[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <2857108.moq831zFsU@bentobox>
Date: Wed, 26 Apr 2017 09:16:54 +0200
From: Sven Eckelmann <sven@...fation.org>
To: Gao Feng <gfree.wind@...mail.com>
Cc: b.a.t.m.a.n@...ts.open-mesh.org, mareklindner@...mailbox.ch,
sw@...onwunderlich.de, a@...table.cc, davem@...emloft.net,
netdev@...r.kernel.org, 'Gao Feng' <fgao@...ai8.com>
Subject: Re: [B.A.T.M.A.N.] [PATCH net] net: batman-adv: Fix possible memleaks when fail to register_netdevice
On Mittwoch, 26. April 2017 14:44:24 CEST Gao Feng wrote:
[...]
> I get it now, thanks.
[...]
> BTW, I think although the batadv_softif_create is legacy, we should fix it
> when it still exists :)
I didn't meant that we should not fix it. I just said that it looks to me like
the fix should look different to ensure that it actually fixes the sysfs and
rtnl link implementation for the batadv interface creation. Right now the
ndo_uninit (when it would be set by batadv) is called in the netdev core
functions when an error happens during the registration. This is not the case
for the destructor. Your patch would not change it. It therefore looks like
you simply have to move the current destructor (without the free_netdev) to
ndo_uninit and change the destructor to free_netdev.
The batadv ops doesn't have a newlink function. It will therefore use the
register_netdevice code path which calls free_netdev on failures. The extra
cleanup you've added in
https://www.mail-archive.com/netdev@vger.kernel.org/msg165253.html can
therefore not work for batman-adv. Actually, it is not touching anything
batman-adv related. The suggestion to change the register_netdevice ->
free_netdev part in rtnl_newlink was new in the reply to the batadv
discussion. It is therefore still an open discussion how it is correctly
fixed.
Kind regards,
Sven
Download attachment "signature.asc" of type "application/pgp-signature" (834 bytes)
Powered by blists - more mailing lists