commit 9c8e0a1b68 upstream.
Timothy Baldwin <timbaldwin@fastmail.co.uk> wrote:
> As per mount_namespaces(7) unprivileged users should not be able to look under mount points:
>
> Mounts that come as a single unit from more privileged mount are locked
> together and may not be separated in a less privileged mount namespace.
>
> However they can:
>
> 1. Create a mount namespace.
> 2. In the mount namespace open a file descriptor to the parent of a mount point.
> 3. Destroy the mount namespace.
> 4. Use the file descriptor to look under the mount point.
>
> I have reproduced this with Linux 4.16.18 and Linux 4.18-rc8.
>
> The setup:
>
> $ sudo sysctl kernel.unprivileged_userns_clone=1
> kernel.unprivileged_userns_clone = 1
> $ mkdir -p A/B/Secret
> $ sudo mount -t tmpfs hide A/B
>
>
> "Secret" is indeed hidden as expected:
>
> $ ls -lR A
> A:
> total 0
> drwxrwxrwt 2 root root 40 Feb 12 21:08 B
>
> A/B:
> total 0
>
>
> The attack revealing "Secret":
>
> $ unshare -Umr sh -c "exec unshare -m ls -lR /proc/self/fd/4/ 4<A"
> /proc/self/fd/4/:
> total 0
> drwxr-xr-x 3 root root 60 Feb 12 21:08 B
>
> /proc/self/fd/4/B:
> total 0
> drwxr-xr-x 2 root root 40 Feb 12 21:08 Secret
>
> /proc/self/fd/4/B/Secret:
> total 0
I tracked this down to put_mnt_ns running passing UMOUNT_SYNC and
disconnecting all of the mounts in a mount namespace. Fix this by
factoring drop_mounts out of drop_collected_mounts and passing
0 instead of UMOUNT_SYNC.
There are two possible behavior differences that result from this.
- No longer setting UMOUNT_SYNC will no longer set MNT_SYNC_UMOUNT on
the vfsmounts being unmounted. This effects the lazy rcu walk by
kicking the walk out of rcu mode and forcing it to be a non-lazy
walk.
- No longer disconnecting locked mounts will keep some mounts around
longer as they stay because the are locked to other mounts.
There are only two users of drop_collected mounts: audit_tree.c and
put_mnt_ns.
In audit_tree.c the mounts are private and there are no rcu lazy walks
only calls to iterate_mounts. So the changes should have no effect
except for a small timing effect as the connected mounts are disconnected.
In put_mnt_ns there may be references from process outside the mount
namespace to the mounts. So the mounts remaining connected will
be the bug fix that is needed. That rcu walks are allowed to continue
appears not to be a problem especially as the rcu walk change was about
an implementation detail not about semantics.
Fixes: 5ff9d8a65c ("vfs: Lock in place mounts from more privileged users")
Reported-by: Timothy Baldwin <timbaldwin@fastmail.co.uk>
Tested-by: Timothy Baldwin <timbaldwin@fastmail.co.uk>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
commit df7342b240 upstream.
Jonathan Calmels from NVIDIA reported that he's able to bypass the
mount visibility security check in place in the Linux kernel by using
a combination of the unbindable property along with the private mount
propagation option to allow a unprivileged user to see a path which
was purposefully hidden by the root user.
Reproducer:
# Hide a path to all users using a tmpfs
root@castiana:~# mount -t tmpfs tmpfs /sys/devices/
root@castiana:~#
# As an unprivileged user, unshare user namespace and mount namespace
stgraber@castiana:~$ unshare -U -m -r
# Confirm the path is still not accessible
root@castiana:~# ls /sys/devices/
# Make /sys recursively unbindable and private
root@castiana:~# mount --make-runbindable /sys
root@castiana:~# mount --make-private /sys
# Recursively bind-mount the rest of /sys over to /mnnt
root@castiana:~# mount --rbind /sys/ /mnt
# Access our hidden /sys/device as an unprivileged user
root@castiana:~# ls /mnt/devices/
breakpoint cpu cstate_core cstate_pkg i915 intel_pt isa kprobe
LNXSYSTM:00 msr pci0000:00 platform pnp0 power software system
tracepoint uncore_arb uncore_cbox_0 uncore_cbox_1 uprobe virtual
Solve this by teaching copy_tree to fail if a mount turns out to be
both unbindable and locked.
Fixes: 5ff9d8a65c ("vfs: Lock in place mounts from more privileged users")
Reported-by: Jonathan Calmels <jcalmels@nvidia.com>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
commit 25d202ed82 upstream.
It was recently pointed out that the one instance of testing MNT_LOCKED
outside of the namespace_sem is in ksys_umount.
Fix that by adding a test inside of do_umount with namespace_sem and
the mount_lock held. As it helps to fail fails the existing test is
maintained with an additional comment pointing out that it may be racy
because the locks are not held.
Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
Fixes: 5ff9d8a65c ("vfs: Lock in place mounts from more privileged users")
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
commit 119e1ef80e upstream.
__legitimize_mnt() has two problems - one is that in case of success
the check of mount_lock is not ordered wrt preceding increment of
refcount, making it possible to have successful __legitimize_mnt()
on one CPU just before the otherwise final mntpu() on another,
with __legitimize_mnt() not seeing mntput() taking the lock and
mntput() not seeing the increment done by __legitimize_mnt().
Solved by a pair of barriers.
Another is that failure of __legitimize_mnt() on the second
read_seqretry() leaves us with reference that'll need to be
dropped by caller; however, if that races with final mntput()
we can end up with caller dropping rcu_read_lock() and doing
mntput() to release that reference - with the first mntput()
having freed the damn thing just as rcu_read_lock() had been
dropped. Solution: in "do mntput() yourself" failure case
grab mount_lock, check if MNT_DOOMED has been set by racing
final mntput() that has missed our increment and if it has -
undo the increment and treat that as "failure, caller doesn't
need to drop anything" case.
It's not easy to hit - the final mntput() has to come right
after the first read_seqretry() in __legitimize_mnt() *and*
manage to miss the increment done by __legitimize_mnt() before
the second read_seqretry() in there. The things that are almost
impossible to hit on bare hardware are not impossible on SMP
KVM, though...
Reported-by: Oleg Nesterov <oleg@redhat.com>
Fixes: 48a066e72d ("RCU'd vsfmounts")
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
[bwh: Backported to 3.16: __legitimize_mnt() has not been split out
from legitimize_mnt(). Adjust the added return statement and
comments accordingly.]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
commit 9ea0a46ca2 upstream.
mntput_no_expire() does the calculation of total refcount under mount_lock;
unfortunately, the decrement (as well as all increments) are done outside
of it, leading to false positives in the "are we dropping the last reference"
test. Consider the following situation:
* mnt is a lazy-umounted mount, kept alive by two opened files. One
of those files gets closed. Total refcount of mnt is 2. On CPU 42
mntput(mnt) (called from __fput()) drops one reference, decrementing component
* After it has looked at component #0, the process on CPU 0 does
mntget(), incrementing component #0, gets preempted and gets to run again -
on CPU 69. There it does mntput(), which drops the reference (component #69)
and proceeds to spin on mount_lock.
* On CPU 42 our first mntput() finishes counting. It observes the
decrement of component #69, but not the increment of component #0. As the
result, the total it gets is not 1 as it should've been - it's 0. At which
point we decide that vfsmount needs to be killed and proceed to free it and
shut the filesystem down. However, there's still another opened file
on that filesystem, with reference to (now freed) vfsmount, etc. and we are
screwed.
It's not a wide race, but it can be reproduced with artificial slowdown of
the mnt_get_count() loop, and it should be easier to hit on SMP KVM setups.
Fix consists of moving the refcount decrement under mount_lock; the tricky
part is that we want (and can) keep the fast case (i.e. mount that still
has non-NULL ->mnt_ns) entirely out of mount_lock. All places that zero
mnt->mnt_ns are dropping some reference to mnt and they call synchronize_rcu()
before that mntput(). IOW, if mntput() observes (under rcu_read_lock())
a non-NULL ->mnt_ns, it is guaranteed that there is another reference yet to
be dropped.
Reported-by: Jann Horn <jannh@google.com>
Tested-by: Jann Horn <jannh@google.com>
Fixes: 48a066e72d ("RCU'd vsfmounts")
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
[bwh: Backported to 3.16: Use ACCESS_ONCE() instead of READ_ONCE()]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
commit 16a34adb93 upstream.
We want it only for the stuff created by SB_KERNMOUNT mounts, *not* for
their copies. As it is, creating a deep stack of bindings of /proc/*/ns/*
somewhere in a new namespace and exiting yields a stack overflow.
Reported-by: Alexander Aring <aring@mojatatu.com>
Bisected-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Tested-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Tested-by: Alexander Aring <aring@mojatatu.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
commit b8850d1fa8 upstream.
The gcc version 4.9.1 compiler complains Even though it isn't possible for
these variables to not get initialized before they are used.
fs/namespace.c: In function ‘SyS_mount’:
fs/namespace.c:2720:8: warning: ‘kernel_dev’ may be used uninitialized in this function [-Wmaybe-uninitialized]
ret = do_mount(kernel_dev, kernel_dir->name, kernel_type, flags,
^
fs/namespace.c:2699:8: note: ‘kernel_dev’ was declared here
char *kernel_dev;
^
fs/namespace.c:2720:8: warning: ‘kernel_type’ may be used uninitialized in this function [-Wmaybe-uninitialized]
ret = do_mount(kernel_dev, kernel_dir->name, kernel_type, flags,
^
fs/namespace.c:2697:8: note: ‘kernel_type’ was declared here
char *kernel_type;
^
Fix the warnings by simplifying copy_mount_string() as suggested by Al Viro.
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Cc: Arnd Bergmann <arnd@arndb.de>
commit d29216842a upstream.
CAI Qian <caiqian@redhat.com> pointed out that the semantics
of shared subtrees make it possible to create an exponentially
increasing number of mounts in a mount namespace.
mkdir /tmp/1 /tmp/2
mount --make-rshared /
for i in $(seq 1 20) ; do mount --bind /tmp/1 /tmp/2 ; done
Will create create 2^20 or 1048576 mounts, which is a practical problem
as some people have managed to hit this by accident.
As such CVE-2016-6213 was assigned.
Ian Kent <raven@themaw.net> described the situation for autofs users
as follows:
> The number of mounts for direct mount maps is usually not very large because of
> the way they are implemented, large direct mount maps can have performance
> problems. There can be anywhere from a few (likely case a few hundred) to less
> than 10000, plus mounts that have been triggered and not yet expired.
>
> Indirect mounts have one autofs mount at the root plus the number of mounts that
> have been triggered and not yet expired.
>
> The number of autofs indirect map entries can range from a few to the common
> case of several thousand and in rare cases up to between 30000 and 50000. I've
> not heard of people with maps larger than 50000 entries.
>
> The larger the number of map entries the greater the possibility for a large
> number of active mounts so it's not hard to expect cases of a 1000 or somewhat
> more active mounts.
So I am setting the default number of mounts allowed per mount
namespace at 100,000. This is more than enough for any use case I
know of, but small enough to quickly stop an exponential increase
in mounts. Which should be perfect to catch misconfigurations and
malfunctioning programs.
For anyone who needs a higher limit this can be changed by writing
to the new /proc/sys/fs/mount-max sysctl.
Tested-by: CAI Qian <caiqian@redhat.com>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
[bwh: Backported to 3.16:
- Use ACCESS_ONCE() instead of READ_ONCE()
- Adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
commit d71ed6c930 upstream.
MNT_LOCKED implies on a child mount implies the child is locked to the
parent. So while looping through the children the children should be
tested (not their parent).
Typically an unshare of a mount namespace locks all mounts together
making both the parent and the slave as locked but there are a few
corner cases where other things work.
Fixes: ceeb0e5d39 ("vfs: Ignore unlocked mounts in fs_fully_visible")
Reported-by: Seth Forshee <seth.forshee@canonical.com>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
[bwh: Backported to 3.16: adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
commit 7236c85e1b upstream.
fs_fully_visible attempts to make fresh mounts of proc and sysfs give
the mounter no more access to proc and sysfs than if they could have
by creating a bind mount. One aspect of proc and sysfs that makes
this particularly tricky is that there are other filesystems that
typically mount on top of proc and sysfs. As those filesystems are
mounted on empty directories in practice it is safe to ignore them.
However testing to ensure filesystems are mounted on empty directories
has not been something the in kernel data structures have supported so
the current test for an empty directory which checks to see
if nlink <= 2 is a bit lacking.
proc and sysfs have recently been modified to use the new empty_dir
infrastructure to create all of their dedicated mount points. Instead
of testing for S_ISDIR(inode->i_mode) && i_nlink <= 2 to see if a
directory is empty, test for is_empty_dir_inode(inode). That small
change guaranteess mounts found on proc and sysfs really are safe to
ignore, because the directories are not only empty but nothing can
ever be added to them. This guarantees there is nothing to worry
about when mounting proc and sysfs.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Luis Henriques <luis.henriques@canonical.com>
commit ceeb0e5d39 upstream.
Limit the mounts fs_fully_visible considers to locked mounts.
Unlocked can always be unmounted so considering them adds hassle
but no security benefit.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Luis Henriques <luis.henriques@canonical.com>
commit 8c6cf9cc82 upstream.
Ignore an existing mount if the locked readonly, nodev or atime
attributes are less permissive than the desired attributes
of the new mount.
On success ensure the new mount locks all of the same readonly, nodev and
atime attributes as the old mount.
The nosuid and noexec attributes are not checked here as this change
is destined for stable and enforcing those attributes causes a
regression in lxc and libvirt-lxc where those applications will not
start and there are no known executables on sysfs or proc and no known
way to create exectuables without code modifications
Fixes: e51db73532 ("userns: Better restrictions on when proc and sysfs can be mounted")
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Luis Henriques <luis.henriques@canonical.com>
commit 1b852bceb0 upstream.
Fresh mounts of proc and sysfs are a very special case that works very
much like a bind mount. Unfortunately the current structure can not
preserve the MNT_LOCK... mount flags. Therefore refactor the logic
into a form that can be modified to preserve those lock bits.
Add a new filesystem flag FS_USERNS_VISIBLE that requires some mount
of the filesystem be fully visible in the current mount namespace,
before the filesystem may be mounted.
Move the logic for calling fs_fully_visible from proc and sysfs into
fs/namespace.c where it has greater access to mount namespace state.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Luis Henriques <luis.henriques@canonical.com>
commit 7e96c1b0e0 upstream.
This fixes a dumb bug in fs_fully_visible that allows proc or sys to
be mounted if there is a bind mount of part of /proc/ or /sys/ visible.
Reported-by: Eric Windisch <ewindisch@docker.com>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Luis Henriques <luis.henriques@canonical.com>
commit cd4a40174b upstream.
The only users of collect_mounts are in audit_tree.c
In audit_trim_trees and audit_add_tree_rule the path passed into
collect_mounts is generated from kern_path passed an audit_tree
pathname which is guaranteed to be an absolute path. In those cases
collect_mounts is obviously intended to work on mounted paths and
if a race results in paths that are unmounted when collect_mounts
it is reasonable to fail early.
The paths passed into audit_tag_tree don't have the absolute path
check. But are used to play with fsnotify and otherwise interact with
the audit_trees, so again operating only on mounted paths appears
reasonable.
Avoid having to worry about what happens when we try and audit
unmounted filesystems by restricting collect_mounts to mounts
that appear in the mount tree.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: Luis Henriques <luis.henriques@canonical.com>
commit e819f15210 upstream.
- Remove the unneeded declaration from pnode.h
- Mark umount_tree static as it has no callers outside of namespace.c
- Define an enumeration of umount_tree's flags.
- Pass umount_tree's flags in by name
This removes the magic numbers 0, 1 and 2 making the code a little
clearer and makes it possible for there to be lazy unmounts that don't
propagate. Which is what __detach_mounts actually wants for example.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
[ luis: backported to 3.16:
- dropped change to __detach_mounts() ]
Signed-off-by: Luis Henriques <luis.henriques@canonical.com>
commit c297abfdf1 upstream.
While reviewing the code of umount_tree I realized that when we append
to a preexisting unmounted list we do not change pprev of the former
first item in the list.
Which means later in namespace_unlock hlist_del_init(&mnt->mnt_hash) on
the former first item of the list will stomp unmounted.first leaving
it set to some random mount point which we are likely to free soon.
This isn't likely to hit, but if it does I don't know how anyone could
track it down.
[ This happened because we don't have all the same operations for
hlist's as we do for normal doubly-linked lists. In particular,
list_splice() is easy on our standard doubly-linked lists, while
hlist_splice() doesn't exist and needs both start/end entries of the
hlist. And commit 38129a13e6 incorrectly open-coded that missing
hlist_splice().
We should think about making these kinds of "mindless" conversions
easier to get right by adding the missing hlist helpers - Linus ]
Fixes: 38129a13e6 switch mnt_hash to hlist
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Luis Henriques <luis.henriques@canonical.com>
commit b2f5d4dc38 upstream.
Forced unmount affects not just the mount namespace but the underlying
superblock as well. Restrict forced unmount to the global root user
for now. Otherwise it becomes possible a user in a less privileged
mount namespace to force the shutdown of a superblock of a filesystem
in a more privileged mount namespace, allowing a DOS attack on root.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Luis Henriques <luis.henriques@canonical.com>
commit 3e1866410f upstream.
Now that remount is properly enforcing the rule that you can't remove
nodev at least sandstorm.io is breaking when performing a remount.
It turns out that there is an easy intuitive solution implicitly
add nodev on remount when nodev was implicitly added on mount.
Tested-by: Cedric Bosdonnat <cbosdonnat@suse.com>
Tested-by: Richard Weinberger <richard@nod.at>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Luis Henriques <luis.henriques@canonical.com>
commit 0d0826019e upstream.
Andy Lutomirski recently demonstrated that when chroot is used to set
the root path below the path for the new ``root'' passed to pivot_root
the pivot_root system call succeeds and leaks mounts.
In examining the code I see that starting with a new root that is
below the current root in the mount tree will result in a loop in the
mount tree after the mounts are detached and then reattached to one
another. Resulting in all kinds of ugliness including a leak of that
mounts involved in the leak of the mount loop.
Prevent this problem by ensuring that the new mount is reachable from
the current root of the mount tree.
[Added stable cc. Fixes CVE-2014-7970. --Andy]
Reported-by: Andy Lutomirski <luto@amacapital.net>
Reviewed-by: Andy Lutomirski <luto@amacapital.net>
Link: http://lkml.kernel.org/r/87bnpmihks.fsf@x220.int.ebiederm.org
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Luis Henriques <luis.henriques@canonical.com>
commit a1480dcc3c upstream.
Accessing do_remount_sb should require global CAP_SYS_ADMIN, but
only one of the two call sites was appropriately protected.
Fixes CVE-2014-7975.
Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 12a5b5294c upstream.
Since 3.14 we had copy_tree() get the shadowing wrong - if we had one
vfsmount shadowing another (i.e. if A is a slave of B, C is mounted
on A/foo, then D got mounted on B/foo creating D' on A/foo shadowed
by C), copy_tree() of A would make a copy of D' shadow the the copy of
C, not the other way around.
It's easy to fix, fortunately - just make sure that mount follows
the one that shadows it in mnt_child as well as in mnt_hash, and when
copy_tree() decides to attach a new mount, check if the last child
it has added to the same parent should be shadowing the new one.
And if it should, just use the same logics commit_tree() has - put the
new mount into the hash and children lists right after the one that
should shadow it.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 81b6b06197 upstream.
We need the parents of victims alive until namespace_unlock() gets to
dput() of the (ex-)mountpoints. However, that screws up the "is it
busy" checks in case when we have shrinkable mounts that need to be
killed. Solution: go ahead and decrement refcounts of parents right
in umount_tree(), increment them again just before dropping rwsem in
namespace_unlock() (and let the loop in the end of namespace_unlock()
finally drop those references for good, as we do now). Parents can't
get freed until we drop rwsem - at least one reference is kept until
then, both in case when parent is among the victims and when it is
not. So they'll still be around when we get to namespace_unlock().
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 88b368f27a upstream.
The check in __propagate_umount() ("has somebody explicitly mounted
something on that slave?") is done *before* taking the already doomed
victims out of the child lists.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit ffbc6f0ead upstream.
Since March 2009 the kernel has treated the state that if no
MS_..ATIME flags are passed then the kernel defaults to relatime.
Defaulting to relatime instead of the existing atime state during a
remount is silly, and causes problems in practice for people who don't
specify any MS_...ATIME flags and to get the default filesystem atime
setting. Those users may encounter a permission error because the
default atime setting does not work.
A default that does not work and causes permission problems is
ridiculous, so preserve the existing value to have a default
atime setting that is always guaranteed to work.
Using the default atime setting in this way is particularly
interesting for applications built to run in restricted userspace
environments without /proc mounted, as the existing atime mount
options of a filesystem can not be read from /proc/mounts.
In practice this fixes user space that uses the default atime
setting on remount that are broken by the permission checks
keeping less privileged users from changing more privileged users
atime settings.
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>
commit 9566d67428 upstream.
While invesgiating the issue where in "mount --bind -oremount,ro ..."
would result in later "mount --bind -oremount,rw" succeeding even if
the mount started off locked I realized that there are several
additional mount flags that should be locked and are not.
In particular MNT_NOSUID, MNT_NODEV, MNT_NOEXEC, and the atime
flags in addition to MNT_READONLY should all be locked. These
flags are all per superblock, can all be changed with MS_BIND,
and should not be changable if set by a more privileged user.
The following additions to the current logic are added in this patch.
- nosuid may not be clearable by a less privileged user.
- nodev may not be clearable by a less privielged user.
- noexec may not be clearable by a less privileged user.
- atime flags may not be changeable by a less privileged user.
The logic with atime is that always setting atime on access is a
global policy and backup software and auditing software could break if
atime bits are not updated (when they are configured to be updated),
and serious performance degradation could result (DOS attack) if atime
updates happen when they have been explicitly disabled. Therefore an
unprivileged user should not be able to mess with the atime bits set
by a more privileged user.
The additional restrictions are implemented with the addition of
MNT_LOCK_NOSUID, MNT_LOCK_NODEV, MNT_LOCK_NOEXEC, and MNT_LOCK_ATIME
mnt flags.
Taken together these changes and the fixes for MNT_LOCK_READONLY
should make it safe for an unprivileged user to create a user
namespace and to call "mount --bind -o remount,... ..." without
the danger of mount flags being changed maliciously.
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>
commit 07b645589d upstream.
There are no races as locked mount flags are guaranteed to never change.
Moving the test into do_remount makes it more visible, and ensures all
filesystem remounts pass the MNT_LOCK_READONLY permission check. This
second case is not an issue today as filesystem remounts are guarded
by capable(CAP_DAC_ADMIN) and thus will always fail in less privileged
mount namespaces, but it could become an issue in the future.
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>
commit a6138db815 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.
Correct this by replacing the mask of mount flags to preserve
with a mask of mount flags that may be changed, and preserve
all others. This ensures that any future bugs with this mask and
remount will fail in an easy to detect way where new mount flags
simply won't change.
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>
Make delayed_free() call free_vfsmnt() so that we don't have two functions
doing the same job. This requires the calls to mnt_free_id() in free_vfsmnt()
to be moved into the callers of that function.
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
new flag in ->f_mode - FMODE_WRITER. Set by do_dentry_open() in case
when it has grabbed write access, checked by __fput() to decide whether
it wants to drop the sucker. Allows to stop bothering with mnt_clone_write()
in alloc_file(), along with fewer special_file() checks.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
The current mainline has copies propagated to *all* nodes, then
tears down the copies we made for nodes that do not contain
counterparts of the desired mountpoint. That sets the right
propagation graph for the copies (at teardown time we move
the slaves of removed node to a surviving peer or directly
to master), but we end up paying a fairly steep price in
useless allocations. It's fairly easy to create a situation
where N calls of mount(2) create exactly N bindings, with
O(N^2) vfsmounts allocated and freed in process.
Fortunately, it is possible to avoid those allocations/freeings.
The trick is to create copies in the right order and find which
one would've eventually become a master with the current algorithm.
It turns out to be possible in O(nodes getting propagation) time
and with no extra allocations at all.
One part is that we need to make sure that eventual master will be
created before its slaves, so we need to walk the propagation
tree in a different order - by peer groups. And iterate through
the peers before dealing with the next group.
Another thing is finding the (earlier) copy that will be a master
of one we are about to create; to do that we are (temporary) marking
the masters of mountpoints we are attaching the copies to.
Either we are in a peer of the last mountpoint we'd dealt with,
or we have the following situation: we are attaching to mountpoint M,
the last copy S_0 had been attached to M_0 and there are sequences
S_0...S_n, M_0...M_n such that S_{i+1} is a master of S_{i},
S_{i} mounted on M{i} and we need to create a slave of the first S_{k}
such that M is getting propagation from M_{k}. It means that the master
of M_{k} will be among the sequence of masters of M. On the
other hand, the nearest marked node in that sequence will either
be the master of M_{k} or the master of M_{k-1} (the latter -
in the case if M_{k-1} is a slave of something M gets propagation
from, but in a wrong peer group).
So we go through the sequence of masters of M until we find
a marked one (P). Let N be the one before it. Then we go through
the sequence of masters of S_0 until we find one (say, S) mounted
on a node D that has P as master and check if D is a peer of N.
If it is, S will be the master of new copy, if not - the master of S
will be.
That's it for the hard part; the rest is fairly simple. Iterator
is in next_group(), handling of one prospective mountpoint is
propagate_one().
It seems to survive all tests and gives a noticably better performance
than the current mainline for setups that are seriously using shared
subtrees.
Cc: stable@vger.kernel.org
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
fixes RCU bug - walking through hlist is safe in face of element moves,
since it's self-terminating. Cyclic lists are not - if we end up jumping
to another hash chain, we'll loop infinitely without ever hitting the
original list head.
[fix for dumb braino folded]
Spotted by: Max Kellermann <mk@cm4all.com>
Cc: stable@vger.kernel.org
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
If the dest_mnt is not shared, propagate_mnt() does nothing -
there's no mounts to propagate to and thus no copies to create.
Might as well don't bother calling it in that case.
Cc: stable@vger.kernel.org
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
* switch allocation to alloc_large_system_hash()
* make sizes overridable by boot parameters (mhash_entries=, mphash_entries=)
* switch mountpoint_hashtable from list_head to hlist_head
Cc: stable@vger.kernel.org
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Here's the big driver core and sysfs patch set for 3.14-rc1.
There's a lot of work here moving sysfs logic out into a "kernfs" to
allow other subsystems to also have a virtual filesystem with the same
attributes of sysfs (handle device disconnect, dynamic creation /
removal as needed / unneeded, etc. This is primarily being done for
the cgroups filesystem, but the goal is to also move debugfs to it when
it is ready, solving all of the known issues in that filesystem as well.
The code isn't completed yet, but all should be stable now (there is a
big section that was reverted due to problems found when testing.)
There's also some other smaller fixes, and a driver core addition that
allows for a "collection" of objects, that the DRM people will be using
soon (it's in this tree to make merges after -rc1 easier.)
All of this has been in linux-next with no reported issues.
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (GNU/Linux)
iEYEABECAAYFAlLdh0cACgkQMUfUDdst+ylv4QCfeDKDgLo4LsaBIIrFSxLoH/c7
UUsAoMPRwA0h8wy+BQcJAg4H4J4maKj3
=0pc0
-----END PGP SIGNATURE-----
Merge tag 'driver-core-3.14-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core
Pull driver core / sysfs patches from Greg KH:
"Here's the big driver core and sysfs patch set for 3.14-rc1.
There's a lot of work here moving sysfs logic out into a "kernfs" to
allow other subsystems to also have a virtual filesystem with the same
attributes of sysfs (handle device disconnect, dynamic creation /
removal as needed / unneeded, etc)
This is primarily being done for the cgroups filesystem, but the goal
is to also move debugfs to it when it is ready, solving all of the
known issues in that filesystem as well. The code isn't completed
yet, but all should be stable now (there is a big section that was
reverted due to problems found when testing)
There's also some other smaller fixes, and a driver core addition that
allows for a "collection" of objects, that the DRM people will be
using soon (it's in this tree to make merges after -rc1 easier)
All of this has been in linux-next with no reported issues"
* tag 'driver-core-3.14-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core: (113 commits)
kernfs: associate a new kernfs_node with its parent on creation
kernfs: add struct dentry declaration in kernfs.h
kernfs: fix get_active failure handling in kernfs_seq_*()
Revert "kernfs: fix get_active failure handling in kernfs_seq_*()"
Revert "kernfs: replace kernfs_node->u.completion with kernfs_root->deactivate_waitq"
Revert "kernfs: remove KERNFS_ACTIVE_REF and add kernfs_lockdep()"
Revert "kernfs: remove KERNFS_REMOVED"
Revert "kernfs: restructure removal path to fix possible premature return"
Revert "kernfs: invoke kernfs_unmap_bin_file() directly from __kernfs_remove()"
Revert "kernfs: remove kernfs_addrm_cxt"
Revert "kernfs: make kernfs_get_active() block if the node is deactivated but not removed"
Revert "kernfs: implement kernfs_{de|re}activate[_self]()"
Revert "kernfs, sysfs, driver-core: implement kernfs_remove_self() and its wrappers"
Revert "pci: use device_remove_file_self() instead of device_schedule_callback()"
Revert "scsi: use device_remove_file_self() instead of device_schedule_callback()"
Revert "s390: use device_remove_file_self() instead of device_schedule_callback()"
Revert "sysfs, driver-core: remove unused {sysfs|device}_schedule_callback_owner()"
Revert "kernfs: remove unnecessary NULL check in __kernfs_remove()"
kernfs: remove unnecessary NULL check in __kernfs_remove()
drivers/base: provide an infrastructure for componentised subsystems
...
We're in the process of separating out core sysfs functionality into
kernfs which will deal with sysfs_dirents directly. This patch
rearranges mount path so that the kernfs and sysfs parts are separate.
* As sysfs_super_info won't be visible outside kernfs proper,
kernfs_super_ns() is added to allow kernfs users to access a
super_block's namespace tag.
* Generic mount operation is separated out into kernfs_mount_ns().
sysfs_mount() now just performs sysfs-specific permission check,
acquires namespace tag, and invokes kernfs_mount_ns().
* Generic superblock release is separated out into kernfs_kill_sb()
which can be used directly as file_system_type->kill_sb(). As sysfs
needs to put the namespace tag, sysfs_kill_sb() wraps
kernfs_kill_sb() with ns tag put.
* sysfs_dir_cachep init and sysfs_inode_init() are separated out into
kernfs_init(). kernfs_init() uses only small amount of memory and
trying to handle and propagate kernfs_init() failure doesn't make
much sense. Use SLAB_PANIC for sysfs_dir_cachep and make
sysfs_inode_init() panic on failure.
After this change, kernfs_init() should be called before
sysfs_init(), fs/namespace.c::mnt_init() modified accordingly.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: linux-fsdevel@vger.kernel.org
Cc: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Gao feng <gaofeng@cn.fujitsu.com> reported that commit
e51db73532
userns: Better restrictions on when proc and sysfs can be mounted
caused a regression on mounting a new instance of proc in a mount
namespace created with user namespace privileges, when binfmt_misc
is mounted on /proc/sys/fs/binfmt_misc.
This is an unintended regression caused by the absolutely bogus empty
directory check in fs_fully_visible. The check fs_fully_visible replaced
didn't even bother to attempt to verify proc was fully visible and
hiding proc files with any kind of mount is rare. So for now fix
the userspace regression by allowing directory with nlink == 1
as /proc/sys/fs/binfmt_misc has.
I will have a better patch but it is not stable material, or
last minute kernel material. So it will have to wait.
Cc: stable@vger.kernel.org
Acked-by: Serge Hallyn <serge.hallyn@canonical.com>
Acked-by: Gao feng <gaofeng@cn.fujitsu.com>
Tested-by: Gao feng <gaofeng@cn.fujitsu.com>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
* RCU-delayed freeing of vfsmounts
* vfsmount_lock replaced with a seqlock (mount_lock)
* sequence number from mount_lock is stored in nameidata->m_seq and
used when we exit RCU mode
* new vfsmount flag - MNT_SYNC_UMOUNT. Set by umount_tree() when its
caller knows that vfsmount will have no surviving references.
* synchronize_rcu() done between unlocking namespace_sem in namespace_unlock()
and doing pending mntput().
* new helper: legitimize_mnt(mnt, seq). Checks the mount_lock sequence
number against seq, then grabs reference to mnt. Then it rechecks mount_lock
again to close the race and either returns success or drops the reference it
has acquired. The subtle point is that in case of MNT_SYNC_UMOUNT we can
simply decrement the refcount and sod off - aforementioned synchronize_rcu()
makes sure that final mntput() won't come until we leave RCU mode. We need
that, since we don't want to end up with some lazy pathwalk racing with
umount() and stealing the final mntput() from it - caller of umount() may
expect it to return only once the fs is shut down and we don't want to break
that. In other cases (i.e. with MNT_SYNC_UMOUNT absent) we have to do
full-blown mntput() in case of mount_lock sequence number mismatch happening
just as we'd grabbed the reference, but in those cases we won't be stealing
the final mntput() from anything that would care.
* mntput_no_expire() doesn't lock anything on the fast path now. Incidentally,
SMP and UP cases are handled the same way - no ifdefs there.
* normal pathname resolution does *not* do any writes to mount_lock. It does,
of course, bump the refcounts of vfsmount and dentry in the very end, but that's
it.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Instead of passing the direction as argument (and checking it on every
step through the hash chain), just have separate __lookup_mnt() and
__lookup_mnt_last(). And use the standard iterators...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
aka br_write_{lock,unlock} of vfsmount_lock. Inlines in fs/mount.h,
vfsmount_lock extern moved over there as well.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
