Re: [PATCH] PCI: quirks: Skip nv_msi_ht_cap_quirk when host bridge 00:00.0 is absent

[Nicolas Chauvet]

Le sam. 24 oct. 2026 à 19:25, Karl Mehltretter
<kmehltretter@xxxxxxxxx> a écrit :
>
> __nv_msi_ht_cap_quirk() runs on all PCI_VENDOR_ID_NVIDIA devices to
> enable or disable HyperTransport MSI mapping relative to a host bridge
> at bus 0, slot 0, function 0.
>
> On NVIDIA Tegra SoCs (such as Tegra210 on Jetson Nano), the PCIe
> controller exposes root ports at 00:01.0 and 00:02.0 with no device at
> 00:00.0, and the root port's HT MSI mapping capability is already
> disabled (found == 1). Looking up 00:00.0 returns NULL and logs a
> false-positive warning on every boot:
>
>   pci 0000:00:02.0: nv_msi_ht_cap_quirk didn't locate host bridge
>
> Return silently when no 00:00.0 device exists on the domain, matching
> nv_ht_enable_msi_mapping().
>
> Assisted-by: LLM
> Signed-off-by: Karl Mehltretter <kmehltretter@xxxxxxxxx>
> ---
> Observed on a Jetson Nano 4GB (Tegra210, p3450-0000) booting v7.3
> (c03c1a14507d), where the Tegra PCIe controller (1003000.pcie)
> enumerates the active root port at 0000:00:02.0 ([10de:0faf]) with no
> 0000:00:00.0 device on bus 00:
>
>   pci 0000:00:02.0: [10de:0faf] type 01 class 0x060400 PCIe Root Port
>   pci 0000:00:02.0: nv_msi_ht_cap_quirk didn't locate host bridge

There is this patch that was well tested on the tegra side (but likely
less on the older x86 nv cases)
https://codeberg.org/libre-tegra/linux/commit/63f7810994f4d2866a27b753ec8d5bc36e84373c