Re: [LTP] [PATCH] read/read02.c: cleanup

[<ch...@su...>]

Hi!
> -#ifndef UCLINUX
> -	    /* Skip since uClinux does not implement memory protection */
> -	    /* the buffer is invalid - EFAULT */
> -	{
> -	&fd3, (void *)-1, EFAULT}
> -#endif
> +	{&badfd, buf, EBADF},
> +	{&fd2, buf, EISDIR},
> +	{&fd3, (void *)-1, EFAULT},
>  };

Removing the ifdefs breaks the testcases on uClinux, please keep them
there.

> -#if !defined(UCLINUX)
> -	bad_addr = mmap(0, 1, PROT_NONE,
> -			MAP_PRIVATE_EXCEPT_UCLINUX | MAP_ANONYMOUS, 0, 0);
> -	if (bad_addr == MAP_FAILED) {
> -		tst_brkm(TBROK, cleanup, "mmap failed");
> +
> +	TEST_ERROR_LOG(TEST_ERRNO);
> +
> +	if (TEST_ERRNO == test->exp_error) {
> +		tst_resm(TPASS | TTERRNO, "expected failure - errno = %d : %s",
> +			 TEST_ERRNO, strerror(TEST_ERRNO));
> +	} else {
> +		tst_resm(TFAIL | TTERRNO,
> +			 "unexpected error - %d : %s - expected %d",
> +			 TEST_ERRNO, strerror(TEST_ERRNO), test->exp_error);
>  	}
> -	TC[2].buf = bad_addr;
> -#endif

And removing the mmap() possibly breaks the testcases on platforms where
there is something mapped on address (void*)-1. Looking at the git log,
this may be case for some 64 bit machines (at least this is what the
patch from 2002 says).

Given that we have quite a lot of testcases that passes addresses that
are expected to generate EFAULT to various syscalls we should unify the
way these are generated.

I would go for creating tst_bad_addr() function that would replace all
the adhoc (and posibly wrong) definitions. I will look into that asap.

-- 
Cyril Hrubis
ch...@su...