Re: [PATCH v8] PCI: hotplug: Add a generic RAS tracepoint for hotplug event

[Matthew W Carlis <mattc@purestorage.com>]

On Thu, 17 Jul 2025, Bjorn Helgaas wrote:
> - slot_name() (which I think comes from make_slot_name(); would you
>   want something else?)

afaik it ends up coming from the Slot Cap Register "Physical Slot Number" bits.
I brought up the slot to just say that I was happy to see it & that it is useful
for our purposes & why.

On Thu, 17 Jul 2025, Lukas Wunner wrote:
>> and IIUC, it would be helpful for you to add:
>> 
>>   - DSP Vendor/Device ID (the Root Port or Switch Downstream Port,
>>     which is relatively static, so seems less useful to me than the
>>     USP/EP would be)
>
> Right, this is already logged in dmesg upon enumeration of the hotplug port,
> as well as available via lspci.

I also agree that the DSP Vendor/Device ID is less useful.

>>   - USP/EP Vendor/Device ID
>
> There's no 1:1 relation between link or presence events on the one hand,
> and enumeration of hotplugged components on the other hand:  The link
> may go up but the kernel may fail to enumerate the component, e.g. because
> it was yanked before it could be enumerated, or because the kernel has run
> out of MMIO space or bus numbers.
> 
> Hence this would have to be logged through a separate tracepoint in
> pciehp_configure_device(), not by changing the tracepoints added here.

Ok I think its reasonable to use a separate tracepoint that would have more
information about the EP.