From owner-freebsd-hackers Fri Mar 17 07:12:42 1995 Return-Path: hackers-owner Received: (from majordom@localhost) by freefall.cdrom.com (8.6.10/8.6.6) id HAA08841 for hackers-outgoing; Fri, 17 Mar 1995 07:12:42 -0800 Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.34]) by freefall.cdrom.com (8.6.10/8.6.6) with ESMTP id HAA08821 for ; Fri, 17 Mar 1995 07:12:26 -0800 Received: (from bde@localhost) by godzilla.zeta.org.au (8.6.9/8.6.9) id BAA21373; Sat, 18 Mar 1995 01:08:20 +1000 Date: Sat, 18 Mar 1995 01:08:20 +1000 From: Bruce Evans Message-Id: <199503171508.BAA21373@godzilla.zeta.org.au> To: bde@zeta.org.au, dufault@hda.com Subject: Re: Multiport serial cards Cc: freebsd-hackers@freefall.cdrom.com Sender: hackers-owner@FreeBSD.org Precedence: bulk >Would you be willing to hazard a guess as to the interrupt latency >for the SIO on a 486DX/33 for FreeBSD 2.x? How are we for keeping >interrupts on in the kernel? Normally 10-20 usec if the serial driver isn't already active. Serial interrupts have highest priority and normally aren't stopped by anything except their own handlers. Of course, the latency is determined by the worst case. The worst case seems to be for braindamaged bus masters. My U34F seems to hog the bus for 15 sector transfer times (250 usec). >Finally - is there an AST/8, or other 8 port 16550 board that works >with the "sio" driver? There will be at least two other light duty >serial ports for this application. An AST/4 plus COM1 and COM2 >will do it, but I would just as soon identify another path. Several unintelligent 8 port and 16 port boards are reported to work. Modem control lines seem to be left out more on 8 port boards than on 16 port ones. sio partly-intentionally uses a dumb polling method that doesn't depend on any board features except the ports not getting in each others way. This increases interrupt overhead a lot for 16 port boards. Bruce