diff options
author | Linus Torvalds <torvalds@linux-foundation.org> | 2024-06-27 10:05:35 -0700 |
---|---|---|
committer | Linus Torvalds <torvalds@linux-foundation.org> | 2024-06-27 10:05:35 -0700 |
commit | fd19d4a492af77b1e8fb0439781a3048d1d1f554 (patch) | |
tree | c293c1a1218fe87c4b6712938352dbc349d0b68e /net/ipv4/tcp_input.c | |
parent | 3c1d29e53d34537063e60f5eafe0482780a1735a (diff) | |
parent | b62cb6a7e83622783100182d9b70e9c70393cfbe (diff) | |
download | linux-fd19d4a492af77b1e8fb0439781a3048d1d1f554.tar.gz linux-fd19d4a492af77b1e8fb0439781a3048d1d1f554.tar.bz2 linux-fd19d4a492af77b1e8fb0439781a3048d1d1f554.zip |
Merge tag 'net-6.10-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Pull networking fixes from Paolo Abeni:
"Including fixes from can, bpf and netfilter.
There are a bunch of regressions addressed here, but hopefully nothing
spectacular. We are still waiting the driver fix from Intel, mentioned
by Jakub in the previous networking pull.
Current release - regressions:
- core: add softirq safety to netdev_rename_lock
- tcp: fix tcp_rcv_fastopen_synack() to enter TCP_CA_Loss for failed
TFO
- batman-adv: fix RCU race at module unload time
Previous releases - regressions:
- openvswitch: get related ct labels from its master if it is not
confirmed
- eth: bonding: fix incorrect software timestamping report
- eth: mlxsw: fix memory corruptions on spectrum-4 systems
- eth: ionic: use dev_consume_skb_any outside of napi
Previous releases - always broken:
- netfilter: fully validate NFT_DATA_VALUE on store to data registers
- unix: several fixes for OoB data
- tcp: fix race for duplicate reqsk on identical SYN
- bpf:
- fix may_goto with negative offset
- fix the corner case with may_goto and jump to the 1st insn
- fix overrunning reservations in ringbuf
- can:
- j1939: recover socket queue on CAN bus error during BAM
transmission
- mcp251xfd: fix infinite loop when xmit fails
- dsa: microchip: monitor potential faults in half-duplex mode
- eth: vxlan: pull inner IP header in vxlan_xmit_one()
- eth: ionic: fix kernel panic due to multi-buffer handling
Misc:
- selftest: unix tests refactor and a lot of new cases added"
* tag 'net-6.10-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (61 commits)
net: mana: Fix possible double free in error handling path
selftest: af_unix: Check SIOCATMARK after every send()/recv() in msg_oob.c.
af_unix: Fix wrong ioctl(SIOCATMARK) when consumed OOB skb is at the head.
selftest: af_unix: Check EPOLLPRI after every send()/recv() in msg_oob.c
selftest: af_unix: Check SIGURG after every send() in msg_oob.c
selftest: af_unix: Add SO_OOBINLINE test cases in msg_oob.c
af_unix: Don't stop recv() at consumed ex-OOB skb.
selftest: af_unix: Add non-TCP-compliant test cases in msg_oob.c.
af_unix: Don't stop recv(MSG_DONTWAIT) if consumed OOB skb is at the head.
af_unix: Stop recv(MSG_PEEK) at consumed OOB skb.
selftest: af_unix: Add msg_oob.c.
selftest: af_unix: Remove test_unix_oob.c.
tracing/net_sched: NULL pointer dereference in perf_trace_qdisc_reset()
netfilter: nf_tables: fully validate NFT_DATA_VALUE on store to data registers
net: usb: qmi_wwan: add Telit FN912 compositions
tcp: fix tcp_rcv_fastopen_synack() to enter TCP_CA_Loss for failed TFO
ionic: use dev_consume_skb_any outside of napi
net: dsa: microchip: fix wrong register write when masking interrupt
Fix race for duplicate reqsk on identical SYN
ibmvnic: Add tx check to prevent skb leak
...
Diffstat (limited to 'net/ipv4/tcp_input.c')
-rw-r--r-- | net/ipv4/tcp_input.c | 45 |
1 files changed, 33 insertions, 12 deletions
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index 01d208e0eef3..2e39cb881e20 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -2782,13 +2782,37 @@ static void tcp_mtup_probe_success(struct sock *sk) NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMTUPSUCCESS); } +/* Sometimes we deduce that packets have been dropped due to reasons other than + * congestion, like path MTU reductions or failed client TFO attempts. In these + * cases we call this function to retransmit as many packets as cwnd allows, + * without reducing cwnd. Given that retransmits will set retrans_stamp to a + * non-zero value (and may do so in a later calling context due to TSQ), we + * also enter CA_Loss so that we track when all retransmitted packets are ACKed + * and clear retrans_stamp when that happens (to ensure later recurring RTOs + * are using the correct retrans_stamp and don't declare ETIMEDOUT + * prematurely). + */ +static void tcp_non_congestion_loss_retransmit(struct sock *sk) +{ + const struct inet_connection_sock *icsk = inet_csk(sk); + struct tcp_sock *tp = tcp_sk(sk); + + if (icsk->icsk_ca_state != TCP_CA_Loss) { + tp->high_seq = tp->snd_nxt; + tp->snd_ssthresh = tcp_current_ssthresh(sk); + tp->prior_ssthresh = 0; + tp->undo_marker = 0; + tcp_set_ca_state(sk, TCP_CA_Loss); + } + tcp_xmit_retransmit_queue(sk); +} + /* Do a simple retransmit without using the backoff mechanisms in * tcp_timer. This is used for path mtu discovery. * The socket is already locked here. */ void tcp_simple_retransmit(struct sock *sk) { - const struct inet_connection_sock *icsk = inet_csk(sk); struct tcp_sock *tp = tcp_sk(sk); struct sk_buff *skb; int mss; @@ -2828,14 +2852,7 @@ void tcp_simple_retransmit(struct sock *sk) * in network, but units changed and effective * cwnd/ssthresh really reduced now. */ - if (icsk->icsk_ca_state != TCP_CA_Loss) { - tp->high_seq = tp->snd_nxt; - tp->snd_ssthresh = tcp_current_ssthresh(sk); - tp->prior_ssthresh = 0; - tp->undo_marker = 0; - tcp_set_ca_state(sk, TCP_CA_Loss); - } - tcp_xmit_retransmit_queue(sk); + tcp_non_congestion_loss_retransmit(sk); } EXPORT_SYMBOL(tcp_simple_retransmit); @@ -6295,8 +6312,7 @@ static bool tcp_rcv_fastopen_synack(struct sock *sk, struct sk_buff *synack, tp->fastopen_client_fail = TFO_DATA_NOT_ACKED; skb_rbtree_walk_from(data) tcp_mark_skb_lost(sk, data); - tcp_xmit_retransmit_queue(sk); - tp->retrans_stamp = 0; + tcp_non_congestion_loss_retransmit(sk); NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPFASTOPENACTIVEFAIL); return true; @@ -7257,7 +7273,12 @@ int tcp_conn_request(struct request_sock_ops *rsk_ops, tcp_rsk(req)->tfo_listener = false; if (!want_cookie) { req->timeout = tcp_timeout_init((struct sock *)req); - inet_csk_reqsk_queue_hash_add(sk, req, req->timeout); + if (unlikely(!inet_csk_reqsk_queue_hash_add(sk, req, + req->timeout))) { + reqsk_free(req); + return 0; + } + } af_ops->send_synack(sk, dst, &fl, req, &foc, !want_cookie ? TCP_SYNACK_NORMAL : |