Userspace mlme and detecting failed authentication attempt

Blaise Gassend blaise at willowgarage.com
Mon Jun 28 15:09:20 EDT 2010


Hi again,

After looking at the latest development version of wpa_supplicant, I
think I have answered many of my own questions.

If I'm understanding things correctly SME is the way things are headed
now for mainstream wpa_supplicant and nl80211 drivers. I'm still curious
about the questions I asked about, but at this point they are more of
academic interest.

Now SME seems to be getting the association timeout events from the
kernel, but so far it isn't very robust. From what I see, it rescans on
each association timeout, and does not blacklist an AP that timed out.
This is stuff I can work on improving. I'm wondering if there is anybody
else out there working on this stuff that I should coordinate with, or
if I should just hack away and submit patches.

Thanks,
Blaise

On Mon, 2010-06-28 at 10:36 -0700, Blaise Gassend wrote:
> Hi,
> 
> I am currently working on using wpa_supplicant as part of a roaming
> streaming video app, and have a few questions that members of this list
> may be able to help with:
> http://www.willowgarage.com/pages/robots/texas/overview
> 
> Right now my main problem is that if association fails, it takes many
> seconds before wpa_supplicant retries an association. I'm exploring
> options to make this better. I have seen that you can adjust the
> association timeout, but would rather a more robust solution where I
> don't have to trade off excessive waiting time with the risk of
> interrupting an association that is still making progress.
> 
> Here are the questions I am wondering about. Even a simple yes/no answer
> would be really helpful in cutting down my search space.
> 
> 1) Is there currently a mechanism for nl80211 to report to
> wpa_supplicant that that the association failed?
> 
> 2) Is there currently a way to get wpa_supplicant to use that mechanism,
> or is that something that I would have to add?
> 
> I see that wpa_supplicant has a userspace mlme option, and my initial
> thought would be easy to hack on if I want to fine-tune the association
> process. According to config.mk that requires a kernel patch (the "vif"
> patch, I believe). From digging through the mailing list, it looks like
> this patch used to be linked to from the wpa_supplicant web page, but
> the link no longer works. I have also seen commits to the kernel code
> that suggest that support for userspace mlme has been removed.
> (http://osdir.com/ml/linux-wireless/2009-03/msg01003.html)
> 
> So more questions:
> 
> 3) Is userspace mlme obsolete?
> 
> 4) Where can the kernel patch be found if I want to use it?
> 
> 5) Apparently the "SME" API may be the replacement for userspace MLME.
> Does wpa_supplicant use the SME API?
> 
> I have seen various threads where people discuss the merits of userspace
> MLME and how many userspace MLME implementations there should be.
> (http://www.mail-archive.com/netdev@vger.kernel.org/msg40483.html for
> example) These were a while back, and I get the feeling that some
> consensus was reached, but I haven't found the consensus written down
> anywhere. 
> 
> 6) Any pointers to what the outcome of that discussion was?
> 
> Thanks for any help,
> Blaise
> 




More information about the HostAP mailing list