mnt: Add tests for unprivileged remount cases that have found to be faulty
authorEric W. Biederman <ebiederm@xmission.com>
Tue, 29 Jul 2014 22:50:44 +0000 (15:50 -0700)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 17 Sep 2014 16:04:00 +0000 (09:04 -0700)
commit db181ce011e3c033328608299cd6fac06ea50130 upstream.

Kenton Varda <kenton@sandstorm.io> discovered that by remounting a
read-only bind mount read-only in a user namespace the
MNT_LOCK_READONLY bit would be cleared, allowing an unprivileged user
to the remount a read-only mount read-write.

Upon review of the code in remount it was discovered that the code allowed
nosuid, noexec, and nodev to be cleared.  It was also discovered that
the code was allowing the per mount atime flags to be changed.

The first naive patch to fix these issues contained the flaw that using
default atime settings when remounting a filesystem could be disallowed.

To avoid this problems in the future add tests to ensure unprivileged
remounts are succeeding and failing at the appropriate times.

Acked-by: Serge E. Hallyn <serge.hallyn@ubuntu.com>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
tools/testing/selftests/Makefile
tools/testing/selftests/mount/Makefile [new file with mode: 0644]
tools/testing/selftests/mount/unprivileged-remount-test.c [new file with mode: 0644]

index 0a63658065f0e9e41285f521665b80b975cfe86c..2cee2b79b4defc8a5bf584d765d46906187b0462 100644 (file)
@@ -4,6 +4,7 @@ TARGETS += efivarfs
 TARGETS += kcmp
 TARGETS += memory-hotplug
 TARGETS += mqueue
+TARGETS += mount
 TARGETS += net
 TARGETS += ptrace
 TARGETS += vm
diff --git a/tools/testing/selftests/mount/Makefile b/tools/testing/selftests/mount/Makefile
new file mode 100644 (file)
index 0000000..337d853
--- /dev/null
@@ -0,0 +1,17 @@
+# Makefile for mount selftests.
+
+all: unprivileged-remount-test
+
+unprivileged-remount-test: unprivileged-remount-test.c
+       gcc -Wall -O2 unprivileged-remount-test.c -o unprivileged-remount-test
+
+# Allow specific tests to be selected.
+test_unprivileged_remount: unprivileged-remount-test
+       @if [ -f /proc/self/uid_map ] ; then ./unprivileged-remount-test ; fi
+
+run_tests: all test_unprivileged_remount
+
+clean:
+       rm -f unprivileged-remount-test
+
+.PHONY: all test_unprivileged_remount
diff --git a/tools/testing/selftests/mount/unprivileged-remount-test.c b/tools/testing/selftests/mount/unprivileged-remount-test.c
new file mode 100644 (file)
index 0000000..1b3ff2f
--- /dev/null
@@ -0,0 +1,242 @@
+#define _GNU_SOURCE
+#include <sched.h>
+#include <stdio.h>
+#include <errno.h>
+#include <string.h>
+#include <sys/types.h>
+#include <sys/mount.h>
+#include <sys/wait.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <fcntl.h>
+#include <grp.h>
+#include <stdbool.h>
+#include <stdarg.h>
+
+#ifndef CLONE_NEWNS
+# define CLONE_NEWNS 0x00020000
+#endif
+#ifndef CLONE_NEWUTS
+# define CLONE_NEWUTS 0x04000000
+#endif
+#ifndef CLONE_NEWIPC
+# define CLONE_NEWIPC 0x08000000
+#endif
+#ifndef CLONE_NEWNET
+# define CLONE_NEWNET 0x40000000
+#endif
+#ifndef CLONE_NEWUSER
+# define CLONE_NEWUSER 0x10000000
+#endif
+#ifndef CLONE_NEWPID
+# define CLONE_NEWPID 0x20000000
+#endif
+
+#ifndef MS_RELATIME
+#define MS_RELATIME (1 << 21)
+#endif
+#ifndef MS_STRICTATIME
+#define MS_STRICTATIME (1 << 24)
+#endif
+
+static void die(char *fmt, ...)
+{
+       va_list ap;
+       va_start(ap, fmt);
+       vfprintf(stderr, fmt, ap);
+       va_end(ap);
+       exit(EXIT_FAILURE);
+}
+
+static void write_file(char *filename, char *fmt, ...)
+{
+       char buf[4096];
+       int fd;
+       ssize_t written;
+       int buf_len;
+       va_list ap;
+
+       va_start(ap, fmt);
+       buf_len = vsnprintf(buf, sizeof(buf), fmt, ap);
+       va_end(ap);
+       if (buf_len < 0) {
+               die("vsnprintf failed: %s\n",
+                   strerror(errno));
+       }
+       if (buf_len >= sizeof(buf)) {
+               die("vsnprintf output truncated\n");
+       }
+
+       fd = open(filename, O_WRONLY);
+       if (fd < 0) {
+               die("open of %s failed: %s\n",
+                   filename, strerror(errno));
+       }
+       written = write(fd, buf, buf_len);
+       if (written != buf_len) {
+               if (written >= 0) {
+                       die("short write to %s\n", filename);
+               } else {
+                       die("write to %s failed: %s\n",
+                               filename, strerror(errno));
+               }
+       }
+       if (close(fd) != 0) {
+               die("close of %s failed: %s\n",
+                       filename, strerror(errno));
+       }
+}
+
+static void create_and_enter_userns(void)
+{
+       uid_t uid;
+       gid_t gid;
+
+       uid = getuid();
+       gid = getgid();
+
+       if (unshare(CLONE_NEWUSER) !=0) {
+               die("unshare(CLONE_NEWUSER) failed: %s\n",
+                       strerror(errno));
+       }
+
+       write_file("/proc/self/uid_map", "0 %d 1", uid);
+       write_file("/proc/self/gid_map", "0 %d 1", gid);
+
+       if (setgroups(0, NULL) != 0) {
+               die("setgroups failed: %s\n",
+                       strerror(errno));
+       }
+       if (setgid(0) != 0) {
+               die ("setgid(0) failed %s\n",
+                       strerror(errno));
+       }
+       if (setuid(0) != 0) {
+               die("setuid(0) failed %s\n",
+                       strerror(errno));
+       }
+}
+
+static
+bool test_unpriv_remount(int mount_flags, int remount_flags, int invalid_flags)
+{
+       pid_t child;
+
+       child = fork();
+       if (child == -1) {
+               die("fork failed: %s\n",
+                       strerror(errno));
+       }
+       if (child != 0) { /* parent */
+               pid_t pid;
+               int status;
+               pid = waitpid(child, &status, 0);
+               if (pid == -1) {
+                       die("waitpid failed: %s\n",
+                               strerror(errno));
+               }
+               if (pid != child) {
+                       die("waited for %d got %d\n",
+                               child, pid);
+               }
+               if (!WIFEXITED(status)) {
+                       die("child did not terminate cleanly\n");
+               }
+               return WEXITSTATUS(status) == EXIT_SUCCESS ? true : false;
+       }
+
+       create_and_enter_userns();
+       if (unshare(CLONE_NEWNS) != 0) {
+               die("unshare(CLONE_NEWNS) failed: %s\n",
+                       strerror(errno));
+       }
+
+       if (mount("testing", "/tmp", "ramfs", mount_flags, NULL) != 0) {
+               die("mount of /tmp failed: %s\n",
+                       strerror(errno));
+       }
+
+       create_and_enter_userns();
+
+       if (unshare(CLONE_NEWNS) != 0) {
+               die("unshare(CLONE_NEWNS) failed: %s\n",
+                       strerror(errno));
+       }
+
+       if (mount("/tmp", "/tmp", "none",
+                 MS_REMOUNT | MS_BIND | remount_flags, NULL) != 0) {
+               /* system("cat /proc/self/mounts"); */
+               die("remount of /tmp failed: %s\n",
+                   strerror(errno));
+       }
+
+       if (mount("/tmp", "/tmp", "none",
+                 MS_REMOUNT | MS_BIND | invalid_flags, NULL) == 0) {
+               /* system("cat /proc/self/mounts"); */
+               die("remount of /tmp with invalid flags "
+                   "succeeded unexpectedly\n");
+       }
+       exit(EXIT_SUCCESS);
+}
+
+static bool test_unpriv_remount_simple(int mount_flags)
+{
+       return test_unpriv_remount(mount_flags, mount_flags, 0);
+}
+
+static bool test_unpriv_remount_atime(int mount_flags, int invalid_flags)
+{
+       return test_unpriv_remount(mount_flags, mount_flags, invalid_flags);
+}
+
+int main(int argc, char **argv)
+{
+       if (!test_unpriv_remount_simple(MS_RDONLY|MS_NODEV)) {
+               die("MS_RDONLY malfunctions\n");
+       }
+       if (!test_unpriv_remount_simple(MS_NODEV)) {
+               die("MS_NODEV malfunctions\n");
+       }
+       if (!test_unpriv_remount_simple(MS_NOSUID|MS_NODEV)) {
+               die("MS_NOSUID malfunctions\n");
+       }
+       if (!test_unpriv_remount_simple(MS_NOEXEC|MS_NODEV)) {
+               die("MS_NOEXEC malfunctions\n");
+       }
+       if (!test_unpriv_remount_atime(MS_RELATIME|MS_NODEV,
+                                      MS_NOATIME|MS_NODEV))
+       {
+               die("MS_RELATIME malfunctions\n");
+       }
+       if (!test_unpriv_remount_atime(MS_STRICTATIME|MS_NODEV,
+                                      MS_NOATIME|MS_NODEV))
+       {
+               die("MS_STRICTATIME malfunctions\n");
+       }
+       if (!test_unpriv_remount_atime(MS_NOATIME|MS_NODEV,
+                                      MS_STRICTATIME|MS_NODEV))
+       {
+               die("MS_RELATIME malfunctions\n");
+       }
+       if (!test_unpriv_remount_atime(MS_RELATIME|MS_NODIRATIME|MS_NODEV,
+                                      MS_NOATIME|MS_NODEV))
+       {
+               die("MS_RELATIME malfunctions\n");
+       }
+       if (!test_unpriv_remount_atime(MS_STRICTATIME|MS_NODIRATIME|MS_NODEV,
+                                      MS_NOATIME|MS_NODEV))
+       {
+               die("MS_RELATIME malfunctions\n");
+       }
+       if (!test_unpriv_remount_atime(MS_NOATIME|MS_NODIRATIME|MS_NODEV,
+                                      MS_STRICTATIME|MS_NODEV))
+       {
+               die("MS_RELATIME malfunctions\n");
+       }
+       if (!test_unpriv_remount(MS_STRICTATIME|MS_NODEV, MS_NODEV,
+                                MS_NOATIME|MS_NODEV))
+       {
+               die("Default atime malfunctions\n");
+       }
+       return EXIT_SUCCESS;
+}