|
From: | Thomas Huth |
Subject: | Re: [PATCH v4] tests/qtest: netdev: test stream and dgram backends |
Date: | Thu, 5 Jan 2023 07:24:37 +0100 |
User-agent: | Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101 Thunderbird/91.13.0 |
On 04/01/2023 21.10, Laurent Vivier wrote:
On 1/4/23 19:37, Thomas Huth wrote:On 03/01/2023 12.00, Laurent Vivier wrote:Signed-off-by: Laurent Vivier <lvivier@redhat.com> Acked-by: Michael S. Tsirkin <mst@redhat.com> --- Notes: v4: - rework EXPECT_STATE() - use g_dir_make_tmp() v3: - Add "-M none" to avoid error: "No machine specified, and there is no default" v2: - Fix ipv6 free port allocation - Check for IPv4, IPv6, AF_UNIX - Use g_mkdtemp() rather than g_file_open_tmp() - Use socketpair() in test_stream_fd()v1: compared to v14 of "qapi: net: add unix socket type support to netdev backend":- use IP addresses 127.0.0.1 and ::1 rather than localhost tests/qtest/meson.build | 2 + tests/qtest/netdev-socket.c | 434 ++++++++++++++++++++++++++++++++++++ 2 files changed, 436 insertions(+) create mode 100644 tests/qtest/netdev-socket.cFYI, the test also does not work on FreeBSD. It hangs on both, FreeBSD 12 and 13:https://cirrus-ci.com/task/5024964768694272?logs=build#L6937 https://cirrus-ci.com/task/5379344567107584?logs=build#L6938Thanks. Are you sure it's this test?
Since it's hanging at the same spot in both jobs, I'm pretty sure, yes.
The "/netdev/stream/inet/ipv6" seems to be the last one of the series (if I compare with previous machines) and it is OK in the logs.I don't understand where it can hang as we have an internal 5 seconds timeout (in EXPECT_STATE()).And I don't understand why we have only /netdev/stream tests and no /netdev/dgram tests.Is it possible to have more details?
You can reproduce the issue locally on a KVM-enabled Linux box by running: make vm-build-freebsd J=4 TARGET_LIST=loongarch64-softmmu \ BUILD_TARGET=check-qtest EXTRA_CONFIGURE_OPTS=--enable-werror \ DEBUG=1 And if I'm running the test manually there, I get this error message:qemu-system-loongarch64: -netdev stream,id=st0,server=true,addr.type=unix,addr.path=/tmp/netdev-socket.8JXDY1/stream_unix_abstract,addr.abstract=on: Parameter 'addr.abstract' is unexpected
Looking at qapi/sockets.json, the "abstract" parameter is only available with CONFIG_LINUX, so this cannot work on FreeBSD, indeed. Maybe it should be switched to CONFIG_POSIX? Otherwise you need to check for Linux in your qtest, I think.
HTH, Thomas
[Prev in Thread] | Current Thread | [Next in Thread] |