Re: [PATCH v2] Add Landlock test for io_uring IORING_OP_OPENAT operation

"Fabio M. De Francesco" <[email protected]>
Newsgroups dev.linux.lists.outreachy
Organization intel
Message-ID <5043999.MHq7AAxBmi@fdefranc-mobl3>
On Wednesday, 27 March 2024 14:20:01 CET Dorine Tipo wrote:
> This patch expands Landlock test coverage to include io_uring operations.
> It introduces a test for IORING_OP_OPENAT with Landlock rules, verifying
> allowed and disallowed access. This mitigates potential security
> vulnerabilities by ensuring Landlock controls access through io_uring.
> 
> It also updates the Makefile to include -luring in the LDLIBS.
> This ensures the test code has access to the necessary liburing
> library for io_uring operations.
> 
> Signed-off-by: Dorine Tipo <[email protected]>
> ---

Hi Dorine,

I'm adding inline a few comments and questions for you. However, since my 
experience with io_uring (inexistent with Landlock) is limited, I won't 
comment on the use of its API.

> Changes since V1:
> V2: - Consolidated two dependent patches in the V1 series into one patch
>       as suggested by <[email protected]>

NIT: No need for an entire email address. Next time"Fabio" will suffice  :)

>     - Updated the subject line to be more descriptive.
> 
>  tools/testing/selftests/landlock/Makefile  |   4 +-
>  tools/testing/selftests/landlock/fs_test.c | 132 +++++++++++++++++++++
>  2 files changed, 134 insertions(+), 2 deletions(-)
> 
> diff --git a/tools/testing/selftests/landlock/Makefile
> b/tools/testing/selftests/landlock/Makefile index
> 348e2dbdb4e0..ab47d1dadb62 100644
> --- a/tools/testing/selftests/landlock/Makefile
> +++ b/tools/testing/selftests/landlock/Makefile
> @@ -13,11 +13,11 @@ TEST_GEN_PROGS := $(src_test:.c=)
>  TEST_GEN_PROGS_EXTENDED := true
> 
>  # Short targets:
> -$(TEST_GEN_PROGS): LDLIBS += -lcap
> +$(TEST_GEN_PROGS): LDLIBS += -lcap -luring
>  $(TEST_GEN_PROGS_EXTENDED): LDFLAGS += -static
> 
>  include ../lib.mk
> 
>  # Targets with $(OUTPUT)/ prefix:
> -$(TEST_GEN_PROGS): LDLIBS += -lcap
> +$(TEST_GEN_PROGS): LDLIBS += -lcap -luring
>  $(TEST_GEN_PROGS_EXTENDED): LDFLAGS += -static
> diff --git a/tools/testing/selftests/landlock/fs_test.c
> b/tools/testing/selftests/landlock/fs_test.c index
> 9a6036fbf289..9c8247995d42 100644
> --- a/tools/testing/selftests/landlock/fs_test.c
> +++ b/tools/testing/selftests/landlock/fs_test.c
> @@ -21,7 +21,10 @@
>  #include <sys/stat.h>
>  #include <sys/sysmacros.h>
>  #include <sys/vfs.h>
> +#include <sys/types.h>
>  #include <unistd.h>
> +#include <liburing.h>
> +#include <linux/io_uring.h>
> 
>  #include "common.h"
> 
> @@ -4874,4 +4877,133 @@ TEST_F_FORK(layout3_fs, release_inodes)
>  	ASSERT_EQ(EACCES, test_open(TMP_DIR, O_RDONLY));
>  }
> 
> +/* Test io_uring openat access control with landlock rules */
> +static int test_ioring_op_openat(struct __test_metadata *const
> _metadata, const __u64 access, const char **paths, const int paths_size)

"access", "paths", and "paths_size" are not used anywhere. Why are there 
three unused arguments in the test_ioring_op_openat() signature?

> +{
> +	struct io_uring ring;
> +	struct io_uring_sqe *sqe;
> +
> +	const char *allowed_paths[] = {
> +		file1_s1d1, file2_s1d1,
> +		file1_s1d2, file2_s1d2,
> +		file1_s1d3, file2_s1d3,
> +		file1_s2d1, file1_s2d2,
> +		file1_s2d3, file2_s2d3,
> +		file1_s3d1,
> +	};
> +
> +	const char *disallowed_paths[] = {
> +		/* dir_s3d2 is a mount point. */
> +		dir_s3d2,
> +		dir_s3d3,
> +	};
> +
> +	/* Test Allowed Access */
> +	const struct rule allowed_rule[] = {
> +		{
> +			.path = allowed_paths[0],
> +			.access = LANDLOCK_ACCESS_FS_READ_FILE |
> +				  LANDLOCK_ACCESS_FS_WRITE_FILE,
> +		},

I suspect that you are forgetting the other elements as well as the final 
NULL element. Don't we need to fill allowed_rule[] with the other pairs of 
allowed paths and their respective access?

> +	};
> +	int allowed_ruleset_fd = create_ruleset(_metadata,
> allowed_rule[0].access, allowed_rule);

Please check how create_ruleset() is implemented. And pay special attention 
at this snippet:

for (i = 0; rules[i].path; i++) {
		add_path_beneath(_metadata, ruleset_fd, rules[i].access,
				 rules[i].path); 

It loops on the elements of rules[], working on the fields of each element 
struct (i.e., "path" and  "access"), but "allowed_rule" is an array with a 
single element. 

Furthermore, I think that you are forgetting a NULL final struct that marks 
the last element of allowed_rule[]. Am I overlooking something?
 
> +	ASSERT_LE(0, allowed_ruleset_fd);
> +
> +	int ret = io_uring_queue_init(32, &ring, 0);
> +
> +	ASSERT_EQ(0, ret);
> +
> +	/* Test each allowed path */
> +	for (int i = 0; i < ARRAY_SIZE(allowed_paths); ++i) {
> +		sqe = io_uring_get_sqe(&ring);
> +		io_uring_prep_openat(sqe, AT_FDCWD, allowed_paths[i], 
O_RDONLY,
> +				     allowed_ruleset_fd);
> +		/* Verify successful SQE preparation */
> +		ASSERT_EQ(0, ret);
> +
> +		if (ret != 0)
> +			return ret;
> +
> +		ret = io_uring_submit(&ring);
> +		/* Verify 1 submission completed */
> +		ASSERT_EQ(1, ret);
> +	}
> +
> +	/* Test Disallowed Access */
> +	const struct rule disallowed_rule[] = {
> +		{
> +			.path = disallowed_paths[0],
> +			.access = 0,
> +		}

Again, what happened to the other disallowed path and to the final NULL 
element of disallowed_rule[]?

> +
> +	};
> +	int disallowed_ruleset_fd = create_ruleset(_metadata,
> disallowed_rule[0].access, disallowed_rule); +
> +	ASSERT_LE(0, disallowed_ruleset_fd);
> +
> +	/* Test each disallowed path */
> +	for (int i = 0; i < ARRAY_SIZE(disallowed_paths); ++i) {
> +		sqe = io_uring_get_sqe(&ring);
> +		io_uring_prep_openat(sqe, AT_FDCWD, disallowed_paths[i], 
O_RDONLY,
> disallowed_ruleset_fd); +		/* Verify successful SQE 
preparation */
> +		ASSERT_EQ(1, ret);
> +
> +		if (ret != 0)
> +			return ret;
> +
> +		ret = io_uring_submit(&ring);
> +		/* Verify 1 submission completed */
> +		ASSERT_EQ(0, ret);
> +	}
> +
> +	/*  Cleanup: close ruleset fds, etc. */
> +	close(allowed_ruleset_fd);
> +	close(disallowed_ruleset_fd);
> +
> +	return 0;
> +}
> +
> +/* clang-format off */
> +FIXTURE(openat_test) {
> +	struct __test_metadata *metadata;
> +	const char *allowed_paths[11];
> +	const char *disallowed_paths[2];
> +};
> +
> +/* clang-format on */
> +
> +FIXTURE_SETUP(openat_test)
> +{
> +	/* initialize metadata, allowed_paths, and disallowed_paths */
> +	self->metadata = _metadata;
> +	const char *temp_allowed_paths[] = {
> +		file1_s1d1, file2_s1d1, file1_s1d2, file2_s1d2,
> +		file1_s1d3, file2_s1d3, file1_s2d1, file1_s2d2,
> +		file1_s2d3, file2_s2d3, file1_s3d1};
> +

Can you please say why you are defining both temp_allowed_paths[] and 
allowed_paths[]? 

> +	memcpy(self->allowed_paths, temp_allowed_paths,
> sizeof(temp_allowed_paths)); +
> +	const char *temp_disallowed_paths[] = {dir_s3d2, dir_s3d3};

I have the same question for temp_disallowed_paths[] and 
disallowed_paths[]. 

I see that they are initialised with the same set of information. Why don't 
you define them only once and pass references where needed?

Thanks,

Fabio 

> +
> +	memcpy(self->disallowed_paths, temp_disallowed_paths,
> sizeof(temp_disallowed_paths)); +}
> +
> +FIXTURE_TEARDOWN(openat_test)
> +{
> +	/* Clean up test environment */
> +}
> +
> +TEST_F_FORK(openat_test, test_ioring_op_openat_allowed)
> +{
> +	test_ioring_op_openat(self->metadata, LANDLOCK_ACCESS_FS_READ_FILE 
|
> +			      LANDLOCK_ACCESS_FS_WRITE_FILE, self-
>allowed_paths,
> +			      ARRAY_SIZE(self->allowed_paths));
> +}
> +
> +TEST_F_FORK(openat_test, test_ioring_op_openat_disallowed)
> +{
> +	test_ioring_op_openat(self->metadata, 0, self->disallowed_paths,
> +			      ARRAY_SIZE(self->disallowed_paths));
> +}
> +
>  TEST_HARNESS_MAIN
> --
> 2.25.1
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.