Re: [PATCH v2 1/3] usb: core: Fix root hub descriptor wBytesPerInterval

[Mathias Nyman <mathias.nyman@linux.intel.com>]

On 5/6/26 13:31, Michal Pecio wrote:
> On Wed, 6 May 2026 13:17:46 +0300, Mathias Nyman wrote:
>> On 5/4/26 12:13, Michal Pecio wrote:
>>> Per USB3 9.6.7, it's "the total number of bytes this endpoint will
>>> transfer every service interval". There seems to be no good reason
>>> to have wBytesPerInterval < wMaxPacketSize - either one is too low
>>> or the other too high. Here, wBytesPerInterval is too low for hubs
>>> with more than 15 ports and xHCI spec allows such root hubs.
>>>    
>>
>> There shouldn't be a USB3 hub with more than 15 ports as
>> USB 3.x specification limits USB3 ports to 15.
>>
>> See USB 3.2 section 10.15.2.1 Hub Descriptor
>> "bNbrPorts:	Number of downstream facing ports that this hub supports. The
>> 		maximum number of ports of ports a hub can support is 15."
>>
>>
>> Hub driver also fails if hub has more than 15 ports.
>> hub.c hub_configure():
>>
>>      maxchild = USB_MAXCHILDREN;
>>           if (hub_is_superspeed(hdev))
>>                   maxchild = min_t(unsigned, maxchild, USB_SS_MAXPORTS);
>>
>>           if (hub->descriptor->bNbrPorts > maxchild) {
>>                   message = "hub has too many ports!";
>>                   ret = -ENODEV;
>>                   goto fail;
>>
>> ch11.h:#define USB_SS_MAXPORTS           15
>>
>> The USB 3.2 specification 10.15.1 'Standard Descriptors for Hub Class' also
>> shows hardcoded values of '2' for both the wBytesPerInterval and wMaxPacketSize
>> for the hub interrupt endpoint.
> 
> Thanks, looks like we should be able to get away with reducing
> wMaxPacketSize then. Currently it's calculated to fit USB_MAXCHILDREN
> i.e. 31 ports, supposedly dictated by the needs of (obsolete) WUSB.
> 
> Any preferences whether to replace USB_MAXCHILDREN with USB_SS_MAXPORTS
> in the existing formula for USB3 hubs or just hardcode 0x02?

Maybe prefer the hardcoded 0x02 value as spec does it as well

Thanks
Mathias