On Fri, Sep 18, 2015 at 02:38:39PM +0000, Shenwei Wang wrote:

quoted

-----Original Message-----
From: Shawn Guo [mailto:shawnguo at kernel.org]
Sent: 2015?9?18? 9:15
To: Wang Shenwei-B38339
Cc: jason at lakedaemon.net; Huang Yongcai-B20788;
linux-kernel at vger.kernel.org; kernel at pengutronix.de; Sudeep Holla;
tglx at linutronix.de; shawn.guo at linaro.org; linux-arm-kernel at lists.infradead.org
Subject: Re: [PATCH v8 2/2] ARM: imx: support suspend states on imx7D

On Tue, Sep 08, 2015 at 03:01:07PM +0000, Shenwei Wang wrote:

quoted

quoted

quoted

 arch/arm/mach-imx/Kconfig        |   1 +
 arch/arm/mach-imx/Makefile       |   2 +
 arch/arm/mach-imx/common.h       |   4 +
 arch/arm/mach-imx/pm-imx7.c      | 917

+++++++++++++++++++++++++++++++++++++++

quoted

 arch/arm/mach-imx/suspend-imx7.S | 529 ++++++++++++++++++++++
 5 files changed, 1453 insertions(+)

1453 lines addition to kernel only for i.MX7D suspend support.  Yes,
this is the way we support suspend on i.MX6, but that's enough, and
we have to stop this somewhere.  I would ask you to take Sudeep's
comment and adopt PSCI for i.MX7D power management.

According to PSCI spec, a supervisory software must be implemented
under either
HVC(EL2) or SMC(EL3) privilege level. I agree that, with this idea, we
could finally move these 1453 lines of codes from kernel into a
platform specific firmware. The problem is we don't have the firmware
so far, and we haven't even had a plan to enable the hypervisor or
secure software on this platform. Because low power consumption is the
key feature of this i.MX7D platform, please continue to consider the current

implementation and get this feature supported first.

Can't U-Boot be considered as a firmware to implement PSCI?  I have seen an
initial ARMv7 PSCI core support for U-Boot added by Marc Zyngier.

The current PSCI support in uboot is not complete. It can now support cpuidle, shutdown, 
and reset. But to support system suspend, there would be a lot more tasks. For example,
even the current wakeup source management codes will have to move into the firmware 
too. It is not a short time task. Please continue to consider this platform dependent implementation
first, meanwhile, we will start to migrate the solution to PSCI. However it will take some time.

I have zero interest to maintain this pile of code in the mainline
kernel.  Take it as a motivation to move to PSCI soon, or you maintain
it on your own as a out-of-tree patch for whatever long time you like.

Shawn