Dan, that was helpful because I found kernel 6.12.57+deb13-amd64 in the boot options and have been running that version three days with no crashing. The version I'm having a regression with is 6.12.63.
To try to narrow the problem down more before filing a bug I was going to connect a kernel debugger, get symbols lined up and wait for it to crash and hopefully find the culprit in the stack trace. But on
https://www.kernel.org/doc/html/v4.14/dev-tools/kgdb.html it shows only serial bus for debugging and the debugger machine doesn't have a serial port. It has a network jack and USB. is there another reliable (it stays connected) way to connect a KD besides over serial?
joel wrote:
I'm a Windows convert and ran bullseye, bookworm and Trixie as a host.
It was reliable with all oses where I could hibernate several times a
day everyday for over a month and the uptime utility reflected that.
With Trixie a week or so ago I ran updates and the machine crashes with
1 day of uptime at most. Its been crashing repeatedly ever since.
When the machine crashes parts of the video gets scrambled and then the
electricity usage spikes up. Reset button, power button (to hibernate),
mouse, keyboard are all unresponsive. How would you investigate to see
if its a known problem?
The first thing I would do is find the previous kernel and boot
from that.
If you're using grub, that should already be in the menu for it
at boot time.
If it continues to crash on the old kernel, it's likely a
hardware problem. If it doesn't:
a) congratulations, you have a workaround
b) you will want to send a bug report to the kernel team
-dsr-