Re: [PATCH v2 for-4.20? 6/6] PCI: drop pci_segments_init()

[Jan Beulich <jbeulich@xxxxxxxx>]

On 04.02.2025 08:45, Jan Beulich wrote:
> On 03.02.2025 18:18, Roger Pau Monné wrote:
>> On Mon, Feb 03, 2025 at 05:27:24PM +0100, Jan Beulich wrote:
>>> --- a/xen/arch/x86/x86_64/mmconfig-shared.c
>>> +++ b/xen/arch/x86/x86_64/mmconfig-shared.c
>>> @@ -402,6 +402,9 @@ void __init acpi_mmcfg_init(void)
>>>  {
>>>      bool valid = true;
>>>  
>>> +    if ( pci_add_segment(0) )
>>> +        panic("Could not initialize PCI segment 0\n");
>>
>> Do you still need the pci_add_segment() call here?
>>
>> pci_add_device() will already add the segments as required, and
>> acpi_parse_mcfg() does also add the segments described in the MCFG.
> 
> Well, in principle you're right. It's more the action in case of
> failure that makes me want to keep it: We won't fare very well on
> about every system if we can't register segment 0.

Thinking about it: Your question may be more applicable on Arm, as I'm
entirely uncertain whether there segment 0 is always going to be needed.
I could well imagine system designers deliberately putting all the
devices elsewhere. Segment 0 always being in use on x86 is more a
heritage thing, after all.

Jan