Vivek Goyal wrote:

>I think if we decide to implement something which allows other policies 
>to co-exist with crash_kexec() then it should be more generic then a
>single function pointer.
>
>Thanks
>Vivek
>
>  
>
A single function pointer function is suggested because it is the 
simpliest compromise
I can thing of which should be able to satisfy all.

The simpliest policy I can think of is
-flip a bit on _dedicated_ hardware (crash notifier)
-launch capture kernel (existing kexec)

Nothing prevents you from implementing multiple policies to be 
executed/selected among from
whatever is called by the single pointer function.


My key point is:
The complexity in my suggestion is a low as it can get, thus reliability of
kexec (hopefully) is unaffected
If crash notifiers is implemented by a complex "management system", I 
might loose
reliability of kexec because of something I basically do not need.

Or to put it in other words, I you need to implement anything complex 
for managing your policies,
you should add it yourself and you yourself is the only one being 
affected by increased complexibility.

./Preben
_______________________________________________
fastboot mailing list
[email protected]
https://lists.osdl.org/mailman/listinfo/fastboot

Reply via email to