Re: Linux Stability & cold.system

Eberhard Burr (
18 Nov 1997 23:19:50 +0100 (Alan Cox) writes:

> > > Anyone else have useful sadistic tricks?
> > Plug a null-modem-cable between two serial ports (both config'd as
> > 16450 at highest baud-rate) and see whether what you put out one of
> > them arrives at the other end. Should generate quite a few
> > interrupts.

> It generates a passable number, but the tend to stay in step TX/RX. Heavy
> networking is better for randomness (or sending form another host). When
> testing serial drivers I also run a strobing carrier signal into the
> serial drivers to check for races. In the 1.2 era this was quite
> effective but 2.0.x seems to have to got the races out

I didn't have the serial driver in mind in particular; I'm sure that
part works very well. However, if you do this, you can generate an
amount of up to ~45k interrupts/second (assuming fifos switched off,
full-duplex and 115kbps) even on a non-networked computer like
mine. Of course you can do something else at the same time (kernel
compile et al) just to see how well anything else works. I suspect a
misplaced cli() will show up quite soon.


Eberhard Burr             finger for PGP key
                          #include <stddisc.h>  ------  electric cookie:
The telephone is a good way to talk to people without having to offer
them a drink.
		-- Fran Lebowitz, "Interview"