[PATCH] l2_packet: fix eloop leak when using l2_packet_none

Jouni Malinen j at w1.fi
Mon Apr 14 16:09:48 EDT 2014


On Mon, Apr 14, 2014 at 01:48:42PM +0200, Maxime Bizon wrote:
> if you don't have CONFIG_L2_PACKET=y, it happens to use l2_packet_none
> implementation automatically
> 
> I don't have CONFIG_RSN_PREAUTH so it is not selected automatically
> 
> the code looked more to me like a "do nothing" layer instead of a
> reference implementation

Ah, I see. Somehow I did not even remember this being a possibility
anymore since l2_packet is used in number of places and wpa_supplicant
uses l2_packet_linux.c by default. Anyway, I see now that this is not
the case for hostapd.

> I've not tested this, but since my eloop patch was applied (assert(fd !=
> -1)), usage of l2_packet_none should not be possible anymore.
> 
> maybe we should leave l2_packet_none as-is or rename it and use empty
> static inline functions when we don't need it ?

It's fine to just skip the eloop_register_read_sock() call while leaving
the example there. I'll fix this.
 
-- 
Jouni Malinen                                            PGP id EFC895FA


More information about the HostAP mailing list