Hello,
On 4/19/06, Eric Lowe <[EMAIL PROTECTED]> wrote:
> Ivan wrote:
> > I've read a lot of the Memory Subsystem in the Solaris Internals Book
> > during my holidays. I'm very interested in the Page Replacement Code
> > (schedpaging, pageout, ....) and I'm wondering whether much work happens
> > today on this part of the kernel?
>
> We're completely re-writing fsflush and the pageout scanner in Solaris as
> part of the memory sets project, which provides physical memory controls
> for resource pools. I have a half-completed blog entry describing the
> project which hopefully I'll get posted in a week or so. In the new
> framework, each memory set (typically associated with a zone) will get its
> own pageout scanner threads and fsflush thread, which will only scan the
> pages in use by that pool (zone). The new scanners are completely
> vnode-based rather than page-based; also fsflush is MUCH more efficient
> since it only needs to scan file vnodes with dirty pages.
I'm already seeing this in cvs.osol.org:
265 pgcnt_t v_npages; /* # pages on this vnode */
266 pgcnt_t v_msnpages; /* # pages charged to v_mset */
267 struct page *v_scanfront; /* scanner front hand */
268 struct page *v_scanback; /* scanner back hand */
Eagerly awaiting that blog post of yours Eric. :)
--
Regards,
Jeremy
_______________________________________________
opensolaris-code mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
