Hi,
If have created an cape for which both CAN interfaces are used. They both work as intended while connected to an bus but if I disconnect one of them the kernel process interrupt irq/42-can0 or irq/43-can1 runs at approx 50% cpu utilization each. Disconnecting both makes the system almost none responding.
Here is from TOP:
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
16755 root -51 0 0 0 0 S 53.9 0.0 27:49.78 irq/42-can0
uname -a
Linux beaglebone 4.14.79-ti-rt-r84 #1 SMP PREEMPT RT Tue Nov 13 20:33:45 UTC 2018 armv7l GNU/Linux
Each CAN interface are initilized using
/sbin/ip link set can0 type can bitrate 500000
And the issues happens instantly if I try cansend canX # on the busses. I have tried /sbin/ip link set can0 type can bitrate 500000 restart-ms 100 but it did not help. As far I know, the CPU usage might be valid since the CAN bus standard specs that it should retry until success. However, if I try sending alot of packages:
cangen can0 -g 5 -I 42A -L 1 -v
can1 42A#E0
can1 42A#BF
can1 42A#9D
can1 42A#8A
can1 42A#8D
can1 42A#74
can1 42A#BB
can1 42A#ED
can1 42A#4B
can1 42A#C5
can1 42A#86
write: No buffer space available. But the irq/42-can0 still utilize 50% usage and does not die. If I stop the interface ifconfig can0 down the irq/42-can0 process disappears but that would require me to create somekind of user space program to check the state and start/stop until reconnected and during this, the process will stlil eat up most of the BBB’s CPU power due to the -51 priority thus not optimal.
Dmesg shows this after the Non buffer space available:
[536657.476659] [] (do_page_fault) from [] (do_PrefetchAbort+0x58/0xd0)
[536657.476667] [] (do_PrefetchAbort) from [] (ret_from_exception+0x0/0x14)
[536657.476671] Exception stack(0xd8fadfb0 to 0xd8fadff8)
[536657.476676] dfa0: 00000000 b6eda0b4 00000000 b6e3e051
[536657.476684] dfc0: 00000001 00000000 b6ed7798 00000001 b6eda0b4 00000001 00484000 00000000
[536657.476692] dfe0: b6e15109 becd2530 b6e14eeb b6e3e050 00070030 ffffffff
[536696.690295] INFO: rcu_preempt detected stalls on CPUs/tasks:
[536696.690313] Tasks blocked on level-0 rcu_node (CPUs 0-0): P16900
[536696.690322] (detected by 0, t=5252 jiffies, g=58839897, c=58839896, q=5053)
[536696.690339] cansend R running task 0 16900 7599 0x00000000
[536696.690397] [] (__schedule) from [] (preempt_schedule_irq+0x58/0x84)
[536696.690413] [] (preempt_schedule_irq) from [] (svc_preempt+0x8/0x2c)
[536696.690435] [] (svc_preempt) from [] (filemap_map_pages+0x1ec/0x38c)
[536696.690449] [] (filemap_map_pages) from [] (handle_mm_fault+0x584/0xc30)
[536696.690461] [] (handle_mm_fault) from [] (do_page_fault+0x34c/0x430)
[536696.690473] [] (do_page_fault) from [] (do_PrefetchAbort+0x58/0xd0)
[536696.690484] [] (do_PrefetchAbort) from [] (ret_from_exception+0x0/0x14)
[536696.690489] Exception stack(0xd8fadfb0 to 0xd8fadff8)
[536696.690497] dfa0: 00000000 00000000 00000000 ffffffff
[536696.690506] dfc0: 00000001 00000000 b6ed7798 00000001 b6eda0b4 00000001 00484000 00000000
[536696.690513] dfe0: b6e15109 becd2530 b6e14ef5 b6e606b8 60070030 ffffffff
Anyone know a good way fix this cpu utilization?
/Kusk