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]

Date:	Mon, 28 Dec 2015 12:56:34 -0800
From:	Rajat Jain <rajatxjain@...il.com>
To:	Nicolai Stange <nicstange@...il.com>
Cc:	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	linux-newbie@...r.kernel.org, Greg Kroah-Hartman <greg@...ah.com>
Subject: Re: debugfs_remove_recursive() while a file is in use by userspace

>
> The current state is that both question are to be answered with "no",
> i.e. debugfs file removal is racy.
>

Ouch! :-(. Thanks for confirming!

Somewhat related, is my understanding correct that the behaviour of
character device layer w.r.t. the two questions that I asked the
answer to my questions would be "yes"?

i.e. cdev_del() would wait

>> (1) until all the currently executing file operation methods
>> (read/write/map etc) have returned?

AND

>> (2) until the user has given up all references (descriptors) to the
>> files (i.e. until release() method has been
>> called)?

Thanks,

Rajat


On Mon, Dec 28, 2015 at 12:42 PM, Nicolai Stange <nicstange@...il.com> wrote:
> Hi Rajat,
>
> Rajat Jain <rajatxjain@...il.com> writes:
>
>> Hi,
>>
>> I wanted to understand the behavior taken when a module calls
>> debugfs_remove_recursive() on a directory, while files under that
>> directory may still be in use by the userspace (for instance an
>> ongoing read / write operation).
>>
>> Does the function wait
>>
>> (1) until all the currently executing file operation methods
>> (read/write/map etc) have returned?
>> OR
>> (2) until the user has given up all references (descriptors) to the
>> files under the directory (i.e. until release() method has been
>> called)?
>
> The current state is that both question are to be answered with "no",
> i.e. debugfs file removal is racy.
>
> I've recently sent a patch addressing this issue:
>   https://lkml.kernel.org/g/87y4dfukzy.fsf@gmail.com
>
> Basically, it turns the answer to your first question into "yes".
> Subsequent reads/writes will return a -EIO.
>
> That patch (series) is still under review though.
>
> Further reference can be found in the *sub*-thread rooted at
>   http://thread.gmane.org/gmane.linux.kernel/1452470/focus=1467314
>
> Best,
>
> Nicolai
--
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/