PM / Hibernate: Include storage keys in hibernation image on s390

For s390 there is one additional byte associated with each page,
the storage key. This byte contains the referenced and changed
bits and needs to be included into the hibernation image.
If the storage keys are not restored to their previous state all
original pages would appear to be dirty. This can cause
inconsistencies e.g. with read-only filesystems.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>

arch/s390/Kconfig

@@ -91,6 +91,7 @@ config S390
 	select HAVE_ARCH_MUTEX_CPU_RELAX
 	select HAVE_ARCH_JUMP_LABEL if !MARCH_G5
 	select HAVE_RCU_TABLE_FREE if SMP
+	select ARCH_SAVE_PAGE_KEYS if HIBERNATION
 	select ARCH_INLINE_SPIN_TRYLOCK
 	select ARCH_INLINE_SPIN_TRYLOCK_BH
 	select ARCH_INLINE_SPIN_LOCK

arch/s390/kernel/suspend.c

@@ -7,6 +7,7 @@
  */
 
 #include <linux/pfn.h>
+#include <linux/mm.h>
 #include <asm/system.h>
 
 /*
@@ -14,6 +15,123 @@
  */
 extern const void __nosave_begin, __nosave_end;
 
+/*
+ * The restore of the saved pages in an hibernation image will set
+ * the change and referenced bits in the storage key for each page.
+ * Overindication of the referenced bits after an hibernation cycle
+ * does not cause any harm but the overindication of the change bits
+ * would cause trouble.
+ * Use the ARCH_SAVE_PAGE_KEYS hooks to save the storage key of each
+ * page to the most significant byte of the associated page frame
+ * number in the hibernation image.
+ */
+
+/*
+ * Key storage is allocated as a linked list of pages.
+ * The size of the keys array is (PAGE_SIZE - sizeof(long))
+ */
+struct page_key_data {
+	struct page_key_data *next;
+	unsigned char data[];
+};
+
+#define PAGE_KEY_DATA_SIZE	(PAGE_SIZE - sizeof(struct page_key_data *))
+
+static struct page_key_data *page_key_data;
+static struct page_key_data *page_key_rp, *page_key_wp;
+static unsigned long page_key_rx, page_key_wx;
+
+/*
+ * For each page in the hibernation image one additional byte is
+ * stored in the most significant byte of the page frame number.
+ * On suspend no additional memory is required but on resume the
+ * keys need to be memorized until the page data has been restored.
+ * Only then can the storage keys be set to their old state.
+ */
+unsigned long page_key_additional_pages(unsigned long pages)
+{
+	return DIV_ROUND_UP(pages, PAGE_KEY_DATA_SIZE);
+}
+
+/*
+ * Free page_key_data list of arrays.
+ */
+void page_key_free(void)
+{
+	struct page_key_data *pkd;
+
+	while (page_key_data) {
+		pkd = page_key_data;
+		page_key_data = pkd->next;
+		free_page((unsigned long) pkd);
+	}
+}
+
+/*
+ * Allocate page_key_data list of arrays with enough room to store
+ * one byte for each page in the hibernation image.
+ */
+int page_key_alloc(unsigned long pages)
+{
+	struct page_key_data *pk;
+	unsigned long size;
+
+	size = DIV_ROUND_UP(pages, PAGE_KEY_DATA_SIZE);
+	while (size--) {
+		pk = (struct page_key_data *) get_zeroed_page(GFP_KERNEL);
+		if (!pk) {
+			page_key_free();
+			return -ENOMEM;
+		}
+		pk->next = page_key_data;
+		page_key_data = pk;
+	}
+	page_key_rp = page_key_wp = page_key_data;
+	page_key_rx = page_key_wx = 0;
+	return 0;
+}
+
+/*
+ * Save the storage key into the upper 8 bits of the page frame number.
+ */
+void page_key_read(unsigned long *pfn)
+{
+	unsigned long addr;
+
+	addr = (unsigned long) page_address(pfn_to_page(*pfn));
+	*(unsigned char *) pfn = (unsigned char) page_get_storage_key(addr);
+}
+
+/*
+ * Extract the storage key from the upper 8 bits of the page frame number
+ * and store it in the page_key_data list of arrays.
+ */
+void page_key_memorize(unsigned long *pfn)
+{
+	page_key_wp->data[page_key_wx] = *(unsigned char *) pfn;
+	*(unsigned char *) pfn = 0;
+	if (++page_key_wx < PAGE_KEY_DATA_SIZE)
+		return;
+	page_key_wp = page_key_wp->next;
+	page_key_wx = 0;
+}
+
+/*
+ * Get the next key from the page_key_data list of arrays and set the
+ * storage key of the page referred by @address. If @address refers to
+ * a "safe" page the swsusp_arch_resume code will transfer the storage
+ * key from the buffer page to the original page.
+ */
+void page_key_write(void *address)
+{
+	page_set_storage_key((unsigned long) address,
+			     page_key_rp->data[page_key_rx], 0);
+	if (++page_key_rx >= PAGE_KEY_DATA_SIZE)
+		return;
+	page_key_rp = page_key_rp->next;
+	page_key_rx = 0;
+}
+
 int pfn_is_nosave(unsigned long pfn)
 {
 	unsigned long nosave_begin_pfn = PFN_DOWN(__pa(&__nosave_begin));

arch/s390/kernel/swsusp_asm64.S

@@ -136,11 +136,14 @@ ENTRY(swsusp_arch_resume)
 0:
 	lg	%r2,8(%r1)
 	lg	%r4,0(%r1)
+	iske	%r0,%r4
 	lghi	%r3,PAGE_SIZE
 	lghi	%r5,PAGE_SIZE
 1:
 	mvcle	%r2,%r4,0
 	jo	1b
+	lg	%r2,8(%r1)
+	sske	%r0,%r2
 	lg	%r1,16(%r1)
 	ltgr	%r1,%r1
 	jnz	0b

include/linux/suspend.h

@@ -386,4 +386,38 @@ static inline void unlock_system_sleep(void)
 }
 #endif
 
+#ifdef CONFIG_ARCH_SAVE_PAGE_KEYS
+/*
+ * The ARCH_SAVE_PAGE_KEYS functions can be used by an architecture
+ * to save/restore additional information to/from the array of page
+ * frame numbers in the hibernation image. For s390 this is used to
+ * save and restore the storage key for each page that is included
+ * in the hibernation image.
+ */
+unsigned long page_key_additional_pages(unsigned long pages);
+int page_key_alloc(unsigned long pages);
+void page_key_free(void);
+void page_key_read(unsigned long *pfn);
+void page_key_memorize(unsigned long *pfn);
+void page_key_write(void *address);
+
+#else /* !CONFIG_ARCH_SAVE_PAGE_KEYS */
+
+static inline unsigned long page_key_additional_pages(unsigned long pages)
+{
+	return 0;
+}
+
+static inline int  page_key_alloc(unsigned long pages)
+{
+	return 0;
+}
+
+static inline void page_key_free(void) {}
+static inline void page_key_read(unsigned long *pfn) {}
+static inline void page_key_memorize(unsigned long *pfn) {}
+static inline void page_key_write(void *address) {}
+
+#endif /* !CONFIG_ARCH_SAVE_PAGE_KEYS */
+
 #endif /* _LINUX_SUSPEND_H */

kernel/power/Kconfig

@@ -65,6 +65,9 @@ config HIBERNATION
 
 	  For more information take a look at <file:Documentation/power/swsusp.txt>.
 
+config ARCH_SAVE_PAGE_KEYS
+	bool
+
 config PM_STD_PARTITION
 	string "Default resume partition"
 	depends on HIBERNATION

kernel/power/snapshot.c

@@ -1339,6 +1339,9 @@ int hibernate_preallocate_memory(void)
 	count += highmem;
 	count -= totalreserve_pages;
 
+	/* Add number of pages required for page keys (s390 only). */
+	size += page_key_additional_pages(saveable);
+
 	/* Compute the maximum number of saveable pages to leave in memory. */
 	max_size = (count - (size + PAGES_FOR_IO)) / 2
 			- 2 * DIV_ROUND_UP(reserved_size, PAGE_SIZE);
@@ -1662,6 +1665,8 @@ pack_pfns(unsigned long *buf, struct memory_bitmap *bm)
 		buf[j] = memory_bm_next_pfn(bm);
 		if (unlikely(buf[j] == BM_END_OF_MAP))
 			break;
+		/* Save page key for data page (s390 only). */
+		page_key_read(buf + j);
 	}
 }
 
@@ -1821,6 +1826,9 @@ static int unpack_orig_pfns(unsigned long *buf, struct memory_bitmap *bm)
 		if (unlikely(buf[j] == BM_END_OF_MAP))
 			break;
 
+		/* Extract and buffer page key for data page (s390 only). */
+		page_key_memorize(buf + j);
+
 		if (memory_bm_pfn_present(bm, buf[j]))
 			memory_bm_set_bit(bm, buf[j]);
 		else
@@ -2223,6 +2231,11 @@ int snapshot_write_next(struct snapshot_handle *handle)
 		if (error)
 			return error;
 
+		/* Allocate buffer for page keys. */
+		error = page_key_alloc(nr_copy_pages);
+		if (error)
+			return error;
+
 	} else if (handle->cur <= nr_meta_pages + 1) {
 		error = unpack_orig_pfns(buffer, &copy_bm);
 		if (error)
@@ -2243,6 +2256,8 @@ int snapshot_write_next(struct snapshot_handle *handle)
 		}
 	} else {
 		copy_last_highmem_page();
+		/* Restore page key for data page (s390 only). */
+		page_key_write(handle->buffer);
 		handle->buffer = get_buffer(&orig_bm, &ca);
 		if (IS_ERR(handle->buffer))
 			return PTR_ERR(handle->buffer);
@@ -2264,6 +2279,9 @@ int snapshot_write_next(struct snapshot_handle *handle)
 void snapshot_write_finalize(struct snapshot_handle *handle)
 {
 	copy_last_highmem_page();
+	/* Restore page key for data page (s390 only). */
+	page_key_write(handle->buffer);
+	page_key_free();
 	/* Free only if we have loaded the image entirely */
 	if (handle->cur > 1 && handle->cur > nr_meta_pages + nr_copy_pages) {
 		memory_bm_free(&orig_bm, PG_UNSAFE_CLEAR);