> > You _really_ want the kernel to know that if some pages have to be free'd
> > for other use, these are the first candidates. So you should
> > cache and even
> > do read-ahead, but it shouldn't "pollute" the page/buffer-cache more than
> > needed.
>
> This can cause major problems if the cache is churning. The page could be
> discarded instants after it was read, before you had a chance to use it.
> This is closer to what you really want, but still not quite right.
Ok. You are right. Maybe "try_to_readahead" should be extended to keep
track of backwards (or forwards if the user reading backwards) usage, as
it does now for forward usage. This way, you keep track of an "active
window" within the file, and the rest can be treated as "use me first",
keeping the cache-trashing footprint of the app to the window-size.
Greetings,
Arjan van de Ven
-
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/