| 1 |
2005. május 6. 00.36 dátummal Casey Allen Shobe - SeattleServer Mailing Lists |
| 2 |
ezt írta: |
| 3 |
> This patch has benefit ONLY to systems with exactly 1GB of RAM. Systems |
| 4 |
> with any less won't hit high memory anyways, and systems with any more are |
| 5 |
> going to have to use high memory support anyways. |
| 6 |
> |
| 7 |
> Regardless of the reasons that the LOWMEM limit is where it is with or |
| 8 |
> without |
| 9 |
|
| 10 |
I've read that it's now in gentoo-dev-sources, and i think it's important to |
| 11 |
say its a configurable option in kernel config and marked as experimental.. |
| 12 |
Many users marked it as stable. |
| 13 |
|
| 14 |
off: I've the latest hardened-sources installed. i found my system in a very |
| 15 |
unstable state (memory leaks), after one week of serious debugging i found it |
| 16 |
was because pax enabled (configured exactly as in the hardened pax howto). |
| 17 |
|
| 18 |
actually the problem was after 0,5-1 hours of uptime all of the ram was |
| 19 |
getting used and the system started to use swap..and programs started to fly |
| 20 |
away with out of memory errors.. av load was about 1-3 |
| 21 |
|
| 22 |
of course the server got the same load from outside and inside apps as it was |
| 23 |
before (avg 0.10-0.30).. the problem appeared after reboot when apache |
| 24 |
started (in threads mode, with about 80-130 threads at startup). Then i |
| 25 |
thought its because apache, reemerged glibc with nptl and the problem |
| 26 |
disappeared.. |
| 27 |
Today i had the same with another thread based traffic-proxy (with nptl too).. |
| 28 |
Now i turned off Pax and as i guessed the problem disappeard (hopely) |
| 29 |
forever.. The system is running with avg 80-130Mb of real memory usage (+- |
| 30 |
buffs) |
| 31 |
|
| 32 |
so, pax has some serius memory leaks and its in the hardened but none the less |
| 33 |
is not in vanilla kernels :) |
| 34 |
|
| 35 |
And there was another buggie with pax.. when i compiled my kernel with highmem |
| 36 |
support, after 1 min of booting in and starting daemons the kernel just |
| 37 |
freezed (oops).. |
| 38 |
|
| 39 |
that was the reason why not to enable highmem and the known lowmem vs highmem |
| 40 |
performance issues. But now i think i stop using pax for a while.. |
| 41 |
|
| 42 |
szab |
| 43 |
|
| 44 |
-- |
| 45 |
gentoo-hardened@g.o mailing list |
Archives