[lkml]   [1998]   [Oct]   [16]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: -j zImage
On Thu, Oct 15, 1998 at 01:38:35PM +0200, Andrea Arcangeli wrote:
> On Wed, 14 Oct 1998, Koblinger Egmont wrote:


> >With 2.1.125, it always crashed after about 2 or 3 minutes. Crashing means
> >here: the only thing I could do is switch between virtual consoles and press
> >alt+sysrq+key. Alt+sysrq+s and u writes "syncing" and "remounting" to the

I have been doing _make -j zImage modules_ for a very long time now but I
have never had any lookups. I realised the change Linus made only when the
compile did not complete successfully because it exceeded my resources (nr
of processes). Last time I tried it I got

make[2]: vfork: Nicht genügend Hauptspeicher verfügbar

which basically means out of memory.

> Apply my OOM fix. Right now the MM subsystem is screwed up since kswapd
> start running all the time when Linux gies near OOM.
> oom-11 is the latest update so far (patches before are buggy btw).
> Apply it to 2.1.125 and your machine will be rock stable.

I have applied your patch, build a new kernel and rebooted. The problem
with out of memory seems to have become worse. This is a
_cat /proc/meminfo_ right after the first oom message:

total: used: free: shared: buffers: cached:
Mem: 131215360 128376832 2838528 123215872 6709248 11931648
Swap: 172650496 15921152 156729344
MemTotal: 128140 kB
MemFree: 2772 kB
MemShared: 120328 kB
Buffers: 6552 kB
Cached: 11652 kB
SwapTotal: 168604 kB
SwapFree: 153056 kB

As you can see I really have plenty of memory still available.

> >I have an intel-486 machine with 32M ram and 52412K swap, but I also tried
> >compiling the kernel with the swap turned off, and it still worked under
> >2.0.36 but caused 2.1.125 to crash.

My system is a dual P133 with 128M of ram and appr. 150M of swap.

> Try with my patch applyed and you will see. In the stock kernel when there
> is no swap the kswapd-deadlock is still more effective since kswapd never
> schedule() due wait_for_io() and so it really load every available CPU
> cycle.

Maybe I should try Rik's addition. Where can I find it?




Well, yeah ... I suppose there's no point in getting greedy, is there?

<< Calvin & Hobbes >>

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
Please read the FAQ at

 \ /
  Last update: 2005-03-22 13:45    [W:0.079 / U:11.624 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site