Re: [RFC PATCH] PCI: Get the hotplug bridge window sizes from DT

[sashiko-bot@kernel.org]

> Most of the PCI hotplug bridges used in Devicetree platforms do not have
> their bridge windows populated by firmware. So the PCI core allocates the
> bridge windows based on the requirements below:
> 
> 1. Total size needed by the downstream devices
> 2. Default size (2MiB for MMIO, 256 bytes for IO) if no devices are
>    available during the initial bus scan
> 3. Command line params like hpmemsize etc...
> 
> If no devices are available during the initial bus scan, then the PCI core
> allocates the default conservative 2MiB MMIO bridge windows. But this size
> sometimes becomes insufficient when a BAR hungry device like a GPU gets
> plugged in later. So users have to resort to command line params to
> increase the window sizes. But this is very inconvenient and causes a lot
> of pain in distro integration.
> 
> To address this issue, parse the 'ranges' property from the bridge DT node
> (if available) and use the sizes for each bridge window. This works well
> because the 'ranges' property is already allowed by the DT binding for PCI
> bridges and it lets system designers pre-define the hotplug bridge window
> size requirements in firmware.
> 
> The DT derived sizes are applied as the additional (hotplug) reservation for
> each window. Command line parameters (hpmemsize etc...) continue to take
> precedence over DT.
> 
> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@oss.qualcomm.com>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260706123133.545104-1-manivannan.sadhasivam@oss.qualcomm.com?part=1