在 2025/7/26 05:09, Bjorn Helgaas 写道:

On Wed, Jul 23, 2025 at 11:31:08AM +0800, Shuai Xue wrote:

quoted

PCIe link speed degradation directly impacts system performance and
often indicates hardware issues such as faulty devices, physical layer
problems, or configuration errors.

To this end, add a RAS tracepoint to monitor link speed changes,
enabling proactive health checks and diagnostic analysis.

The output is like below:

$ echo 1 > /sys/kernel/debug/tracing/events/pci/pci_link_event/enable
$ cat /sys/kernel/debug/tracing/trace_pipe
cat /sys/kernel/debug/tracing/trace_pipe
            <...>-119     [002] .....   125.776171: pci_hp_event: 0000:00:03.0 slot:30, event:CARD_PRESENT

            <...>-119     [002] .....   125.776197: pci_hp_event: 0000:00:03.0 slot:30, event:LINK_UP

    irq/57-pciehp-119     [002] .....   125.904335: pcie_link_event: 0000:00:03.0 type:4, reason:4, cur_bus_speed:2.5 GT/s PCIe, max_bus_speed:16.0 GT/s PCIe, width:1, flit_mode:0, status:DLLLA

    irq/57-pciehp-119     [002] .....   125.907051: pcie_link_event: 0000:00:03.0 type:4, reason:0, cur_bus_speed:2.5 GT/s PCIe, max_bus_speed:16.0 GT/s PCIe, width:1, flit_mode:0, status:DLLLA

I guess this example would actually require both of these enables, right?

   echo 1 > /sys/kernel/debug/tracing/events/pci/pci_hp_event/enable
   echo 1 > /sys/kernel/debug/tracing/events/pci/pci_link_event/enable

Yes, you're absolutely right. I'll correct the commit log to show both
commands.

(As a side note, echo 1 > /sys/kernel/debug/tracing/events/pci/enable
would enable both events at once.)

(echo 1 > /sys/kernel/debug/tracing/events/pci/enable will enable both
of these event.)

quoted

Suggested-by: Bjorn Helgaas <helgaas@kernel.org>

I don't think I've suggested anything that really warrants this ;)

Fair enough! I'll drop the Suggested-by tag.

quoted

...

@@ -292,7 +292,7 @@ int pciehp_check_link_status(struct controller *ctrl)
  {
  	struct pci_dev *pdev = ctrl_dev(ctrl);
  	bool found;
-	u16 lnk_status, linksta2;
+	u16 lnk_status;
  
  	if (!pcie_wait_for_link(pdev, true)) {
  		ctrl_info(ctrl, "Slot(%s): No link\n", slot_name(ctrl));

@@ -319,8 +319,7 @@ int pciehp_check_link_status(struct controller *ctrl)
  		return -1;
  	}
  
-	pcie_capability_read_word(pdev, PCI_EXP_LNKSTA2, &linksta2);
-	__pcie_update_link_speed(ctrl->pcie->port->subordinate, lnk_status, linksta2);
+	pcie_update_link_speed(ctrl->pcie->port->subordinate, PCIE_HOTPLUG);

It kind of bugs me that the hot-add flow reads LNKSTA three times and
generates both pci_hp_event LINK_UP and link_event tracepoints:

   pciehp_handle_presence_or_link_change
     link_active = pciehp_check_link_active()
       pcie_capability_read_word(PCI_EXP_LNKSTA)
     if (link_active)
       ctrl_info(ctrl, "Slot(%s): Link Up\n")
       trace_pci_hp_event(PCI_HOTPLUG_LINK_UP)
       pciehp_enable_slot
         __pciehp_enable_slot
           board_added
             pciehp_check_link_status
               pcie_capability_read_word(PCI_EXP_LNKSTA)
               pcie_update_link_speed
                 pcie_capability_read_word(PCI_EXP_LNKSTA)
                 pcie_capability_read_word(PCI_EXP_LNKSTA2)
                 trace_pcie_link_event(<REASON>)

Maybe there are good reasons for reading LNKSTA three times, but it
does make me raise my eyebrows.  Not that this is a performance path,
but it just offends my sense of propriety.

And maybe we need both a bare LINK_UP event and a link_event with all
the details, but again it seems a little weird to me that there are
two tracepoints when there's really only one event and we know all the
link_event information from the very first LNKSTA read.

Bjorn

I understand your concern about the multiple LNKSTA reads and the
apparent duplication. Please see comments from Lukas.

Best Regards,
Shuai