From: davidsen@sixhub.UUCP (Wm E. Davidsen Jr) Subject: Re: Are any SIMMs cheap these day$ ? Date: 9 Aug 1993 14:33:02 GMT
In article <1993Aug7.101410.361@umibox.hanse.de> root@umibox.hanse.de (Bernd Meyer) writes:
| With parity errors just bringing your machine to a sudden halt, you can
| corrupt all your filesystems and lose all your data. But at least you'll
| know why :-)
There's no reason to bring the system down just because you catch the
parity error. If it's in the kernel you probably should, and the
decision to write buffers which may be corrupted vs just going down is
one you have to make. If the error is in the user area it gets more
interesting.
If the memory is mapped to (read only) text you can mark the physical
memory as bad, log a message, grab a fresh copy of the the text off the
disk (it becomes a page fault) and continue. Otherwise you are probably
safer killing the user process and marking the memory.
If you ignore the error you will have bad results, corrupted data in
memory, etc. Over the course of time your system will "get sick" as
pervasive corruption take place. I prefer a working computer or none at
all, with recovery and logging if the problem is in user space. Having
chased a few subtle errors in my life, I really want what comes out of
my computer to be the best the software will produce.
--
Bill Davidsen, davidsen%sixhub.uucp@uunet.uu.net
TMR Associates, +1 518-370-5654
C programming, training, data gathering, porting to open systems,
heterogeneous environments, computer controlled housing, custom software