LKML Archive on lore.kernel.org
help / color / mirror / Atom feed
From: Martin Schwidefsky <schwidefsky@de.ibm.com>
To: linux-kernel@vger.kernel.org, linux-s390@vger.kernel.org
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Subject: [patch 34/38] split page_test_and_clear_dirty.
Date: Fri, 27 Apr 2007 16:05:37 +0200 [thread overview]
Message-ID: <20070427140519.911265225@de.ibm.com> (raw)
In-Reply-To: <20070427140503.087958775@de.ibm.com>
[-- Attachment #1: 137-test-clear-dirty.diff --]
[-- Type: text/plain, Size: 4300 bytes --]
From: Martin Schwidefsky <schwidefsky@de.ibm.com>
The page_test_and_clear_dirty primitive really consists of two
operations, page_test_dirty and the page_clear_dirty. The combination
of the two is not an atomic operation, so it makes more sense to have
two separate operations instead of one.
In addition to the improved readability of the s390 version of
SetPageUptodate, it now avoids the page_test_dirty operation which is
an insert-storage-key-extended (iske) instruction which is an expensive
operation.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
---
include/asm-generic/pgtable.h | 11 +++++++++--
include/asm-s390/pgtable.h | 15 ++++++++-------
include/linux/page-flags.h | 2 +-
mm/rmap.c | 8 ++++++--
4 files changed, 24 insertions(+), 12 deletions(-)
Index: quilt-2.6/include/asm-generic/pgtable.h
===================================================================
--- quilt-2.6.orig/include/asm-generic/pgtable.h 2007-04-27 16:01:49.000000000 +0200
+++ quilt-2.6/include/asm-generic/pgtable.h 2007-04-27 16:05:01.000000000 +0200
@@ -139,8 +139,15 @@
#define pte_same(A,B) (pte_val(A) == pte_val(B))
#endif
-#ifndef __HAVE_ARCH_PAGE_TEST_AND_CLEAR_DIRTY
-#define page_test_and_clear_dirty(page) (0)
+#ifndef __HAVE_ARCH_PAGE_TEST_DIRTY
+#define page_test_dirty(page) (0)
+#endif
+
+#ifndef __HAVE_ARCH_PAGE_CLEAR_DIRTY
+#define page_clear_dirty(page) do { } while (0)
+#endif
+
+#ifndef __HAVE_ARCH_PAGE_TEST_DIRTY
#define pte_maybe_dirty(pte) pte_dirty(pte)
#else
#define pte_maybe_dirty(pte) (1)
Index: quilt-2.6/include/asm-s390/pgtable.h
===================================================================
--- quilt-2.6.orig/include/asm-s390/pgtable.h 2007-04-27 16:01:49.000000000 +0200
+++ quilt-2.6/include/asm-s390/pgtable.h 2007-04-27 16:05:01.000000000 +0200
@@ -753,14 +753,14 @@
* should therefore only be called if it is not mapped in any
* address space.
*/
-static inline int page_test_and_clear_dirty(struct page *page)
+static inline int page_test_dirty(struct page *page)
{
- unsigned long physpage = page_to_phys(page);
- int skey = page_get_storage_key(physpage);
+ return (page_get_storage_key(page_to_phys(page)) & _PAGE_CHANGED) != 0;
+}
- if (skey & _PAGE_CHANGED)
- page_set_storage_key(physpage, skey & ~_PAGE_CHANGED);
- return skey & _PAGE_CHANGED;
+static inline void page_clear_dirty(struct page *page)
+{
+ page_set_storage_key(page_to_phys(page), PAGE_DEFAULT_KEY);
}
/*
@@ -953,7 +953,8 @@
#define __HAVE_ARCH_PTEP_CLEAR_FLUSH
#define __HAVE_ARCH_PTEP_SET_WRPROTECT
#define __HAVE_ARCH_PTE_SAME
-#define __HAVE_ARCH_PAGE_TEST_AND_CLEAR_DIRTY
+#define __HAVE_ARCH_PAGE_TEST_DIRTY
+#define __HAVE_ARCH_PAGE_CLEAR_DIRTY
#define __HAVE_ARCH_PAGE_TEST_AND_CLEAR_YOUNG
#include <asm-generic/pgtable.h>
Index: quilt-2.6/include/linux/page-flags.h
===================================================================
--- quilt-2.6.orig/include/linux/page-flags.h 2007-04-27 16:01:49.000000000 +0200
+++ quilt-2.6/include/linux/page-flags.h 2007-04-27 16:05:01.000000000 +0200
@@ -133,7 +133,7 @@
static inline void SetPageUptodate(struct page *page)
{
if (!test_and_set_bit(PG_uptodate, &page->flags))
- page_test_and_clear_dirty(page);
+ page_clear_dirty(page);
}
#else
#define SetPageUptodate(page) set_bit(PG_uptodate, &(page)->flags)
Index: quilt-2.6/mm/rmap.c
===================================================================
--- quilt-2.6.orig/mm/rmap.c 2007-04-27 16:01:49.000000000 +0200
+++ quilt-2.6/mm/rmap.c 2007-04-27 16:05:01.000000000 +0200
@@ -498,8 +498,10 @@
struct address_space *mapping = page_mapping(page);
if (mapping)
ret = page_mkclean_file(mapping, page);
- if (page_test_and_clear_dirty(page))
+ if (page_test_dirty(page)) {
+ page_clear_dirty(page);
ret = 1;
+ }
}
return ret;
@@ -605,8 +607,10 @@
* Leaving it set also helps swapoff to reinstate ptes
* faster for those pages still in swapcache.
*/
- if (page_test_and_clear_dirty(page))
+ if (page_test_dirty(page)) {
+ page_clear_dirty(page);
set_page_dirty(page);
+ }
__dec_zone_page_state(page,
PageAnon(page) ? NR_ANON_PAGES : NR_FILE_MAPPED);
}
--
blue skies,
Martin.
"Reality continues to ruin my life." - Calvin.
next prev parent reply other threads:[~2007-04-27 14:09 UTC|newest]
Thread overview: 53+ messages / expand[flat|nested] mbox.gz Atom feed top
2007-04-27 14:05 [patch 00/38] s390 patches for 2.6.22 Martin Schwidefsky
2007-04-27 14:05 ` [patch 01/38] memory detection: stop at first memory hole Martin Schwidefsky
2007-04-27 14:05 ` [patch 02/38] cio/ipl: Clean interface between cio and ipl code Martin Schwidefsky
2007-04-27 14:05 ` [patch 03/38] cio: Introduce struct chp_id Martin Schwidefsky
2007-04-27 14:05 ` [patch 04/38] cio: Allow 0 and 1 as input for channel path status attribute Martin Schwidefsky
2007-04-27 14:05 ` [patch 05/38] cio: Introduce separate files for channel-path related code Martin Schwidefsky
2007-04-27 14:05 ` [patch 06/38] cio: observe chpid valid flag Martin Schwidefsky
2007-04-27 14:05 ` [patch 07/38] cio: Clean up online_store Martin Schwidefsky
2007-04-27 14:05 ` [patch 08/38] cio: Channel-path configure function Martin Schwidefsky
2007-04-27 14:05 ` [patch 09/38] cio: Use add_uevent_var Martin Schwidefsky
2007-04-27 14:05 ` [patch 10/38] cio: Re-start path verification after aborting internal I/O Martin Schwidefsky
2007-04-27 14:05 ` [patch 11/38] cio: replace subchannel evaluation queue with bitmap Martin Schwidefsky
2007-04-27 14:05 ` [patch 12/38] cio: fix subchannel channel-path data usage Martin Schwidefsky
2007-04-27 14:05 ` [patch 13/38] cio: Dont call css_update_ssd_info from interrupt context Martin Schwidefsky
2007-04-27 14:05 ` [patch 14/38] cio: ccwgroup register vs. unregister Martin Schwidefsky
2007-04-27 14:05 ` [patch 15/38] cio: cm_enable memory leak Martin Schwidefsky
2007-04-27 14:05 ` [patch 16/38] cio: Unregister ccw devices directly Martin Schwidefsky
2007-04-27 14:05 ` [patch 17/38] System call cleanup Martin Schwidefsky
2007-04-27 14:05 ` [patch 18/38] Improved oops output Martin Schwidefsky
2007-04-27 14:25 ` Christoph Hellwig
2007-04-27 15:24 ` Martin Schwidefsky
2007-04-27 16:01 ` Chuck Ebbert
2007-04-27 16:15 ` Martin Schwidefsky
2007-04-27 16:21 ` Chuck Ebbert
2007-04-27 18:14 ` Andi Kleen
2007-04-27 20:56 ` Martin Schwidefsky
2007-04-27 21:10 ` Chuck Ebbert
2007-04-27 14:05 ` [patch 19/38] Use generic bug Martin Schwidefsky
2007-04-27 14:05 ` [patch 20/38] Minor fault path optimization Martin Schwidefsky
2007-04-28 4:49 ` Paul Mackerras
2007-04-28 8:34 ` Christoph Hellwig
2007-04-30 0:57 ` Paul Mackerras
2007-04-30 10:56 ` [PATCH] powerpc: minor " Christoph Hellwig
2007-05-02 10:45 ` [patch 20/38] Minor " Martin Schwidefsky
2007-05-02 10:49 ` Christoph Hellwig
2007-04-27 14:05 ` [patch 21/38] No execute support cleanup Martin Schwidefsky
2007-04-27 14:05 ` [patch 22/38] Get rid of console setup functions Martin Schwidefsky
2007-04-27 14:05 ` [patch 23/38] Improved kernel stack overflow checking Martin Schwidefsky
2007-04-27 14:05 ` [patch 24/38] dasd: Add sysfs attribute status and generate uevents Martin Schwidefsky
2007-04-27 14:05 ` [patch 25/38] dasd: Add ipldev parameter Martin Schwidefsky
2007-04-27 14:05 ` [patch 26/38] zfcpdump support Martin Schwidefsky
2007-04-27 14:05 ` [patch 27/38] ctc: kmalloc->kzalloc/casting cleanups Martin Schwidefsky
2007-04-27 14:05 ` [patch 28/38] sclp: initialize early Martin Schwidefsky
2007-04-27 14:05 ` [patch 29/38] vmlogrdr: stop IUCV connection in vmlogrdr_release Martin Schwidefsky
2007-04-27 14:05 ` [patch 30/38] sclp: fix coding style Martin Schwidefsky
2007-04-27 14:05 ` [patch 31/38] crypto: cleanup Martin Schwidefsky
2007-04-27 14:05 ` [patch 32/38] vtime: cleanup per_cpu usage Martin Schwidefsky
2007-04-27 14:05 ` [patch 33/38] Processor degradation notification Martin Schwidefsky
2007-04-27 14:05 ` Martin Schwidefsky [this message]
2007-04-27 14:05 ` [patch 35/38] Switch etr from tasklet to workqueue Martin Schwidefsky
2007-04-27 14:05 ` [patch 36/38] Remove debugging junk Martin Schwidefsky
2007-04-27 14:05 ` [patch 37/38] Clean up smp code in preparation for some larger changes Martin Schwidefsky
2007-04-27 14:05 ` [patch 38/38] SPIN_LOCK_UNLOCKED cleanup in drivers/s390 Martin Schwidefsky
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20070427140519.911265225@de.ibm.com \
--to=schwidefsky@de.ibm.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-s390@vger.kernel.org \
--subject='Re: [patch 34/38] split page_test_and_clear_dirty.' \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).