linux-stable.git/net/core, branch linux-5.14.y

bpf, cgroups: Fix cgroup v2 fallback on v1/v2 mixed mode

2021-11-17T10:04:46+00:00

[ Upstream commit 8520e224f547cd070c7c8f97b1fc6d58cff7ccaa ]

Fix cgroup v1 interference when non-root cgroup v2 BPF programs are used.
Back in the days, commit bd1060a1d671 ("sock, cgroup: add sock->sk_cgroup")
embedded per-socket cgroup information into sock->sk_cgrp_data and in order
to save 8 bytes in struct sock made both mutually exclusive, that is, when
cgroup v1 socket tagging (e.g. net_cls/net_prio) is used, then cgroup v2
falls back to the root cgroup in sock_cgroup_ptr() (&cgrp_dfl_root.cgrp).

The assumption made was "there is no reason to mix the two and this is in line
with how legacy and v2 compatibility is handled" as stated in bd1060a1d671.
However, with Kubernetes more widely supporting cgroups v2 as well nowadays,
this assumption no longer holds, and the possibility of the v1/v2 mixed mode
with the v2 root fallback being hit becomes a real security issue.

Many of the cgroup v2 BPF programs are also used for policy enforcement, just
to pick _one_ example, that is, to programmatically deny socket related system
calls like connect(2) or bind(2). A v2 root fallback would implicitly cause
a policy bypass for the affected Pods.

In production environments, we have recently seen this case due to various
circumstances: i) a different 3rd party agent and/or ii) a container runtime
such as [0] in the user's environment configuring legacy cgroup v1 net_cls
tags, which triggered implicitly mentioned root fallback. Another case is
Kubernetes projects like kind [1] which create Kubernetes nodes in a container
and also add cgroup namespaces to the mix, meaning programs which are attached
to the cgroup v2 root of the cgroup namespace get attached to a non-root
cgroup v2 path from init namespace point of view. And the latter's root is
out of reach for agents on a kind Kubernetes node to configure. Meaning, any
entity on the node setting cgroup v1 net_cls tag will trigger the bypass
despite cgroup v2 BPF programs attached to the namespace root.

Generally, this mutual exclusiveness does not hold anymore in today's user
environments and makes cgroup v2 usage from BPF side fragile and unreliable.
This fix adds proper struct cgroup pointer for the cgroup v2 case to struct
sock_cgroup_data in order to address these issues; this implicitly also fixes
the tradeoffs being made back then with regards to races and refcount leaks
as stated in bd1060a1d671, and removes the fallback, so that cgroup v2 BPF
programs always operate as expected.

  [0] https://github.com/nestybox/sysbox/
  [1] https://kind.sigs.k8s.io/

Fixes: bd1060a1d671 ("sock, cgroup: add sock->sk_cgroup")
Signed-off-by: Daniel Borkmann 
Signed-off-by: Alexei Starovoitov 
Acked-by: Stanislav Fomichev 
Acked-by: Tejun Heo 
Link: https://lore.kernel.org/bpf/20210913230759.2313-1-daniel@iogearbox.net
Signed-off-by: Sasha Levin

net, neigh: Enable state migration between NUD_PERMANENT and NTF_USE

2021-11-17T10:04:46+00:00

[ Upstream commit 3dc20f4762c62d3b3f0940644881ed818aa7b2f5 ]

Currently, it is not possible to migrate a neighbor entry between NUD_PERMANENT
state and NTF_USE flag with a dynamic NUD state from a user space control plane.
Similarly, it is not possible to add/remove NTF_EXT_LEARNED flag from an existing
neighbor entry in combination with NTF_USE flag.

This is due to the latter directly calling into neigh_event_send() without any
meta data updates as happening in __neigh_update(). Thus, to enable this use
case, extend the latter with a NEIGH_UPDATE_F_USE flag where we break the
NUD_PERMANENT state in particular so that a latter neigh_event_send() is able
to re-resolve a neighbor entry.

Before fix, NUD_PERMANENT -> NUD_* & NTF_USE:

  # ./ip/ip n replace 192.168.178.30 dev enp5s0 lladdr f4:8c:50:5e:71:9a
  # ./ip/ip n
  192.168.178.30 dev enp5s0 lladdr f4:8c:50:5e:71:9a PERMANENT
  [...]
  # ./ip/ip n replace 192.168.178.30 dev enp5s0 use extern_learn
  # ./ip/ip n
  192.168.178.30 dev enp5s0 lladdr f4:8c:50:5e:71:9a PERMANENT
  [...]

As can be seen, despite the admin-triggered replace, the entry remains in the
NUD_PERMANENT state.

After fix, NUD_PERMANENT -> NUD_* & NTF_USE:

  # ./ip/ip n replace 192.168.178.30 dev enp5s0 lladdr f4:8c:50:5e:71:9a
  # ./ip/ip n
  192.168.178.30 dev enp5s0 lladdr f4:8c:50:5e:71:9a PERMANENT
  [...]
  # ./ip/ip n replace 192.168.178.30 dev enp5s0 use extern_learn
  # ./ip/ip n
  192.168.178.30 dev enp5s0 lladdr f4:8c:50:5e:71:9a extern_learn REACHABLE
  [...]
  # ./ip/ip n
  192.168.178.30 dev enp5s0 lladdr f4:8c:50:5e:71:9a extern_learn STALE
  [...]
  # ./ip/ip n replace 192.168.178.30 dev enp5s0 lladdr f4:8c:50:5e:71:9a
  # ./ip/ip n
  192.168.178.30 dev enp5s0 lladdr f4:8c:50:5e:71:9a PERMANENT
  [...]

After the fix, the admin-triggered replace switches to a dynamic state from
the NTF_USE flag which triggered a new neighbor resolution. Likewise, we can
transition back from there, if needed, into NUD_PERMANENT.

Similar before/after behavior can be observed for below transitions:

Before fix, NTF_USE -> NTF_USE | NTF_EXT_LEARNED -> NTF_USE:

  # ./ip/ip n replace 192.168.178.30 dev enp5s0 use
  # ./ip/ip n
  192.168.178.30 dev enp5s0 lladdr f4:8c:50:5e:71:9a REACHABLE
  [...]
  # ./ip/ip n replace 192.168.178.30 dev enp5s0 use extern_learn
  # ./ip/ip n
  192.168.178.30 dev enp5s0 lladdr f4:8c:50:5e:71:9a REACHABLE
  [...]

After fix, NTF_USE -> NTF_USE | NTF_EXT_LEARNED -> NTF_USE:

  # ./ip/ip n replace 192.168.178.30 dev enp5s0 use
  # ./ip/ip n
  192.168.178.30 dev enp5s0 lladdr f4:8c:50:5e:71:9a REACHABLE
  [...]
  # ./ip/ip n replace 192.168.178.30 dev enp5s0 use extern_learn
  # ./ip/ip n
  192.168.178.30 dev enp5s0 lladdr f4:8c:50:5e:71:9a extern_learn REACHABLE
  [...]
  # ./ip/ip n replace 192.168.178.30 dev enp5s0 use
  # ./ip/ip n
  192.168.178.30 dev enp5s0 lladdr f4:8c:50:5e:71:9a REACHABLE
  [..]

Signed-off-by: Daniel Borkmann 
Acked-by: Roopa Prabhu 
Signed-off-by: David S. Miller 
Signed-off-by: Sasha Levin

bpf, sockmap: sk_skb data_end access incorrect when src_reg = dst_reg

2021-11-17T10:04:42+00:00

[ Upstream commit b2c4618162ec615a15883a804cce7e27afecfa58 ]

The current conversion of skb->data_end reads like this:

  ; data_end = (void*)(long)skb->data_end;
   559: (79) r1 = *(u64 *)(r2 +200)   ; r1  = skb->data
   560: (61) r11 = *(u32 *)(r2 +112)  ; r11 = skb->len
   561: (0f) r1 += r11
   562: (61) r11 = *(u32 *)(r2 +116)
   563: (1f) r1 -= r11

But similar to the case in 84f44df664e9 ("bpf: sock_ops sk access may stomp
registers when dst_reg = src_reg"), the code will read an incorrect skb->len
when src == dst. In this case we end up generating this xlated code:

  ; data_end = (void*)(long)skb->data_end;
   559: (79) r1 = *(u64 *)(r1 +200)   ; r1  = skb->data
   560: (61) r11 = *(u32 *)(r1 +112)  ; r11 = (skb->data)->len
   561: (0f) r1 += r11
   562: (61) r11 = *(u32 *)(r1 +116)
   563: (1f) r1 -= r11

... where line 560 is the reading 4B of (skb->data + 112) instead of the
intended skb->len Here the skb pointer in r1 gets set to skb->data and the
later deref for skb->len ends up following skb->data instead of skb.

This fixes the issue similarly to the patch mentioned above by creating an
additional temporary variable and using to store the register when dst_reg =
src_reg. We name the variable bpf_temp_reg and place it in the cb context for
sk_skb. Then we restore from the temp to ensure nothing is lost.

Fixes: 16137b09a66f2 ("bpf: Compute data_end dynamically with JIT code")
Signed-off-by: Jussi Maki 
Signed-off-by: John Fastabend 
Signed-off-by: Daniel Borkmann 
Reviewed-by: Jakub Sitnicki 
Link: https://lore.kernel.org/bpf/20211103204736.248403-6-john.fastabend@gmail.com
Signed-off-by: Sasha Levin

bpf: sockmap, strparser, and tls are reusing qdisc_skb_cb and colliding

2021-11-17T10:04:42+00:00

[ Upstream commit e0dc3b93bd7bcff8c3813d1df43e0908499c7cf0 ]

Strparser is reusing the qdisc_skb_cb struct to stash the skb message handling
progress, e.g. offset and length of the skb. First this is poorly named and
inherits a struct from qdisc that doesn't reflect the actual usage of cb[] at
this layer.

But, more importantly strparser is using the following to access its metadata.

  (struct _strp_msg *)((void *)skb->cb + offsetof(struct qdisc_skb_cb, data))

Where _strp_msg is defined as:

  struct _strp_msg {
        struct strp_msg            strp;                 /*     0     8 */
        int                        accum_len;            /*     8     4 */

        /* size: 12, cachelines: 1, members: 2 */
        /* last cacheline: 12 bytes */
  };

So we use 12 bytes of ->data[] in struct. However in BPF code running parser
and verdict the user has read capabilities into the data[] array as well. Its
not too problematic, but we should not be exposing internal state to BPF
program. If its really needed then we can use the probe_read() APIs which allow
reading kernel memory. And I don't believe cb[] layer poses any API breakage by
moving this around because programs can't depend on cb[] across layers.

In order to fix another issue with a ctx rewrite we need to stash a temp
variable somewhere. To make this work cleanly this patch builds a cb struct
for sk_skb types called sk_skb_cb struct. Then we can use this consistently
in the strparser, sockmap space. Additionally we can start allowing ->cb[]
write access after this.

Fixes: 604326b41a6fb ("bpf, sockmap: convert to generic sk_msg interface")
Signed-off-by: John Fastabend 
Signed-off-by: Daniel Borkmann 
Tested-by: Jussi Maki 
Reviewed-by: Jakub Sitnicki 
Link: https://lore.kernel.org/bpf/20211103204736.248403-5-john.fastabend@gmail.com
Signed-off-by: Sasha Levin

skmsg: Lose offset info in sk_psock_skb_ingress

2021-11-17T10:04:19+00:00

[ Upstream commit 7303524e04af49a47991e19f895c3b8cdc3796c7 ]

If sockmap enable strparser, there are lose offset info in
sk_psock_skb_ingress(). If the length determined by parse_msg function is not
skb->len, the skb will be converted to sk_msg multiple times, and userspace
app will get the data multiple times.

Fix this by get the offset and length from strp_msg. And as Cong suggested,
add one bit in skb->_sk_redir to distinguish enable or disable strparser.

Fixes: 604326b41a6fb ("bpf, sockmap: convert to generic sk_msg interface")
Signed-off-by: Liu Jian 
Signed-off-by: Daniel Borkmann 
Reviewed-by: Cong Wang 
Acked-by: John Fastabend 
Link: https://lore.kernel.org/bpf/20211029141216.211899-1-liujian56@huawei.com
Signed-off-by: Sasha Levin

net: stream: don't purge sk_error_queue in sk_stream_kill_queues()

2021-11-17T10:04:07+00:00

[ Upstream commit 24bcbe1cc69fa52dc4f7b5b2456678ed464724d8 ]

sk_stream_kill_queues() can be called on close when there are
still outstanding skbs to transmit. Those skbs may try to queue
notifications to the error queue (e.g. timestamps).
If sk_stream_kill_queues() purges the queue without taking
its lock the queue may get corrupted, and skbs leaked.

This shows up as a warning about an rmem leak:

WARNING: CPU: 24 PID: 0 at net/ipv4/af_inet.c:154 inet_sock_destruct+0x...

The leak is always a multiple of 0x300 bytes (the value is in
%rax on my builds, so RAX: 0000000000000300). 0x300 is truesize of
an empty sk_buff. Indeed if we dump the socket state at the time
of the warning the sk_error_queue is often (but not always)
corrupted. The ->next pointer points back at the list head,
but not the ->prev pointer. Indeed we can find the leaked skb
by scanning the kernel memory for something that looks like
an skb with ->sk = socket in question, and ->truesize = 0x300.
The contents of ->cb[] of the skb confirms the suspicion that
it is indeed a timestamp notification (as generated in
__skb_complete_tx_timestamp()).

Removing purging of sk_error_queue should be okay, since
inet_sock_destruct() does it again once all socket refs
are gone. Eric suggests this may cause sockets that go
thru disconnect() to maintain notifications from the
previous incarnations of the socket, but that should be
okay since the race was there anyway, and disconnect()
is not exactly dependable.

Thanks to Jonathan Lemon and Omar Sandoval for help at various
stages of tracing the issue.

Fixes: cb9eff097831 ("net: new user space API for time stamping of incoming and outgoing packets")
Signed-off-by: Jakub Kicinski 
Reviewed-by: Eric Dumazet 
Signed-off-by: David S. Miller 
Signed-off-by: Sasha Levin

net, neigh: Fix NTF_EXT_LEARNED in combination with NTF_USE

2021-11-17T10:04:06+00:00

[ Upstream commit e4400bbf5b15750e1b59bf4722d18d99be60c69f ]

The NTF_EXT_LEARNED neigh flag is usually propagated back to user space
upon dump of the neighbor table. However, when used in combination with
NTF_USE flag this is not the case despite exempting the entry from the
garbage collector. This results in inconsistent state since entries are
typically marked in neigh->flags with NTF_EXT_LEARNED, but here they are
not. Fix it by propagating the creation flag to ___neigh_create().

Before fix:

  # ./ip/ip n replace 192.168.178.30 dev enp5s0 use extern_learn
  # ./ip/ip n
  192.168.178.30 dev enp5s0 lladdr f4:8c:50:5e:71:9a REACHABLE
  [...]

After fix:

  # ./ip/ip n replace 192.168.178.30 dev enp5s0 use extern_learn
  # ./ip/ip n
  192.168.178.30 dev enp5s0 lladdr f4:8c:50:5e:71:9a extern_learn REACHABLE
  [...]

Fixes: 9ce33e46531d ("neighbour: support for NTF_EXT_LEARNED flag")
Signed-off-by: Daniel Borkmann 
Acked-by: Roopa Prabhu 
Signed-off-by: David S. Miller 
Signed-off-by: Sasha Levin

net: net_namespace: Fix undefined member in key_remove_domain()

2021-11-17T10:03:58+00:00

[ Upstream commit aed0826b0cf2e488900ab92193893e803d65c070 ]

The key_domain member in struct net only exists if we define CONFIG_KEYS.
So we should add the define when we used key_domain.

Fixes: 9b242610514f ("keys: Network namespace domain tag")
Signed-off-by: Yajun Deng 
Signed-off-by: David S. Miller 
Signed-off-by: Sasha Levin

net-sysfs: try not to restart the syscall if it will fail eventually

2021-11-17T10:03:49+00:00

[ Upstream commit 146e5e733310379f51924111068f08a3af0db830 ]

Due to deadlocks in the networking subsystem spotted 12 years ago[1],
a workaround was put in place[2] to avoid taking the rtnl lock when it
was not available and restarting the syscall (back to VFS, letting
userspace spin). The following construction is found a lot in the net
sysfs and sysctl code:

  if (!rtnl_trylock())
          return restart_syscall();

This can be problematic when multiple userspace threads use such
interfaces in a short period, making them to spin a lot. This happens
for example when adding and moving virtual interfaces: userspace
programs listening on events, such as systemd-udevd and NetworkManager,
do trigger actions reading files in sysfs. It gets worse when a lot of
virtual interfaces are created concurrently, say when creating
containers at boot time.

Returning early without hitting the above pattern when the syscall will
fail eventually does make things better. While it is not a fix for the
issue, it does ease things.

[1] https://lore.kernel.org/netdev/49A4D5D5.5090602@trash.net/
    https://lore.kernel.org/netdev/m14oyhis31.fsf@fess.ebiederm.org/
    and https://lore.kernel.org/netdev/20090226084924.16cb3e08@nehalam/
[2] Rightfully, those deadlocks are *hard* to solve.

Signed-off-by: Antoine Tenart 
Reviewed-by: Paolo Abeni 
Signed-off-by: David S. Miller 
Signed-off-by: Sasha Levin

net: sched: update default qdisc visibility after Tx queue cnt changes

2021-11-17T10:03:44+00:00

[ Upstream commit 1e080f17750d1083e8a32f7b350584ae1cd7ff20 ]

mq / mqprio make the default child qdiscs visible. They only do
so for the qdiscs which are within real_num_tx_queues when the
device is registered. Depending on order of calls in the driver,
or if user space changes config via ethtool -L the number of
qdiscs visible under tc qdisc show will differ from the number
of queues. This is confusing to users and potentially to system
configuration scripts which try to make sure qdiscs have the
right parameters.

Add a new Qdisc_ops callback and make relevant qdiscs TTRT.

Note that this uncovers the "shortcut" created by
commit 1f27cde313d7 ("net: sched: use pfifo_fast for non real queues")
The default child qdiscs beyond initial real_num_tx are always
pfifo_fast, no matter what the sysfs setting is. Fixing this
gets a little tricky because we'd need to keep a reference
on whatever the default qdisc was at the time of creation.
In practice this is likely an non-issue the qdiscs likely have
to be configured to non-default settings, so whatever user space
is doing such configuration can replace the pfifos... now that
it will see them.

Reported-by: Matthew Massey 
Reviewed-by: Dave Taht 
Signed-off-by: Jakub Kicinski 
Signed-off-by: David S. Miller 
Signed-off-by: Sasha Levin