First problem observed:
There may be a problem with a memory leak in Linux-2.1.49. This is the
SMP version.
Conditions: No network daemons, only two getties, only
the root file-system mounted (1.2gb, 91% full):
MemTotal: 46900 kB
MemFree: 35532 kB
MemShared: 6428 kB
Buffers: 1796 kB
Cached: 4448 kB
SwapTotal: 100344 kB
SwapFree: 100344 kB
I execute `ls -R /` then I have this memory available:
MemTotal: 46900 kB
MemFree: 1016 kB
MemShared: 6432 kB
Buffers: 11732 kB
Cached: 4600 kB
SwapTotal: 100344 kB
SwapFree: 100344 kB
I execute `ls -R /` again. Then I have this memory available:
MemTotal: 46900 kB
MemFree: 768 kB
MemShared: 6432 kB
Buffers: 12228 kB
Cached: 4608 kB
SwapTotal: 100344 kB
SwapFree: 100336 kB
I execute `ls -R /` again. Then I have this memory available:
MemTotal: 46900 kB
MemFree: 468 kB
MemShared: 6432 kB
Buffers: 12348 kB
Cached: 1724 kB
SwapTotal: 100344 kB
SwapFree: 100124 kB
Free memory goes down, swap usage goes up. If I execute this is a loop
from a shell-script for about 3 hours, I can no longer fork and the shell
fails.
Second problem observed:
Ftp between two linux machines, both running Linux 2.1.49. If I
attempt to 'get' two files in a row, without waiting for several
seconds, I get:
bind: Address already in use.
This might be caused by sockets remaining in their "CLOSING" state for
a long time, i.e.:
Active Internet connections
Proto Recv-Q Send-Q Local Address Foreign Address (State) User
tcp 1 1 chaos.analogic.co:smtp joshua.fuller.edu:1846 CLOSING root
Active UNIX domain sockets
Proto RefCnt Flags Type State Path
This mail from linux-kernel was received over an hour ago. It should not
still be in the "CLOSING" state. It will eventually go away but after
other states such as "TIME_WAIT", etc.
Cheers,
DJ
Richard B. Johnson
Analogic Corporation
Penguin : Linux version 2.1.49 on an i586 machine (66.15 BogoMips).
Warning : It's hard to stay on the trailing edge of technology.
Linux : Engineering tool
Windows : Typewriter