This series improves error handling in the passive TFO test by (1) fixing a broken behavior when the child processes failed (or timed out), and (2) adding more error handlng code in the test program.
The first patch fixes the behavior that the test didn't report failure even if the server or the client process exited with non-zero status. The second patch adds error handling code in the test program to improve reliability of the test.
This series was split out from the following series to address the feedback from Andrew Lunn: https://lore.kernel.org/netdev/cover.1767032397.git.yk@y-koj.net/
ChangeLog ========= v2 (this version): - Fix a typo in the patch description - Rephrase the patch description in imperative mood - Add error handling for fprintf() reflecting Markus Elfring's feedback v1: https://lore.kernel.org/netdev/cover.1768207347.git.yk@y-koj.net/
Yohei Kojima (2): selftests: net: fix passive TFO test to fail if child processes failed selftests: net: improve error handling in passive TFO test
tools/testing/selftests/net/tfo.c | 13 +++++++++---- tools/testing/selftests/net/tfo_passive.sh | 13 ++++++++++++- 2 files changed, 21 insertions(+), 5 deletions(-)
Improve the passive TFO test to report failure if the server or the client timed out or exited with non-zero status.
Before this commit, TFO test didn't fail even if exit(EXIT_FAILURE) is added to the first line of the run_server() and run_client() functions.
Signed-off-by: Yohei Kojima yk@y-koj.net --- tools/testing/selftests/net/tfo_passive.sh | 13 ++++++++++++- 1 file changed, 12 insertions(+), 1 deletion(-)
diff --git a/tools/testing/selftests/net/tfo_passive.sh b/tools/testing/selftests/net/tfo_passive.sh index a4550511830a..f116f888b794 100755 --- a/tools/testing/selftests/net/tfo_passive.sh +++ b/tools/testing/selftests/net/tfo_passive.sh @@ -85,12 +85,15 @@ timeout -k 1s 30s ip netns exec nssv ./tfo \ -s \ -p ${SERVER_PORT} \ -o ${out_file}& +server_pid="$!"
wait_local_port_listen nssv ${SERVER_PORT} tcp
ip netns exec nscl ./tfo -c -h ${SERVER_IP} -p ${SERVER_PORT} +client_exit_status="$?"
-wait +wait "$server_pid" +server_exit_status="$?"
res=$(cat $out_file) rm $out_file @@ -101,6 +104,14 @@ if [ "$res" = "0" ]; then exit 1 fi
+if [ "$client_exit_status" -ne 0 ] || [ "$server_exit_status" -ne 0 ]; then + # Note: timeout(1) exits with 124 if it timed out + echo "client exited with ${client_exit_status}" + echo "server exited with ${server_exit_status}" + cleanup_ns + exit 1 +fi + echo "$NSIM_SV_FD:$NSIM_SV_IFIDX" > $NSIM_DEV_SYS_UNLINK
echo $NSIM_CL_ID > $NSIM_DEV_SYS_DEL
Improve the error handling in passive TFO test to check the return value from sendto(), and to fail if read() or fprintf() failed.
Signed-off-by: Yohei Kojima yk@y-koj.net --- tools/testing/selftests/net/tfo.c | 13 +++++++++---- 1 file changed, 9 insertions(+), 4 deletions(-)
diff --git a/tools/testing/selftests/net/tfo.c b/tools/testing/selftests/net/tfo.c index 8d82140f0f76..3b1ee2d3d417 100644 --- a/tools/testing/selftests/net/tfo.c +++ b/tools/testing/selftests/net/tfo.c @@ -82,8 +82,10 @@ static void run_server(void) error(1, errno, "getsockopt(SO_INCOMING_NAPI_ID)");
if (read(connfd, buf, 64) < 0) - perror("read()"); - fprintf(outfile, "%d\n", opt); + error(1, errno, "read()"); + + if (fprintf(outfile, "%d\n", opt) < 0) + error(1, errno, "fprintf()");
fclose(outfile); close(connfd); @@ -92,14 +94,17 @@ static void run_server(void)
static void run_client(void) { - int fd; + int fd, ret; char *msg = "Hello, world!";
fd = socket(AF_INET6, SOCK_STREAM, 0); if (fd == -1) error(1, errno, "socket()");
- sendto(fd, msg, strlen(msg), MSG_FASTOPEN, (struct sockaddr *)&cfg_addr, sizeof(cfg_addr)); + ret = sendto(fd, msg, strlen(msg), MSG_FASTOPEN, + (struct sockaddr *)&cfg_addr, sizeof(cfg_addr)); + if (ret < 0) + error(1, errno, "sendto()");
close(fd); }
…
+++ b/tools/testing/selftests/net/tfo.c @@ -82,8 +82,10 @@ static void run_server(void)
…
if (read(connfd, buf, 64) < 0)
perror("read()");- fprintf(outfile, "%d\n", opt);
error(1, errno, "read()");- if (fprintf(outfile, "%d\n", opt) < 0)
error(1, errno, "fprintf()");fclose(outfile); close(connfd);
…
Why was error detection omitted for close() calls here so far?
https://pubs.opengroup.org/onlinepubs/9799919799/functions/fclose.html
Regards, Markus
On Tue, Jan 13, 2026 at 03:48:11PM +0100, Markus Elfring wrote:
…
+++ b/tools/testing/selftests/net/tfo.c @@ -82,8 +82,10 @@ static void run_server(void)
…
if (read(connfd, buf, 64) < 0)
perror("read()");- fprintf(outfile, "%d\n", opt);
error(1, errno, "read()");- if (fprintf(outfile, "%d\n", opt) < 0)
error(1, errno, "fprintf()");fclose(outfile); close(connfd);
…
Why was error detection omitted for close() calls here so far?
Because I believe that checking the return value of fclose() would not provide additional value in this test case, which is focused on testing the behavior of passive TFO.
I understand that fclose() could fail there, but considering the trade-off between test reliability and code complexity (which increases review and maintenance costs), I think checking the return value there does not provide benefits to justify the added complexity. In fact, as far as I can see, none of the existing tests in selftests/net check the fclose() return value.
Thank you, Yohei
https://pubs.opengroup.org/onlinepubs/9799919799/functions/fclose.html
Regards, Markus
linux-kselftest-mirror@lists.linaro.org
-
Markus Elfring -
Yohei Kojima