this issue is agx xavier/jetpack4.6;
there are some system reboot but no kernel Oops log in customer production environment.
the device don’t have some valid log and the system maintenance capability is relatively weak.
so we try to add kernel kdump feature to solve it.
we have some tests:
1、for jp4.6/kernel 4.9, add some CONFIG, but can’t allocate memory for crashkernel.
2、for jp5.0.2/kernel 5.10, it is the same thing and can’t allocate memory.
I’ve never tested on a Jetson, but it looks like the kernel command line needs this during load: crashkernel=auto
I have no idea if that works on a Jetson, but to see if that is part of the command line, examine: cat /proc/cmdline
I usually advise that if you are going to experiment with a kernel, and if it might render the system unable to boot, to do so via an alternate boot entry (and then select that entry on boot via serial console). Basically though, look at the APPEND key/value pair in “/boot/extlinux/extlinux.conf”. This is a space-delimited way of adding to the kernel command line. You could put this at the end of that line (it is a long line, do not add any newline): crashkernel=auto
Try it again, and verify you see this first in /proc/cmdline.
Beyond the above I don’t know. It is certainly possible this differs in ARMv8, or just on Jetsons. Someone from NVIDIA might be required to test and comment on kdump.
