Re: [PATCH 2/2] PCI: uniphier: Add checking whether PERST# is deasserted

[Andrew Murray]

On Fri, Nov 08, 2019 at 04:30:27PM +0900, Kunihiko Hayashi wrote:
> Hi Andrew,
> +CC Kishon
> 
> On Thu, 7 Nov 2019 12:46:17 +0000 <andrew.murray@xxxxxxx> wrote:
> 
> > On Thu, Nov 07, 2019 at 08:52:39PM +0900, Kunihiko Hayashi wrote:
> > > Hi Andrew,
> > > Thank you for your comments.
> > > 
> > > On Thu, 7 Nov 2019 10:02:08 +0000 <andrew.murray@xxxxxxx> wrote:
> > > 
> > > > On Thu, Nov 07, 2019 at 01:58:15PM +0900, Kunihiko Hayashi wrote:
> > > > > When PERST# is asserted once, EP configuration will be initialized.
> > > > 
> > > > I don't quite understand this - does the EP/RC mode depend on how often
> > > > PERST# is toggled?
> > > 
> > > I think of connecting this RC controller and EP based on `Linux PCI
> > > endpoint framework' in another machine.
> > > 
> > > While this RC driver is probing, the EP driver might be also probing and
> > > configurating itself using configfs. If PERST# is toggled after the EP
> > > has done its configuration, this configuration will be lost.
> > > 
> > > I expect that the EP configurates after RC has toggled PERST#, however,
> > > there is no way to synchronize both of them.
> > > 
> > 
> > OK I understand where you are coming from now. Please ensure the commit
> > message gives this rationale.
> 
> I'll explain about that in the commit message next.
> 
> > However, If I understand correctly, doesn't your solution only work some
> > of the time? What happens if you boot both machines at the same time,
> > and PERST# isn't asserted prior to the kernel booting?
> 
> I think it contains an annoying problem.
> 
> If PERST# isn't toggled prior to the kernel booting, PERST# remains asserted
> and the RC driver can't access PCI bus.
> 
> As a result, this patch works and deasserts PERST# (and EP configuration will
> be lost). So boot sequence needs to include deasserting PERST#.
> 
> > The only way you can ensure the EP is started after the RC is initialised
> > is to start the EP after the RC is initialised.
> 
> Yes, it's the only soution for now.
> 
> > I'm not sure what the solution is here, but it feels like this approach
> > only partially solves it.
> 
> Surely relying on outside of the driver doesn't seem to be a complete solution.
> 
> If there is the way that `Linux PCI endpoint framework' assumes,
> I'd like to follow it, however, I can't find the other way.

Indeed, this must be a common problem - many host controller drivers in the
tree toggle perst on start up.

Keen for any feedback from Kishon on this.

Thanks,

Andrew Murray

> 
> Thank you,
> 
> ---
> Best Regards,
> Kunihiko Hayashi
> 
>