On Thu, Mar 20, 2014 at 02:03:22AM +0000, Ashwin Chaugule wrote:
Hi Mark,
On 14 March 2014 05:52, Mark Rutland mark.rutland@arm.com wrote:
psci_function_id[PSCI_FN_MIGRATE] = PSCI_ID_CPU_MIGRATE;psci_ops.migrate = psci_migrate;It would also be a good idea to hook up AFFINITY_INFO to a cpu_kill callback from the start, to ensure implementors bother to implement it, and such that we can make kexec safe(r).
If we rely on it from the start it's more likely that it will actually get implemented, and it will highlight any dodgy PSCI 0.2 implementations while there's a chance they can be fixed.
Although I sent v1 of this patch, I haven't forgotten about this part. Had a go at it, but it turned out quite messy for ARM32. I was trying to get cpu_kill via smp_ops.
I'm not sure I follow. Do you have an example patch or tree that would highlight the difficulty?
I'll revisit this again hopefully tomorrow and see if there's anything better.
Good to hear!
Cheers, Mark.