Re: Linux warns `Unknown NUMA node; performance will be reduced`

[Bjorn Helgaas <helgaas@kernel.org>]

On Mon, Jun 10, 2024 at 10:27:37PM +0200, Paul Menzel wrote:
> Am 10.06.24 um 21:42 schrieb Bjorn Helgaas:
> > On Sun, Jun 09, 2024 at 10:31:05AM +0200, Paul Menzel wrote:
> > > On the servers below Linux warns:
> > > 
> > >       Unknown NUMA node; performance will be reduced
> > 
> > This warning was added by ad5086108b9f ("PCI: Warn if no host bridge
> > NUMA node info"), which appeared in v5.5, so I assume this isn't new.
> > 
> > That commit log says:
> > 
> >    In pci_call_probe(), we try to run driver probe functions on the node where
> >    the device is attached.  If we don't know which node the device is attached
> >    to, the driver will likely run on the wrong node.  This will still work,
> >    but performance will not be as good as it could be.
> > 
> >    On NUMA systems, warn if we don't know which node a PCI host bridge is
> >    attached to.  This is likely an indication that ACPI didn't supply a _PXM
> >    method or the DT didn't supply a "numa-node-id" property.
> > 
> > I assume these are all ACPI systems, so likely missing _PXM.  An
> > acpidump could confirm this.
> 
> I created an issue in the Linux Kernel Bugzilla [1] and attached the output
> of `acpidump` on a Dell PowerEdge T630 there. The DSDT contains:
> 
>         Device (PCI1)
>         {
>         […]
>             Method (_PXM, 0, NotSerialized)  // _PXM: Device Proximity
>             {
>                 If ((CLOD == 0x00))
>                 {
>                     Return (0x01)
>                 }
>                 Else
>                 {
>                     Return (0x02)
>                 }
>             }
>         […]
>         }

This machine (the T630, from your first message) has several PCI host
bridges:

  ACPI: PCI Root Bridge [UNC1] (domain 0000 [bus ff])
  pci_bus 0000:ff: root bus resource [bus ff]

  ACPI: PCI Root Bridge [UNC0] (domain 0000 [bus 7f])
  pci_bus 0000:7f: root bus resource [bus 7f]

  ACPI: PCI Root Bridge [PCI0] (domain 0000 [bus 00-7e])
  pci_bus 0000:00: root bus resource [io  0x0000-0x03bb window]
  pci_bus 0000:00: root bus resource [io  0x03bc-0x03df window]
  pci_bus 0000:00: root bus resource [io  0x03e0-0x0cf7 window]
  pci_bus 0000:00: root bus resource [io  0x1000-0x7fff window]
  pci_bus 0000:00: root bus resource [mem 0x000a0000-0x000bffff window]
  pci_bus 0000:00: root bus resource [mem 0x90000000-0xc7ffbfff window]
  pci_bus 0000:00: root bus resource [mem 0x38000000000-0x3bfffffffff window]
  pci_bus 0000:00: root bus resource [bus 00-7e]

  ACPI: PCI Root Bridge [PCI1] (domain 0000 [bus 80-fe])
  pci_bus 0000:80: root bus resource [io  0x8000-0xffff window]
  pci_bus 0000:80: root bus resource [mem 0xc8000000-0xfbffbfff window]
  pci_bus 0000:80: root bus resource [mem 0x3c000000000-0x3ffffffffff window]
  pci_bus 0000:80: root bus resource [bus 80-fe]

PCI0 and PCI1 lead to all your normal PCI devices, they both
implement _PXM, and they have all the usual apertures for PCI I/O and
MMIO space where device BARs live.

UNC0 and UNC1 lead to these special chipset devices, they don't
implement _PXM, and they don't have any resources except the bus
number.  The devices on bus 7f and ff can only be used via config
space accesses, and I have no idea what they are used for.

> [1]: https://bugzilla.kernel.org/show_bug.cgi?id=218951