Desaster report

Martin Dalecki (dalecki@cs.net.pl)
Sat, 23 Oct 1999 12:24:32 +0200


After I got the time to try out the recent devel. kernels here are
my observations:

2.3.21: Works quite fine. However if I start an application which is
leaking memmory like a sive, the whole systems goes into a starvation
until the whole virtual memmory is exhausted.

2.3.22: Just dumps core upon apmd startup during the overall system
initialization. I didn't investigate further.

2.3.23:
1. Compilation:

a) bttv.c line 125 gives me a message about invalid operads to |

b) serial.c contains get_zeroed_page which is refferencing
__io_virt_debug.
This is prevendting the serial driver from working as a module.

BTW. The overall idea behind the kernel included system tables
as a means of a public interface is slowly but surely starting to look
like
a bad joke. At the current state one could as well fetch them
all just out of System.map as well and keep theyr maintainance
entierly in user space.

2. find /usr/src/linux -name "*.h"
works find.

find /usr/src/linux -name "*.h" -exec grep get_zeroed_page /dev/null {}
\;

crashed the system.

Under X11 the systems first starves, and after some time the X11
server get's killed and so on until there isn't much of the systme
living.

Under the console the systm is going to crash too, however I can
see the follogin messages (plenty of them in fact...)

VM: killing process klogd
....

and after some time (out 1 minute)

VML killing process init

sic.

The setup I have is quite usual: K6III/550 MHZ 64MB RAM and so on.
The only mybe unusual thing about it is the fact that I'm running
the swap partintion on a separate drive - which indeed makes perfectly
sense performance wise. Oh and yes the disk system is IDE.

No wonder none feels oblidget do send driver updates as long as long
the overall systems remains shaky like this. (I for one
don't like to shoot at moving targets...)
I for personal wouldn't even bother to try tracking the developement.
Note that you are supposed to be in feature freeze mode by now, and
during the developement of the previous version there wasn't ever
something like a code freeze before the release.

--Marcin

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/