Subject: Re: siginfo, signal queues, memory issues
To: Lord Isildur <mrfusion@uranium.vaxpower.org>
From: Greywolf <greywolf@starwolf.com>
List: tech-kern
Date: 11/29/2002 15:57:20
On Thu, 28 Nov 2002, Lord Isildur wrote:

# On Thu, 28 Nov 2002, Jaromir Dolecek wrote:
# > I agree with Soda on rest of items. Particularily, there indeed
# > needs to be system limit on number of queued signals. I'm not sure
#
# a per-process limit should be enforced as well, to prevent a more
# limited-scope DoS.. and perhaps if we get fancy, reserve a slot in the
# per-process quota for a process running as root to send to, to prevent
# processes from rendering themselves more unkillable by keeping their own
# signal queues full , thereby rejecting KILLs, XCPUs, STOPs, and the like?

I think that KILL and STOP should be sent straightaway OOB, bypassing
the signal queue, considering that, well, what the hell is a process
going to do about KILL and STOP when it receives them?  "Cannot be
caught, blocked or ignored" seems pretty unambiguous to me.

# Isildur

				--*greywolf;
--
NetBSD:  Can we risk exposing the humans to our advanced technology?