On 2017-05-31 07:27, Stephen Boyd wrote:
> On 05/30, Kiran Gunda wrote:
>> From: Abhijeet Dharmapurikar <[email protected]>
>>
>> We see a unmapped irqs trigger right around bootup. This could
>> likely be because the bootloader exited leaving the interrupts
>> in an unknown or unhandled state. Ack and mask the interrupt
>> if one is found. A request_irq later will unmask it and also
>> setup proper mapping structures.
>
> Do we have systems where this is causing an interrupt storm due
> to a level high interrupt or something? Just plain acking and
> masking irqs at boot if we don't have an irq descriptor created
> yet doesn't sound like a good idea, because we'll lose all
> interrupts that happen before this driver probes?
>
Yes. There were instances of an interrupt storm without this patch.
There is an RT_STATUS register in the peripheral address space which
maintains the real time state and can be read by the client driver
before it registers for the irq. Few client drivers such as charger
already doing this.
>>
>> Also the current driver ensures that no read/write transaction
>> is in progress while it makes changes to the interrupt regions.
>> This is not necessary because read/writes over spmi and arbiter
>> interrupt control are independent operations. Hence, remove the
>> synchronized accesses to interrupt region.
>
> That's another patch for this paragraph.
This patch is already pulled in to Greg's tree.
On 06/06, [email protected] wrote:
> On 2017-05-31 07:27, Stephen Boyd wrote:
> >On 05/30, Kiran Gunda wrote:
> >>From: Abhijeet Dharmapurikar <[email protected]>
> >>
> >>We see a unmapped irqs trigger right around bootup. This could
> >>likely be because the bootloader exited leaving the interrupts
> >>in an unknown or unhandled state. Ack and mask the interrupt
> >>if one is found. A request_irq later will unmask it and also
> >>setup proper mapping structures.
> >
> >Do we have systems where this is causing an interrupt storm due
> >to a level high interrupt or something? Just plain acking and
> >masking irqs at boot if we don't have an irq descriptor created
> >yet doesn't sound like a good idea, because we'll lose all
> >interrupts that happen before this driver probes?
> >
> Yes. There were instances of an interrupt storm without this patch.
> There is an RT_STATUS register in the peripheral address space which
> maintains the real time state and can be read by the client driver
> before it registers for the irq. Few client drivers such as charger
> already doing this.
So you're saying that drivers need to read RT_STATUS to know if
they have a pending interrupt before requesting it? That sounds
bogus.
--
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project
On 2017-06-13 07:41, Stephen Boyd wrote:
> On 06/06, [email protected] wrote:
>> On 2017-05-31 07:27, Stephen Boyd wrote:
>> >On 05/30, Kiran Gunda wrote:
>> >>From: Abhijeet Dharmapurikar <[email protected]>
>> >>
>> >>We see a unmapped irqs trigger right around bootup. This could
>> >>likely be because the bootloader exited leaving the interrupts
>> >>in an unknown or unhandled state. Ack and mask the interrupt
>> >>if one is found. A request_irq later will unmask it and also
>> >>setup proper mapping structures.
>> >
>> >Do we have systems where this is causing an interrupt storm due
>> >to a level high interrupt or something? Just plain acking and
>> >masking irqs at boot if we don't have an irq descriptor created
>> >yet doesn't sound like a good idea, because we'll lose all
>> >interrupts that happen before this driver probes?
>> >
>> Yes. There were instances of an interrupt storm without this patch.
>> There is an RT_STATUS register in the peripheral address space which
>> maintains the real time state and can be read by the client driver
>> before it registers for the irq. Few client drivers such as charger
>> already doing this.
>
> So you're saying that drivers need to read RT_STATUS to know if
> they have a pending interrupt before requesting it? That sounds
> bogus.
In many cases a PMIC module driver does need to know the initial state
of
the hardware during driver initialization and the RT_STATUS register
indicates
this (irrespective of the IRQ_EN status). The reset value of IRQ_EN =
0 and if
an event occurs before we request an IRQ, there is no way to know the
initial state
even after we request this irq as there will be no pending interrupt
because the
HW default value of IRQ_EN = 0. This can be known only through reading
the RT_STATUS.
I understand your concern of clearing the IRQ before we request it does
not seem
a clean way, do you have any other recommendation of this could be
handled ?