Date: Wed, 04 Mar 1998 20:07:55 -0800 (PST) From: Simon Shapiro <shimon@simon-shapiro.org> To: Mike Smith <mike@smith.net.au> Cc: current@FreeBSD.ORG, Matthew Thyer <Matthew.Thyer@dsto.defence.gov.au> Subject: Re: silo overflows (Was Re: 3.0-RELEASE?) Message-ID: <XFMail.980304200755.shimon@simon-shapiro.org> In-Reply-To: <199803050256.SAA23678@dingo.cdrom.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On 05-Mar-98 Mike Smith wrote:
...
> But without any testing or hard numbers, your claim above is
> meaningless. You haven't given any context, or demonstrated with
> similar machines and loads side-by-side that there has been any
> quantifiable change in system behaviour. Perhaps its your usage
> profile that's changed?
I have to disagre, Mike. It may be hard to reproduce, or practically
impossible to track (lost interrupts are nasty. First the are totally
asynchronous, and second, they are missing... :-) But, here is a test setup
for you that will exhonorate the sio driver;
a. Hook two boxes via null modem.
b. Boot to single use mode.
c. Set the baud rate to 1200
d. Assume /dev/cuaa00 on both sides, with a full null modem
e. Generate a barberpole file on the remote machine (go to
ftp:simon-shapiro.org/crash/tools, and get blob.c)
make blob;install -c -m 755 blob /usr/local/bin; blob > blob.out
f. On the remote machine do ``cat < /dev/cuaa00 > output_file''
g. On the local machine do ``cat blob.out > /dev/cua00''
h. ``diff blob.out output_file'' Any difference means the driver is
broken.
i. bump the baud rate to 115,200 and repeat the domp.
j. Repeat the whole thing but in the other direction.
k. Go to multi-user mode and repeat steps f-j
l. Start loading the system. First with memory traffic;
dd if=/dev/zero of=/dev/null bs=64k is good.
m. Upgrade to disk traffic; Replace /dev/zero with /dev/sd0c or something
like that. If you want to be nasty to your disk subsystem, pick up st.c
from the same place and run it instead of dd.
n. Add network traffic.
If Mike is correct, this test will pass with flying colors. If it doesm it
is time to look elsewhere. I saw quite few spltty's around on things that
are not Teletype devices :-)
> It has been. We know about it. But nobody that cares enough about it
> has taken the time to analyse the problem. I tried to suggest, in my
> response to Simon, how one might go about doing this, in the
> (apparently vain) hope that someone might take me up on it.
You find me an employer that agrees that I work on these things and I give
you my word of honor that I will get this, and any other bug you give me
fixed.
> Shall we say that slagging the 'sio' driver certainly isn't the best way
> to start?
I have been preached to frequently on this list. Now is my turn; Raising
an issue with a piece of code is NOT criticism of the author, nor of the
maintainer, nor of the project's qulity of management. It is simply an
issue. I think I speak for many on this list when I say that we are at awe
at the talent and dedication expressed in this project. It is simply, that
when we see something broken, we ask ``Is it broken or is my brain in
neutral?'' FreeBSD is growing to be an extreamly complex engineering
product. Canvassing it all is impossible, thus old issues re-surface.
That's all.
----------
Sincerely Yours,
Simon Shapiro
Shimon@Simon-Shapiro.ORG Voice: 503.799.2313
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?XFMail.980304200755.shimon>