From owner-freebsd-stable Fri Jan 8 11:59:38 1999 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id LAA16518 for freebsd-stable-outgoing; Fri, 8 Jan 1999 11:59:38 -0800 (PST) (envelope-from owner-freebsd-stable@FreeBSD.ORG) Received: from wattres.Watt.COM (wattres.Watt.COM [207.33.154.225]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id LAA16477 for ; Fri, 8 Jan 1999 11:59:32 -0800 (PST) (envelope-from steve@Watt.COM) Received: (from steve@localhost) by wattres.Watt.COM (8.8.8/8.8.8) id LAA14579; Fri, 8 Jan 1999 11:58:43 -0800 (PST) (envelope-from steve) Date: Fri, 8 Jan 1999 11:58:42 -0800 From: Steve Watt To: Marc Giannoni Cc: freebsd-stable@FreeBSD.ORG, Mark Murray , Chris Dillon Subject: Re: IRQ entropy causes panics? Message-ID: <19990108115842.A14261@wattres.Watt.COM> Reply-To: Steve Watt References: <19990108094015.A10590@wattres.Watt.COM> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii X-Mailer: Mutt 0.95i In-Reply-To: ; from Marc Giannoni on Fri, Jan 08, 1999 at 02:57:26PM -0500 X-Callsign: KD6GGD Sender: owner-freebsd-stable@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG On Fri, Jan 08, 1999 at 02:57:26PM -0500, Marc Giannoni wrote: > > Hooking interrupts is a "real time" issue. There is no way an OS, however > will written, can tolerate flagrant additions to it's interrupt latencies. > ("j. random sysadmin run-amok") I'm extremely well versed in the issues -- I work in the kernel engineering group at Lynx Real Time Systems. There are ways to minimize real-time impacts ("it's just code"), and there are ways to prevent reentrant corruption. I haven't looked at the IRQ entropy code yet, so I'm not in a position to say if it _could_ be improved. > I'm suprised that the randomizer code can hook interrupts at all and still > keep the OS stable. This stuff is some pretty clever code. You are actually > modifying the kernel's core execution by enabling these hooks! Is there a microtime() call already made at each interrupt entry? If not, then yeah, that would be adding a huge latency. Something on the order of 5uS on a faster ISA system I tested on (when I added similar microtime()-like stuff to LynxOS). I haven't looked to see if it's gotten any better with recent chipsets, though -- that's just way too much cost for us RTOS guys at interrupt time, so I didn't do that. > Trust me when I say that there probably is no "Underlying Problem". > It's pretty amazing that this thing even exists. I'd still contend, from an engineering standpoint, that such things shouldn't destabilize a system that way. If it's inherently unstable, then say so. On the gripping hand, yup, it's a *really* nifty idea. -- Steve Watt KD6GGD PP-ASEL-IA ICBM: 121W 56' 58.1" / 37N 20' 14.2" Internet: steve @ Watt.COM Whois: SW32 Free time? There's no such thing. It just comes in varying prices... To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message