Path: utzoo!utgpu!jarvis.csri.toronto.edu!clyde.concordia.ca!uunet!zephyr.ens.tek.com!tektronix!reed!busker!p8.f14.n105.z1.FIDONET.ORG!Howard.Spindel
From: Howard.Spindel@p8.f14.n105.z1.FIDONET.ORG (Howard Spindel)
Newsgroups: comp.sys.ibm.pc
Subject: Re: Request for info on Disk Caches
Message-ID: <1414.25C69DAC@busker.FIDONET.ORG>
Date: 30 Jan 90 19:58:16 GMT
Sender: ufgate@busker.FIDONET.ORG (newsout1.26)
Organization: FidoNet node 1:105/14.8
Lines: 28

In an article of <27 Jan 90 00:25:32 GMT>, jmerrill@jarthur.Claremont.EDU (Jason Merrill) writes:

 >I've found that disk caches intefere with communications; I have a 19200
 >baud link to a local system, and when I was using a cache I would 
 >frequently
 >lose characters.  When I removed the cache, the problem went away.
 >Incidentally, the DSZ docs list disk caches as one possible culprit for 
 >Data
 >Overflows...

On 286 machines disk cache programs frequently use extended memory.  
In order to use extended memory the cache program must briefly put
the machine into protected mode.  To switch from protected mode to
real mode requires resetting the processor.  While protected mode
is active and during the reset interrupts are disabled.  An interrupt
driven communications program is unable to respond to received characters
and receive overflow may occur.

This problem may be somewhat alleviated by using 16550 UART chips as
they have internal buffering.  The problem does not exist with 386
machines.



--  
Howard Spindel - via FidoNet node 1:105/14
	    UUCP: ...!{uunet!oresoft, tektronix!reed}!busker!14.8!Howard.Spindel
	    ARPA: Howard.Spindel@p8.f14.n105.z1.FIDONET.ORG