On 23-09-07 14:33:29, tomer.maimon@xxxxxxxxxxx wrote: > Hi Peter > > My name is Tomer I am working in Nuvoton system in the Linux kernel, our project developing BMC SoC for servers. > In Nuvoton BMC (NPCM) there are ten identical Synopsys USB 2.0 device controllers called NPCM UDC. > > Lately we started to work with USB Chipidea driver to add NPCM UDC as well. > The NPCM BMC using only the UDC and not the USB host. > > Using the default compatible<https://elixir.bootlin.com/linux/v5.15.130/C/ident/compatible> = "chipidea,usb2<https://elixir.bootlin.com/linux/v5.15.130/B/ident/chipidea%2Cusb2>" didn't work for us since: > > 1. The USB_MODE should be set at boot or after reset since the reset The USB_MODE is incorrect (0x15002) > > [cid:image002.jpg@01D9E1B1.691030B0] > > > It solved by setting USB_MODE during NPCM reset ci_hdrc_npcm_notify_event, now using unique npcm-udc compatible :) > > > > 2. vbus_active parameter don't change, stay 0. > > The Device Control Capability Parameters Register (DCCPARAMS) is a read only register that indicate the module is only DC (Device Capable) > > This is why the driver didn't indicate the driver is OTG > > https://elixir.bootlin.com/linux/v6.5.2/source/drivers/usb/chipidea/core.c#L948 > > ci<https://elixir.bootlin.com/linux/v6.5.2/C/ident/ci>->is_otg<https://elixir.bootlin.com/linux/v6.5.2/C/ident/is_otg> = (hw_read<https://elixir.bootlin.com/linux/v6.5.2/C/ident/hw_read>(ci<https://elixir.bootlin.com/linux/v6.5.2/C/ident/ci>, CAP_DCCPARAMS<https://elixir.bootlin.com/linux/v6.5.2/C/ident/CAP_DCCPARAMS>, > > DCCPARAMS_DC<https://elixir.bootlin.com/linux/v6.5.2/C/ident/DCCPARAMS_DC> | DCCPARAMS_HC<https://elixir.bootlin.com/linux/v6.5.2/C/ident/DCCPARAMS_HC>) > > == (DCCPARAMS_DC<https://elixir.bootlin.com/linux/v6.5.2/C/ident/DCCPARAMS_DC> | DCCPARAMS_HC<https://elixir.bootlin.com/linux/v6.5.2/C/ident/DCCPARAMS_HC>)); > > Why otg is set only when DC and HC is set? > > By enabling ci<https://elixir.bootlin.com/linux/v6.5.2/C/ident/ci>->is_otg<https://elixir.bootlin.com/linux/v6.5.2/C/ident/is_otg> = true we see we need to set the extcon in the device tree, why we need extcon to handle the vbus? Can the vbus be permanent? > > > > Even after setting extcon vbus to dummy GPIO we succeed to modify the vbus_active parameter to is_active but it didn't worked in the end because it a dummy GPIO that not related to the USB vbus. > > > > BTW, > > If we adding ci->vbus_active = true at probe stage the UDC Chipidea driver works fine. > > https://elixir.bootlin.com/linux/v6.5.2/source/drivers/usb/chipidea/core.c#L1123 > > > > Appreciate if you could you advise how should we overcome this issue After loading the gadget driver, try to force "connect" udc using /sys entry. See the below code at file: drivers/usb/gadget/udc/core.c static ssize_t soft_connect_store(struct device *dev, struct device_attribute *attr, const char *buf, size_t n) Your issue is probably due to the UDC don't know it is connected, so DP is not pulled up. -- Thanks, Peter Chen
