[PATCH v2 3/4] PCI: Add link down handling for host bridges

[Manivannan Sadhasivam]

On Thu, Apr 17, 2025 at 02:41:55PM +0530, Krishna Chaitanya Chundru wrote:
> 
> 
> On 4/17/2025 1:24 PM, Manivannan Sadhasivam wrote:
> > On Wed, Apr 16, 2025 at 11:21:49PM +0530, Krishna Chaitanya Chundru wrote:
> > > 
> > > 
> > > On 4/16/2025 9:59 PM, Manivannan Sadhasivam via B4 Relay wrote:
> > > > From: Manivannan Sadhasivam <manivannan.sadhasivam at linaro.org>
> > > > 
> > > > The PCI link, when down, needs to be recovered to bring it back. But that
> > > > cannot be done in a generic way as link recovery procedure is specific to
> > > > host bridges. So add a new API pci_host_handle_link_down() that could be
> > > > called by the host bridge drivers when the link goes down.
> > > > 
> > > > The API will iterate through all the slots and calls the pcie_do_recovery()
> > > > function with 'pci_channel_io_frozen' as the state. This will result in the
> > > > execution of the AER Fatal error handling code. Since the link down
> > > > recovery is pretty much the same as AER Fatal error handling,
> > > > pcie_do_recovery() helper is reused here. First the AER error_detected
> > > > callback will be triggered for the bridge and the downstream devices. Then,
> > > > pcie_do_slot_reset() will be called for each slots, which will reset the
> > > > slots using 'reset_slot' callback to recover the link. Once that's done,
> > > > resume message will be broadcasted to the bridge and the downstream devices
> > > > indicating successful link recovery.
> > > > 
> > > > In case if the AER support is not enabled in the kernel, only
> > > > pci_bus_error_reset() will be called for each slots as there is no way we
> > > > could inform the drivers about link recovery.
> > > > 
> > > The PCIe endpoint drivers are registering with err_handlers and they
> > > will be invoked only from pcie_do_recovery, but there are getting built
> > > by default irrespective of AER is enabled or not.
> > > 
> > 
> > AER is *one* of the functionalities of an endpoint. And the endpoint could
> > mostly work without AER reporting (except for AER fatal/non-fatal where recovery
> > need to be performed by the host). So it wouldn't make sense to add AER
> > dependency for them.
> > 
> > > Does it make sense to built err.c irrespective of AER is enabled or not
> > > to use common logic without the need of having dependency on AER.
> > > 
> > 
> > Well, yes and no. Right now, only DPC reuses the err handlers except AER. But
> > DPC driver itself is functional dependent on AER. So I don't think it is really
> > required to build err.c independent of AER. But I will try to rework the code in
> > the future for fixing things like 'AER' prefix added to logs and such.
> > 
> Right now we have DPC & AER to use this pcie_do_recovery(), now we are
> adding supporting for controller reported error (Link down) not sure if
> there will be newer ways to report errors in future.
> 
> May be not in this series, in future better to de-couple err.c from
> AER as err.c. As the sources of error reporting is not limited to AER
> or DPC alone now.
> 

Yes, that's part of my plan.

> > > Also since err.c is tied with AER, DPC also had a hard requirement
> > > to enable AER which is not needed technically.
> > > 
> > 
> > DPC driver is functional dependent on AER.
> I got a impression by seeing below statement that DPC can work
> independently.
> As per spec 6 sec 6.2.11.2, DPC error signaling "A DPC-capable
> Downstream Port must support ERR_COR signaling, independent of whether
> it supports Advanced Error Reporting (AER) or not".
> 

That's why I intentionally said 'DPC driver' not 'DPC'. The driver has the
dependency, not the feature.

> In fact it can work if AER is not enabled also, but will not have full
> functionality of DPC.
> 

Right. That's why I said functionally dependent.

- Mani

-- 
மணிவண்ணன் சதாசிவம்