On 06/09/16 17:15, Manish Jaggi wrote:

quoted

quoted

In my setup the GICD_CTRL[RWP] bit is not cleared in the
crashkernels' distributor init function.

Which instance is failing? The initial one (just after the initial
disable)? Or the one called from gic_dist_config()?

In crash kernel, when the GICD_CTRL is set to 0x0, RWP is not getting clear.
And is never cleared for any subsequent writes.

That's weird. It means writes are still pending, and never drained. What
happens if you put a dsb(sy) in the wait_for_rwp() loop?

Thanks,

	M.
-- 
Jazz is not dead. It just smells funny...