Re: [LTP] [PATCH v3 09/11] Add statmount04 test

[Cyril Hrubis <chrubis@suse.cz>]

Hi!
> This test verifies that statmount() is correctly reading propagation
> from what mount in current namespace using STATMOUNT_PROPAGATE_FROM.
> ---
>  runtest/syscalls                                  |   1 +
>  testcases/kernel/syscalls/statmount/.gitignore    |   1 +
>  testcases/kernel/syscalls/statmount/statmount04.c | 131 ++++++++++++++++++++++
>  3 files changed, 133 insertions(+)
> 
> diff --git a/runtest/syscalls b/runtest/syscalls
> index 6eba317de..81c2f9740 100644
> --- a/runtest/syscalls
> +++ b/runtest/syscalls
> @@ -1541,6 +1541,7 @@ stat04_64 symlink01 -T stat04_64
>  statmount01 statmount01
>  statmount02 statmount02
>  statmount03 statmount03
> +statmount04 statmount04
>  
>  statfs01 statfs01
>  statfs01_64 statfs01_64
> diff --git a/testcases/kernel/syscalls/statmount/.gitignore b/testcases/kernel/syscalls/statmount/.gitignore
> index 2a02bf721..e720050b5 100644
> --- a/testcases/kernel/syscalls/statmount/.gitignore
> +++ b/testcases/kernel/syscalls/statmount/.gitignore
> @@ -1,3 +1,4 @@
>  statmount01
>  statmount02
>  statmount03
> +statmount04
> diff --git a/testcases/kernel/syscalls/statmount/statmount04.c b/testcases/kernel/syscalls/statmount/statmount04.c
> new file mode 100644
> index 000000000..6ca7ab4d0
> --- /dev/null
> +++ b/testcases/kernel/syscalls/statmount/statmount04.c
> @@ -0,0 +1,131 @@
> +// SPDX-License-Identifier: GPL-2.0-or-later
> +/*
> + * Copyright (C) 2024 SUSE LLC Andrea Cervesato <andrea.cervesato@suse.com>
> + */
> +
> +/*\
> + * [Description]
> + *
> + * This test verifies that statmount() is correctly reading propagation from
> + * what mount in current namespace using STATMOUNT_PROPAGATE_FROM.
> + *
> + * [Algorithm]
> + *
> + * * create a mount point
> + * * propagate a mounted folder inside the mount point
> + * * run statmount() on the mount point using STATMOUNT_PROPAGATE_FROM
> + * * read results and check propagated_from parameter contains the propagated
> + *   folder ID

Again dashes please.

> + */
> +
> +#include "statmount.h"
> +#include "lapi/stat.h"
> +#include "lapi/sched.h"
> +#include "tst_safe_stdio.h"
> +
> +#define MNTPOINT "mntpoint"
> +#define DIR_A MNTPOINT "/LTP_DIR_A"
> +#define DIR_C_SUBFOLDER "/LTP_DIR_A/propagated"
> +#define DIR_C (MNTPOINT DIR_C_SUBFOLDER)
> +#define DIR_B MNTPOINT "/LTP_DIR_B"
> +#define DIR_D MNTPOINT "/LTP_DIR_B/propagated"
> +
> +static uint64_t peer_group_id;
> +static uint64_t dird_id;
> +static struct statmount *st_mount;
> +
> +static int read_peer_group(void)
> +{
> +	FILE *file;
> +	char line[PATH_MAX];
> +	char mroot[PATH_MAX];
> +	int group = -1;
> +
> +	file = SAFE_FOPEN("/proc/self/mountinfo", "r");
> +
> +	while (fgets(line, sizeof(line), file)) {
> +		if (sscanf(line, "%*d %*d %*d:%*d %s %*s %*s shared:%d", mroot, &group) != 2)
> +			continue;
> +
> +		if (strcmp(mroot, DIR_C_SUBFOLDER) == 0)
> +			break;
> +	}
> +
> +	if (group == -1)
> +		tst_brk(TBROK, "Can't reed peer group ID for %s", DIR_C_SUBFOLDER);
                                        ^
					read

Also I'm wondering if we simply cannot call statmount() to get basic
info for DIR_C_SUBFOLDER and get the group id from there instead of
parsing proc.

> +	return group;
> +}
> +
> +static void run(void)
> +{
> +	memset(st_mount, 0, sizeof(struct statmount));
> +
> +	TST_EXP_PASS(statmount(
> +		dird_id,
> +		STATMOUNT_PROPAGATE_FROM,
> +		st_mount,
> +		sizeof(struct statmount),
> +		0));

Here as well, strange formatting.

> +	if (TST_RET == -1)
> +		return;
> +
> +	TST_EXP_EQ_LI(st_mount->mask, STATMOUNT_PROPAGATE_FROM);
> +	TST_EXP_EQ_LI(st_mount->size, sizeof(struct statmount));
> +	TST_EXP_EQ_LI(st_mount->propagate_from, peer_group_id);
> +}
> +
> +static void setup(void)
> +{
> +	struct statx sx;
> +
> +	/* create DIR_A / DIR_C structure with DIR_C mounted */
> +	SAFE_MKDIR(DIR_A, 0700);
> +	SAFE_MOUNT(DIR_A, DIR_A, "none", MS_BIND, NULL);
> +	SAFE_MOUNT("none", DIR_A, "none", MS_SHARED, NULL);
> +
> +	SAFE_MKDIR(DIR_C, 0700);
> +	SAFE_MOUNT(DIR_C, DIR_C, "none", MS_BIND, NULL);
> +	SAFE_MOUNT("none", DIR_C, "none", MS_SHARED, NULL);
> +
> +	/* DIR_A mounts into DIR_B. DIR_D is propagated */
> +	SAFE_MKDIR(DIR_B, 0700);
> +	SAFE_MOUNT(DIR_A, DIR_B, "none", MS_BIND, NULL);
> +	SAFE_MOUNT("none", DIR_B, "none", MS_SLAVE, NULL);
> +
> +	SAFE_STATX(AT_FDCWD, DIR_D, 0, STATX_MNT_ID_UNIQUE, &sx);
> +	dird_id = sx.stx_mnt_id;
> +
> +	peer_group_id = read_peer_group();
> +}
> +
> +static void cleanup(void)
> +{
> +	if (tst_is_mounted(DIR_C))
> +		SAFE_UMOUNT(DIR_C);
> +
> +	if (tst_is_mounted(DIR_B))
> +		SAFE_UMOUNT(DIR_B);
> +
> +	if (tst_is_mounted(DIR_A))
> +		SAFE_UMOUNT(DIR_A);

Again I wouldn't bet that umounting once is enough here. If my mental
image is right we have stacked several mounts on each directory.

I wonder how many cycled would this code do?

	while (tst_is_mounted(DIR_A))
		SAFE_UMOUNT(DIR_A);

> +}
> +
> +static struct tst_test test = {
> +	.test_all = run,
> +	.setup = setup,
> +	.cleanup = cleanup,
> +	.min_kver = "6.8",
> +	.mount_device = 1,
> +	.mntpoint = MNTPOINT,
> +	.all_filesystems = 1,
> +	.skip_filesystems = (const char *const []) {
> +		"fuse",
> +		NULL
> +	},
> +	.bufs = (struct tst_buffers []) {
> +		{&st_mount, .size = sizeof(struct statmount)},
> +		{}
> +	}
> +};
> 
> -- 
> 2.35.3
> 
> 
> -- 
> Mailing list info: https://lists.linux.it/listinfo/ltp

-- 
Cyril Hrubis
chrubis@suse.cz

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp