segfault in debug mode, level 0x10 (SAE_DEBUG_CRYPTO_VERB) about authsae

[bcopeland]

Hmm, well, it would help to get a core file or at least get the addresses resolved to line numbers.

That said, these days you should probably be using wpa_supplicant to run a secure mesh.

[jkogels]

Thanks for your reply.
Do you mean the core dump file? That one is not created, last lines logged are "segfault at.."
The meshd-80211 will crash in several seconds. So I can fetch some info from strace if it's helpfull? The fault is very easy to reproduce, just set the debug level in the authsae.cfg to 16.

Other issue is a mem leakage, after a while the proces has allocated all the mem available.

I like your suggestion to use an other way to secure the mesh, wpa_supplicant.
Regarding this suggestion to use wpa_supplicant, it does raise some questions.
What is the advantage of using wpa_supplicant versus the authsae daemon for secure mesh? And is it well supported and maintained in the github?

At least I will setup a config of the wpa_supplicant on my current meshd hardware.
I found your article on wpa_supplicant : https://bobcopeland.com/blog/2014/09/wpas-mesh/
A good start for setting up the wpa_supplicant.

Kind regards,
Jeffrey

[bcopeland]

Yes, if you could get a core dump file that would be useful (try ulimit -c unlimited) or a backtrace (run mesh inside gdb, and do bt once the segfault occurs). I'll try your debug level shortly.

wpa_supplicant is probably better supported than authsae at this point. It is maintained inside the upstream hostapd repository (see https://w1.fi/), so use that instead of any github trees you may find.

[alexgrin]

Jeffrey, you can also try running authsae via strace for a quick glance at what's going on - "strace -v meshd-80211 -c <your.conf>" I'll try to run with your config as well shortly.

[alexgrin]

I tracked down the explosion to debug print that was trying to put data into NULLed print buffer, see the pull request above. My local build stopped breaking with the patch applied.

[jkogels]

Pulled the latest master with the merged fix in it, this is running oke now! Thanks for the fast response and fix.
Jeffrey