Re: [PATCH v2 1/4] PCI: dw-rockchip: Do not enumerate bus before endpoint devices are ready

[Bjorn Helgaas <helgaas@kernel.org>]

On Thu, Jun 05, 2025 at 02:28:41PM +0200, Niklas Cassel wrote:
> On Wed, Jun 04, 2025 at 01:44:45PM -0500, Bjorn Helgaas wrote:
> > On Wed, Jun 04, 2025 at 10:40:09PM +0530, Manivannan Sadhasivam wrote:
> > 
> > > > If we add a 100 ms sleep after wait_for_link(), then I suggest that we
> > > > also reduce LINK_WAIT_SLEEP_MS to something shorter.
> > > 
> > > No. The 900ms sleep is to make sure that we wait 1s before erroring out
> > > assuming that the device is not present. This is mandated by the spec. So
> > > irrespective of the delay we add *after* link up, we should try to detect the
> > > link up for ~1s.
> > 
> > I think it would be sensible for dw_pcie_wait_for_link() to check for
> > link up more frequently, i.e., reduce LINK_WAIT_SLEEP_MS and increase
> > LINK_WAIT_MAX_RETRIES.
> > 
> > If LINK_WAIT_SLEEP_MS * LINK_WAIT_MAX_RETRIES is for the 1.0s
> > mentioned in sec 6.6.1, seems like maybe we should make a generic
> > #define for it so we could include the spec reference and use it
> > across all drivers.  And resolve the question of 900ms vs 1000ms.
> 
> Like Bjorn mentioned, when I wrote reduce LINK_WAIT_SLEEP_MS,
> I simply meant that we should poll for link up more frequently.
> 
> But yes, if we reduce LINK_WAIT_SLEEP_MS we should bump
> LINK_WAIT_MAX_RETRIES to not change the current max wait time.
> 
> 
> Bjorn, should I send something out after -rc1, or did you want
> to work on this yourself?

Yes, please post something after -rc1.  Given the number of drivers
and the much smaller number of msleep() calls, I suspect lots of
drivers have similar problems.

Bjorn