ap_vlan_open exits hwsim tests

[Eduardo Abinader]

Actually, I'm using vm/kernel-config which enables
CONFIG_DEBUG_KOBJECT_RELEASE. Should we change this in default
vm/kernel-config? So, the workaround for select timeout is ok to change
meanwhile?

Just now, I'm testing with kernel option disabled.


Thanks.


On Wed, Mar 12, 2014 at 5:31 AM, Jouni Malinen <j at w1.fi> wrote:

> On Wed, Mar 12, 2014 at 10:58:46AM +0200, Jouni Malinen wrote:
> > On Wed, Mar 12, 2014 at 12:43:33AM -0700, Jacob Minshall wrote:
> > > I have found the offending commit
> > >
> > > 89c86a64cd056e283323710c9ddf6f7090a450c8
> > > kobject: delay kobject release for random time
>
> > That said, this is test functionality and it does not look like one
> > would enable CONFIG_DEBUG_KOBJECT_RELEASE=y for a production build.. I
> > didn't actually have that enabled in my kernel which would explain why I
> > don't see the issue. As a quick workaround, I'd assume you can just
> > remove that from the kernel build used for hwsim tests for now. I'll see
> > what happens when I do the opposite.. :)
>
> It looks like this kernel debug options makes NL80211_CMD_DEL_INTERFACE
> and SIOCGIFBR(BRCTL_DEL_BRIDGE) block for 1-8 seconds. ap_vlan_open ends
> up having to run four of these operations during hostapd deinit path on
> "REMOVE wlan3" control interface operation. Consequently, this is very
> likely to hit the 10 second timeout.
>
> IMHO, the kernel is misbehaving from user space view point, so blindly
> increasing the timeout is not really desirable. That said, it would be
> good to confirm that this kind of blocking behavior on interface removal
> is indeed expected result of that kernel commit and if so, make it
> possible to run the hwsim test cases even with this behavior in place.
> wpaspy.py could be changed to accept a timeout value for the request()
> command and the hostapd REMOVE commands could use larger timeout value
> to workaround this.
>
> --
> Jouni Malinen                                            PGP id EFC895FA
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.shmoo.com/pipermail/hostap/attachments/20140312/b3218a09/attachment.htm>