!350 [sync] PR-347: sync waiting when primary process not start already

From: @openeuler-sync-bot Reviewed-by: @wu-changsheng Signed-off-by: @wu-changsheng
2023-04-21 03:33:19 +00:00 · 2023-04-21 03:33:19 +00:00 · 5c62cf29c8
commit 5c62cf29c8
parent 27eb7a3937 e89e1afa50
3 changed files with 209 additions and 1 deletions
--- a/0215-gazelle-send-recv-thread-bind-numa.patch
+++ b/0215-gazelle-send-recv-thread-bind-numa.patch
@ -0,0 +1,65 @@
+From b658a975bb76d3b39edb7fd73ab5815c302bce02 Mon Sep 17 00:00:00 2001
+From: kircher <majun65@huawei.com>
+Date: Tue, 21 Mar 2023 11:08:46 +0800
+Subject: [PATCH] gazelle send/recv thread bind numa
+
+---
+ src/lstack/core/lstack_lwip.c | 21 +++++++++++++++++++++
+ 1 file changed, 21 insertions(+)
+
+diff --git a/src/lstack/core/lstack_lwip.c b/src/lstack/core/lstack_lwip.c
+index 60abfe8..7355d7b 100644
+--- a/src/lstack/core/lstack_lwip.c
+++ b/src/lstack/core/lstack_lwip.c
+@@ -97,6 +97,7 @@ static void reset_sock_data(struct lwip_sock *sock)
+     sock->events = 0;
+     sock->call_num = 0;
+     sock->remain_len = 0;
+    sock->already_bind_numa = 0;
+ 
+     if (sock->recv_lastdata) {
+         pbuf_free(sock->recv_lastdata);
+@@ -882,6 +883,21 @@ ssize_t gazelle_same_node_ring_send(struct lwip_sock *sock, const void *buf, siz
+     return act_len;
+ }
+ 
+PER_THREAD uint16_t stack_sock_num[GAZELLE_MAX_STACK_NUM] = {0};
+PER_THREAD uint16_t max_sock_stack = 0;
+
+static void thread_bind_stack(struct lwip_sock *sock) {
+    if (likely(!sock->stack || sock->already_bind_numa)) {
+        return;
+    }
+    sock->already_bind_numa = 1;
+    stack_sock_num[sock->stack->stack_idx]++;
+    if (stack_sock_num[sock->stack->stack_idx] > max_sock_stack) {
+        max_sock_stack = stack_sock_num[sock->stack->stack_idx];
+	bind_to_stack_numa(sock->stack);
+    }
+}
+
+ ssize_t gazelle_send(int32_t fd, const void *buf, size_t len, int32_t flags)
+ {
+     if (buf == NULL) {
+@@ -893,6 +909,9 @@ ssize_t gazelle_send(int32_t fd, const void *buf, size_t len, int32_t flags)
+     }
+ 
+     struct lwip_sock *sock = get_socket_by_fd(fd);
+
+    thread_bind_stack(sock);
+
+     if (sock->same_node_tx_ring != NULL) {
+         return gazelle_same_node_ring_send(sock, buf, len, flags);
+     }
+@@ -973,6 +992,8 @@ ssize_t read_stack_data(int32_t fd, void *buf, size_t len, int32_t flags)
+         return 0;
+     }
+ 
+    thread_bind_stack(sock);
+
+     if (sock->same_node_rx_ring != NULL) {
+         return gazelle_same_node_ring_recv(sock, buf, len, flags);
+     }
+-- 
+2.33.0
+
--- a/0216-waiting-when-primary-process-not-start-already.patch
+++ b/0216-waiting-when-primary-process-not-start-already.patch
@ -0,0 +1,137 @@
+From f17b74371b14b97521462724069812fc9728b9ed Mon Sep 17 00:00:00 2001
+From: kircher <majun65@huawei.com>
+Date: Tue, 21 Mar 2023 15:36:52 +0800
+Subject: [PATCH] waiting when primary process not start already
+
+---
+ src/common/gazelle_opt.h                  |  1 +
+ src/lstack/api/lstack_signal.c            |  7 +++--
+ src/lstack/core/lstack_control_plane.c    |  6 ++++
+ src/lstack/core/lstack_init.c             | 34 +++++++++++++++++++++++
+ src/lstack/include/lstack_control_plane.h |  1 +
+ 5 files changed, 47 insertions(+), 2 deletions(-)
+
+diff --git a/src/common/gazelle_opt.h b/src/common/gazelle_opt.h
+index 4c0eef3..fe0483b 100644
+--- a/src/common/gazelle_opt.h
+++ b/src/common/gazelle_opt.h
+@@ -85,6 +85,7 @@
+ #define GAZELLE_SOCK_FILENAME_MAXLEN    128
+ 
+ #define GAZELLE_RUN_DIR                  "/var/run/gazelle/"
+#define GAZELLE_PRIMARY_START_PATH       "/var/run/gazelle/gazelle_primary"
+ #define GAZELLE_FILE_PERMISSION          0700
+ 
+ #define SEND_TIME_WAIT_NS 20000
+diff --git a/src/lstack/api/lstack_signal.c b/src/lstack/api/lstack_signal.c
+index e73bc61..03118f5 100644
+--- a/src/lstack/api/lstack_signal.c
+++ b/src/lstack/api/lstack_signal.c
+@@ -57,12 +57,15 @@ static inline bool match_hijack_signal(int sig)
+ static void lstack_sig_default_handler(int sig)
+ {
+     LSTACK_LOG(ERR, LSTACK, "lstack dumped，caught signal：%d\n", sig);
+-    dump_stack();
+-    lwip_exit();
+    if (get_global_cfg_params() && get_global_cfg_params()->is_primary) {
+        delete_primary_path();
+    }
+     if (!use_ltran()) {
+         dpdk_kni_release();
+     }
+     control_fd_close();
+    dump_stack();
+    lwip_exit();
+     (void)kill(getpid(), sig);
+ }
+ 
+diff --git a/src/lstack/core/lstack_control_plane.c b/src/lstack/core/lstack_control_plane.c
+index bc2c55d..e705cae 100644
+--- a/src/lstack/core/lstack_control_plane.c
+++ b/src/lstack/core/lstack_control_plane.c
+@@ -473,6 +473,12 @@ void control_fd_close(void)
+     }
+ }
+ 
+void delete_primary_path(void) {
+    if (!remove(GAZELLE_PRIMARY_START_PATH)) {
+        LSTACK_LOG(ERR, LSTACK, "delete %s failed\n", GAZELLE_PRIMARY_START_PATH);
+    }
+}
+
+ int32_t control_init_client(bool is_reconnect)
+ {
+     int32_t ret;
+diff --git a/src/lstack/core/lstack_init.c b/src/lstack/core/lstack_init.c
+index e8fa0dc..3537002 100644
+--- a/src/lstack/core/lstack_init.c
+++ b/src/lstack/core/lstack_init.c
+@@ -151,6 +151,32 @@ bool select_thread_path(void)
+     return true;
+ }
+ 
+static void check_process_start(void) {
+    if (get_global_cfg_params()->is_primary) {
+        return;
+    }
+
+    while (!fopen(GAZELLE_PRIMARY_START_PATH, "r")) {
+	printf("please make sure the primary process start already!\n");
+	sleep(1);
+    }
+}
+
+static int32_t set_process_start_flag(void) {
+    if (!get_global_cfg_params()->is_primary) {
+        return 0;
+    }
+
+    FILE *fp = NULL;
+    fp = fopen(GAZELLE_PRIMARY_START_PATH, "w");
+    if (fp == NULL) {
+        LSTACK_PRE_LOG(LSTACK_ERR, "set primary proceaa start flag failed!\n");
+	return -1;
+    }
+    (void)fclose(fp);
+    return 0;
+}
+
+ static int32_t check_process_conflict(void)
+ {
+     int32_t ret;
+@@ -340,6 +366,10 @@ __attribute__((constructor)) void gazelle_network_init(void)
+     }
+     LSTACK_PRE_LOG(LSTACK_INFO, "cfg_init success\n");
+ 
+    /*
+     * check primary process start */
+    check_process_start();
+
+     /*
+     * check conflict */
+     if (check_process_conflict() < 0) {
+@@ -402,6 +432,10 @@ __attribute__((constructor)) void gazelle_network_init(void)
+         set_kni_ip_mac();
+     }
+ 
+    if (set_process_start_flag() != 0) {
+        LSTACK_EXIT(1, "set_process_start_flag failed\n");
+    }
+
+     posix_api->ues_posix = 0;
+     LSTACK_LOG(INFO, LSTACK, "gazelle_network_init success\n");
+     rte_smp_mb();
+diff --git a/src/lstack/include/lstack_control_plane.h b/src/lstack/include/lstack_control_plane.h
+index d631ea6..aed5443 100644
+--- a/src/lstack/include/lstack_control_plane.h
+++ b/src/lstack/include/lstack_control_plane.h
+@@ -32,5 +32,6 @@ bool get_register_state(void);
+ void thread_register_phase1(struct rpc_msg *msg);
+ void thread_register_phase2(struct rpc_msg *msg);
+ void control_fd_close(void);
+void delete_primary_path(void);
+ 
+ #endif /* GAZELLE_CONTROL_PLANE_H */
+-- 
+2.33.0
+
--- a/gazelle.spec
+++ b/gazelle.spec
@ -2,7 +2,7 @@

 Name:          gazelle
 Version:       1.0.1
-Release:       53
+Release:       54
 Summary:       gazelle is a high performance user-mode stack
 License:       MulanPSL-2.0
 URL:           https://gitee.com/openeuler/gazelle
@ -229,6 +229,8 @@ Patch9211:     0211-refactor-mbuf-private-data.patch
 Patch9212:     0212-fix-kernel-scoket-select-path-error.patch
 Patch9213:     0213-discard-wakeup_num-parameter.patch
 Patch9214:     0214-fix-parse-args-error.patch
+Patch9215:     0215-gazelle-send-recv-thread-bind-numa.patch
+Patch9216:     0216-waiting-when-primary-process-not-start-already.patch

 %description
 %{name} is a high performance user-mode stack.
@ -269,6 +271,10 @@ install -Dpm 0640 %{_builddir}/%{name}-%{version}/src/ltran/ltran.conf       %{b
 %config(noreplace) %{conf_path}/ltran.conf

 %changelog
+* Thu Apr 20 UTC wu-changsheng <wuchangsheng2@huawei.com> - 1.0.1-54
+- waiting when primary process not start already
+- gazelle send/recv thread bind numa
+
 * Tue Mar 21 2023 jiangheng12 <jiangheng14@huawei.com> - 1.0.1-53
 - fix parse args error
 - discard wakeup_num parameter