Re: [PATCH] 2.6.4-rc2-mm1: vm-split-active-lists

[Nick Piggin <piggin@cyberone.com.au>]

Andrew Morton wrote:

>Nick Piggin <piggin@cyberone.com.au> wrote:
>
>>Just had a try of doing things like updatedb and dd if=/dev/zero of=./blah
>>It is pretty swappy I guess.
>>
>
>You'll need to bring the scanning priority back into the picture: don't
>move mapped pages down onto the inactive list at low scanning priorities. 
>And that eans retaining the remember-the-priority-from-last-time logic.
>
>Otherwise it's inevitable that even a `cat monster_file > /dev/null' will
>eventually swap out everything it can.
>
>

Hmm I dunno. At mapped_page_cost 8, I don't think it is swappy enough
that your desktop users will be running into problems. I need to write
4GB of file to push out 70MB of swap here (256MB RAM). And not much of
that swap has come back in, by the way...

>>By the way, I would be interested to know the rationale behind
>>mark_page_accessed as it is without this patch, also what is it doing in
>>rmap.c (I know hardly anything actually uses page_test_and_clear_young, but
>>still). It seems to me like it only serves to make VM behaviour harder to
>>understand, but I'm probably missing something. Andrew?
>>
>
>hm, that's left-over code which is pretty pointless now.
>
>
>	if (page_test_and_clear_young(page))
>		mark_page_accessed(page);
>
>	if (TestClearPageReferenced(page))
>		referenced++;
>
>The pages in here are never on the LRU, so all the mark_page_accessed()
>will do is to set PG_Referenced.  And we immediately clear it again.  So
>the mark_page_accessed() can be replaced with referenced++.
>
>
>

Yep, see the patch I'd attached before.