Re: [PATCH] pci: make reset_subordinate hotplug safe

[Keith Busch <kbusch@kernel.org>]

On Wed, Jan 14, 2026 at 01:18:28PM -0600, Bjorn Helgaas wrote:
> [+cc Alex, Lukas]
> 
> On Wed, Jan 14, 2026 at 10:58:21AM -0800, Keith Busch wrote:
> > From: Keith Busch <kbusch@kernel.org>
> > 
> > Use the slot reset method when resetting the bridge if the bus contains
> > hotplug slots. This fixes spurious hotplug events the secondary bus
> > reset triggers by utilizing slot's specific reset callback that ignores
> > link events.
> 
> What do these spurious events look like in dmesg?

Sure, after starting the reset_subordinate:

 # echo 1 > /sys/bus/pci/devices/0000:00:03.1/reset_subordinate

Kernel dmesg currently show something like this:

  pci 0000:01:03.4: save config 0x30: 0x00000000
  pci 0000:01:03.4: save config 0x34: 0x00000040
  pci 0000:01:03.4: save config 0x38: 0x00000000
  pci 0000:01:03.4: save config 0x3c: 0x000001ff
  pcieport 0000:00:03.1: pciehp: pending interrupts 0x0100 from Slot Status
  pcieport 0000:00:03.1: pciehp: Slot(33): Link Down
  pci 0000:01:03.4: PME# disabled
  pcieport 0000:00:03.1: waiting 100 ms for downstream link, after activation
  pcieport 0000:00:03.1: pciehp: pending interrupts 0x0008 from Slot Status
  pcieport 0000:00:03.1: pciehp: pending interrupts 0x0008 from Slot Status
  pcieport 0000:00:03.1: pciehp: pending interrupts 0x0100 from Slot Status
  pci 0000:01:00.0: disconnected; not waiting
  pci 0000:01:00.0: restore config 0x3c: 0xffffffff -> 0x000001ff

The desirable behavior we'd expect to see instead would inclue this
message:

  pcieport 0000:00:03.1: pciehp: Slot(33): Link Down/Up ignored