On Wed, 23 Sep 2015, Wang Dongsheng wrote:

quoted

-----Original Message-----
From: linux-kernel-owner@vger.kernel.org [mailto:linux-kernel-
owner@vger.kernel.org] On Behalf Of Thomas Gleixner
Sent: Wednesday, September 23, 2015 11:49 AM
To: Wang Dongsheng-B40534
Cc: Wood Scott-B07421; Sudeep Holla; linux-pm@vger.kernel.org; linux-
kernel@vger.kernel.org; Rafael J. Wysocki; Benjamin Herrenschmidt; Paul
Mackerras; Michael Ellerman; Jia Hongtao-B38951; Marc Zyngier; linuxppc-
dev@lists.ozlabs.org
Subject: RE: [PATCH 04/17] powerpc: mpic: use IRQCHIP_SKIP_SET_WAKE instead of
redundant mpic_irq_set_wake

Can you please fix you mail client to get rid of that silly copy of
the mail header?

quoted

On Wed, 23 Sep 2015, Wang Dongsheng wrote:

quoted

quoted

On Mon, 2015-09-21 at 16:47 +0100, Sudeep Holla wrote:

quoted

mpic_irq_set_wake return -ENXIO for non FSL MPIC and sets IRQF_NO_SUSPEND
flag for FSL ones. enable_irq_wake already returns -ENXIO if irq_set_wak
is not implemented. Also there's no need to set the IRQF_NO_SUSPEND flag
as it doesn't guarantee wakeup for that interrupt.

Non-freescale return -ENXIO, is there any issue? If non-freescale
platform does not support it, but IPs still use
enable/disable_irq_wake, we should return a error number.

You can just set IRQCHIP_SKIP_SET_WAKE for FSL chips and not for the
others.

quoted

@Scott:
If set this flag we cannot keep a irq as a wakeup source when system going to
SUSPEND or MEM.

irq_set_wake() means we can set this irq as a wake source.
IRQCHIP_SKIP_SET_WAKE is ignore irq_set_wake() feature.

Nonsense. IRQCHIP_SKIP_SET_WAKE merily tells the core not to bail on
!chip->irq_set_wake(), but its still marking the interrupt as wakeup
source and therefor not masking it on suspend.

Sorry, I just check irq_set_irq_wake() code, right, IRQCHIP_SKIP_SET_WAKE also can
going to irqd_set to mask IRQD_WAKEUP_STATE.

Yes, this flag just skip the irq_set_wake() not this feature.

And just for completeness. That commit 5ff04b7287d87c 'powerpc/mpic:
add irq_set_wake support' is another example of trainwreck engineering.

    desc->action->flags |= IRQF_NO_SUSPEND;

This is not only horribly avoiding any of the existing APIs, it's also
broken as hell. desc->action can be NULL when that is called.

It seems fleascale is hell bent to fiddle with the guts of the core
code mindlessly. See commit c866cda47f2c

Yours grumpy

      tglx