File linux-2.6-mm-suspend-improvements.patch of Package kernel

Overview Repositories Revisions Requests Users Attributes Meta

File linux-2.6-mm-suspend-improvements.patch of Package kernel (Revision cf0079dcee8f8d4336adf1ec51657db0)

Currently displaying revision cf0079dcee8f8d4336adf1ec51657db0 , Show latest

swsusp-write-timer.patch
swsusp-write-speedup.patch
swsusp-read-timer.patch
swsusp-read-speedup.patch
swsusp-read-speedup-fix.patch
swsusp-read-speedup-cleanup.patch
swsusp-read-speedup-cleanup-2.patch
swsusp-clean-up-browsing-of-pfns.patch
swsusp-struct-snapshot_handle-cleanup.patch
make-swsusp-avoid-memory-holes-and-reserved-memory-regions-on-x86_64.patch
disable-cpu-hotplug-during-suspend-2.patch
From: Andrew Morton <akpm@osdl.org>

Add some instrumentation to the swsusp writeout code to show what bandwidth
we're achieving.

Cc: Pavel Machek <pavel@ucw.cz>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Signed-off-by: Andrew Morton <akpm@osdl.org>
---

kernel/power/swap.c |   32 +++++++++++++++++++++++++++++---
 1 file changed, 29 insertions(+), 3 deletions(-)

diff -puN kernel/power/swap.c~swsusp-write-timer kernel/power/swap.c
--- a/kernel/power/swap.c~swsusp-write-timer
+++ a/kernel/power/swap.c
@@ -146,6 +146,26 @@ static void release_swap_writer(struct s
 	handle->bitmap = NULL;
 }
 
+static void show_speed(struct timeval *start, struct timeval *stop,
+			unsigned nr_pages, char *msg)
+{
+	s64 elapsed_centisecs64;
+	int centisecs;
+	int k;
+	int kps;
+
+	elapsed_centisecs64 = timeval_to_ns(stop) - timeval_to_ns(start);
+	do_div(elapsed_centisecs64, NSEC_PER_SEC / 100);
+	centisecs = elapsed_centisecs64;
+	if (centisecs == 0)
+		centisecs = 1;	/* avoid div-by-zero */
+	k = nr_pages * (PAGE_SIZE / 1024);
+	kps = (k * 100) / centisecs;
+	printk("%s %d kbytes in %d.%02d seconds (%d.%02d MB/s)\n", msg, k,
+			centisecs / 100, centisecs % 100,
+			kps / 1000, (kps % 1000) / 10);
+}
+
 static int get_swap_writer(struct swap_map_handle *handle)
 {
 	handle->cur = (struct swap_map_page *)get_zeroed_page(GFP_KERNEL);
@@ -206,17 +226,21 @@ static int flush_swap_writer(struct swap
 
 static int save_image(struct swap_map_handle *handle,
                       struct snapshot_handle *snapshot,
-                      unsigned int nr_pages)
+                      unsigned int nr_to_write)
 {
 	unsigned int m;
 	int ret;
 	int error = 0;
+	int nr_pages;
+	struct timeval start;
+	struct timeval stop;
 
-	printk("Saving image data pages (%u pages) ...     ", nr_pages);
-	m = nr_pages / 100;
+	printk("Saving image data pages (%u pages) ...     ", nr_to_write);
+	m = nr_to_write / 100;
 	if (!m)
 		m = 1;
 	nr_pages = 0;
+	do_gettimeofday(&start);
 	do {
 		ret = snapshot_read_next(snapshot, PAGE_SIZE);
 		if (ret > 0) {
@@ -228,8 +252,10 @@ static int save_image(struct swap_map_ha
 			nr_pages++;
 		}
 	} while (ret > 0);
+	do_gettimeofday(&stop);
 	if (!error)
 		printk("\b\b\b\bdone\n");
+	show_speed(&start, &stop, nr_to_write, "Wrote");
 	return error;
 }
 
_
From: Andrew Morton <akpm@osdl.org>

Switch the swsusp writeout code from 4k-at-a-time to 4MB-at-a-time.

Crufty old PIII testbox:
	12.9 MB/s -> 20.9 MB/s

Sony Vaio:
	14.7 MB/s -> 26.5 MB/s

The implementation is crude.  A better one would use larger BIOs, but wouldn't
gain any performance.

The memcpys will be mostly pipelined with the IO and basically come for free.

The ENOMEM path has not been tested.  It should be.

Cc: Pavel Machek <pavel@ucw.cz>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Signed-off-by: Andrew Morton <akpm@osdl.org>
---

include/linux/swap.h |    5 +-
 kernel/power/swap.c  |   93 +++++++++++++++++++++++++++++++++--------
 mm/page_io.c         |   23 +++++++---
 3 files changed, 97 insertions(+), 24 deletions(-)

diff -puN include/linux/swap.h~swsusp-write-speedup include/linux/swap.h
--- a/include/linux/swap.h~swsusp-write-speedup
+++ a/include/linux/swap.h
@@ -12,6 +12,8 @@
 
 struct notifier_block;
 
+struct bio;
+
 #define SWAP_FLAG_PREFER	0x8000	/* set if swap priority specified */
 #define SWAP_FLAG_PRIO_MASK	0x7fff
 #define SWAP_FLAG_PRIO_SHIFT	0
@@ -215,7 +217,8 @@ extern void swap_unplug_io_fn(struct bac
 /* linux/mm/page_io.c */
 extern int swap_readpage(struct file *, struct page *);
 extern int swap_writepage(struct page *page, struct writeback_control *wbc);
-extern int rw_swap_page_sync(int, swp_entry_t, struct page *);
+extern int rw_swap_page_sync(int rw, swp_entry_t entry, struct page *page,
+				struct bio **bio_chain);
 
 /* linux/mm/swap_state.c */
 extern struct address_space swapper_space;
diff -puN kernel/power/swap.c~swsusp-write-speedup kernel/power/swap.c
--- a/kernel/power/swap.c~swsusp-write-speedup
+++ a/kernel/power/swap.c
@@ -49,18 +49,16 @@ static int mark_swapfiles(swp_entry_t st
 {
 	int error;
 
-	rw_swap_page_sync(READ,
-			  swp_entry(root_swap, 0),
-			  virt_to_page((unsigned long)&swsusp_header));
+	rw_swap_page_sync(READ, swp_entry(root_swap, 0),
+			  virt_to_page((unsigned long)&swsusp_header), NULL);
 	if (!memcmp("SWAP-SPACE",swsusp_header.sig, 10) ||
 	    !memcmp("SWAPSPACE2",swsusp_header.sig, 10)) {
 		memcpy(swsusp_header.orig_sig,swsusp_header.sig, 10);
 		memcpy(swsusp_header.sig,SWSUSP_SIG, 10);
 		swsusp_header.image = start;
-		error = rw_swap_page_sync(WRITE,
-					  swp_entry(root_swap, 0),
-					  virt_to_page((unsigned long)
-						       &swsusp_header));
+		error = rw_swap_page_sync(WRITE, swp_entry(root_swap, 0),
+				virt_to_page((unsigned long)&swsusp_header),
+				NULL);
 	} else {
 		pr_debug("swsusp: Partition is not swap space.\n");
 		error = -ENODEV;
@@ -88,16 +86,37 @@ static int swsusp_swap_check(void) /* Th
  *	write_page - Write one page to given swap location.
  *	@buf:		Address we're writing.
  *	@offset:	Offset of the swap page we're writing to.
+ *	@bio_chain:	Link the next write BIO here
  */
 
-static int write_page(void *buf, unsigned long offset)
+static int write_page(void *buf, unsigned long offset, struct bio **bio_chain)
 {
 	swp_entry_t entry;
 	int error = -ENOSPC;
 
 	if (offset) {
+		struct page *page = virt_to_page(buf);
+
+		if (bio_chain) {
+			/*
+			 * Whether or not we successfully allocated a copy page,
+			 * we take a ref on the page here.  It gets undone in
+			 * wait_on_bio_chain().
+			 */
+			struct page *page_copy;
+			page_copy = alloc_page(GFP_ATOMIC);
+			if (page_copy == NULL) {
+				WARN_ON_ONCE(1);
+				bio_chain = NULL;	/* Go synchronous */
+				get_page(page);
+			} else {
+				memcpy(page_address(page_copy),
+					page_address(page), PAGE_SIZE);
+				page = page_copy;
+			}
+		}
 		entry = swp_entry(root_swap, offset);
-		error = rw_swap_page_sync(WRITE, entry, virt_to_page(buf));
+		error = rw_swap_page_sync(WRITE, entry, page, bio_chain);
 	}
 	return error;
 }
@@ -185,37 +204,68 @@ static int get_swap_writer(struct swap_m
 	return 0;
 }
 
-static int swap_write_page(struct swap_map_handle *handle, void *buf)
+static int wait_on_bio_chain(struct bio **bio_chain)
 {
-	int error;
+	struct bio *bio;
+	struct bio *next_bio;
+	int ret = 0;
+
+	if (bio_chain == NULL)
+		return 0;
+
+	bio = *bio_chain;
+	while (bio) {
+		struct page *page;
+
+		next_bio = bio->bi_private;
+		page = bio->bi_io_vec[0].bv_page;
+		wait_on_page_locked(page);
+		if (!PageUptodate(page) || PageError(page))
+			ret = -EIO;
+		put_page(page);
+		bio_put(bio);
+		bio = next_bio;
+	}
+	*bio_chain = NULL;
+	return ret;
+}
+
+static int swap_write_page(struct swap_map_handle *handle, void *buf,
+				struct bio **bio_chain)
+{
+	int error = 0;
 	unsigned long offset;
 
 	if (!handle->cur)
 		return -EINVAL;
 	offset = alloc_swap_page(root_swap, handle->bitmap);
-	error = write_page(buf, offset);
+	error = write_page(buf, offset, bio_chain);
 	if (error)
 		return error;
 	handle->cur->entries[handle->k++] = offset;
 	if (handle->k >= MAP_PAGE_ENTRIES) {
+		error = wait_on_bio_chain(bio_chain);
+		if (error)
+			goto out;
 		offset = alloc_swap_page(root_swap, handle->bitmap);
 		if (!offset)
 			return -ENOSPC;
 		handle->cur->next_swap = offset;
-		error = write_page(handle->cur, handle->cur_swap);
+		error = write_page(handle->cur, handle->cur_swap, NULL);
 		if (error)
-			return error;
+			goto out;
 		memset(handle->cur, 0, PAGE_SIZE);
 		handle->cur_swap = offset;
 		handle->k = 0;
 	}
-	return 0;
+out:
+	return error;
 }
 
 static int flush_swap_writer(struct swap_map_handle *handle)
 {
 	if (handle->cur && handle->cur_swap)
-		return write_page(handle->cur, handle->cur_swap);
+		return write_page(handle->cur, handle->cur_swap, NULL);
 	else
 		return -EINVAL;
 }
@@ -232,6 +282,8 @@ static int save_image(struct swap_map_ha
 	int ret;
 	int error = 0;
 	int nr_pages;
+	int err2;
+	struct bio *bio;
 	struct timeval start;
 	struct timeval stop;
 
@@ -240,11 +292,13 @@ static int save_image(struct swap_map_ha
 	if (!m)
 		m = 1;
 	nr_pages = 0;
+	bio = NULL;
 	do_gettimeofday(&start);
 	do {
 		ret = snapshot_read_next(snapshot, PAGE_SIZE);
 		if (ret > 0) {
-			error = swap_write_page(handle, data_of(*snapshot));
+			error = swap_write_page(handle, data_of(*snapshot),
+						&bio);
 			if (error)
 				break;
 			if (!(nr_pages % m))
@@ -252,8 +306,11 @@ static int save_image(struct swap_map_ha
 			nr_pages++;
 		}
 	} while (ret > 0);
+	err2 = wait_on_bio_chain(&bio);
 	do_gettimeofday(&stop);
 	if (!error)
+		error = err2;
+	if (!error)
 		printk("\b\b\b\bdone\n");
 	show_speed(&start, &stop, nr_to_write, "Wrote");
 	return error;
@@ -307,7 +364,7 @@ int swsusp_write(void)
 	error = get_swap_writer(&handle);
 	if (!error) {
 		unsigned long start = handle.cur_swap;
-		error = swap_write_page(&handle, header);
+		error = swap_write_page(&handle, header, NULL);
 		if (!error)
 			error = save_image(&handle, &snapshot,
 					header->pages - 1);
diff -puN mm/page_io.c~swsusp-write-speedup mm/page_io.c
--- a/mm/page_io.c~swsusp-write-speedup
+++ a/mm/page_io.c
@@ -137,10 +137,12 @@ out:
  * We use end_swap_bio_read() even for writes, because it happens to do what
  * we want.
  */
-int rw_swap_page_sync(int rw, swp_entry_t entry, struct page *page)
+int rw_swap_page_sync(int rw, swp_entry_t entry, struct page *page,
+			struct bio **bio_chain)
 {
 	struct bio *bio;
 	int ret = 0;
+	int bio_rw;
 
 	lock_page(page);
 
@@ -151,11 +153,22 @@ int rw_swap_page_sync(int rw, swp_entry_
 		goto out;
 	}
 
-	submit_bio(rw | (1 << BIO_RW_SYNC), bio);
-	wait_on_page_locked(page);
+	bio_rw = rw;
+	if (!bio_chain)
+		bio_rw |= (1 << BIO_RW_SYNC);
+	if (bio_chain)
+		bio_get(bio);
+	submit_bio(bio_rw, bio);
+	if (bio_chain == NULL) {
+		wait_on_page_locked(page);
 
-	if (!PageUptodate(page) || PageError(page))
-		ret = -EIO;
+		if (!PageUptodate(page) || PageError(page))
+			ret = -EIO;
+	}
+	if (bio_chain) {
+		bio->bi_private = *bio_chain;
+		*bio_chain = bio;
+	}
 out:
 	return ret;
 }
_
From: Andrew Morton <akpm@osdl.org>

Add some instrumentation to the swsusp readin code to show what bandwidth
we're achieving.

Cc: Pavel Machek <pavel@ucw.cz>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Signed-off-by: Andrew Morton <akpm@osdl.org>
---

kernel/power/swap.c |    5 +++++
 1 file changed, 5 insertions(+)

diff -puN kernel/power/swap.c~swsusp-read-timer kernel/power/swap.c
--- a/kernel/power/swap.c~swsusp-read-timer
+++ a/kernel/power/swap.c
@@ -522,12 +522,15 @@ static int load_image(struct swap_map_ha
 	unsigned int m;
 	int ret;
 	int error = 0;
+	struct timeval start;
+	struct timeval stop;
 
 	printk("Loading image data pages (%u pages) ...     ", nr_pages);
 	m = nr_pages / 100;
 	if (!m)
 		m = 1;
 	nr_pages = 0;
+	do_gettimeofday(&start);
 	do {
 		ret = snapshot_write_next(snapshot, PAGE_SIZE);
 		if (ret > 0) {
@@ -539,11 +542,13 @@ static int load_image(struct swap_map_ha
 			nr_pages++;
 		}
 	} while (ret > 0);
+	do_gettimeofday(&stop);
 	if (!error) {
 		printk("\b\b\b\bdone\n");
 		if (!snapshot_image_loaded(snapshot))
 			error = -ENODATA;
 	}
+	show_speed(&start, &stop, nr_pages, "Read");
 	return error;
 }
 
_
From: Andrew Morton <akpm@osdl.org>

Implement async reads for swsusp resuming.

Crufty old PIII testbox:
	15.7 MB/s -> 20.3 MB/s

Sony Vaio:
	14.6 MB/s -> 33.3 MB/s

I didn't implement the post-resume bio_set_pages_dirty().  I don't really
understand why resume needs to run set_page_dirty() against these pages.

It might be a worry that this code modifies PG_Uptodate, PG_Error and
PG_Locked against the image pages.  Can this possibly affect the resumed-into
kernel?  Hopefully not, if we're atomically restoring its mem_map?

Cc: Pavel Machek <pavel@ucw.cz>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Signed-off-by: Andrew Morton <akpm@osdl.org>
---

diff -puN include/linux/swap.h~swsusp-read-speedup include/linux/swap.h
--- a/include/linux/swap.h~swsusp-read-speedup
+++ a/include/linux/swap.h
@@ -219,6 +219,7 @@ extern int swap_readpage(struct file *, 
 extern int swap_writepage(struct page *page, struct writeback_control *wbc);
 extern int rw_swap_page_sync(int rw, swp_entry_t entry, struct page *page,
 				struct bio **bio_chain);
+extern int end_swap_bio_read(struct bio *bio, unsigned int bytes_done, int err);
 
 /* linux/mm/swap_state.c */
 extern struct address_space swapper_space;
diff -puN kernel/power/power.h~swsusp-read-speedup kernel/power/power.h
--- a/kernel/power/power.h~swsusp-read-speedup
+++ a/kernel/power/power.h
@@ -58,6 +58,7 @@ struct snapshot_handle {
 	struct pbe	*pbe, *last_pbe;
 	void		*buffer;
 	unsigned int	buf_offset;
+	int		sync_read;
 };
 
 #define data_of(handle)	((handle).buffer + (handle).buf_offset)
diff -puN kernel/power/snapshot.c~swsusp-read-speedup kernel/power/snapshot.c
--- a/kernel/power/snapshot.c~swsusp-read-speedup
+++ a/kernel/power/snapshot.c
@@ -314,7 +314,7 @@ static unsigned int unsafe_pages;
  *	and we count them using unsafe_pages
  */
 
-static inline void *alloc_image_page(gfp_t gfp_mask, int safe_needed)
+static void *alloc_image_page(gfp_t gfp_mask, int safe_needed)
 {
 	void *res;
 
@@ -828,13 +828,16 @@ int snapshot_write_next(struct snapshot_
 	}
 	if (!handle->offset)
 		handle->buffer = buffer;
+	handle->sync_read = 1;
 	if (handle->prev < handle->page) {
 		if (!handle->prev) {
-			error = load_header(handle, (struct swsusp_info *)buffer);
+			error = load_header(handle,
+					(struct swsusp_info *)buffer);
 			if (error)
 				return error;
 		} else if (handle->prev <= nr_meta_pages) {
-			handle->pbe = unpack_orig_addresses(buffer, handle->pbe);
+			handle->pbe = unpack_orig_addresses(buffer,
+							handle->pbe);
 			if (!handle->pbe) {
 				error = prepare_image(handle);
 				if (error)
@@ -842,10 +845,12 @@ int snapshot_write_next(struct snapshot_
 				handle->pbe = pagedir_nosave;
 				handle->last_pbe = NULL;
 				handle->buffer = get_buffer(handle);
+				handle->sync_read = 0;
 			}
 		} else {
 			handle->pbe = handle->pbe->next;
 			handle->buffer = get_buffer(handle);
+			handle->sync_read = 0;
 		}
 		handle->prev = handle->page;
 	}
diff -puN kernel/power/swap.c~swsusp-read-speedup kernel/power/swap.c
--- a/kernel/power/swap.c~swsusp-read-speedup
+++ a/kernel/power/swap.c
@@ -22,6 +22,7 @@
 #include <linux/device.h>
 #include <linux/buffer_head.h>
 #include <linux/bio.h>
+#include <linux/blkdev.h>
 #include <linux/swap.h>
 #include <linux/swapops.h>
 #include <linux/pm.h>
@@ -214,6 +215,8 @@ static int wait_on_bio_chain(struct bio 
 		return 0;
 
 	bio = *bio_chain;
+	if (bio == NULL)
+		return 0;
 	while (bio) {
 		struct page *page;
 
@@ -349,7 +353,8 @@ int swsusp_write(void)
 	int error;
 
 	if ((error = swsusp_swap_check())) {
-		printk(KERN_ERR "swsusp: Cannot find swap device, try swapon -a.\n");
+		printk(KERN_ERR "swsusp: Cannot find swap device, try "
+				"swapon -a.\n");
 		return error;
 	}
 	memset(&snapshot, 0, sizeof(struct snapshot_handle));
@@ -415,7 +420,8 @@ static struct block_device *resume_bdev;
  *	Then submit it and wait.
  */
 
-static int submit(int rw, pgoff_t page_off, void *page)
+static int submit(int rw, pgoff_t page_off, struct page *page,
+			struct bio **bio_chain)
 {
 	int error = 0;
 	struct bio *bio;
@@ -427,31 +433,41 @@ static int submit(int rw, pgoff_t page_o
 	bio->bi_bdev = resume_bdev;
 	bio->bi_end_io = end_io;
 
-	if (bio_add_page(bio, virt_to_page(page), PAGE_SIZE, 0) < PAGE_SIZE) {
-		printk("swsusp: ERROR: adding page to bio at %ld\n",page_off);
+	if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) {
+		printk("swsusp: ERROR: adding page to bio at %ld\n", page_off);
 		error = -EFAULT;
 		goto Done;
 	}
 
-	atomic_set(&io_done, 1);
-	submit_bio(rw | (1 << BIO_RW_SYNC), bio);
-	while (atomic_read(&io_done))
-		yield();
-	if (rw == READ)
-		bio_set_pages_dirty(bio);
+	if (bio_chain == NULL) {
+		atomic_set(&io_done, 1);
+		submit_bio(rw | (1 << BIO_RW_SYNC), bio);
+		while (atomic_read(&io_done))
+			yield();
+		if (rw == READ)
+			bio_set_pages_dirty(bio);
+		bio_put(bio);
+	} else {
+		lock_page(page);
+		get_page(page);
+		bio_get(bio);
+		bio->bi_end_io = end_swap_bio_read;
+		bio->bi_private = *bio_chain;
+		*bio_chain = bio;
+		submit_bio(rw, bio);
+	}
  Done:
-	bio_put(bio);
 	return error;
 }
 
-static int bio_read_page(pgoff_t page_off, void *page)
+static int bio_read_page(pgoff_t page_off, void *addr, struct bio **bio_chain)
 {
-	return submit(READ, page_off, page);
+	return submit(READ, page_off, virt_to_page(addr), bio_chain);
 }
 
-static int bio_write_page(pgoff_t page_off, void *page)
+static int bio_write_page(pgoff_t page_off, void *addr)
 {
-	return submit(WRITE, page_off, page);
+	return submit(WRITE, page_off, virt_to_page(addr), NULL);
 }
 
 /**
@@ -476,7 +492,7 @@ static int get_swap_reader(struct swap_m
 	handle->cur = (struct swap_map_page *)get_zeroed_page(GFP_ATOMIC);
 	if (!handle->cur)
 		return -ENOMEM;
-	error = bio_read_page(swp_offset(start), handle->cur);
+	error = bio_read_page(swp_offset(start), handle->cur, NULL);
 	if (error) {
 		release_swap_reader(handle);
 		return error;
@@ -485,7 +501,8 @@ static int get_swap_reader(struct swap_m
 	return 0;
 }
 
-static int swap_read_page(struct swap_map_handle *handle, void *buf)
+static int swap_read_page(struct swap_map_handle *handle, void *buf,
+				struct bio **bio_chain)
 {
 	unsigned long offset;
 	int error;
@@ -495,16 +512,17 @@ static int swap_read_page(struct swap_ma
 	offset = handle->cur->entries[handle->k];
 	if (!offset)
 		return -EFAULT;
-	error = bio_read_page(offset, buf);
+	error = bio_read_page(offset, buf, bio_chain);
 	if (error)
 		return error;
 	if (++handle->k >= MAP_PAGE_ENTRIES) {
+		error = wait_on_bio_chain(bio_chain);
 		handle->k = 0;
 		offset = handle->cur->next_swap;
 		if (!offset)
 			release_swap_reader(handle);
-		else
-			error = bio_read_page(offset, handle->cur);
+		else if (!error)
+			error = bio_read_page(offset, handle->cur, NULL);
 	}
 	return error;
 }
@@ -517,38 +535,55 @@ static int swap_read_page(struct swap_ma
 
 static int load_image(struct swap_map_handle *handle,
                       struct snapshot_handle *snapshot,
-                      unsigned int nr_pages)
+                      unsigned int nr_to_read)
 {
 	unsigned int m;
-	int ret;
 	int error = 0;
 	struct timeval start;
 	struct timeval stop;
+	struct bio *bio;
+	int err2;
+	unsigned nr_pages;
+	int sync = 0;
+	int async = 0;
 
-	printk("Loading image data pages (%u pages) ...     ", nr_pages);
-	m = nr_pages / 100;
+	printk("Loading image data pages (%u pages) ...     ", nr_to_read);
+	m = nr_to_read / 100;
 	if (!m)
 		m = 1;
 	nr_pages = 0;
+	bio = NULL;
 	do_gettimeofday(&start);
-	do {
-		ret = snapshot_write_next(snapshot, PAGE_SIZE);
-		if (ret > 0) {
-			error = swap_read_page(handle, data_of(*snapshot));
-			if (error)
-				break;
-			if (!(nr_pages % m))
-				printk("\b\b\b\b%3d%%", nr_pages / m);
-			nr_pages++;
+	for ( ; ; ) {
+		error = snapshot_write_next(snapshot, PAGE_SIZE);
+		if (error <= 0)
+			break;
+		error = swap_read_page(handle, data_of(*snapshot), &bio);
+		if (error)
+			break;
+		if (snapshot->sync_read) {
+			error = wait_on_bio_chain(&bio);
+			sync++;
+		} else {
+			async++;
 		}
-	} while (ret > 0);
+		if (error)
+			break;
+		if (!(nr_pages % m))
+			printk("\b\b\b\b%3d%%", nr_pages / m);
+		nr_pages++;
+	}
+	err2 = wait_on_bio_chain(&bio);
 	do_gettimeofday(&stop);
+	if (!error)
+		error = err2;
 	if (!error) {
 		printk("\b\b\b\bdone\n");
 		if (!snapshot_image_loaded(snapshot))
 			error = -ENODATA;
 	}
-	show_speed(&start, &stop, nr_pages, "Read");
+	show_speed(&start, &stop, nr_to_read, "Read");
+	printk("sync=%d async=%d\n", sync, async);
 	return error;
 }
 
@@ -571,7 +606,7 @@ int swsusp_read(void)
 	header = (struct swsusp_info *)data_of(snapshot);
 	error = get_swap_reader(&handle, swsusp_header.image);
 	if (!error)
-		error = swap_read_page(&handle, header);
+		error = swap_read_page(&handle, header, NULL);
 	if (!error)
 		error = load_image(&handle, &snapshot, header->pages - 1);
 	release_swap_reader(&handle);
@@ -597,7 +632,7 @@ int swsusp_check(void)
 	if (!IS_ERR(resume_bdev)) {
 		set_blocksize(resume_bdev, PAGE_SIZE);
 		memset(&swsusp_header, 0, sizeof(swsusp_header));
-		if ((error = bio_read_page(0, &swsusp_header)))
+		if ((error = bio_read_page(0, &swsusp_header, NULL)))
 			return error;
 		if (!memcmp(SWSUSP_SIG, swsusp_header.sig, 10)) {
 			memcpy(swsusp_header.sig, swsusp_header.orig_sig, 10);
diff -puN mm/page_io.c~swsusp-read-speedup mm/page_io.c
--- a/mm/page_io.c~swsusp-read-speedup
+++ a/mm/page_io.c
@@ -59,7 +59,7 @@ static int end_swap_bio_write(struct bio
 	return 0;
 }
 
-static int end_swap_bio_read(struct bio *bio, unsigned int bytes_done, int err)
+int end_swap_bio_read(struct bio *bio, unsigned int bytes_done, int err)
 {
 	const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
 	struct page *page = bio->bi_io_vec[0].bv_page;
_
From: Andrew Morton <akpm@osdl.org>

Fix an error-path BIO leak.

Cc: Pavel Machek <pavel@ucw.cz>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Signed-off-by: Andrew Morton <akpm@osdl.org>
---

kernel/power/swap.c |    8 +++-----
 1 file changed, 3 insertions(+), 5 deletions(-)

diff -puN kernel/power/swap.c~swsusp-read-speedup-fix kernel/power/swap.c
--- a/kernel/power/swap.c~swsusp-read-speedup-fix
+++ a/kernel/power/swap.c
@@ -423,7 +423,6 @@ static struct block_device *resume_bdev;
 static int submit(int rw, pgoff_t page_off, struct page *page,
 			struct bio **bio_chain)
 {
-	int error = 0;
 	struct bio *bio;
 
 	bio = bio_alloc(GFP_ATOMIC, 1);
@@ -435,8 +434,8 @@ static int submit(int rw, pgoff_t page_o
 
 	if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) {
 		printk("swsusp: ERROR: adding page to bio at %ld\n", page_off);
-		error = -EFAULT;
-		goto Done;
+		bio_put(bio);
+		return -EFAULT;
 	}
 
 	if (bio_chain == NULL) {
@@ -456,8 +455,7 @@ static int submit(int rw, pgoff_t page_o
 		*bio_chain = bio;
 		submit_bio(rw, bio);
 	}
- Done:
-	return error;
+	return 0;
 }
 
 static int bio_read_page(pgoff_t page_off, void *addr, struct bio **bio_chain)
_
From: Andrew Morton <akpm@osdl.org>

Kill off the polled-mode read completion code, clean a few little things up.

Cc: Pavel Machek <pavel@ucw.cz>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Signed-off-by: Andrew Morton <akpm@osdl.org>
---

kernel/power/swap.c |   39 ++++++++-------------------------------
 1 file changed, 8 insertions(+), 31 deletions(-)

diff -puN kernel/power/swap.c~swsusp-read-speedup-cleanup kernel/power/swap.c
--- a/kernel/power/swap.c~swsusp-read-speedup-cleanup
+++ a/kernel/power/swap.c
@@ -386,27 +386,6 @@ int swsusp_write(void)
 	return error;
 }
 
-/*
- *	Using bio to read from swap.
- *	This code requires a bit more work than just using buffer heads
- *	but, it is the recommended way for 2.5/2.6.
- *	The following are to signal the beginning and end of I/O. Bios
- *	finish asynchronously, while we want them to happen synchronously.
- *	A simple atomic_t, and a wait loop take care of this problem.
- */
-
-static atomic_t io_done = ATOMIC_INIT(0);
-
-static int end_io(struct bio *bio, unsigned int num, int err)
-{
-	if (!test_bit(BIO_UPTODATE, &bio->bi_flags)) {
-		printk(KERN_ERR "I/O error reading swsusp image.\n");
-		return -EIO;
-	}
-	atomic_set(&io_done, 0);
-	return 0;
-}
-
 static struct block_device *resume_bdev;
 
 /**
@@ -414,12 +393,12 @@ static struct block_device *resume_bdev;
  *	@rw:	READ or WRITE.
  *	@off	physical offset of page.
  *	@page:	page we're reading or writing.
+ *	@bio_chain: list of pending biod (for async reading)
  *
  *	Straight from the textbook - allocate and initialize the bio.
- *	If we're writing, make sure the page is marked as dirty.
- *	Then submit it and wait.
+ *	If we're reading, make sure the page is marked as dirty.
+ *	Then submit it and, if @bio_chain == NULL, wait.
  */
-
 static int submit(int rw, pgoff_t page_off, struct page *page,
 			struct bio **bio_chain)
 {
@@ -430,7 +409,7 @@ static int submit(int rw, pgoff_t page_o
 		return -ENOMEM;
 	bio->bi_sector = page_off * (PAGE_SIZE >> 9);
 	bio->bi_bdev = resume_bdev;
-	bio->bi_end_io = end_io;
+	bio->bi_end_io = end_swap_bio_read;
 
 	if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) {
 		printk("swsusp: ERROR: adding page to bio at %ld\n", page_off);
@@ -438,19 +417,17 @@ static int submit(int rw, pgoff_t page_o
 		return -EFAULT;
 	}
 
+	lock_page(page);
+	bio_get(bio);
+
 	if (bio_chain == NULL) {
-		atomic_set(&io_done, 1);
 		submit_bio(rw | (1 << BIO_RW_SYNC), bio);
-		while (atomic_read(&io_done))
-			yield();
+		wait_on_page_locked(page);
 		if (rw == READ)
 			bio_set_pages_dirty(bio);
 		bio_put(bio);
 	} else {
-		lock_page(page);
 		get_page(page);
-		bio_get(bio);
-		bio->bi_end_io = end_swap_bio_read;
 		bio->bi_private = *bio_chain;
 		*bio_chain = bio;
 		submit_bio(rw, bio);
_
From: Andrew Morton <akpm@osdl.org>

Remove left-over diagnostic stuff.

Cc: Pavel Machek <pavel@ucw.cz>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Signed-off-by: Andrew Morton <akpm@osdl.org>
---

kernel/power/swap.c |    9 +--------
 1 file changed, 1 insertion(+), 8 deletions(-)

diff -puN kernel/power/swap.c~swsusp-read-speedup-cleanup-2 kernel/power/swap.c
--- a/kernel/power/swap.c~swsusp-read-speedup-cleanup-2
+++ a/kernel/power/swap.c
@@ -519,8 +519,6 @@ static int load_image(struct swap_map_ha
 	struct bio *bio;
 	int err2;
 	unsigned nr_pages;
-	int sync = 0;
-	int async = 0;
 
 	printk("Loading image data pages (%u pages) ...     ", nr_to_read);
 	m = nr_to_read / 100;
@@ -536,12 +534,8 @@ static int load_image(struct swap_map_ha
 		error = swap_read_page(handle, data_of(*snapshot), &bio);
 		if (error)
 			break;
-		if (snapshot->sync_read) {
+		if (snapshot->sync_read)
 			error = wait_on_bio_chain(&bio);
-			sync++;
-		} else {
-			async++;
-		}
 		if (error)
 			break;
 		if (!(nr_pages % m))
@@ -558,7 +552,6 @@ static int load_image(struct swap_map_ha
 			error = -ENODATA;
 	}
 	show_speed(&start, &stop, nr_to_read, "Read");
-	printk("sync=%d async=%d\n", sync, async);
 	return error;
 }
 
_
From: "Rafael J. Wysocki" <rjw@sisk.pl>

Clean up some loops over pfns for each zone in snapshot.c: reduce the
number of additions to perform, rework detection of saveable pages and make
the code a bit less difficult to understand, hopefully.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Acked-by: Pavel Machek <pavel@ucw.cz>
Signed-off-by: Andrew Morton <akpm@osdl.org>
---

kernel/power/snapshot.c |   62 +++++++++++++++++++-------------------
 1 file changed, 32 insertions(+), 30 deletions(-)

diff -puN kernel/power/snapshot.c~swsusp-clean-up-browsing-of-pfns kernel/power/snapshot.c
--- a/kernel/power/snapshot.c~swsusp-clean-up-browsing-of-pfns
+++ a/kernel/power/snapshot.c
@@ -156,7 +156,7 @@ static inline int save_highmem(void) {re
 static inline int restore_highmem(void) {return 0;}
 #endif
 
-static int pfn_is_nosave(unsigned long pfn)
+static inline int pfn_is_nosave(unsigned long pfn)
 {
 	unsigned long nosave_begin_pfn = __pa(&__nosave_begin) >> PAGE_SHIFT;
 	unsigned long nosave_end_pfn = PAGE_ALIGN(__pa(&__nosave_end)) >> PAGE_SHIFT;
@@ -167,43 +167,43 @@ static int pfn_is_nosave(unsigned long p
  *	saveable - Determine whether a page should be cloned or not.
  *	@pfn:	The page
  *
- *	We save a page if it's Reserved, and not in the range of pages
- *	statically defined as 'unsaveable', or if it isn't reserved, and
- *	isn't part of a free chunk of pages.
+ *	We save a page if it isn't Nosave, and is not in the range of pages
+ *	statically defined as 'unsaveable', and it
+ *	isn't a part of a free chunk of pages.
  */
 
-static int saveable(struct zone *zone, unsigned long *zone_pfn)
+static struct page *saveable_page(unsigned long pfn)
 {
-	unsigned long pfn = *zone_pfn + zone->zone_start_pfn;
 	struct page *page;
 
 	if (!pfn_valid(pfn))
-		return 0;
+		return NULL;
 
 	page = pfn_to_page(pfn);
-	BUG_ON(PageReserved(page) && PageNosave(page));
+
 	if (PageNosave(page))
-		return 0;
+		return NULL;
 	if (PageReserved(page) && pfn_is_nosave(pfn))
-		return 0;
+		return NULL;
 	if (PageNosaveFree(page))
-		return 0;
+		return NULL;
 
-	return 1;
+	return page;
 }
 
 unsigned int count_data_pages(void)
 {
 	struct zone *zone;
-	unsigned long zone_pfn;
+	unsigned long pfn, max_zone_pfn;
 	unsigned int n = 0;
 
 	for_each_zone (zone) {
 		if (is_highmem(zone))
 			continue;
 		mark_free_pages(zone);
-		for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn)
-			n += saveable(zone, &zone_pfn);
+		max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages;
+		for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++)
+			n += !!saveable_page(pfn);
 	}
 	return n;
 }
@@ -211,7 +211,7 @@ unsigned int count_data_pages(void)
 static void copy_data_pages(struct pbe *pblist)
 {
 	struct zone *zone;
-	unsigned long zone_pfn;
+	unsigned long pfn, max_zone_pfn;
 	struct pbe *pbe, *p;
 
 	pbe = pblist;
@@ -224,13 +224,14 @@ static void copy_data_pages(struct pbe *
 			SetPageNosaveFree(virt_to_page(p));
 		for_each_pbe (p, pblist)
 			SetPageNosaveFree(virt_to_page(p->address));
-		for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn) {
-			if (saveable(zone, &zone_pfn)) {
-				struct page *page;
+		max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages;
+		for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) {
+			struct page *page = saveable_page(pfn);
+
+			if (page) {
 				long *src, *dst;
 				int n;
 
-				page = pfn_to_page(zone_pfn + zone->zone_start_pfn);
 				BUG_ON(!pbe);
 				pbe->orig_address = (unsigned long)page_address(page);
 				/* copy_page and memcpy are not usable for copying task structs. */
@@ -383,13 +384,14 @@ static struct pbe *alloc_pagedir(unsigne
 void swsusp_free(void)
 {
 	struct zone *zone;
-	unsigned long zone_pfn;
+	unsigned long pfn, max_zone_pfn;
 
 	for_each_zone(zone) {
-		for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn)
-			if (pfn_valid(zone_pfn + zone->zone_start_pfn)) {
-				struct page *page;
-				page = pfn_to_page(zone_pfn + zone->zone_start_pfn);
+		max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages;
+		for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++)
+			if (pfn_valid(pfn)) {
+				struct page *page = pfn_to_page(pfn);
+
 				if (PageNosave(page) && PageNosaveFree(page)) {
 					ClearPageNosave(page);
 					ClearPageNosaveFree(page);
@@ -598,7 +600,7 @@ int snapshot_read_next(struct snapshot_h
 static int mark_unsafe_pages(struct pbe *pblist)
 {
 	struct zone *zone;
-	unsigned long zone_pfn;
+	unsigned long pfn, max_zone_pfn;
 	struct pbe *p;
 
 	if (!pblist) /* a sanity check */
@@ -606,10 +608,10 @@ static int mark_unsafe_pages(struct pbe 
 
 	/* Clear page flags */
 	for_each_zone (zone) {
-		for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn)
-			if (pfn_valid(zone_pfn + zone->zone_start_pfn))
-				ClearPageNosaveFree(pfn_to_page(zone_pfn +
-					zone->zone_start_pfn));
+		max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages;
+		for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++)
+			if (pfn_valid(pfn))
+				ClearPageNosaveFree(pfn_to_page(pfn));
 	}
 
 	/* Mark orig addresses */
_
From: "Rafael J. Wysocki" <rjw@sisk.pl>

Add comments describing struct snapshot_handle and its members, change the
confusing name of its member 'page' to 'cur'.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Cc: Pavel Machek <pavel@ucw.cz>
Signed-off-by: Andrew Morton <akpm@osdl.org>
---

kernel/power/power.h    |   64 +++++++++++++++++++++++++++++++++-----
 kernel/power/snapshot.c |   40 +++++++++++------------
 2 files changed, 76 insertions(+), 28 deletions(-)

diff -puN kernel/power/power.h~swsusp-struct-snapshot_handle-cleanup kernel/power/power.h
--- a/kernel/power/power.h~swsusp-struct-snapshot_handle-cleanup
+++ a/kernel/power/power.h
@@ -50,17 +50,65 @@ extern asmlinkage int swsusp_arch_resume
 
 extern unsigned int count_data_pages(void);
 
+/**
+ *	Auxiliary structure used for reading the snapshot image data and
+ *	metadata from and writing them to the list of page backup entries
+ *	(PBEs) which is the main data structure of swsusp.
+ *
+ *	Using struct snapshot_handle we can transfer the image, including its
+ *	metadata, as a continuous sequence of bytes with the help of
+ *	snapshot_read_next() and snapshot_write_next().
+ *
+ *	The code that writes the image to a storage or transfers it to
+ *	the user land is required to use snapshot_read_next() for this
+ *	purpose and it should not make any assumptions regarding the internal
+ *	structure of the image.  Similarly, the code that reads the image from
+ *	a storage or transfers it from the user land is required to use
+ *	snapshot_write_next().
+ *
+ *	This may allow us to change the internal structure of the image
+ *	in the future with considerably less effort.
+ */
+
 struct snapshot_handle {
-	loff_t		offset;
-	unsigned int	page;
-	unsigned int	page_offset;
-	unsigned int	prev;
-	struct pbe	*pbe, *last_pbe;
-	void		*buffer;
-	unsigned int	buf_offset;
-	int		sync_read;
+	loff_t		offset;	/* number of the last byte ready for reading
+				 * or writing in the sequence
+				 */
+	unsigned int	cur;	/* number of the block of PAGE_SIZE bytes the
+				 * next operation will refer to (ie. current)
+				 */
+	unsigned int	cur_offset;	/* offset with respect to the current
+					 * block (for the next operation)
+					 */
+	unsigned int	prev;	/* number of the block of PAGE_SIZE bytes that
+				 * was the current one previously
+				 */
+	struct pbe	*pbe;	/* PBE that corresponds to 'buffer' */
+	struct pbe	*last_pbe;	/* When the image is restored (eg. read
+					 * from disk) we can store some image
+					 * data directly in the page frames
+					 * in which they were before suspend.
+					 * In such a case the PBEs that
+					 * correspond to them will be unused.
+					 * This is the last PBE, so far, that
+					 * does not correspond to such data.
+					 */
+	void		*buffer;	/* address of the block to read from
+					 * or write to
+					 */
+	unsigned int	buf_offset;	/* location to read from or write to,
+					 * given as a displacement from 'buffer'
+					 */
+	int		sync_read;	/* Set to one to notify the caller of
+					 * snapshot_write_next() that it may
+					 * need to call wait_on_bio_chain()
+					 */
 };
 
+/* This macro returns the address from/to which the caller of
+ * snapshot_read_next()/snapshot_write_next() is allowed to
+ * read/write data after the function returns
+ */
 #define data_of(handle)	((handle).buffer + (handle).buf_offset)
 
 extern int snapshot_read_next(struct snapshot_handle *handle, size_t count);
diff -puN kernel/power/snapshot.c~swsusp-struct-snapshot_handle-cleanup kernel/power/snapshot.c
--- a/kernel/power/snapshot.c~swsusp-struct-snapshot_handle-cleanup
+++ a/kernel/power/snapshot.c
@@ -555,7 +555,7 @@ static inline struct pbe *pack_orig_addr
 
 int snapshot_read_next(struct snapshot_handle *handle, size_t count)
 {
-	if (handle->page > nr_meta_pages + nr_copy_pages)
+	if (handle->cur > nr_meta_pages + nr_copy_pages)
 		return 0;
 	if (!buffer) {
 		/* This makes the buffer be freed by swsusp_free() */
@@ -568,8 +568,8 @@ int snapshot_read_next(struct snapshot_h
 		handle->buffer = buffer;
 		handle->pbe = pagedir_nosave;
 	}
-	if (handle->prev < handle->page) {
-		if (handle->page <= nr_meta_pages) {
+	if (handle->prev < handle->cur) {
+		if (handle->cur <= nr_meta_pages) {
 			handle->pbe = pack_orig_addresses(buffer, handle->pbe);
 			if (!handle->pbe)
 				handle->pbe = pagedir_nosave;
@@ -577,15 +577,15 @@ int snapshot_read_next(struct snapshot_h
 			handle->buffer = (void *)handle->pbe->address;
 			handle->pbe = handle->pbe->next;
 		}
-		handle->prev = handle->page;
+		handle->prev = handle->cur;
 	}
-	handle->buf_offset = handle->page_offset;
-	if (handle->page_offset + count >= PAGE_SIZE) {
-		count = PAGE_SIZE - handle->page_offset;
-		handle->page_offset = 0;
-		handle->page++;
+	handle->buf_offset = handle->cur_offset;
+	if (handle->cur_offset + count >= PAGE_SIZE) {
+		count = PAGE_SIZE - handle->cur_offset;
+		handle->cur_offset = 0;
+		handle->cur++;
 	} else {
-		handle->page_offset += count;
+		handle->cur_offset += count;
 	}
 	handle->offset += count;
 	return count;
@@ -820,7 +820,7 @@ int snapshot_write_next(struct snapshot_
 {
 	int error = 0;
 
-	if (handle->prev && handle->page > nr_meta_pages + nr_copy_pages)
+	if (handle->prev && handle->cur > nr_meta_pages + nr_copy_pages)
 		return 0;
 	if (!buffer) {
 		/* This makes the buffer be freed by swsusp_free() */
@@ -831,7 +831,7 @@ int snapshot_write_next(struct snapshot_
 	if (!handle->offset)
 		handle->buffer = buffer;
 	handle->sync_read = 1;
-	if (handle->prev < handle->page) {
+	if (handle->prev < handle->cur) {
 		if (!handle->prev) {
 			error = load_header(handle,
 					(struct swsusp_info *)buffer);
@@ -854,15 +854,15 @@ int snapshot_write_next(struct snapshot_
 			handle->buffer = get_buffer(handle);
 			handle->sync_read = 0;
 		}
-		handle->prev = handle->page;
+		handle->prev = handle->cur;
 	}
-	handle->buf_offset = handle->page_offset;
-	if (handle->page_offset + count >= PAGE_SIZE) {
-		count = PAGE_SIZE - handle->page_offset;
-		handle->page_offset = 0;
-		handle->page++;
+	handle->buf_offset = handle->cur_offset;
+	if (handle->cur_offset + count >= PAGE_SIZE) {
+		count = PAGE_SIZE - handle->cur_offset;
+		handle->cur_offset = 0;
+		handle->cur++;
 	} else {
-		handle->page_offset += count;
+		handle->cur_offset += count;
 	}
 	handle->offset += count;
 	return count;
@@ -871,5 +871,5 @@ int snapshot_write_next(struct snapshot_
 int snapshot_image_loaded(struct snapshot_handle *handle)
 {
 	return !(!handle->pbe || handle->pbe->next || !nr_copy_pages ||
-		handle->page <= nr_meta_pages + nr_copy_pages);
+		handle->cur <= nr_meta_pages + nr_copy_pages);
 }
_
From: "Rafael J. Wysocki" <rjw@sisk.pl>

On x86_64 machines with more than 2 GB of RAM there are large memory gaps
(with no corresponding kernel virtual addresses) and reserved memory
regions between areas of usable physical RAM.  Moreover, if CONFIG_FLATMEM
is set, they appear within the normal zone.  swsusp should not try to save
them, so the corresponding page structs have to be marked as 'nosave'.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Cc: Mel Gorman <mel@csn.ul.ie>
Acked-by: Pavel Machek <pavel@ucw.cz>
Signed-off-by: Andrew Morton <akpm@osdl.org>
---

arch/x86_64/kernel/e820.c  |   48 +++++++++++++++++++++++++++++++++++
 arch/x86_64/kernel/setup.c |    1 
 include/asm-x86_64/e820.h  |    1 
 3 files changed, 50 insertions(+)

diff -puN arch/x86_64/kernel/e820.c~make-swsusp-avoid-memory-holes-and-reserved-memory-regions-on-x86_64 arch/x86_64/kernel/e820.c
--- a/arch/x86_64/kernel/e820.c~make-swsusp-avoid-memory-holes-and-reserved-memory-regions-on-x86_64
+++ a/arch/x86_64/kernel/e820.c
@@ -16,6 +16,7 @@
 #include <linux/string.h>
 #include <linux/kexec.h>
 #include <linux/module.h>
+#include <linux/mm.h>
 
 #include <asm/pgtable.h>
 #include <asm/page.h>
@@ -293,6 +294,53 @@ 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
+ * suspend and suspend to RAM.
+ *
+ * This function requires the e820 map to be sorted and without any
+ * overlapping entries and assumes the first e820 area to be RAM.
+ */
+void __init e820_mark_nosave_regions(void)
+{
+	int i;
+	unsigned long paddr;
+
+	paddr = round_down(e820.map[0].addr + e820.map[0].size, PAGE_SIZE);
+	for (i = 1; i < e820.nr_map; i++) {
+		struct e820entry *ei = &e820.map[i];
+
+		if (paddr < ei->addr)
+			e820_mark_nosave_range(paddr,
+					round_up(ei->addr, PAGE_SIZE));
+
+		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);
+
+		if (paddr >= (end_pfn << PAGE_SHIFT))
+			break;
+	}
+}
+
 /* 
  * Add a memory region to the kernel e820 map.
  */ 
diff -puN arch/x86_64/kernel/setup.c~make-swsusp-avoid-memory-holes-and-reserved-memory-regions-on-x86_64 arch/x86_64/kernel/setup.c
--- a/arch/x86_64/kernel/setup.c~make-swsusp-avoid-memory-holes-and-reserved-memory-regions-on-x86_64
+++ a/arch/x86_64/kernel/setup.c
@@ -678,6 +678,7 @@ void __init setup_arch(char **cmdline_p)
 	 */
 	probe_roms();
 	e820_reserve_resources(); 
+	e820_mark_nosave_regions();
 
 	request_resource(&iomem_resource, &video_ram_resource);
 
diff -puN include/asm-x86_64/e820.h~make-swsusp-avoid-memory-holes-and-reserved-memory-regions-on-x86_64 include/asm-x86_64/e820.h
--- a/include/asm-x86_64/e820.h~make-swsusp-avoid-memory-holes-and-reserved-memory-regions-on-x86_64
+++ a/include/asm-x86_64/e820.h
@@ -46,6 +46,7 @@ extern void setup_memory_region(void);
 extern void contig_e820_setup(void); 
 extern unsigned long e820_end_of_ram(void);
 extern void e820_reserve_resources(void);
+extern void e820_mark_nosave_regions(void);
 extern void e820_print_map(char *who);
 extern int e820_any_mapped(unsigned long start, unsigned long end, unsigned type);
 extern int e820_all_mapped(unsigned long start, unsigned long end, unsigned type);
_
From: "Rafael J. Wysocki" <rjw@sisk.pl>

The current suspend code has to be run on one CPU, so we use the CPU
hotplug to take the non-boot CPUs offline on SMP machines.  However, we
should also make sure that these CPUs will not be enabled by someone else
after we have disabled them.

The functions disable_nonboot_cpus() and enable_nonboot_cpus() are moved to
kernel/cpu.c, because they now refer to some stuff in there that should
better be static.  Also it's better if disable_nonboot_cpus() returns an
error instead of panicking if something goes wrong, and
enable_nonboot_cpus() has no reason to panic(), because the CPUs may have
been enabled by the userland before it tries to take them online.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Acked-by: Pavel Machek <pavel@ucw.cz>
Signed-off-by: Andrew Morton <akpm@osdl.org>
---

diff -puN include/linux/cpu.h~disable-cpu-hotplug-during-suspend-2 include/linux/cpu.h
--- a/include/linux/cpu.h~disable-cpu-hotplug-during-suspend-2
+++ a/include/linux/cpu.h
@@ -89,4 +89,12 @@ int cpu_down(unsigned int cpu);
 static inline int cpu_is_offline(int cpu) { return 0; }
 #endif
 
+#ifdef CONFIG_SUSPEND_SMP
+extern int disable_nonboot_cpus(void);
+extern void enable_nonboot_cpus(void);
+#else
+static inline int disable_nonboot_cpus(void) { return 0; }
+static inline void enable_nonboot_cpus(void) {}
+#endif
+
 #endif /* _LINUX_CPU_H_ */
diff -puN include/linux/suspend.h~disable-cpu-hotplug-during-suspend-2 include/linux/suspend.h
--- a/include/linux/suspend.h~disable-cpu-hotplug-during-suspend-2
+++ a/include/linux/suspend.h
@@ -57,14 +57,6 @@ static inline int software_suspend(void)
 }
 #endif /* CONFIG_PM */
 
-#ifdef CONFIG_SUSPEND_SMP
-extern void disable_nonboot_cpus(void);
-extern void enable_nonboot_cpus(void);
-#else
-static inline void disable_nonboot_cpus(void) {}
-static inline void enable_nonboot_cpus(void) {}
-#endif
-
 void save_processor_state(void);
 void restore_processor_state(void);
 struct saved_context;
diff -puN kernel/cpu.c~disable-cpu-hotplug-during-suspend-2 kernel/cpu.c
--- a/kernel/cpu.c~disable-cpu-hotplug-during-suspend-2
+++ a/kernel/cpu.c
@@ -21,6 +21,11 @@ static DEFINE_MUTEX(cpu_bitmask_lock);
 
 static __cpuinitdata BLOCKING_NOTIFIER_HEAD(cpu_chain);
 
+/* If set, cpu_up and cpu_down will return -EBUSY and do nothing.
+ * Should always be manipulated under cpu_add_remove_lock
+ */
+static int cpu_hotplug_disabled;
+
 #ifdef CONFIG_HOTPLUG_CPU
 
 /* Crappy recursive lock-takers in cpufreq! Complain loudly about idiots */
@@ -108,30 +113,25 @@ static int take_cpu_down(void *unused)
 	return 0;
 }
 
-int cpu_down(unsigned int cpu)
+/* Requires cpu_add_remove_lock to be held */
+static int _cpu_down(unsigned int cpu)
 {
 	int err;
 	struct task_struct *p;
 	cpumask_t old_allowed, tmp;
 
-	mutex_lock(&cpu_add_remove_lock);
-	if (num_online_cpus() == 1) {
-		err = -EBUSY;
-		goto out;
-	}
+	if (num_online_cpus() == 1)
+		return -EBUSY;
 
-	if (!cpu_online(cpu)) {
-		err = -EINVAL;
-		goto out;
-	}
+	if (!cpu_online(cpu))
+		return -EINVAL;
 
 	err = blocking_notifier_call_chain(&cpu_chain, CPU_DOWN_PREPARE,
 						(void *)(long)cpu);
 	if (err == NOTIFY_BAD) {
 		printk("%s: attempt to take down CPU %u failed\n",
 				__FUNCTION__, cpu);
-		err = -EINVAL;
-		goto out;
+		return -EINVAL;
 	}
 
 	/* Ensure that we are not runnable on dying cpu */
@@ -179,22 +179,32 @@ out_thread:
 	err = kthread_stop(p);
 out_allowed:
 	set_cpus_allowed(current, old_allowed);
-out:
+	return err;
+}
+
+int cpu_down(unsigned int cpu)
+{
+	int err = 0;
+
+	mutex_lock(&cpu_add_remove_lock);
+	if (cpu_hotplug_disabled)
+		err = -EBUSY;
+	else
+		err = _cpu_down(cpu);
+
 	mutex_unlock(&cpu_add_remove_lock);
 	return err;
 }
 #endif /*CONFIG_HOTPLUG_CPU*/
 
-int __devinit cpu_up(unsigned int cpu)
+/* Requires cpu_add_remove_lock to be held */
+static int __devinit _cpu_up(unsigned int cpu)
 {
 	int ret;
 	void *hcpu = (void *)(long)cpu;
 
-	mutex_lock(&cpu_add_remove_lock);
-	if (cpu_online(cpu) || !cpu_present(cpu)) {
-		ret = -EINVAL;
-		goto out;
-	}
+	if (cpu_online(cpu) || !cpu_present(cpu))
+		return -EINVAL;
 
 	ret = blocking_notifier_call_chain(&cpu_chain, CPU_UP_PREPARE, hcpu);
 	if (ret == NOTIFY_BAD) {
@@ -219,7 +229,95 @@ out_notify:
 	if (ret != 0)
 		blocking_notifier_call_chain(&cpu_chain,
 				CPU_UP_CANCELED, hcpu);
+
+	return ret;
+}
+
+int __devinit cpu_up(unsigned int cpu)
+{
+	int err = 0;
+
+	mutex_lock(&cpu_add_remove_lock);
+	if (cpu_hotplug_disabled)
+		err = -EBUSY;
+	else
+		err = _cpu_up(cpu);
+
+	mutex_unlock(&cpu_add_remove_lock);
+	return err;
+}
+
+#ifdef CONFIG_SUSPEND_SMP
+static cpumask_t frozen_cpus;
+
+int disable_nonboot_cpus(void)
+{
+	int cpu, first_cpu, error;
+
+	mutex_lock(&cpu_add_remove_lock);
+	first_cpu = first_cpu(cpu_present_map);
+	if (!cpu_online(first_cpu)) {
+		error = _cpu_up(first_cpu);
+		if (error) {
+			printk(KERN_ERR "Could not bring CPU%d up.\n",
+				first_cpu);
+			goto out;
+		}
+	}
+	error = set_cpus_allowed(current, cpumask_of_cpu(first_cpu));
+	if (error) {
+		printk(KERN_ERR "Could not run on CPU%d\n", first_cpu);
+		goto out;
+	}
+	/* We take down all of the non-boot CPUs in one shot to avoid races
+	 * with the userspace trying to use the CPU hotplug at the same time
+	 */
+	cpus_clear(frozen_cpus);
+	printk("Disabling non-boot CPUs ...\n");
+	for_each_online_cpu(cpu) {
+		if (cpu == first_cpu)
+			continue;
+		error = _cpu_down(cpu);
+		if (!error) {
+			cpu_set(cpu, frozen_cpus);
+			printk("CPU%d is down\n", cpu);
+		} else {
+			printk(KERN_ERR "Error taking CPU%d down: %d\n",
+				cpu, error);
+			break;
+		}
+	}
+	if (!error) {
+		BUG_ON(num_online_cpus() > 1);
+		/* Make sure the CPUs won't be enabled by someone else */
+		cpu_hotplug_disabled = 1;
+	} else {
+		printk(KERN_ERR "Non-boot CPUs are not disabled");
+	}
 out:
 	mutex_unlock(&cpu_add_remove_lock);
-	return ret;
+	return error;
+}
+
+void enable_nonboot_cpus(void)
+{
+	int cpu, error;
+
+	/* Allow everyone to use the CPU hotplug again */
+	mutex_lock(&cpu_add_remove_lock);
+	cpu_hotplug_disabled = 0;
+	mutex_unlock(&cpu_add_remove_lock);
+
+	printk("Enabling non-boot CPUs ...\n");
+	for_each_cpu_mask(cpu, frozen_cpus) {
+		error = cpu_up(cpu);
+		if (!error) {
+			printk("CPU%d is up\n", cpu);
+			continue;
+		}
+		printk(KERN_WARNING "Error taking CPU%d up: %d\n",
+			cpu, error);
+	}
+	cpus_clear(frozen_cpus);
 }
+#endif
diff -puN kernel/power/disk.c~disable-cpu-hotplug-during-suspend-2 kernel/power/disk.c
--- a/kernel/power/disk.c~disable-cpu-hotplug-during-suspend-2
+++ a/kernel/power/disk.c
@@ -18,6 +18,7 @@
 #include <linux/fs.h>
 #include <linux/mount.h>
 #include <linux/pm.h>
+#include <linux/cpu.h>
 
 #include "power.h"
 
@@ -72,7 +73,10 @@ static int prepare_processes(void)
 	int error;
 
 	pm_prepare_console();
-	disable_nonboot_cpus();
+
+	error = disable_nonboot_cpus();
+	if (error)
+		goto enable_cpus;
 
 	if (freeze_processes()) {
 		error = -EBUSY;
@@ -84,6 +88,7 @@ static int prepare_processes(void)
 		return 0;
 thaw:
 	thaw_processes();
+enable_cpus:
 	enable_nonboot_cpus();
 	pm_restore_console();
 	return error;
diff -puN kernel/power/main.c~disable-cpu-hotplug-during-suspend-2 kernel/power/main.c
--- a/kernel/power/main.c~disable-cpu-hotplug-during-suspend-2
+++ a/kernel/power/main.c
@@ -16,6 +16,7 @@
 #include <linux/init.h>
 #include <linux/pm.h>
 #include <linux/console.h>
+#include <linux/cpu.h>
 
 #include "power.h"
 
@@ -51,7 +52,7 @@ void pm_set_ops(struct pm_ops * ops)
 
 static int suspend_prepare(suspend_state_t state)
 {
-	int error = 0;
+	int error;
 	unsigned int free_pages;
 
 	if (!pm_ops || !pm_ops->enter)
@@ -63,12 +64,9 @@ static int suspend_prepare(suspend_state
 
 	pm_prepare_console();
 
-	disable_nonboot_cpus();
-
-	if (num_online_cpus() != 1) {
-		error = -EPERM;
+	error = disable_nonboot_cpus();
+	if (error)
 		goto Enable_cpu;
-	}
 
 	if (freeze_processes()) {
 		error = -EAGAIN;
diff -puN kernel/power/Makefile~disable-cpu-hotplug-during-suspend-2 kernel/power/Makefile
--- a/kernel/power/Makefile~disable-cpu-hotplug-during-suspend-2
+++ a/kernel/power/Makefile
@@ -7,6 +7,4 @@ obj-y				:= main.o process.o console.o
 obj-$(CONFIG_PM_LEGACY)		+= pm.o
 obj-$(CONFIG_SOFTWARE_SUSPEND)	+= swsusp.o disk.o snapshot.o swap.o user.o
 
-obj-$(CONFIG_SUSPEND_SMP)	+= smp.o
-
 obj-$(CONFIG_MAGIC_SYSRQ)	+= poweroff.o
diff -puN kernel/power/smp.c~disable-cpu-hotplug-during-suspend-2 /dev/null
--- a/kernel/power/smp.c
+++ /dev/null
@@ -1,62 +0,0 @@
-/*
- * drivers/power/smp.c - Functions for stopping other CPUs.
- *
- * Copyright 2004 Pavel Machek <pavel@suse.cz>
- * Copyright (C) 2002-2003 Nigel Cunningham <ncunningham@clear.net.nz>
- *
- * This file is released under the GPLv2.
- */
-
-#undef DEBUG
-
-#include <linux/smp_lock.h>
-#include <linux/interrupt.h>
-#include <linux/suspend.h>
-#include <linux/module.h>
-#include <linux/cpu.h>
-#include <asm/atomic.h>
-#include <asm/tlbflush.h>
-
-/* This is protected by pm_sem semaphore */
-static cpumask_t frozen_cpus;
-
-void disable_nonboot_cpus(void)
-{
-	int cpu, error;
-
-	error = 0;
-	cpus_clear(frozen_cpus);
-	printk("Freezing cpus ...\n");
-	for_each_online_cpu(cpu) {
-		if (cpu == 0)
-			continue;
-		error = cpu_down(cpu);
-		if (!error) {
-			cpu_set(cpu, frozen_cpus);
-			printk("CPU%d is down\n", cpu);
-			continue;
-		}
-		printk("Error taking cpu %d down: %d\n", cpu, error);
-	}
-	BUG_ON(raw_smp_processor_id() != 0);
-	if (error)
-		panic("cpus not sleeping");
-}
-
-void enable_nonboot_cpus(void)
-{
-	int cpu, error;
-
-	printk("Thawing cpus ...\n");
-	for_each_cpu_mask(cpu, frozen_cpus) {
-		error = cpu_up(cpu);
-		if (!error) {
-			printk("CPU%d is up\n", cpu);
-			continue;
-		}
-		printk("Error taking cpu %d up: %d\n", cpu, error);
-		panic("Not enough cpus");
-	}
-	cpus_clear(frozen_cpus);
-}
-
diff -puN kernel/power/user.c~disable-cpu-hotplug-during-suspend-2 kernel/power/user.c
--- a/kernel/power/user.c~disable-cpu-hotplug-during-suspend-2
+++ a/kernel/power/user.c
@@ -19,6 +19,7 @@
 #include <linux/swapops.h>
 #include <linux/pm.h>
 #include <linux/fs.h>
+#include <linux/cpu.h>
 
 #include <asm/uaccess.h>
 
@@ -139,12 +140,15 @@ static int snapshot_ioctl(struct inode *
 		if (data->frozen)
 			break;
 		down(&pm_sem);
-		disable_nonboot_cpus();
-		if (freeze_processes()) {
-			thaw_processes();
-			enable_nonboot_cpus();
-			error = -EBUSY;
+		error = disable_nonboot_cpus();
+		if (!error) {
+			error = freeze_processes();
+			if (error) {
+				thaw_processes();
+				error = -EBUSY;
+			}
 		}
+		enable_nonboot_cpus();
 		up(&pm_sem);
 		if (!error)
 			data->frozen = 1;
_

Places

File linux-2.6-mm-suspend-improvements.patch of Package kernel (Revision cf0079dcee8f8d4336adf1ec51657db0)

Places