Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1479753 > unrolled thread

[PATCH] PM / Hibernate: allow hibernation with PAGE_POISONING_ZERO

Back to article view | Back to linux.kernel

Contents

  [PATCH] PM / Hibernate: allow hibernation with PAGE_POISONING_ZERO Anisse Astier <anisse@astier.eu> - 2016-09-09 10:50 +0200
    Re: [PATCH] PM / Hibernate: allow hibernation with PAGE_POISONING_ZERO Kees Cook <keescook@chromium.org> - 2016-09-09 20:30 +0200
    Re: [PATCH] PM / Hibernate: allow hibernation with  PAGE_POISONING_ZERO Pavel Machek <pavel@ucw.cz> - 2016-09-12 13:40 +0200
      Re: [PATCH] PM / Hibernate: allow hibernation with  PAGE_POISONING_ZERO Pavel Machek <pavel@ucw.cz> - 2016-09-12 19:50 +0200
    Re: [PATCH] PM / Hibernate: allow hibernation with PAGE_POISONING_ZERO "Rafael J. Wysocki" <rjw@rjwysocki.net> - 2016-09-14 03:00 +0200

#1479753 — [PATCH] PM / Hibernate: allow hibernation with PAGE_POISONING_ZERO

PAGE_POISONING_ZERO disables zeroing new pages on alloc, they are
poisoned (zeroed) as they become available.
In the hibernate use case, free pages will appear in the system without
being cleared, left there by the loading kernel.

This patch will make sure free pages are cleared on resume when
PAGE_POISONING_ZERO is enabled. We free the pages just after resume
because we can't do it later: going through any device resume code might
allocate some memory and invalidate the free pages bitmap.

Thus we don't need to disable hibernation when PAGE_POISONING_ZERO is
enabled.

Signed-off-by: Anisse Astier <anisse@astier.eu>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Laura Abbott <labbott@fedoraproject.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Rafael J. Wysocki <rjw@rjwysocki.net>
---
 kernel/power/hibernate.c | 21 +++------------------
 kernel/power/power.h     |  2 ++
 kernel/power/snapshot.c  | 22 ++++++++++++++++++++++
 mm/Kconfig.debug         |  2 --
 4 files changed, 27 insertions(+), 20 deletions(-)

diff --git a/kernel/power/hibernate.c b/kernel/power/hibernate.c
index 33c79b6..b26dbc4 100644
--- a/kernel/power/hibernate.c
+++ b/kernel/power/hibernate.c
@@ -306,8 +306,10 @@ static int create_image(int platform_mode)
 	if (error)
 		printk(KERN_ERR "PM: Error %d creating hibernation image\n",
 			error);
-	if (!in_suspend)
+	if (!in_suspend) {
 		events_check_enabled = false;
+		clear_free_pages();
+	}
 
 	platform_leave(platform_mode);
 
@@ -1189,22 +1191,6 @@ static int __init nohibernate_setup(char *str)
 	return 1;
 }
 
-static int __init page_poison_nohibernate_setup(char *str)
-{
-#ifdef CONFIG_PAGE_POISONING_ZERO
-	/*
-	 * The zeroing option for page poison skips the checks on alloc.
-	 * since hibernation doesn't save free pages there's no way to
-	 * guarantee the pages will still be zeroed.
-	 */
-	if (!strcmp(str, "on")) {
-		pr_info("Disabling hibernation due to page poisoning\n");
-		return nohibernate_setup(str);
-	}
-#endif
-	return 1;
-}
-
 __setup("noresume", noresume_setup);
 __setup("resume_offset=", resume_offset_setup);
 __setup("resume=", resume_setup);
@@ -1212,4 +1198,3 @@ __setup("hibernate=", hibernate_setup);
 __setup("resumewait", resumewait_setup);
 __setup("resumedelay=", resumedelay_setup);
 __setup("nohibernate", nohibernate_setup);
-__setup("page_poison=", page_poison_nohibernate_setup);
diff --git a/kernel/power/power.h b/kernel/power/power.h
index 242d8b8..56d1d0d 100644
--- a/kernel/power/power.h
+++ b/kernel/power/power.h
@@ -110,6 +110,8 @@ extern int create_basic_memory_bitmaps(void);
 extern void free_basic_memory_bitmaps(void);
 extern int hibernate_preallocate_memory(void);
 
+extern void clear_free_pages(void);
+
 /**
  *	Auxiliary structure used for reading the snapshot image data and
  *	metadata from and writing them to the list of page backup entries
diff --git a/kernel/power/snapshot.c b/kernel/power/snapshot.c
index b022284..4f0f060 100644
--- a/kernel/power/snapshot.c
+++ b/kernel/power/snapshot.c
@@ -1132,6 +1132,28 @@ void free_basic_memory_bitmaps(void)
 	pr_debug("PM: Basic memory bitmaps freed\n");
 }
 
+void clear_free_pages(void)
+{
+#ifdef CONFIG_PAGE_POISONING_ZERO
+	struct memory_bitmap *bm = free_pages_map;
+	unsigned long pfn;
+
+	if (WARN_ON(!(free_pages_map)))
+		return;
+
+	memory_bm_position_reset(bm);
+	pfn = memory_bm_next_pfn(bm);
+	while (pfn != BM_END_OF_MAP) {
+		if (pfn_valid(pfn))
+			clear_highpage(pfn_to_page(pfn));
+
+		pfn = memory_bm_next_pfn(bm);
+	}
+	memory_bm_position_reset(bm);
+	pr_info("PM: free pages cleared after restore\n");
+#endif /* PAGE_POISONING_ZERO */
+}
+
 /**
  * snapshot_additional_pages - Estimate the number of extra pages needed.
  * @zone: Memory zone to carry out the computation for.
diff --git a/mm/Kconfig.debug b/mm/Kconfig.debug
index 22f4cd9..afcc550 100644
--- a/mm/Kconfig.debug
+++ b/mm/Kconfig.debug
@@ -76,8 +76,6 @@ config PAGE_POISONING_ZERO
 	   no longer necessary to write zeros when GFP_ZERO is used on
 	   allocation.
 
-	   Enabling page poisoning with this option will disable hibernation
-
 	   If unsure, say N
 	bool
 
-- 
2.7.4

[toc] | [next] | [standalone]

#1480246

On Fri, Sep 9, 2016 at 1:43 AM, Anisse Astier <anisse@astier.eu> wrote:
> PAGE_POISONING_ZERO disables zeroing new pages on alloc, they are
> poisoned (zeroed) as they become available.
> In the hibernate use case, free pages will appear in the system without
> being cleared, left there by the loading kernel.
>
> This patch will make sure free pages are cleared on resume when
> PAGE_POISONING_ZERO is enabled. We free the pages just after resume
> because we can't do it later: going through any device resume code might
> allocate some memory and invalidate the free pages bitmap.
>
> Thus we don't need to disable hibernation when PAGE_POISONING_ZERO is
> enabled.
>
> Signed-off-by: Anisse Astier <anisse@astier.eu>
> Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
> Cc: Laura Abbott <labbott@fedoraproject.org>
> Cc: Mel Gorman <mgorman@suse.de>
> Cc: Rafael J. Wysocki <rjw@rjwysocki.net>

Cool!

Reviewed-by: Kees Cook <keescook@chromium.org>

-Kees

-- 
Kees Cook
Nexus Security

[toc] | [prev] | [next] | [standalone]

#1481159 — Re: [PATCH] PM / Hibernate: allow hibernation with PAGE_POISONING_ZERO

On Fri 2016-09-09 10:43:32, Anisse Astier wrote:
> PAGE_POISONING_ZERO disables zeroing new pages on alloc, they are
> poisoned (zeroed) as they become available.
> In the hibernate use case, free pages will appear in the system without
> being cleared, left there by the loading kernel.
> 
> This patch will make sure free pages are cleared on resume when
> PAGE_POISONING_ZERO is enabled. We free the pages just after resume
> because we can't do it later: going through any device resume code might
> allocate some memory and invalidate the free pages bitmap.
> 
> Thus we don't need to disable hibernation when PAGE_POISONING_ZERO is
> enabled.
> 
> Signed-off-by: Anisse Astier <anisse@astier.eu>
> Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
> Cc: Laura Abbott <labbott@fedoraproject.org>
> Cc: Mel Gorman <mgorman@suse.de>
> Cc: Rafael J. Wysocki <rjw@rjwysocki.net>

Looks reasonable to me.

Acked-by: Pavel Machek <pavel@ucw.cz>

Actually.... this takes basically zero time come. Do we want to do it
unconditionally?

(Yes, it is free memory, but for sake of debugging, I guess zeros are
preffered to random content that changed during hibernation.)

(But that does not change the Ack.)

Best regards,
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

[toc] | [prev] | [next] | [standalone]

#1481750 — Re: [PATCH] PM / Hibernate: allow hibernation with PAGE_POISONING_ZERO

Hi!

On Mon 2016-09-12 17:19:54, Anisse Astier wrote:
> Le 12 sept. 2016 13:32, "Pavel Machek" <pavel@ucw.cz> a écrit :
> >
> > On Fri 2016-09-09 10:43:32, Anisse Astier wrote:
> > > PAGE_POISONING_ZERO disables zeroing new pages on alloc, they are
> > > poisoned (zeroed) as they become available.
> > > In the hibernate use case, free pages will appear in the system without
> > > being cleared, left there by the loading kernel.
> > >
> > > This patch will make sure free pages are cleared on resume when
> > > PAGE_POISONING_ZERO is enabled. We free the pages just after resume
> > > because we can't do it later: going through any device resume code might
> > > allocate some memory and invalidate the free pages bitmap.
> > >
> > > Thus we don't need to disable hibernation when PAGE_POISONING_ZERO is
> > > enabled.
> > >
> > > Signed-off-by: Anisse Astier <anisse@astier.eu>
> > > Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
> > > Cc: Laura Abbott <labbott@fedoraproject.org>
> > > Cc: Mel Gorman <mgorman@suse.de>
> > > Cc: Rafael J. Wysocki <rjw@rjwysocki.net>
> >
> > Looks reasonable to me.
> >
> > Acked-by: Pavel Machek <pavel@ucw.cz>
> >
> > Actually.... this takes basically zero time come. Do we want to do it
> > unconditionally?
> >
> > (Yes, it is free memory, but for sake of debugging, I guess zeros are
> > preffered to random content that changed during hibernation.)
> >
> > (But that does not change the Ack.)
> >
> > Best regards,
> >
> Pavel
> > --
> 
> I have no opposition on doing this unconditionally. I can send a v2 as soon
> as I get closer to a computer.

Actually, I'd keep this one as is, when it works and there are no
problems for a release or so, we can delete the ifdefs.

Thanks!
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

[toc] | [prev] | [next] | [standalone]

#1482859

On Friday, September 09, 2016 10:43:32 AM Anisse Astier wrote:
> PAGE_POISONING_ZERO disables zeroing new pages on alloc, they are
> poisoned (zeroed) as they become available.
> In the hibernate use case, free pages will appear in the system without
> being cleared, left there by the loading kernel.
> 
> This patch will make sure free pages are cleared on resume when
> PAGE_POISONING_ZERO is enabled. We free the pages just after resume
> because we can't do it later: going through any device resume code might
> allocate some memory and invalidate the free pages bitmap.
> 
> Thus we don't need to disable hibernation when PAGE_POISONING_ZERO is
> enabled.
> 
> Signed-off-by: Anisse Astier <anisse@astier.eu>
> Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
> Cc: Laura Abbott <labbott@fedoraproject.org>
> Cc: Mel Gorman <mgorman@suse.de>
> Cc: Rafael J. Wysocki <rjw@rjwysocki.net>

Applied (with the tags from Pavel and Kees).

Thanks,
Rafael

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web