Do not block forever if the tracee died right after recieving the fd
before having a chance to report the syscall exit to us.
diff --git a/src/AutoRemoteSyscalls.cc b/src/AutoRemoteSyscalls.cc
index 776d517..5678cfd 100644
--- a/src/AutoRemoteSyscalls.cc
+++ b/src/AutoRemoteSyscalls.cc
@@ -544,10 +544,14 @@
return their_fd;
}
-static int recvmsg_socket(ScopedFd& sock) {
+static int recvmsg_socket(ScopedFd& sock, bool blocking=true) {
fd_message<NativeArch> msg;
struct msghdr *msgp = (struct msghdr*)&msg.msg;
- if (0 > recvmsg(sock, msgp, MSG_CMSG_CLOEXEC)) {
+ int flag = MSG_CMSG_CLOEXEC;
+ if (!blocking) {
+ flag |= MSG_DONTWAIT;
+ }
+ if (0 > recvmsg(sock, msgp, flag)) {
return -1;
}
@@ -604,8 +608,13 @@
child_recvmsg<Arch>(*this, task()->session().tracee_fd_number());
if (child_syscall_result == -ESRCH) {
/* The child did not receive the message. Read it out of the socket
- buffer so it doesn't get read by another child later! */
- int fd = recvmsg_socket(task()->session().tracee_socket_receiver_fd());
+ * buffer so it doesn't get read by another child later!
+ * Note that if the child was killed, we might get an error
+ * (i.e. process exit stop before syscall exit stop) even
+ * if the fd is already removed from the socket.
+ * Because of this, we use a non-blocking recvmsg to avoid blocking forever.
+ */
+ int fd = recvmsg_socket(task()->session().tracee_socket_receiver_fd(), false);
if (fd >= 0) {
ASSERT(t, fd != our_fd.get()) << "This should always return a fresh fd!";
close(fd);
