Re: [PATCH 03/11] PCI: aardvark: Add support for DLLSC and hotplug interrupt

[Lorenzo Pieralisi <lpieralisi@kernel.org>]

On Tue, Sep 27, 2022 at 01:13:57PM +0200, Pali Rohár wrote:
> Hello! Just briefly from my side, but Marek would probably answer it
> better.
> 
> On Tuesday 27 September 2022 10:29:26 Lorenzo Pieralisi wrote:
> > Better, certainly. Question, also related to Marc's query. Do you
> > rely on the hotplug (emulated IRQ) to be run _before_ carrying on
> > with PCI config space accesses following a link-up detection ?
> 
> During PCI config space access is PCI core code holding atomic raw spin
> lock, so link-up check from PCI config space can throw emulated HP IRQ
> only _after_ config space is finished (when IRQs are unmasked again). So
> it happens after (not before).
> 
> > How was the jiffies + 1 expiration time determined ?
> 
> jiffies + 1 was chosen as the earliest possible time when HP IRQ can be
> thrown. Somebody said to me (year or more ago, no remember who and
> when) that I cannot use just "jiffies", I have to use "jiffies + 1", so
> timer would be scheduled after my call finish, which is after PCI config
> space access finish. jiffies + 1 should be the earliest possible time
> with the highest priority.
> 
> > I assume you
> > want to run the emulated HP IRQ asap - the question though is
> > how fast should it be ?
> 
> HP IRQ should be thrown _ASAP_ when we know that link is up, so PCIe HP
> driver can handle it and do its job. Just like for hardware which fully
> and correctly supports link up HP IRQs.

My question really is - what are we expecting the HP core code to fix up
?

And related, is it safe to carry on with the PCI config space access
till the IRQ is actually emulated to carry out those actions ?

Lorenzo

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel