[][HIGH][kernel][mt7988][eth][Fix issue with rcu_sched self-detected stall for the USXGMII out-of-bnad link down recovery handler]
[Description]
Fix issue with rcu_sched self-detected stall for the USXGMII out-of-band
link down recovery handler.
When phylink calls pcs_link_up, the USXGMII driver will reconfigure PCS
to ensure RX performance. However, during the PCS reconfiguration, the
CPU may jump to the ISR due to an interrupt and execute timer-related
procedures (USXGMII link poll). Subsequently, the CPU call spin_lock()
once more in mtk_usxgmii_pcs_config(), leading to a CPU hang.
Therefore, we use spin_lock_irqsave to ensure that the PCS configuration
procedures will be uninterrupted.
Without this patch, the kernel might encounter the rcu_sched
self-detected stall issue when executing /etc/init.d/network restart.
[Release-log]
N/A
Change-Id: If53bed057d378c314a61297c29c0e5d365c4cb50
Reviewed-on: https://gerrit.mediatek.inc/c/openwrt/feeds/mtk_openwrt_feeds/+/9169707
1 file changed