4 ms·
> Is there really no better way to do this than to have an entire second kernel > ready to take over? Like a more specialized piece of code that only handles
by worthless-trash 3y ago
> Is there really no better way to do this than to have an entire second kernel
> ready to take over? Like a more specialized piece of code that only handles
> kernel coredumps?
So the requirements of this software are: It must be able to boot a system, write to the specific sections of memory, have some method of initializing hardware, some ability to write the other memory to disk or network connections. It sounds like you're talking about an OS.
I believe this is exactly what KDUMP does.
- loeg 3y agoThe crash dumper only needs to write to disk or network. The other steps are just things the kexec process does to achieve that aim.
- worthless-trash 3y agoNo, this is not true. If you just boot like a regular operating system, you will trash the crashed kernel making the dump useless.
- loeg 3y agoA crash dumper does not have to be implemented with a reboot; this is just something Linux chooses to do. I have actually worked on kernel code that performs crash dumps to disk and over the network, if you find that kind of expertise relevant.
- worthless-trash 3y agoI really do. I believe kexec/kdump does execute in place and does not go through a full reboot, am I wrong ?