Thread (37 messages) 37 messages, 5 authors, 2025-07-26

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

From: Lukas Wunner <lukas@wunner.de>
Date: 2025-05-20 10:44:46
Also in: linux-edac, linux-pci, lkml

On Tue, May 20, 2025 at 01:07:28PM +0300, Ilpo Järvinen wrote:
On Tue, 20 May 2025, Shuai Xue wrote:
quoted
, and the format of "Link Speed changed" is a bit different from 
"pci_hp_event".
The difference is only because when the Link is down, there's no Link
Speed (obviously). Whenever a new device is hotplugged and it comes up, 
there's also Link Speed for it which can be included into the trace event. 

I think the trace event should have some special value for the fields that 
are N/A due to Link being off. While it would be possible to create 
separate events for speed changes and hotplug, I don't see any pros in 
that approach over just having the N/A fields marked as such when the Link 
is Down.
Link speed changes and device plug/unplug events are orthogonal,
I don't think they should be mixed together in the same event.

A link speed event can be signaled simultaneously to a plug event
and then user space can decide in which type of event it's
interested in.

That also avoids the awkwardness of having N/A values for the
link speed on unplug.

Thanks,

Lukas
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help