On Thu, 22 Jun 2000, Andrea Arcangeli wrote:
> On Thu, 22 Jun 2000, Rik van Riel wrote:
>
> >*nod* Then again, if we're in serious overload situation we
> >will just *have* to wait somewhere. There is no possibility
> >to just get some memory in a magic way.
>
> The balance_dirty() mechanims make sure that half of the
> cache+free_memory remains clean. Thus we could block there even
> if there's something clean in the next entry of the LRU.
Indeed. This means that we should have bdflush start the writing
*before* we run out of request slots or have too many dirty pages.
> I think instead of blocking there we could try to let know
> balance_dirty that the amount of dirty stuff in the system
> should decrease because there's memory pressure and we need to
> reclaim also half of such cache _fast_.
This is exactly what I am doing in my patch, except that I'd
*really* like to keep the "hint" to bdflush that we want the
oldest pages (from the point of view of page replacement)
flushed so that we can make a better effort at freeing the
right pages. Not doing that is effectively saying that you
no longer care about freeing the right page from memory...
I'm currently experimenting with maxlaunder logic, more or
less like FreeBSD has.
regards,
Rik
-- The Internet is not a network of computers. It is a network of people. That is its real strength.Wanna talk about the kernel? irc.openprojects.net / #kernelnewbies http://www.conectiva.com/ http://www.surriel.com/
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Fri Jun 23 2000 - 21:00:24 EST