The ifdef'ed Linux kernel code uses the 1 based port number, whereas U-Boot puts a 0 based port number into the register. The reason the 0 based port number apparently works can probably be taken from the USB 2.0 spec:
8.4.2.2 Start-Split Transaction Token ... The host must correctly set the port field for single and multiple TT hub implementations. A single TT hub implementation *may ignore* the port field.
Actually, as far as I understand, a multi TT hub defaults to single TT (bAlternateSetting: 0) until switched via SetInterface, so even "port 42" would work.
The change was verified by hardcoding the port number to a wrong value, SPLIT transactions kept working (although using a DWC2 instead of MUSB). Tested hubs are the RPi onboard SMC9514 and an external "05e3:0608 Genesys Logic, Inc. USB-2.0 4-Port HUB". The former is a multi TT hub, the latter single TT only.
Addendum: Tested on sunxi/MUSB by Hans de Goede
Signed-off-by: Stefan Brüns stefan.bruens@rwth-aachen.de Reviewed-by: Hans de Goede hdegoede@redhat.com Tested-by: Hans de Goede hdegoede@redhat.com --- v2: - renamed function to usb_find_usb2_hub_address_port() - put musb port numbering change into separate patch v3: - do one assignment per line v4: - fix error in musb code (udev => urb->dev) - added Reviewed-by: et al
drivers/usb/musb-new/musb_host.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/usb/musb-new/musb_host.c b/drivers/usb/musb-new/musb_host.c index 70f8a99..2515447 100644 --- a/drivers/usb/musb-new/musb_host.c +++ b/drivers/usb/musb-new/musb_host.c @@ -2098,7 +2098,7 @@ int musb_urb_enqueue( &hubaddr, &portnr); qh->h_addr_reg = hubaddr; - qh->h_port_reg = portnr - 1; + qh->h_port_reg = portnr; } #endif }