[PATCH v3] powerpc/powernv: Add queue mechanism for early messages

[Michael Ellerman]

On Mon, 2018-05-21 at 02:04:38 UTC, Deb McLemore wrote:
> Problem being solved is when issuing a BMC soft poweroff during IPL,
> the poweroff was being lost so the machine would not poweroff.
> 
> Opal messages were being received before the opal-power code
> registered its notifiers.
> 
> Alternatives discussed (option #3 was chosen):
> 
> 1 - Have opal_message_init() explicitly call opal_power_control_init()
> before it dequeues any OPAL messages (i.e. before we register the
> opal-msg IRQ handler).
> 
> 2 - Introduce concept of critical message types and when we register
> handlers we track which message types have a registered handler,
> then defer the opal-msg IRQ registration until we have a handler
> registered for all the critical types.
> 
> 3 - Buffering messages, if we receive a message and do not yet
> have a handler for that type, store the message and replay when
> a handler for that type is registered.
> 
> Signed-off-by: Deb McLemore <debmc at linux.vnet.ibm.com>

Applied to powerpc next, thanks.

https://git.kernel.org/powerpc/c/a9336ddf448b1cba3080195cec2287af3907236c

cheers