From: kutcha@eos.acm.rpi.edu (Phillip Rzewski) Subject: Crashing machine: >16M mem patch Date: 11 Apr 1993 21:43:11 GMT
I recently upgraded the memory on my machine from 8 megs in 1 meg simms to 32
megs in 4 meg simms. Naturally I would like to be able to use all of those 32
megs from within linux. Of course that meant enabling the >16M patch option in
the kernel configuration.
Trouble is, when I enable it, I can't use my machine for long without it
breaking. I looked through the FAQ and the only reference I found to this sort
of thing was "you have to patch the kernel to get >16M". It said nothing like
"it will break on <this type of machine>" or anything. I'm not sure if this is
rarely tested code, but I may have found a bug. My setup is a 386/25dx, 32 megs
of memory in 4 meg simms (I think it's called "x9" type memory, not sure), and
the "mainboard " (motherboard?) is a "PD 386", for what that's worth. I looked
through the whole manual and found no reference to whether the bus was EISA or
ISA, but since I bought the machine a couple years ago and didn't know anything
much about buses at the time, they probably gave me an ISA.
Here's some idea of how I crashed the machine. I made myself two boot disks,
one with the default "only up 16 megs" and the other with the ">16 megs" patch
enabled. I booted with the >16M one and went into X. I had a kermit open and a
couple of xterms running bash, one of which I had top in. I then ran idraw once
to see what top was saying. It accepted that one. I ran a second idraw and it
accepted that one. So now I had two open at once. I was about to go to make a
third when he machine froze for a couple seconds and then rebooted by itself. I
crashed it similarly by compiling gnutar and opening xv and emacs and a couple
other things (for that one it just froze totally and I had to reboot myself),
but that would be harder to repeat exactly as I did it the first time. The
amount of free memory on top was still above the 16 meg level when I was
crashing the machine. It was usually in the 17000k range.
To then get some comparison, I booted with the other disk so I was limited to
only the first 16 megs of memory. Right off the bat I opened 4 idraws with no
problems whatsoever. As I am typing this message I am using the same session I
once had the idraws on. It is quite obvious that the difference between with
the patch an without the patch is: one works, one doesn't (for me).
I guess the questions that need to be asked are: has anyone else had this
happen? has anyone else done these tests and had it work? if so what is your
config? is it simply a case that no one actually has 32 megs of memory and
hence this code never gets tested? if my problem is serious is there one person
in particular i should be telling this to?