Another issue I'm facing and happens very often with 2 PCI cards connected. It happens less often with only one card connected. When this happens device is unresponsive and the only way to recover is to cut power. I got this log on UART.
[ 1613.607084] rcu: 1-...0: (3 ticks this GP) idle=7b2/1/0x4000000000000000 softirq=11591/11593 fqs=1050
[ 1613.616521] rcu: 3-...0: (3 ticks this GP) idle=602/1/0x4000000000000000 softirq=11114/11116 fqs=1050
[ 1613.625956] (detected by 0, t=2102 jiffies, g=14753, q=42)
[ 1613.631554] Task dump for CPU 1:
[ 1613.634796] task:kworker/1:1 state:R running task stack: 0 pid: 5 628 ppid: 2 flags:0x0000000a
[ 1613.644774] Workqueue: events dbs_work_handler
[ 1613.649239] Call trace:
[ 1613.651700] __switch_to+0x9c/0xfc
[ 1613.655119] process_one_work+0x1f0/0x380
[ 1613.659148] worker_thread+0x70/0x4c4
[ 1613.662828] kthread+0x120/0x124
[ 1613.666071] ret_from_fork+0x10/0x20
[ 1613.669661] Task dump for CPU 3:
[ 1613.672903] task:kworker/3:6 state:R running task stack: 0 pid: 759 ppid: 2 flags:0x0000000a
[ 1613.682874] Workqueue: events_freezable_power_ thermal_zone_device_check
[ 1613.689604] Call trace:
[ 1613.692062] __switch_to+0x9c/0xfc
[ 1613.695480] 0xffffff8100b46900
[ 1676.647648] rcu: INFO: rcu_sched detected stalls on CPUs/tasks:
[ 1676.653606] rcu: 1-...0: (3 ticks this GP) idle=7b2/1/0x4000000000000000 softirq=11591/11593 fqs=4197
[ 1676.663043] rcu: 3-...0: (3 ticks this GP) idle=602/1/0x4000000000000000 softirq=11114/11116 fqs=4197
[ 1676.672478] (detected by 0, t=8407 jiffies, g=14753, q=220)
[ 1676.678162] Task dump for CPU 1:
[ 1676.681404] task:kworker/1:1 state:R running task stack: 0 pid: 5 628 ppid: 2 flags:0x0000000a
[ 1676.691382] Workqueue: events dbs_work_handler
[ 1676.695848] Call trace:
[ 1676.698309] __switch_to+0x9c/0xfc
[ 1676.701728] process_one_work+0x1f0/0x380
[ 1676.705756] worker_thread+0x70/0x4c4
[ 1676.709435] kthread+0x120/0x124
[ 1676.712678] ret_from_fork+0x10/0x20
[ 1676.716267] Task dump for CPU 3:
[ 1676.719508] task:kworker/3:6 state:R running task stack: 0 pid: 759 ppid: 2 flags:0x0000000a
[ 1676.729479] Workqueue: events_freezable_power_ thermal_zone_device_check
[ 1676.736210] Call trace:
[ 1676.738668] __switch_to+0x9c/0xfc
[ 1676.742086] 0xffffff8100b46900