On 12/02/2014 00:15, Ted Unangst wrote:
I think the patch is broken though. It should add a close() if !kflag.
I will submit the updated patch.
Additionally, logic of opening/closing of the listening socket isn't clear.
With lflag and kflag, it reopens listening socket on every iteration for
family!=AF_UNIX, but not for family==AF_UNIX.
I think, with kflag it should keep listening socket the whole time for
all families, and have it with listen backlog>1. And without kflag it
should close the listening socket as soon as possible and have listen
backlog=1.
Should be no difference by family there.
But if there is some reason for this, this reason is worth being
mentioned in the comments.
Yuri
