netdev - Re: [PATCH net 0/2] net: Fix oops on state free after lwt module unload

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: <916bada8-febc-7da9-022a-34dee4010e8b@brocade.com>
Date:   Fri, 20 Jan 2017 20:21:45 +0000
From:   Robert Shearman <rshearma@...cade.com>
To:     David Miller <davem@...emloft.net>
CC:     <netdev@...r.kernel.org>, <tom@...bertland.com>,
        <roopa@...ulusnetworks.com>
Subject: Re: [PATCH net 0/2] net: Fix oops on state free after lwt module
 unload

On 20/01/17 17:03, David Miller wrote:
> From: Robert Shearman <rshearma@...cade.com>
> Date: Wed, 18 Jan 2017 15:32:01 +0000
>
>> This patchset fixes an oops in lwtstate_free and a memory leak that
>> would otherwise be exposed by ensuring that references are taken on
>> modules that need to stay around to clean up lwt state. To faciliate
>> this all ops that implement destroy_state and that can be configured
>> to build as a module are changed specify the owner module in the
>> ops. The intersection of those two sets is just ila at the moment.
>
> Two things:
>
> 1) Under no circumstances should we allow a lwtunnel ops implementing
>    module to unload while there is a rule using those ops which is
>    alive.
>
>    Therefore, we should not special case the destroy op.  We should
>    unconditionally grab the module reference.
>
> 2) Please add the new 'owner' field and add an appropriate assignment
>    for ops->owner to _every_ lwtunnel implementation, and do so in
>    your first patch.  Please do not only do this for ILA.
>
> Thanks.

Very clear, makes sense, will do.

Thanks,
Rob