[LTP] [PATCH v4] unshare03.c: Add test coverage for dup_fd() failure handling in unshare_fd()

[Wei Gao]

On Tue, Mar 04, 2025 at 09:31:03AM +0100, Andrea Cervesato wrote:
> Hi,
> 
> thanks for editing the last version. A couple of comments below.
> 
> On 3/4/25 05:06, Wei Gao via ltp wrote:
> > +		nr_open + 1024);
> > +
> > +	SAFE_DUP2(2, nr_open + 64);
> > +
> > +	if (!SAFE_CLONE(&args)) {
> > +		SAFE_FILE_PRINTF(FS_NR_OPEN, "%d", nr_open);
> > +		TST_EXP_FAIL(unshare(CLONE_FILES), EMFILE);
> > +		TST_CHECKPOINT_WAKE(0);
> There's no need to use synchronization mechanisms because at the end of the
> test we call tst_reap_children() waiting for all the children to be
> completed.
There are some race condition happen and trigger failure sometimes if you do not
use this synchronization(Currently no idea why this happen). 
Rerun case 30 times can trigger 2 or 3 cases failed if not use synchronization in my env.
> > +		exit(0);
> > +	}
> > +
> > +	TST_CHECKPOINT_WAIT(0);
> > +}
> > +
> > +static void setup(void)
> > +{
> > +	clone3_supported_by_kernel();
> > +}
> > +
> > +static struct tst_test test = {
> > +	.forks_child = 1,
> > +	.needs_root = 1,
> > +	.test_all = run,
> > +	.setup = setup,
> > +	.needs_checkpoints = 1,
> > +	.save_restore = (const struct tst_path_val[]) {
> > +		{FS_NR_OPEN, NULL, TST_SR_TCONF},
> > +		{}
> > +	},
> > +};
> > +
> > +#else
> > +TST_TEST_TCONF("unshare syscall is undefined.");
> > +#endif
> 
> The rest looks good. If you want I can edit the checkpoint thing and merge
> this patch.
> 
> Kind regards,
> Andrea Cervesato
>