lkml.org 
[lkml]   [2015]   [Nov]   [8]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
From
SubjectRe: [PATCH 4/4] mm: prepare page_referenced() and page_idle to new THP refcounting
On Fri, Nov 06, 2015 at 02:39:00PM -0800, Andrew Morton wrote:
> On Fri, 6 Nov 2015 12:29:21 +0200 "Kirill A. Shutemov" <kirill@shutemov.name> wrote:
>
> > > page_mapcount() is getting pretty bad too.
> >
> > Do you want me to uninline slow path (PageCompound())?
>
> I guess so. Uninlining all of page_mapcount() does this:
>
> gcc-4.4.4:
>
> text data bss dec hex filename
> 973702 273954 831512 2079168 1fb9c0 mm/built-in.o-before
> 970148 273954 831000 2075102 1fa9de mm/built-in.o-after
>
> That's quite a bit of bloat.
>
> I don't know why bss changed; this usually (always?) happens. Seems
> bogus.

Here it is.

From 4bd3af3b6b9498254bd71e8288721dcff641156c Mon Sep 17 00:00:00 2001
From: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Date: Mon, 9 Nov 2015 01:34:08 +0200
Subject: [PATCH] mm: uninline slowpath of page_mapcount()

Let's move page_mapcount() part for compound page into mm/util.c.

make allyesconfig:

text data bss dec hex filename
188515051 153360535 85458720 427334306 19789aa2 vmlinux.o.before
188512917 153356439 85458720 427328076 1978824c vmlinux.o.after

Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
---
include/linux/mm.h | 14 +++++---------
mm/util.c | 14 ++++++++++++++
2 files changed, 19 insertions(+), 9 deletions(-)

diff --git a/include/linux/mm.h b/include/linux/mm.h
index f874d2a1d1a6..72edbbec7b91 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -417,19 +417,15 @@ static inline void page_mapcount_reset(struct page *page)
atomic_set(&(page)->_mapcount, -1);
}

+int __page_mapcount(struct page *page);
+
static inline int page_mapcount(struct page *page)
{
- int ret;
VM_BUG_ON_PAGE(PageSlab(page), page);

- ret = atomic_read(&page->_mapcount) + 1;
- if (PageCompound(page)) {
- page = compound_head(page);
- ret += atomic_read(compound_mapcount_ptr(page)) + 1;
- if (PageDoubleMap(page))
- ret--;
- }
- return ret;
+ if (unlikely(PageCompound(page)))
+ return __page_mapcount(page);
+ return atomic_read(&page->_mapcount) + 1;
}

#ifdef CONFIG_TRANSPARENT_HUGEPAGE
diff --git a/mm/util.c b/mm/util.c
index 902b65a43899..68535c0bb9da 100644
--- a/mm/util.c
+++ b/mm/util.c
@@ -376,6 +376,20 @@ struct address_space *page_mapping(struct page *page)
return mapping;
}

+/* Slow path of page_mapcount() for compound pages */
+int __page_mapcount(struct page *page)
+{
+ int ret;
+
+ page = compound_head(page);
+ ret = atomic_read(&page->_mapcount) + 1;
+ ret += atomic_read(compound_mapcount_ptr(page)) + 1;
+ if (PageDoubleMap(page))
+ ret--;
+ return ret;
+}
+EXPORT_SYMBOL_GPL(__page_mapcount);
+
int overcommit_ratio_handler(struct ctl_table *table, int write,
void __user *buffer, size_t *lenp,
loff_t *ppos)
--
Kirill A. Shutemov

\
 
 \ /
  Last update: 2015-11-09 01:01    [W:0.949 / U:0.788 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site