HELP! When I run my IP kernel, my FPGA will disconnect. Whatever it is ethernet port or the way with USB port

Hello everyone:
I made a CNN IP kernel with vitis HLS, and let it into bitstream file at last. But there is a question I never met befor: When I run my IP Kernel with jupyter, FPGA will lose connection and the board looks no change(for example: led goes red or give me some tips like warnings on the screen).
You must reboot FPGA if you wanna connect again.
And then I changed the way to use USB port to connect. This time ,it gives me some messages on the screen:

[ 268.315970] rcu: INFO: rcu_sched detected stalls on CPUs/tasks:
[ 268.321895] rcu: All QSes seen, last rcu_sched kthread activity 5252 (4294959 375-4294954123), jiffies_till_next_fqs=1, root ->qsmask 0x0
[ 268.334273] rcu: rcu_sched kthread starved for 5252 jiffies! g16249 f0x2 RCU_ GP_WAIT_FQS(5) ->state=0x200 ->cpu=1
[ 268.344521] rcu: RCU grace-period kthread stack dump:
[ 331.335963] rcu: INFO: rcu_sched detected stalls on CPUs/tasks:
[ 331.341876] rcu: All QSes seen, last rcu_sched kthread activity 21007 (4294975130-4294954123), jiffies_till_next_fqs=1, root ->qsmask 0x0
[ 331.354316] rcu: rcu_sched kthread starved for 21007 jiffies! g16249 f0x2 RCU_GP_WAIT_FQS(5) ->state=0x200 ->cpu=1
[ 331.364645] rcu: RCU grace-period kthread stack dump:

This messages are given before it disconnect, just once. When I tried again to run the IP with connecting in USB, these messagers do not appear anymore.
If you met question like this ever or knew what happened, please tell me. Thanks

Hi there,

What you’re describing sounds like a board hang or freeze (you can use these keywords to look up similar issues). We typically see board hangs when you try write or access a memory you’re not supposed to. If you’re using MMIO to talk to your IP block, I recommend making sure that the addresses in the vivado address editor are assigned and match the addresses you’re using in your python code.

Nobody will be able to give you much advice without seeing your vivado design, what code you ran to reproduce the issue or what board you’re using.

Thanks
Shawn

4 Likes

Thanks for your reply very much. Yesterday, I resolved this question.
I found there is a unconnect IO named “ap_ce” on my IP before, and I don`t know where it need to connect.
And then, I compared other IP I made, I found that caused unconnect IO “ap_ce” exist is I chosed a selection in vitis HLS :

Solution -> solution settings -> General -> config_interface -> clock_enable.

When I cancled this choice, then the IO “ap_ce” is gone and the IP could run normally.
I don`t know why, at the beginning I just consider that is a selection of autoconnection_clock_signal.
Thank you!

1 Like

Thanks for reporting back with a solution!