Skip to contentSkip to conversation

swsusp: Stop using page flags

PATCH In review Rafael J. Wysocki 4 participants 24 messages Mar 12, 2007 – Mar 20, 2007

Subsystems (5)
lore
Share & tools
lore
#0 lore raw

cover swsusp: Stop using page flags

Hi,

The following three patches make swsusp use its own data structures for memory management instead of special page flags, so that these page flags can be used for other purposes.

Greetings,
Rafael

Signature · 3 lines
If you don't have the time to read,
you don't have the time or the tools to write.
		- Stephen King
#1 lore raw

PATCH 1/3 swsusp: Use inline functions for changing page flags

From: Rafael J. Wysocki <[email protected]>

Replace direct invocations of SetPageNosave(), SetPageNosaveFree() etc. with calls to inline functions that can be changed in subsequent patches without modifying the code calling them.

---
 include/linux/suspend.h |   33 +++++++++++++++++++++++++++++++++
 kernel/power/snapshot.c |   48 +++++++++++++++++++++++++-----------------------
 mm/page_alloc.c         |    6 +++---
 3 files changed, 61 insertions(+), 26 deletions(-)
include/linux/suspend.h +33 -0
Index: linux-2.6.21-rc2/include/linux/suspend.h
===================================================================
@@ -8,6 +8,7 @@
 #include <linux/notifier.h>
 #include <linux/init.h>
 #include <linux/pm.h>
+#include <linux/mm.h>
 
 /* struct pbe is used for creating lists of pages that should be restored
  * atomically during the resume from disk, because the page frames they have
@@ -49,6 +50,38 @@ void __save_processor_state(struct saved
 void __restore_processor_state(struct saved_context *ctxt);
 unsigned long get_safe_page(gfp_t gfp_mask);
 
+/* Page management functions for the software suspend (swsusp) */
+
+static inline void swsusp_set_page_forbidden(struct page *page)
+{
+	SetPageNosave(page);
+}
+
+static inline int swsusp_page_is_forbidden(struct page *page)
+{
+	return PageNosave(page);
+}
+
+static inline void swsusp_unset_page_forbidden(struct page *page)
+{
+	ClearPageNosave(page);
+}
+
+static inline void swsusp_set_page_free(struct page *page)
+{
+	SetPageNosaveFree(page);
+}
+
+static inline int swsusp_page_is_free(struct page *page)
+{
+	return PageNosaveFree(page);
+}
+
+static inline void swsusp_unset_page_free(struct page *page)
+{
+	ClearPageNosaveFree(page);
+}
+
 /*
  * XXX: We try to keep some more pages free so that I/O operations succeed
  * without paging. Might this be more?
kernel/power/snapshot.c +25 -23

136 lines not on this page · raw message

mm/page_alloc.c +3 -3
Index: linux-2.6.21-rc2/mm/page_alloc.c
===================================================================
@@ -770,8 +770,8 @@ void mark_free_pages(struct zone *zone)
 		if (pfn_valid(pfn)) {
 			struct page *page = pfn_to_page(pfn);
 
-			if (!PageNosave(page))
-				ClearPageNosaveFree(page);
+			if (!swsusp_page_is_forbidden(page))
+				swsusp_unset_page_free(page);
 		}
 
 	for (order = MAX_ORDER - 1; order >= 0; --order)
@@ -780,7 +780,7 @@ void mark_free_pages(struct zone *zone)
 
 			pfn = page_to_pfn(list_entry(curr, struct page, lru));
 			for (i = 0; i < (1UL << order); i++)
-				SetPageNosaveFree(pfn_to_page(pfn + i));
+				swsusp_set_page_free(pfn_to_page(pfn + i));
 		}
 
 	spin_unlock_irqrestore(&zone->lock, flags);
#2 lore raw

PATCH 2/3 swsusp: Do not use page flags

In reply to Rafael J. Wysocki

From: Rafael J. Wysocki <[email protected]>

Make swsusp use memory bitmaps instead of page flags for marking 'nosave' and free pages. This allows us to 'recycle' two page flags that can be used for other purposes. Also, the memory needed to store the bitmaps is allocated when necessary (ie. before the suspend) and freed after the resume which is more reasonable.

The patch is designed to minimize the amount of changes and there are some nice simplifications and optimizations possible on top of it. I am going to implement them separately in the future.

---
 arch/x86_64/kernel/e820.c |   26 +---
 include/linux/suspend.h   |   58 +++-------
 kernel/power/disk.c       |   23 +++-
 kernel/power/power.h      |    2 
 kernel/power/snapshot.c   |  250 +++++++++++++++++++++++++++++++++++++++++++---
 kernel/power/user.c       |    4 
 6 files changed, 281 insertions(+), 82 deletions(-)
include/linux/suspend.h +18 -40
Index: linux-2.6.21-rc3/include/linux/suspend.h
===================================================================
@@ -24,63 +24,41 @@ struct pbe {
 extern void drain_local_pages(void);
 extern void mark_free_pages(struct zone *zone);
 
-#ifdef CONFIG_PM
-/* kernel/power/swsusp.c */
-extern int software_suspend(void);
-
-#if defined(CONFIG_VT) && defined(CONFIG_VT_CONSOLE)
+#if defined(CONFIG_PM) && defined(CONFIG_VT) && defined(CONFIG_VT_CONSOLE)
 extern int pm_prepare_console(void);
 extern void pm_restore_console(void);
 #else
 static inline int pm_prepare_console(void) { return 0; }
 static inline void pm_restore_console(void) {}
-#endif /* defined(CONFIG_VT) && defined(CONFIG_VT_CONSOLE) */
+#endif
+
+#if defined(CONFIG_PM) && defined(CONFIG_SOFTWARE_SUSPEND)
+/* kernel/power/swsusp.c */
+extern int software_suspend(void);
+/* kernel/power/snapshot.c */
+extern void __init register_nosave_region(unsigned long, unsigned long);
+extern int swsusp_page_is_forbidden(struct page *);
+extern void swsusp_set_page_free(struct page *);
+extern void swsusp_unset_page_free(struct page *);
+extern unsigned long get_safe_page(gfp_t gfp_mask);
 #else
 static inline int software_suspend(void)
 {
 	printk("Warning: fake suspend called\n");
 	return -ENOSYS;
 }
-#endif /* CONFIG_PM */
+
+static inline void register_nosave_region(unsigned long b, unsigned long e) {}
+static inline int swsusp_page_is_forbidden(struct page *p) { return 0; }
+static inline void swsusp_set_page_free(struct page *p) {}
+static inline void swsusp_unset_page_free(struct page *p) {}
+#endif /* defined(CONFIG_PM) && defined(CONFIG_SOFTWARE_SUSPEND) */
 
 void save_processor_state(void);
 void restore_processor_state(void);
 struct saved_context;
 void __save_processor_state(struct saved_context *ctxt);
 void __restore_processor_state(struct saved_context *ctxt);
-unsigned long get_safe_page(gfp_t gfp_mask);
-
-/* Page management functions for the software suspend (swsusp) */
-
-static inline void swsusp_set_page_forbidden(struct page *page)
-{
-	SetPageNosave(page);
-}
-
-static inline int swsusp_page_is_forbidden(struct page *page)
-{
-	return PageNosave(page);
-}
-
-static inline void swsusp_unset_page_forbidden(struct page *page)
-{
-	ClearPageNosave(page);
-}
-
-static inline void swsusp_set_page_free(struct page *page)
-{
-	SetPageNosaveFree(page);
-}
-
-static inline int swsusp_page_is_free(struct page *page)
-{
-	return PageNosaveFree(page);
-}
-
-static inline void swsusp_unset_page_free(struct page *page)
-{
-	ClearPageNosaveFree(page);
-}
 
 /*
  * XXX: We try to keep some more pages free so that I/O operations succeed
kernel/power/snapshot.c +234 -16

299 lines not on this page · raw message

arch/x86_64/kernel/e820.c +6 -20
Index: linux-2.6.21-rc3/arch/x86_64/kernel/e820.c
===================================================================
@@ -17,6 +17,8 @@
 #include <linux/kexec.h>
 #include <linux/module.h>
 #include <linux/mm.h>
+#include <linux/suspend.h>
+#include <linux/pfn.h>
 
 #include <asm/pgtable.h>
 #include <asm/page.h>
@@ -255,22 +257,6 @@ void __init e820_reserve_resources(void)
 	}
 }
 
-/* Mark pages corresponding to given address range as nosave */
-static void __init
-e820_mark_nosave_range(unsigned long start, unsigned long end)
-{
-	unsigned long pfn, max_pfn;
-
-	if (start >= end)
-		return;
-
-	printk("Nosave address range: %016lx - %016lx\n", start, end);
-	max_pfn = end >> PAGE_SHIFT;
-	for (pfn = start >> PAGE_SHIFT; pfn < max_pfn; pfn++)
-		if (pfn_valid(pfn))
-			SetPageNosave(pfn_to_page(pfn));
-}
-
 /*
  * Find the ranges of physical addresses that do not correspond to
  * e820 RAM areas and mark the corresponding pages as nosave for software
@@ -289,13 +275,13 @@ void __init e820_mark_nosave_regions(voi
 		struct e820entry *ei = &e820.map[i];
 
 		if (paddr < ei->addr)
-			e820_mark_nosave_range(paddr,
-					round_up(ei->addr, PAGE_SIZE));
+			register_nosave_region(PFN_DOWN(paddr),
+						PFN_UP(ei->addr));
 
 		paddr = round_down(ei->addr + ei->size, PAGE_SIZE);
 		if (ei->type != E820_RAM)
-			e820_mark_nosave_range(round_up(ei->addr, PAGE_SIZE),
-					paddr);
+			register_nosave_region(PFN_UP(ei->addr),
+						PFN_DOWN(paddr));
 
 		if (paddr >= (end_pfn << PAGE_SHIFT))
 			break;
kernel/power/power.h +2 -0
Index: linux-2.6.21-rc3/kernel/power/power.h
===================================================================
@@ -49,6 +49,8 @@ extern sector_t swsusp_resume_block;
 extern asmlinkage int swsusp_arch_suspend(void);
 extern asmlinkage int swsusp_arch_resume(void);
 
+extern int create_basic_memory_bitmaps(void);
+extern void free_basic_memory_bitmaps(void);
 extern unsigned int count_data_pages(void);
 
 /**
kernel/power/user.c +4 -0
Index: linux-2.6.21-rc3/kernel/power/user.c
===================================================================
@@ -52,6 +52,9 @@ static int snapshot_open(struct inode *i
 	if ((filp->f_flags & O_ACCMODE) == O_RDWR)
 		return -ENOSYS;
 
+	if(create_basic_memory_bitmaps())
+		return -ENOMEM;
+
 	nonseekable_open(inode, filp);
 	data = &snapshot_state;
 	filp->private_data = data;
@@ -77,6 +80,7 @@ static int snapshot_release(struct inode
 	struct snapshot_data *data;
 
 	swsusp_free();
+	free_basic_memory_bitmaps();
 	data = filp->private_data;
 	free_all_swap_pages(data->swap, data->bitmap);
 	free_bitmap(data->bitmap);
kernel/power/disk.c +17 -6
Index: linux-2.6.21-rc3/kernel/power/disk.c
===================================================================
@@ -127,14 +127,19 @@ int pm_suspend_disk(void)
 		mdelay(5000);
 		goto Thaw;
 	}
+	/* Allocate memory management structures */
+	error = create_basic_memory_bitmaps();
+	if (error)
+		goto Thaw;
+
 	/* Free memory before shutting down devices. */
 	error = swsusp_shrink_memory();
 	if (error)
-		goto Thaw;
+		goto Finish;
 
 	error = platform_prepare();
 	if (error)
-		goto Thaw;
+		goto Finish;
 
 	suspend_console();
 	error = device_suspend(PMSG_FREEZE);
@@ -169,7 +174,7 @@ int pm_suspend_disk(void)
 			power_down(pm_disk_mode);
 		else {
 			swsusp_free();
-			goto Thaw;
+			goto Finish;
 		}
 	} else {
 		pr_debug("PM: Image restored successfully.\n");
@@ -182,6 +187,8 @@ int pm_suspend_disk(void)
 	platform_finish();
 	device_resume();
 	resume_console();
+ Finish:
+	free_basic_memory_bitmaps();
  Thaw:
 	unprepare_processes();
 	return error;
@@ -227,13 +234,15 @@ static int software_resume(void)
 	}
 
 	pr_debug("PM: Checking swsusp image.\n");
-
 	error = swsusp_check();
 	if (error)
-		goto Done;
+		goto Unlock;
 
-	pr_debug("PM: Preparing processes for restore.\n");
+	error = create_basic_memory_bitmaps();
+	if (error)
+		goto Unlock;
 
+	pr_debug("PM: Preparing processes for restore.\n");
 	error = prepare_processes();
 	if (error) {
 		swsusp_close();
@@ -275,7 +284,9 @@ static int software_resume(void)
 	printk(KERN_ERR "PM: Restore failed, recovering.\n");
 	unprepare_processes();
  Done:
+	free_basic_memory_bitmaps();
 	/* For success case, the suspend path will release the lock */
+ Unlock:
 	mutex_unlock(&pm_mutex);
 	pr_debug("PM: Resume from disk failed.\n");
 	return 0;
maintainer
#3 lore raw

Re: [PATCH 2/3] swsusp: Do not use page flags

Quoted 12 lines · On Mon, 12 Mar 2007 22:19:20 +0100 "Rafael J. Wysocki" <[email protected]> wrote:
kernel/power/snapshot.c +11 -0
+int create_basic_memory_bitmaps(void)
+{
+	struct memory_bitmap *bm1, *bm2;
+	int error = 0;
+
+	BUG_ON(forbidden_pages_map || free_pages_map);
+
+	bm1 = kzalloc(sizeof(struct memory_bitmap), GFP_ATOMIC);
+	if (!bm1)
+		return -ENOMEM;
+

kernel/power/snapshot.c static unsigned long memory_bm_next_pfn( · line 739

Quoted 10 lines
kernel/power/snapshot.c +10 -0
+	error = memory_bm_create(bm1, GFP_ATOMIC | __GFP_COLD, PG_ANY);
+	if (error)
+		goto Free_first_object;
+
+	bm2 = kzalloc(sizeof(struct memory_bitmap), GFP_ATOMIC);
+	if (!bm2)
+		goto Free_first_bitmap;
+
+	error = memory_bm_create(bm2, GFP_ATOMIC | __GFP_COLD, PG_ANY);
+	if (error)

What is the risk that we'll go OOM here? GFP_ATOMIC is rather unreliable.

And why _does_ suspend use GFP_ATOMIC all over the place?

#4 lore raw

Re: [PATCH 2/3] swsusp: Do not use page flags

Quoted 16 lines · Andrew Morton
Quoted 16 lines · On Mon, 12 Mar 2007 22:19:20 +0100 "Rafael J. Wysocki" <[email protected]> wrote:
kernel/power/snapshot.c +15 -0
+int create_basic_memory_bitmaps(void)
+{
+	struct memory_bitmap *bm1, *bm2;
+	int error = 0;
+
+	BUG_ON(forbidden_pages_map || free_pages_map);
+
+	bm1 = kzalloc(sizeof(struct memory_bitmap), GFP_ATOMIC);
+	if (!bm1)
+		return -ENOMEM;
+
+	error = memory_bm_create(bm1, GFP_ATOMIC | __GFP_COLD, PG_ANY);
+	if (error)
+		goto Free_first_object;
+
Quoted 8 lines

kernel/power/snapshot.c static unsigned long memory_bm_next_pfn( · line 739

Quoted 6 lines
kernel/power/snapshot.c +6 -0
+	bm2 = kzalloc(sizeof(struct memory_bitmap), GFP_ATOMIC);
+	if (!bm2)
+		goto Free_first_bitmap;
+
+	error = memory_bm_create(bm2, GFP_ATOMIC | __GFP_COLD, PG_ANY);
+	if (error)

What is the risk that we'll go OOM here? GFP_ATOMIC is rather unreliable.

Well, this can be called after processes (including kswapd) has been frozen. We can't go to sleep at this point.

Quoted 1 line

And why _does_ suspend use GFP_ATOMIC all over the place?

Generally, because it cannot sleep.

maintainer
#5 lore raw

Re: [PATCH 2/3] swsusp: Do not use page flags

Quoted 26 lines · On Thu, 15 Mar 2007 22:05:53 +0100 "Rafael J. Wysocki" <[email protected]> wrote:
Quoted 16 lines · Andrew Morton
Quoted 16 lines · On Mon, 12 Mar 2007 22:19:20 +0100 "Rafael J. Wysocki" <[email protected]> wrote:
kernel/power/snapshot.c +15 -0
+int create_basic_memory_bitmaps(void)
+{
+	struct memory_bitmap *bm1, *bm2;
+	int error = 0;
+
+	BUG_ON(forbidden_pages_map || free_pages_map);
+
+	bm1 = kzalloc(sizeof(struct memory_bitmap), GFP_ATOMIC);
+	if (!bm1)
+		return -ENOMEM;
+
+	error = memory_bm_create(bm1, GFP_ATOMIC | __GFP_COLD, PG_ANY);
+	if (error)
+		goto Free_first_object;
+
Quoted 8 lines

kernel/power/snapshot.c static unsigned long memory_bm_next_pfn( · line 739

Quoted 6 lines
kernel/power/snapshot.c +6 -0
+	bm2 = kzalloc(sizeof(struct memory_bitmap), GFP_ATOMIC);
+	if (!bm2)
+		goto Free_first_bitmap;
+
+	error = memory_bm_create(bm2, GFP_ATOMIC | __GFP_COLD, PG_ANY);
+	if (error)

What is the risk that we'll go OOM here? GFP_ATOMIC is rather unreliable.

Quoted 2 lines

Well, this can be called after processes (including kswapd) has been frozen. We can't go to sleep at this point.

So it _is_ unreliable?

Quoted 3 lines
Quoted 1 line

And why _does_ suspend use GFP_ATOMIC all over the place?

Generally, because it cannot sleep.

Why not?

#6 lore raw

Re: [PATCH 2/3] swsusp: Do not use page flags

Quoted 3 lines · Andrew Morton
Quoted 2 lines
Quoted 1 line

And why _does_ suspend use GFP_ATOMIC all over the place?

Generally, because it cannot sleep.

Why not?

I guess it's simply beucase of kswapd being already frozen, so there is no chance that once GFP_KERNEL allocation goes to sleep, it is going to get any free pages eventually ... ?

Signature · 1 line
Jiri Kosina
maintainer
#7 lore raw

Re: [PATCH 2/3] swsusp: Do not use page flags

Quoted 9 lines · On Thu, 15 Mar 2007 23:19:02 +0100 (CET) Jiri Kosina <[email protected]> wrote:
Quoted 3 lines · Andrew Morton
Quoted 2 lines
Quoted 1 line

And why _does_ suspend use GFP_ATOMIC all over the place?

Generally, because it cannot sleep.

Why not?

I guess it's simply beucase of kswapd being already frozen, so there is no chance that once GFP_KERNEL allocation goes to sleep, it is going to get any free pages eventually ... ?

No, things should run fine with a dead kswapd.

There are reasons why we can't call into filesystems from there, but GFP_NOIO will ensure that and it is heaps better than GFP_ATOMIC.

#8 lore raw

Re: [PATCH 2/3] swsusp: Do not use page flags

Quoted 9 lines · Andrew Morton
Quoted 6 lines · On Thu, 15 Mar 2007 23:19:02 +0100 (CET) Jiri Kosina <[email protected]> wrote:
Quoted 3 lines · Andrew Morton
Quoted 2 lines
Quoted 1 line

And why _does_ suspend use GFP_ATOMIC all over the place?

Generally, because it cannot sleep.

Why not?

Quoted 8 lines
Quoted 3 lines

I guess it's simply beucase of kswapd being already frozen, so there is no chance that once GFP_KERNEL allocation goes to sleep, it is going to get any free pages eventually ... ?

No, things should run fine with a dead kswapd.

There are reasons why we can't call into filesystems from there, but GFP_NOIO will ensure that and it is heaps better than GFP_ATOMIC.

In fact the role of swsusp_shrink_memory() is to ensure that our subsequent atomic allocations won't fail.

Still, the particular allocations in create_basic_memory_bitmaps() are made before we call swsusp_shrink_memory(), so it's better to use GFP_NOIO in there.

I'll prepare a patch for that on top of the current series.

#9 lore raw

Re: [PATCH 2/3] swsusp: Do not use page flags

In reply to Andrew Morton

Hi!

Quoted 25 lines
Quoted 25 lines
Quoted 24 lines
Quoted 12 lines · On Mon, 12 Mar 2007 22:19:20 +0100 "Rafael J. Wysocki" <[email protected]> wrote:
kernel/power/snapshot.c +11 -0
+int create_basic_memory_bitmaps(void)
+{
+	struct memory_bitmap *bm1, *bm2;
+	int error = 0;
+
+	BUG_ON(forbidden_pages_map || free_pages_map);
+
+	bm1 = kzalloc(sizeof(struct memory_bitmap), GFP_ATOMIC);
+	if (!bm1)
+		return -ENOMEM;
+

kernel/power/snapshot.c static unsigned long memory_bm_next_pfn( · line 739

Quoted 10 lines
kernel/power/snapshot.c +10 -0
+	error = memory_bm_create(bm1, GFP_ATOMIC | __GFP_COLD, PG_ANY);
+	if (error)
+		goto Free_first_object;
+
+	bm2 = kzalloc(sizeof(struct memory_bitmap), GFP_ATOMIC);
+	if (!bm2)
+		goto Free_first_bitmap;
+
+	error = memory_bm_create(bm2, GFP_ATOMIC | __GFP_COLD, PG_ANY);
+	if (error)

What is the risk that we'll go OOM here? GFP_ATOMIC is rather unreliable.

Quoted 4 lines
Quoted 2 lines

Well, this can be called after processes (including kswapd) has been frozen. We can't go to sleep at this point.

So it _is_ unreliable?

We are careful to leave some memory aside for suspend... We actually free memory at beggining of suspend, and there's some simple "add few percent for our overhead" there.
Pavel

Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
maintainer
#10 lore raw

Re: [PATCH 2/3] swsusp: Do not use page flags

In reply to Rafael J. Wysocki
Quoted 9 lines · On Mon, 12 Mar 2007 22:19:20 +0100 "Rafael J. Wysocki" <[email protected]> wrote:

Make swsusp use memory bitmaps instead of page flags for marking 'nosave' and free pages. This allows us to 'recycle' two page flags that can be used for other purposes. Also, the memory needed to store the bitmaps is allocated when necessary (ie. before the suspend) and freed after the resume which is more reasonable.

The patch is designed to minimize the amount of changes and there are some nice simplifications and optimizations possible on top of it. I am going to implement them separately in the future.

Blows up with ia64 allmodconfig due to CONFIG_PM=y, CONFIG_SOFTWARE_SUSPEND=n:

kernel/power/main.c:223: error: redefinition of 'software_suspend' include/linux/suspend.h:46: error: previous definition of 'software_suspend' was here

I had a look at fixing it, but it's unobvious why we're compiling most of kernel/power/main.c when CONFIG_SOFTWARE_SUSPEND=n so I'll send this series back for repair please.

#11 lore raw

Re: [PATCH 2/3] swsusp: Do not use page flags

Hi!

Quoted 4 lines
Quoted 3 lines

The patch is designed to minimize the amount of changes and there are some nice simplifications and optimizations possible on top of it. I am going to implement them separately in the future.

Quoted 8 lines

Blows up with ia64 allmodconfig due to CONFIG_PM=y, CONFIG_SOFTWARE_SUSPEND=n:

kernel/power/main.c:223: error: redefinition of 'software_suspend' include/linux/suspend.h:46: error: previous definition of 'software_suspend' was here

I had a look at fixing it, but it's unobvious why we're compiling most of kernel/power/main.c when CONFIG_SOFTWARE_SUSPEND=n so I'll send this series back for repair please.

main.c is needed for suspend-to-ram, too.
Pavel

Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
#12 lore raw

Re: [PATCH 2/3] swsusp: Do not use page flags

In reply to Andrew Morton
Quoted 13 lines · Andrew Morton
Quoted 9 lines · On Mon, 12 Mar 2007 22:19:20 +0100 "Rafael J. Wysocki" <[email protected]> wrote:

Make swsusp use memory bitmaps instead of page flags for marking 'nosave' and free pages. This allows us to 'recycle' two page flags that can be used for other purposes. Also, the memory needed to store the bitmaps is allocated when necessary (ie. before the suspend) and freed after the resume which is more reasonable.

The patch is designed to minimize the amount of changes and there are some nice simplifications and optimizations possible on top of it. I am going to implement them separately in the future.

Quoted 8 lines

Blows up with ia64 allmodconfig due to CONFIG_PM=y, CONFIG_SOFTWARE_SUSPEND=n:

kernel/power/main.c:223: error: redefinition of 'software_suspend' include/linux/suspend.h:46: error: previous definition of 'software_suspend' was here

I had a look at fixing it, but it's unobvious why we're compiling most of kernel/power/main.c when CONFIG_SOFTWARE_SUSPEND=n so I'll send this series back for repair please.

Well, it was sufficient to add #ifdef CONFIG_SOFTWARE_SUSPEND around the definition of software_suspend() in kernel/power/main.c.

The following series of patches compiles with CONFIG_SOFTWARE_SUSPEND=n too.

Greetings,
Rafael

#13 lore raw

PATCH 1/5 swsusp: Use inline functions for changing page flags

From: Rafael J. Wysocki <[email protected]>

Replace direct invocations of SetPageNosave(), SetPageNosaveFree() etc. with calls to inline functions that can be changed in subsequent patches without modifying the code calling them.

---
 include/linux/suspend.h |   33 +++++++++++++++++++++++++++++++++
 kernel/power/snapshot.c |   48 +++++++++++++++++++++++++-----------------------
 mm/page_alloc.c         |    6 +++---
 3 files changed, 61 insertions(+), 26 deletions(-)
include/linux/suspend.h +33 -0
Index: linux-2.6.21-rc4-mm1/include/linux/suspend.h
===================================================================
@@ -8,6 +8,7 @@
 #include <linux/notifier.h>
 #include <linux/init.h>
 #include <linux/pm.h>
+#include <linux/mm.h>
 
 /* struct pbe is used for creating lists of pages that should be restored
  * atomically during the resume from disk, because the page frames they have
@@ -49,6 +50,38 @@ void __save_processor_state(struct saved
 void __restore_processor_state(struct saved_context *ctxt);
 unsigned long get_safe_page(gfp_t gfp_mask);
 
+/* Page management functions for the software suspend (swsusp) */
+
+static inline void swsusp_set_page_forbidden(struct page *page)
+{
+	SetPageNosave(page);
+}
+
+static inline int swsusp_page_is_forbidden(struct page *page)
+{
+	return PageNosave(page);
+}
+
+static inline void swsusp_unset_page_forbidden(struct page *page)
+{
+	ClearPageNosave(page);
+}
+
+static inline void swsusp_set_page_free(struct page *page)
+{
+	SetPageNosaveFree(page);
+}
+
+static inline int swsusp_page_is_free(struct page *page)
+{
+	return PageNosaveFree(page);
+}
+
+static inline void swsusp_unset_page_free(struct page *page)
+{
+	ClearPageNosaveFree(page);
+}
+
 /*
  * XXX: We try to keep some more pages free so that I/O operations succeed
  * without paging. Might this be more?
kernel/power/snapshot.c +25 -23

136 lines not on this page · raw message

mm/page_alloc.c +3 -3
Index: linux-2.6.21-rc4-mm1/mm/page_alloc.c
===================================================================
@@ -1050,8 +1050,8 @@ void mark_free_pages(struct zone *zone)
 		if (pfn_valid(pfn)) {
 			struct page *page = pfn_to_page(pfn);
 
-			if (!PageNosave(page))
-				ClearPageNosaveFree(page);
+			if (!swsusp_page_is_forbidden(page))
+				swsusp_unset_page_free(page);
 		}
 
 	for_each_migratetype_order(order, t) {
@@ -1060,7 +1060,7 @@ void mark_free_pages(struct zone *zone)
 
 			pfn = page_to_pfn(list_entry(curr, struct page, lru));
 			for (i = 0; i < (1UL << order); i++)
-				SetPageNosaveFree(pfn_to_page(pfn + i));
+				swsusp_set_page_free(pfn_to_page(pfn + i));
 		}
 	}
 
#14 lore raw

PATCH 2/5 swsusp: do not use page flags

In reply to Rafael J. Wysocki

From: Rafael J. Wysocki <[email protected]>

Make swsusp use memory bitmaps instead of page flags for marking 'nosave' and free pages. This allows us to 'recycle' two page flags that can be used for other purposes. Also, the memory needed to store the bitmaps is allocated when necessary (ie. before the suspend) and freed after the resume which is more reasonable.

The patch is designed to minimize the amount of changes and there are some nice simplifications and optimizations possible on top of it. I am going to implement them separately in the future.

---
 arch/x86_64/kernel/e820.c |   26 +---
 include/linux/suspend.h   |   58 +++-------
 kernel/power/disk.c       |   23 +++-
 kernel/power/main.c       |    2 
 kernel/power/power.h      |    2 
 kernel/power/snapshot.c   |  250 +++++++++++++++++++++++++++++++++++++++++++---
 kernel/power/user.c       |    4 
 7 files changed, 283 insertions(+), 82 deletions(-)
include/linux/suspend.h +18 -40
Index: linux-2.6.21-rc4-mm1/include/linux/suspend.h
===================================================================
@@ -24,63 +24,41 @@ struct pbe {
 extern void drain_local_pages(void);
 extern void mark_free_pages(struct zone *zone);
 
-#ifdef CONFIG_PM
-/* kernel/power/swsusp.c */
-extern int software_suspend(void);
-
-#if defined(CONFIG_VT) && defined(CONFIG_VT_CONSOLE)
+#if defined(CONFIG_PM) && defined(CONFIG_VT) && defined(CONFIG_VT_CONSOLE)
 extern int pm_prepare_console(void);
 extern void pm_restore_console(void);
 #else
 static inline int pm_prepare_console(void) { return 0; }
 static inline void pm_restore_console(void) {}
-#endif /* defined(CONFIG_VT) && defined(CONFIG_VT_CONSOLE) */
+#endif
+
+#if defined(CONFIG_PM) && defined(CONFIG_SOFTWARE_SUSPEND)
+/* kernel/power/swsusp.c */
+extern int software_suspend(void);
+/* kernel/power/snapshot.c */
+extern void __init register_nosave_region(unsigned long, unsigned long);
+extern int swsusp_page_is_forbidden(struct page *);
+extern void swsusp_set_page_free(struct page *);
+extern void swsusp_unset_page_free(struct page *);
+extern unsigned long get_safe_page(gfp_t gfp_mask);
 #else
 static inline int software_suspend(void)
 {
 	printk("Warning: fake suspend called\n");
 	return -ENOSYS;
 }
-#endif /* CONFIG_PM */
+
+static inline void register_nosave_region(unsigned long b, unsigned long e) {}
+static inline int swsusp_page_is_forbidden(struct page *p) { return 0; }
+static inline void swsusp_set_page_free(struct page *p) {}
+static inline void swsusp_unset_page_free(struct page *p) {}
+#endif /* defined(CONFIG_PM) && defined(CONFIG_SOFTWARE_SUSPEND) */
 
 void save_processor_state(void);
 void restore_processor_state(void);
 struct saved_context;
 void __save_processor_state(struct saved_context *ctxt);
 void __restore_processor_state(struct saved_context *ctxt);
-unsigned long get_safe_page(gfp_t gfp_mask);
-
-/* Page management functions for the software suspend (swsusp) */
-
-static inline void swsusp_set_page_forbidden(struct page *page)
-{
-	SetPageNosave(page);
-}
-
-static inline int swsusp_page_is_forbidden(struct page *page)
-{
-	return PageNosave(page);
-}
-
-static inline void swsusp_unset_page_forbidden(struct page *page)
-{
-	ClearPageNosave(page);
-}
-
-static inline void swsusp_set_page_free(struct page *page)
-{
-	SetPageNosaveFree(page);
-}
-
-static inline int swsusp_page_is_free(struct page *page)
-{
-	return PageNosaveFree(page);
-}
-
-static inline void swsusp_unset_page_free(struct page *page)
-{
-	ClearPageNosaveFree(page);
-}
 
 /*
  * XXX: We try to keep some more pages free so that I/O operations succeed
kernel/power/snapshot.c +234 -16

299 lines not on this page · raw message

arch/x86_64/kernel/e820.c +6 -20
Index: linux-2.6.21-rc4-mm1/arch/x86_64/kernel/e820.c
===================================================================
@@ -17,6 +17,8 @@
 #include <linux/kexec.h>
 #include <linux/module.h>
 #include <linux/mm.h>
+#include <linux/suspend.h>
+#include <linux/pfn.h>
 
 #include <asm/pgtable.h>
 #include <asm/page.h>
@@ -255,22 +257,6 @@ void __init e820_reserve_resources(void)
 	}
 }
 
-/* Mark pages corresponding to given address range as nosave */
-static void __init
-e820_mark_nosave_range(unsigned long start, unsigned long end)
-{
-	unsigned long pfn, max_pfn;
-
-	if (start >= end)
-		return;
-
-	printk("Nosave address range: %016lx - %016lx\n", start, end);
-	max_pfn = end >> PAGE_SHIFT;
-	for (pfn = start >> PAGE_SHIFT; pfn < max_pfn; pfn++)
-		if (pfn_valid(pfn))
-			SetPageNosave(pfn_to_page(pfn));
-}
-
 /*
  * Find the ranges of physical addresses that do not correspond to
  * e820 RAM areas and mark the corresponding pages as nosave for software
@@ -289,13 +275,13 @@ void __init e820_mark_nosave_regions(voi
 		struct e820entry *ei = &e820.map[i];
 
 		if (paddr < ei->addr)
-			e820_mark_nosave_range(paddr,
-					round_up(ei->addr, PAGE_SIZE));
+			register_nosave_region(PFN_DOWN(paddr),
+						PFN_UP(ei->addr));
 
 		paddr = round_down(ei->addr + ei->size, PAGE_SIZE);
 		if (ei->type != E820_RAM)
-			e820_mark_nosave_range(round_up(ei->addr, PAGE_SIZE),
-					paddr);
+			register_nosave_region(PFN_UP(ei->addr),
+						PFN_DOWN(paddr));
 
 		if (paddr >= (end_pfn << PAGE_SHIFT))
 			break;
kernel/power/power.h +2 -0
Index: linux-2.6.21-rc4-mm1/kernel/power/power.h
===================================================================
@@ -49,6 +49,8 @@ extern sector_t swsusp_resume_block;
 extern asmlinkage int swsusp_arch_suspend(void);
 extern asmlinkage int swsusp_arch_resume(void);
 
+extern int create_basic_memory_bitmaps(void);
+extern void free_basic_memory_bitmaps(void);
 extern unsigned int count_data_pages(void);
 
 /**
kernel/power/user.c +4 -0
Index: linux-2.6.21-rc4-mm1/kernel/power/user.c
===================================================================
@@ -52,6 +52,9 @@ static int snapshot_open(struct inode *i
 	if ((filp->f_flags & O_ACCMODE) == O_RDWR)
 		return -ENOSYS;
 
+	if(create_basic_memory_bitmaps())
+		return -ENOMEM;
+
 	nonseekable_open(inode, filp);
 	data = &snapshot_state;
 	filp->private_data = data;
@@ -77,6 +80,7 @@ static int snapshot_release(struct inode
 	struct snapshot_data *data;
 
 	swsusp_free();
+	free_basic_memory_bitmaps();
 	data = filp->private_data;
 	free_all_swap_pages(data->swap, data->bitmap);
 	free_bitmap(data->bitmap);
kernel/power/disk.c +17 -6
Index: linux-2.6.21-rc4-mm1/kernel/power/disk.c
===================================================================
@@ -128,14 +128,19 @@ int pm_suspend_disk(void)
 		mdelay(5000);
 		goto Thaw;
 	}
+	/* Allocate memory management structures */
+	error = create_basic_memory_bitmaps();
+	if (error)
+		goto Thaw;
+
 	/* Free memory before shutting down devices. */
 	error = swsusp_shrink_memory();
 	if (error)
-		goto Thaw;
+		goto Finish;
 
 	error = platform_prepare();
 	if (error)
-		goto Thaw;
+		goto Finish;
 
 	suspend_console();
 	error = device_suspend(PMSG_FREEZE);
@@ -170,7 +175,7 @@ int pm_suspend_disk(void)
 			power_down(pm_disk_mode);
 		else {
 			swsusp_free();
-			goto Thaw;
+			goto Finish;
 		}
 	} else {
 		pr_debug("PM: Image restored successfully.\n");
@@ -183,6 +188,8 @@ int pm_suspend_disk(void)
 	platform_finish();
 	device_resume();
 	resume_console();
+ Finish:
+	free_basic_memory_bitmaps();
  Thaw:
 	unprepare_processes();
 	return error;
@@ -228,13 +235,15 @@ static int software_resume(void)
 	}
 
 	pr_debug("PM: Checking swsusp image.\n");
-
 	error = swsusp_check();
 	if (error)
-		goto Done;
+		goto Unlock;
 
-	pr_debug("PM: Preparing processes for restore.\n");
+	error = create_basic_memory_bitmaps();
+	if (error)
+		goto Unlock;
 
+	pr_debug("PM: Preparing processes for restore.\n");
 	error = prepare_processes();
 	if (error) {
 		swsusp_close();
@@ -277,7 +286,9 @@ static int software_resume(void)
 	printk(KERN_ERR "PM: Restore failed, recovering.\n");
 	unprepare_processes();
  Done:
+	free_basic_memory_bitmaps();
 	/* For success case, the suspend path will release the lock */
+ Unlock:
 	mutex_unlock(&pm_mutex);
 	pr_debug("PM: Resume from disk failed.\n");
 	return 0;
kernel/power/main.c +2 -0
Index: linux-2.6.21-rc4-mm1/kernel/power/main.c
===================================================================
@@ -215,6 +215,7 @@ static int enter_state(suspend_state_t s
 	return error;
 }
 
+#ifdef CONFIG_SOFTWARE_SUSPEND
 /*
  * This is main interface to the outside world. It needs to be
  * called from process context.
@@ -223,6 +224,7 @@ int software_suspend(void)
 {
 	return enter_state(PM_SUSPEND_DISK);
 }
+#endif
 
 
 /**
#15 lore raw

PATCH 3/5 mm: remove unused page flags

In reply to Rafael J. Wysocki

From: Rafael J. Wysocki <[email protected]>

Remove the two page flags that were previously used by swsusp and are no longer needed.

---
 include/linux/page-flags.h |   12 ------------
 1 file changed, 12 deletions(-)
include/linux/page-flags.h +0 -12
Index: linux-2.6.21-rc4-mm1/include/linux/page-flags.h
===================================================================
@@ -82,13 +82,11 @@
 #define PG_private		11	/* If pagecache, has fs-private data */
 
 #define PG_writeback		12	/* Page is under writeback */
-#define PG_nosave		13	/* Used for system suspend/resume */
 #define PG_compound		14	/* Part of a compound page */
 #define PG_swapcache		15	/* Swap page: swp_entry_t in private */
 
 #define PG_mappedtodisk		16	/* Has blocks allocated on-disk */
 #define PG_reclaim		17	/* To be reclaimed asap */
-#define PG_nosave_free		18	/* Used for system suspend/resume */
 #define PG_buddy		19	/* Page is free, on buddy lists */
 #define PG_booked		20	/* Has blocks reserved on-disk */
 
@@ -195,16 +193,6 @@ static inline void SetPageUptodate(struc
 #define TestClearPageWriteback(page) test_and_clear_bit(PG_writeback,	\
 							&(page)->flags)
 
-#define PageNosave(page)	test_bit(PG_nosave, &(page)->flags)
-#define SetPageNosave(page)	set_bit(PG_nosave, &(page)->flags)
-#define TestSetPageNosave(page)	test_and_set_bit(PG_nosave, &(page)->flags)
-#define ClearPageNosave(page)		clear_bit(PG_nosave, &(page)->flags)
-#define TestClearPageNosave(page)	test_and_clear_bit(PG_nosave, &(page)->flags)
-
-#define PageNosaveFree(page)	test_bit(PG_nosave_free, &(page)->flags)
-#define SetPageNosaveFree(page)	set_bit(PG_nosave_free, &(page)->flags)
-#define ClearPageNosaveFree(page)		clear_bit(PG_nosave_free, &(page)->flags)
-
 #define PageBuddy(page)		test_bit(PG_buddy, &(page)->flags)
 #define __SetPageBuddy(page)	__set_bit(PG_buddy, &(page)->flags)
 #define __ClearPageBuddy(page)	__clear_bit(PG_buddy, &(page)->flags)
#16 lore raw

PATCH 4/5 swsusp: fix error paths in snapshot_open

In reply to Rafael J. Wysocki

From: Rafael J. Wysocki <[email protected]>

We forget to increase device_available if there's an error in snapshot_open(), so the snapshot device cannot be open at all after snapshot_open() has returned an error.

---
 kernel/power/user.c |   10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)
kernel/power/user.c +6 -4
Index: linux-2.6.21-rc4-mm1/kernel/power/user.c
===================================================================
@@ -49,12 +49,14 @@ static int snapshot_open(struct inode *i
 	if (!atomic_add_unless(&device_available, -1, 0))
 		return -EBUSY;
 
-	if ((filp->f_flags & O_ACCMODE) == O_RDWR)
+	if ((filp->f_flags & O_ACCMODE) == O_RDWR) {
+		atomic_inc(&device_available);
 		return -ENOSYS;
-
-	if(create_basic_memory_bitmaps())
+	}
+	if(create_basic_memory_bitmaps()) {
+		atomic_inc(&device_available);
 		return -ENOMEM;
-
+	}
 	nonseekable_open(inode, filp);
 	data = &snapshot_state;
 	filp->private_data = data;
#17 lore raw

Re: [PATCH 4/5] swsusp: fix error paths in snapshot_open

Hi!

Quoted 3 lines

We forget to increase device_available if there's an error in snapshot_open(), so the snapshot device cannot be open at all after snapshot_open() has returned an error.

Actually, this should go to the beggining of series, as it is (non-critical) bugfix.

Pavel

Quoted 19 lines
kernel/power/user.c +6 -4
@@ -49,12 +49,14 @@ static int snapshot_open(struct inode *i
 	if (!atomic_add_unless(&device_available, -1, 0))
 		return -EBUSY;
 
-	if ((filp->f_flags & O_ACCMODE) == O_RDWR)
+	if ((filp->f_flags & O_ACCMODE) == O_RDWR) {
+		atomic_inc(&device_available);
 		return -ENOSYS;
-
-	if(create_basic_memory_bitmaps())
+	}
+	if(create_basic_memory_bitmaps()) {
+		atomic_inc(&device_available);
 		return -ENOMEM;
-
+	}
 	nonseekable_open(inode, filp);
 	data = &snapshot_state;
 	filp->private_data = data;
Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
#18 lore raw

Re: [PATCH 4/5] swsusp: fix error paths in snapshot_open

Quoted 8 lines · Pavel Machek

Hi!

Quoted 3 lines

We forget to increase device_available if there's an error in snapshot_open(), so the snapshot device cannot be open at all after snapshot_open() has returned an error.

Actually, this should go to the beggining of series, as it is (non-critical) bugfix.

Well, yes.

I've just kept the original order. OTOH, I don't think it's as urgent as to go into 2.6.21 ("been there forever" kind of thing).

Rafael

#19 lore raw

Re: [PATCH 4/5] swsusp: fix error paths in snapshot_open

Hi!

Quoted 11 lines
Quoted 6 lines
Quoted 3 lines

We forget to increase device_available if there's an error in snapshot_open(), so the snapshot device cannot be open at all after snapshot_open() has returned an error.

Actually, this should go to the beggining of series, as it is (non-critical) bugfix.

Well, yes.

I've just kept the original order. OTOH, I don't think it's as urgent as to go into 2.6.21 ("been there forever" kind of thing).

No, it is not urgent enough for 2.6.21... But I have secret plan... trying to push bitmaps+non-bugfixes for swsusp to 2.6.23, and have swsusp/s2ram stabilize during 2.6.22. Way too much stuff happened in 2.6.21 series.
Pavel

Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
#20 lore raw

Re: [PATCH 4/5] swsusp: fix error paths in snapshot_open

Quoted 11 lines · Pavel Machek

Hi!

Quoted 9 lines
Quoted 6 lines
Quoted 3 lines

We forget to increase device_available if there's an error in snapshot_open(), so the snapshot device cannot be open at all after snapshot_open() has returned an error.

Actually, this should go to the beggining of series, as it is (non-critical) bugfix.

Well, yes.

Quoted 7 lines
Quoted 2 lines

I've just kept the original order. OTOH, I don't think it's as urgent as to go into 2.6.21 ("been there forever" kind of thing).

No, it is not urgent enough for 2.6.21... But I have secret plan... trying to push bitmaps+non-bugfixes for swsusp to 2.6.23, and have swsusp/s2ram stabilize during 2.6.22. Way too much stuff happened in 2.6.21 series.

OK by me, but I think we'll have to tell Andrew which patches should wait for 2.6.23 anyway. ;-)

Rafael

#21 lore raw

PATCH 5/5 swsusp: Use GFP_KERNEL for creating basic data structures

In reply to Rafael J. Wysocki

From: Rafael J. Wysocki <[email protected]>

Make swsusp call create_basic_memory_bitmaps() before processes are frozen, so that GFP_KERNEL allocations can be made in it. Additionally, ensure that the swsusp's userland interface won't be used while either pm_suspend_disk() or software_resume() is being executed.

---
 kernel/power/disk.c     |   37 ++++++++++++++++++++++++++-----------
 kernel/power/power.h    |    3 +++
 kernel/power/snapshot.c |    8 ++++----
 kernel/power/user.c     |   10 +++++-----
 4 files changed, 38 insertions(+), 20 deletions(-)
kernel/power/disk.c +26 -11
Index: linux-2.6.21-rc4-mm1/kernel/power/disk.c
===================================================================
@@ -119,28 +119,33 @@ int pm_suspend_disk(void)
 {
 	int error;
 
+	/* The snapshot device should not be opened while we're running */
+	if (!atomic_add_unless(&snapshot_device_available, -1, 0))
+		return -EBUSY;
+
+	/* Allocate memory management structures */
+	error = create_basic_memory_bitmaps();
+	if (error)
+		goto Exit;
+
 	error = prepare_processes();
 	if (error)
-		return error;
+		goto Finish;
 
 	if (pm_disk_mode == PM_DISK_TESTPROC) {
 		printk("swsusp debug: Waiting for 5 seconds.\n");
 		mdelay(5000);
 		goto Thaw;
 	}
-	/* Allocate memory management structures */
-	error = create_basic_memory_bitmaps();
-	if (error)
-		goto Thaw;
 
 	/* Free memory before shutting down devices. */
 	error = swsusp_shrink_memory();
 	if (error)
-		goto Finish;
+		goto Thaw;
 
 	error = platform_prepare();
 	if (error)
-		goto Finish;
+		goto Thaw;
 
 	suspend_console();
 	error = device_suspend(PMSG_FREEZE);
@@ -175,7 +180,7 @@ int pm_suspend_disk(void)
 			power_down(pm_disk_mode);
 		else {
 			swsusp_free();
-			goto Finish;
+			goto Thaw;
 		}
 	} else {
 		pr_debug("PM: Image restored successfully.\n");
@@ -188,10 +193,12 @@ int pm_suspend_disk(void)
 	platform_finish();
 	device_resume();
 	resume_console();
- Finish:
-	free_basic_memory_bitmaps();
  Thaw:
 	unprepare_processes();
+ Finish:
+	free_basic_memory_bitmaps();
+ Exit:
+	atomic_inc(&snapshot_device_available);
 	return error;
 }
 
@@ -239,9 +246,15 @@ static int software_resume(void)
 	if (error)
 		goto Unlock;
 
+	/* The snapshot device should not be opened while we're running */
+	if (!atomic_add_unless(&snapshot_device_available, -1, 0)) {
+		error = -EBUSY;
+		goto Unlock;
+	}
+
 	error = create_basic_memory_bitmaps();
 	if (error)
-		goto Unlock;
+		goto Finish;
 
 	pr_debug("PM: Preparing processes for restore.\n");
 	error = prepare_processes();
@@ -288,6 +301,8 @@ static int software_resume(void)
 	unprepare_processes();
  Done:
 	free_basic_memory_bitmaps();
+ Finish:
+	atomic_inc(&snapshot_device_available);
 	/* For success case, the suspend path will release the lock */
  Unlock:
 	mutex_unlock(&pm_mutex);
kernel/power/power.h +3 -0
Index: linux-2.6.21-rc4-mm1/kernel/power/power.h
===================================================================
@@ -141,6 +141,9 @@ struct resume_swap_area {
 #define PMOPS_ENTER	2
 #define PMOPS_FINISH	3
 
+/* If unset, the snapshot device cannot be open. */
+extern atomic_t snapshot_device_available;
+
 /**
  *	The bitmap is used for tracing allocated swap pages
  *
kernel/power/snapshot.c +4 -4
Index: linux-2.6.21-rc4-mm1/kernel/power/snapshot.c
===================================================================
@@ -723,19 +723,19 @@ int create_basic_memory_bitmaps(void)
 
 	BUG_ON(forbidden_pages_map || free_pages_map);
 
-	bm1 = kzalloc(sizeof(struct memory_bitmap), GFP_ATOMIC);
+	bm1 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL);
 	if (!bm1)
 		return -ENOMEM;
 
-	error = memory_bm_create(bm1, GFP_ATOMIC | __GFP_COLD, PG_ANY);
+	error = memory_bm_create(bm1, GFP_KERNEL, PG_ANY);
 	if (error)
 		goto Free_first_object;
 
-	bm2 = kzalloc(sizeof(struct memory_bitmap), GFP_ATOMIC);
+	bm2 = kzalloc(sizeof(struct memory_bitmap), GFP_KERNEL);
 	if (!bm2)
 		goto Free_first_bitmap;
 
-	error = memory_bm_create(bm2, GFP_ATOMIC | __GFP_COLD, PG_ANY);
+	error = memory_bm_create(bm2, GFP_KERNEL, PG_ANY);
 	if (error)
 		goto Free_second_object;
 
kernel/power/user.c +5 -5
Index: linux-2.6.21-rc4-mm1/kernel/power/user.c
===================================================================
@@ -40,21 +40,21 @@ static struct snapshot_data {
 	char platform_suspend;
 } snapshot_state;
 
-static atomic_t device_available = ATOMIC_INIT(1);
+atomic_t snapshot_device_available = ATOMIC_INIT(1);
 
 static int snapshot_open(struct inode *inode, struct file *filp)
 {
 	struct snapshot_data *data;
 
-	if (!atomic_add_unless(&device_available, -1, 0))
+	if (!atomic_add_unless(&snapshot_device_available, -1, 0))
 		return -EBUSY;
 
 	if ((filp->f_flags & O_ACCMODE) == O_RDWR) {
-		atomic_inc(&device_available);
+		atomic_inc(&snapshot_device_available);
 		return -ENOSYS;
 	}
 	if(create_basic_memory_bitmaps()) {
-		atomic_inc(&device_available);
+		atomic_inc(&snapshot_device_available);
 		return -ENOMEM;
 	}
 	nonseekable_open(inode, filp);
@@ -92,7 +92,7 @@ static int snapshot_release(struct inode
 		enable_nonboot_cpus();
 		mutex_unlock(&pm_mutex);
 	}
-	atomic_inc(&device_available);
+	atomic_inc(&snapshot_device_available);
 	return 0;
 }
 
#22 lore raw

Re: [PATCH 5/5] swsusp: Use GFP_KERNEL for creating basic data structures

Hi!

Quoted 8 lines

From: Rafael J. Wysocki <[email protected]>

Make swsusp call create_basic_memory_bitmaps() before processes are frozen, so that GFP_KERNEL allocations can be made in it. Additionally, ensure that the swsusp's userland interface won't be used while either pm_suspend_disk() or software_resume() is being executed.

ACK, but I'd prefer it to go before the bitmap patches, as it can be viewed as bugfix... ok, it probably does not matter much in this case.

Pavel

Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
#23 lore raw

PATCH 3/3 mm: Remove unused page flags

In reply to Rafael J. Wysocki

From: Rafael J. Wysocki <[email protected]>

Remove the two page flags that were previously used by swsusp and are no longer needed.

---
 include/linux/page-flags.h |   12 ------------
 1 file changed, 12 deletions(-)
include/linux/page-flags.h +0 -12
Index: linux-2.6.21-rc3/include/linux/page-flags.h
===================================================================
@@ -82,13 +82,11 @@
 #define PG_private		11	/* If pagecache, has fs-private data */
 
 #define PG_writeback		12	/* Page is under writeback */
-#define PG_nosave		13	/* Used for system suspend/resume */
 #define PG_compound		14	/* Part of a compound page */
 #define PG_swapcache		15	/* Swap page: swp_entry_t in private */
 
 #define PG_mappedtodisk		16	/* Has blocks allocated on-disk */
 #define PG_reclaim		17	/* To be reclaimed asap */
-#define PG_nosave_free		18	/* Used for system suspend/resume */
 #define PG_buddy		19	/* Page is free, on buddy lists */
 
 /* PG_owner_priv_1 users should have descriptive aliases */
@@ -214,16 +212,6 @@ static inline void SetPageUptodate(struc
 		ret;							\
 	})
 
-#define PageNosave(page)	test_bit(PG_nosave, &(page)->flags)
-#define SetPageNosave(page)	set_bit(PG_nosave, &(page)->flags)
-#define TestSetPageNosave(page)	test_and_set_bit(PG_nosave, &(page)->flags)
-#define ClearPageNosave(page)		clear_bit(PG_nosave, &(page)->flags)
-#define TestClearPageNosave(page)	test_and_clear_bit(PG_nosave, &(page)->flags)
-
-#define PageNosaveFree(page)	test_bit(PG_nosave_free, &(page)->flags)
-#define SetPageNosaveFree(page)	set_bit(PG_nosave_free, &(page)->flags)
-#define ClearPageNosaveFree(page)		clear_bit(PG_nosave_free, &(page)->flags)
-
 #define PageBuddy(page)		test_bit(PG_buddy, &(page)->flags)
 #define __SetPageBuddy(page)	__set_bit(PG_buddy, &(page)->flags)
 #define __ClearPageBuddy(page)	__clear_bit(PG_buddy, &(page)->flags)