On Tue, Jan 15, 2019 at 08:18:25PM +0000, Peter Oh wrote: > On 01/15/2019 03:43 AM, Masashi Honma wrote: > > This reverts commit 2564184440d9d6041d11a8c7d50b31368634c3bd. > > > > commit 2564184440d9d6041d11a8c7d50b31368634c3bd 'mesh: Apply > > channel attributes before setup interface' causes memory > > leakage of auto test mesh_secure_ocv_mix_legacy. > > > If the patch is doing right thing, but has a bug, isn't it a right way > to fix the bug other than revert it? It would be, but after trying to find a clean way of fixing this, it does not look like there is any easy and sufficiently clean way for doing that and I applied this patch to revert the earlier commit. The main issue here seems to be in hostapd_setup_bss() not getting called when the completion happens through a callback function, e.g., from HT40 coex scan in the case of the hwsim test case mesh_secure_ocv_mix_legacy. That leaves hapd->started to be 0 and hostapd_free_hapd_data() does not free hapd->wpa_auth. I'd assume there could very well be other issues in hostapd_setup_bss() and hostapd_free_hapd_data() being skipped, so this needs this a more complete fix with the mesh-completion-from-callback patch. But since the latest version of that patch broke functionality (without an unacceptable other patch), it looks like more work is needed here before the proper fix can be put in place. -- Jouni Malinen PGP id EFC895FA _______________________________________________ Hostap mailing list Hostap@xxxxxxxxxxxxxxxxxxx http://lists.infradead.org/mailman/listinfo/hostap
