wifi: ath11k: Fix deadlock during WoWLAN suspend
We are seeing system hangs during WoWLAN suspend, and get below two stacks: Stack1: [ffffb02cc1557b20] __schedule at ffffffff8bb10860 [ffffb02cc1557ba8] schedule at ffffffff8bb10f24 [ffffb02cc1557bb8] schedule_timeout at ffffffff8bb16d88 [ffffb02cc1557c30] wait_for_completion at ffffffff8bb11778 [ffffb02cc1557c78] __flush_work at ffffffff8b0b30cd [ffffb02cc1557cf0] __cancel_work_timer at ffffffff8b0b33ad [ffffb02cc1557d60] ath11k_mac_drain_tx at ffffffffc0c1f0ca [ath11k] [ffffb02cc1557d70] ath11k_wow_op_suspend at ffffffffc0c5201e [ath11k] [ffffb02cc1557da8] __ieee80211_suspend at ffffffffc11e2bd3 [mac80211] [ffffb02cc1557dd8] wiphy_suspend at ffffffffc0f901ac [cfg80211] [ffffb02cc1557e08] dpm_run_callback at ffffffff8b75118a [ffffb02cc1557e38] __device_suspend at ffffffff8b751630 [ffffb02cc1557e70] async_suspend at ffffffff8b7519ea [ffffb02cc1557e88] async_run_entry_fn at ffffffff8b0bf4ce [ffffb02cc1557ea8] process_one_work at ffffffff8b0b1a24 [ffffb02cc1557ee0] worker_thread at ffffffff8b0b1c4a [ffffb02cc1557f18] kthread at ffffffff8b0b9cb8 [ffffb02cc1557f50] ret_from_fork at ffffffff8b001d32 Stack2: [ffffb02cc00b7d18] __schedule at ffffffff8bb10860 [ffffb02cc00b7da0] schedule at ffffffff8bb10f24 [ffffb02cc00b7db0] schedule_preempt_disabled at ffffffff8bb112b4 [ffffb02cc00b7db8] __mutex_lock at ffffffff8bb127ea [ffffb02cc00b7e38] ath11k_mgmt_over_wmi_tx_work at ffffffffc0c1aa44 [ath11k] [ffffb02cc00b7ea8] process_one_work at ffffffff8b0b1a24 [ffffb02cc00b7ee0] worker_thread at ffffffff8b0b1c4a [ffffb02cc00b7f18] kthread at ffffffff8b0b9cb8 [ffffb02cc00b7f50] ret_from_fork at ffffffff8b001d32 From the first stack, ath11k_mac_drain_tx calls cancel_work_sync(&ar->wmi_mgmt_tx_work) and waits all packets to be sent out or dropped. However, we find from Stack2 that this work item is blocked because ar->conf_mutex is already held by ath11k_wow_op_suspend. Fix this issue by moving ath11k_mac_wait_tx_complete to the start of ath11k_wow_op_suspend where ar->conf_mutex has not been acquired. And this change also makes the logic in ath11k_wow_op_suspend match the logic in ath11k_mac_op_start and ath11k_mac_op_stop. Tested-on: WCN6855 hw2.0 PCI WLAN.HSP.1.1-03125-QCAHSPSWPL_V1_V2_SILICONZ_LITE-3 Signed-off-by: Baochen Qiang <quic_bqiang@quicinc.com> Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com> Link: https://lore.kernel.org/r/20220919021435.2459-1-quic_bqiang@quicinc.com
Please register or sign in to comment