Re: 回复: 回复: Re: [PATCH v20 3/5] usb: misc: Add onboard_usb_hub driver

["mka@chromium.org" <mka@chromium.org>]

On Tue, Mar 01, 2022 at 02:30:00AM +0000, Linyu Yuan (QUIC) wrote:
> > From: mka@chromium.org <mka@chromium.org>
> > Sent: Tuesday, March 1, 2022 2:29 AM
> > To: Linyu Yuan (QUIC) <quic_linyyuan@quicinc.com>
> > Cc: gregkh@linuxfoundation.org; Tao Wang (Consultant) (QUIC)
> > <quic_wat@quicinc.com>; balbi@kernel.org; devicetree@vger.kernel.org;
> > dianders@chromium.org; frowand.list@gmail.com; hadess@hadess.net;
> > krzk@kernel.org; linux-kernel@vger.kernel.org; linux-usb@vger.kernel.org;
> > mathias.nyman@intel.com; michal.simek@xilinx.com;
> > peter.chen@kernel.org; ravisadineni@chromium.org; robh+dt@kernel.org;
> > rogerq@kernel.org; stern@rowland.harvard.edu; swboyd@chromium.org
> > Subject: Re: 回复: 回复: Re: [PATCH v20 3/5] usb: misc: Add
> > onboard_usb_hub driver
> > 
> > >
> > > Hi Greg and mka,
> > >
> > > Let's make it clear that we are talking about once this driver is approved
> > into usb tree,
> > > If we use different USB HUB which have VID/PID not defined in this driver,
> > > We need to update this driver.
> > >
> > > But if we defined VID/PID in device tree(for a specific board, manufacture
> > should know VID/PID from HUB it used),
> > > dynamic parsed by the driver,  then we don't need to change this driver
> > (increase VID/PID table).
> > 
> > As per my earlier reply, the kernel/USB core uses the VID:PID reported
> > by the USB device, the compatible string in the device tree is purely
> > informational. That's not something that could be changed by this
> > driver.
> I can't fully understand this comment,  could you please share step if we want to add a new HUB support, what should we do ? nothing ?

Add the VID:PID and compatible strings to onboard_usb_hub.c, analogous
to those for the RTS5411 and RTS5414. More work will be needed if the
hub needs a special power up or power down sequence (multiple regulators,
GPIOs, ...)

> If do nothing, can we remove id_table from  onboard_hub_usbdev_driver  ?
> > 
> > And even if the VID:PID from the device tree was used: how is the
> > kernel supposed to know that the onboard_hub driver should be
> > probed for a given VID:PID from the device tree, without listing
> > the VID:PID (or compatible string) in the driver (which is what
> > you seem to seek to avoid)?
> In my opinion, if it need update VID/PID table in this driver to support a new HUB,
> we can parse VID/PID from device tree and create dynamic VID/PID entry to id_table of onboard_hub_usbdev_driver.
> 
> Hope you can understand what I said.

Not really.

I doubt that what you are suggesting would work. The easiest thing
to convince people would probably be to send a patch (based on this
one) with a working implementation of your idea.