Le 04.05.2019 à 18.34, Stefan Sperling a écrit :
On Fri, May 03, 2019 at 09:37:31PM +0200, [email protected] wrote:
Synopsis:       Atheros card AR5B91 crash on mode hostap on OBSD 6.5
Category:       driver
Environment:
        System      : OpenBSD 6.5
        Details     : OpenBSD 6.5 (GENERIC.MP) #0: Wed Apr 24 23:38:54 CEST 2019
                         
[email protected]:/usr/src/sys/arch/amd64/compile/GENERIC.MP

        Architecture: OpenBSD.amd64
        Machine     : amd64
Description:
        As OpenBSD 6.5 was released and installed, the athn driver on the card 
(AR5B91) kept crashing.
        The driver would simply echo "device timeout" on the console and 
eventually, kernel would crash.
Crash how? You must include a ddb trace when reporting kernel crashes.

First time I encourted it, I had at least three crashs. But when I indeed made the report, I could not provide a ddb.

Now, I reproduce the bug, the kernel indeed does not crash but hold on stubbornly. This is annoying.

        It reveals after tests that the hostap mode is guilty. Note that no 
previous release made any trouble with this hardware.
How-To-Repeat:
        Putting the device up on hostap mode.
Fix:
        Putting the device down (ifconfig athn0 down) would stop it.
        Putting the device out of the hostap mode (ifconfig athn0 media 
-mediaopt hostap) while running.
athn0 at pci6 dev 0 function 0 "Atheros AR9281" rev 0x01: apic 2 int 16
athn0: AR9280 rev 2 (1T2R), ROM rev 16, address 00:24:2b:72:d1:df
I have the same device running in hostap mode on 6.5 and I am
continously using it without issues.

I don't see this problem locally.

Can you please try to narrow down the circumstances which trigger
the problem a bit further?

"device timeout" means that a frame could not be sent.
Are you somehow creating conditions where the this could occur?
is this problem triggered by particular client devices?

Not that I am aware of.

This happens as soon as I do anything with the device (Yes I know, I wish I could provide more info).

Actually, for the whole purpose of testing, I disabled the two actions I thought could trigger : I disabled dhcpcd ipv6 address assignement (I put the address statically in config' instead) and I did not start dnsmasq.

I also changed physically the card into the other card slot, still timeout.

Clients outside cannot actually connect to the wifi.

Reply via email to