Re: tmpfs swapoff oddity

From: Christoph Rohland (cr@sap.com)
Date: Thu Feb 08 2001 - 05:17:16 EST


Hi Mike,

On Thu, 8 Feb 2001, Mike Galbraith wrote:
> Hi Christoph,
>
> While testing Jens' loop-4 patch (and not being able to find
> any way to lock it up), I stumbled onto a strange behavior.
>
> I set up an interleaved swap with one swap partition, and one
> swapfile in a loopback mounted reiserfs - populated tmpfs with
> a kernel tree and did hefty make -j kernel builds to generate
> much I/O. Afterward (bored), I figured I'd bounce the data in
> tmpfs back and forth between swap containers with swapoff. It
> took much longer than I expected.

Oh, the swapoff handling in Linux is much less then suboptimal. So I
would expect that.

To explain: For every single page on swap we scan all processes vmas
and all shm/tmpfs objects swap tables :-( And we have to hold some
locks for that...

Greetings
                Christoph

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



This archive was generated by hypermail 2b29 : Thu Feb 15 2001 - 21:00:11 EST