Re: [Patch v4] PCI: hv: Fix NUMA node assignment when kernel boots with custom NUMA topology

[Wei Liu <wei.liu@kernel.org>]

On Wed, Jan 26, 2022 at 05:43:34PM -0800, longli@linuxonhyperv.com wrote:
> From: Long Li <longli@microsoft.com>
> 
> When kernel boots with a NUMA topology with some NUMA nodes offline, the PCI
> driver should only set an online NUMA node on the device. This can happen
> during KDUMP where some NUMA nodes are not made online by the KDUMP kernel.
> 
> This patch also fixes the case where kernel is booting with "numa=off".
> 
> Fixes: 999dd956d838 ("PCI: hv: Add support for protocol 1.3 and support PCI_BUS_RELATIONS2")
> Signed-off-by: Long Li <longli@microsoft.com>
> Reviewed-by: Michael Kelley <mikelley@microsoft.com>

CC PCI maintainers.

Hi Bjorn and Lorenzo

Are you going to pick this up? This is a fix we would like to see
upstream sooner rather than later.

I can pick this up via hyperv-fixes if that suits you, since this only
affect pci-hyperv.c.

Thanks,
Wei.

> ---
> Change log:
> v2: use numa_map_to_online_node() to assign a node to device (suggested by
> Michael Kelly <mikelley@microsoft.com>)
> v3: add "Fixes" and check for num_possible_nodes()
> v4: fix commit message format
> 
>  drivers/pci/controller/pci-hyperv.c | 13 +++++++++++--
>  1 file changed, 11 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/pci/controller/pci-hyperv.c b/drivers/pci/controller/pci-hyperv.c
> index 20ea2ee330b8..ae0bc2fee4ca 100644
> --- a/drivers/pci/controller/pci-hyperv.c
> +++ b/drivers/pci/controller/pci-hyperv.c
> @@ -2155,8 +2155,17 @@ static void hv_pci_assign_numa_node(struct hv_pcibus_device *hbus)
>  		if (!hv_dev)
>  			continue;
>  
> -		if (hv_dev->desc.flags & HV_PCI_DEVICE_FLAG_NUMA_AFFINITY)
> -			set_dev_node(&dev->dev, hv_dev->desc.virtual_numa_node);
> +		if (hv_dev->desc.flags & HV_PCI_DEVICE_FLAG_NUMA_AFFINITY &&
> +		    hv_dev->desc.virtual_numa_node < num_possible_nodes())
> +			/*
> +			 * The kernel may boot with some NUMA nodes offline
> +			 * (e.g. in a KDUMP kernel) or with NUMA disabled via
> +			 * "numa=off". In those cases, adjust the host provided
> +			 * NUMA node to a valid NUMA node used by the kernel.
> +			 */
> +			set_dev_node(&dev->dev,
> +				     numa_map_to_online_node(
> +					     hv_dev->desc.virtual_numa_node));
>  
>  		put_pcichild(hv_dev);
>  	}
> -- 
> 2.25.1
>