Skip to contentSkip to conversation

fix fanotify anonymous pipe testcases

PATCH In review Mete Durlu 4 participants 6 messages Mar 12, 2024 – Mar 14, 2024

lore
Share & tools
lore

Mete Durlu

author
#0 lore raw

v2 cover fix fanotify anonymous pipe testcases

* v1 -> v2:
- accept both EINVAL and EACCES when SElinux is configured. Checking SElinux policies would be too much effort.
- replace TST_EXP_FD_OR_FAIL with TST_EXP_FAIL since fanotify_mark() returns 0 on success and -1 with errno on failure.

* v0 -> v1:
- sent wrong patch

Mete Durlu (1):
fanotify14: fix anonymous pipe testcases

 .../kernel/syscalls/fanotify/fanotify14.c     | 32 +++++++++++++++++--
 1 file changed, 29 insertions(+), 3 deletions(-)
Signature · 5 lines
2.44.0


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
author
#1 lore raw

PATCH v2 1/1 fanotify14: fix anonymous pipe testcases

When SElinux is in enforcing state and SEpolicies disallow anonymous pipe usage with fanotify_mark(), related fanotify14 testcases fail with EACCES instead of EINVAL. Accept both errnos when SElinux is in enforcing state to correctly evaluate test results.

Replace TST_EXP_FD_OR_FAIL with TST_EXP_FAIL when testing fanotify_mark() as it returns -1 on failure and 0 on success not a file descriptor.

---
 .../kernel/syscalls/fanotify/fanotify14.c     | 32 +++++++++++++++++--
 1 file changed, 29 insertions(+), 3 deletions(-)
testcases/kernel/syscalls/fanotify/fanotify14.c +29 -3
index d02d81495..52c327dff 100644
@@ -27,12 +27,14 @@
 #define _GNU_SOURCE
 #include "tst_test.h"
 #include <errno.h>
+#include <stdlib.h>
 
 #ifdef HAVE_SYS_FANOTIFY_H
 #include "fanotify.h"
 
 #define MNTPOINT "mntpoint"
 #define FILE1 MNTPOINT"/file1"
+#define SELINUX_STATUS_PATH "/sys/fs/selinux/enforce"
 
 /*
  * List of inode events that are only available when notification group is
@@ -240,6 +242,19 @@ static struct test_case_t {
 	},
 };
 
+static int is_selinux_enforcing(void)
+{
+	char res;
+	int fd;
+
+	fd = open(SELINUX_STATUS_PATH, O_RDONLY);
+	if (fd <= 0)
+		return 0;
+	SAFE_READ(1, fd, &res, 1);
+	SAFE_CLOSE(fd);
+	return atoi(&res);
+}
+
 static void do_test(unsigned int number)
 {
 	struct test_case_t *tc = &test_cases[number];
@@ -275,17 +290,28 @@ static void do_test(unsigned int number)
 	/* Set mark on non-dir only when expecting error ENOTDIR */
 	const char *path = tc->expected_errno == ENOTDIR ? FILE1 : MNTPOINT;
 	int dirfd = AT_FDCWD;
+	int se_enforcing = 0;
 
 	if (tc->pfd) {
 		dirfd = tc->pfd[0];
 		path = NULL;
+		se_enforcing = is_selinux_enforcing();
 	}
 
 	tst_res(TINFO, "Testing %s with %s",
 		tc->mark.desc, tc->mask.desc);
-	TST_EXP_FD_OR_FAIL(fanotify_mark(fanotify_fd, FAN_MARK_ADD | tc->mark.flags,
-					 tc->mask.flags, dirfd, path),
-					 tc->expected_errno);
+
+	if (tc->pfd && se_enforcing) {
+		const int exp_errs[] = {tc->expected_errno, EACCES};
+
+		TST_EXP_FAIL_ARR(fanotify_mark(fanotify_fd, FAN_MARK_ADD | tc->mark.flags,
+				 tc->mask.flags, dirfd, path),
+				 exp_errs);
+	} else {
+		TST_EXP_FAIL(fanotify_mark(fanotify_fd, FAN_MARK_ADD | tc->mark.flags,
+						 tc->mask.flags, dirfd, path),
+						 tc->expected_errno);
+	}
 
 	/*
 	 * ENOTDIR are errors for events/flags not allowed on a non-dir inode.
Signature · 5 lines
2.44.0


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
#2 lore raw

Re: [LTP] [PATCH v2 1/1] fanotify14: fix anonymous pipe testcases

Quoted 74 lines · Mete Durlu

When SElinux is in enforcing state and SEpolicies disallow anonymous pipe usage with fanotify_mark(), related fanotify14 testcases fail with EACCES instead of EINVAL. Accept both errnos when SElinux is in enforcing state to correctly evaluate test results.

Replace TST_EXP_FD_OR_FAIL with TST_EXP_FAIL when testing fanotify_mark() as it returns -1 on failure and 0 on success not a file descriptor.

---
 .../kernel/syscalls/fanotify/fanotify14.c     | 32 +++++++++++++++++--
 1 file changed, 29 insertions(+), 3 deletions(-)
testcases/kernel/syscalls/fanotify/fanotify14.c +20 -3
index d02d81495..52c327dff 100644
@@ -27,12 +27,14 @@
 #define _GNU_SOURCE
 #include "tst_test.h"
 #include <errno.h>
+#include <stdlib.h>
 
 #ifdef HAVE_SYS_FANOTIFY_H
 #include "fanotify.h"
 
 #define MNTPOINT "mntpoint"
 #define FILE1 MNTPOINT"/file1"
+#define SELINUX_STATUS_PATH "/sys/fs/selinux/enforce"
 
 /*
  * List of inode events that are only available when notification group is
@@ -240,6 +242,19 @@ static struct test_case_t {
        },
 };
 
+static int is_selinux_enforcing(void)
+{
+       char res;
+       int fd;
+
+       fd = open(SELINUX_STATUS_PATH, O_RDONLY);
+       if (fd <= 0)
+               return 0;
+       SAFE_READ(1, fd, &res, 1);
+       SAFE_CLOSE(fd);
+       return atoi(&res);
+}
+
 static void do_test(unsigned int number)
 {
        struct test_case_t *tc = &test_cases[number];
@@ -275,17 +290,28 @@ static void do_test(unsigned int number)
        /* Set mark on non-dir only when expecting error ENOTDIR */
        const char *path = tc->expected_errno == ENOTDIR ? FILE1 : MNTPOINT;
        int dirfd = AT_FDCWD;
+       int se_enforcing = 0;
 
        if (tc->pfd) {
                dirfd = tc->pfd[0];
                path = NULL;
+               se_enforcing = is_selinux_enforcing();
        }
 
        tst_res(TINFO, "Testing %s with %s",
                tc->mark.desc, tc->mask.desc);
-       TST_EXP_FD_OR_FAIL(fanotify_mark(fanotify_fd, FAN_MARK_ADD | tc->mark.flags,
-                                        tc->mask.flags, dirfd, path),
-                                        tc->expected_errno);
+
+       if (tc->pfd && se_enforcing) {
+               const int exp_errs[] = {tc->expected_errno, EACCES};

testcases/kernel/syscalls/fanotify/fanotify14.c static void do_test(unsigned int number) · line 315

Quoted 10 lines
testcases/kernel/syscalls/fanotify/fanotify14.c +9 -0
+
+               TST_EXP_FAIL_ARR(fanotify_mark(fanotify_fd, FAN_MARK_ADD | tc->mark.flags,
+                                tc->mask.flags, dirfd, path),
+                                exp_errs);
+       } else {
+               TST_EXP_FAIL(fanotify_mark(fanotify_fd, FAN_MARK_ADD | tc->mark.flags,
+                                                tc->mask.flags, dirfd, path),
+                                                tc->expected_errno);
+       }
 

This looks fine to me, but on second thought I am not sure how important it is to special case se_enforcing.
We could probably always check for either error value.

Let's see what Jan and Petr think.

Thanks,
Amir.

Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
Reviewed-by
#3 lore raw

Re: [LTP] [PATCH v2 1/1] fanotify14: fix anonymous pipe testcases

Hi all,
...

Quoted 9 lines
Quoted 8 lines

static void do_test(unsigned int number)

 {
        struct test_case_t *tc = &test_cases[number];
(quoted diff) +1 -0
@@ -275,17 +290,28 @@ static void do_test(unsigned int number)
        /* Set mark on non-dir only when expecting error ENOTDIR */
        const char *path = tc->expected_errno == ENOTDIR ? FILE1 : MNTPOINT;
        int dirfd = AT_FDCWD;
+       int se_enforcing = 0;
Quoted 4 lines

Review static void do_test(unsigned int number) · line 297

Quoted 4 lines
(quoted diff) +1 -0
        if (tc->pfd) {
                dirfd = tc->pfd[0];
                path = NULL;
+               se_enforcing = is_selinux_enforcing();

nit: this check should be in the setup function to be done only once. (by default it's done twice, because we have 2 testcases with pfd, we support -iN parameter, thus it's actually 2*N.). I'll fix it before merge.

Quoted 16 lines
Quoted 16 lines
(quoted diff) +9 -3
        }
 
        tst_res(TINFO, "Testing %s with %s",
                tc->mark.desc, tc->mask.desc);
-       TST_EXP_FD_OR_FAIL(fanotify_mark(fanotify_fd, FAN_MARK_ADD | tc->mark.flags,
-                                        tc->mask.flags, dirfd, path),
-                                        tc->expected_errno);
+
+       if (tc->pfd && se_enforcing) {
+               const int exp_errs[] = {tc->expected_errno, EACCES};
+
+               TST_EXP_FAIL_ARR(fanotify_mark(fanotify_fd, FAN_MARK_ADD | tc->mark.flags,
+                                tc->mask.flags, dirfd, path),
+                                exp_errs);
+       } else {
+               TST_EXP_FAIL(fanotify_mark(fanotify_fd, FAN_MARK_ADD | tc->mark.flags,
Quoted 8 lines

Review static void do_test(unsigned int number) · line 313

Quoted 3 lines
(quoted diff) +3 -0
+                                                tc->mask.flags, dirfd, path),
+                                                tc->expected_errno);
+       }

This looks fine to me, but on second thought I am not sure how important it is to special case se_enforcing.
We could probably always check for either error value.

I don't mind explicitly testing EACCES with SELinux. @Jan WDYT?

With a diff below (I can change it before merge + I would do the following work to integrate this into the LTP C API):

Kind regards,
Petr

kernel/syscalls/fanotify/fanotify14.c +3 -2
index 52c327dff..89d59c8b2 100644
@@ -49,6 +49,7 @@ static int pipes[2] = {-1, -1};
 static int fanotify_fd;
 static int ignore_mark_unsupported;
 static int filesystem_mark_unsupported;
+static int se_enforcing;
 static unsigned int supported_init_flags;
 
 struct test_case_flags_t {
@@ -290,12 +291,10 @@ static void do_test(unsigned int number)
 	/* Set mark on non-dir only when expecting error ENOTDIR */
 	const char *path = tc->expected_errno == ENOTDIR ? FILE1 : MNTPOINT;
 	int dirfd = AT_FDCWD;
-	int se_enforcing = 0;
 
 	if (tc->pfd) {
 		dirfd = tc->pfd[0];
 		path = NULL;
-		se_enforcing = is_selinux_enforcing();
 	}
 
 	tst_res(TINFO, "Testing %s with %s",
@@ -360,6 +359,8 @@ static void do_setup(void)
 	SAFE_FILE_PRINTF(FILE1, "0");
 	/* Create anonymous pipes to place marks on */
 	SAFE_PIPE2(pipes, O_CLOEXEC);
+
+	se_enforcing = is_selinux_enforcing();
 }
 
 static void do_cleanup(void)
Quoted 4 lines

Let's see what Jan and Petr think.

Thanks,
Amir.

Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
Reviewed-by
#4 lore raw

Re: [LTP] [PATCH v2 1/1] fanotify14: fix anonymous pipe testcases

Quoted 44 lines · Petr Vorel

Hi all,
...

Quoted 9 lines
Quoted 8 lines

static void do_test(unsigned int number)

 {
        struct test_case_t *tc = &test_cases[number];
kernel/syscalls/fanotify/fanotify14.c +1 -0
@@ -275,17 +290,28 @@ static void do_test(unsigned int number)
        /* Set mark on non-dir only when expecting error ENOTDIR */
        const char *path = tc->expected_errno == ENOTDIR ? FILE1 : MNTPOINT;
        int dirfd = AT_FDCWD;
+       int se_enforcing = 0;
Quoted 4 lines

kernel/syscalls/fanotify/fanotify14.c static void do_test(unsigned int number)

Quoted 4 lines
kernel/syscalls/fanotify/fanotify14.c +1 -0
        if (tc->pfd) {
                dirfd = tc->pfd[0];
                path = NULL;
+               se_enforcing = is_selinux_enforcing();

nit: this check should be in the setup function to be done only once. (by default it's done twice, because we have 2 testcases with pfd, we support -iN parameter, thus it's actually 2*N.). I'll fix it before merge.

Quoted 16 lines
Quoted 16 lines
(quoted diff) +9 -3
        }
 
        tst_res(TINFO, "Testing %s with %s",
                tc->mark.desc, tc->mask.desc);
-       TST_EXP_FD_OR_FAIL(fanotify_mark(fanotify_fd, FAN_MARK_ADD | tc->mark.flags,
-                                        tc->mask.flags, dirfd, path),
-                                        tc->expected_errno);
+
+       if (tc->pfd && se_enforcing) {
+               const int exp_errs[] = {tc->expected_errno, EACCES};
+
+               TST_EXP_FAIL_ARR(fanotify_mark(fanotify_fd, FAN_MARK_ADD | tc->mark.flags,
+                                tc->mask.flags, dirfd, path),
+                                exp_errs);
+       } else {
+               TST_EXP_FAIL(fanotify_mark(fanotify_fd, FAN_MARK_ADD | tc->mark.flags,
Quoted 8 lines

Review static void do_test(unsigned int number) · line 313

Quoted 3 lines
(quoted diff) +3 -0
+                                                tc->mask.flags, dirfd, path),
+                                                tc->expected_errno);
+       }

This looks fine to me, but on second thought I am not sure how important it is to special case se_enforcing.
We could probably always check for either error value.

Quoted 5 lines

I don't mind explicitly testing EACCES with SELinux. @Jan WDYT?

With a diff below (I can change it before merge + I would do the following work to integrate this into the LTP C API):

Yes, looks fine to me as well. I don't feel strongly whether we should accept EACCESS unconditionally or only with SELinux. I suspect eventually we might need to accept it unconditionally as there may be other security modules that would block addition of the mark. But let's see what the future brings. So feel free to add:

Honza

Signature · 5 lines
Jan Kara <[email protected]>
SUSE Labs, CR

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
author
#5 lore raw

Re: [LTP] [PATCH v2 1/1] fanotify14: fix anonymous pipe testcases

In reply to Petr Vorel

On 3/13/24 08:26, Petr Vorel wrote:

Hi,

thanks for the review. I can send a v3 with the suggested changes if that will make things easier. Just let me know.

Quoted 9 lines
Quoted 4 lines
Quoted 4 lines
         if (tc->pfd) {
                 dirfd = tc->pfd[0];
                 path = NULL;

+ se_enforcing = is_selinux_enforcing();

nit: this check should be in the setup function to be done only once. (by default it's done twice, because we have 2 testcases with pfd, we support -iN parameter, thus it's actually 2*N.). I'll fix it before merge.

Quoted 1 line
Quoted 1 line
         }

Nice catch! I fully forgot that there was a setup function while I was trying to find the best TST_ macro to use.

Quoted 70 lines
Quoted 14 lines
Quoted 14 lines

tst_res(TINFO, "Testing %s with %s",
tc->mark.desc, tc->mask.desc);

(quoted diff) +9 -3
-       TST_EXP_FD_OR_FAIL(fanotify_mark(fanotify_fd, FAN_MARK_ADD | tc->mark.flags,
-                                        tc->mask.flags, dirfd, path),
-                                        tc->expected_errno);
+
+       if (tc->pfd && se_enforcing) {
+               const int exp_errs[] = {tc->expected_errno, EACCES};
+
+               TST_EXP_FAIL_ARR(fanotify_mark(fanotify_fd, FAN_MARK_ADD | tc->mark.flags,
+                                tc->mask.flags, dirfd, path),
+                                exp_errs);
+       } else {
+               TST_EXP_FAIL(fanotify_mark(fanotify_fd, FAN_MARK_ADD | tc->mark.flags,
Quoted 8 lines

Review

Quoted 3 lines
(quoted diff) +3 -0
+                                                tc->mask.flags, dirfd, path),
+                                                tc->expected_errno);
+       }

This looks fine to me, but on second thought I am not sure how important it is to special case se_enforcing.
We could probably always check for either error value.

I don't mind explicitly testing EACCES with SELinux. @Jan WDYT?

With a diff below (I can change it before merge + I would do the following work to integrate this into the LTP C API):

Kind regards,
Petr

kernel/syscalls/fanotify/fanotify14.c +3 -2
index 52c327dff..89d59c8b2 100644
@@ -49,6 +49,7 @@ static int pipes[2] = {-1, -1};
  static int fanotify_fd;
  static int ignore_mark_unsupported;
  static int filesystem_mark_unsupported;
+static int se_enforcing;
  static unsigned int supported_init_flags;
 
  struct test_case_flags_t {
@@ -290,12 +291,10 @@ static void do_test(unsigned int number)
  	/* Set mark on non-dir only when expecting error ENOTDIR */
  	const char *path = tc->expected_errno == ENOTDIR ? FILE1 : MNTPOINT;
  	int dirfd = AT_FDCWD;
-	int se_enforcing = 0;
 
  	if (tc->pfd) {
  		dirfd = tc->pfd[0];
  		path = NULL;
-		se_enforcing = is_selinux_enforcing();
  	}
 
  	tst_res(TINFO, "Testing %s with %s",
@@ -360,6 +359,8 @@ static void do_setup(void)
  	SAFE_FILE_PRINTF(FILE1, "0");
  	/* Create anonymous pipes to place marks on */
  	SAFE_PIPE2(pipes, O_CLOEXEC);
+
+	se_enforcing = is_selinux_enforcing();
  }
 
  static void do_cleanup(void)
Quoted 4 lines

Let's see what Jan and Petr think.

Thanks,
Amir.

Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp