Hi Petr
I think this is just a simple test of some APIs, but some test cases are not standardized and cause errors like "Segmentation fault" during testing. I think it is necessary to fix these errors or delete these tests.
Kind regards,
Dong
-----?ʼ?ԭ??-----
??????: Petr Vorel [mailto:[email protected]]
????ʱ??: 2021??6??21?? 15:59
?ռ???: James Dong (??????) <[email protected]>
????: [email protected]; Alexey Kodanev <[email protected]>; Steve Dickson <[email protected]>; [email protected]; [email protected]
????: Re: [LTP] [PATCH] fix rpc_suite/rpc:add check returned value
Hi all,
[Cc libtirpc ML and Steve]
> "Segmentation fault (core dumped)" due to the failure of svcfd_create
> during the rpc test, so you need to check the return value of the
> "svcfd_create" function
I'm not sure what is the value of TI-RPC tests. IMHO really messy code does not in the end cover much of libtirpc functionality. That's why I'm thinking to propose deleting whole testcases/network/rpc/rpc-tirpc/. libtirpc is being used in nfs-utils, thus it'd deserve to have some testing, but IMHO this should be libtirpc.
I'm not planning to dive into the technology to understand it enough be able to written the tests from scratch and I'm not aware of anybody else planning it.
> Signed-off-by: dongshijiang <[email protected]>
> ---
> .../rpc/rpc_createdestroy_svc_destroy/rpc_svc_destroy.c | 5 +++++
> .../rpc_createdestroy_svc_destroy/rpc_svc_destroy_stress.c | 5 +++++
> .../rpc/rpc_regunreg_xprt_register/rpc_xprt_register.c | 5 +++++
> .../rpc/rpc_regunreg_xprt_unregister/rpc_xprt_unregister.c | 5 +++++
> 4 files changed, 20 insertions(+)
> diff --git
> a/testcases/network/rpc/rpc-tirpc/tests_pack/rpc_suite/rpc/rpc_created
> estroy_svc_destroy/rpc_svc_destroy.c
> b/testcases/network/rpc/rpc-tirpc/tests_pack/rpc_suite/rpc/rpc_created
> estroy_svc_destroy/rpc_svc_destroy.c
> index 60b96cec3..3557c0068 100644
> ---
> a/testcases/network/rpc/rpc-tirpc/tests_pack/rpc_suite/rpc/rpc_created
> estroy_svc_destroy/rpc_svc_destroy.c
> +++ b/testcases/network/rpc/rpc-tirpc/tests_pack/rpc_suite/rpc/rpc_cre
> +++ atedestroy_svc_destroy/rpc_svc_destroy.c
> @@ -46,6 +46,11 @@ int main(void)
> //First of all, create a server
> svcr = svcfd_create(fd, 0, 0);
> +
> + //check returned value
> + if ((SVCXPRT *) svcr == NULL) {
> + return test_status;
> + }
> //Then call destroy macro
> svc_destroy(svcr);
> diff --git
> a/testcases/network/rpc/rpc-tirpc/tests_pack/rpc_suite/rpc/rpc_created
> estroy_svc_destroy/rpc_svc_destroy_stress.c
> b/testcases/network/rpc/rpc-tirpc/tests_pack/rpc_suite/rpc/rpc_created
> estroy_svc_destroy/rpc_svc_destroy_stress.c
> index ecd145393..5a4331f4d 100644
> ---
> a/testcases/network/rpc/rpc-tirpc/tests_pack/rpc_suite/rpc/rpc_created
> estroy_svc_destroy/rpc_svc_destroy_stress.c
> +++ b/testcases/network/rpc/rpc-tirpc/tests_pack/rpc_suite/rpc/rpc_cre
> +++ atedestroy_svc_destroy/rpc_svc_destroy_stress.c
> @@ -55,6 +55,11 @@ int main(int argn, char *argc[])
> //First of all, create a server
> for (i = 0; i < nbCall; i++) {
> svcr = svcfd_create(fd, 0, 0);
> +
> + //check returned value
> + if ((SVCXPRT *) svcr == NULL)
> + continue;
> + svcr = NULL;
man svc_destroy(3) states that it deallocates private data structures, including xprt itself.
Kind regards,
Petr
> //Then call destroy macro
> svc_destroy(svcr);
> diff --git
> a/testcases/network/rpc/rpc-tirpc/tests_pack/rpc_suite/rpc/rpc_regunre
> g_xprt_register/rpc_xprt_register.c
> b/testcases/network/rpc/rpc-tirpc/tests_pack/rpc_suite/rpc/rpc_regunre
> g_xprt_register/rpc_xprt_register.c
> index da3b93022..de4df15f1 100644
> ---
> a/testcases/network/rpc/rpc-tirpc/tests_pack/rpc_suite/rpc/rpc_regunre
> g_xprt_register/rpc_xprt_register.c
> +++ b/testcases/network/rpc/rpc-tirpc/tests_pack/rpc_suite/rpc/rpc_reg
> +++ unreg_xprt_register/rpc_xprt_register.c
> @@ -48,6 +48,11 @@ int main(void)
> //create a server
> svcr = svcfd_create(fd, 1024, 1024);
> +
> + //check returned value
> + if ((SVCXPRT *) svcr == NULL) {
> + return test_status;
> + }
> //call routine
> xprt_register(svcr);
> diff --git
> a/testcases/network/rpc/rpc-tirpc/tests_pack/rpc_suite/rpc/rpc_regunre
> g_xprt_unregister/rpc_xprt_unregister.c
> b/testcases/network/rpc/rpc-tirpc/tests_pack/rpc_suite/rpc/rpc_regunre
> g_xprt_unregister/rpc_xprt_unregister.c
> index d0b7a20d4..fbaec25ad 100644
> ---
> a/testcases/network/rpc/rpc-tirpc/tests_pack/rpc_suite/rpc/rpc_regunre
> g_xprt_unregister/rpc_xprt_unregister.c
> +++ b/testcases/network/rpc/rpc-tirpc/tests_pack/rpc_suite/rpc/rpc_reg
> +++ unreg_xprt_unregister/rpc_xprt_unregister.c
> @@ -52,6 +52,11 @@ int main(int argn, char *argc[])
> //create a server
> svcr = svcfd_create(fd, 1024, 1024);
> +
> + //check returned value
> + if ((SVCXPRT *) svcr == NULL) {
> + return test_status;
> + }
> xprt_register(svcr);
> //call routine
Hi Dong,
> Hi Petr
> I think this is just a simple test of some APIs, but some test cases are not standardized and cause errors like "Segmentation fault" during testing. I think it is necessary to fix these errors or delete these tests.
Sure this fix can get in. I saw issues with some tests on openSUSE, but that's a
separate problem (I was not able to find the problem thus report it.
> Kind regards,
> Dong
> > +++ b/testcases/network/rpc/rpc-tirpc/tests_pack/rpc_suite/rpc/rpc_createdestroy_svc_destroy/rpc_svc_destroy.c
> > @@ -46,6 +46,11 @@ int main(void)
> > //First of all, create a server
> > svcr = svcfd_create(fd, 0, 0);
> > +
> > + //check returned value
> > + if ((SVCXPRT *) svcr == NULL) {
IMHO casting is not required, right? Just
if (svcr == NULL) {
Kind regards,
Petr