linux-kernel - Re: [PATCH v2 2/2] PCI: Add PCI_BRIDGE_NO

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <918da2d4-dd80-4fd2-8767-679d22b6d929@nvidia.com>
Date: Thu, 18 Dec 2025 15:15:58 +0100
From: Nirmoy Das <nirmoyd@...dia.com>
To: Robin Murphy <robin.murphy@....com>, Bjorn Helgaas <bhelgaas@...gle.com>
CC: <linux-pci@...r.kernel.org>, <linux-kernel@...r.kernel.org>, "Jason
 Gunthorpe" <jgg@...dia.com>, Will Deacon <will@...nel.org>, Joerg Roedel
	<joro@...tes.org>, <iommu@...ts.linux.dev>, <jammy_huang@...eedtech.com>,
	<mochs@...dia.com>
Subject: Re: [PATCH v2 2/2] PCI: Add PCI_BRIDGE_NO_ALIASES quirk for ASPEED
 AST1150


On 17.12.25 19:07, Robin Murphy wrote:
> On 17/12/2025 3:45 pm, Nirmoy Das wrote:
>> ASPEED BMC controllers have VGA and USB functions behind a PCIe-to-PCI
>> bridge that causes them to share the same stream ID:
>>
>>    [e0]---00.0-[e1-e2]----00.0-[e2]--+-00.0  ASPEED Graphics Family
>>                                      \-02.0  ASPEED USB Controller
>>
>> Both devices get stream ID 0x5e200 due to bridge aliasing, causing the
>> USB controller to be rejected with 'Aliasing StreamID unsupported'.
>>
>> Per ASPEED, the AST1150 doesn't use a real PCI bus and always forwards
>> the original requester ID from downstream devices rather than replacing
>> it with any alias.
>>
>> Add a new PCI_DEV_FLAGS_PCI_BRIDGE_NO_ALIASES flag and apply it to the
>> AST1150.
>
> Looks reasonable to me;
>
> Reviewed-by: Robin Murphy <robin.murphy@....com>
>
> (Super-nit: perhaps s/ALIASES/ALIAS/ to neatly mirror 
> PCI_DEV_FLAG_PCIE_BRIDGE_ALIAS, which amusingly it's pretty much the 
> exact opposite of, but I'll leave that to Bjorn's discretion)


Thanks Robin! Happy to resend with s/ALIASES/ALIAS/ if Bjorn prefers.

>
>> Suggested-by: Jason Gunthorpe <jgg@...dia.com>
>> Signed-off-by: Nirmoy Das <nirmoyd@...dia.com>
>> ---
>> v2:
>>    - Use new PCI_DEV_FLAGS_PCI_BRIDGE_NO_ALIASES flag instead of
>>      PCI_DEV_FLAGS_BRIDGE_XLATE_ROOT to only skip aliasing at this
>>      bridge, not stop the entire upstream alias walk (Jason Gunthorpe)
>>
>>   drivers/pci/quirks.c | 10 ++++++++++
>>   drivers/pci/search.c |  2 ++
>>   include/linux/pci.h  |  5 +++++
>>   3 files changed, 17 insertions(+)
>>
>> diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c
>> index b9c252aa6fe0..a37b7305ae5f 100644
>> --- a/drivers/pci/quirks.c
>> +++ b/drivers/pci/quirks.c
>> @@ -4453,6 +4453,16 @@ 
>> DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_BROADCOM, 0x9000,
>>   DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_BROADCOM, 0x9084,
>>                   quirk_bridge_cavm_thrx2_pcie_root);
>>   +/*
>> + * AST1150 doesn't use a real PCI bus and always forwards the 
>> requester ID
>> + * from downstream devices.
>> + */
>> +static void quirk_aspeed_pci_bridge_no_aliases(struct pci_dev *pdev)
>> +{
>> +    pdev->dev_flags |= PCI_DEV_FLAGS_PCI_BRIDGE_NO_ALIASES;
>> +}
>> +DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_ASPEED, 0x1150, 
>> quirk_aspeed_pci_bridge_no_aliases);
>> +
>>   /*
>>    * Intersil/Techwell TW686[4589]-based video capture cards have an 
>> empty (zero)
>>    * class code.  Fix it.
>> diff --git a/drivers/pci/search.c b/drivers/pci/search.c
>> index 53840634fbfc..2f44444ae22f 100644
>> --- a/drivers/pci/search.c
>> +++ b/drivers/pci/search.c
>> @@ -86,6 +86,8 @@ int pci_for_each_dma_alias(struct pci_dev *pdev,
>>               case PCI_EXP_TYPE_DOWNSTREAM:
>>                   continue;
>>               case PCI_EXP_TYPE_PCI_BRIDGE:
>> +                if (tmp->dev_flags & 
>> PCI_DEV_FLAGS_PCI_BRIDGE_NO_ALIASES)
>> +                    continue;
>>                   ret = fn(tmp,
>>                        PCI_DEVID(tmp->subordinate->number,
>>                              PCI_DEVFN(0, 0)), data);
>> diff --git a/include/linux/pci.h b/include/linux/pci.h
>> index b16127c6a7b4..963da06ef193 100644
>> --- a/include/linux/pci.h
>> +++ b/include/linux/pci.h
>> @@ -248,6 +248,11 @@ enum pci_dev_flags {
>>       PCI_DEV_FLAGS_HAS_MSI_MASKING = (__force pci_dev_flags_t) (1 << 
>> 12),
>>       /* Device requires write to PCI_MSIX_ENTRY_DATA before any MSIX 
>> reads */
>>       PCI_DEV_FLAGS_MSIX_TOUCH_ENTRY_DATA_FIRST = (__force 
>> pci_dev_flags_t) (1 << 13),
>> +    /*
>> +     * PCIe to PCI bridge does not create RID aliases because the 
>> bridge is
>> +     * integrated with the downstream devices and doesn't use real PCI.
>> +     */
>> +    PCI_DEV_FLAGS_PCI_BRIDGE_NO_ALIASES = (__force pci_dev_flags_t) 
>> (1 << 14),
>>   };
>>     enum pci_irq_reroute_variant {
>