[PATCH V5 3/3] PCI: Mask and unmask hotplug interrupts during reset

On 7/3/2018 9:59 AM, Lukas Wunner wrote:

On Tue, Jul 03, 2018 at 09:31:24AM -0400, Sinan Kaya wrote:

quoted

Issue is observing hotplug link down event in the middle of AER recovery
as in my previous reply.

If we mask hotplug interrupts before secondary bus reset via my patch,
then hotplug driver will not observe both link up and link down interrupts.

If we don't mask hotplug interrupts, we have a race condition.

I assume that a bus reset not only causes a link and presence event but
also clears the Presence Detect State bit in the Slot Status register
and the Data Link Layer Link Active bit in the Link Status register
momentarily.

pciehp may access those two bits concurrently to the AER driver
performing a slot reset.  So it may not be sufficient to mask
the interrupt.

I've posted this patch to address the issue:
https://patchwork.ozlabs.org/patch/930391/

Very interesting!

I missed this completely.

I know for a fact that bus reset clears the Data Link Layer Active bit
as soon as link goes down. It gets set again following link up.

Presence detect depends on the HW implementation. 

QDT root ports don't change presence detect for instance since nobody
actually removed the card.

If an implementation supports in-band presence detect, the answer is yes.
As soon as the link goes down, presence detect bit will get cleared until
recovery.

It sounds like we need to update your lock change with my proposal.

lock() in mask_irq() and unlock() in unmask_irq()

Thanks,

Lukas

-- 
Sinan Kaya
Qualcomm Datacenter Technologies, Inc. as an affiliate of Qualcomm Technologies, Inc.
Qualcomm Technologies, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project.