Fedor Pchelkin pchelkin@ispras.ru wrote:
Now that nullfunc skbs are recycled in a separate work item in the driver, the following race during initialization and processing of those skbs might lead to noticeable bugs:
Waiting thread Completing thread
rtw89_core_send_nullfunc() rtw89_core_tx_write_link() ... rtw89_pci_txwd_submit() skb_data->wait = NULL /* add skb to the queue */ skb_queue_tail(&txwd->queue, skb) rtw89_pci_napi_poll() ... rtw89_pci_release_txwd_skb() /* get skb from the queue */ skb_unlink(skb, &txwd->queue) rtw89_pci_tx_status() rtw89_core_tx_wait_complete() /* use incorrect skb_data->wait */ rtw89_core_tx_kick_off_and_wait() /* assign skb_data->wait but too late */
How will we receive tx completion before TX kick off? (see the original code below)
[...]
diff --git a/drivers/net/wireless/realtek/rtw89/core.c b/drivers/net/wireless/realtek/rtw89/core.c index 438930b65631..1efe4bb09262 100644 --- a/drivers/net/wireless/realtek/rtw89/core.c +++ b/drivers/net/wireless/realtek/rtw89/core.c @@ -1094,22 +1094,13 @@ int rtw89_core_tx_kick_off_and_wait(struct rtw89_dev *rtwdev, struct sk_buff *sk int qsel, unsigned int timeout) { struct rtw89_tx_skb_data *skb_data = RTW89_TX_SKB_CB(skb);
struct rtw89_tx_wait_info *wait;
struct rtw89_tx_wait_info *wait = wiphy_dereference(rtwdev->hw->wiphy,
skb_data->wait);
Can't we just pass 'wait' by function argument?
unsigned long time_left; int ret = 0; lockdep_assert_wiphy(rtwdev->hw->wiphy);
wait = kzalloc(sizeof(*wait), GFP_KERNEL);
if (!wait) {
rtw89_core_tx_kick_off(rtwdev, qsel);
return 0;
}
init_completion(&wait->completion);
wait->skb = skb;
rcu_assign_pointer(skb_data->wait, wait);
Here, original code prepares completion before TX kick off. How it could be a problem? Do I miss something?
rtw89_core_tx_kick_off(rtwdev, qsel); time_left = wait_for_completion_timeout(&wait->completion, msecs_to_jiffies(timeout));