linux-kernel - Re: [PATCH] module: Fix race condition between load and unload module

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: <CA+55aFxhwYO8wrHXHH6ry0pfpPHURsjN_aotT0xsDwZZZj7wvQ@mail.gmail.com>
Date:	Sat, 13 Apr 2013 10:53:50 -0700
From:	Linus Torvalds <torvalds@...ux-foundation.org>
To:	Anatol Pomozov <anatol.pomozov@...il.com>
Cc:	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
	Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
	Salman Qazi <sqazi@...gle.com>,
	Rusty Russell <rusty@...tcorp.com.au>,
	Al Viro <viro@...iv.linux.org.uk>
Subject: Re: [PATCH] module: Fix race condition between load and unload module

On Sat, Apr 13, 2013 at 8:41 AM, Anatol Pomozov
<anatol.pomozov@...il.com> wrote:
>
> Does it make sense to move it to a separate function in kref.h?
>
> /** Useful when kref_get is racing with kref_put and refcounter might be 0 */
> int kref_get_not_zero(kref* ref) {
>     return atomic_inc_not_zero(&kref->refcount);
> }

It turns out we have that, except it's called "unless_zero", because
it uses "atomic_add_unless(x,1,0)", rather than the simplified
"atomic_inc_not_zero(x)".

> or maybe instead change default behavior of kref_get() to
> atomic_inc_not_zero and force callers check the return value from
> kref_get()?

That would be painful, and _most_ users should have a preexisting
refcount. So it's probably better in the long run to just keep the
warning (but perhaps fix it to be SMP-safe). So I think the part of
your patch that made kref_get() use atomic_inc_return() is probably a
good idea regardless.

Also, I changed my patch to be minimal, and not change other users of
kobject_get(). So other users (not kset_find_obj()) will continue to
get the warning, and kset_find_obj() uses the safe version. So this is
what I'm planning on committing as the minimal patch and marking for
stable. The rest (including that atomic_inc_return() in kref_get)
would be cleanup.

Can you give this a quick test?

                       Linus

Download attachment "patch.diff" of type "application/octet-stream" (802 bytes)