linux.git/net/sched, branch v6.10

net/sched: Fix UAF when resolving a clash

2024-07-11T10:07:54+00:00

KASAN reports the following UAF:

 BUG: KASAN: slab-use-after-free in tcf_ct_flow_table_process_conn+0x12b/0x380 [act_ct]
 Read of size 1 at addr ffff888c07603600 by task handler130/6469

 Call Trace:
  
  dump_stack_lvl+0x48/0x70
  print_address_description.constprop.0+0x33/0x3d0
  print_report+0xc0/0x2b0
  kasan_report+0xd0/0x120
  __asan_load1+0x6c/0x80
  tcf_ct_flow_table_process_conn+0x12b/0x380 [act_ct]
  tcf_ct_act+0x886/0x1350 [act_ct]
  tcf_action_exec+0xf8/0x1f0
  fl_classify+0x355/0x360 [cls_flower]
  __tcf_classify+0x1fd/0x330
  tcf_classify+0x21c/0x3c0
  sch_handle_ingress.constprop.0+0x2c5/0x500
  __netif_receive_skb_core.constprop.0+0xb25/0x1510
  __netif_receive_skb_list_core+0x220/0x4c0
  netif_receive_skb_list_internal+0x446/0x620
  napi_complete_done+0x157/0x3d0
  gro_cell_poll+0xcf/0x100
  __napi_poll+0x65/0x310
  net_rx_action+0x30c/0x5c0
  __do_softirq+0x14f/0x491
  __irq_exit_rcu+0x82/0xc0
  irq_exit_rcu+0xe/0x20
  common_interrupt+0xa1/0xb0
  
  
  asm_common_interrupt+0x27/0x40

 Allocated by task 6469:
  kasan_save_stack+0x38/0x70
  kasan_set_track+0x25/0x40
  kasan_save_alloc_info+0x1e/0x40
  __kasan_krealloc+0x133/0x190
  krealloc+0xaa/0x130
  nf_ct_ext_add+0xed/0x230 [nf_conntrack]
  tcf_ct_act+0x1095/0x1350 [act_ct]
  tcf_action_exec+0xf8/0x1f0
  fl_classify+0x355/0x360 [cls_flower]
  __tcf_classify+0x1fd/0x330
  tcf_classify+0x21c/0x3c0
  sch_handle_ingress.constprop.0+0x2c5/0x500
  __netif_receive_skb_core.constprop.0+0xb25/0x1510
  __netif_receive_skb_list_core+0x220/0x4c0
  netif_receive_skb_list_internal+0x446/0x620
  napi_complete_done+0x157/0x3d0
  gro_cell_poll+0xcf/0x100
  __napi_poll+0x65/0x310
  net_rx_action+0x30c/0x5c0
  __do_softirq+0x14f/0x491

 Freed by task 6469:
  kasan_save_stack+0x38/0x70
  kasan_set_track+0x25/0x40
  kasan_save_free_info+0x2b/0x60
  ____kasan_slab_free+0x180/0x1f0
  __kasan_slab_free+0x12/0x30
  slab_free_freelist_hook+0xd2/0x1a0
  __kmem_cache_free+0x1a2/0x2f0
  kfree+0x78/0x120
  nf_conntrack_free+0x74/0x130 [nf_conntrack]
  nf_ct_destroy+0xb2/0x140 [nf_conntrack]
  __nf_ct_resolve_clash+0x529/0x5d0 [nf_conntrack]
  nf_ct_resolve_clash+0xf6/0x490 [nf_conntrack]
  __nf_conntrack_confirm+0x2c6/0x770 [nf_conntrack]
  tcf_ct_act+0x12ad/0x1350 [act_ct]
  tcf_action_exec+0xf8/0x1f0
  fl_classify+0x355/0x360 [cls_flower]
  __tcf_classify+0x1fd/0x330
  tcf_classify+0x21c/0x3c0
  sch_handle_ingress.constprop.0+0x2c5/0x500
  __netif_receive_skb_core.constprop.0+0xb25/0x1510
  __netif_receive_skb_list_core+0x220/0x4c0
  netif_receive_skb_list_internal+0x446/0x620
  napi_complete_done+0x157/0x3d0
  gro_cell_poll+0xcf/0x100
  __napi_poll+0x65/0x310
  net_rx_action+0x30c/0x5c0
  __do_softirq+0x14f/0x491

The ct may be dropped if a clash has been resolved but is still passed to
the tcf_ct_flow_table_process_conn function for further usage. This issue
can be fixed by retrieving ct from skb again after confirming conntrack.

Fixes: 0cc254e5aa37 ("net/sched: act_ct: Offload connections with commit action")
Co-developed-by: Gerald Yang 
Signed-off-by: Gerald Yang 
Signed-off-by: Chengen Du 
Link: https://patch.msgid.link/20240710053747.13223-1-chengen.du@canonical.com
Signed-off-by: Paolo Abeni

bpf: Fix too early release of tcx_entry

2024-07-08T21:07:31+00:00

Pedro Pinto and later independently also Hyunwoo Kim and Wongi Lee reported
an issue that the tcx_entry can be released too early leading to a use
after free (UAF) when an active old-style ingress or clsact qdisc with a
shared tc block is later replaced by another ingress or clsact instance.

Essentially, the sequence to trigger the UAF (one example) can be as follows:

  1. A network namespace is created
  2. An ingress qdisc is created. This allocates a tcx_entry, and
     &tcx_entry->miniq is stored in the qdisc's miniqp->p_miniq. At the
     same time, a tcf block with index 1 is created.
  3. chain0 is attached to the tcf block. chain0 must be connected to
     the block linked to the ingress qdisc to later reach the function
     tcf_chain0_head_change_cb_del() which triggers the UAF.
  4. Create and graft a clsact qdisc. This causes the ingress qdisc
     created in step 1 to be removed, thus freeing the previously linked
     tcx_entry:

     rtnetlink_rcv_msg()
       => tc_modify_qdisc()
         => qdisc_create()
           => clsact_init() [a]
         => qdisc_graft()
           => qdisc_destroy()
             => __qdisc_destroy()
               => ingress_destroy() [b]
                 => tcx_entry_free()
                   => kfree_rcu() // tcx_entry freed

  5. Finally, the network namespace is closed. This registers the
     cleanup_net worker, and during the process of releasing the
     remaining clsact qdisc, it accesses the tcx_entry that was
     already freed in step 4, causing the UAF to occur:

     cleanup_net()
       => ops_exit_list()
         => default_device_exit_batch()
           => unregister_netdevice_many()
             => unregister_netdevice_many_notify()
               => dev_shutdown()
                 => qdisc_put()
                   => clsact_destroy() [c]
                     => tcf_block_put_ext()
                       => tcf_chain0_head_change_cb_del()
                         => tcf_chain_head_change_item()
                           => clsact_chain_head_change()
                             => mini_qdisc_pair_swap() // UAF

There are also other variants, the gist is to add an ingress (or clsact)
qdisc with a specific shared block, then to replace that qdisc, waiting
for the tcx_entry kfree_rcu() to be executed and subsequently accessing
the current active qdisc's miniq one way or another.

The correct fix is to turn the miniq_active boolean into a counter. What
can be observed, at step 2 above, the counter transitions from 0->1, at
step [a] from 1->2 (in order for the miniq object to remain active during
the replacement), then in [b] from 2->1 and finally [c] 1->0 with the
eventual release. The reference counter in general ranges from [0,2] and
it does not need to be atomic since all access to the counter is protected
by the rtnl mutex. With this in place, there is no longer a UAF happening
and the tcx_entry is freed at the correct time.

Fixes: e420bed02507 ("bpf: Add fd-based tcx multi-prog infra with link support")
Reported-by: Pedro Pinto 
Co-developed-by: Pedro Pinto 
Signed-off-by: Pedro Pinto 
Signed-off-by: Daniel Borkmann 
Cc: Hyunwoo Kim 
Cc: Wongi Lee 
Cc: Martin KaFai Lau 
Link: https://lore.kernel.org/r/20240708133130.11609-1-daniel@iogearbox.net
Signed-off-by: Martin KaFai Lau

sched: act_ct: add netns into the key of tcf_ct_flow_table

2024-06-18T13:24:24+00:00

zones_ht is a global hashtable for flow_table with zone as key. However,
it does not consider netns when getting a flow_table from zones_ht in
tcf_ct_init(), and it means an act_ct action in netns A may get a
flow_table that belongs to netns B if it has the same zone value.

In Shuang's test with the TOPO:

  tcf2_c <---> tcf2_sw1 <---> tcf2_sw2 <---> tcf2_s

tcf2_sw1 and tcf2_sw2 saw the same flow and used the same flow table,
which caused their ct entries entering unexpected states and the
TCP connection not able to end normally.

This patch fixes the issue simply by adding netns into the key of
tcf_ct_flow_table so that an act_ct action gets a flow_table that
belongs to its own netns in tcf_ct_init().

Note that for easy coding we don't use tcf_ct_flow_table.nf_ft.net,
as the ct_ft is initialized after inserting it to the hashtable in
tcf_ct_flow_table_get() and also it requires to implement several
functions in rhashtable_params including hashfn, obj_hashfn and
obj_cmpfn.

Fixes: 64ff70b80fd4 ("net/sched: act_ct: Offload established connections to flow table")
Reported-by: Shuang Li 
Signed-off-by: Xin Long 
Reviewed-by: Simon Horman 
Link: https://lore.kernel.org/r/1db5b6cc6902c5fc6f8c6cbd85494a2008087be5.1718488050.git.lucien.xin@gmail.com
Signed-off-by: Paolo Abeni

net/sched: act_api: fix possible infinite loop in tcf_idr_check_alloc()

2024-06-18T10:28:01+00:00

syzbot found hanging tasks waiting on rtnl_lock [1]

A reproducer is available in the syzbot bug.

When a request to add multiple actions with the same index is sent, the
second request will block forever on the first request. This holds
rtnl_lock, and causes tasks to hang.

Return -EAGAIN to prevent infinite looping, while keeping documented
behavior.

[1]

INFO: task kworker/1:0:5088 blocked for more than 143 seconds.
Not tainted 6.9.0-rc4-syzkaller-00173-g3cdb45594619 #0
"echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
task:kworker/1:0 state:D stack:23744 pid:5088 tgid:5088 ppid:2 flags:0x00004000
Workqueue: events_power_efficient reg_check_chans_work
Call Trace:

context_switch kernel/sched/core.c:5409 [inline]
__schedule+0xf15/0x5d00 kernel/sched/core.c:6746
__schedule_loop kernel/sched/core.c:6823 [inline]
schedule+0xe7/0x350 kernel/sched/core.c:6838
schedule_preempt_disabled+0x13/0x30 kernel/sched/core.c:6895
__mutex_lock_common kernel/locking/mutex.c:684 [inline]
__mutex_lock+0x5b8/0x9c0 kernel/locking/mutex.c:752
wiphy_lock include/net/cfg80211.h:5953 [inline]
reg_leave_invalid_chans net/wireless/reg.c:2466 [inline]
reg_check_chans_work+0x10a/0x10e0 net/wireless/reg.c:2481

Fixes: 0190c1d452a9 ("net: sched: atomically check-allocate action")
Reported-by: syzbot+b87c222546179f4513a7@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=b87c222546179f4513a7
Signed-off-by: David Ruth 
Reviewed-by: Jamal Hadi Salim 
Link: https://lore.kernel.org/r/20240614190326.1349786-1-druth@chromium.org
Signed-off-by: Paolo Abeni

net/sched: initialize noop_qdisc owner

2024-06-11T02:36:49+00:00

When the noop_qdisc owner isn't initialized, then it will be 0,
so packets will erroneously be regarded as having been subject
to recursion as long as only CPU 0 queues them. For non-SMP,
that's all packets, of course. This causes a change in what's
reported to userspace, normally noop_qdisc would drop packets
silently, but with this change the syscall returns -ENOBUFS if
RECVERR is also set on the socket.

Fix this by initializing the owner field to -1, just like it
would be for dynamically allocated qdiscs by qdisc_alloc().

Fixes: 0f022d32c3ec ("net/sched: Fix mirred deadlock on device recursion")
Signed-off-by: Johannes Berg 
Reviewed-by: Eric Dumazet 
Link: https://lore.kernel.org/r/20240607175340.786bfb938803.I493bf8422e36be4454c08880a8d3703cea8e421a@changeid
Signed-off-by: Jakub Kicinski

net/sched: taprio: always validate TCA_TAPRIO_ATTR_PRIOMAP

2024-06-05T22:54:51+00:00

If one TCA_TAPRIO_ATTR_PRIOMAP attribute has been provided,
taprio_parse_mqprio_opt() must validate it, or userspace
can inject arbitrary data to the kernel, the second time
taprio_change() is called.

First call (with valid attributes) sets dev->num_tc
to a non zero value.

Second call (with arbitrary mqprio attributes)
returns early from taprio_parse_mqprio_opt()
and bad things can happen.

Fixes: a3d43c0d56f1 ("taprio: Add support adding an admin schedule")
Reported-by: Noam Rathaus 
Signed-off-by: Eric Dumazet 
Acked-by: Vinicius Costa Gomes 
Reviewed-by: Vladimir Oltean 
Link: https://lore.kernel.org/r/20240604181511.769870-1-edumazet@google.com
Signed-off-by: Jakub Kicinski

net: sched: sch_multiq: fix possible OOB write in multiq_tune()

2024-06-05T09:50:19+00:00

q->bands will be assigned to qopt->bands to execute subsequent code logic
after kmalloc. So the old q->bands should not be used in kmalloc.
Otherwise, an out-of-bounds write will occur.

Fixes: c2999f7fb05b ("net: sched: multiq: don't call qdisc_put() while holding tree lock")
Signed-off-by: Hangyu Hua 
Acked-by: Cong Wang 
Signed-off-by: David S. Miller

net/sched: taprio: extend minimum interval restriction to entire cycle too

2024-05-29T02:46:41+00:00

It is possible for syzbot to side-step the restriction imposed by the
blamed commit in the Fixes: tag, because the taprio UAPI permits a
cycle-time different from (and potentially shorter than) the sum of
entry intervals.

We need one more restriction, which is that the cycle time itself must
be larger than N * ETH_ZLEN bit times, where N is the number of schedule
entries. This restriction needs to apply regardless of whether the cycle
time came from the user or was the implicit, auto-calculated value, so
we move the existing "cycle == 0" check outside the "if "(!new->cycle_time)"
branch. This way covers both conditions and scenarios.

Add a selftest which illustrates the issue triggered by syzbot.

Fixes: b5b73b26b3ca ("taprio: Fix allowing too small intervals")
Reported-by: syzbot+a7d2b1d5d1af83035567@syzkaller.appspotmail.com
Closes: https://lore.kernel.org/netdev/0000000000007d66bc06196e7c66@google.com/
Signed-off-by: Vladimir Oltean 
Link: https://lore.kernel.org/r/20240527153955.553333-2-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski

net/sched: taprio: make q->picos_per_byte available to fill_sched_entry()

2024-05-29T02:46:41+00:00

In commit b5b73b26b3ca ("taprio: Fix allowing too small intervals"), a
comparison of user input against length_to_duration(q, ETH_ZLEN) was
introduced, to avoid RCU stalls due to frequent hrtimers.

The implementation of length_to_duration() depends on q->picos_per_byte
being set for the link speed. The blamed commit in the Fixes: tag has
moved this too late, so the checks introduced above are ineffective.
The q->picos_per_byte is zero at parse_taprio_schedule() ->
parse_sched_list() -> parse_sched_entry() -> fill_sched_entry() time.

Move the taprio_set_picos_per_byte() call as one of the first things in
taprio_change(), before the bulk of the netlink attribute parsing is
done. That's because it is needed there.

Add a selftest to make sure the issue doesn't get reintroduced.

Fixes: 09dbdf28f9f9 ("net/sched: taprio: fix calculation of maximum gate durations")
Signed-off-by: Vladimir Oltean 
Reviewed-by: Eric Dumazet 
Link: https://lore.kernel.org/r/20240527153955.553333-1-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski

net/sched: adjust device watchdog timer to detect stopped queue at right time

2024-05-10T03:24:13+00:00

Applications are sensitive to long network latency, particularly
heartbeat monitoring ones. Longer the tx timeout recovery higher the
risk with such applications on a production machines. This patch
remedies, yet honoring device set tx timeout.

Modify watchdog next timeout to be shorter than the device specified.
Compute the next timeout be equal to device watchdog timeout less the
how long ago queue stop had been done. At next watchdog timeout tx
timeout handler is called into if still in stopped state. Either called
or not called, restore the watchdog timeout back to device specified.

Signed-off-by: Praveen Kumar Kannoju 
Link: https://lore.kernel.org/r/20240508133617.4424-1-praveen.kannoju@oracle.com
Signed-off-by: Jakub Kicinski