WHC/CST-agent
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix: stop parliament accept starvation on outbound dial (0.9.22)

Browse Source 
Peer dial used per-step read timeouts so a stuck handshake could block the
single-threaded daemon for tens of seconds; inbound hellos piled up unread
(CLOSE-WAIT). Use a wall-clock dial budget, drain accepts before each dial,
skip self, and only one side initiates idle polls.

Co-authored-by: Cursor <cursoragent@cursor.com>
This commit is contained in:
WHC
2026-10-29 13:19:03 +08:00
co-authored by Cursor
parent 2f96b050d6
commit de2eaf87f7
5 changed files with 129 additions and 45 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/peer_auth.c
+72 -21
View File
@@ -13,6 +13,33 @@
#include <string.h>
#include <time.h>
#ifdef _WIN32
#include <windows.h>
#else
#include <sys/time.h>
#endif
/* Wall-clock deadline so a multi-step handshake cannot stretch to
* N_steps * timeout_ms (that starved the daemon accept loop). */
static int64_t peer_now_ms(void) {
#ifdef _WIN32
return (int64_t)GetTickCount64();
#else
struct timeval tv;
gettimeofday(&tv, NULL);
return (int64_t)tv.tv_sec * 1000 + (int64_t)tv.tv_usec / 1000;
#endif
}
static int peer_remain_ms(int64_t deadline_ms) {
int64_t left = deadline_ms - peer_now_ms();
if (left <= 0)
return 0;
if (left > 600000)
return 600000;
return (int)left;
}
static void hex_encode(const uint8_t *in, size_t n, char *out) {
static const char *H = "0123456789abcdef";
for (size_t i = 0; i < n; i++) {
@@ -379,48 +406,60 @@ static int recv_sealed(cst_peer_session_t *sess, uint8_t **out, size_t *out_n,
static int handshake_as_client(os_socket_t s, const cst_realm_cert_t *self,
const uint8_t self_sk[64],
const uint8_t realm_pk[32], int timeout_ms,
const uint8_t realm_pk[32], int64_t deadline_ms,
cst_realm_cert_t *remote) {
if (send_hello(s, self, timeout_ms) != 0)
int t;
if ((t = peer_remain_ms(deadline_ms)) <= 0 || send_hello(s, self, t) != 0)
return -1;
uint8_t nonce[32];
if (read_chal(s, timeout_ms, nonce) != 0)
if ((t = peer_remain_ms(deadline_ms)) <= 0 || read_chal(s, t, nonce) != 0)
return -1;
if (send_auth(s, self_sk, nonce, timeout_ms) != 0)
if ((t = peer_remain_ms(deadline_ms)) <= 0 ||
send_auth(s, self_sk, nonce, t) != 0)
return -1;
if (read_hello(s, realm_pk, timeout_ms, remote) != 0)
if ((t = peer_remain_ms(deadline_ms)) <= 0 ||
read_hello(s, realm_pk, t, remote) != 0)
return -1;
if (fill_random(nonce, 32) != 0)
return -1;
if (send_chal(s, nonce, timeout_ms) != 0)
if ((t = peer_remain_ms(deadline_ms)) <= 0 || send_chal(s, nonce, t) != 0)
return -1;
if (read_auth(s, remote, nonce, timeout_ms) != 0)
if ((t = peer_remain_ms(deadline_ms)) <= 0 ||
read_auth(s, remote, nonce, t) != 0)
return -1;
if (send_ok(s, timeout_ms) != 0 || read_ok(s, timeout_ms) != 0)
if ((t = peer_remain_ms(deadline_ms)) <= 0 || send_ok(s, t) != 0)
return -1;
if ((t = peer_remain_ms(deadline_ms)) <= 0 || read_ok(s, t) != 0)
return -1;
return 0;
}
static int handshake_as_server(os_socket_t s, const cst_realm_cert_t *self,
const uint8_t self_sk[64],
const uint8_t realm_pk[32], int timeout_ms,
const uint8_t realm_pk[32], int64_t deadline_ms,
cst_realm_cert_t *remote) {
if (read_hello(s, realm_pk, timeout_ms, remote) != 0)
int t;
if ((t = peer_remain_ms(deadline_ms)) <= 0 ||
read_hello(s, realm_pk, t, remote) != 0)
return -1;
uint8_t nonce[32];
if (fill_random(nonce, 32) != 0)
return -1;
if (send_chal(s, nonce, timeout_ms) != 0)
if ((t = peer_remain_ms(deadline_ms)) <= 0 || send_chal(s, nonce, t) != 0)
return -1;
if (read_auth(s, remote, nonce, timeout_ms) != 0)
if ((t = peer_remain_ms(deadline_ms)) <= 0 ||
read_auth(s, remote, nonce, t) != 0)
return -1;
if (send_hello(s, self, timeout_ms) != 0)
if ((t = peer_remain_ms(deadline_ms)) <= 0 || send_hello(s, self, t) != 0)
return -1;
if (read_chal(s, timeout_ms, nonce) != 0)
if ((t = peer_remain_ms(deadline_ms)) <= 0 || read_chal(s, t, nonce) != 0)
return -1;
if (send_auth(s, self_sk, nonce, timeout_ms) != 0)
if ((t = peer_remain_ms(deadline_ms)) <= 0 ||
send_auth(s, self_sk, nonce, t) != 0)
return -1;
if (read_ok(s, timeout_ms) != 0 || send_ok(s, timeout_ms) != 0)
if ((t = peer_remain_ms(deadline_ms)) <= 0 || read_ok(s, t) != 0)
return -1;
if ((t = peer_remain_ms(deadline_ms)) <= 0 || send_ok(s, t) != 0)
return -1;
return 0;
}
@@ -433,15 +472,19 @@ int cst_peer_dial(const char *host, int port, const cst_realm_cert_t *self,
err[0] = 0;
if (!host || !self || !self_sk || !realm_pk || !out)
return -1;
if (timeout_ms <= 0)
timeout_ms = 3000;
int64_t deadline = peer_now_ms() + timeout_ms;
os_net_init();
os_socket_t s = os_net_connect(host, port, timeout_ms);
int t = peer_remain_ms(deadline);
os_socket_t s = os_net_connect(host, port, t);
if (s == OS_SOCKET_BAD) {
if (err && err_sz)
snprintf(err, err_sz, "connect failed");
return -1;
}
cst_realm_cert_t remote;
if (handshake_as_client(s, self, self_sk, realm_pk, timeout_ms, &remote) !=
if (handshake_as_client(s, self, self_sk, realm_pk, deadline, &remote) !=
0) {
if (err && err_sz)
snprintf(err, err_sz, "peer handshake failed");
@@ -453,7 +496,8 @@ int cst_peer_dial(const char *host, int port, const cst_realm_cert_t *self,
snprintf(out->host, sizeof(out->host), "%s", host);
out->port = port;
out->sock = (intptr_t)s;
if (peer_kx(s, 1, out->aead_key, timeout_ms) != 0) {
t = peer_remain_ms(deadline);
if (t <= 0 || peer_kx(s, 1, out->aead_key, t) != 0) {
if (err && err_sz)
snprintf(err, err_sz, "peer key exchange failed");
cst_peer_close(out);
@@ -471,15 +515,21 @@ int cst_peer_accept(intptr_t listen_sock, const cst_realm_cert_t *self,
err[0] = 0;
if (!self || !self_sk || !realm_pk || !out)
return -1;
if (timeout_ms <= 0)
timeout_ms = 3000;
int timed = 0;
/* Accept wait uses the caller budget; handshake+kx share a fresh deadline
* after the connection is taken, so a long accept wait does not shrink the
* crypto exchange to zero. */
os_socket_t s = os_net_accept((os_socket_t)listen_sock, timeout_ms, &timed);
if (s == OS_SOCKET_BAD) {
if (err && err_sz)
snprintf(err, err_sz, timed ? "accept timeout" : "accept failed");
return -1;
}
int64_t deadline = peer_now_ms() + (timeout_ms < 3000 ? 3000 : timeout_ms);
cst_realm_cert_t remote;
if (handshake_as_server(s, self, self_sk, realm_pk, timeout_ms, &remote) !=
if (handshake_as_server(s, self, self_sk, realm_pk, deadline, &remote) !=
0) {
if (err && err_sz)
snprintf(err, err_sz, "peer handshake failed");
@@ -491,7 +541,8 @@ int cst_peer_accept(intptr_t listen_sock, const cst_realm_cert_t *self,
os_net_peer(s, out->host, sizeof(out->host));
out->port = 0;
out->sock = (intptr_t)s;
if (peer_kx(s, 0, out->aead_key, timeout_ms) != 0) {
int t = peer_remain_ms(deadline);
if (t <= 0 || peer_kx(s, 0, out->aead_key, t) != 0) {
if (err && err_sz)
snprintf(err, err_sz, "peer key exchange failed");
cst_peer_close(out);