linux-kernel - Re: [PATCH] usb: gadget: ffs: don't allow to open with O

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: <xa1tlhi49hnw.fsf@mina86.com>
Date:	Tue, 07 Apr 2015 15:44:51 +0200
From:	Michal Nazarewicz <mina86@...a86.com>
To:	Robert Baldyga <r.baldyga@...sung.com>, balbi@...com
Cc:	gregkh@...uxfoundation.org, linux-usb@...r.kernel.org,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH] usb: gadget: ffs: don't allow to open with O_NONBLOCK flag

On Fri, Apr 03 2015, Robert Baldyga <r.baldyga@...sung.com> wrote:
> Hi Michal,
>
> On 04/01/2015 05:17 PM, Michal Nazarewicz wrote:
>> On Wed, Apr 01 2015, Robert Baldyga <r.baldyga@...sung.com> wrote:
>>> FunctionFS can't support O_NONBLOCK because read/write operatons are
>>> directly translated into USB requests which are asynchoronous, so we
>>> can't know how long we will have to wait for request completion. For
>>> this reason in case of open with O_NONBLOCK flag we return
>>> -EWOULDBLOCK.
>> 
>> ‘can’t’ is a bit strong of a word here though.  It can, but in a few
>> cases it doesn’t.
>> 
>> It kinda saddens me that this undoes all the lines of code that were put
>> into the file to support O_NONBLOCK (e.g. FFS_NO_SETUP path of
>> ffs_ep0_read).
>> 
>> I’m also worried this may break existing applications which, for better
>> or worse, open the file with O_NONBLOCK.
>> 
>> Most importantly though, this does not stop users from using fcntl to
>> set O_NONBLOCK, so if you really want to stop O_NONBLOCK from being set,
>> that path should be checked as well (if possible).
>
> I want rather to inform users that non-blocking i/o wouldn't work for
> epfiles. Indeed we can handle O_NONBLOCK for ep0 (for the same reason we
> can have poll), but for other epfiles there is no way to check if
> read/write operation can end up in short time.

There is potentially a way to implement O_NONBLOCK for epfiles.  This
would require adding a new state property to epfile and moving
completion structure from stack to epfile.  In pseudo code we would
have:

epfile->state = FREE

epfile_io(direction)
	if epfile->state == FREE
		queue(ep->req)
		epfile->state = PENDING

	if epfile->state == PENDING
		if O_NONBLOCK
			return -EAGAIN
		wait
		epfile->state = FINISHED

	// epfile->state == FINISHED
	copy data
	epfile->state = FREE
	return

I think this is the only ‘technically correct’ solution, but I dunno if
it is worth implementing especially since AIO is available.
        
-- 
Best regards,                                         _     _
.o. | Liege of Serenely Enlightened Majesty of      o' \,=./ `o
..o | Computer Science,  Michał “mina86” Nazarewicz    (o o)
ooo +--<mpn@...gle.com>--<xmpp:mina86@...ber.org>--ooO--(_)--Ooo--
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/