[PATCH] Add epoll option for better performance

[Masashi Honma]

By your review, I realized that I should use more safer way than modifying
CONFIG_ELOOP. Because modifying it can cause trouble to various
functionality includes older version wpa_supplicant/hostapd. So I modified
like that.

> Could you please confirm that this combination of the last two patches
> have been tested to work with epoll?

I have already tested these patches before sending.
I have tested with select/poll/epoll and wpa_supplicant/hostapd by
WPA-PSK connection.
I have not tested on Android.

2014-05-13 23:14 GMT+09:00 Jouni Malinen <j at w1.fi>:
> On Tue, May 13, 2014 at 09:35:48AM +0900, Masashi Honma wrote:
>> > On its own, this does not seem to be changing anything, so I'm not going to
>> > apply it without some real justification showing how this is going to be
>> > used with the following changes (epoll).
>>
>> This is the following patch.
>
> Hmm.. This seems to be changing the design completely from where the
> changes were previously going.. While there is nothing particularly
> wrong with this approach, it is a bit difficult to understand why there
> was such a change in the design rather than a minor change to fix the
> previous version to not end up commenting out both options in a way that
> resulted in a busy loop.
>
> I guess you don't care too much about the exact mechanism, so for now,
> we might as well go with this addition of CONFIG_ELOOP_EPOLL=y and leave
> the cleanup of CONFIG_ELOOP for future, if anyone cares enough about it.
> Could you please confirm that this combination of the last two patches
> have been tested to work with epoll?
>
> --
> Jouni Malinen                                            PGP id EFC895FA
> _______________________________________________
> HostAP mailing list
> HostAP at lists.shmoo.com
> http://lists.shmoo.com/mailman/listinfo/hostap