BUG: Memory usage, 2.4.10 meaning of "free" and cached.

Elmer Joandi (elmer@linking.ee)
Tue, 9 Oct 2001 04:55:53 +0200 (EET)


------=_?hOe4ym7uCUUmuF/Ihy2ac+g'JRBhu6Lcs6I93'8k6_1+.9Kt((9d7+PLMc6
Content-Type: text/plain; charset=iso-8859-15
Content-Transfer-Encoding: 8bit

About 1 G of RAM
Squid, mysql, bunch of other daemons and programs configured to take about 500MB(lots of
chroot boxes).
/tmp 's mounted as tmpfs

With 2.4.2-ac28(also tmpfs) there was 500MB cached(or somethimes on random occasions
lots of buffered). And swap being used, 300MB after pushing the computer to swap.

Now, with 2.4.10....

There is everything used (only 120M cached, 5M buffered) ,exept swap(10M),
When starting applications like soffice,mozilla and other hogs, the cached number
... grows...(and free degrades a bit if it is free at particular moment).
hovever, numbers of memory usage are much higher than with 2.4.2-ac28
stoping hogs gives us some free memory (looks like someone installed RAM on the fly
:) (not total of it, of course :) ) )

Now, taking a 300MB file with joe, opening it... the cached grows(probably as joe
makes temporary on tmpfs), swap usage grows minimally.
Quiting joe... free grows (but swap is still used minimally, 13M) and now there is plenty
of free and lots of cached(i.e. same amount of programs and memory used by
applications, but, compared to start, there has come about some hundred megs of free
memory from somewhere)

1. It seems that the meaning of "free" or "free+cached+buffered" in top has changed over the
time a bit ? It seems that all of cached is not reported as cached, and after stressing memory
and freeing those hidden cached there is more "free+cached+buffered" at the end of
memory-stressing programs run despite that no other programs have released any signifficant
amount of memory.
2. How to tell kernel to use swap more agressively ?

------=_?hOe4ym7uCUUmuF/Ihy2ac+g'JRBhu6Lcs6I93'8k6_1+.9Kt((9d7+PLMc6
Content-Type: application/octet-stream;name=""
Content-Disposition: attachment; filename=""
Content-Transfer-Encoding: base64

DQo=

------=_?hOe4ym7uCUUmuF/Ihy2ac+g'JRBhu6Lcs6I93'8k6_1+.9Kt((9d7+PLMc6--

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