* [HEADS-UP]: mount(8) non-root user mount changes
@ 2019-11-19 14:34 Karel Zak
2019-12-03 14:43 ` Karel Zak
0 siblings, 1 reply; 2+ messages in thread
From: Karel Zak @ 2019-11-19 14:34 UTC (permalink / raw)
To: util-linux
Hi,
for years users are unhappy with too strict mount/umount rules for
non-root users. The rules makes mount(8) useless for use-cases where
root permissions are unnecessary. This patch changes the game.
Please, review, complain, ...
Karel
From 6497f2d99e9cabee3531e644ba4dcffd14532200 Mon Sep 17 00:00:00 2001
From: Karel Zak <kzak@redhat.com>
Date: Tue, 19 Nov 2019 14:58:20 +0100
Subject: [PATCH] mount: no exit on EPERM, continue without suid
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
The current libmount assumes that mount(8) and umount(8) are suid
binaries. For this reason it implements internal rules which
restrict what is allowed for non-root users. Unfortunately, it's
out of reality for some use-cases where root permissions are no
required. Nice example are fuse filesystems.
So, the current situation is to call exit() always when mount, umount or
libmount are unsure with non-root user rights. This patch removes the
exit() call and replaces it with suid permissions drop, after that it
continues as usually. It means after suid-drop all depend on kernel
and no another security rule is used by libmount (simply because any
rule is no more necessary).
Example:
old version:
$ mount -t fuse.sshfs kzak@192.168.111.1:/home/kzak /home/kzak/mnt
mount: only root can use "--types" option
new version:
$ mount -t fuse.sshfs kzak@192.168.111.1:/home/kzak /home/kzak/mnt
kzak@192.168.111.1's password:
$ findmnt /home/kzak/mnt
TARGET SOURCE FSTYPE OPTIONS
/home/kzak/mnt kzak@192.168.111.1:/home/kzak fuse.sshfs rw,nosuid,nodev,relatime,user_id=1000,group_id=1000
$ umount /home/kzak/mnt
$ echo $?
0
Note that fuse user umount is supported since v2.34 due to user_id= in
kernel mount table.
Signed-off-by: Karel Zak <kzak@redhat.com>
---
libmount/docs/libmount-sections.txt | 1 +
libmount/src/context.c | 25 +++++++++++++++++
libmount/src/libmount.h.in | 1 +
libmount/src/libmount.sym | 1 +
sys-utils/mount.8 | 6 ++++
sys-utils/mount.c | 43 +++++++++++++++++------------
sys-utils/umount.8 | 19 +++++++++++++
sys-utils/umount.c | 37 +++++++++++++++----------
8 files changed, 102 insertions(+), 31 deletions(-)
diff --git a/libmount/docs/libmount-sections.txt b/libmount/docs/libmount-sections.txt
index 990c0394f..0bba9f835 100644
--- a/libmount/docs/libmount-sections.txt
+++ b/libmount/docs/libmount-sections.txt
@@ -40,6 +40,7 @@ mnt_context_enable_rwonly_mount
mnt_context_enable_sloppy
mnt_context_enable_verbose
mnt_context_forced_rdonly
+mnt_context_force_unrestricted
mnt_context_get_cache
mnt_context_get_excode
mnt_context_get_fs
diff --git a/libmount/src/context.c b/libmount/src/context.c
index 977842f7d..72a0c5a7b 100644
--- a/libmount/src/context.c
+++ b/libmount/src/context.c
@@ -426,6 +426,31 @@ int mnt_context_is_restricted(struct libmnt_context *cxt)
return cxt->restricted;
}
+/**
+ * mnt_context_force_unrestricted:
+ * @cxt: mount context
+ *
+ * This function is DANGEROURS as it disables all security policies in libmount.
+ * Don't use if not sure. It removes "restricted" flag from the context, so
+ * libmount will use the current context as for root user.
+ *
+ * This function is designed for case you have no any suid permissions, so you
+ * can depend on kernel.
+ *
+ * Returns: 0 on success, negative number in case of error.
+ *
+ * Since: 2.35
+ */
+int mnt_context_force_unrestricted(struct libmnt_context *cxt)
+{
+ if (mnt_context_is_restricted(cxt)) {
+ DBG(CXT, ul_debugobj(cxt, "force UNRESTRICTED"));
+ cxt->restricted = 0;
+ }
+
+ return 0;
+}
+
/**
* mnt_context_set_optsmode
* @cxt: mount context
diff --git a/libmount/src/libmount.h.in b/libmount/src/libmount.h.in
index 19d4c5b53..ba54cf25d 100644
--- a/libmount/src/libmount.h.in
+++ b/libmount/src/libmount.h.in
@@ -705,6 +705,7 @@ extern void mnt_free_context(struct libmnt_context *cxt);
extern int mnt_reset_context(struct libmnt_context *cxt);
extern int mnt_context_is_restricted(struct libmnt_context *cxt)
__ul_attribute__((nonnull));
+extern int mnt_context_force_unrestricted(struct libmnt_context *cxt);
extern int mnt_context_init_helper(struct libmnt_context *cxt,
int action, int flags);
diff --git a/libmount/src/libmount.sym b/libmount/src/libmount.sym
index b9a4c0d22..792d11753 100644
--- a/libmount/src/libmount.sym
+++ b/libmount/src/libmount.sym
@@ -352,6 +352,7 @@ MOUNT_2.34 {
} MOUNT_2.33;
MOUNT_2_35 {
+ mnt_context_force_unrestricted;
mnt_context_get_target_prefix;
mnt_context_set_target_prefix;
} MOUNT_2.34;
diff --git a/sys-utils/mount.8 b/sys-utils/mount.8
index 9d31d8245..a6231c7c1 100644
--- a/sys-utils/mount.8
+++ b/sys-utils/mount.8
@@ -315,6 +315,12 @@ program is executed. It's strongly recommended to use a valid mountpoint to
specify filesystem, otherwise \fBmount\fR may fail. For example it's bad idea
to use NFS or CIFS source on command line.
.PP
+Since version 2.35 \fBmount\fR command does not exit when user permissions are
+inadequate by internal libmount security rules. It drops suid permissions
+and continue as regular non-root user. It allows to support use-cases where
+root permissions are not necessary (e.g. fuse filesystems, user namespaces,
+etc).
+.PP
For more details, see
.BR fstab (5).
Only the user that mounted a filesystem can unmount it again.
diff --git a/sys-utils/mount.c b/sys-utils/mount.c
index 08da9e6a5..5842bc2ec 100644
--- a/sys-utils/mount.c
+++ b/sys-utils/mount.c
@@ -47,23 +47,24 @@
static int mk_exit_code(struct libmnt_context *cxt, int rc);
-static void __attribute__((__noreturn__)) exit_non_root(const char *option)
+static void suid_drop(struct libmnt_context *cxt)
{
const uid_t ruid = getuid();
const uid_t euid = geteuid();
- if (ruid == 0 && euid != 0) {
- /* user is root, but setuid to non-root */
- if (option)
- errx(MNT_EX_USAGE, _("only root can use \"--%s\" option "
- "(effective UID is %u)"),
- option, euid);
- errx(MNT_EX_USAGE, _("only root can do that "
- "(effective UID is %u)"), euid);
+ if (ruid != 0 && euid == 0) {
+ if (setgid(getgid()) < 0)
+ err(MNT_EX_FAIL, _("setgid() failed"));
+
+ if (setuid(getuid()) < 0)
+ err(MNT_EX_FAIL, _("setuid() failed"));
}
- if (option)
- errx(MNT_EX_USAGE, _("only root can use \"--%s\" option"), option);
- errx(MNT_EX_USAGE, _("only root can do that"));
+
+ /* be paranoid and check it, setuid(0) has to fail */
+ if (ruid != 0 && setuid(0) == 0)
+ errx(MNT_EX_FAIL, _("drop permissions failed."));
+
+ mnt_context_force_unrestricted(cxt);
}
static void __attribute__((__noreturn__)) mount_print_version(void)
@@ -672,7 +673,7 @@ int main(int argc, char **argv)
!strchr("hlLUVvrist", c) &&
c != MOUNT_OPT_TARGET &&
c != MOUNT_OPT_SOURCE)
- exit_non_root(option_to_longopt(c, longopts));
+ suid_drop(cxt);
err_exclusive_options(c, longopts, excl, excl_st);
@@ -872,7 +873,7 @@ int main(int argc, char **argv)
/* Non-root users are allowed to use -t to print_all(),
but not to mount */
if (mnt_context_is_restricted(cxt) && types)
- exit_non_root("types");
+ suid_drop(cxt);
if (oper && (types || all || mnt_context_get_source(cxt))) {
warnx(_("bad usage"));
@@ -905,7 +906,7 @@ int main(int argc, char **argv)
if (mnt_context_is_restricted(cxt) &&
mnt_context_get_source(cxt) &&
mnt_context_get_target(cxt))
- exit_non_root(NULL);
+ suid_drop(cxt);
} else if (argc == 1 && (!mnt_context_get_source(cxt) ||
!mnt_context_get_target(cxt))) {
@@ -933,7 +934,7 @@ int main(int argc, char **argv)
if (mnt_context_is_restricted(cxt) &&
mnt_context_get_source(cxt) &&
mnt_context_get_target(cxt))
- exit_non_root(NULL);
+ suid_drop(cxt);
} else if (argc == 2 && !mnt_context_get_source(cxt)
&& !mnt_context_get_target(cxt)) {
@@ -941,7 +942,7 @@ int main(int argc, char **argv)
* D) mount <source> <target>
*/
if (mnt_context_is_restricted(cxt))
- exit_non_root(NULL);
+ suid_drop(cxt);
mnt_context_set_source(cxt, argv[0]);
mnt_context_set_target(cxt, argv[1]);
@@ -963,6 +964,14 @@ int main(int argc, char **argv)
mnt_context_set_optsmode(cxt, MNT_OMODE_NOTAB);
rc = mnt_context_mount(cxt);
+
+ if (rc == -EPERM
+ && mnt_context_is_restricted(cxt)
+ && !mnt_context_syscall_called(cxt)) {
+ /* Try it again without permissions */
+ suid_drop(cxt);
+ rc = mnt_context_mount(cxt);
+ }
rc = mk_exit_code(cxt, rc);
if (rc == MNT_EX_SUCCESS && mnt_context_is_verbose(cxt))
diff --git a/sys-utils/umount.8 b/sys-utils/umount.8
index f94d2f41f..9bec521e7 100644
--- a/sys-utils/umount.8
+++ b/sys-utils/umount.8
@@ -190,6 +190,25 @@ Display version information and exit.
.TP
.BR \-h , " \-\-help"
Display help text and exit.
+.SH "NON-SUPERUSER UMOUNTS"
+Normally, only the superuser can umount filesystems.
+However, when
+.I fstab
+contains the
+.B user
+option on a line, anybody can umount the corresponding filesystem. For more details see
+.BR mount (8)
+man page.
+.PP
+Since version 2.34 \fBumount\fR command allows to perform umount operation also
+for fuse filesystems if kernel mount table contains user's ID. In this case fstab
+user= mount option is not required.
+.PP
+Since version 2.35 \fBumount\fR command does not exit when user permissions are
+inadequate by internal libmount security rules. It drops suid permissions
+and continue as regular non-root user. It allows to support use-cases where
+root permissions are not necessary (e.g. fuse filesystems, user namespaces,
+etc).
.SH "LOOP DEVICE"
The
.B umount
diff --git a/sys-utils/umount.c b/sys-utils/umount.c
index 397e0ebfc..74d87d671 100644
--- a/sys-utils/umount.c
+++ b/sys-utils/umount.c
@@ -112,24 +112,24 @@ static void __attribute__((__noreturn__)) usage(void)
exit(MNT_EX_SUCCESS);
}
-static void __attribute__((__noreturn__)) exit_non_root(const char *option)
+static void suid_drop(struct libmnt_context *cxt)
{
const uid_t ruid = getuid();
const uid_t euid = geteuid();
- if (ruid == 0 && euid != 0) {
- /* user is root, but setuid to non-root */
- if (option)
- errx(MNT_EX_USAGE,
- _("only root can use \"--%s\" option "
- "(effective UID is %u)"),
- option, euid);
- errx(MNT_EX_USAGE, _("only root can do that "
- "(effective UID is %u)"), euid);
+ if (ruid != 0 && euid == 0) {
+ if (setgid(getgid()) < 0)
+ err(MNT_EX_FAIL, _("setgid() failed"));
+
+ if (setuid(getuid()) < 0)
+ err(MNT_EX_FAIL, _("setuid() failed"));
}
- if (option)
- errx(MNT_EX_USAGE, _("only root can use \"--%s\" option"), option);
- errx(MNT_EX_USAGE, _("only root can do that"));
+
+ /* be paranoid and check it, setuid(0) has to fail */
+ if (ruid != 0 && setuid(0) == 0)
+ errx(MNT_EX_FAIL, _("drop permissions failed."));
+
+ mnt_context_force_unrestricted(cxt);
}
static void success_message(struct libmnt_context *cxt)
@@ -220,6 +220,15 @@ static int umount_one(struct libmnt_context *cxt, const char *spec)
err(MNT_EX_SYSERR, _("failed to set umount target"));
rc = mnt_context_umount(cxt);
+
+ if (rc == -EPERM
+ && mnt_context_is_restricted(cxt)
+ && !mnt_context_syscall_called(cxt)) {
+ /* Failed somewhere in libmount, drop perms and try it again */
+ suid_drop(cxt);
+ rc = mnt_context_umount(cxt);
+ }
+
rc = mk_exit_code(cxt, rc);
if (rc == MNT_EX_SUCCESS && mnt_context_is_verbose(cxt))
@@ -494,7 +503,7 @@ int main(int argc, char **argv)
/* only few options are allowed for non-root users */
if (mnt_context_is_restricted(cxt) && !strchr("hdilqVv", c))
- exit_non_root(option_to_longopt(c, longopts));
+ suid_drop(cxt);
err_exclusive_options(c, longopts, excl, excl_st);
--
2.21.0
^ permalink raw reply related [flat|nested] 2+ messages in thread
* Re: [HEADS-UP]: mount(8) non-root user mount changes
2019-11-19 14:34 [HEADS-UP]: mount(8) non-root user mount changes Karel Zak
@ 2019-12-03 14:43 ` Karel Zak
0 siblings, 0 replies; 2+ messages in thread
From: Karel Zak @ 2019-12-03 14:43 UTC (permalink / raw)
To: util-linux
On Tue, Nov 19, 2019 at 03:34:11PM +0100, Karel Zak wrote:
> for years users are unhappy with too strict mount/umount rules for
> non-root users. The rules makes mount(8) useless for use-cases where
> root permissions are unnecessary. This patch changes the game.
Merged, we will see... ;-) BTW, I'd like to make -rc1 next week.
Karel
--
Karel Zak <kzak@redhat.com>
http://karelzak.blogspot.com
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2019-12-03 14:43 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-11-19 14:34 [HEADS-UP]: mount(8) non-root user mount changes Karel Zak
2019-12-03 14:43 ` Karel Zak
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).