Re: [PATCH v2] PCI: Use downstream bridges for distributing resources

[Kai-Heng Feng <kaihengf@nvidia.com>]

On 2025/1/20 2:39 PM, Kai-Heng Feng wrote:
> 
> 
> On 2024/12/18 7:12 PM, Kai-Heng Feng wrote:
>> Hi Bjorn,
>>
>> On 2024/12/4 1:48 PM, Mika Westerberg wrote:
>>> External email: Use caution opening links or attachments
>>>
>>>
>>> On Wed, Dec 04, 2024 at 10:24:57AM +0800, Kai-Heng Feng wrote:
>>>> Commit 7180c1d08639 ("PCI: Distribute available resources for root
>>>> buses, too") breaks BAR assignment on some devcies:
>>>> [   10.021193] pci 0006:03:00.0: BAR 0 [mem 0x6300c0000000-0x6300c1ffffff 
>>>> 64bit pref]: assigned
>>>> [   10.029880] pci 0006:03:00.1: BAR 0 [mem 0x6300c2000000-0x6300c3ffffff 
>>>> 64bit pref]: assigned
>>>> [   10.038561] pci 0006:03:00.2: BAR 0 [mem size 0x00800000 64bit pref]: 
>>>> can't assign; no space
>>>> [   10.047191] pci 0006:03:00.2: BAR 0 [mem size 0x00800000 64bit pref]: 
>>>> failed to assign
>>>> [   10.055285] pci 0006:03:00.0: VF BAR 0 [mem size 0x02000000 64bit pref]: 
>>>> can't assign; no space
>>>> [   10.064180] pci 0006:03:00.0: VF BAR 0 [mem size 0x02000000 64bit pref]: 
>>>> failed to assign
>>>> [   10.072543] pci 0006:03:00.1: VF BAR 0 [mem size 0x02000000 64bit pref]: 
>>>> can't assign; no space
>>>> [   10.081437] pci 0006:03:00.1: VF BAR 0 [mem size 0x02000000 64bit pref]: 
>>>> failed to assign
>>>>
>>>> The apertures of domain 0006 before the commit:
>>>> 6300c0000000-63ffffffffff : PCI Bus 0006:00
>>>>    6300c0000000-6300c9ffffff : PCI Bus 0006:01
>>>>      6300c0000000-6300c9ffffff : PCI Bus 0006:02
>>>>        6300c0000000-6300c8ffffff : PCI Bus 0006:03
>>>>          6300c0000000-6300c1ffffff : 0006:03:00.0
>>>>            6300c0000000-6300c1ffffff : mlx5_core
>>>>          6300c2000000-6300c3ffffff : 0006:03:00.1
>>>>            6300c2000000-6300c3ffffff : mlx5_core
>>>>          6300c4000000-6300c47fffff : 0006:03:00.2
>>>>          6300c4800000-6300c67fffff : 0006:03:00.0
>>>>          6300c6800000-6300c87fffff : 0006:03:00.1
>>>>        6300c9000000-6300c9bfffff : PCI Bus 0006:04
>>>>          6300c9000000-6300c9bfffff : PCI Bus 0006:05
>>>>            6300c9000000-6300c91fffff : PCI Bus 0006:06
>>>>            6300c9200000-6300c93fffff : PCI Bus 0006:07
>>>>            6300c9400000-6300c95fffff : PCI Bus 0006:08
>>>>            6300c9600000-6300c97fffff : PCI Bus 0006:09
>>>>
>>>> After the commit:
>>>> 6300c0000000-63ffffffffff : PCI Bus 0006:00
>>>>    6300c0000000-6300c9ffffff : PCI Bus 0006:01
>>>>      6300c0000000-6300c9ffffff : PCI Bus 0006:02
>>>>        6300c0000000-6300c43fffff : PCI Bus 0006:03
>>>>          6300c0000000-6300c1ffffff : 0006:03:00.0
>>>>            6300c0000000-6300c1ffffff : mlx5_core
>>>>          6300c2000000-6300c3ffffff : 0006:03:00.1
>>>>            6300c2000000-6300c3ffffff : mlx5_core
>>>>        6300c4400000-6300c4dfffff : PCI Bus 0006:04
>>>>          6300c4400000-6300c4dfffff : PCI Bus 0006:05
>>>>            6300c4400000-6300c45fffff : PCI Bus 0006:06
>>>>            6300c4600000-6300c47fffff : PCI Bus 0006:07
>>>>            6300c4800000-6300c49fffff : PCI Bus 0006:08
>>>>            6300c4a00000-6300c4bfffff : PCI Bus 0006:09
>>>>
>>>> We can see that the window of 0006:03 gets shrunken too much and 0006:04
>>>> eats away the window for 0006:03:00.2.
>>>>
>>>> The offending commit distributes the upstream bridge's resources
>>>> multiple times to every downstream bridges, hence makes the aperture
>>>> smaller than desired because calculation of io_per_b, mmio_per_b and
>>>> mmio_pref_per_b becomes incorrect.
>>>>
>>>> Instead, distributing downstream bridges' own resources to resolve the
>>>> issue.
>>>>
>>>> Link: https://bugzilla.kernel.org/show_bug.cgi?id=219540
>>>> Cc: Carol Soto <csoto@nvidia.com>
>>>> Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
>>>> Cc: Chris Chiu <chris.chiu@canonical.com>
>>>> Cc: Mika Westerberg <mika.westerberg@linux.intel.com>
>>>
>>> Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com>
>>
>> Is it possible to pickup this up in 6.13 window?
> 
> Now 6.14 window is open, is it possible to include this fix?

Another gentle ping...

> 
> Kai-Heng
> 
>>
>> Kai-Heng
>