Merge git://git.kernel.org/pub/scm/linux/kernel/git/agk/linux-2.6-dm

* git://git.kernel.org/pub/scm/linux/kernel/git/agk/linux-2.6-dm:
  dm: tidy local_init
  dm: remove unused flush_all
  dm raid1: separate region_hash interface part1
  dm: mark split bio as cloned
  dm crypt: remove waitqueue
  dm crypt: fix async split
  dm crypt: tidy sector
  dm: remove dm header from targets
  dm: publish array_too_big
  dm exception store: fix misordered writes
  dm exception store: refactor zero_area
  dm snapshot: drop unused last_percent
  dm snapshot: fix primary_pe race
  dm kcopyd: avoid queue shuffle

drivers/md/Makefile

@@ -34,7 +34,7 @@ obj-$(CONFIG_DM_CRYPT)		+= dm-crypt.o
 obj-$(CONFIG_DM_DELAY)		+= dm-delay.o
 obj-$(CONFIG_DM_MULTIPATH)	+= dm-multipath.o dm-round-robin.o
 obj-$(CONFIG_DM_SNAPSHOT)	+= dm-snapshot.o
-obj-$(CONFIG_DM_MIRROR)		+= dm-mirror.o dm-log.o
+obj-$(CONFIG_DM_MIRROR)		+= dm-mirror.o dm-log.o dm-region-hash.o
 obj-$(CONFIG_DM_ZERO)		+= dm-zero.o
 
 quiet_cmd_unroll = UNROLL  $@

drivers/md/dm-crypt.c

@@ -23,7 +23,7 @@
 #include <asm/page.h>
 #include <asm/unaligned.h>
 
-#include "dm.h"
+#include <linux/device-mapper.h>
 
 #define DM_MSG_PREFIX "crypt"
 #define MESG_STR(x) x, sizeof(x)
@@ -56,6 +56,7 @@ struct dm_crypt_io {
 	atomic_t pending;
 	int error;
 	sector_t sector;
+	struct dm_crypt_io *base_io;
 };
 
 struct dm_crypt_request {
@@ -93,7 +94,6 @@ struct crypt_config {
 
 	struct workqueue_struct *io_queue;
 	struct workqueue_struct *crypt_queue;
-	wait_queue_head_t writeq;
 
 	/*
 	 * crypto related data
@@ -534,6 +534,7 @@ static struct dm_crypt_io *crypt_io_alloc(struct dm_target *ti,
 	io->base_bio = bio;
 	io->sector = sector;
 	io->error = 0;
+	io->base_io = NULL;
 	atomic_set(&io->pending, 0);
 
 	return io;
@@ -547,6 +548,7 @@ static void crypt_inc_pending(struct dm_crypt_io *io)
 /*
  * One of the bios was finished. Check for completion of
  * the whole request and correctly clean up the buffer.
+ * If base_io is set, wait for the last fragment to complete.
  */
 static void crypt_dec_pending(struct dm_crypt_io *io)
 {
@@ -555,7 +557,14 @@ static void crypt_dec_pending(struct dm_crypt_io *io)
 	if (!atomic_dec_and_test(&io->pending))
 		return;
 
-	bio_endio(io->base_bio, io->error);
+	if (likely(!io->base_io))
+		bio_endio(io->base_bio, io->error);
+	else {
+		if (io->error && !io->base_io->error)
+			io->base_io->error = io->error;
+		crypt_dec_pending(io->base_io);
+	}
+
 	mempool_free(io, cc->io_pool);
 }
 
@@ -646,10 +655,7 @@ static void kcryptd_io_read(struct dm_crypt_io *io)
 static void kcryptd_io_write(struct dm_crypt_io *io)
 {
 	struct bio *clone = io->ctx.bio_out;
-	struct crypt_config *cc = io->target->private;
-
 	generic_make_request(clone);
-	wake_up(&cc->writeq);
 }
 
 static void kcryptd_io(struct work_struct *work)
@@ -688,7 +694,6 @@ static void kcryptd_crypt_write_io_submit(struct dm_crypt_io *io,
 	BUG_ON(io->ctx.idx_out < clone->bi_vcnt);
 
 	clone->bi_sector = cc->start + io->sector;
-	io->sector += bio_sectors(clone);
 
 	if (async)
 		kcryptd_queue_io(io);
@@ -700,16 +705,18 @@ static void kcryptd_crypt_write_convert(struct dm_crypt_io *io)
 {
 	struct crypt_config *cc = io->target->private;
 	struct bio *clone;
+	struct dm_crypt_io *new_io;
 	int crypt_finished;
 	unsigned out_of_pages = 0;
 	unsigned remaining = io->base_bio->bi_size;
+	sector_t sector = io->sector;
 	int r;
 
 	/*
 	 * Prevent io from disappearing until this function completes.
 	 */
 	crypt_inc_pending(io);
-	crypt_convert_init(cc, &io->ctx, NULL, io->base_bio, io->sector);
+	crypt_convert_init(cc, &io->ctx, NULL, io->base_bio, sector);
 
 	/*
 	 * The allocated buffers can be smaller than the whole bio,
@@ -726,6 +733,7 @@ static void kcryptd_crypt_write_convert(struct dm_crypt_io *io)
 		io->ctx.idx_out = 0;
 
 		remaining -= clone->bi_size;
+		sector += bio_sectors(clone);
 
 		crypt_inc_pending(io);
 		r = crypt_convert(cc, &io->ctx);
@@ -741,6 +749,8 @@ static void kcryptd_crypt_write_convert(struct dm_crypt_io *io)
 			 */
 			if (unlikely(r < 0))
 				break;
+
+			io->sector = sector;
 		}
 
 		/*
@@ -750,8 +760,33 @@ static void kcryptd_crypt_write_convert(struct dm_crypt_io *io)
 		if (unlikely(out_of_pages))
 			congestion_wait(WRITE, HZ/100);
 
-		if (unlikely(remaining))
-			wait_event(cc->writeq, !atomic_read(&io->ctx.pending));
+		/*
+		 * With async crypto it is unsafe to share the crypto context
+		 * between fragments, so switch to a new dm_crypt_io structure.
+		 */
+		if (unlikely(!crypt_finished && remaining)) {
+			new_io = crypt_io_alloc(io->target, io->base_bio,
+						sector);
+			crypt_inc_pending(new_io);
+			crypt_convert_init(cc, &new_io->ctx, NULL,
+					   io->base_bio, sector);
+			new_io->ctx.idx_in = io->ctx.idx_in;
+			new_io->ctx.offset_in = io->ctx.offset_in;
+
+			/*
+			 * Fragments after the first use the base_io
+			 * pending count.
+			 */
+			if (!io->base_io)
+				new_io->base_io = io;
+			else {
+				new_io->base_io = io->base_io;
+				crypt_inc_pending(io->base_io);
+				crypt_dec_pending(io);
+			}
+
+			io = new_io;
+		}
 	}
 
 	crypt_dec_pending(io);
@@ -1078,7 +1113,6 @@ static int crypt_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 		goto bad_crypt_queue;
 	}
 
-	init_waitqueue_head(&cc->writeq);
 	ti->private = cc;
 	return 0;
 

drivers/md/dm-delay.c

@@ -13,7 +13,8 @@
 #include <linux/bio.h>
 #include <linux/slab.h>
 
-#include "dm.h"
+#include <linux/device-mapper.h>
+
 #include "dm-bio-list.h"
 
 #define DM_MSG_PREFIX "delay"

drivers/md/dm-exception-store.c

@@ -7,7 +7,6 @@
  * This file is released under the GPL.
  */
 
-#include "dm.h"
 #include "dm-snap.h"
 
 #include <linux/mm.h>
@@ -104,6 +103,11 @@ struct pstore {
 	 */
 	void *area;
 
+	/*
+	 * An area of zeros used to clear the next area.
+	 */
+	void *zero_area;
+
 	/*
 	 * Used to keep track of which metadata area the data in
 	 * 'chunk' refers to.
@@ -149,13 +153,22 @@ static int alloc_area(struct pstore *ps)
 	if (!ps->area)
 		return r;
 
+	ps->zero_area = vmalloc(len);
+	if (!ps->zero_area) {
+		vfree(ps->area);
+		return r;
+	}
+	memset(ps->zero_area, 0, len);
+
 	return 0;
 }
 
 static void free_area(struct pstore *ps)
 {
 	vfree(ps->area);
 	ps->area = NULL;
+	vfree(ps->zero_area);
+	ps->zero_area = NULL;
 }
 
 struct mdata_req {
@@ -220,25 +233,41 @@ static chunk_t area_location(struct pstore *ps, chunk_t area)
  * Read or write a metadata area.  Remembering to skip the first
  * chunk which holds the header.
  */
-static int area_io(struct pstore *ps, chunk_t area, int rw)
+static int area_io(struct pstore *ps, int rw)
 {
 	int r;
 	chunk_t chunk;
 
-	chunk = area_location(ps, area);
+	chunk = area_location(ps, ps->current_area);
 
 	r = chunk_io(ps, chunk, rw, 0);
 	if (r)
 		return r;
 
-	ps->current_area = area;
 	return 0;
 }
 
-static int zero_area(struct pstore *ps, chunk_t area)
+static void zero_memory_area(struct pstore *ps)
 {
 	memset(ps->area, 0, ps->snap->chunk_size << SECTOR_SHIFT);
-	return area_io(ps, area, WRITE);
+}
+
+static int zero_disk_area(struct pstore *ps, chunk_t area)
+{
+	struct dm_io_region where = {
+		.bdev = ps->snap->cow->bdev,
+		.sector = ps->snap->chunk_size * area_location(ps, area),
+		.count = ps->snap->chunk_size,
+	};
+	struct dm_io_request io_req = {
+		.bi_rw = WRITE,
+		.mem.type = DM_IO_VMA,
+		.mem.ptr.vma = ps->zero_area,
+		.client = ps->io_client,
+		.notify.fn = NULL,
+	};
+
+	return dm_io(&io_req, 1, &where, NULL);
 }
 
 static int read_header(struct pstore *ps, int *new_snapshot)
@@ -411,15 +440,14 @@ static int insert_exceptions(struct pstore *ps, int *full)
 
 static int read_exceptions(struct pstore *ps)
 {
-	chunk_t area;
 	int r, full = 1;
 
 	/*
 	 * Keeping reading chunks and inserting exceptions until
 	 * we find a partially full area.
 	 */
-	for (area = 0; full; area++) {
-		r = area_io(ps, area, READ);
+	for (ps->current_area = 0; full; ps->current_area++) {
+		r = area_io(ps, READ);
 		if (r)
 			return r;
 
@@ -428,6 +456,8 @@ static int read_exceptions(struct pstore *ps)
 			return r;
 	}
 
+	ps->current_area--;
+
 	return 0;
 }
 
@@ -486,12 +516,13 @@ static int persistent_read_metadata(struct exception_store *store)
 			return r;
 		}
 
-		r = zero_area(ps, 0);
+		ps->current_area = 0;
+		zero_memory_area(ps);
+		r = zero_disk_area(ps, 0);
 		if (r) {
-			DMWARN("zero_area(0) failed");
+			DMWARN("zero_disk_area(0) failed");
 			return r;
 		}
-
 	} else {
 		/*
 		 * Sanity checks.
@@ -551,7 +582,6 @@ static void persistent_commit(struct exception_store *store,
 			      void (*callback) (void *, int success),
 			      void *callback_context)
 {
-	int r;
 	unsigned int i;
 	struct pstore *ps = get_info(store);
 	struct disk_exception de;
@@ -572,33 +602,41 @@ static void persistent_commit(struct exception_store *store,
 	cb->context = callback_context;
 
 	/*
-	 * If there are no more exceptions in flight, or we have
-	 * filled this metadata area we commit the exceptions to
-	 * disk.
+	 * If there are exceptions in flight and we have not yet
+	 * filled this metadata area there's nothing more to do.
 	 */
-	if (atomic_dec_and_test(&ps->pending_count) ||
-	    (ps->current_committed == ps->exceptions_per_area)) {
-		r = area_io(ps, ps->current_area, WRITE);
-		if (r)
-			ps->valid = 0;
+	if (!atomic_dec_and_test(&ps->pending_count) &&
+	    (ps->current_committed != ps->exceptions_per_area))
+		return;
 
-		/*
-		 * Have we completely filled the current area ?
-		 */
-		if (ps->current_committed == ps->exceptions_per_area) {
-			ps->current_committed = 0;
-			r = zero_area(ps, ps->current_area + 1);
-			if (r)
-				ps->valid = 0;
-		}
+	/*
+	 * If we completely filled the current area, then wipe the next one.
+	 */
+	if ((ps->current_committed == ps->exceptions_per_area) &&
+	     zero_disk_area(ps, ps->current_area + 1))
+		ps->valid = 0;
 
-		for (i = 0; i < ps->callback_count; i++) {
-			cb = ps->callbacks + i;
-			cb->callback(cb->context, r == 0 ? 1 : 0);
-		}
+	/*
+	 * Commit exceptions to disk.
+	 */
+	if (ps->valid && area_io(ps, WRITE))
+		ps->valid = 0;
 
-		ps->callback_count = 0;
+	/*
+	 * Advance to the next area if this one is full.
+	 */
+	if (ps->current_committed == ps->exceptions_per_area) {
+		ps->current_committed = 0;
+		ps->current_area++;
+		zero_memory_area(ps);
 	}
+
+	for (i = 0; i < ps->callback_count; i++) {
+		cb = ps->callbacks + i;
+		cb->callback(cb->context, ps->valid);
+	}
+
+	ps->callback_count = 0;
 }
 
 static void persistent_drop(struct exception_store *store)

drivers/md/dm-io.c

@@ -5,7 +5,7 @@
  * This file is released under the GPL.
  */
 
-#include "dm.h"
+#include <linux/device-mapper.h>
 
 #include <linux/bio.h>
 #include <linux/mempool.h>