Return to BSD News archive
Path: sserve!newshost.anu.edu.au!harbinger.cc.monash.edu.au!yarrina.connect.com.au!warrane.connect.com.au!godzilla.zeta.org.au!not-for-mail From: bde@zeta.org.au (Bruce Evans) Newsgroups: comp.unix.bsd.freebsd.misc Subject: Re: 1.1.5.1 Serial Quirks/Phenomena??? Date: 24 Jul 1995 15:14:22 +1000 Organization: Kralizec Dialup Unix Sydney - +61-2-837-1183, v.32bis v.42bis Lines: 39 Message-ID: <3uva7e$2d6@godzilla.zeta.org.au> References: <DBvwHr.L5p@agora.rdrop.com> NNTP-Posting-Host: godzilla.zeta.org.au In article <DBvwHr.L5p@agora.rdrop.com>, Craig Keenan <sundans@agora.rdrop.com> wrote: >happens later.... I have observed maybe a silo overflow error about once >a month, but not enough to really make a difference. What I am concerned >about are the many 'interrupt-level overflows' I get.... sometimes about >200/day !!! From the sio manpage, I see these are problems with the >bottom half of the sio dirver.... and I don't think I am losing any data >At least my customers aren't saying anything :) > So do you think the problem is with the STB 4-Com shitting out flaky >interrupt signalss, confusing the sio driver??? Or could it be the AST >486/66 (Bravo LC, I think...) is flipping out? I have heard about using No, they are all software problems and/or general machine overload problems. The bottom half of the driver is sometimes locked out a little longer than was planned for from talking with the top half. This problem can normally be avoided by using hardware handshaking (stty crtscts). >IDE drives with multiport serial causing the silo's (due to IDE >bus-hoggin), but I really would like to clean up my syslog always saying >I have interrupt-level overflow errors... Under FreeBSD, SCSI drives can hog the bus in a way that causes silo overruns, but IDE drives can't (hoggish SCSI controllers can prevent serial interrupts for 250 usec or more, but IDE transfers are under cpu control and can't prevent serial interrupts for more than 1 or 2 usec). However, IDE transfers can hog the bus in a way that prevents the top half of the sio driver from running. They can keep interrupting faster than the data can be transferred. This is more of a problem under 2.0.5 because the transfer size is often 64K which takes 20-30 msec to transfer. Interrupt-level overruns occur if the top half is blocked for more than about 20 msec. >Also, how ereliable is the >serial in 2.0.5R??? I would like to upgrade to 2.0.5 (or maybe 2.1, time Slightly less reliable than 1.1.5 for opening and closing connections. The same for moving data. -- Bruce Evans bde@zeta.org.au