linux-kernel - Re: [PATCH 1/2] usb: configfs: allow UDC binding rule configured as binding to *any* UDC

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]

Date:	Thu, 05 May 2016 09:31:33 +0200
From:	Krzysztof Opasiak <k.opasiak@...sung.com>
To:	"Du, Changbin" <changbin.du@...el.com>,
	"balbi@...nel.org" <balbi@...nel.org>
Cc:	"gregkh@...uxfoundation.org" <gregkh@...uxfoundation.org>,
	"corbet@....net" <corbet@....net>,
	"linux-usb@...r.kernel.org" <linux-usb@...r.kernel.org>,
	"linux-doc@...r.kernel.org" <linux-doc@...r.kernel.org>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	"Du, Changbin" <changbin.du@...il.com>,
	"Du@...r.kernel.org" <Du@...r.kernel.org>
Subject: Re: [PATCH 1/2] usb: configfs: allow UDC binding rule configured as
 binding to *any* UDC

Hi,

On 05/05/2016 07:46 AM, Du, Changbin wrote:
> Hi,
>>> On most platforms, there is only one device controller available.
>>> In this case, we desn't care the UDC's name. So let's ignore the
>>> name by setting 'UDC' to 'any'.
>>
>> Hmm libubsgx allows to do this for a very long time. You simply pass
>> NULL instead of pointer to usbg_udc.
>>
>> It is also possible to do this from command line, just simply:
>>
>> $ echo `ls -1 /sys/class/udc | head -n 1` > UDC
>>
>> So if we can easily do this from user space what's the benefit of adding
>> this special "any" keyword to kernel?
>>
> Well, it is just for *easy to use*. Looking up /sys/class/udc mostly
> can be skipped. The UDC core support this convenience behavior, 
> so why don't we export it with a little change?
> 

Well, I'm not sure if any configfs interface has been proposed as easy
to use from cmd line. I think they all has been proposed as  *usable*
from cmd line but not necessarily *easy to use*.

That's why most of configfs clients has some support in userspace. For
example for target there is a taget-cli and for usb gadget we have
libusbg/libusbgx.

So the functionality which you proposed here is not only already
implemented in libusbgx but also can be easily achieved from cmd line
like I showed above.

In addition this patch will break existing userspace tools (at least
libusbgx for sure) as it assumes that:

cat UDC

should return an empty string or an valid UDC name which can be found
inside /sys/class/udc.

After this patch the kernel can return some kind of magic string "any"
which obviously will cannot be found in udc dir.

>>> And also we can change UDC name
>>> at any time if it is not binded (no need set to "" first).
>>>
>>
>> Not sure if:
>>
>> $ echo "" > UDC
>>
>> is really a problem. Personally I'm quite used to situation in which I
>> have to turn the light off before turning it on once again;)
>>
> That is not a problem. But just avoid pseudo 'busy'. If gadget is not 
> bind, it is free to reconfigure it. So seem no need block re-configuration.
> 

What do you mean pseudo 'busy'? If we do:

echo <udc-name> > UDC

then gadget should be really bound to some udc and potentially really busy.

> In a word, this patch is just an improvement, not to fix any issues or
> add new function.

So it doesn't add any new functionality and breaks existing user space
tools.

Cheers,
-- 
Krzysztof Opasiak
Samsung R&D Institute Poland
Samsung Electronics