[LTP] [RFC PATCH 0/1] brk: use direct syscalls

[Petr Vorel]

> Hello LTP maintainers,

> This patch slightly reworks the brk01 and brk02 tests to use direct
> syscalls with tst_syscall rather than calling through the libc.

> While running LTP on a musl-based distribution, we noticed that the brk
> tests were failing. It turns out that Musl prevents the use of brk
> with their wrapper: it always returns an error.
> This isn't too egregious as using brk is deprecated in favor of malloc
> and it isn't part of POSIX anymore since POSIX.1-2001, but it _is_
> different from glibc's beavior, which allows using it.

> This patch allows proper testing of brk's functionality, independent of
> libc specifics, and thus allows the tests to pass on Musl-based
> distributions like Alpine.

> Do you think this is a correct approach for LTP ?
> From what I could see there are a few tests that use tst_syscall directly
> and it doesn't affect the logic much for brk.

LGTM.

I wonder if it makes sense to add .test_variants [1] for glibc and uclibc,
e.g. for brk01():

void verify_brk(void)
{
	if (tst_variant) {
		tst_res(TINFO, "Testing sbrk()");
		cur_brk = (uintptr_t)sbrk(0);
	} else {
		tst_res(TINFO, "Testing __NR_brk");
		cur_brk = (void *)tst_syscall(__NR_brk, 0);
	}

}

struct tst_test test = {
	...
#ifdef __GLIBC__
	.test_variants = 2,
#else
	.test_variants = 1,
#endif
	...

Not sure if it should be testeed also on android, i.e:
#if defined(__GLIBC__) || defined(__ANDROID__)

Kind regards,
Petr

[1] https://github.com/linux-test-project/ltp/wiki/C-Test-API#130-testing-similar-syscalls-in-one-test

> Green build:
> https://github.com/Teo-CD/ltp/actions/runs/3563193507

> This was discovered in the context of the Morello project[0].
> [0]: https://www.morello-project.org/

> Teo Couprie Diaz (1):
>   syscalls/brk: use direct syscall

>  testcases/kernel/syscalls/brk/brk01.c | 15 ++++++---------
>  testcases/kernel/syscalls/brk/brk02.c | 14 ++++++--------
>  2 files changed, 12 insertions(+), 17 deletions(-)