[v2] irqchip: ls-extirq: add flag IRQCHIP_SKIP_SET_WAKE to remove call trace

[Biwen Li (OSS)]

> -----Original Message-----
> From: Marc Zyngier <maz at kernel.org>
> Sent: 2021年1月27日 19:38
> To: Biwen Li (OSS) <biwen.li at oss.nxp.com>
> Cc: mark.rutland at arm.com; Leo Li <leoyang.li at nxp.com>; tglx at linutronix.de;
> jason at lakedaemon.net; linux-kernel at vger.kernel.org; Jiafei Pan
> <jiafei.pan at nxp.com>; linux-arm-kernel at lists.infradead.org; Ran Wang
> <ran.wang_1 at nxp.com>; Biwen Li <biwen.li at nxp.com>
> Subject: Re: [v2] irqchip: ls-extirq: add flag IRQCHIP_SKIP_SET_WAKE to
> remove call trace
> 
> On 2021-01-27 08:58, Biwen Li wrote:
> > From: Biwen Li <biwen.li at nxp.com>
> >
> > Add flag IRQCHIP_SKIP_SET_WAKE to remove call trace as follow, ...
> > [   45.605239] Unbalanced IRQ 120 wake disable
> > [   45.609445] WARNING: CPU: 0 PID: 1124 at kernel/irq/manage.c:800
> > irq_set_irq_wake+0x154/0x1a0
> > ...
> > [   45.645141] pstate: 60000085 (nZCv daIf -PAN -UAO -TCO BTYPE=--)
> > [   45.651144] pc : irq_set_irq_wake+0x154/0x1a0
> > [   45.655497] lr : irq_set_irq_wake+0x154/0x1a0
> > ...
> > [   45.742825] Call trace:
> > [   45.745268]  irq_set_irq_wake+0x154/0x1a0
> > [   45.749278]  ds3232_resume+0x38/0x50
> >
> > On ls2088ardb:
> > In suspend progress(# echo mem > /sys/power/state),
> >
> pm_suspend()->suspend_devices_and_enter()->dpm_suspend()->device_suspe
> > nd()
> > ->ds3232_suspend()->enable_irq_wake()->irq_set_irq_wake()
> > ->set_irq_wake_real(), return -ENXIO, there get
> > "Cannot set wakeup source" in ds3232_suspend().
> >
> > In resume progress(wakeup by flextimer)
> > dpm_resume_end()->dpm_resume()
> > ->device_resume()->ds3232_resume()
> > ->disable_irq_wake()->irq_set_irq_wake()
> > ->set_irq_wake_real(), there get
> > kernel call trace(Unbalanced IRQ 120 wake
> > disable)
> 
> This is again paraphrasing the stack trace instead of explaining the problem it
> fixes. How about:
> 
> "The ls-extirq driver doesn't implement the irq_set_wake()
>   callback, while being wake-up capable. This results in
>   ugly behaviours across suspend/resume cycles.
> 
>   Advertise this by adding IRQCHIP_SKIP_SET_WAKE to the irqchip
>   flags"
> 
> The subject line should be fixed along the same lines, and a Fixes: tag added.
Okay, got it. Thanks. Will update in v3.
> 
>          M.
> --
> Jazz is not dead. It just smells funny...