Re: [PATCH 2/9] mm: workingset: tell cache transitions from workingset thrashing

On Thu, Aug 02, 2018 at 08:57:31AM +0200, peter enderborg wrote:

On 08/01/2018 05:13 PM, Johannes Weiner wrote:

quoted

diff --git a/include/linux/page-flags.h b/include/linux/page-flags.h
index e34a27727b9a..7af1c3c15d8e 100644
--- a/include/linux/page-flags.h
+++ b/include/linux/page-flags.h

@@ -69,13 +69,14 @@
  */
 enum pageflags {
 	PG_locked,		/* Page is locked. Don't touch. */
-	PG_error,
 	PG_referenced,
 	PG_uptodate,
 	PG_dirty,
 	PG_lru,
 	PG_active,
+	PG_workingset,
 	PG_waiters,		/* Page has waiters, check its waitqueue. Must be bit #7 and in the same byte as "PG_locked" */
+	PG_error,
 	PG_slab,
 	PG_owner_priv_1,	/* Owner use. If pagecache, fs may use*/
 	PG_arch_1,

@@ -280,6 +281,8 @@ PAGEFLAG(Dirty, dirty, PF_HEAD) TESTSCFLAG(Dirty, dirty, PF_HEAD)

Any reason why the PG_error was moved? And dont you need to do some handling of this flag in proc/fs/page.c ?
Some KFP_WORKINGSET ?

I wanted PG_workingset next to PG_active as they both describe how hot
the page is, but PG_waiters needs to remain with the same bit number.

As far as fs/proc/page.c and include/uapi/linux/kernel-page-flags.h
go, that's a good point and we'll probably want to make that available
to userspace eventually. But I'm not super eager to make a brandnew
page flag user ABI right away. Let's give the code that uses it some
wider exposure first and maybe publish it a few release cycles later.

Thanks