packet: avoid re-registering fanout hook after device unregister
Subsystems (3)
PATCH v2 1/1 packet: avoid re-registering fanout hook after device unregister
packet_set_ring() temporarily detaches a socket from packet delivery while reconfiguring its ring. It records the previous running state, clears po->num, unregisters the protocol hook when needed, drops po->bind_lock, and later restores po->num and re-registers the hook from the saved was_running value.
That unlocked window can race with NETDEV_UNREGISTER. The notifier can observe the socket as not running, skip __unregister_prot_hook(), and invalidate the per-socket binding by setting po->ifindex to -1 and clearing po->prot_hook.dev. A one-member fanout group can still retain its shared fanout hook device pointer. When packet_set_ring() resumes, re-registering solely from the stale was_running state can re-add the fanout hook after the device has been unregistered.
Treat po->ifindex == -1 as an invalidated binding after reacquiring po->bind_lock. Restore po->num as before, but do not re-register the hook if device unregister already detached the socket.
---
Changes in v2:
- Add Fixes and Reported-by tags.
- Fix the incorrect Signed-off-by tag from v1.
Trail of Bits has a PoC that achieves local privilege escalation using this bug on a custom kernel config with CONFIG_LIST_HARDENED disabled, which can be shared further if needed.
net/packet/af_packet.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-)
net/packet/af_packet.c +5 -1
@@ -4561,7 +4561,11 @@ static int packet_set_ring(struct sock *sk, union tpacket_req_u *req_u,spin_lock (&po->bind_lock);WRITE_ONCE (po->num, num); -if (was_running) +/* +* NETDEV_UNREGISTER may have invalidated the binding while bind_lock +* was dropped above. Do not re-add a fanout hook to a dead device. +*/ +if (was_running &&READ_ONCE (po->ifindex) !=-1 )register_prot_hook (sk);spin_unlock (&po->bind_lock);
Signature · 1 line
2.43.0
Quoted 14 lines · David Lee
packet_set_ring()temporarily detaches a socket from packet delivery while reconfiguring its ring. It records the previous running state, clearspo->num, unregisters the protocol hook when needed, dropspo->bind_lock, and later restorespo->numand re-registers the hook from the savedwas_runningvalue.That unlocked window can race with
NETDEV_UNREGISTER. The notifier can observe the socket as not running, skip__unregister_prot_hook(), and invalidate the per-socket binding by settingpo->ifindexto -1 and clearingpo->prot_hook.dev. A one-member fanout group can still retain its shared fanout hook device pointer. Whenpacket_set_ring()resumes, re-registering solely from the stalewas_runningstate can re-add the fanout hook after the device has been unregistered.
Quoted 7 lines
Treat
po->ifindex== -1 as an invalidated binding after reacquiringpo->bind_lock. Restorepo->numas before, but do not re-register the hook if device unregister already detached the socket.Fixes dc99f600698d ("packet: Add fanout support.")Reported-by David Lee <[email protected]>Signed-off-by David Lee <[email protected]>
My suggestion of a Reported-by was becaused the Signed-off-by was from a different author. Both tags from the same author is generally discouraged.
Perhaps also add a Link to the discussion in the first patch, which adds some more rationale. The crux to me is that ifindex -1 is not just the default for detached sockets, it is a special invalidated condition.
With those asides
Quoted 31 lines
Assisted-by Codex:gpt-5---
Changes in v2:
- Add Fixes and Reported-by tags.
- Fix the incorrect Signed-off-by tag from v1.Trail of Bits has a PoC that achieves local privilege escalation using this bug on a custom kernel config with
CONFIG_LIST_HARDENEDdisabled, which can be shared further if needed.net/packet/af_packet.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-)net/packet/af_packet.c +5
-1@@ -4561,7 +4561,11 @@ static int packet_set_ring(struct sock *sk, union tpacket_req_u *req_u,spin_lock (&po->bind_lock);WRITE_ONCE (po->num, num); -if (was_running) +/* +* NETDEV_UNREGISTER may have invalidated the binding while bind_lock +* was dropped above. Do not re-add a fanout hook to a dead device. +*/ +if (was_running &&READ_ONCE (po->ifindex) !=-1 )register_prot_hook (sk);spin_unlock (&po->bind_lock);Signature · 1 line
2.43.0