tcp: tcp_tx_timestamp() fix
Subsystems (4)
cover tcp: tcp_tx_timestamp() fix
Fix an old bug in tcp_tx_timestamp().
Add one corresponding packetdrill test.
Eric Dumazet (2):
tcp: tcp_tx_timestamp() must look at the rtx queue
selftest: packetdrill: add tcp_timestamping_tcp_tx_timestamp_bug.pkt
net/ipv4/tcp.c | 3 + .../tcp_timestamping_tcp_tx_timestamp_bug.pkt | 70 +++++++++++++++++++ 2 files changed, 73 insertions(+) create mode 100644 tools/testing/selftests/net/packetdrill/tcp_timestamping_tcp_tx_timestamp_bug.pkt
Signature · 1 line
2.53.0.rc1.217.geba53bf80e-goog
PATCH 1/2 tcp: tcp_tx_timestamp() must look at the rtx queue
tcp_tx_timestamp() is only called at the end of tcp_sendmsg_locked() before the final tcp_push().
By the time it is called, it is possible all the copied data has been sent already (transmit queue is empty).
If this is the case, use the last skb in the rtx queue.
--- net/ipv4/tcp.c | 3 +++ 1 file changed, 3 insertions(+)
net/ipv4/tcp.c +3 -0
@@ -501,6 +501,9 @@ static void tcp_tx_timestamp(struct sock *sk, struct sockcm_cookie *sockc)struct sk_buff *skb =tcp_write_queue_tail (sk);u32 tsflags = sockc->tsflags; +if (unlikely (!skb)) + skb =skb_rb_last (&sk->tcp_rtx_queue); +if (tsflags && skb) {struct skb_shared_info *shinfo =skb_shinfo (skb);struct tcp_skb_cb *tcb =TCP_SKB_CB (skb);
Signature · 1 line
2.53.0.rc1.217.geba53bf80e-goog
Re: [PATCH net-next 1/2] tcp: tcp_tx_timestamp() must look at the rtx queue
Quoted 11 lines · Eric Dumazet
tcp_tx_timestamp()is only called at the end oftcp_sendmsg_locked()before the finaltcp_push().By the time it is called, it is possible all the copied data has been sent already (transmit queue is empty).
If this is the case, use the last skb in the rtx queue.
Fixes 75c119afe14f ("tcp: implement rb-tree based retransmit queue")Signed-off-by Eric Dumazet <[email protected]>
Great, I failed to spot this case... It could be reproduced when 'wait_for_space' is triggered. I strongly feel that more packetdrill tests are needed to observe the behavior of socket timestamping.
+ Willem
Thanks,
Jason
PATCH 2/2 selftest: packetdrill: add tcp_timestamping_tcp_tx_timestamp_bug.pkt
In reply to Eric DumazetTest tcp_tx_timestamp() behavior after ("tcp: tcp_tx_timestamp() must look at the rtx queue").
Without the fix, this new test fails like this:
tcp_timestamping_tcp_tx_timestamp_bug.pkt:55: runtime error in recvmsg call: Expected result 0 but got -1 with errno 11 (Resource temporarily unavailable)
--- .../tcp_timestamping_tcp_tx_timestamp_bug.pkt | 70 +++++++++++++++++++ 1 file changed, 70 insertions(+) create mode 100644 tools/testing/selftests/net/packetdrill/tcp_timestamping_tcp_tx_timestamp_bug.pkt
tools/testing/selftests/net/packetdrill/tcp_timestamping_tcp_tx_timestamp_bug.pkt new file +70 -0
@@ -0,0 +1,70 @@ +// SPDX-License-Identifier: GPL-2.0 +// Test after "tcp: tcp_tx_timestamp() must look at the rtx queue" + +// This test is about receiving the SCM_TSTAMP_ACK, +// we do not care about its SCM_TIMESTAMPING precision. +--tolerance_usecs=1000000 + +`./defaults.sh +sysctl -q net.ipv4.tcp_min_tso_segs=70 +` + +// Create a socket and set it to non-blocking. + 0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3 + +0 fcntl(3, F_GETFL) = 0x2 (flags O_RDWR) + +0 fcntl(3, F_SETFL, O_RDWR|O_NONBLOCK) = 0 + +// Establish connection and verify that there was no error. + +0 connect(3, ..., ...) = -1 EINPROGRESS (Operation now in progress) + +0 > S 0:0(0) <mss 1460,sackOK,TS val 100 ecr 0,nop,wscale 8> ++.010 < S. 0:0(0) ack 1 win 65535 <mss 1000,sackOK,TS val 700 ecr 100,nop,wscale 7> + +0 > . 1:1(0) ack 1 <nop,nop,TS val 200 ecr 700> + +0 getsockopt(3, SOL_SOCKET, SO_ERROR, [0], [4]) = 0 + +0 setsockopt(3, SOL_SOCKET, SO_SNDBUF, [30000], 4) = 0 + + +0 write(3, ..., 9880) = 9880 + +0 > P. 1:9881(9880) ack 1 <nop,nop,TS val 200 ecr 700> ++.010 < . 1:1(0) ack 9881 win 10000 <nop,nop,TS val 701 ecr 200> + + +0 write(3, ..., 19760) = 19760 + +0 > P. 9881:29641(19760) ack 1 <nop,nop,TS val 201 ecr 701> ++.010 < . 1:1(0) ack 29641 win 10000 <nop,nop,TS val 702 ecr 201> + + +0 write(3, ..., 39520) = 39520 + +0 > P. 29641:69161(39520) ack 1 <nop,nop,TS val 202 ecr 702> ++.010 < . 1:1(0) ack 69161 win 10000 <nop,nop,TS val 703 ecr 202> + +// One more write to increase cwnd + +0 write(3, ..., 79040) = 79040 + +0 > P. 69161:108681(39520) ack 1 <nop,nop,TS val 203 ecr 703> + +0 > P. 108681:148201(39520) ack 1 <nop,nop,TS val 203 ecr 703> ++.010 < . 1:1(0) ack 148201 win 1000 <nop,nop,TS val 704 ecr 203> + + +0 setsockopt(3, SOL_SOCKET, SO_TIMESTAMPING, + [SOF_TIMESTAMPING_TX_ACK | SOF_TIMESTAMPING_SOFTWARE | + SOF_TIMESTAMPING_OPT_ID], 4) = 0 + +// We have one write filling one skb +// last byte can not be stored because of our small SO_SNDBUF + +0 write(3, ..., 65209) = 65208 + +0 > P. 148201:213409(65208) ack 1 <nop,nop,TS val 204 ecr 704> ++.010 < . 1:1(0) ack 213409 win 1000 <nop,nop,TS val 705 ecr 204> + +// SCM_TSTAMP_ACK should be received after the last ack at +// t=60ms. + +0 recvmsg(3, {msg_name(...)=..., + msg_iov(1)=[{...,0}], + msg_flags=MSG_ERRQUEUE|MSG_TRUNC, + msg_control=[ + {cmsg_level=SOL_SOCKET, + cmsg_type=SCM_TIMESTAMPING, + cmsg_data={scm_sec=0,scm_nsec=60000000}}, + {cmsg_level=CMSG_LEVEL_IP, + cmsg_type=CMSG_TYPE_RECVERR, + cmsg_data={ee_errno=ENOMSG, + ee_origin=SO_EE_ORIGIN_TIMESTAMPING, + ee_type=0, + ee_code=0, + ee_info=SCM_TSTAMP_ACK, + ee_data=65207}} + ]}, MSG_ERRQUEUE) = 0
Signature · 1 line
2.53.0.rc1.217.geba53bf80e-goog
Re: [PATCH net-next 2/2] selftest: packetdrill: add tcp_timestamping_tcp_tx_timestamp_bug.pkt
Quoted 9 lines · Eric Dumazet
Test
tcp_tx_timestamp()behavior after ("tcp:tcp_tx_timestamp()must look at the rtx queue").Without the fix, this new test fails like this:
tcp_timestamping_tcp_tx_timestamp_bug.pkt:55: runtime error in recvmsg call: Expected result 0 but got -1 with errno 11 (Resource temporarily unavailable)
Signed-off-by Eric Dumazet <[email protected]>
Thanks for the test!
+Willem
Thanks,
Jason
This series was applied to netdev/net-next.git (main) by Jakub Kicinski <[email protected]>:
Hello:
This series was applied to netdev/net-next.git (main) by Jakub Kicinski <[email protected]>:
Quoted 9 lines · On Tue, 27 Jan 2026 12:38:26 +0000 you wrote:
Fix an old bug in
tcp_tx_timestamp().Add one corresponding packetdrill test.
Eric Dumazet (2):
tcp:tcp_tx_timestamp()must look at the rtx queue
selftest: packetdrill: add tcp_timestamping_tcp_tx_timestamp_bug.pkt[...]
Here is the summary with links:
- [net-next,1/2] tcp: tcp_tx_timestamp() must look at the rtx queue
https://git.kernel.org/netdev/net-next/c/838eb9687691 - [net-next,2/2] selftest: packetdrill: add tcp_timestamping_tcp_tx_timestamp_bug.pkt https://git.kernel.org/netdev/net-next/c/6080d525aba8
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html