On 03/09/2016 12:26 PM, Li Zhijian wrote: > > > On 03/09/2016 09:36 AM, Wen Congyang wrote: >> On 03/08/2016 05:54 PM, Peter Maydell wrote: >>> On 8 March 2016 at 16:06, Zhang Chen <zhangchen.f...@cn.fujitsu.com> >>> wrote: >>>> I found the reason for this problem is that >>>> unix_connect() have not connect to sock_path before iov_send(). >>>> It need time to establish connection. so can we fix it with usleep() >>>> like this: >>>> >>>> recv_sock = unix_connect(sock_path, NULL); >>>> g_assert_cmpint(recv_sock, !=, -1); >>>> + usleep(1000); >>>> >>>> ret = iov_send(send_sock[0], iov, 2, 0, sizeof(size) + >>>> sizeof(send_buf)); >>>> g_assert_cmpint(ret, ==, sizeof(send_buf) + sizeof(size)); >>>> close(send_sock[0]); >>>> >>>> ret = qemu_recv(recv_sock, &len, sizeof(len), 0); >>> >>> I would prefer it if we could find a way to fix this race >>> reliably rather than just inserting a delay and hoping it >>> is sufficient. Otherwise the test is likely to be unreliable >>> if run on a heavily loaded or slow machine. >> >> Yes, but there is no way to know when tcp_chr_accept() is called. Add >> a event >> to notify it? >> >> Thanks >> Wen Congyang >> > > Hi, Jason, PMM > As Congyang said that this is a bug of testcase instead of filter-mirror. > Maybe we should re-wrok the testcase, for example > - using -chardev pipe instead of -chardev socket, because we are > intend to test the packet mirror fuction instead of -chardev socket > > How about that ?
I still prefer to use socket since it will be more similar to its real usage (e.g COLO). Btw, I haven't investigated deeply, but at least qmp command trick should work. But what's better is to passing the pre-connected fd (e.g the socket created by socketpair()) to chardev, not sure we can do this.