Patchwork PM / Hibernate: Don't mark pages dirty when reading pages while thawing

login
register
mail settings
Submitter Theodore Ts'o
Date Jan. 13, 2011, 5:59 a.m.
Message ID <1294898351-26156-1-git-send-email-tytso@mit.edu>
Download mbox | patch
Permalink /patch/78660/
State Not Applicable
Headers show

Comments

Theodore Ts'o - Jan. 13, 2011, 5:59 a.m.
Everything was sync'ed before the hibernation, so no pages could be
dirty.  So this causes a lot of wasted I/O activity right after
resuming from hibernation.

Worse, it also causes pages from files that were opened read/only to
be marked writeble which makes them subject to writeback.  This was
discovered when ext4 was changed to so that the jinode pointer was not
initialized unless the file was opened read/write, and this caused
things to blow up.  But that just unmasked a problem, since the pages
belonging to the file in question should have never been marked dirty
in the first place.  It increases the chances the text blocks for
executables like /usr/bin/killall will get corrupted when they are
needlessly written, and of course it means extra write cycles to the
SSD.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: Sebastian Ott <sebott@linux.vnet.ibm.com>
Cc: linux-pm@lists.linux-foundation.org
---
 kernel/power/block_io.c |    2 --
 1 files changed, 0 insertions(+), 2 deletions(-)
Sebastian Ott - Jan. 13, 2011, 12:36 p.m.
Hi,

On Thu, 13 Jan 2011, Theodore Ts'o wrote:
> Everything was sync'ed before the hibernation, so no pages could be
> dirty.  So this causes a lot of wasted I/O activity right after
> resuming from hibernation.
> 
> Worse, it also causes pages from files that were opened read/only to
> be marked writeble which makes them subject to writeback.  This was
> discovered when ext4 was changed to so that the jinode pointer was not
> initialized unless the file was opened read/write, and this caused
> things to blow up.  But that just unmasked a problem, since the pages
> belonging to the file in question should have never been marked dirty
> in the first place.  It increases the chances the text blocks for
> executables like /usr/bin/killall will get corrupted when they are
> needlessly written, and of course it means extra write cycles to the
> SSD.
> 
> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
> Cc: Sebastian Ott <sebott@linux.vnet.ibm.com>
> Cc: linux-pm@lists.linux-foundation.org
> ---
>  kernel/power/block_io.c |    2 --
>  1 files changed, 0 insertions(+), 2 deletions(-)
> 
> diff --git a/kernel/power/block_io.c b/kernel/power/block_io.c
> index 83bbc7c..108a4f3 100644
> --- a/kernel/power/block_io.c
> +++ b/kernel/power/block_io.c
> @@ -49,8 +49,6 @@ static int submit(int rw, struct block_device *bdev, sector_t sector,
>  	if (bio_chain == NULL) {
>  		submit_bio(bio_rw, bio);
>  		wait_on_page_locked(page);
> -		if (rw == READ)
> -			bio_set_pages_dirty(bio);
>  		bio_put(bio);
>  	} else {
>  		if (rw == READ)
> -- 
> 1.7.3.1
> 
> 

I did some test with this patch applied, but sadly it didn't help.
The testcase was reduced to one hibernation followed by a sync.

Regards,
Sebastian
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

diff --git a/kernel/power/block_io.c b/kernel/power/block_io.c
index 83bbc7c..108a4f3 100644
--- a/kernel/power/block_io.c
+++ b/kernel/power/block_io.c
@@ -49,8 +49,6 @@  static int submit(int rw, struct block_device *bdev, sector_t sector,
 	if (bio_chain == NULL) {
 		submit_bio(bio_rw, bio);
 		wait_on_page_locked(page);
-		if (rw == READ)
-			bio_set_pages_dirty(bio);
 		bio_put(bio);
 	} else {
 		if (rw == READ)