netdev - Re: [PATCH] bonding: remove sysfs before removing devices

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:	Sat, 06 Apr 2013 01:29:48 +0200
From:	Nikolay Aleksandrov <nikolay@...hat.com>
To:	Veaceslav Falico <vfalico@...hat.com>
CC:	netdev@...r.kernel.org, fubar@...ibm.com, andy@...yhouse.net,
	davem@...emloft.net
Subject: Re: [PATCH] bonding: remove sysfs before removing devices

On 04/06/2013 01:21 AM, Veaceslav Falico wrote:
> On Sat, Apr 06, 2013 at 12:15:11AM +0200, Nikolay Aleksandrov wrote:
>> Hi,
>> Sorry for the late reply but I was travelling this week. In my
>> opinion this
>> fix is wrong because in bond_uninit() (called by rtnl_link_unregister)
>> you have:
>> list_del(&bond->bond_list);
>> which is linked in the bond_net dev_list which is freed by
>> unregister_pernet_subsys.
>
> Yep, you're right, I've hit it recently with the patch applied, and now
> working on it. However, I still think that the idea of the patch is
> correct
> - i.e. to first disable sysfs (especially bonding_masters) and only
>  afterwards to start removing everything else. Or, obviously, to finally
> add normal locking to sysfs functions.
>
> Anyway, this corruption is really rare, so I'll wait for your fix next
> week.
>
Well, there's no need for that because you'll have to iterate over all
"net"s to do it properly. Since we already have code that does it
(unregister_pernet_subsys), my fix is to kill off any "left" bond
devices in bond_net_exit() _after_ destroying the bonding_masters
sysfs entry for that net. This way we preserve the code structure and avoid
duplicating a loop over all nets.
This is the fix we discussed a week ago.
I'd be happy to hear any comments on it before posting :-)

Of course stopping the whole bonding sysfs handling is also an alternative.

--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html