linux-kernel - Re: [PATCH 06/15] usb: serial: f81232: use usb_control_msg_recv() and usb_control_msg

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <X8oGN9G+i9rQM9eX@localhost>
Date:   Fri, 4 Dec 2020 10:49:43 +0100
From:   Johan Hovold <johan@...nel.org>
To:     Himadri Pandya <himadrispandya@...il.com>
Cc:     johan@...nel.org, gregkh@...uxfoundation.org,
        linux-usb@...r.kernel.org, linux-kernel@...r.kernel.org,
        linux-kernel-mentees@...ts.linuxfoundation.org
Subject: Re: [PATCH 06/15] usb: serial: f81232: use usb_control_msg_recv()
 and usb_control_msg_send()

On Wed, Nov 04, 2020 at 12:16:54PM +0530, Himadri Pandya wrote:
> The new usb_control_msg_recv() and usb_control_msg_send() nicely wraps
> usb_control_msg() with proper error check. Hence use the wrappers
> instead of calling usb_control_msg() directly.
> 
> Signed-off-by: Himadri Pandya <himadrispandya@...il.com>
> ---
>  drivers/usb/serial/f81232.c | 88 ++++++++-----------------------------
>  1 file changed, 18 insertions(+), 70 deletions(-)
> 
> diff --git a/drivers/usb/serial/f81232.c b/drivers/usb/serial/f81232.c
> index 0c7eacc630e0..78107feef8a8 100644
> --- a/drivers/usb/serial/f81232.c
> +++ b/drivers/usb/serial/f81232.c
> @@ -139,71 +139,36 @@ static int calc_baud_divisor(speed_t baudrate, speed_t clockrate)
>  static int f81232_get_register(struct usb_serial_port *port, u16 reg, u8 *val)
>  {
>  	int status;
> -	u8 *tmp;
>  	struct usb_device *dev = port->serial->dev;
>  
> -	tmp = kmalloc(sizeof(*val), GFP_KERNEL);
> -	if (!tmp)
> -		return -ENOMEM;
> -
> -	status = usb_control_msg(dev,
> -				usb_rcvctrlpipe(dev, 0),
> -				F81232_REGISTER_REQUEST,
> -				F81232_GET_REGISTER,
> -				reg,
> -				0,
> -				tmp,
> -				sizeof(*val),
> -				USB_CTRL_GET_TIMEOUT);
> -	if (status != sizeof(*val)) {
> +	status = usb_control_msg_recv(dev, 0, F81232_REGISTER_REQUEST,
> +				      F81232_GET_REGISTER, reg, 0, val,
> +				      sizeof(*val), USB_CTRL_GET_TIMEOUT,
> +				      GFP_KERNEL);

This driver use tabs for indentation consistently so please stick to
that style. That should also amount to a smaller patch (and less noise
in the logs, e.g. for git blame).

> +	if (status) {
>  		dev_err(&port->dev, "%s failed status: %d\n", __func__, status);
>  
> -		if (status < 0)
> -			status = usb_translate_errors(status);
> -		else
> -			status = -EIO;
> -	} else {
> -		status = 0;
> -		*val = *tmp;
> +		status = usb_translate_errors(status);
>  	}
>  
> -	kfree(tmp);
>  	return status;
>  }

> @@ -866,32 +831,16 @@ static int f81534a_ctrl_set_register(struct usb_interface *intf, u16 reg,
>  	struct usb_device *dev = interface_to_usbdev(intf);
>  	int retry = F81534A_ACCESS_REG_RETRY;
>  	int status;
> -	u8 *tmp;
> -
> -	tmp = kmemdup(val, size, GFP_KERNEL);
> -	if (!tmp)
> -		return -ENOMEM;
>  
>  	while (retry--) {
> -		status = usb_control_msg(dev,
> -					usb_sndctrlpipe(dev, 0),
> -					F81232_REGISTER_REQUEST,
> -					F81232_SET_REGISTER,
> -					reg,
> -					0,
> -					tmp,
> -					size,
> -					USB_CTRL_SET_TIMEOUT);
> -		if (status < 0) {
> +		status = usb_control_msg_send(dev, 0, F81232_REGISTER_REQUEST,
> +					      F81232_SET_REGISTER, reg, 0, val,
> +					      size, USB_CTRL_SET_TIMEOUT, GFP_KERNEL);
> +		if (status) {
> +			/* Retry on error or short transfers */
>  			status = usb_translate_errors(status);
>  			if (status == -EIO)
>  				continue;

So this now depends on the new helper returning an errno on short
transfers which gets mapped to -EIO by usb_translate_errors(). This
works currently (and with my suggested change to use -EREMOTEIO) but
could possibly lead to subtle regressions later.

I don't think we need to worry about it, but just wanted to highlight
the kind of impact these changes can have.

Note that this change also replaces a single memdup with a memdup+memcpy
for each iteration. But as this helper is used rarely (e.g. probe,
resume and disconnect), and hopefully the retries are exceptions, the
overhead could be acceptable.

> -		} else if (status != size) {
> -			/* Retry on short transfers */
> -			status = -EIO;
> -			continue;
> -		} else {
> -			status = 0;
>  		}
>  
>  		break;
> @@ -902,7 +851,6 @@ static int f81534a_ctrl_set_register(struct usb_interface *intf, u16 reg,
>  				reg, status);
>  	}
>  
> -	kfree(tmp);
>  	return status;
>  }

Johan