Event loop stall bug in hostapd-0.7.3

[Ben Greear]

On 01/05/2012 04:08 PM, Bryan Phillippe wrote:
> Hi All,
>
> For the last several weeks, I've been troubleshooting a problem with hostapd-0.7.3 where it gets stuck somewhere in the event loop and clients are no longer able to connect to the AP until I kill and restart hostapd.  I'm not sure what the repro is yet; this seems to happen between 3 to 10 days of uptime on the AP, and only for the busiest APs (I'm running the same version of hostapd on my home network and have never seen the problem).

'strace' often helps debug this kind of thing if it's stuck in a blocking system call
(and should be smaller and probably easier to install than gdb).

gdb would be a big help as well, of course...especially if it's hanging in a tight
loop.

The only hang I ever remember hitting was when supplicant was blocked on reading from
it's unix command pipe.  I've posted a patch to make that non-blocking, but unless
you are doing wpa_cli commands it is unlikely you are hitting that problem.

Have you tried using the latest top-of-tree hostapd?  Seems like it has some recent
improvements for timeouts and such.

Thanks,
Ben

-- 
Ben Greear <greearb at candelatech.com>
Candela Technologies Inc  http://www.candelatech.com