Path: utzoo!utgpu!water!watmath!clyde!att!osu-cis!tut.cis.ohio-state.edu!bloom-beacon!think!ames!pasteur!ucbvax!decwrl!sun!pitstop!sundc!seismo!uunet!munnari!otc!metro!ipso!runx!brucee From: brucee@runx.ips.oz (Bruce Evans) Newsgroups: comp.os.minix Subject: Re: insufficient buffer size in tty code; linewrap Message-ID: <1779@runx.ips.oz> Date: 7 Oct 88 19:53:41 GMT References: Reply-To: brucee@runx.OZ (Bruce Evans) Organization: RUNX Un*x Timeshare. Sydney, Australia. Lines: 30 Charles Hedrick writes: >good line utilization.) The interrupt level buffer was a surprise. >The buffer had 100 characters, which at 2400 baud should have lasted >for almost 1/2 second. Surely the tty task should get to run within >1/2 second. About all I can think is that there's some problem in the >way communication is being done such that when the floppy starts its >motor (which takes 1/2 sec, if remember correctly), the tty task gets >locked out for that period. Note that this is not an interrupt It must be the little-discussed FS bottleneck. A user calls FS, FS calls FLOPPY, and FLOPPY can take up to 3/4 sec waiting for the motor. Meanwhile FS is blocked. TTY buffers fill up because no user task can get through FS to unload them. This is a fancy form of busy waiting and is unacceptable for 1 millisec delays, let alone 750. It will be difficult to fix. This has been ignored for too long :-(. >[Politics of getting line wrap into kernel] Line wrap is essential while running dumb tty style programs, and cleverer programs should be able to change it. >[Fixes for makefile] Don't forget to fix at_makefile and pc_makefile as well as makefile. All the duplications should be eliminated by having dependencies for things like atkernel and xtkernel. No underlines, please. This eliminates wini.c too. Bruce Evans Internet: brucee@runx.ips.oz.au UUCP: uunet!runx.ips.oz.au!brucee