This is an automated email from the ASF dual-hosted git repository. acassis pushed a commit to branch master in repository https://gitbox.apache.org/repos/asf/nuttx.git
commit da5a493f58e0866b754e7cbec3810ce9cd1bdd32 Author: Marco Casaroli <[email protected]> AuthorDate: Fri Nov 20 17:12:24 2026 +0100 net/tcp: Do not lose the wakeup of a send that waits on the send limit. With CONFIG_NET_SEND_BUFSIZE > 0, psock_tcp_send() compares the queued bytes with the limit while it holds the device and connection locks, and then waits on snd_sem. conn_dev_sem_timedwait() releases the locks before the wait. tcp_sendbuffer_notify() posted snd_sem only when a thread already waited on it (count < 0). When the ACKs that empty the write queue arrive between the release of the locks and the wait (for example on the work queue of a network driver that waited for the device lock), the post was lost. If no other ACK came, the sender waited until its SO_SNDTIMEO time-out, or for ever. Post also when the count is 0. A count that no sender waited for only makes a sender check the limit once more. Assisted-by: Claude Code:claude-opus-5-5 Signed-off-by: Marco Casaroli <[email protected]> --- net/tcp/tcp_send_buffered.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/net/tcp/tcp_send_buffered.c b/net/tcp/tcp_send_buffered.c index 2b88be23384..e5d1fa2d84b 100644 --- a/net/tcp/tcp_send_buffered.c +++ b/net/tcp/tcp_send_buffered.c @@ -1770,8 +1770,13 @@ void tcp_sendbuffer_notify(FAR struct tcp_conn_s *conn) { int val = 0; + /* Post also when no thread waits yet: a sender can be between its check + * of the limit and its wait, with the locks released. A count that no + * sender waited for only makes a sender check the limit once more. + */ + nxsem_get_value(&conn->snd_sem, &val); - if (val < 0) + if (val <= 0) { nxsem_post(&conn->snd_sem); }
