[PATCH v2 6/6] nl80211: Use global netlink objects.

Ben Greear greearb at candelatech.com
Mon Oct 24 14:26:11 EDT 2011


On 10/22/2011 10:57 AM, Jouni Malinen wrote:
> On Thu, Oct 20, 2011 at 01:21:24PM -0700, greearb at candelatech.com wrote:
>> Netlink sockets can be shared among all driver instances,
>> saving lots of sockets, spurious log messages, memory,
>> and CPU usage when using multiple interfaces in a single
>> process.
>
> I split this into pieces to make it easier to review and address
> conflict with other pending work. The patches are here:
> http://w1.fi/p/ben/
> (the patch numbers below refer to these six patches and not the six that
> happened to be in this series you sent to the list)
>
> I applied 1/6 and 2/6 now, i.e., global nl_cb and nl/nl80211 for
> commands.
>
> 5/6 is clearly in conflict with what Johannes is working on, so I don't
> want to push that in at this point. And anyway, its effect should be
> pretty minimal since Probe Request reporting is currently used only for
> P2P Listen state and there is unlikely to be more than one of those
> running (unless you start simulating large number of P2P devices, I'd
> guess).
>
> 4/6 is somewhat less clear. I think we may end up having to split
> process_event() into parts since it is now used for events that could be
> global and for some cases like Action frames where it would likely
> benefit from being per drv (or bss). I hope that Johannes can address
> this in whatever way fits with his patch set. 3/6 is no-op preparation
> for 4/6, so that is in the same category.

Just FYI:  I pulled down your latest, applied these and ran some quick tests.

Seems to work fine.

Thanks,
Ben

-- 
Ben Greear <greearb at candelatech.com>
Candela Technologies Inc  http://www.candelatech.com



More information about the HostAP mailing list