lkml.org 
[lkml]   [1996]   [Sep]   [13]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
Date
Subjectswapping races

Hi,

Has anyone else observed these kinds of problems during/after heavy
swapping activity?

1. Random g++ crashes, which go away as soon as you retry.
2. System hang/freeze just after a successful "swapoff -a".

My setup is:
i486/33, 8 MB RAM, 10 MB swap partition + 7 MB swap file,
2 IDE disks, Mitsumi CD-ROM,
kernel 2.0.6 without networking.

Detailed description:

1. During compilation of several hundred C++ files, g++ crashes on some
of them with signal 11 or "internal compiler error". The compilation
of each of these files needs about 10 to 17 megabytes memory. It swaps
heavily.
No way to reproduce the crashes: When "make" is restarted immediately
afterwards, in the same environment, from the same "bash" process,
it works fine - but then a similar error occurs a couple of files later.
Under these circumstances, it doesn't seem to be g++ bug: Even
the nastiest g++ bugs are usually reproducible.
It is probably not a hardware problem: Bad RAM? No, gcc compilations
(which need much less memory, hence rarely swap) don't show up these
problems. Bad disk? No, I haven't had a single bad sector on these
disks in 4 years.

This happens with kernel 2.0.6 as well as will 2.0.17.

2. In a situation where everything should fit in memory. "swapoff -a" goes
like a breeze, "free" reports no used swap, but then - just a couple
of seconds later - the mouse freezes, only Ctrl-Alt-+/- still work,
and another couple of seconds later the system is completely frozen.
The "swapoff" apparently had bad interaction with the X server.

Bruno
<haible@ma2s2.mathematik.uni-karlsruhe.de>

\
 
 \ /
  Last update: 2005-03-22 13:37    [W:0.029 / U:0.040 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site