linux-kernel - Re: [PATCH net] net: mvpp2: debugfs: fix problem with previous memory leak fix

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] [day] [month] [year] [list]

Date:   Sun, 25 Sep 2022 09:55:10 +0200
From:   Marcin Wojtas <mw@...ihalf.com>
To:     "Russell King (Oracle)" <linux@...linux.org.uk>
Cc:     Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        netdev@...r.kernel.org, linux-kernel@...r.kernel.org,
        "David S. Miller" <davem@...emloft.net>,
        Eric Dumazet <edumazet@...gle.com>,
        Jakub Kicinski <kuba@...nel.org>,
        Paolo Abeni <pabeni@...hat.com>, stable <stable@...nel.org>
Subject: Re: [PATCH net] net: mvpp2: debugfs: fix problem with previous memory
 leak fix

niedz., 25 wrz 2022 o 08:58 Russell King (Oracle)
<linux@...linux.org.uk> napisał(a):
>
> On Sun, Sep 25, 2022 at 01:27:06AM +0200, Marcin Wojtas wrote:
> > Hi Russell,
> >
> > czw., 22 wrz 2022 o 19:08 Marcin Wojtas <mw@...ihalf.com> napisał(a):
> > >
> > > Hi,
> > >
> > > Thank you both for the patches.
> > >
> > >
> > > czw., 22 wrz 2022 o 18:05 Russell King (Oracle)
> > > <linux@...linux.org.uk> napisał(a):
> > > >
> > > > On Wed, Sep 21, 2022 at 01:44:44PM +0200, Greg Kroah-Hartman wrote:
> > > > > In commit fe2c9c61f668 ("net: mvpp2: debugfs: fix memory leak when using
> > > > > debugfs_lookup()"), if the module is unloaded, the directory will still
> > > > > be present if the module is loaded again and creating the directory will
> > > > > fail, causing the creation of additional child debugfs entries for the
> > > > > individual devices to fail.
> > > > >
> > > > > As this module never cleaned up the root directory it created, even when
> > > > > loaded, and unloading/loading a module is not a normal operation, none
> > > > > of would normally happen.
> > > > >
> > > > > To clean all of this up, use a tiny reference counted structure to hold
> > > > > the root debugfs directory for the driver, and then clean it up when the
> > > > > last user of it is removed from the system.  This should resolve the
> > > > > previously reported problems, and the original memory leak that
> > > > > fe2c9c61f668 ("net: mvpp2: debugfs: fix memory leak when using
> > > > > debugfs_lookup()") was attempting to fix.
> > > >
> > > > For the record... I have a better fix for this, but I haven't been able
> > > > to get it into a state suitable for submission yet.
> > > >
> > > > http://www.home.armlinux.org.uk/~rmk/misc/mvpp2-debugfs.diff
> > > >
> > > > Not yet against the net tree. Might have time tomorrow to do that, not
> > > > sure at the moment. Medical stuff is getting in the way. :(
> > > >
> > >
> > > I'd lean towards this version - it is a bit more compact. I'll try to
> > > test that tomorrow or during the weekend.
> > >
> >
> > I improved the patch compile and work (tested on my CN913x board).
> > Feel free to submit (if you wish, I can do it too - please let me know
> > your preference):
> > https://github.com/semihalf-wojtas-marcin/Linux-Kernel/commit/0abb75115ffb2772f595bb3346573e27e650018b
>
> I don't see what the compile fixes were in that - it looks like my patch

I overlooked drivers/net/ethernet/marvell/mvpp2/mvpp2.h diff (I
applied patch manually), but I double-checked and it was there, so you
are right.

> ported onto current -net. Obvious changes:

Yes, it's applied on top of net/net-next.

>
> - moved mvpp2_dbgfs_exit() declaration from after mvpp2_dbgfs_cleanup()
>   to before.
> - moved definition of mvpp2_root to the top of the file (as no effect
>   on the code.)

These are irrelevant. The most important part was:

--- a/drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c
+++ b/drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c
@@ -7714,8 +7714,8 @@ module_init(mvpp2_driver_init);

 static void __exit mvpp2_driver_exit(void)
 {
-       mvpp2_dbgfs_exit();
        platform_driver_unregister(&mvpp2_driver);
+       mvpp2_dbgfs_exit();
 }
 module_exit(mvpp2_driver_exit);

Otherwise, I observed NULL-pointer dereference when
mvpp2_remove()->mvpp2_dbgfs_cleanup().

>
> and the change to port it:
>
> - dropped my mvpp2_dbgfs_init() hunk (because it's different in -net)
> - removed static declaration of mvpp2_root in mvpp2_dbgfs_init()
>
> I'm not seeing any other changes.
>
> Note that Sasha has submitted a revert of Greg's original patch for
> mainline, so my original patch should apply as-is if that revert
> happens - and I don't see any compile issues with it.
>

Sure. In order to avoid misunderstandings, I stop here and leave the
submission to you.

Thanks,
Marcin