On Tue, Jun 11, 2019 at 09:29:23AM +0000, Regupathy, Rajaram wrote: > > > > -----Original Message----- > > From: Greg KH [mailto:gregkh@xxxxxxxxxxxxxxxxxxx] > > Sent: Monday, June 10, 2019 7:46 PM > > To: Mathias Nyman <mathias.nyman@xxxxxxxxxxxxxxx> > > Cc: Pandey, Prabhat Chand <prabhat.chand.pandey@xxxxxxxxx>; linux- > > usb@xxxxxxxxxxxxxxx; Nyman, Mathias <mathias.nyman@xxxxxxxxx>; > > Regupathy, Rajaram <rajaram.regupathy@xxxxxxxxx>; K V, Abhilash > > <abhilash.k.v@xxxxxxxxx>; Balaji, M <m.balaji@xxxxxxxxx> > > Subject: Re: [PATCH 4/5] usb: xhci: dbc: Add a dbc raw driver to provide a raw > > interface on DbC > > Importance: High > > > > On Mon, Jun 10, 2019 at 04:53:51PM +0300, Mathias Nyman wrote: > > > On 7.6.2019 17.21, Greg KH wrote: > > > > On Fri, Jun 07, 2019 at 12:03:05PM +0530, Prabhat Chand Pandey wrote: > > > > > From: Abhilash K V <abhilash.k.v@xxxxxxxxx> > > > > > > > > > > This patch provides a raw device interface on xhci Debug capability. > > > > > > > > What is a "raw device"? > > > > > > > > > This abstracts dbc functionality to user space inorder to > > > > > facilitate various frameworks to utilize xhci debug capability. > > > > > > > > I do not understand this sentance at all. Please provide a lot more > > > > information. > > > > > > > > > It helps to render the target as an usb debug class device on host > > > > > and establish an usb connection by providing two bulk endpoints. > > > > > > > > provide bulk endpoints where? To send data where? This is very > > > > confusing and does not make any sense to me... > > > > > > > > > > > > > > > > > > [don't dynamically allocate tiny space for name only -Mathias] > > > > > Signed-off-by: Rajaram Regupathy <rajaram.regupathy@xxxxxxxxx> > > > > > Signed-off-by: Prabhat Chand Pandey > > > > > <prabhat.chand.pandey@xxxxxxxxx> > > > > > Signed-off-by: Abhilash K V <abhilash.k.v@xxxxxxxxx> > > > > > Acked-by: Mathias Nyman <mathias.nyman@xxxxxxxxxxxxxxx> > > > > > --- > > > ... > > > > > > > > So you have a new char device, with a undocumented and unknown > > > > format of data flowing across it to the device. How in the world > > > > are we supposed to use this thing? Where is it documented? What > > > > does it do? How can you use it? > > We had captured all information in patch 5/5 patch in the documentation part. > We could always improve the documentation. Please let us know The documentation needs work, see my comments on that. Also, I don't think you answered these basic questions there, like "what is the data format", and "how is this supposed to be used". > > > > I don't mean to be so harsh here, but come on people, this stuff > > > > needs a lot more background documentation, information, and > > > > explaination as to exactly why in the world we need any of this, and what it > > even does! > > > > > > > > Also, you need to fix the code, it doesn't work as pointed out in a > > > > few places :) > > > > > > > > > > Thanks for going through this. > > > It's now clear this is far from ready. > > > I need to re-evaluate my position on this, not just the code and the > > > documentation, but the usefulness of it all. > > > > What is this even supposed to be used for? What is the application for it? I > > couldn't determine that at all, what am I missing? > > A typical use case is ADB for x86 Android systems or similar user > space class(debug) drivers that can leverage xHCI.DbC capability for > debug purpose. Why does adb need a "high speed" interface? And do you need special hardware to access this? Do you need patches on the adb side for this? > The larger goal here is to have DbC as a unified debug infrastructure for different debug methods like KGDB or early printk and leverage the benefits of a dedicated debug infra (DbC) brings in. Have you modified kgdb for this? Do you have patches for that? Who can use this interface in the "real world", is it only developers that have access to the special hardware dongle? Or can anyone use this on their laptops for getting console access in a way that is somehow "better" than the existing interface? And just how much "faster" is all of this than the current tty interface? What is lacking in the tty interface today that you need this new, special one? Can you just not fix any bottleneck in the tty driver if you are not properly saturating the bus? thanks, greg k-h
