netdev - RE: [RFC 1/4] Add usb_get_address and usb_set

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] [day] [month] [year] [list]

Date:   Thu, 22 Aug 2019 17:14:30 +0000
From:   <Charles.Hyde@...lteam.com>
To:     <oneukum@...e.com>, <gregkh@...uxfoundation.org>
CC:     <Mario.Limonciello@...l.com>, <nic_swsd@...ltek.com>,
        <linux-acpi@...r.kernel.org>, <linux-usb@...r.kernel.org>,
        <netdev@...r.kernel.org>
Subject: RE: [RFC 1/4] Add usb_get_address and usb_set_address support

> > <snipped>
> > >
> > > This is a VERY cdc-net-specific function.  It is not a "generic" USB
> > > function at all.  Why does it belong in the USB core?  Shouldn't it
> > > live in the code that handles the other cdc-net-specific logic?
> > >
> > > thanks,
> > >
> > > greg k-h
> >
> >
> > Thank you for this feedback, Greg.  I was not sure about adding this to
> message.c, because of the USB_CDC_GET_NET_ADDRESS.  I had found
> references to SET_ADDRESS in the USB protocol at
> https://wiki.osdev.org/Universal_Serial_Bus#USB_Protocol.  If one wanted a
> generic USB function for SET_ADDRESS, to be used for both sending a MAC
> address and receiving one, how would you suggest this be implemented?  This
> is a legit question because I am curious.
> 
> Your implementation was, except for missing error handling, usable.
> The problem is where you put it. CDC messages exist only for CDC devices. Now
> it is true that there is no generic CDC driver.
> Creating a module just for that would cost more memory than it saves in most
> cases.
> But MACs are confined to network devices. Hence the functionality can be put
> into usbnet. It should not be put into any individual driver, so that every
> network driver can use it without duplication.
> 
> > Your feedback led to moving the functionality into cdc_ncm.c for today's
> testing, and removing all changes from messages.c, usb.h, usbnet.c, and
> usbnet.h.  This may be where I end up long term, but I would like to learn if
> there is a possible solution that could live in message.c and be callable from
> other USB-to-Ethernet aware drivers.
> 
> All those drivers use usbnet. Hence there it should be.
> 
> 	Regards
> 		Oliver


Some of the drivers in drivers/net/usb/ do call functions in drivers/net/usb/usbnet, but not all.  As Greg pointed out, the USB change I developed is cdc specific, so putting it into usbnet would raise the same concerns Greg mentioned.  Leaving my newest implementation in cdc_ncm.c will be most appropriate, as it also fits with what other drivers in this folder have done.  My original code was rather short sighted, at best.

Charles