[LTP] [PATCH v3] madvise11.c:Check loadable module before rmmod

[Wei Gao]

On Mon, Mar 13, 2023 at 01:37:12PM +0100, Cyril Hrubis wrote:
> Hi!
> > > That does not solve the problem completely though, if we have a kernel
> > > where the hwpoinson_inject is set to N in config the test will attempt
> > > to rmmod it and get different error.
> > 
> > 
> > I have tested on the kernel which set to N in config and the test will report:
> > tst_test.c:1180: TCONF: hwpoison_inject driver not available
> > 
> > I think it should caused by following configuration of test case:
> >         .needs_drivers = (const char *const []) {
> >                 HW_MODULE,
> >                 NULL
> >         },
> > 
> > So the scenario of "N in kernel config" already handled by LTP framework, i have
> > to say LTP frame work already do a lot of things which i have no idea... xD
> 
> Ah, missed that part as well.
> 
> Looking at lib/tst_kernel.c we can also easily add tst_buildin_driver()
> function into the LTP library with just:
> 
> diff --git a/lib/tst_kernel.c b/lib/tst_kernel.c
> index ecf4b917e..6000522b7 100644
> --- a/lib/tst_kernel.c
> +++ b/lib/tst_kernel.c
> @@ -153,6 +153,11 @@ static int tst_check_driver_(const char *driver)
>         return -1;
>  }
> 
> +int tst_buildin_driver(const char *driver)
> +{
> +       return !tst_search_driver(driver, "modules.buildin");
> +}
> +
>  int tst_check_driver(const char *driver)
>  {
>  #ifdef __ANDROID__
> 

Try use above implementation but i found another TWARN : (

localhost:/home/ltp/testcases/kernel/syscalls/madvise # ./madvise11
tst_test.c:1560: TINFO: Timeout per run is 0h 01m 00s
madvise11.c:396: TINFO: Spawning 5 threads, with a total of 800 memory pages
madvise11.c:165: TINFO: Thread [3] returned 0, succeeded.
madvise11.c:165: TINFO: Thread [1] returned 0, succeeded.
madvise11.c:165: TINFO: Thread [2] returned 0, succeeded.
madvise11.c:165: TINFO: Thread [4] returned 0, succeeded.
madvise11.c:165: TINFO: Thread [0] returned 0, succeeded.
madvise11.c:199: TPASS: soft-offline / mmap race still clean
madvise11.c:327: TWARN: open(/sys/kernel/debug/hwpoison/unpoison-pfn,1,0000) failed: ENOENT (2) !!!!


> 
> -- 
> Cyril Hrubis
> chrubis@suse.cz