eal: fix data race in multi-process support

If DPDK is built with thread sanitizer it reports a race
in setting of multiprocess file descriptor. The fix is to
use atomic operations when updating mp_fd.

Build:
$ meson -Db_sanitize=address build
$ ninja -C build

Simple example:
$ .build/app/dpdk-testpmd -l 1-3 --no-huge
EAL: Detected CPU lcores: 16
EAL: Detected NUMA nodes: 1
EAL: Static memory layout is selected, amount of reserved memory can be adjusted with -m or --socket-mem
EAL: Detected static linkage of DPDK
EAL: Multi-process socket /run/user/1000/dpdk/rte/mp_socket
EAL: Selected IOVA mode 'VA'
testpmd: No probed ethernet devices
testpmd: create a new mbuf pool <mb_pool_0>: n=163456, size=2176, socket=0
testpmd: preferred mempool ops selected: ring_mp_mc
EAL: Error - exiting with code: 1
  Cause: Creation of mbuf pool for socket 0 failed: Cannot allocate memory
==================
WARNING: ThreadSanitizer: data race (pid=87245)
  Write of size 4 at 0x558e04d8ff70 by main thread:
    #0 rte_mp_channel_cleanup <null> (dpdk-testpmd+0x1e7d30c)
    #1 rte_eal_cleanup <null> (dpdk-testpmd+0x1e85929)
    #2 rte_exit <null> (dpdk-testpmd+0x1e5bc0a)
    #3 mbuf_pool_create.cold <null> (dpdk-testpmd+0x274011)
    #4 main <null> (dpdk-testpmd+0x5cc15d)

  Previous read of size 4 at 0x558e04d8ff70 by thread T2:
    #0 mp_handle <null> (dpdk-testpmd+0x1e7c439)
    #1 ctrl_thread_init <null> (dpdk-testpmd+0x1e6ee1e)

  As if synchronized via sleep:
    #0 nanosleep libsanitizer/tsan/tsan_interceptors_posix.cpp:366
    #1 get_tsc_freq <null> (dpdk-testpmd+0x1e92ff9)
    #2 set_tsc_freq <null> (dpdk-testpmd+0x1e6f2fc)
    #3 rte_eal_timer_init <null> (dpdk-testpmd+0x1e931a4)
    #4 rte_eal_init.cold <null> (dpdk-testpmd+0x29e578)
    DPDK#5 main <null> (dpdk-testpmd+0x5cbc45)

  Location is global 'mp_fd' of size 4 at 0x558e04d8ff70 (dpdk-testpmd+0x000003122f70)

  Thread T2 'rte_mp_handle' (tid=87248, running) created by main thread at:
    #0 pthread_create libsanitizer/tsan/tsan_interceptors_posix.cpp:969
    #1 rte_ctrl_thread_create <null> (dpdk-testpmd+0x1e6efd0)
    #2 rte_mp_channel_init.cold <null> (dpdk-testpmd+0x29cb7c)
    #3 rte_eal_init <null> (dpdk-testpmd+0x1e8662e)
    #4 main <null> (dpdk-testpmd+0x5cbc45)

SUMMARY: ThreadSanitizer: data race (app/dpdk-testpmd+0x1e7d30c) in rte_mp_channel_cleanup
==================
ThreadSanitizer: reported 1 warnings

Fixes: bacaa27 ("eal: add channel for multi-process communication")
Cc: [email protected]

Signed-off-by: Stephen Hemminger <[email protected]>
Acked-by: Anatoly Burakov <[email protected]>
Reviewed-by: Chengwen Feng <[email protected]>

Author: shemminger

lib/eal/common/eal_common_proc.c

@@ -260,7 +260,7 @@ rte_mp_action_unregister(const char *name)
 }
 
 static int
-read_msg(struct mp_msg_internal *m, struct sockaddr_un *s)
+read_msg(int fd, struct mp_msg_internal *m, struct sockaddr_un *s)
 {
 	int msglen;
 	struct iovec iov;
@@ -281,7 +281,7 @@ read_msg(struct mp_msg_internal *m, struct sockaddr_un *s)
 	msgh.msg_controllen = sizeof(control);
 
 retry:
-	msglen = recvmsg(mp_fd, &msgh, 0);
+	msglen = recvmsg(fd, &msgh, 0);
 
 	/* zero length message means socket was closed */
 	if (msglen == 0)
@@ -390,11 +390,12 @@ mp_handle(void *arg __rte_unused)
 {
 	struct mp_msg_internal msg;
 	struct sockaddr_un sa;
+	int fd;
 
-	while (mp_fd >= 0) {
+	while ((fd = __atomic_load_n(&mp_fd, __ATOMIC_RELAXED)) >= 0) {
 		int ret;
 
-		ret = read_msg(&msg, &sa);
+		ret = read_msg(fd, &msg, &sa);
 		if (ret <= 0)
 			break;
 
@@ -638,9 +639,8 @@ rte_mp_channel_init(void)
 			NULL, mp_handle, NULL) < 0) {
 		RTE_LOG(ERR, EAL, "failed to create mp thread: %s\n",
 			strerror(errno));
-		close(mp_fd);
 		close(dir_fd);
-		mp_fd = -1;
+		close(__atomic_exchange_n(&mp_fd, -1, __ATOMIC_RELAXED));
 		return -1;
 	}
 
@@ -656,11 +656,10 @@ rte_mp_channel_cleanup(void)
 {
 	int fd;
 
-	if (mp_fd < 0)
+	fd = __atomic_exchange_n(&mp_fd, -1, __ATOMIC_RELAXED);
+	if (fd < 0)
 		return;
 
-	fd = mp_fd;
-	mp_fd = -1;
 	pthread_cancel(mp_handle_tid);
 	pthread_join(mp_handle_tid, NULL);
 	close_socket_fd(fd);