When device is running and the interface status is changed, the gpf issue
is triggered. The problem triggering process is as follows:
Thread A: Thread B
ieee80211_runtime_change_iftype() process_one_work()
... ...
ieee80211_do_stop() ...
... ...
sdata->bss = NULL ...
... ieee80211_subif_start_xmit()
ieee80211_multicast_to_unicast
//!sdata->bss->multicast_to_unicast
cause gpf issue
When the interface status is changed, the sending queue continues to send
packets. After the bss is set to NULL, the bss is accessed. As a result,
this causes a general-protection-fault issue.
The following is the stack information:
general protection fault, probably for non-canonical address
0xdffffc000000002f: 0000 [#1] PREEMPT SMP KASAN
KASAN: null-ptr-deref in range [0x0000000000000178-0x000000000000017f]
Workqueue: mld mld_ifc_work
RIP: 0010:ieee80211_subif_start_xmit+0x25b/0x1310
Call Trace:
<TASK>
dev_hard_start_xmit+0x1be/0x990
__dev_queue_xmit+0x2c9a/0x3b60
ip6_finish_output2+0xf92/0x1520
ip6_finish_output+0x6af/0x11e0
ip6_output+0x1ed/0x540
mld_sendpack+0xa09/0xe70
mld_ifc_work+0x71c/0xdb0
process_one_work+0x9bf/0x1710
worker_thread+0x665/0x1080
kthread+0x2e4/0x3a0
ret_from_fork+0x1f/0x30
</TASK>
Fixes: f856373e2f31 ("wifi: mac80211: do not wake queues on a vif that is being stopped")
Reported-by: [email protected]
Signed-off-by: Zhengchao Shao <[email protected]>
---
v3: revise format
v2: replace the netif call with test_bit()
---
net/mac80211/tx.c | 5 +++++
1 file changed, 5 insertions(+)
diff --git a/net/mac80211/tx.c b/net/mac80211/tx.c
index a364148149f9..874f2a4d831d 100644
--- a/net/mac80211/tx.c
+++ b/net/mac80211/tx.c
@@ -4418,6 +4418,11 @@ netdev_tx_t ieee80211_subif_start_xmit(struct sk_buff *skb,
if (likely(!is_multicast_ether_addr(eth->h_dest)))
goto normal;
+ if (unlikely(!ieee80211_sdata_running(sdata))) {
+ kfree_skb(skb);
+ return NETDEV_TX_OK;
+ }
+
if (unlikely(ieee80211_multicast_to_unicast(skb, dev))) {
struct sk_buff_head queue;
--
2.17.1