linux-kernel - Re: debugfs_remove_recursive() while a file is in use by userspace

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: <CAA93t1o9We6EFgvJ+7dPndptcurp=ZTw=jfb0ek_hhmB21tRkA@mail.gmail.com>
Date:	Mon, 28 Dec 2015 13:11:53 -0800
From:	Rajat Jain <rajatxjain@...il.com>
To:	Greg Kroah-Hartman <greg@...ah.com>
Cc:	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	linux-newbie@...r.kernel.org
Subject: Re: debugfs_remove_recursive() while a file is in use by userspace

On Mon, Dec 28, 2015 at 12:58 PM, Greg Kroah-Hartman <greg@...ah.com> wrote:
> On Mon, Dec 28, 2015 at 12:51:32PM -0800, Rajat Jain wrote:
>> Thanks Greg and Al for the quick turnaround.
>>
>> Essentially I have a device that supports something called "contexts"
>> that can be "created" and "destroyed" during the life of the device. I
>> want to expose some debug files for the context when it is created,
>> and destroy the files when the context is destroyed. However, I'm not
>> sure how do I ensure that the user is not in the middle of reading /
>> writing / mmaping to those files. Also how do I know that user is
>> still not holding a reference to the file structure.
>
> You don't.
>
>> It seems like debugfs is currently not a good choice for this? Would
>> you recommend me to any other fs or subsystem that I should use for
>> this?
>
> What exactly do you need to export to userspace and for what purpose?
> For debugging-only stuff, sure, use debugfs, but don't rely on it for
> any "real" tools, only your own debugging.

I'm actually writing a driver that would expose a "dummy device" to a
real driver. The dummy driver relies on user space to feed in the
device attributes (no of supported contexts etc). I am now thinking
that a character device interface to user space may actually be a
better choice. Question: Does cdev_del() ensure that all references to
the file are dropped before it returns?

Thanks,

Rajat

>
>> Would hanging those files under the sysfs node for the device sound
>> any better (by representing each "context" using an embedded kobject)?
>
> That would ensure that things work properly.  But you don't need a whole
> kobject, just use a named group and a subdir will be created properly
> for you.
>
> good luck,
>
> greg k-h
--
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/