On Mon, 14 Dec 1998 11:57:00 +0000, Neil Conway
<nconway.list@ukaea.org.uk> said:
> I've mentioned this a couple of times in posts. After talking to Rik
> and Stephen, it seems that this it totally normal though I still don't
> completely understand it, and is related in some way to swap-cache.
> Now I think Stephen agrees that this at least needs to be documented and
> reported differently by the system so that 'free' doesn't report swap
> and cache being used when they aren't really (at least in the case of
> the swap, reporting it used is bogus).
The "cache" reporting is correct. Anyone who reads the wrong things
into the kernel's cache size just needs to be reeducated: document it,
yes, but don't fix the accounting.
As for the lazy swap cleanup, there is an obvious fix which is a post-2.2
item because of the subtle way it changes VM semantics: we need to
avoid acounting the swap cache as a reference to the swap entry. This
is more tricky than it sounds, because it means we need to start
accounting readonly swap-cached pages mapped in VM as still having a
swap reference on disk (currently, such pages rely on the swap cache's
reference to the swap entry to ensure that the on-disk swap page is not
reused). That changes too many bits of VM too subtly to be done during
2.2 codefreeze. Any other solution is just the wrong solution.
--Stephen
-
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/