[PATCH v2 0/2] mm/memfd: fix sysctl MEMFD_NOEXEC_SCOPE_NOEXEC_ENFORCED

[PATCH v2 0/2] mm/memfd: fix sysctl MEMFD_NOEXEC_SCOPE_NOEXEC_ENFORCED

[jeffxu]

From: Jeff Xu <jeffxu@google.com>

When sysctl vm.memfd_noexec is 2 (MEMFD_NOEXEC_SCOPE_NOEXEC_ENFORCED),
memfd_create(.., MFD_EXEC) should fail.

This complies with how MEMFD_NOEXEC_SCOPE_NOEXEC_ENFORCED is
defined - "memfd_create() without MFD_NOEXEC_SEAL will be rejected"

Thanks to Dominique Martinet <asmadeus@codewreck.org> who reported the bug.
see [1] for context.

[1] https://lore.kernel.org/linux-mm/CABi2SkXUX_QqTQ10Yx9bBUGpN1wByOi_=gZU6WEy5a8MaQY3Jw@mail.gmail.com/T/

History:
V2: fix build error when CONFIG_SYSCTL is not defined. 

V1: initial version
https://lore.kernel.org/linux-mm/20230630031721.623955-3-jeffxu@google.com/T/

Jeff Xu (2):
  mm/memfd: sysctl: fix MEMFD_NOEXEC_SCOPE_NOEXEC_ENFORCED
  selftests/memfd: sysctl: fix MEMFD_NOEXEC_SCOPE_NOEXEC_ENFORCED

 mm/memfd.c                                 | 57 +++++++++++++---------
 tools/testing/selftests/memfd/memfd_test.c |  5 ++
 2 files changed, 38 insertions(+), 24 deletions(-)

-- 
2.41.0.255.g8b1d071c50-goog

[PATCH v2 1/2] mm/memfd: sysctl: fix MEMFD_NOEXEC_SCOPE_NOEXEC_ENFORCED

[jeffxu]

From: Jeff Xu <jeffxu@google.com>

When vm.memfd_noexec is 2 (MEMFD_NOEXEC_SCOPE_NOEXEC_ENFORCED),
memfd_create(.., MFD_EXEC) should fail.

This complies with how MEMFD_NOEXEC_SCOPE_NOEXEC_ENFORCED is
defined - "memfd_create() without MFD_NOEXEC_SEAL will be rejected"

Fixes: 105ff5339f49 ("mm/memfd: add MFD_NOEXEC_SEAL and MFD_EXEC")
Reported-by: Dominique Martinet <asmadeus@codewreck.org>
Closes: https://lore.kernel.org/linux-mm/CABi2SkXUX_QqTQ10Yx9bBUGpN1wByOi_=gZU6WEy5a8MaQY3Jw@mail.gmail.com/T/
Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202306301351.kkbSegQW-lkp@intel.com/
Signed-off-by: Jeff Xu <jeffxu@google.com>
---
 mm/memfd.c | 57 +++++++++++++++++++++++++++++++-----------------------
 1 file changed, 33 insertions(+), 24 deletions(-)

diff --git a/mm/memfd.c b/mm/memfd.c
index e763e76f1106..0bdbd2335af7 100644
--- a/mm/memfd.c
+++ b/mm/memfd.c
@@ -268,6 +268,36 @@ long memfd_fcntl(struct file *file, unsigned int cmd, unsigned int arg)
 
 #define MFD_ALL_FLAGS (MFD_CLOEXEC | MFD_ALLOW_SEALING | MFD_HUGETLB | MFD_NOEXEC_SEAL | MFD_EXEC)
 
+static int check_sysctl_memfd_noexec(unsigned int *flags)
+{
+#ifdef CONFIG_SYSCTL
+	char comm[TASK_COMM_LEN];
+	int sysctl = MEMFD_NOEXEC_SCOPE_EXEC;
+	struct pid_namespace *ns;
+
+	ns = task_active_pid_ns(current);
+	if (ns)
+		sysctl = ns->memfd_noexec_scope;
+
+	if (!(*flags & (MFD_EXEC | MFD_NOEXEC_SEAL))) {
+		if (sysctl == MEMFD_NOEXEC_SCOPE_NOEXEC_SEAL)
+			*flags |= MFD_NOEXEC_SEAL;
+		else
+			*flags |= MFD_EXEC;
+	}
+
+	if (*flags & MFD_EXEC && sysctl >= MEMFD_NOEXEC_SCOPE_NOEXEC_ENFORCED) {
+		pr_warn_once(
+			"memfd_create(): MFD_NOEXEC_SEAL is enforced, pid=%d '%s'\n",
+			task_pid_nr(current), get_task_comm(comm, current));
+
+		return -EACCES;
+	}
+#endif
+
+	return 0;
+}
+
 SYSCALL_DEFINE2(memfd_create,
 		const char __user *, uname,
 		unsigned int, flags)
@@ -294,35 +324,14 @@ SYSCALL_DEFINE2(memfd_create,
 		return -EINVAL;
 
 	if (!(flags & (MFD_EXEC | MFD_NOEXEC_SEAL))) {
-#ifdef CONFIG_SYSCTL
-		int sysctl = MEMFD_NOEXEC_SCOPE_EXEC;
-		struct pid_namespace *ns;
-
-		ns = task_active_pid_ns(current);
-		if (ns)
-			sysctl = ns->memfd_noexec_scope;
-
-		switch (sysctl) {
-		case MEMFD_NOEXEC_SCOPE_EXEC:
-			flags |= MFD_EXEC;
-			break;
-		case MEMFD_NOEXEC_SCOPE_NOEXEC_SEAL:
-			flags |= MFD_NOEXEC_SEAL;
-			break;
-		default:
-			pr_warn_once(
-				"memfd_create(): MFD_NOEXEC_SEAL is enforced, pid=%d '%s'\n",
-				task_pid_nr(current), get_task_comm(comm, current));
-			return -EINVAL;
-		}
-#else
-		flags |= MFD_EXEC;
-#endif
 		pr_warn_once(
 			"memfd_create() without MFD_EXEC nor MFD_NOEXEC_SEAL, pid=%d '%s'\n",
 			task_pid_nr(current), get_task_comm(comm, current));
 	}
 
+	if (check_sysctl_memfd_noexec(&flags) < 0)
+		return -EACCES;
+
 	/* length includes terminating zero */
 	len = strnlen_user(uname, MFD_NAME_MAX_LEN + 1);
 	if (len <= 0)
-- 
2.41.0.255.g8b1d071c50-goog

[PATCH v2 2/2] selftests/memfd: sysctl: fix MEMFD_NOEXEC_SCOPE_NOEXEC_ENFORCED

[jeffxu]

From: Jeff Xu <jeffxu@google.com>

Add selftest for sysctl vm.memfd_noexec is 2
(MEMFD_NOEXEC_SCOPE_NOEXEC_ENFORCED)

memfd_create(.., MFD_EXEC) should fail in this case.

Reported-by: Dominique Martinet <asmadeus@codewreck.org>
Closes: https://lore.kernel.org/linux-mm/CABi2SkXUX_QqTQ10Yx9bBUGpN1wByOi_=gZU6WEy5a8MaQY3Jw@mail.gmail.com/T/
Signed-off-by: Jeff Xu <jeffxu@google.com>
---
 tools/testing/selftests/memfd/memfd_test.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/tools/testing/selftests/memfd/memfd_test.c b/tools/testing/selftests/memfd/memfd_test.c
index dba0e8ba002f..dbdd9ec5e397 100644
--- a/tools/testing/selftests/memfd/memfd_test.c
+++ b/tools/testing/selftests/memfd/memfd_test.c
@@ -1147,6 +1147,11 @@ static void test_sysctl_child(void)
 	sysctl_assert_write("2");
 	mfd_fail_new("kern_memfd_sysctl_2",
 		MFD_CLOEXEC | MFD_ALLOW_SEALING);
+	mfd_fail_new("kern_memfd_sysctl_2_MFD_EXEC",
+		MFD_CLOEXEC | MFD_EXEC);
+	fd = mfd_assert_new("", 0, MFD_NOEXEC_SEAL);
+	close(fd);
+
 	sysctl_fail_write("0");
 	sysctl_fail_write("1");
 }
-- 
2.41.0.255.g8b1d071c50-goog