Re: [PATCH] LSM: Add security_socket_post_accept() and security_socket_post_recv_datagram().

[Greg Lindahl <greg@blekko.com>]

On Tue, Apr 21, 2009 at 11:46:18PM -0700, David Miller wrote:

> EAGAIN does not happen if the application calls poll(), gets
> an indication that connections are available, and then
> immediately calls accept() on the indicated FD.

I have observed it recently and historically, and not by calling
accept() repeatedly. I don't know what you mean by "immediately" since I
don't think you're advocating race conditions; the other end can
always exit/reset/whatever between the poll() and the accept().

> If overly anal apps "code for it" that is entirely besides the point.
> What we have to be concerned for, from a kernel behavioral standpoint,
> is that some apps "might not code for it".  This is why we don't
> change behavior.

I am suggesting that you survey actual apps. If you find that they're
all overly anal, then maybe you've learned something about EAGAIN
already happening today. I assure you that the co-worker who stuck in
the "ignore EAGAIN without logging it" only did so because he saw it
fairly frequently. He's that way.

-- greg