOTELY_ABORTED; - rxrpc_abort_calls(conn, RXRPC_CALL_REMOTELY_ABORTED, - abort_code, -ECONNABORTED); + rxrpc_abort_calls(conn, RXRPC_CALL_REMOTELY_ABORTED); return -ECONNABORTED; case RXRPC_PACKET_TYPE_CHALLENGE: -- cgit v1.2.3-59-g8ed1b From c1e15b4944c9fa7fbbb74f7a5920a1e31b4b965a Mon Sep 17 00:00:00 2001 From: David Howells Date: Mon, 8 Oct 2018 15:46:25 +0100 Subject: rxrpc: Fix the packet reception routine The rxrpc_input_packet() function and its call tree was built around the assumption that data_ready() handler called from UDP to inform a kernel service that there is data to be had was non-reentrant. This means that certain locking could be dispensed with. This, however, turns out not to be the case with a multi-queue network card that can deliver packets to multiple cpus simultaneously. Each of those cpus can be in the rxrpc_input_packet() function at the same time. Fix by adding or changing some structure members: (1) Add peer->rtt_input_lock to serialise access to the RTT buffer. (2) Make conn->service_id into a 32-bit variable so that it can be cmpxchg'd on all arches. (3) Add call->input_lock to serialise access to the Rx/Tx state. Note that although the Rx and Tx states are (almost) entirely separate, there's no point completing the separation and having separate locks since it's a bi-phasal RPC protocol rather than a bi-direction streaming protocol. Data transmission and data reception do not take place simultaneously on any particular call. and making the following functional changes: (1) In rxrpc_input_data(), hold call->input_lock around the core to prevent simultaneous producing of packets into the Rx ring and updating of tracking state for a particular call. (2) In rxrpc_input_ping_response(), only read call->ping_serial once, and check it before checking RXRPC_CALL_PINGING as that's a cheaper test. The bit test and bit clear can then be combined. No further locking is needed here. (3) In rxrpc_input_ack(), take call->input_lock after we've parsed much of the ACK packet. The superseded ACK check is then done both before and after the lock is taken. The handing of ackinfo data is split, parsing before the lock is taken and processing with it held. This is keyed on rxMTU being non-zero. Congestion management is also done within the locked section. (4) In rxrpc_input_ackall(), take call->input_lock around the Tx window rotation. The ACKALL packet carries no information and is only really useful after all packets have been transmitted since it's imprecise. (5) In rxrpc_input_implicit_end_call(), we use rx->incoming_lock to prevent calls being simultaneously implicitly ended on two cpus and also to prevent any races with incoming call setup. (6) In rxrpc_input_packet(), use cmpxchg() to effect the service upgrade on a connection. It is only permitted to happen once for a connection. (7) In rxrpc_new_incoming_call(), we have to recheck the routing inside rx->incoming_lock to see if someone else set up the call, connection or peer whilst we were getting there. We can't trust the values from the earlier routing check unless we pin refs on them - which we want to avoid. Further, we need to allow for an incoming call to have its state changed on another CPU between us making it live and us adjusting it because the conn is now in the RXRPC_CONN_SERVICE state. (8) In rxrpc_peer_add_rtt(), take peer->rtt_input_lock around the access to the RTT buffer. Don't need to lock around setting peer->rtt. For reference, the inventory of state-accessing or state-altering functions used by the packet input procedure is: > rxrpc_input_packet() * PACKET CHECKING * ROUTING > rxrpc_post_packet_to_local() > rxrpc_find_connection_rcu() - uses RCU > rxrpc_lookup_peer_rcu() - uses RCU > rxrpc_find_service_conn_rcu() - uses RCU > idr_find() - uses RCU * CONNECTION-LEVEL PROCESSING - Service upgrade - Can only happen once per conn ! Changed to use cmpxchg > rxrpc_post_packet_to_conn() - Setting conn->hi_serial - Probably safe not using locks - Maybe use cmpxchg * CALL-LEVEL PROCESSING > Old-call checking > rxrpc_input_implicit_end_call() > rxrpc_call_completed() > rxrpc_queue_call() ! Need to take rx->incoming_lock > __rxrpc_disconnect_call() > rxrpc_notify_socket() > rxrpc_new_incoming_call() - Uses rx->incoming_lock for the entire process - Might be able to drop this earlier in favour of the call lock > rxrpc_incoming_call() ! Conflicts with rxrpc_input_implicit_end_call() > rxrpc_send_ping() - Don't need locks to check rtt state > rxrpc_propose_ACK * PACKET DISTRIBUTION > rxrpc_input_call_packet() > rxrpc_input_data() * QUEUE DATA PACKET ON CALL > rxrpc_reduce_call_timer() - Uses timer_reduce() ! Needs call->input_lock() > rxrpc_receiving_reply() ! Needs locking around ack state > rxrpc_rotate_tx_window() > rxrpc_end_tx_phase() > rxrpc_proto_abort() > rxrpc_input_dup_data() - Fills the Rx buffer - rxrpc_propose_ACK() - rxrpc_notify_socket() > rxrpc_input_ack() * APPLY ACK PACKET TO CALL AND DISCARD PACKET > rxrpc_input_ping_response() - Probably doesn't need any extra locking ! Need READ_ONCE() on call->ping_serial > rxrpc_input_check_for_lost_ack() - Takes call->lock to consult Tx buffer > rxrpc_peer_add_rtt() ! Needs to take a lock (peer->rtt_input_lock) ! Could perhaps manage with cmpxchg() and xadd() instead > rxrpc_input_requested_ack - Consults Tx buffer ! Probably needs a lock > rxrpc_peer_add_rtt() > rxrpc_propose_ack() > rxrpc_input_ackinfo() - Changes call->tx_winsize ! Use cmpxchg to handle change ! Should perhaps track serial number - Uses peer->lock to record MTU specification changes > rxrpc_proto_abort() ! Need to take call->input_lock > rxrpc_rotate_tx_window() > rxrpc_end_tx_phase() > rxrpc_input_soft_acks() - Consults the Tx buffer > rxrpc_congestion_management() - Modifies the Tx annotations ! Needs call->input_lock() > rxrpc_queue_call() > rxrpc_input_abort() * APPLY ABORT PACKET TO CALL AND DISCARD PACKET > rxrpc_set_call_completion() > rxrpc_notify_socket() > rxrpc_input_ackall() * APPLY ACKALL PACKET TO CALL AND DISCARD PACKET ! Need to take call->input_lock > rxrpc_rotate_tx_window() > rxrpc_end_tx_phase() > rxrpc_reject_packet() There are some functions used by the above that queue the packet, after which the procedure is terminated: - rxrpc_post_packet_to_local() - local->event_queue is an sk_buff_head - local->processor is a work_struct - rxrpc_post_packet_to_conn() - conn->rx_queue is an sk_buff_head - conn->processor is a work_struct - rxrpc_reject_packet() - local->reject_queue is an sk_buff_head - local->processor is a work_struct And some that offload processing to process context: - rxrpc_notify_socket() - Uses RCU lock - Uses call->notify_lock to call call->notify_rx - Uses call->recvmsg_lock to queue recvmsg side - rxrpc_queue_call() - call->processor is a work_struct - rxrpc_propose_ACK() - Uses call->lock to wrap __rxrpc_propose_ACK() And a bunch that complete a call, all of which use call->state_lock to protect the call state: - rxrpc_call_completed() - rxrpc_set_call_completion() - rxrpc_abort_call() - rxrpc_proto_abort() - Also uses rxrpc_queue_call() Fixes: 17926a79320a ("[AF_RXRPC]: Provide secure RxRPC sockets for use by userspace and kernel both") Signed-off-by: David Howells --- net/rxrpc/ar-internal.h | 7 +-- net/rxrpc/call_accept.c | 21 ++++++--- net/rxrpc/call_object.c | 1 + net/rxrpc/input.c | 120 +++++++++++++++++++++++++++++++----------------- net/rxrpc/peer_event.c | 5 ++ net/rxrpc/peer_object.c | 1 + 6 files changed, 105 insertions(+), 50 deletions(-) (limited to 'net/rxrpc') diff --git a/net/rxrpc/ar-internal.h b/net/rxrpc/ar-internal.h index 45307463b7dd..a6e6cae82c30 100644 --- a/net/rxrpc/ar-internal.h +++ b/net/rxrpc/ar-internal.h @@ -302,6 +302,7 @@ struct rxrpc_peer { /* calculated RTT cache */ #define RXRPC_RTT_CACHE_SIZE 32 + spinlock_t rtt_input_lock; /* RTT lock for input routine */ ktime_t rtt_last_req; /* Time of last RTT request */ u64 rtt; /* Current RTT estimate (in nS) */ u64 rtt_sum; /* Sum of cache contents */ @@ -447,7 +448,7 @@ struct rxrpc_connection { atomic_t serial; /* packet serial number counter */ unsigned int hi_serial; /* highest serial number received */ u32 security_nonce; /* response re-use preventer */ - u16 service_id; /* Service ID, possibly upgraded */ + u32 service_id; /* Service ID, possibly upgraded */ u8 size_align; /* data size alignment (for security) */ u8 security_size; /* security header size */ u8 security_ix; /* security type */ @@ -635,6 +636,8 @@ struct rxrpc_call { bool tx_phase; /* T if transmission phase, F if receive phase */ u8 nr_jumbo_bad; /* Number of jumbo dups/exceeds-windows */ + spinlock_t input_lock; /* Lock for packet input to this call */ + /* receive-phase ACK management */ u8 ackr_reason; /* reason to ACK */ u16 ackr_skew; /* skew on packet being ACK'd */ @@ -720,8 +723,6 @@ int rxrpc_service_prealloc(struct rxrpc_sock *, gfp_t); void rxrpc_discard_prealloc(struct rxrpc_sock *); struct rxrpc_call *rxrpc_new_incoming_call(struct rxrpc_local *, struct rxrpc_sock *, - struct rxrpc_peer *, - struct rxrpc_connection *, struct sk_buff *); void rxrpc_accept_incoming_calls(struct rxrpc_local *); struct rxrpc_call *rxrpc_accept_call(struct rxrpc_sock *, unsigned long, diff --git a/net/rxrpc/call_accept.c b/net/rxrpc/call_accept.c index 1c4ebc0cb25b..652e314de38e 100644 --- a/net/rxrpc/call_accept.c +++ b/net/rxrpc/call_accept.c @@ -333,11 +333,11 @@ static struct rxrpc_call *rxrpc_alloc_incoming_call(struct rxrpc_sock *rx, */ struct rxrpc_call *rxrpc_new_incoming_call(struct rxrpc_local *local, struct rxrpc_sock *rx, - struct rxrpc_peer *peer, - struct rxrpc_connection *conn, struct sk_buff *skb) { struct rxrpc_skb_priv *sp = rxrpc_skb(skb); + struct rxrpc_connection *conn; + struct rxrpc_peer *peer; struct rxrpc_call *call; _enter(""); @@ -354,6 +354,13 @@ struct rxrpc_call *rxrpc_new_incoming_call(struct rxrpc_local *local, goto out; } + /* The peer, connection and call may all have sprung into existence due + * to a duplicate packet being handled on another CPU in parallel, so + * we have to recheck the routing. However, we're now holding + * rx->incoming_lock, so the values should remain stable. + */ + conn = rxrpc_find_connection_rcu(local, skb, &peer); + call = rxrpc_alloc_incoming_call(rx, local, peer, conn, skb); if (!call) { skb->mark = RXRPC_SKB_MARK_REJECT_BUSY; @@ -396,10 +403,12 @@ struct rxrpc_call *rxrpc_new_incoming_call(struct rxrpc_local *local, case RXRPC_CONN_SERVICE: write_lock(&call->state_lock); - if (rx->discard_new_call) - call->state = RXRPC_CALL_SERVER_RECV_REQUEST; - else - call->state = RXRPC_CALL_SERVER_ACCEPTING; + if (call->state < RXRPC_CALL_COMPLETE) { + if (rx->discard_new_call) + call->state = RXRPC_CALL_SERVER_RECV_REQUEST; + else + call->state = RXRPC_CALL_SERVER_ACCEPTING; + } write_unlock(&call->state_lock); break; diff --git a/net/rxrpc/call_object.c b/net/rxrpc/call_object.c index 0ca2c2dfd196..8f1a8f85b1f9 100644 --- a/net/rxrpc/call_object.c +++ b/net/rxrpc/call_object.c @@ -138,6 +138,7 @@ struct rxrpc_call *rxrpc_alloc_call(struct rxrpc_sock *rx, gfp_t gfp, init_waitqueue_head(&call->waitq); spin_lock_init(&call->lock); spin_lock_init(&call->notify_lock); + spin_lock_init(&call->input_lock); rwlock_init(&call->state_lock); atomic_set(&call->usage, 1); call->debug_id = debug_id; diff --git a/net/rxrpc/input.c b/net/rxrpc/input.c index 04213a65c1ac..570b49d2da42 100644 --- a/net/rxrpc/input.c +++ b/net/rxrpc/input.c @@ -459,13 +459,15 @@ static void rxrpc_input_data(struct rxrpc_call *call, struct sk_buff *skb, } } + spin_lock(&call->input_lock); + /* Received data implicitly ACKs all of the request packets we sent * when we're acting as a client. */ if ((state == RXRPC_CALL_CLIENT_SEND_REQUEST || state == RXRPC_CALL_CLIENT_AWAIT_REPLY) && !rxrpc_receiving_reply(call)) - return; + goto unlock; call->ackr_prev_seq = seq; @@ -495,12 +497,16 @@ next_subpacket: if (flags & RXRPC_LAST_PACKET) { if (test_bit(RXRPC_CALL_RX_LAST, &call->flags) && - seq != call->rx_top) - return rxrpc_proto_abort("LSN", call, seq); + seq != call->rx_top) { + rxrpc_proto_abort("LSN", call, seq); + goto unlock; + } } else { if (test_bit(RXRPC_CALL_RX_LAST, &call->flags) && - after_eq(seq, call->rx_top)) - return rxrpc_proto_abort("LSA", call, seq); + after_eq(seq, call->rx_top)) { + rxrpc_proto_abort("LSA", call, seq); + goto unlock; + } } trace_rxrpc_rx_data(call->debug_id, seq, serial, flags, annotation); @@ -567,8 +573,10 @@ next_subpacket: skip: offset += len; if (flags & RXRPC_JUMBO_PACKET) { - if (skb_copy_bits(skb, offset, &flags, 1) < 0) - return rxrpc_proto_abort("XJF", call, seq); + if (skb_copy_bits(skb, offset, &flags, 1) < 0) { + rxrpc_proto_abort("XJF", call, seq); + goto unlock; + } offset += sizeof(struct rxrpc_jumbo_header); seq++; serial++; @@ -608,6 +616,9 @@ ack: trace_rxrpc_notify_socket(call->debug_id, serial); rxrpc_notify_socket(call); } + +unlock: + spin_unlock(&call->input_lock); _leave(" [queued]"); } @@ -694,15 +705,14 @@ static void rxrpc_input_ping_response(struct rxrpc_call *call, ping_time = call->ping_time; smp_rmb(); - ping_serial = call->ping_serial; + ping_serial = READ_ONCE(call->ping_serial); if (orig_serial == call->acks_lost_ping) rxrpc_input_check_for_lost_ack(call); - if (!test_bit(RXRPC_CALL_PINGING, &call->flags) || - before(orig_serial, ping_serial)) + if (before(orig_serial, ping_serial) || + !test_and_clear_bit(RXRPC_CALL_PINGING, &call->flags)) return; - clear_bit(RXRPC_CALL_PINGING, &call->flags); if (after(orig_serial, ping_serial)) return; @@ -869,24 +879,31 @@ static void rxrpc_input_ack(struct rxrpc_call *call, struct sk_buff *skb, } /* Discard any out-of-order or duplicate ACKs. */ - if (before_eq(sp->hdr.serial, call->acks_latest)) { - _debug("discard ACK %d <= %d", - sp->hdr.serial, call->acks_latest); + if (before_eq(sp->hdr.serial, call->acks_latest)) return; - } + + buf.info.rxMTU = 0; + ioffset = offset + nr_acks + 3; + if (skb->len >= ioffset + sizeof(buf.info) && + skb_copy_bits(skb, ioffset, &buf.info, sizeof(buf.info)) < 0) + return rxrpc_proto_abort("XAI", call, 0); + + spin_lock(&call->input_lock); + + /* Discard any out-of-order or duplicate ACKs. */ + if (before_eq(sp->hdr.serial, call->acks_latest)) + goto out; call->acks_latest_ts = skb->tstamp; call->acks_latest = sp->hdr.serial; /* Parse rwind and mtu sizes if provided. */ - ioffset = offset + nr_acks + 3; - if (skb->len >= ioffset + sizeof(buf.info)) { - if (skb_copy_bits(skb, ioffset, &buf.info, sizeof(buf.info)) < 0) - return rxrpc_proto_abort("XAI", call, 0); + if (buf.info.rxMTU) rxrpc_input_ackinfo(call, skb, &buf.info); - } - if (first_soft_ack == 0) - return rxrpc_proto_abort("AK0", call, 0); + if (first_soft_ack == 0) { + rxrpc_proto_abort("AK0", call, 0); + goto out; + } /* Ignore ACKs unless we are or have just been transmitting. */ switch (READ_ONCE(call->state)) { @@ -896,25 +913,31 @@ static void rxrpc_input_ack(struct rxrpc_call *call, struct sk_buff *skb, case RXRPC_CALL_SERVER_AWAIT_ACK: break; default: - return; + goto out; } if (before(hard_ack, call->tx_hard_ack) || - after(hard_ack, call->tx_top)) - return rxrpc_proto_abort("AKW", call, 0); - if (nr_acks > call->tx_top - hard_ack) - return rxrpc_proto_abort("AKN", call, 0); + after(hard_ack, call->tx_top)) { + rxrpc_proto_abort("AKW", call, 0); + goto out; + } + if (nr_acks > call->tx_top - hard_ack) { + rxrpc_proto_abort("AKN", call, 0); + goto out; + } if (after(hard_ack, call->tx_hard_ack)) { if (rxrpc_rotate_tx_window(call, hard_ack, &summary)) { rxrpc_end_tx_phase(call, false, "ETA"); - return; + goto out; } } if (nr_acks > 0) { - if (skb_copy_bits(skb, offset, buf.acks, nr_acks) < 0) - return rxrpc_proto_abort("XSA", call, 0); + if (skb_copy_bits(skb, offset, buf.acks, nr_acks) < 0) { + rxrpc_proto_abort("XSA", call, 0); + goto out; + } rxrpc_input_soft_acks(call, buf.acks, first_soft_ack, nr_acks, &summary); } @@ -927,7 +950,9 @@ static void rxrpc_input_ack(struct rxrpc_call *call, struct sk_buff *skb, false, true, rxrpc_propose_ack_ping_for_lost_reply); - return rxrpc_congestion_management(call, skb, &summary, acked_serial); + rxrpc_congestion_management(call, skb, &summary, acked_serial); +out: + spin_unlock(&call->input_lock); } /* @@ -940,8 +965,12 @@ static void rxrpc_input_ackall(struct rxrpc_call *call, struct sk_buff *skb) _proto("Rx ACKALL %%%u", sp->hdr.serial); + spin_lock(&call->input_lock); + if (rxrpc_rotate_tx_window(call, call->tx_top, &summary)) rxrpc_end_tx_phase(call, false, "ETL"); + + spin_unlock(&call->input_lock); } /* @@ -1024,18 +1053,19 @@ static void rxrpc_input_call_packet(struct rxrpc_call *call, } /* - * Handle a new call on a channel implicitly completing the preceding call on - * that channel. + * Handle a new service call on a channel implicitly completing the preceding + * call on that channel. This does not apply to client conns. * * TODO: If callNumber > call_id + 1, renegotiate security. */ -static void rxrpc_input_implicit_end_call(struct rxrpc_connection *conn, +static void rxrpc_input_implicit_end_call(struct rxrpc_sock *rx, + struct rxrpc_connection *conn, struct rxrpc_call *call) { switch (READ_ONCE(call->state)) { case RXRPC_CALL_SERVER_AWAIT_ACK: rxrpc_call_completed(call); - break; + /* Fall through */ case RXRPC_CALL_COMPLETE: break; default: @@ -1043,11 +1073,13 @@ static void rxrpc_input_implicit_end_call(struct rxrpc_connection *conn, set_bit(RXRPC_CALL_EV_ABORT, &call->events); rxrpc_queue_call(call); } + trace_rxrpc_improper_term(call); break; } - trace_rxrpc_improper_term(call); + spin_lock(&rx->incoming_lock); __rxrpc_disconnect_call(conn, call); + spin_unlock(&rx->incoming_lock); rxrpc_notify_socket(call); } @@ -1244,10 +1276,16 @@ int rxrpc_input_packet(struct sock *udp_sk, struct sk_buff *skb) goto wrong_security; if (sp->hdr.serviceId != conn->service_id) { - if (!test_bit(RXRPC_CONN_PROBING_FOR_UPGRADE, &conn->flags) || - conn->service_id != conn->params.service_id) + int old_id; + + if (!test_bit(RXRPC_CONN_PROBING_FOR_UPGRADE, &conn->flags)) + goto reupgrade; + old_id = cmpxchg(&conn->service_id, conn->params.service_id, + sp->hdr.serviceId); + + if (old_id != conn->params.service_id && + old_id != sp->hdr.serviceId) goto reupgrade; - conn->service_id = sp->hdr.serviceId; } if (sp->hdr.callNumber == 0) { @@ -1305,7 +1343,7 @@ int rxrpc_input_packet(struct sock *udp_sk, struct sk_buff *skb) if (rxrpc_to_client(sp)) goto reject_packet; if (call) - rxrpc_input_implicit_end_call(conn, call); + rxrpc_input_implicit_end_call(rx, conn, call); call = NULL; } @@ -1325,7 +1363,7 @@ int rxrpc_input_packet(struct sock *udp_sk, struct sk_buff *skb) goto bad_message; if (sp->hdr.seq != 1) goto discard; - call = rxrpc_new_incoming_call(local, rx, peer, conn, skb); + call = rxrpc_new_incoming_call(local, rx, skb); if (!call) goto reject_packet; rxrpc_send_ping(call, skb, skew); diff --git a/net/rxrpc/peer_event.c b/net/rxrpc/peer_event.c index f3e6fc670da2..05b51bdbdd41 100644 --- a/net/rxrpc/peer_event.c +++ b/net/rxrpc/peer_event.c @@ -301,6 +301,8 @@ void rxrpc_peer_add_rtt(struct rxrpc_call *call, enum rxrpc_rtt_rx_trace why, if (rtt < 0) return; + spin_lock(&peer->rtt_input_lock); + /* Replace the oldest datum in the RTT buffer */ sum -= peer->rtt_cache[cursor]; sum += rtt; @@ -312,6 +314,8 @@ void rxrpc_peer_add_rtt(struct rxrpc_call *call, enum rxrpc_rtt_rx_trace why, peer->rtt_usage = usage; } + spin_unlock(&peer->rtt_input_lock); + /* Now recalculate the average */ if (usage == RXRPC_RTT_CACHE_SIZE) { avg = sum / RXRPC_RTT_CACHE_SIZE; @@ -320,6 +324,7 @@ void rxrpc_peer_add_rtt(struct rxrpc_call *call, enum rxrpc_rtt_rx_trace why, do_div(avg, usage); } + /* Don't need to update this under lock */ peer->rtt = avg; trace_rxrpc_rtt_rx(call, why, send_serial, resp_serial, rtt, usage, avg); diff --git a/net/rxrpc/peer_object.c b/net/rxrpc/peer_object.c index 2d39eaf19620..5691b7d266ca 100644 --- a/net/rxrpc/peer_object.c +++ b/net/rxrpc/peer_object.c @@ -225,6 +225,7 @@ struct rxrpc_peer *rxrpc_alloc_peer(struct rxrpc_local *local, gfp_t gfp) peer->service_conns = RB_ROOT; seqlock_init(&peer->service_conn_lock); spin_lock_init(&peer->lock); + spin_lock_init(&peer->rtt_input_lock); peer->debug_id = atomic_inc_return(&rxrpc_debug_id); if (RXRPC_TX_SMSS > 2190) -- cgit v1.2.3-59-g8ed1b