Re: [PATCH 1/2] usb: serial: Add MaxLinear/Exar USB to Serial driver

[Manivannan Sadhasivam <mani@kernel.org>]

On Wed, Apr 29, 2020 at 11:29:04AM +0200, Greg KH wrote:
> On Wed, Apr 29, 2020 at 01:10:26PM +0530, Manivannan Sadhasivam wrote:
> > Hi Greg,
> > 
> > On Wed, Apr 29, 2020 at 09:20:36AM +0200, Greg KH wrote:
> > > On Wed, Apr 29, 2020 at 01:26:50AM +0530, mani@kernel.org wrote:
> > > > From: Manivannan Sadhasivam <mani@kernel.org>
> > > > 
> > > > Add support for MaxLinear/Exar USB to Serial converters. This driver
> > > > only supports XR21V141X series but provision has been made to support
> > > > other series in future.
> > > > 
> > > > This driver is inspired from the initial one submitted by Patong Yang:
> > > > 
> > > > https://patchwork.kernel.org/patch/10543261/
> > > > 
> > > > While the initial driver was a custom tty USB driver exposing whole
> > > > new serial interface ttyXRUSBn, this version is completely based on USB
> > > > serial core thus exposing the interfaces as ttyUSBn. This will avoid
> > > > the overhead of exposing a new USB serial interface which the userspace
> > > > tools are unaware of.
> > > 
> > > Nice work!
> > > 
> > > Some comments below:
> > > 
> > > > +// SPDX-License-Identifier: GPL-2.0+
> > > > +/*
> > > > + * MaxLinear/Exar USB to Serial driver
> > > > + *
> > > > + * Based on initial driver written by Patong Yang <patong.mxl@gmail.com>
> > > > + *
> > > > + * Copyright (c) 2020 Manivannan Sadhasivam <mani@kernel.org>
> > > > + */
> > > > +
> > > > +#include <linux/kernel.h>
> > > > +#include <linux/module.h>
> > > > +#include <linux/slab.h>
> > > > +#include <linux/tty.h>
> > > > +#include <linux/usb.h>
> > > > +#include <linux/usb/serial.h>
> > > > +
> > > > +#include "xr_serial.h"
> > > 
> > > No need for a .h file for a single .c file.
> > > 
> > 
> > Yeah but since this driver can support multiple series of XR chips (they
> > might have separate register definitions and such), I thought it is a good
> > idea to have a header file to keep the driver sane. But can club it to the
> > source file for now.
> 
> Don't worry about future stuff, focus on what you need to do now :)
> 

Alright, will do :)

> > > > +static int xr_get_reg(struct usb_serial_port *port, u8 block, u16 reg,
> > > > +		      u16 *val)
> > > > +{
> > > > +	struct usb_serial *serial = port->serial;
> > > > +	struct xr_port_private *port_priv = usb_get_serial_port_data(port);
> > > > +	void *dmabuf;
> > > > +	int ret = -EINVAL;
> > > > +
> > > > +	dmabuf = kmalloc(sizeof(reg), GFP_KERNEL);
> > > 
> > > So that is 2 bytes?
> > > 
> > 
> > Explanation below...
> > 
> > > > +	if (!dmabuf)
> > > > +		return -ENOMEM;
> > > > +
> > > > +	if (port_priv->idProduct == XR21V141X_ID) {
> > > > +		/* XR21V141X uses custom command for reading UART registers */
> > > > +		ret = usb_control_msg(serial->dev,
> > > > +				      usb_rcvctrlpipe(serial->dev, 0),
> > > > +				      XR_GET_XR21V141X,
> > > > +				      USB_DIR_IN | USB_TYPE_VENDOR, 0,
> > > > +				      reg | (block << 8), dmabuf,
> > > > +				      port_priv->reg_width,
> > > > +				      USB_CTRL_SET_TIMEOUT);
> > > > +	}
> > > > +
> > > > +	if (ret == port_priv->reg_width) {
> > > > +		memcpy(val, dmabuf, port_priv->reg_width);
> > > 
> > > But here you copy ->reg_width bytes in?  How do you know val can hold
> > > that much?  It's only set to be 1, so you copy 1 byte to a 16bit value?
> > > What part of the 16bits did you just copy those 8 bits to (hint, think
> > > cpu endian issues...)
> > > 
> > > That feels really really odd and a bit broken.
> > > 
> > 
> > Right. The reason is, the other series which can be supported by this driver
> > have different register widths. For instance XR2280x. I haven't used them
> > personally but seen this in initial driver. So I just used the max u16 type
> > to make the reg_{set/get} routines work with those.
> 
> Drop the whole "different register width" stuff for now, as the driver
> does not support it and it adds additional complexity that is hard to
> review for no good reason.  If you want to add support for new devices
> later, _then_ we can add support for that.
> 
> Don't over-engineer :)
> 

Sure!

> > But agree, I should've used le16_to_cpu() cast to avoid endian issues.
> 
> You have to, the code is broken as-is right now.
> 

We don't need the conversion if u8 is used everywhere :)

Thanks,
Mani

> > If you think this hack is not required now, I can just use u8 and worry about
> > compatibility later.
> 
> Please do so.
> 
> thanks,
> 
> greg k-h