We sometimes observe a 'deadly embrace' type deadlock occurring
between mutually connected sockets on the same node. This happens
when the one-hour peer supervision timers happen to expire
simultaneously in both sockets.
The scenario is as follows:
CPU 1: CPU 2:
Acked. This will only be relevant in clusters > 1000 nodes, which I must admit
I haven't tested yet.
///jon
> -Original Message-
> From: Dan Carpenter [mailto:dan.carpen...@oracle.com]
> Sent: Friday, 17 June, 2016 05:22
> To: Jon Maloy
> Cc: Ying Xue; David S. Miller; net...@vger.kerne
From: Dan Carpenter
Date: Fri, 17 Jun 2016 12:22:26 +0300
> "up_map" is a u64 type but we're not using the high 32 bits.
>
> Fixes: 35c55c9877f8 ('tipc: add neighbor monitoring framework')
> Signed-off-by: Dan Carpenter
Applied.
From: Jon Maloy
Date: Fri, 17 Jun 2016 06:35:57 -0400
> We sometimes observe a 'deadly embrace' type deadlock occurring
> between mutually connected sockets on the same node. This happens
> when the one-hour peer supervision timers happen to expire
> simultaneously in both sockets.
>
> The scena