Re: [PATCH net] bonding: fix two race conditions in bond_store_updelay/downdelay

[Nikolay Aleksandrov <nikolay@redhat.com>]

On 11/13/2013 06:03 PM, Eric Dumazet wrote:
> On Wed, 2013-11-13 at 17:07 +0100, Nikolay Aleksandrov wrote:
>> This patch fixes two race conditions between bond_store_updelay/downdelay
>> and bond_store_miimon which could lead to division by zero as miimon can
>> be set to 0 while either updelay/downdelay are being set and thus miss the
>> zero check in the beginning, the zero div happens because updelay/downdelay
>> are stored as new_value / bond->params.miimon. Use rtnl to synchronize with
>> miimon setting.
> 
> It seems a bit heavy duty to take rtnl for this.
> 
> Using ACCESS_ONCE() in bonding_store_updelay()/bonding_store_downdelay()
> should be enough ?
> 
> int miimon = ACCESS_ONCE(bond->params.miimon);
> 
> 
> 
> --
> To unsubscribe from this list: send the line "unsubscribe netdev" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 
Hi Eric,
I thought about this version too, but downdelay/updelay can be changed in other
places (e.g., store_miimon) and the resulting downdelay/updelay value might not
be the right one.
Correct me if I'm wrong, but this is what I have in mind (miimon = 100, updelay
= 200):
set miimon to 300 and concurrently set updelay to 400, we might endup leaving
updelay to 400 because the old value of miimon is used in the calculation in
store_updelay even though when changing miimon updelay/downdelay get adjusted,
they might get adjusted by store_updelay/downdelay to a wrong value afterwards.

 Nik