Re: [PATCH v2 03/22] PCI: Inherit bus numbers from previous kernel during Live Update

[Pranjal Shrivastava <praan@google.com>]

On Thu, Jan 29, 2026 at 09:24:50PM +0000, David Matlack wrote:
> Inherit bus numbers from the previous kernel during a Live Update when
> one or more PCI devices are being preserved. This is necessary so that
> preserved devices can DMA through the IOMMU during a Live Update
> (changing bus numbers would break IOMMU translation).
> 
> Signed-off-by: David Matlack <dmatlack@google.com>
> ---
>  drivers/pci/probe.c | 21 ++++++++++++++++++---
>  1 file changed, 18 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c
> index af6356c5a156..ca6e5f79debb 100644
> --- a/drivers/pci/probe.c
> +++ b/drivers/pci/probe.c
> @@ -1351,6 +1351,20 @@ static bool pci_ea_fixed_busnrs(struct pci_dev *dev, u8 *sec, u8 *sub)
>  	return true;
>  }
>  
> +static bool pci_assign_all_busses(void)
> +{
> +	/*
> +	 * During a Live Update where devices are preserved by the previous
> +	 * kernel, inherit all bus numbers assigned by the previous kernel. Bus
> +	 * numbers must remain stable for preserved devices so that they can
> +	 * perform DMA during the Live Update uninterrupted.
> +	 */
> +	if (pci_liveupdate_incoming_nr_devices())
> +		return false;

Following the comment on Patch 2 regarding propagating errors, the check
if (pci_liveupdate_incoming_nr_devices()) should be made explicit to 
distinguish between "Preservation Active" and "Retrieval Failed".

> +
> +	return pcibios_assign_all_busses();
> +}
> +
>  /*
>   * pci_scan_bridge_extend() - Scan buses behind a bridge
>   * @bus: Parent bus the bridge is on
> @@ -1378,6 +1392,7 @@ static int pci_scan_bridge_extend(struct pci_bus *bus, struct pci_dev *dev,
>  				  int max, unsigned int available_buses,
>  				  int pass)
>  {
> +	bool assign_all_busses = pci_assign_all_busses();
>  	struct pci_bus *child;
>  	int is_cardbus = (dev->hdr_type == PCI_HEADER_TYPE_CARDBUS);
>  	u32 buses, i, j = 0;
> @@ -1424,7 +1439,7 @@ static int pci_scan_bridge_extend(struct pci_bus *bus, struct pci_dev *dev,
>  	pci_write_config_word(dev, PCI_BRIDGE_CONTROL,
>  			      bctl & ~PCI_BRIDGE_CTL_MASTER_ABORT);
>  
> -	if ((secondary || subordinate) && !pcibios_assign_all_busses() &&
> +	if ((secondary || subordinate) && !assign_all_busses &&
>  	    !is_cardbus && !broken) {
>  		unsigned int cmax, buses;
>  
> @@ -1467,7 +1482,7 @@ static int pci_scan_bridge_extend(struct pci_bus *bus, struct pci_dev *dev,
>  		 * do in the second pass.
>  		 */
>  		if (!pass) {
> -			if (pcibios_assign_all_busses() || broken || is_cardbus)
> +			if (assign_all_busses || broken || is_cardbus)
>  
>  				/*
>  				 * Temporarily disable forwarding of the
> @@ -1542,7 +1557,7 @@ static int pci_scan_bridge_extend(struct pci_bus *bus, struct pci_dev *dev,
>  							max+i+1))
>  					break;
>  				while (parent->parent) {
> -					if ((!pcibios_assign_all_busses()) &&
> +					if (!assign_all_busses &&
>  					    (parent->busn_res.end > max) &&
>  					    (parent->busn_res.end <= max+i)) {
>  						j = 1;

Looks like we over-ride the pci=assign-busses boot param here. 
We should document how this change affects the pci=assign-busses kernel
command line. If both are present, the inheritance required by LUO would
likely take precedence to prevent DMA corruption, but a doc update & a 
warning to the user would be nice.

Thanks,
Praan