On 15-1-2022 15:28, Eugene Grosbein wrote:
15.01.2022 21:06, Willem Jan Withagen wrote:
On 14-1-2022 17:41, Eugene Grosbein wrote:
14.01.2022 23:34, Willem Jan Withagen wrote
That is it..... it is "dead" after that
Does 14-CURRENT have KDB stuff preloaded, I could give it a shot.
I'm not sure, you'll have to try it yourself.
Otherwise I'd have to build 11.4 and go into KDB from there.
What are the first things you need to know?
https://docs.freebsd.org/en/books/developers-handbook/book/
Look at the section "10.6. Debugging Deadlocks" as the problem looks like some
deadlock (no panic, no progress).
I have tried all incarnations on that page.
Even found
https://gist.github.com/bijanebrahimi/f2eb0c620d81aa6234e121a0ddd88cc2
which I think has the info more organized.
Can it be that the deadlock is too early for the DDB break to be triggered?
Maybe, can't tell for sure. I'm afraid you'll need to dig it more:
read the code and add more "prints" to the kernel.
Got this a bit further by adding this to the config:
options ACPI_DEBUG
options VERBOSE_SYSINIT=1
But now it tells me:
Table 'FACP' at 0xdffc0200
Table 'APIC' at 0xdffc0390
Table 'MCFG' at 0xdffc0460
Table 'OEMB' at 0xdffce040
Table 'SRAT' at 0xdffc50f0
ACPI: No SLIT table found
done.
subsystem 1000000
vm_mem_init(0)...
So I guess that the ACPI stuff is done, and we're going to look at
init-ing the subsystems.
But how do I translate the number to a name?
I guess it is the virtual memory init, since there is only on place that
calls vm_mem_init.
And that is from the sysinit chain.
But I would expect things to be really bad if the system hangs in the VM
initialisation?
--WjW
