[ALT-PATCH v3 9/10] PM / Hibernate: Call flush_icache_range() on pages restored in-place

[James Morse]

Some architectures require code written to memory as if it were data to be
'cleaned' from any data caches before the processor can fetch them as new
instructions.

During resume from hibernate, the snapshot code copies some pages directly,
meaning these architectures do not get a chance to perform their cache
maintenance. Modify the read and decompress code to call
flush_icache_range() on all pages that are restored, so that the restored
in-place pages are guaranteed to be executable on these architectures.

Signed-off-by: James Morse <james.morse at arm.com>
CC: Michael Ellerman <mpe at ellerman.id.au>

---
Hi,

This is an alternative version of 9/10 [0] requested by Pavel. It isn't
possible to know which pages are being restored in place, so this code
cleans all the pages restored. This is unnecessary 99.95% of the time,
as the pages will be copied (and cleaned again) by arch code walking
restore_pblist.

I may be being over-cautions, but this looks like it will have the most
impact on powerpc, where the cache will be cleaned to memory, a page at
a time.

Patch 10/10 of the series won't build with this 9/10, I will repost the
series if we chose to use this version.


James

[0] http://www.spinics.net/lists/arm-kernel/msg463600.html

 kernel/power/swap.c | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)

diff --git a/kernel/power/swap.c b/kernel/power/swap.c
index 12cd989dadf6..a30645d2e93f 100644
--- a/kernel/power/swap.c
+++ b/kernel/power/swap.c
@@ -37,6 +37,14 @@
 #define HIBERNATE_SIG	"S1SUSPEND"
 
 /*
+ * When reading an {un,}compressed image, we may restore pages in place,
+ * in which case some architectures need these pages cleaning before they
+ * can be executed. We don't know which pages these may be, so clean the lot.
+ */
+bool clean_pages_on_read = false;
+bool clean_pages_on_decompress = false;
+
+/*
  *	The swap map is a data structure used for keeping track of each page
  *	written to a swap partition.  It consists of many swap_map_page
  *	structures that contain each an array of MAP_PAGE_ENTRIES swap entries.
@@ -241,6 +249,9 @@ static void hib_end_io(struct bio *bio)
 
 	if (bio_data_dir(bio) == WRITE)
 		put_page(page);
+	else if (clean_pages_on_read)
+		flush_icache_range((unsigned long)page_address(page),
+				   (unsigned long)page_address(page) + PAGE_SIZE);
 
 	if (bio->bi_error && !hb->error)
 		hb->error = bio->bi_error;
@@ -1049,6 +1060,7 @@ static int load_image(struct swap_map_handle *handle,
 
 	hib_init_batch(&hb);
 
+	clean_pages_on_read = true;
 	printk(KERN_INFO "PM: Loading image data pages (%u pages)...\n",
 		nr_to_read);
 	m = nr_to_read / 10;
@@ -1124,6 +1136,10 @@ static int lzo_decompress_threadfn(void *data)
 		d->unc_len = LZO_UNC_SIZE;
 		d->ret = lzo1x_decompress_safe(d->cmp + LZO_HEADER, d->cmp_len,
 		                               d->unc, &d->unc_len);
+		if (clean_pages_on_decompress)
+			flush_icache_range((unsigned long)d->unc,
+					   (unsigned long)d->unc + d->unc_len);
+
 		atomic_set(&d->stop, 1);
 		wake_up(&d->done);
 	}
@@ -1189,6 +1205,8 @@ static int load_image_lzo(struct swap_map_handle *handle,
 	}
 	memset(crc, 0, offsetof(struct crc_data, go));
 
+	clean_pages_on_decompress = true;
+
 	/*
 	 * Start the decompression threads.
 	 */
-- 
2.6.2