[Devel] [PATCH vz10 v2] selftests/core: fix unshare_test with large fs.nr_open

[Konstantin Khorenko]

On 6/18/26 19:46, Eva Kurchatova wrote:
> The test assumes fs.nr_open is close to the default 1048576, but some
> systems set it much higher (e.g. 1073741816). In that case, dup2() to
> nr_open + 64 requires the kernel to allocate a file descriptor table
> with ~1 billion entries, which fails with ENOMEM.
> 
> Fix by setting fs.nr_open to a known reasonable base value (1048576)
> for the duration of the test, rather than using whatever the system
> happens to have.
> 
> Signed-off-by: Eva Kurchatova <eva.kurchatova at virtuozzo.com>
> 
> https://virtuozzo.atlassian.net/browse/VSTOR-132444
> Feature: fix selftests
> ---
>  tools/testing/selftests/core/unshare_test.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/tools/testing/selftests/core/unshare_test.c b/tools/testing/selftests/core/unshare_test.c
> index 7fec9dfb1b0e..f4a4aac6afc4 100644
> --- a/tools/testing/selftests/core/unshare_test.c
> +++ b/tools/testing/selftests/core/unshare_test.c
> @@ -40,6 +40,10 @@ TEST(unshare_EMFILE)
>  
>  	ASSERT_EQ(sscanf(buf, "%d", &nr_open), 1);
>  
> +	/* Cap nr_open to avoid ENOMEM from a huge fd table allocation */
> +	if (nr_open > 1024 * 1024)
> +		n = sprintf(buf, "%d\n", (nr_open = 1024 * 1024));
> +

If your tests modifies some configuration (and this test does set fs.nr_open), it mus resurrect the original value after the run.

>  	ASSERT_EQ(0, getrlimit(RLIMIT_NOFILE, &rlimit));
>  
>  	/* bump fs.nr_open */