[RFC] UBI: MLC Support v0

[RFC] UBI: MLC Support v0

[Richard Weinberger]

as announced in [1] I'd like to submit the very first patch series to
have MLC support in UBI (and therefore also UBIFS).
The code has not changed much since the announcement but now the patches
are in a more reviewable shape. Still not mainline ready but better.
I'm sending this now as I'm on vacation until June 15th and hope that
we get some reviews on the concept.
The series consists of three parts:
	a) Generic UBI rework to make the MLC implementation possible.
	b) SLC emulation mode to access only lower pages on MLC NAND.
	c) LEB consolidation to exploit MLC NAND's higher pages.

[1] http://lists.infradead.org/pipermail/linux-mtd/2016-April/067322.html

Thanks,
//richard

---

David Gstir (1):
      ubi: Add debugfs knob to force LEB consolidation

Richard Weinberger (12):
      ubi: Undo "UBI: modify ubi_wl_flush function to clear work queue for a lnum"
      ubi: Rework UBI worker
      ubi: auto re-size after UBI thread is ready
      ubi: Kill ubi->alc_mutex
      ubi: Get rid of __schedule_ubi_work()
      ubi: Remove tst_disable_bgt debugfs knob
      ubi: Move work related functions to work.c
      ubi: Remove lnum and vol_id from erase work
      ubi: Remove usless debug info from wear_leveling_worker()
      ubi: SLC mode
      ubi: LEB consolidation
      ubi: Add debugfs knob to trigger LEB consolidation

 drivers/mtd/ubi/Kconfig       |   4 +
 drivers/mtd/ubi/Makefile      |   3 +-
 drivers/mtd/ubi/attach.c      | 407 ++++++++++++++++++--------------
 drivers/mtd/ubi/build.c       |  41 ++--
 drivers/mtd/ubi/cdev.c        |  14 +-
 drivers/mtd/ubi/consolidate.c | 516 ++++++++++++++++++++++++++++++++++++++++
 drivers/mtd/ubi/debug.c       |  49 ++--
 drivers/mtd/ubi/debug.h       |  19 +-
 drivers/mtd/ubi/eba.c         | 429 ++++++++++++++++++++++++++++------
 drivers/mtd/ubi/fastmap-wl.c  |  33 ++-
 drivers/mtd/ubi/fastmap.c     | 265 ++++++++++++++-------
 drivers/mtd/ubi/io.c          | 426 +++++++++++++++++++++++++--------
 drivers/mtd/ubi/kapi.c        |  43 +---
 drivers/mtd/ubi/ubi-media.h   |  18 +-
 drivers/mtd/ubi/ubi.h         | 221 ++++++++++++++----
 drivers/mtd/ubi/upd.c         |   7 +-
 drivers/mtd/ubi/vmt.c         | 106 +++++----
 drivers/mtd/ubi/vtbl.c        |  87 ++++---
 drivers/mtd/ubi/wl.c          | 531 +++++++++++++++---------------------------
 drivers/mtd/ubi/work.c        | 332 ++++++++++++++++++++++++++
 include/linux/mtd/ubi.h       |   1 -
 21 files changed, 2554 insertions(+), 998 deletions(-)

[PATCH 01/13] ubi: Undo "UBI: modify ubi_wl_flush function to clear work queue for a lnum"

[Richard Weinberger]

This patch undoes 62f384552b6756cf1ea71f8762d1e97dc77dbd90
("UBI: modify ubi_wl_flush function to clear work queue for a lnum"),
it was part of a larger series which tried to add encryption and secure undelete
to UBIFS. But except from this preparation patch, the series went never upstream.

While this commit offers very little functionality it makes maintenance hard as it
complicates the UBI worker logic. Let's rip it in favor to maintenance.

Signed-off-by: Richard Weinberger <richard@nod.at>
---
 drivers/mtd/ubi/cdev.c  |  2 +-
 drivers/mtd/ubi/kapi.c  | 29 +------------------------
 drivers/mtd/ubi/ubi.h   |  2 +-
 drivers/mtd/ubi/upd.c   |  4 ++--
 drivers/mtd/ubi/vmt.c   |  2 +-
 drivers/mtd/ubi/wl.c    | 58 ++++++++++++++++---------------------------------
 include/linux/mtd/ubi.h |  1 -
 7 files changed, 25 insertions(+), 73 deletions(-)

diff --git a/drivers/mtd/ubi/cdev.c b/drivers/mtd/ubi/cdev.c
index ee2b74d..7e49de9 100644
--- a/drivers/mtd/ubi/cdev.c
+++ b/drivers/mtd/ubi/cdev.c
@@ -495,7 +495,7 @@ static long vol_cdev_ioctl(struct file *file, unsigned int cmd,
 		if (err)
 			break;
 
-		err = ubi_wl_flush(ubi, UBI_ALL, UBI_ALL);
+		err = ubi_wl_flush(ubi);
 		break;
 	}
 
diff --git a/drivers/mtd/ubi/kapi.c b/drivers/mtd/ubi/kapi.c
index 348dbbc..11ca859 100644
--- a/drivers/mtd/ubi/kapi.c
+++ b/drivers/mtd/ubi/kapi.c
@@ -624,7 +624,7 @@ int ubi_leb_erase(struct ubi_volume_desc *desc, int lnum)
 	if (err)
 		return err;
 
-	return ubi_wl_flush(ubi, vol->vol_id, lnum);
+	return ubi_wl_flush(ubi);
 }
 EXPORT_SYMBOL_GPL(ubi_leb_erase);
 
@@ -777,33 +777,6 @@ int ubi_sync(int ubi_num)
 }
 EXPORT_SYMBOL_GPL(ubi_sync);
 
-/**
- * ubi_flush - flush UBI work queue.
- * @ubi_num: UBI device to flush work queue
- * @vol_id: volume id to flush for
- * @lnum: logical eraseblock number to flush for
- *
- * This function executes all pending works for a particular volume id / logical
- * eraseblock number pair. If either value is set to %UBI_ALL, then it acts as
- * a wildcard for all of the corresponding volume numbers or logical
- * eraseblock numbers. It returns zero in case of success and a negative error
- * code in case of failure.
- */
-int ubi_flush(int ubi_num, int vol_id, int lnum)
-{
-	struct ubi_device *ubi;
-	int err = 0;
-
-	ubi = ubi_get_device(ubi_num);
-	if (!ubi)
-		return -ENODEV;
-
-	err = ubi_wl_flush(ubi, vol_id, lnum);
-	ubi_put_device(ubi);
-	return err;
-}
-EXPORT_SYMBOL_GPL(ubi_flush);
-
 BLOCKING_NOTIFIER_HEAD(ubi_notifiers);
 
 /**
diff --git a/drivers/mtd/ubi/ubi.h b/drivers/mtd/ubi/ubi.h
index 61d4e99..c472aad 100644
--- a/drivers/mtd/ubi/ubi.h
+++ b/drivers/mtd/ubi/ubi.h
@@ -854,7 +854,7 @@ int self_check_eba(struct ubi_device *ubi, struct ubi_attach_info *ai_fastmap,
 int ubi_wl_get_peb(struct ubi_device *ubi);
 int ubi_wl_put_peb(struct ubi_device *ubi, int vol_id, int lnum,
 		   int pnum, int torture);
-int ubi_wl_flush(struct ubi_device *ubi, int vol_id, int lnum);
+int ubi_wl_flush(struct ubi_device *ubi);
 int ubi_wl_scrub_peb(struct ubi_device *ubi, int pnum);
 int ubi_wl_init(struct ubi_device *ubi, struct ubi_attach_info *ai);
 void ubi_wl_close(struct ubi_device *ubi);
diff --git a/drivers/mtd/ubi/upd.c b/drivers/mtd/ubi/upd.c
index 0134ba3..6799cc4 100644
--- a/drivers/mtd/ubi/upd.c
+++ b/drivers/mtd/ubi/upd.c
@@ -149,7 +149,7 @@ int ubi_start_update(struct ubi_device *ubi, struct ubi_volume *vol,
 	}
 
 	if (bytes == 0) {
-		err = ubi_wl_flush(ubi, UBI_ALL, UBI_ALL);
+		err = ubi_wl_flush(ubi);
 		if (err)
 			return err;
 
@@ -361,7 +361,7 @@ int ubi_more_update_data(struct ubi_device *ubi, struct ubi_volume *vol,
 
 	ubi_assert(vol->upd_received <= vol->upd_bytes);
 	if (vol->upd_received == vol->upd_bytes) {
-		err = ubi_wl_flush(ubi, UBI_ALL, UBI_ALL);
+		err = ubi_wl_flush(ubi);
 		if (err)
 			return err;
 		/* The update is finished, clear the update marker */
diff --git a/drivers/mtd/ubi/vmt.c b/drivers/mtd/ubi/vmt.c
index 10059df..714e0b3 100644
--- a/drivers/mtd/ubi/vmt.c
+++ b/drivers/mtd/ubi/vmt.c
@@ -237,7 +237,7 @@ int ubi_create_volume(struct ubi_device *ubi, struct ubi_mkvol_req *req)
 	 * Finish all pending erases because there may be some LEBs belonging
 	 * to the same volume ID.
 	 */
-	err = ubi_wl_flush(ubi, vol_id, UBI_ALL);
+	err = ubi_wl_flush(ubi);
 	if (err)
 		goto out_acc;
 
diff --git a/drivers/mtd/ubi/wl.c b/drivers/mtd/ubi/wl.c
index 959c7b12..a54c3f4 100644
--- a/drivers/mtd/ubi/wl.c
+++ b/drivers/mtd/ubi/wl.c
@@ -1331,54 +1331,23 @@ retry:
 /**
  * ubi_wl_flush - flush all pending works.
  * @ubi: UBI device description object
- * @vol_id: the volume id to flush for
- * @lnum: the logical eraseblock number to flush for
  *
- * This function executes all pending works for a particular volume id /
- * logical eraseblock number pair. If either value is set to %UBI_ALL, then it
- * acts as a wildcard for all of the corresponding volume numbers or logical
- * eraseblock numbers. It returns zero in case of success and a negative error
- * code in case of failure.
+ * This function returns zero in case of success and a negative error code in
+ * case of failure.
  */
-int ubi_wl_flush(struct ubi_device *ubi, int vol_id, int lnum)
+int ubi_wl_flush(struct ubi_device *ubi)
 {
 	int err = 0;
-	int found = 1;
 
 	/*
 	 * Erase while the pending works queue is not empty, but not more than
 	 * the number of currently pending works.
 	 */
-	dbg_wl("flush pending work for LEB %d:%d (%d pending works)",
-	       vol_id, lnum, ubi->works_count);
-
-	while (found) {
-		struct ubi_work *wrk, *tmp;
-		found = 0;
-
-		down_read(&ubi->work_sem);
-		spin_lock(&ubi->wl_lock);
-		list_for_each_entry_safe(wrk, tmp, &ubi->works, list) {
-			if ((vol_id == UBI_ALL || wrk->vol_id == vol_id) &&
-			    (lnum == UBI_ALL || wrk->lnum == lnum)) {
-				list_del(&wrk->list);
-				ubi->works_count -= 1;
-				ubi_assert(ubi->works_count >= 0);
-				spin_unlock(&ubi->wl_lock);
-
-				err = wrk->func(ubi, wrk, 0);
-				if (err) {
-					up_read(&ubi->work_sem);
-					return err;
-				}
-
-				spin_lock(&ubi->wl_lock);
-				found = 1;
-				break;
-			}
-		}
-		spin_unlock(&ubi->wl_lock);
-		up_read(&ubi->work_sem);
+	dbg_wl("flush (%d pending works)", ubi->works_count);
+	while (ubi->works_count) {
+		err = do_work(ubi);
+		if (err)
+			return err;
 	}
 
 	/*
@@ -1388,6 +1357,17 @@ int ubi_wl_flush(struct ubi_device *ubi, int vol_id, int lnum)
 	down_write(&ubi->work_sem);
 	up_write(&ubi->work_sem);
 
+	/*
+	 * And in case last was the WL worker and it canceled the LEB
+	 * movement, flush again.
+	 */
+	while (ubi->works_count) {
+		dbg_wl("flush more (%d pending works)", ubi->works_count);
+		err = do_work(ubi);
+		if (err)
+			return err;
+	}
+
 	return err;
 }
 
diff --git a/include/linux/mtd/ubi.h b/include/linux/mtd/ubi.h
index 1e271cb..2fe0f20 100644
--- a/include/linux/mtd/ubi.h
+++ b/include/linux/mtd/ubi.h
@@ -260,7 +260,6 @@ int ubi_leb_unmap(struct ubi_volume_desc *desc, int lnum);
 int ubi_leb_map(struct ubi_volume_desc *desc, int lnum);
 int ubi_is_mapped(struct ubi_volume_desc *desc, int lnum);
 int ubi_sync(int ubi_num);
-int ubi_flush(int ubi_num, int vol_id, int lnum);
 
 /*
  * This function is the same as the 'ubi_leb_read()' function, but it does not
-- 
2.7.3

[PATCH 02/13] ubi: Rework UBI worker

[Richard Weinberger]

>From now on all UBI work is done exclusively within the worker thread.
This relaxes the locking situation a lot. It also allows us to run work
without an extra lock held.
Waiting for work is now implemented using a struct completition.

Signed-off-by: Richard Weinberger <richard@nod.at>
---
 drivers/mtd/ubi/build.c      |   5 +-
 drivers/mtd/ubi/fastmap-wl.c |  12 +-
 drivers/mtd/ubi/fastmap.c    |   4 +-
 drivers/mtd/ubi/ubi.h        |  19 +++-
 drivers/mtd/ubi/upd.c        |   1 +
 drivers/mtd/ubi/wl.c         | 254 +++++++++++++++++++++++++++++++------------
 6 files changed, 211 insertions(+), 84 deletions(-)

diff --git a/drivers/mtd/ubi/build.c b/drivers/mtd/ubi/build.c
index 16baeb5..ee3c833 100644
--- a/drivers/mtd/ubi/build.c
+++ b/drivers/mtd/ubi/build.c
@@ -979,6 +979,9 @@ int ubi_attach_mtd_dev(struct mtd_info *mtd, int ubi_num,
 	if (!ubi->fm_buf)
 		goto out_free;
 #endif
+	ubi->thread_enabled = 1;
+	ubi->thread_suspended = 1;
+
 	err = ubi_attach(ubi, 0);
 	if (err) {
 		ubi_err(ubi, "failed to attach mtd%d, error %d",
@@ -1032,7 +1035,7 @@ int ubi_attach_mtd_dev(struct mtd_info *mtd, int ubi_num,
 	 * checks @ubi->thread_enabled. Otherwise we may fail to wake it up.
 	 */
 	spin_lock(&ubi->wl_lock);
-	ubi->thread_enabled = 1;
+	ubi->thread_suspended = 0;
 	wake_up_process(ubi->bgt_thread);
 	spin_unlock(&ubi->wl_lock);
 
diff --git a/drivers/mtd/ubi/fastmap-wl.c b/drivers/mtd/ubi/fastmap-wl.c
index 30d3999..c740095 100644
--- a/drivers/mtd/ubi/fastmap-wl.c
+++ b/drivers/mtd/ubi/fastmap-wl.c
@@ -182,14 +182,12 @@ void ubi_refill_pools(struct ubi_device *ubi)
  */
 static int produce_free_peb(struct ubi_device *ubi)
 {
-	int err;
-
-	while (!ubi->free.rb_node && ubi->works_count) {
+	while (!ubi->free.rb_node) {
 		dbg_wl("do one work synchronously");
-		err = do_work(ubi);
-
-		if (err)
-			return err;
+		if (!wl_do_one_work_sync(ubi)) {
+			/* Nothing to do. We have to give up. */
+			return -ENOSPC;
+		}
 	}
 
 	return 0;
diff --git a/drivers/mtd/ubi/fastmap.c b/drivers/mtd/ubi/fastmap.c
index 990898b..3d5e674 100644
--- a/drivers/mtd/ubi/fastmap.c
+++ b/drivers/mtd/ubi/fastmap.c
@@ -1607,11 +1607,11 @@ int ubi_update_fastmap(struct ubi_device *ubi)
 		new_fm->e[0] = tmp_e;
 	}
 
-	down_write(&ubi->work_sem);
+	ubi_wl_suspend_work(ubi);
 	down_write(&ubi->fm_eba_sem);
 	ret = ubi_write_fastmap(ubi, new_fm);
 	up_write(&ubi->fm_eba_sem);
-	up_write(&ubi->work_sem);
+	ubi_wl_resume_work(ubi);
 
 	if (ret)
 		goto err;
diff --git a/drivers/mtd/ubi/ubi.h b/drivers/mtd/ubi/ubi.h
index c472aad..4486869 100644
--- a/drivers/mtd/ubi/ubi.h
+++ b/drivers/mtd/ubi/ubi.h
@@ -30,6 +30,8 @@
 #include <linux/mutex.h>
 #include <linux/rwsem.h>
 #include <linux/spinlock.h>
+#include <linux/completion.h>
+#include <linux/kref.h>
 #include <linux/fs.h>
 #include <linux/cdev.h>
 #include <linux/device.h>
@@ -481,8 +483,8 @@ struct ubi_debug_info {
  *	     @erroneous, @erroneous_peb_count, @fm_work_scheduled, @fm_pool,
  *	     and @fm_wl_pool fields
  * @move_mutex: serializes eraseblock moves
- * @work_sem: used to wait for all the scheduled works to finish and prevent
- * new works from being submitted
+ * @work_mutex: used to protect the worker thread and block it temporary
+ * @cur_work: Pointer to the currently executed work
  * @wl_scheduled: non-zero if the wear-leveling was scheduled
  * @lookuptbl: a table to quickly find a &struct ubi_wl_entry object for any
  *             physical eraseblock
@@ -493,6 +495,7 @@ struct ubi_debug_info {
  * @works_count: count of pending works
  * @bgt_thread: background thread description object
  * @thread_enabled: if the background thread is enabled
+ * @thread_suspended: if the background thread is suspended
  * @bgt_name: background thread name
  *
  * @flash_size: underlying MTD device size (in bytes)
@@ -587,7 +590,8 @@ struct ubi_device {
 	int pq_head;
 	spinlock_t wl_lock;
 	struct mutex move_mutex;
-	struct rw_semaphore work_sem;
+	struct mutex work_mutex;
+	struct ubi_work *cur_work;
 	int wl_scheduled;
 	struct ubi_wl_entry **lookuptbl;
 	struct ubi_wl_entry *move_from;
@@ -597,6 +601,7 @@ struct ubi_device {
 	int works_count;
 	struct task_struct *bgt_thread;
 	int thread_enabled;
+	int thread_suspended;
 	char bgt_name[sizeof(UBI_BGT_NAME_PATTERN)+2];
 
 	/* I/O sub-system's stuff */
@@ -752,6 +757,9 @@ struct ubi_attach_info {
  * struct ubi_work - UBI work description data structure.
  * @list: a link in the list of pending works
  * @func: worker function
+ * @ret: return value of the worker function
+ * @comp: completion to wait on a work
+ * @ref: reference counter for work objects
  * @e: physical eraseblock to erase
  * @vol_id: the volume ID on which this erasure is being performed
  * @lnum: the logical eraseblock number
@@ -767,6 +775,9 @@ struct ubi_attach_info {
 struct ubi_work {
 	struct list_head list;
 	int (*func)(struct ubi_device *ubi, struct ubi_work *wrk, int shutdown);
+	int ret;
+	struct completion comp;
+	struct kref ref;
 	/* The below fields are only relevant to erasure works */
 	struct ubi_wl_entry *e;
 	int vol_id;
@@ -865,6 +876,8 @@ int ubi_wl_put_fm_peb(struct ubi_device *ubi, struct ubi_wl_entry *used_e,
 int ubi_is_erase_work(struct ubi_work *wrk);
 void ubi_refill_pools(struct ubi_device *ubi);
 int ubi_ensure_anchor_pebs(struct ubi_device *ubi);
+void ubi_wl_suspend_work(struct ubi_device *ubi);
+void ubi_wl_resume_work(struct ubi_device *ubi);
 
 /* io.c */
 int ubi_io_read(const struct ubi_device *ubi, void *buf, int pnum, int offset,
diff --git a/drivers/mtd/ubi/upd.c b/drivers/mtd/ubi/upd.c
index 6799cc4..b7901ce 100644
--- a/drivers/mtd/ubi/upd.c
+++ b/drivers/mtd/ubi/upd.c
@@ -364,6 +364,7 @@ int ubi_more_update_data(struct ubi_device *ubi, struct ubi_volume *vol,
 		err = ubi_wl_flush(ubi);
 		if (err)
 			return err;
+
 		/* The update is finished, clear the update marker */
 		err = clear_update_marker(ubi, vol, vol->upd_bytes);
 		if (err)
diff --git a/drivers/mtd/ubi/wl.c b/drivers/mtd/ubi/wl.c
index a54c3f4..8e6e8cf 100644
--- a/drivers/mtd/ubi/wl.c
+++ b/drivers/mtd/ubi/wl.c
@@ -192,6 +192,28 @@ static void wl_entry_destroy(struct ubi_device *ubi, struct ubi_wl_entry *e)
 }
 
 /**
+ * destroy_work - destroy an UBI work.
+ * @ref: kref object
+ *
+ * This function is called by kref upon the last reference is gone.
+ */
+static void destroy_work(struct kref *ref)
+{
+	struct ubi_work *wrk = container_of(ref, struct ubi_work, ref);
+
+	kfree(wrk);
+}
+
+/**
+ * wl_work_suspended - Check whether UBI work is suspended.
+ * @e: the wear-leveling entry to add
+ */
+static bool wl_work_suspended(struct ubi_device *ubi)
+{
+	return ubi->thread_suspended || !ubi->thread_enabled;
+}
+
+/**
  * do_work - do one pending work.
  * @ubi: UBI device description object
  *
@@ -205,17 +227,12 @@ static int do_work(struct ubi_device *ubi)
 
 	cond_resched();
 
-	/*
-	 * @ubi->work_sem is used to synchronize with the workers. Workers take
-	 * it in read mode, so many of them may be doing works at a time. But
-	 * the queue flush code has to be sure the whole queue of works is
-	 * done, and it takes the mutex in write mode.
-	 */
-	down_read(&ubi->work_sem);
+	mutex_lock(&ubi->work_mutex);
 	spin_lock(&ubi->wl_lock);
-	if (list_empty(&ubi->works)) {
+	ubi_assert(!ubi->cur_work);
+	if (list_empty(&ubi->works) || wl_work_suspended(ubi)) {
 		spin_unlock(&ubi->wl_lock);
-		up_read(&ubi->work_sem);
+		mutex_unlock(&ubi->work_mutex);
 		return 0;
 	}
 
@@ -223,7 +240,9 @@ static int do_work(struct ubi_device *ubi)
 	list_del(&wrk->list);
 	ubi->works_count -= 1;
 	ubi_assert(ubi->works_count >= 0);
+	ubi->cur_work = wrk;
 	spin_unlock(&ubi->wl_lock);
+	mutex_unlock(&ubi->work_mutex);
 
 	/*
 	 * Call the worker function. Do not touch the work structure
@@ -231,13 +250,92 @@ static int do_work(struct ubi_device *ubi)
 	 * time by the worker function.
 	 */
 	err = wrk->func(ubi, wrk, 0);
+	wrk->ret = err;
 	if (err)
 		ubi_err(ubi, "work failed with error code %d", err);
-	up_read(&ubi->work_sem);
+
+	spin_lock(&ubi->wl_lock);
+	ubi->cur_work = NULL;
+	spin_unlock(&ubi->wl_lock);
+
+	complete_all(&wrk->comp);
+
+	spin_lock(&ubi->wl_lock);
+	kref_put(&wrk->ref, destroy_work);
+	spin_unlock(&ubi->wl_lock);
 
 	return err;
 }
 
+void ubi_wl_suspend_work(struct ubi_device *ubi)
+{
+	struct ubi_work *wrk = NULL;
+
+	mutex_lock(&ubi->work_mutex);
+	spin_lock(&ubi->wl_lock);
+
+	wrk = ubi->cur_work;
+	if (wrk)
+		kref_get(&wrk->ref);
+
+	ubi->thread_suspended = 1;
+
+	spin_unlock(&ubi->wl_lock);
+	mutex_unlock(&ubi->work_mutex);
+
+	if (wrk) {
+		wait_for_completion(&wrk->comp);
+		spin_lock(&ubi->wl_lock);
+		kref_put(&wrk->ref, destroy_work);
+		spin_unlock(&ubi->wl_lock);
+	}
+}
+
+void ubi_wl_resume_work(struct ubi_device *ubi)
+{
+	ubi->thread_suspended = 0;
+	wake_up_process(ubi->bgt_thread);
+}
+
+/**
+ * wl_do_one_work_sync - Run one work in sync.
+ * @ubi: UBI device description object
+ *
+ * This function joins one work and waits for it.
+ * Call it when you run out of free LEBs need to wait for one.
+ * It returns false if no pending work was found to join, true otherwise.
+ */
+static bool wl_do_one_work_sync(struct ubi_device *ubi)
+{
+	struct ubi_work *wrk;
+	bool success = false;
+
+	mutex_lock(&ubi->work_mutex);
+	spin_lock(&ubi->wl_lock);
+	if (ubi->cur_work)
+		wrk = ubi->cur_work;
+	else
+		wrk = list_first_entry_or_null(&ubi->works,
+			struct ubi_work, list);
+
+	if (wrk)
+		kref_get(&wrk->ref);
+	spin_unlock(&ubi->wl_lock);
+	mutex_unlock(&ubi->work_mutex);
+
+	if (wrk) {
+		wait_for_completion(&wrk->comp);
+		if (wrk->ret == 0)
+			success = true;
+
+		spin_lock(&ubi->wl_lock);
+		kref_put(&wrk->ref, destroy_work);
+		spin_unlock(&ubi->wl_lock);
+	}
+
+	return success;
+}
+
 /**
  * in_wl_tree - check if wear-leveling entry is present in a WL RB-tree.
  * @e: the wear-leveling entry to check
@@ -531,21 +629,31 @@ repeat:
 	spin_unlock(&ubi->wl_lock);
 }
 
+static bool wl_work_suspended(struct ubi_device *ubi)
+{
+	return ubi->thread_suspended || !ubi->thread_enabled;
+}
+
 /**
  * __schedule_ubi_work - schedule a work.
  * @ubi: UBI device description object
  * @wrk: the work to schedule
  *
  * This function adds a work defined by @wrk to the tail of the pending works
- * list. Can only be used if ubi->work_sem is already held in read mode!
+ * list. Can only be used if ubi->work_mutex is already held.
  */
 static void __schedule_ubi_work(struct ubi_device *ubi, struct ubi_work *wrk)
 {
+	ubi_assert(ubi->thread_enabled);
+
 	spin_lock(&ubi->wl_lock);
+	INIT_LIST_HEAD(&wrk->list);
+	kref_init(&wrk->ref);
+	init_completion(&wrk->comp);
 	list_add_tail(&wrk->list, &ubi->works);
 	ubi_assert(ubi->works_count >= 0);
 	ubi->works_count += 1;
-	if (ubi->thread_enabled && !ubi_dbg_is_bgt_disabled(ubi))
+	if (!wl_work_suspended(ubi) && !ubi_dbg_is_bgt_disabled(ubi))
 		wake_up_process(ubi->bgt_thread);
 	spin_unlock(&ubi->wl_lock);
 }
@@ -560,9 +668,9 @@ static void __schedule_ubi_work(struct ubi_device *ubi, struct ubi_work *wrk)
  */
 static void schedule_ubi_work(struct ubi_device *ubi, struct ubi_work *wrk)
 {
-	down_read(&ubi->work_sem);
+	mutex_lock(&ubi->work_mutex);
 	__schedule_ubi_work(ubi, wrk);
-	up_read(&ubi->work_sem);
+	mutex_unlock(&ubi->work_mutex);
 }
 
 static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk,
@@ -616,16 +724,20 @@ static int __erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk);
 static int do_sync_erase(struct ubi_device *ubi, struct ubi_wl_entry *e,
 			 int vol_id, int lnum, int torture)
 {
-	struct ubi_work wl_wrk;
+	struct ubi_work *wl_wrk;
 
 	dbg_wl("sync erase of PEB %i", e->pnum);
 
-	wl_wrk.e = e;
-	wl_wrk.vol_id = vol_id;
-	wl_wrk.lnum = lnum;
-	wl_wrk.torture = torture;
+	wl_wrk = ubi_alloc_work(ubi);
+	if (!wl_wrk)
+		return -ENOMEM;
+
+	wl_wrk->e = e;
+	wl_wrk->vol_id = vol_id;
+	wl_wrk->lnum = lnum;
+	wl_wrk->torture = torture;
 
-	return __erase_worker(ubi, &wl_wrk);
+	return __erase_worker(ubi, wl_wrk);
 }
 
 static int ensure_wear_leveling(struct ubi_device *ubi, int nested);
@@ -652,7 +764,6 @@ static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk,
 	struct ubi_vid_hdr *vid_hdr;
 	int dst_leb_clean = 0;
 
-	kfree(wrk);
 	if (shutdown)
 		return 0;
 
@@ -1160,13 +1271,11 @@ static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk,
 		struct ubi_wl_entry *e = wl_wrk->e;
 
 		dbg_wl("cancel erasure of PEB %d EC %d", e->pnum, e->ec);
-		kfree(wl_wrk);
 		wl_entry_destroy(ubi, e);
 		return 0;
 	}
 
 	ret = __erase_worker(ubi, wl_wrk);
-	kfree(wl_wrk);
 	return ret;
 }
 
@@ -1332,43 +1441,33 @@ retry:
  * ubi_wl_flush - flush all pending works.
  * @ubi: UBI device description object
  *
- * This function returns zero in case of success and a negative error code in
- * case of failure.
  */
 int ubi_wl_flush(struct ubi_device *ubi)
 {
-	int err = 0;
+	int ret = 0;
+	struct ubi_work *wrk = NULL;
 
-	/*
-	 * Erase while the pending works queue is not empty, but not more than
-	 * the number of currently pending works.
-	 */
 	dbg_wl("flush (%d pending works)", ubi->works_count);
-	while (ubi->works_count) {
-		err = do_work(ubi);
-		if (err)
-			return err;
-	}
 
-	/*
-	 * Make sure all the works which have been done in parallel are
-	 * finished.
-	 */
-	down_write(&ubi->work_sem);
-	up_write(&ubi->work_sem);
+	/* Find the last entry in the work list and wait for it. */
+	mutex_lock(&ubi->work_mutex);
+	spin_lock(&ubi->wl_lock);
+	if (!list_empty(&ubi->works)) {
+		wrk = list_last_entry(&ubi->works, struct ubi_work, list);
+		kref_get(&wrk->ref);
+	}
+	spin_unlock(&ubi->wl_lock);
+	mutex_unlock(&ubi->work_mutex);
 
-	/*
-	 * And in case last was the WL worker and it canceled the LEB
-	 * movement, flush again.
-	 */
-	while (ubi->works_count) {
-		dbg_wl("flush more (%d pending works)", ubi->works_count);
-		err = do_work(ubi);
-		if (err)
-			return err;
+	if (wrk) {
+		wait_for_completion(&wrk->comp);
+		ret = wrk->ret;
+		spin_lock(&ubi->wl_lock);
+		kref_put(&wrk->ref, destroy_work);
+		spin_unlock(&ubi->wl_lock);
 	}
 
-	return err;
+	return ret;
 }
 
 /**
@@ -1403,6 +1502,24 @@ static void tree_destroy(struct ubi_device *ubi, struct rb_root *root)
 	}
 }
 
+static void __shutdown_work(struct ubi_device *ubi, int error)
+{
+	struct ubi_work *wrk;
+
+	while (!list_empty(&ubi->works)) {
+		wrk = list_entry(ubi->works.next, struct ubi_work, list);
+		list_del(&wrk->list);
+		wrk->func(ubi, wrk, 1);
+		wrk->ret = error;
+		complete_all(&wrk->comp);
+		spin_lock(&ubi->wl_lock);
+		kref_put(&wrk->ref, destroy_work);
+		spin_unlock(&ubi->wl_lock);
+		ubi->works_count -= 1;
+		ubi_assert(ubi->works_count >= 0);
+	}
+}
+
 /**
  * ubi_thread - UBI background thread.
  * @u: the UBI device description object pointer
@@ -1427,7 +1544,7 @@ int ubi_thread(void *u)
 
 		spin_lock(&ubi->wl_lock);
 		if (list_empty(&ubi->works) || ubi->ro_mode ||
-		    !ubi->thread_enabled || ubi_dbg_is_bgt_disabled(ubi)) {
+		    wl_work_suspended(ubi) || ubi_dbg_is_bgt_disabled(ubi)) {
 			set_current_state(TASK_INTERRUPTIBLE);
 			spin_unlock(&ubi->wl_lock);
 			schedule();
@@ -1444,8 +1561,9 @@ int ubi_thread(void *u)
 				 * Too many failures, disable the thread and
 				 * switch to read-only mode.
 				 */
-				ubi_msg(ubi, "%s: %d consecutive failures",
+				ubi_err(ubi, "%s: %d consecutive failures",
 					ubi->bgt_name, WL_MAX_FAILURES);
+				__shutdown_work(ubi, -EROFS);
 				ubi_ro_mode(ubi);
 				ubi->thread_enabled = 0;
 				continue;
@@ -1464,20 +1582,12 @@ int ubi_thread(void *u)
  * shutdown_work - shutdown all pending works.
  * @ubi: UBI device description object
  */
-static void shutdown_work(struct ubi_device *ubi)
+static void shutdown_work(struct ubi_device *ubi, int error)
 {
 #ifdef CONFIG_MTD_UBI_FASTMAP
 	flush_work(&ubi->fm_work);
 #endif
-	while (!list_empty(&ubi->works)) {
-		struct ubi_work *wrk;
-
-		wrk = list_entry(ubi->works.next, struct ubi_work, list);
-		list_del(&wrk->list);
-		wrk->func(ubi, wrk, 1);
-		ubi->works_count -= 1;
-		ubi_assert(ubi->works_count >= 0);
-	}
+	__shutdown_work(ubi, error);
 }
 
 /**
@@ -1499,7 +1609,7 @@ int ubi_wl_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
 	ubi->used = ubi->erroneous = ubi->free = ubi->scrub = RB_ROOT;
 	spin_lock_init(&ubi->wl_lock);
 	mutex_init(&ubi->move_mutex);
-	init_rwsem(&ubi->work_sem);
+	mutex_init(&ubi->work_mutex);
 	ubi->max_ec = ai->max_ec;
 	INIT_LIST_HEAD(&ubi->works);
 
@@ -1615,7 +1725,7 @@ int ubi_wl_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
 	return 0;
 
 out_free:
-	shutdown_work(ubi);
+	shutdown_work(ubi, err);
 	tree_destroy(ubi, &ubi->used);
 	tree_destroy(ubi, &ubi->free);
 	tree_destroy(ubi, &ubi->scrub);
@@ -1648,7 +1758,7 @@ void ubi_wl_close(struct ubi_device *ubi)
 {
 	dbg_wl("close the WL sub-system");
 	ubi_fastmap_close(ubi);
-	shutdown_work(ubi);
+	shutdown_work(ubi, 0);
 	protection_queue_destroy(ubi);
 	tree_destroy(ubi, &ubi->used);
 	tree_destroy(ubi, &ubi->erroneous);
@@ -1777,17 +1887,19 @@ static struct ubi_wl_entry *get_peb_for_wl(struct ubi_device *ubi)
  */
 static int produce_free_peb(struct ubi_device *ubi)
 {
-	int err;
+	ubi_assert(spin_is_locked(&ubi->wl_lock));
 
-	while (!ubi->free.rb_node && ubi->works_count) {
+	while (!ubi->free.rb_node) {
 		spin_unlock(&ubi->wl_lock);
 
 		dbg_wl("do one work synchronously");
-		err = do_work(ubi);
+		if (!wl_do_one_work_sync(ubi)) {
+			spin_lock(&ubi->wl_lock);
+			/* Nothing to do. We have to give up. */
+			return -ENOSPC;
+		}
 
 		spin_lock(&ubi->wl_lock);
-		if (err)
-			return err;
 	}
 
 	return 0;
-- 
2.7.3

[PATCH 03/13] ubi: auto re-size after UBI thread is ready

[Richard Weinberger]

As all work is now done within the UBI thread
re-size cannot performed before UBI thread is ready.

Signed-off-by: Richard Weinberger <richard@nod.at>
---
 drivers/mtd/ubi/build.c | 14 ++++++++------
 1 file changed, 8 insertions(+), 6 deletions(-)

diff --git a/drivers/mtd/ubi/build.c b/drivers/mtd/ubi/build.c
index ee3c833..1448f05 100644
--- a/drivers/mtd/ubi/build.c
+++ b/drivers/mtd/ubi/build.c
@@ -989,12 +989,6 @@ int ubi_attach_mtd_dev(struct mtd_info *mtd, int ubi_num,
 		goto out_free;
 	}
 
-	if (ubi->autoresize_vol_id != -1) {
-		err = autoresize(ubi, ubi->autoresize_vol_id);
-		if (err)
-			goto out_detach;
-	}
-
 	err = uif_init(ubi, &ref);
 	if (err)
 		goto out_detach;
@@ -1039,10 +1033,18 @@ int ubi_attach_mtd_dev(struct mtd_info *mtd, int ubi_num,
 	wake_up_process(ubi->bgt_thread);
 	spin_unlock(&ubi->wl_lock);
 
+	if (ubi->autoresize_vol_id != -1) {
+		err = autoresize(ubi, ubi->autoresize_vol_id);
+		if (err)
+			goto out_kthread;
+	}
+
 	ubi_devices[ubi_num] = ubi;
 	ubi_notify_all(ubi, UBI_VOLUME_ADDED, NULL);
 	return ubi_num;
 
+out_kthread:
+	kthread_stop(ubi->bgt_thread);
 out_debugfs:
 	ubi_debugfs_exit_dev(ubi);
 out_uif:
-- 
2.7.3

[PATCH 04/13] ubi: Kill ubi->alc_mutex

[Richard Weinberger]

With our new work model work can never run in parallel.
So, the serialization mutex is in vain

Signed-off-by: Richard Weinberger <richard@nod.at>
---
 drivers/mtd/ubi/eba.c | 6 ------
 drivers/mtd/ubi/ubi.h | 2 --
 2 files changed, 8 deletions(-)

diff --git a/drivers/mtd/ubi/eba.c b/drivers/mtd/ubi/eba.c
index 5780dd1..32938aa 100644
--- a/drivers/mtd/ubi/eba.c
+++ b/drivers/mtd/ubi/eba.c
@@ -920,9 +920,6 @@ write_error:
  * data, which has to be aligned. This function guarantees that in case of an
  * unclean reboot the old contents is preserved. Returns zero in case of
  * success and a negative error code in case of failure.
- *
- * UBI reserves one LEB for the "atomic LEB change" operation, so only one
- * LEB change may be done at a time. This is ensured by @ubi->alc_mutex.
  */
 int ubi_eba_atomic_leb_change(struct ubi_device *ubi, struct ubi_volume *vol,
 			      int lnum, const void *buf, int len)
@@ -949,7 +946,6 @@ int ubi_eba_atomic_leb_change(struct ubi_device *ubi, struct ubi_volume *vol,
 	if (!vid_hdr)
 		return -ENOMEM;
 
-	mutex_lock(&ubi->alc_mutex);
 	err = leb_write_lock(ubi, vol_id, lnum);
 	if (err)
 		goto out_mutex;
@@ -1006,7 +1002,6 @@ retry:
 out_leb_unlock:
 	leb_write_unlock(ubi, vol_id, lnum);
 out_mutex:
-	mutex_unlock(&ubi->alc_mutex);
 	ubi_free_vid_hdr(ubi, vid_hdr);
 	return err;
 
@@ -1389,7 +1384,6 @@ int ubi_eba_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
 	dbg_eba("initialize EBA sub-system");
 
 	spin_lock_init(&ubi->ltree_lock);
-	mutex_init(&ubi->alc_mutex);
 	ubi->ltree = RB_ROOT;
 
 	ubi->global_sqnum = ai->max_sqnum + 1;
diff --git a/drivers/mtd/ubi/ubi.h b/drivers/mtd/ubi/ubi.h
index 4486869..efdff07 100644
--- a/drivers/mtd/ubi/ubi.h
+++ b/drivers/mtd/ubi/ubi.h
@@ -454,7 +454,6 @@ struct ubi_debug_info {
  * @global_sqnum: global sequence number
  * @ltree_lock: protects the lock tree and @global_sqnum
  * @ltree: the lock tree
- * @alc_mutex: serializes "atomic LEB change" operations
  *
  * @fm_disabled: non-zero if fastmap is disabled (default)
  * @fm: in-memory data structure of the currently used fastmap
@@ -565,7 +564,6 @@ struct ubi_device {
 	unsigned long long global_sqnum;
 	spinlock_t ltree_lock;
 	struct rb_root ltree;
-	struct mutex alc_mutex;
 
 	/* Fastmap stuff */
 	int fm_disabled;
-- 
2.7.3

[PATCH 05/13] ubi: Get rid of __schedule_ubi_work()

[Richard Weinberger]

Now we can call schedule_ubi_work() directly in a worker
function without the need of a nested parameter and the
helper.

Signed-off-by: Richard Weinberger <richard@nod.at>
---
 drivers/mtd/ubi/wl.c | 45 ++++++++++++---------------------------------
 1 file changed, 12 insertions(+), 33 deletions(-)

diff --git a/drivers/mtd/ubi/wl.c b/drivers/mtd/ubi/wl.c
index 8e6e8cf..0ed6e39 100644
--- a/drivers/mtd/ubi/wl.c
+++ b/drivers/mtd/ubi/wl.c
@@ -629,23 +629,19 @@ repeat:
 	spin_unlock(&ubi->wl_lock);
 }
 
-static bool wl_work_suspended(struct ubi_device *ubi)
-{
-	return ubi->thread_suspended || !ubi->thread_enabled;
-}
-
 /**
- * __schedule_ubi_work - schedule a work.
+ * schedule_ubi_work - schedule a work.
  * @ubi: UBI device description object
  * @wrk: the work to schedule
  *
  * This function adds a work defined by @wrk to the tail of the pending works
- * list. Can only be used if ubi->work_mutex is already held.
+ * list.
  */
-static void __schedule_ubi_work(struct ubi_device *ubi, struct ubi_work *wrk)
+static void schedule_ubi_work(struct ubi_device *ubi, struct ubi_work *wrk)
 {
 	ubi_assert(ubi->thread_enabled);
 
+	mutex_lock(&ubi->work_mutex);
 	spin_lock(&ubi->wl_lock);
 	INIT_LIST_HEAD(&wrk->list);
 	kref_init(&wrk->ref);
@@ -656,20 +652,6 @@ static void __schedule_ubi_work(struct ubi_device *ubi, struct ubi_work *wrk)
 	if (!wl_work_suspended(ubi) && !ubi_dbg_is_bgt_disabled(ubi))
 		wake_up_process(ubi->bgt_thread);
 	spin_unlock(&ubi->wl_lock);
-}
-
-/**
- * schedule_ubi_work - schedule a work.
- * @ubi: UBI device description object
- * @wrk: the work to schedule
- *
- * This function adds a work defined by @wrk to the tail of the pending works
- * list.
- */
-static void schedule_ubi_work(struct ubi_device *ubi, struct ubi_work *wrk)
-{
-	mutex_lock(&ubi->work_mutex);
-	__schedule_ubi_work(ubi, wrk);
 	mutex_unlock(&ubi->work_mutex);
 }
 
@@ -740,7 +722,7 @@ static int do_sync_erase(struct ubi_device *ubi, struct ubi_wl_entry *e,
 	return __erase_worker(ubi, wl_wrk);
 }
 
-static int ensure_wear_leveling(struct ubi_device *ubi, int nested);
+static int ensure_wear_leveling(struct ubi_device *ubi);
 /**
  * wear_leveling_worker - wear-leveling worker function.
  * @ubi: UBI device description object
@@ -1026,7 +1008,7 @@ out_not_moved:
 
 	ubi_free_vid_hdr(ubi, vid_hdr);
 	if (dst_leb_clean) {
-		ensure_wear_leveling(ubi, 1);
+		ensure_wear_leveling(ubi);
 	} else {
 		err = do_sync_erase(ubi, e2, vol_id, lnum, torture);
 		if (err)
@@ -1069,13 +1051,12 @@ out_cancel:
 /**
  * ensure_wear_leveling - schedule wear-leveling if it is needed.
  * @ubi: UBI device description object
- * @nested: set to non-zero if this function is called from UBI worker
  *
  * This function checks if it is time to start wear-leveling and schedules it
  * if yes. This function returns zero in case of success and a negative error
  * code in case of failure.
  */
-static int ensure_wear_leveling(struct ubi_device *ubi, int nested)
+static int ensure_wear_leveling(struct ubi_device *ubi)
 {
 	int err = 0;
 	struct ubi_wl_entry *e1;
@@ -1122,10 +1103,8 @@ static int ensure_wear_leveling(struct ubi_device *ubi, int nested)
 
 	wrk->anchor = 0;
 	wrk->func = &wear_leveling_worker;
-	if (nested)
-		__schedule_ubi_work(ubi, wrk);
-	else
-		schedule_ubi_work(ubi, wrk);
+	schedule_ubi_work(ubi, wrk);
+
 	return err;
 
 out_cancel:
@@ -1173,7 +1152,7 @@ static int __erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk)
 		serve_prot_queue(ubi);
 
 		/* And take care about wear-leveling */
-		err = ensure_wear_leveling(ubi, 1);
+		err = ensure_wear_leveling(ubi);
 		return err;
 	}
 
@@ -1434,7 +1413,7 @@ retry:
 	 * Technically scrubbing is the same as wear-leveling, so it is done
 	 * by the WL worker.
 	 */
-	return ensure_wear_leveling(ubi, 0);
+	return ensure_wear_leveling(ubi);
 }
 
 /**
@@ -1718,7 +1697,7 @@ int ubi_wl_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
 	ubi->rsvd_pebs += reserved_pebs;
 
 	/* Schedule wear-leveling if needed */
-	err = ensure_wear_leveling(ubi, 0);
+	err = ensure_wear_leveling(ubi);
 	if (err)
 		goto out_free;
 
-- 
2.7.3

[PATCH 06/13] ubi: Remove tst_disable_bgt debugfs knob

[Richard Weinberger]

With the new worker model we hard depend on the worker
and disabling it will lockup UBI.
So, this knob makes no longer sense.

Signed-off-by: Richard Weinberger <richard@nod.at>
---
 drivers/mtd/ubi/debug.c | 11 -----------
 drivers/mtd/ubi/debug.h | 12 ------------
 drivers/mtd/ubi/ubi.h   |  4 ----
 drivers/mtd/ubi/wl.c    |  4 ++--
 4 files changed, 2 insertions(+), 29 deletions(-)

diff --git a/drivers/mtd/ubi/debug.c b/drivers/mtd/ubi/debug.c
index f101a49..ed23009 100644
--- a/drivers/mtd/ubi/debug.c
+++ b/drivers/mtd/ubi/debug.c
@@ -277,8 +277,6 @@ static ssize_t dfs_file_read(struct file *file, char __user *user_buf,
 		val = d->chk_io;
 	else if (dent == d->dfs_chk_fastmap)
 		val = d->chk_fastmap;
-	else if (dent == d->dfs_disable_bgt)
-		val = d->disable_bgt;
 	else if (dent == d->dfs_emulate_bitflips)
 		val = d->emulate_bitflips;
 	else if (dent == d->dfs_emulate_io_failures)
@@ -372,8 +370,6 @@ static ssize_t dfs_file_write(struct file *file, const char __user *user_buf,
 		d->chk_io = val;
 	else if (dent == d->dfs_chk_fastmap)
 		d->chk_fastmap = val;
-	else if (dent == d->dfs_disable_bgt)
-		d->disable_bgt = val;
 	else if (dent == d->dfs_emulate_bitflips)
 		d->emulate_bitflips = val;
 	else if (dent == d->dfs_emulate_io_failures)
@@ -449,13 +445,6 @@ int ubi_debugfs_init_dev(struct ubi_device *ubi)
 		goto out_remove;
 	d->dfs_chk_fastmap = dent;
 
-	fname = "tst_disable_bgt";
-	dent = debugfs_create_file(fname, S_IWUSR, d->dfs_dir, (void *)ubi_num,
-				   &dfs_fops);
-	if (IS_ERR_OR_NULL(dent))
-		goto out_remove;
-	d->dfs_disable_bgt = dent;
-
 	fname = "tst_emulate_bitflips";
 	dent = debugfs_create_file(fname, S_IWUSR, d->dfs_dir, (void *)ubi_num,
 				   &dfs_fops);
diff --git a/drivers/mtd/ubi/debug.h b/drivers/mtd/ubi/debug.h
index eb8985e..47c447d 100644
--- a/drivers/mtd/ubi/debug.h
+++ b/drivers/mtd/ubi/debug.h
@@ -66,18 +66,6 @@ int ubi_debugfs_init_dev(struct ubi_device *ubi);
 void ubi_debugfs_exit_dev(struct ubi_device *ubi);
 
 /**
- * ubi_dbg_is_bgt_disabled - if the background thread is disabled.
- * @ubi: UBI device description object
- *
- * Returns non-zero if the UBI background thread is disabled for testing
- * purposes.
- */
-static inline int ubi_dbg_is_bgt_disabled(const struct ubi_device *ubi)
-{
-	return ubi->dbg.disable_bgt;
-}
-
-/**
  * ubi_dbg_is_bitflip - if it is time to emulate a bit-flip.
  * @ubi: UBI device description object
  *
diff --git a/drivers/mtd/ubi/ubi.h b/drivers/mtd/ubi/ubi.h
index efdff07..107a9d9 100644
--- a/drivers/mtd/ubi/ubi.h
+++ b/drivers/mtd/ubi/ubi.h
@@ -374,7 +374,6 @@ struct ubi_wl_entry;
  * @chk_gen: if UBI general extra checks are enabled
  * @chk_io: if UBI I/O extra checks are enabled
  * @chk_fastmap: if UBI fastmap extra checks are enabled
- * @disable_bgt: disable the background task for testing purposes
  * @emulate_bitflips: emulate bit-flips for testing purposes
  * @emulate_io_failures: emulate write/erase failures for testing purposes
  * @emulate_power_cut: emulate power cut for testing purposes
@@ -386,7 +385,6 @@ struct ubi_wl_entry;
  * @dfs_chk_gen: debugfs knob to enable UBI general extra checks
  * @dfs_chk_io: debugfs knob to enable UBI I/O extra checks
  * @dfs_chk_fastmap: debugfs knob to enable UBI fastmap extra checks
- * @dfs_disable_bgt: debugfs knob to disable the background task
  * @dfs_emulate_bitflips: debugfs knob to emulate bit-flips
  * @dfs_emulate_io_failures: debugfs knob to emulate write/erase failures
  * @dfs_emulate_power_cut: debugfs knob to emulate power cuts
@@ -397,7 +395,6 @@ struct ubi_debug_info {
 	unsigned int chk_gen:1;
 	unsigned int chk_io:1;
 	unsigned int chk_fastmap:1;
-	unsigned int disable_bgt:1;
 	unsigned int emulate_bitflips:1;
 	unsigned int emulate_io_failures:1;
 	unsigned int emulate_power_cut:2;
@@ -409,7 +406,6 @@ struct ubi_debug_info {
 	struct dentry *dfs_chk_gen;
 	struct dentry *dfs_chk_io;
 	struct dentry *dfs_chk_fastmap;
-	struct dentry *dfs_disable_bgt;
 	struct dentry *dfs_emulate_bitflips;
 	struct dentry *dfs_emulate_io_failures;
 	struct dentry *dfs_emulate_power_cut;
diff --git a/drivers/mtd/ubi/wl.c b/drivers/mtd/ubi/wl.c
index 0ed6e39..a80f019 100644
--- a/drivers/mtd/ubi/wl.c
+++ b/drivers/mtd/ubi/wl.c
@@ -649,7 +649,7 @@ static void schedule_ubi_work(struct ubi_device *ubi, struct ubi_work *wrk)
 	list_add_tail(&wrk->list, &ubi->works);
 	ubi_assert(ubi->works_count >= 0);
 	ubi->works_count += 1;
-	if (!wl_work_suspended(ubi) && !ubi_dbg_is_bgt_disabled(ubi))
+	if (!wl_work_suspended(ubi))
 		wake_up_process(ubi->bgt_thread);
 	spin_unlock(&ubi->wl_lock);
 	mutex_unlock(&ubi->work_mutex);
@@ -1523,7 +1523,7 @@ int ubi_thread(void *u)
 
 		spin_lock(&ubi->wl_lock);
 		if (list_empty(&ubi->works) || ubi->ro_mode ||
-		    wl_work_suspended(ubi) || ubi_dbg_is_bgt_disabled(ubi)) {
+		    wl_work_suspended(ubi)) {
 			set_current_state(TASK_INTERRUPTIBLE);
 			spin_unlock(&ubi->wl_lock);
 			schedule();
-- 
2.7.3

[PATCH 07/13] ubi: Move work related functions to work.c

[Richard Weinberger]

Signed-off-by: Richard Weinberger <richard@nod.at>
---
 drivers/mtd/ubi/Makefile     |   2 +-
 drivers/mtd/ubi/cdev.c       |   2 +-
 drivers/mtd/ubi/fastmap-wl.c |   4 +-
 drivers/mtd/ubi/fastmap.c    |   4 +-
 drivers/mtd/ubi/kapi.c       |   2 +-
 drivers/mtd/ubi/ubi.h        |  16 ++-
 drivers/mtd/ubi/upd.c        |   4 +-
 drivers/mtd/ubi/vmt.c        |   2 +-
 drivers/mtd/ubi/wl.c         | 309 +-----------------------------------------
 drivers/mtd/ubi/work.c       | 316 +++++++++++++++++++++++++++++++++++++++++++
 10 files changed, 343 insertions(+), 318 deletions(-)
 create mode 100644 drivers/mtd/ubi/work.c

diff --git a/drivers/mtd/ubi/Makefile b/drivers/mtd/ubi/Makefile
index 4e3c3d7..e9d4b1d 100644
--- a/drivers/mtd/ubi/Makefile
+++ b/drivers/mtd/ubi/Makefile
@@ -1,7 +1,7 @@
 obj-$(CONFIG_MTD_UBI) += ubi.o
 
 ubi-y += vtbl.o vmt.o upd.o build.o cdev.o kapi.o eba.o io.o wl.o attach.o
-ubi-y += misc.o debug.o
+ubi-y += work.o misc.o debug.o
 ubi-$(CONFIG_MTD_UBI_FASTMAP) += fastmap.o
 ubi-$(CONFIG_MTD_UBI_BLOCK) += block.o
 
diff --git a/drivers/mtd/ubi/cdev.c b/drivers/mtd/ubi/cdev.c
index 7e49de9..833c0a82 100644
--- a/drivers/mtd/ubi/cdev.c
+++ b/drivers/mtd/ubi/cdev.c
@@ -495,7 +495,7 @@ static long vol_cdev_ioctl(struct file *file, unsigned int cmd,
 		if (err)
 			break;
 
-		err = ubi_wl_flush(ubi);
+		err = ubi_work_flush(ubi);
 		break;
 	}
 
diff --git a/drivers/mtd/ubi/fastmap-wl.c b/drivers/mtd/ubi/fastmap-wl.c
index c740095..f6dc426 100644
--- a/drivers/mtd/ubi/fastmap-wl.c
+++ b/drivers/mtd/ubi/fastmap-wl.c
@@ -184,7 +184,7 @@ static int produce_free_peb(struct ubi_device *ubi)
 {
 	while (!ubi->free.rb_node) {
 		dbg_wl("do one work synchronously");
-		if (!wl_do_one_work_sync(ubi)) {
+		if (!ubi_work_join_one(ubi)) {
 			/* Nothing to do. We have to give up. */
 			return -ENOSPC;
 		}
@@ -301,7 +301,7 @@ int ubi_ensure_anchor_pebs(struct ubi_device *ubi)
 
 	wrk->anchor = 1;
 	wrk->func = &wear_leveling_worker;
-	schedule_ubi_work(ubi, wrk);
+	ubi_schedule_work(ubi, wrk);
 	return 0;
 }
 
diff --git a/drivers/mtd/ubi/fastmap.c b/drivers/mtd/ubi/fastmap.c
index 3d5e674..c878313 100644
--- a/drivers/mtd/ubi/fastmap.c
+++ b/drivers/mtd/ubi/fastmap.c
@@ -1607,11 +1607,11 @@ int ubi_update_fastmap(struct ubi_device *ubi)
 		new_fm->e[0] = tmp_e;
 	}
 
-	ubi_wl_suspend_work(ubi);
+	ubi_work_suspend(ubi);
 	down_write(&ubi->fm_eba_sem);
 	ret = ubi_write_fastmap(ubi, new_fm);
 	up_write(&ubi->fm_eba_sem);
-	ubi_wl_resume_work(ubi);
+	ubi_work_resume(ubi);
 
 	if (ret)
 		goto err;
diff --git a/drivers/mtd/ubi/kapi.c b/drivers/mtd/ubi/kapi.c
index 11ca859..dc315c2 100644
--- a/drivers/mtd/ubi/kapi.c
+++ b/drivers/mtd/ubi/kapi.c
@@ -624,7 +624,7 @@ int ubi_leb_erase(struct ubi_volume_desc *desc, int lnum)
 	if (err)
 		return err;
 
-	return ubi_wl_flush(ubi);
+	return ubi_work_flush(ubi);
 }
 EXPORT_SYMBOL_GPL(ubi_leb_erase);
 
diff --git a/drivers/mtd/ubi/ubi.h b/drivers/mtd/ubi/ubi.h
index 107a9d9..0f57786 100644
--- a/drivers/mtd/ubi/ubi.h
+++ b/drivers/mtd/ubi/ubi.h
@@ -859,19 +859,25 @@ int self_check_eba(struct ubi_device *ubi, struct ubi_attach_info *ai_fastmap,
 int ubi_wl_get_peb(struct ubi_device *ubi);
 int ubi_wl_put_peb(struct ubi_device *ubi, int vol_id, int lnum,
 		   int pnum, int torture);
-int ubi_wl_flush(struct ubi_device *ubi);
 int ubi_wl_scrub_peb(struct ubi_device *ubi, int pnum);
 int ubi_wl_init(struct ubi_device *ubi, struct ubi_attach_info *ai);
 void ubi_wl_close(struct ubi_device *ubi);
-int ubi_thread(void *u);
 struct ubi_wl_entry *ubi_wl_get_fm_peb(struct ubi_device *ubi, int anchor);
 int ubi_wl_put_fm_peb(struct ubi_device *ubi, struct ubi_wl_entry *used_e,
 		      int lnum, int torture);
-int ubi_is_erase_work(struct ubi_work *wrk);
 void ubi_refill_pools(struct ubi_device *ubi);
 int ubi_ensure_anchor_pebs(struct ubi_device *ubi);
-void ubi_wl_suspend_work(struct ubi_device *ubi);
-void ubi_wl_resume_work(struct ubi_device *ubi);
+
+/* work.c */
+int ubi_thread(void *u);
+int ubi_is_erase_work(struct ubi_work *wrk);
+void ubi_work_suspend(struct ubi_device *ubi);
+void ubi_work_resume(struct ubi_device *ubi);
+void ubi_schedule_work(struct ubi_device *ubi, struct ubi_work *wrk);
+void ubi_work_close(struct ubi_device *ubi, int error);
+struct ubi_work *ubi_alloc_work(struct ubi_device *ubi);
+int ubi_work_flush(struct ubi_device *ubi);
+bool ubi_work_join_one(struct ubi_device *ubi);
 
 /* io.c */
 int ubi_io_read(const struct ubi_device *ubi, void *buf, int pnum, int offset,
diff --git a/drivers/mtd/ubi/upd.c b/drivers/mtd/ubi/upd.c
index b7901ce..ffaface 100644
--- a/drivers/mtd/ubi/upd.c
+++ b/drivers/mtd/ubi/upd.c
@@ -149,7 +149,7 @@ int ubi_start_update(struct ubi_device *ubi, struct ubi_volume *vol,
 	}
 
 	if (bytes == 0) {
-		err = ubi_wl_flush(ubi);
+		err = ubi_work_flush(ubi);
 		if (err)
 			return err;
 
@@ -361,7 +361,7 @@ int ubi_more_update_data(struct ubi_device *ubi, struct ubi_volume *vol,
 
 	ubi_assert(vol->upd_received <= vol->upd_bytes);
 	if (vol->upd_received == vol->upd_bytes) {
-		err = ubi_wl_flush(ubi);
+		err = ubi_work_flush(ubi);
 		if (err)
 			return err;
 
diff --git a/drivers/mtd/ubi/vmt.c b/drivers/mtd/ubi/vmt.c
index 714e0b3..8a2e081 100644
--- a/drivers/mtd/ubi/vmt.c
+++ b/drivers/mtd/ubi/vmt.c
@@ -237,7 +237,7 @@ int ubi_create_volume(struct ubi_device *ubi, struct ubi_mkvol_req *req)
 	 * Finish all pending erases because there may be some LEBs belonging
 	 * to the same volume ID.
 	 */
-	err = ubi_wl_flush(ubi);
+	err = ubi_work_flush(ubi);
 	if (err)
 		goto out_acc;
 
diff --git a/drivers/mtd/ubi/wl.c b/drivers/mtd/ubi/wl.c
index a80f019..35196c8 100644
--- a/drivers/mtd/ubi/wl.c
+++ b/drivers/mtd/ubi/wl.c
@@ -129,12 +129,6 @@
  */
 #define WL_FREE_MAX_DIFF (2*UBI_WL_THRESHOLD)
 
-/*
- * Maximum number of consecutive background thread failures which is enough to
- * switch to read-only mode.
- */
-#define WL_MAX_FAILURES 32
-
 static int self_check_ec(struct ubi_device *ubi, int pnum, int ec);
 static int self_check_in_wl_tree(const struct ubi_device *ubi,
 				 struct ubi_wl_entry *e, struct rb_root *root);
@@ -192,151 +186,6 @@ static void wl_entry_destroy(struct ubi_device *ubi, struct ubi_wl_entry *e)
 }
 
 /**
- * destroy_work - destroy an UBI work.
- * @ref: kref object
- *
- * This function is called by kref upon the last reference is gone.
- */
-static void destroy_work(struct kref *ref)
-{
-	struct ubi_work *wrk = container_of(ref, struct ubi_work, ref);
-
-	kfree(wrk);
-}
-
-/**
- * wl_work_suspended - Check whether UBI work is suspended.
- * @e: the wear-leveling entry to add
- */
-static bool wl_work_suspended(struct ubi_device *ubi)
-{
-	return ubi->thread_suspended || !ubi->thread_enabled;
-}
-
-/**
- * do_work - do one pending work.
- * @ubi: UBI device description object
- *
- * This function returns zero in case of success and a negative error code in
- * case of failure.
- */
-static int do_work(struct ubi_device *ubi)
-{
-	int err;
-	struct ubi_work *wrk;
-
-	cond_resched();
-
-	mutex_lock(&ubi->work_mutex);
-	spin_lock(&ubi->wl_lock);
-	ubi_assert(!ubi->cur_work);
-	if (list_empty(&ubi->works) || wl_work_suspended(ubi)) {
-		spin_unlock(&ubi->wl_lock);
-		mutex_unlock(&ubi->work_mutex);
-		return 0;
-	}
-
-	wrk = list_entry(ubi->works.next, struct ubi_work, list);
-	list_del(&wrk->list);
-	ubi->works_count -= 1;
-	ubi_assert(ubi->works_count >= 0);
-	ubi->cur_work = wrk;
-	spin_unlock(&ubi->wl_lock);
-	mutex_unlock(&ubi->work_mutex);
-
-	/*
-	 * Call the worker function. Do not touch the work structure
-	 * after this call as it will have been freed or reused by that
-	 * time by the worker function.
-	 */
-	err = wrk->func(ubi, wrk, 0);
-	wrk->ret = err;
-	if (err)
-		ubi_err(ubi, "work failed with error code %d", err);
-
-	spin_lock(&ubi->wl_lock);
-	ubi->cur_work = NULL;
-	spin_unlock(&ubi->wl_lock);
-
-	complete_all(&wrk->comp);
-
-	spin_lock(&ubi->wl_lock);
-	kref_put(&wrk->ref, destroy_work);
-	spin_unlock(&ubi->wl_lock);
-
-	return err;
-}
-
-void ubi_wl_suspend_work(struct ubi_device *ubi)
-{
-	struct ubi_work *wrk = NULL;
-
-	mutex_lock(&ubi->work_mutex);
-	spin_lock(&ubi->wl_lock);
-
-	wrk = ubi->cur_work;
-	if (wrk)
-		kref_get(&wrk->ref);
-
-	ubi->thread_suspended = 1;
-
-	spin_unlock(&ubi->wl_lock);
-	mutex_unlock(&ubi->work_mutex);
-
-	if (wrk) {
-		wait_for_completion(&wrk->comp);
-		spin_lock(&ubi->wl_lock);
-		kref_put(&wrk->ref, destroy_work);
-		spin_unlock(&ubi->wl_lock);
-	}
-}
-
-void ubi_wl_resume_work(struct ubi_device *ubi)
-{
-	ubi->thread_suspended = 0;
-	wake_up_process(ubi->bgt_thread);
-}
-
-/**
- * wl_do_one_work_sync - Run one work in sync.
- * @ubi: UBI device description object
- *
- * This function joins one work and waits for it.
- * Call it when you run out of free LEBs need to wait for one.
- * It returns false if no pending work was found to join, true otherwise.
- */
-static bool wl_do_one_work_sync(struct ubi_device *ubi)
-{
-	struct ubi_work *wrk;
-	bool success = false;
-
-	mutex_lock(&ubi->work_mutex);
-	spin_lock(&ubi->wl_lock);
-	if (ubi->cur_work)
-		wrk = ubi->cur_work;
-	else
-		wrk = list_first_entry_or_null(&ubi->works,
-			struct ubi_work, list);
-
-	if (wrk)
-		kref_get(&wrk->ref);
-	spin_unlock(&ubi->wl_lock);
-	mutex_unlock(&ubi->work_mutex);
-
-	if (wrk) {
-		wait_for_completion(&wrk->comp);
-		if (wrk->ret == 0)
-			success = true;
-
-		spin_lock(&ubi->wl_lock);
-		kref_put(&wrk->ref, destroy_work);
-		spin_unlock(&ubi->wl_lock);
-	}
-
-	return success;
-}
-
-/**
  * in_wl_tree - check if wear-leveling entry is present in a WL RB-tree.
  * @e: the wear-leveling entry to check
  * @root: the root of the tree
@@ -629,32 +478,6 @@ repeat:
 	spin_unlock(&ubi->wl_lock);
 }
 
-/**
- * schedule_ubi_work - schedule a work.
- * @ubi: UBI device description object
- * @wrk: the work to schedule
- *
- * This function adds a work defined by @wrk to the tail of the pending works
- * list.
- */
-static void schedule_ubi_work(struct ubi_device *ubi, struct ubi_work *wrk)
-{
-	ubi_assert(ubi->thread_enabled);
-
-	mutex_lock(&ubi->work_mutex);
-	spin_lock(&ubi->wl_lock);
-	INIT_LIST_HEAD(&wrk->list);
-	kref_init(&wrk->ref);
-	init_completion(&wrk->comp);
-	list_add_tail(&wrk->list, &ubi->works);
-	ubi_assert(ubi->works_count >= 0);
-	ubi->works_count += 1;
-	if (!wl_work_suspended(ubi))
-		wake_up_process(ubi->bgt_thread);
-	spin_unlock(&ubi->wl_lock);
-	mutex_unlock(&ubi->work_mutex);
-}
-
 static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk,
 			int shutdown);
 
@@ -689,7 +512,8 @@ static int schedule_erase(struct ubi_device *ubi, struct ubi_wl_entry *e,
 	wl_wrk->lnum = lnum;
 	wl_wrk->torture = torture;
 
-	schedule_ubi_work(ubi, wl_wrk);
+	ubi_schedule_work(ubi, wl_wrk);
+
 	return 0;
 }
 
@@ -1103,7 +927,7 @@ static int ensure_wear_leveling(struct ubi_device *ubi)
 
 	wrk->anchor = 0;
 	wrk->func = &wear_leveling_worker;
-	schedule_ubi_work(ubi, wrk);
+	ubi_schedule_work(ubi, wrk);
 
 	return err;
 
@@ -1417,39 +1241,6 @@ retry:
 }
 
 /**
- * ubi_wl_flush - flush all pending works.
- * @ubi: UBI device description object
- *
- */
-int ubi_wl_flush(struct ubi_device *ubi)
-{
-	int ret = 0;
-	struct ubi_work *wrk = NULL;
-
-	dbg_wl("flush (%d pending works)", ubi->works_count);
-
-	/* Find the last entry in the work list and wait for it. */
-	mutex_lock(&ubi->work_mutex);
-	spin_lock(&ubi->wl_lock);
-	if (!list_empty(&ubi->works)) {
-		wrk = list_last_entry(&ubi->works, struct ubi_work, list);
-		kref_get(&wrk->ref);
-	}
-	spin_unlock(&ubi->wl_lock);
-	mutex_unlock(&ubi->work_mutex);
-
-	if (wrk) {
-		wait_for_completion(&wrk->comp);
-		ret = wrk->ret;
-		spin_lock(&ubi->wl_lock);
-		kref_put(&wrk->ref, destroy_work);
-		spin_unlock(&ubi->wl_lock);
-	}
-
-	return ret;
-}
-
-/**
  * tree_destroy - destroy an RB-tree.
  * @ubi: UBI device description object
  * @root: the root of the tree to destroy
@@ -1481,94 +1272,6 @@ static void tree_destroy(struct ubi_device *ubi, struct rb_root *root)
 	}
 }
 
-static void __shutdown_work(struct ubi_device *ubi, int error)
-{
-	struct ubi_work *wrk;
-
-	while (!list_empty(&ubi->works)) {
-		wrk = list_entry(ubi->works.next, struct ubi_work, list);
-		list_del(&wrk->list);
-		wrk->func(ubi, wrk, 1);
-		wrk->ret = error;
-		complete_all(&wrk->comp);
-		spin_lock(&ubi->wl_lock);
-		kref_put(&wrk->ref, destroy_work);
-		spin_unlock(&ubi->wl_lock);
-		ubi->works_count -= 1;
-		ubi_assert(ubi->works_count >= 0);
-	}
-}
-
-/**
- * ubi_thread - UBI background thread.
- * @u: the UBI device description object pointer
- */
-int ubi_thread(void *u)
-{
-	int failures = 0;
-	struct ubi_device *ubi = u;
-
-	ubi_msg(ubi, "background thread \"%s\" started, PID %d",
-		ubi->bgt_name, task_pid_nr(current));
-
-	set_freezable();
-	for (;;) {
-		int err;
-
-		if (kthread_should_stop())
-			break;
-
-		if (try_to_freeze())
-			continue;
-
-		spin_lock(&ubi->wl_lock);
-		if (list_empty(&ubi->works) || ubi->ro_mode ||
-		    wl_work_suspended(ubi)) {
-			set_current_state(TASK_INTERRUPTIBLE);
-			spin_unlock(&ubi->wl_lock);
-			schedule();
-			continue;
-		}
-		spin_unlock(&ubi->wl_lock);
-
-		err = do_work(ubi);
-		if (err) {
-			ubi_err(ubi, "%s: work failed with error code %d",
-				ubi->bgt_name, err);
-			if (failures++ > WL_MAX_FAILURES) {
-				/*
-				 * Too many failures, disable the thread and
-				 * switch to read-only mode.
-				 */
-				ubi_err(ubi, "%s: %d consecutive failures",
-					ubi->bgt_name, WL_MAX_FAILURES);
-				__shutdown_work(ubi, -EROFS);
-				ubi_ro_mode(ubi);
-				ubi->thread_enabled = 0;
-				continue;
-			}
-		} else
-			failures = 0;
-
-		cond_resched();
-	}
-
-	dbg_wl("background thread \"%s\" is killed", ubi->bgt_name);
-	return 0;
-}
-
-/**
- * shutdown_work - shutdown all pending works.
- * @ubi: UBI device description object
- */
-static void shutdown_work(struct ubi_device *ubi, int error)
-{
-#ifdef CONFIG_MTD_UBI_FASTMAP
-	flush_work(&ubi->fm_work);
-#endif
-	__shutdown_work(ubi, error);
-}
-
 /**
  * ubi_wl_init - initialize the WL sub-system using attaching information.
  * @ubi: UBI device description object
@@ -1704,7 +1407,7 @@ int ubi_wl_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
 	return 0;
 
 out_free:
-	shutdown_work(ubi, err);
+	ubi_work_close(ubi, err);
 	tree_destroy(ubi, &ubi->used);
 	tree_destroy(ubi, &ubi->free);
 	tree_destroy(ubi, &ubi->scrub);
@@ -1737,7 +1440,7 @@ void ubi_wl_close(struct ubi_device *ubi)
 {
 	dbg_wl("close the WL sub-system");
 	ubi_fastmap_close(ubi);
-	shutdown_work(ubi, 0);
+	ubi_work_close(ubi, 0);
 	protection_queue_destroy(ubi);
 	tree_destroy(ubi, &ubi->used);
 	tree_destroy(ubi, &ubi->erroneous);
@@ -1872,7 +1575,7 @@ static int produce_free_peb(struct ubi_device *ubi)
 		spin_unlock(&ubi->wl_lock);
 
 		dbg_wl("do one work synchronously");
-		if (!wl_do_one_work_sync(ubi)) {
+		if (!ubi_work_join_one(ubi)) {
 			spin_lock(&ubi->wl_lock);
 			/* Nothing to do. We have to give up. */
 			return -ENOSPC;
diff --git a/drivers/mtd/ubi/work.c b/drivers/mtd/ubi/work.c
new file mode 100644
index 0000000..d2f8ba8
--- /dev/null
+++ b/drivers/mtd/ubi/work.c
@@ -0,0 +1,316 @@
+#include <linux/slab.h>
+#include <linux/crc32.h>
+#include <linux/freezer.h>
+#include <linux/kthread.h>
+#include "ubi.h"
+
+/*
+ * Maximum number of consecutive background thread failures which is enough to
+ * switch to read-only mode.
+ */
+#define WORK_MAX_FAILURES 32
+
+/**
+ * work_suspended - Check whether UBI work is suspended.
+ */
+static bool work_suspended(struct ubi_device *ubi)
+{
+	return ubi->thread_suspended || !ubi->thread_enabled;
+}
+
+
+/**
+ * ubi_schedule_work - schedule a work.
+ * @ubi: UBI device description object
+ * @wrk: the work to schedule
+ *
+ * This function adds a work defined by @wrk to the tail of the pending works
+ * list.
+ */
+void ubi_schedule_work(struct ubi_device *ubi, struct ubi_work *wrk)
+{
+	ubi_assert(ubi->thread_enabled);
+
+	mutex_lock(&ubi->work_mutex);
+	spin_lock(&ubi->wl_lock);
+	list_add_tail(&wrk->list, &ubi->works);
+	ubi_assert(ubi->works_count >= 0);
+	ubi->works_count += 1;
+	if (!work_suspended(ubi))
+		wake_up_process(ubi->bgt_thread);
+	spin_unlock(&ubi->wl_lock);
+	mutex_unlock(&ubi->work_mutex);
+}
+
+struct ubi_work *ubi_alloc_work(struct ubi_device *ubi)
+{
+	struct ubi_work *wrk;
+
+	wrk = kzalloc(sizeof(*wrk), GFP_NOFS);
+	if (!wrk)
+		return NULL;
+
+	INIT_LIST_HEAD(&wrk->list);
+	kref_init(&wrk->ref);
+	init_completion(&wrk->comp);
+
+	return wrk;
+}
+
+/**
+ * destroy_work - destroy an UBI work.
+ * @ref: kref object
+ *
+ * This function is called by kref upon the last reference is gone.
+ */
+static void destroy_work(struct kref *ref)
+{
+	struct ubi_work *wrk = container_of(ref, struct ubi_work, ref);
+
+	kfree(wrk);
+}
+
+static void shutdown_work(struct ubi_device *ubi, int error)
+{
+	struct ubi_work *wrk;
+
+	while (!list_empty(&ubi->works)) {
+		wrk = list_entry(ubi->works.next, struct ubi_work, list);
+		list_del_init(&wrk->list);
+		wrk->func(ubi, wrk, 1);
+		wrk->ret = error;
+		complete_all(&wrk->comp);
+		spin_lock(&ubi->wl_lock);
+		kref_put(&wrk->ref, destroy_work);
+		spin_unlock(&ubi->wl_lock);
+		ubi->works_count -= 1;
+		ubi_assert(ubi->works_count >= 0);
+	}
+}
+
+/**
+ * ubi_work_close - shutdown all pending works.
+ * @ubi: UBI device description object
+ */
+void ubi_work_close(struct ubi_device *ubi, int error)
+{
+#ifdef CONFIG_MTD_UBI_FASTMAP
+	flush_work(&ubi->fm_work);
+#endif
+	shutdown_work(ubi, error);
+}
+
+/**
+ * do_work - do one pending work.
+ * @ubi: UBI device description object
+ *
+ * This function returns zero in case of success and a negative error code in
+ * case of failure.
+ */
+static int do_work(struct ubi_device *ubi)
+{
+	int err;
+	struct ubi_work *wrk;
+
+	cond_resched();
+
+	mutex_lock(&ubi->work_mutex);
+	spin_lock(&ubi->wl_lock);
+	ubi_assert(!ubi->cur_work);
+	if (list_empty(&ubi->works) || work_suspended(ubi)) {
+		spin_unlock(&ubi->wl_lock);
+		mutex_unlock(&ubi->work_mutex);
+		return 0;
+	}
+
+	wrk = list_entry(ubi->works.next, struct ubi_work, list);
+	list_del_init(&wrk->list);
+	ubi->works_count -= 1;
+	ubi_assert(ubi->works_count >= 0);
+	ubi->cur_work = wrk;
+	spin_unlock(&ubi->wl_lock);
+	mutex_unlock(&ubi->work_mutex);
+
+	/*
+	 * Call the worker function. Do not touch the work structure
+	 * after this call as it will have been freed or reused by that
+	 * time by the worker function.
+	 */
+	err = wrk->func(ubi, wrk, 0);
+	wrk->ret = err;
+	if (err)
+		ubi_err(ubi, "work failed with error code %d", err);
+
+	spin_lock(&ubi->wl_lock);
+	ubi->cur_work = NULL;
+	spin_unlock(&ubi->wl_lock);
+
+	complete_all(&wrk->comp);
+
+	spin_lock(&ubi->wl_lock);
+	kref_put(&wrk->ref, destroy_work);
+	spin_unlock(&ubi->wl_lock);
+
+	return err;
+}
+
+void ubi_work_suspend(struct ubi_device *ubi)
+{
+	struct ubi_work *wrk = NULL;
+
+	mutex_lock(&ubi->work_mutex);
+	spin_lock(&ubi->wl_lock);
+
+	wrk = ubi->cur_work;
+	if (wrk)
+		kref_get(&wrk->ref);
+
+	ubi->thread_suspended = 1;
+
+	spin_unlock(&ubi->wl_lock);
+	mutex_unlock(&ubi->work_mutex);
+
+	if (wrk) {
+		wait_for_completion(&wrk->comp);
+		spin_lock(&ubi->wl_lock);
+		kref_put(&wrk->ref, destroy_work);
+		spin_unlock(&ubi->wl_lock);
+	}
+}
+
+void ubi_work_resume(struct ubi_device *ubi)
+{
+	ubi->thread_suspended = 0;
+	wake_up_process(ubi->bgt_thread);
+}
+
+/**
+ * ubi_work_join_one - Run one work in sync.
+ * @ubi: UBI device description object
+ *
+ * This function joins one work and waits for it.
+ * Call it when you run out of free LEBs need to wait for one.
+ * It returns false if no pending work was found to join, true otherwise.
+ */
+bool ubi_work_join_one(struct ubi_device *ubi)
+{
+	struct ubi_work *wrk;
+	bool success = false;
+
+	mutex_lock(&ubi->work_mutex);
+	spin_lock(&ubi->wl_lock);
+	if (ubi->cur_work)
+		wrk = ubi->cur_work;
+	else
+		wrk = list_first_entry_or_null(&ubi->works,
+			struct ubi_work, list);
+
+	if (wrk)
+		kref_get(&wrk->ref);
+	spin_unlock(&ubi->wl_lock);
+	mutex_unlock(&ubi->work_mutex);
+
+	if (wrk) {
+		wait_for_completion(&wrk->comp);
+		if (wrk->ret == 0)
+			success = true;
+
+		spin_lock(&ubi->wl_lock);
+		kref_put(&wrk->ref, destroy_work);
+		spin_unlock(&ubi->wl_lock);
+	}
+
+	return success;
+}
+
+/**
+ * ubi_work_flush - flush all pending works.
+ * @ubi: UBI device description object
+ *
+ */
+int ubi_work_flush(struct ubi_device *ubi)
+{
+	int ret = 0;
+	struct ubi_work *wrk = NULL;
+
+	dbg_wl("flush (%d pending works)", ubi->works_count);
+
+	/* Find the last entry in the work list and wait for it. */
+	mutex_lock(&ubi->work_mutex);
+	spin_lock(&ubi->wl_lock);
+	if (!list_empty(&ubi->works)) {
+		wrk = list_last_entry(&ubi->works, struct ubi_work, list);
+		kref_get(&wrk->ref);
+	}
+	spin_unlock(&ubi->wl_lock);
+	mutex_unlock(&ubi->work_mutex);
+
+	if (wrk) {
+		wait_for_completion(&wrk->comp);
+		ret = wrk->ret;
+		spin_lock(&ubi->wl_lock);
+		kref_put(&wrk->ref, destroy_work);
+		spin_unlock(&ubi->wl_lock);
+	}
+
+	return ret;
+}
+
+/**
+ * ubi_thread - UBI background thread.
+ * @u: the UBI device description object pointer
+ */
+int ubi_thread(void *u)
+{
+	int failures = 0;
+	struct ubi_device *ubi = u;
+
+	ubi_msg(ubi, "background thread \"%s\" started, PID %d",
+		ubi->bgt_name, task_pid_nr(current));
+
+	set_freezable();
+	for (;;) {
+		int err;
+
+		if (kthread_should_stop())
+			break;
+
+		if (try_to_freeze())
+			continue;
+
+		spin_lock(&ubi->wl_lock);
+		if (list_empty(&ubi->works) || ubi->ro_mode ||
+		    work_suspended(ubi)) {
+			set_current_state(TASK_INTERRUPTIBLE);
+			spin_unlock(&ubi->wl_lock);
+			schedule();
+			continue;
+		}
+		spin_unlock(&ubi->wl_lock);
+
+		err = do_work(ubi);
+		if (err) {
+			ubi_err(ubi, "%s: work failed with error code %d",
+				ubi->bgt_name, err);
+			if (failures++ > WORK_MAX_FAILURES) {
+				/*
+				 * Too many failures, disable the thread and
+				 * switch to read-only mode.
+				 */
+				ubi_err(ubi, "%s: %d consecutive failures",
+					ubi->bgt_name, WORK_MAX_FAILURES);
+				shutdown_work(ubi, -EROFS);
+				ubi_ro_mode(ubi);
+				ubi->thread_enabled = 0;
+				continue;
+			}
+		} else
+			failures = 0;
+
+		cond_resched();
+	}
+
+	dbg_wl("background thread \"%s\" is killed", ubi->bgt_name);
+	return 0;
+}
+
-- 
2.7.3

[PATCH 08/13] ubi: Remove lnum and vol_id from erase work

[Richard Weinberger]

This has no value except more or less useful debug
output. And with MLC NAND a PEB can carry more than
one lnum/vol_id.

Signed-off-by: Richard Weinberger <richard@nod.at>
---
 drivers/mtd/ubi/eba.c        | 16 ++++++-------
 drivers/mtd/ubi/fastmap-wl.c |  2 +-
 drivers/mtd/ubi/ubi.h        |  5 +---
 drivers/mtd/ubi/wl.c         | 55 ++++++++++++++++++++++++--------------------
 4 files changed, 40 insertions(+), 38 deletions(-)

diff --git a/drivers/mtd/ubi/eba.c b/drivers/mtd/ubi/eba.c
index 32938aa..229be7c 100644
--- a/drivers/mtd/ubi/eba.c
+++ b/drivers/mtd/ubi/eba.c
@@ -343,7 +343,7 @@ int ubi_eba_unmap_leb(struct ubi_device *ubi, struct ubi_volume *vol,
 	down_read(&ubi->fm_eba_sem);
 	vol->eba_tbl[lnum] = UBI_LEB_UNMAPPED;
 	up_read(&ubi->fm_eba_sem);
-	err = ubi_wl_put_peb(ubi, vol_id, lnum, pnum, 0);
+	err = ubi_wl_put_peb(ubi, pnum, 0);
 
 out_unlock:
 	leb_write_unlock(ubi, vol_id, lnum);
@@ -633,7 +633,7 @@ retry:
 
 	vol->eba_tbl[lnum] = new_pnum;
 	up_read(&ubi->fm_eba_sem);
-	ubi_wl_put_peb(ubi, vol_id, lnum, pnum, 1);
+	ubi_wl_put_peb(ubi, vol_id, 1);
 
 	ubi_msg(ubi, "data was successfully recovered");
 	return 0;
@@ -641,7 +641,7 @@ retry:
 out_unlock:
 	mutex_unlock(&ubi->buf_mutex);
 out_put:
-	ubi_wl_put_peb(ubi, vol_id, lnum, new_pnum, 1);
+	ubi_wl_put_peb(ubi, new_pnum, 1);
 	ubi_free_vid_hdr(ubi, vid_hdr);
 	return err;
 
@@ -651,7 +651,7 @@ write_error:
 	 * get another one.
 	 */
 	ubi_warn(ubi, "failed to write to PEB %d", new_pnum);
-	ubi_wl_put_peb(ubi, vol_id, lnum, new_pnum, 1);
+	ubi_wl_put_peb(ubi, new_pnum, 1);
 	if (++tries > UBI_IO_RETRIES) {
 		ubi_free_vid_hdr(ubi, vid_hdr);
 		return err;
@@ -772,7 +772,7 @@ write_error:
 	 * eraseblock, so just put it and request a new one. We assume that if
 	 * this physical eraseblock went bad, the erase code will handle that.
 	 */
-	err = ubi_wl_put_peb(ubi, vol_id, lnum, pnum, 1);
+	err = ubi_wl_put_peb(ubi, pnum, 1);
 	if (err || ++tries > UBI_IO_RETRIES) {
 		ubi_ro_mode(ubi);
 		leb_write_unlock(ubi, vol_id, lnum);
@@ -894,7 +894,7 @@ write_error:
 		return err;
 	}
 
-	err = ubi_wl_put_peb(ubi, vol_id, lnum, pnum, 1);
+	err = ubi_wl_put_peb(ubi, pnum, 1);
 	if (err || ++tries > UBI_IO_RETRIES) {
 		ubi_ro_mode(ubi);
 		leb_write_unlock(ubi, vol_id, lnum);
@@ -994,7 +994,7 @@ retry:
 	up_read(&ubi->fm_eba_sem);
 
 	if (old_pnum >= 0) {
-		err = ubi_wl_put_peb(ubi, vol_id, lnum, old_pnum, 0);
+		err = ubi_wl_put_peb(ubi, old_pnum, 0);
 		if (err)
 			goto out_leb_unlock;
 	}
@@ -1016,7 +1016,7 @@ write_error:
 		goto out_leb_unlock;
 	}
 
-	err = ubi_wl_put_peb(ubi, vol_id, lnum, pnum, 1);
+	err = ubi_wl_put_peb(ubi, pnum, 1);
 	if (err || ++tries > UBI_IO_RETRIES) {
 		ubi_ro_mode(ubi);
 		goto out_leb_unlock;
diff --git a/drivers/mtd/ubi/fastmap-wl.c b/drivers/mtd/ubi/fastmap-wl.c
index f6dc426..cafa7b0 100644
--- a/drivers/mtd/ubi/fastmap-wl.c
+++ b/drivers/mtd/ubi/fastmap-wl.c
@@ -342,7 +342,7 @@ int ubi_wl_put_fm_peb(struct ubi_device *ubi, struct ubi_wl_entry *fm_e,
 	spin_unlock(&ubi->wl_lock);
 
 	vol_id = lnum ? UBI_FM_DATA_VOLUME_ID : UBI_FM_SB_VOLUME_ID;
-	return schedule_erase(ubi, e, vol_id, lnum, torture);
+	return schedule_erase(ubi, e, torture);
 }
 
 /**
diff --git a/drivers/mtd/ubi/ubi.h b/drivers/mtd/ubi/ubi.h
index 0f57786..c901cc0 100644
--- a/drivers/mtd/ubi/ubi.h
+++ b/drivers/mtd/ubi/ubi.h
@@ -774,8 +774,6 @@ struct ubi_work {
 	struct kref ref;
 	/* The below fields are only relevant to erasure works */
 	struct ubi_wl_entry *e;
-	int vol_id;
-	int lnum;
 	int torture;
 	int anchor;
 };
@@ -857,8 +855,7 @@ int self_check_eba(struct ubi_device *ubi, struct ubi_attach_info *ai_fastmap,
 
 /* wl.c */
 int ubi_wl_get_peb(struct ubi_device *ubi);
-int ubi_wl_put_peb(struct ubi_device *ubi, int vol_id, int lnum,
-		   int pnum, int torture);
+int ubi_wl_put_peb(struct ubi_device *ubi, int pnum, int torture);
 int ubi_wl_scrub_peb(struct ubi_device *ubi, int pnum);
 int ubi_wl_init(struct ubi_device *ubi, struct ubi_attach_info *ai);
 void ubi_wl_close(struct ubi_device *ubi);
diff --git a/drivers/mtd/ubi/wl.c b/drivers/mtd/ubi/wl.c
index 35196c8..7d42739 100644
--- a/drivers/mtd/ubi/wl.c
+++ b/drivers/mtd/ubi/wl.c
@@ -481,19 +481,36 @@ repeat:
 static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk,
 			int shutdown);
 
+static struct ubi_work *ubi_alloc_erase_work(struct ubi_device *ubi,
+					     struct ubi_wl_entry *e,
+					     int torture)
+{
+	struct ubi_work *wl_wrk;
+
+	ubi_assert(e);
+
+	wl_wrk = ubi_alloc_work(ubi);
+	if (!wl_wrk)
+		return NULL;
+
+	wl_wrk->func = &erase_worker;
+	wl_wrk->e = e;
+	wl_wrk->torture = torture;
+
+	return wl_wrk;
+}
+
 /**
  * schedule_erase - schedule an erase work.
  * @ubi: UBI device description object
  * @e: the WL entry of the physical eraseblock to erase
- * @vol_id: the volume ID that last used this PEB
- * @lnum: the last used logical eraseblock number for the PEB
  * @torture: if the physical eraseblock has to be tortured
  *
  * This function returns zero in case of success and a %-ENOMEM in case of
  * failure.
  */
 static int schedule_erase(struct ubi_device *ubi, struct ubi_wl_entry *e,
-			  int vol_id, int lnum, int torture)
+			  int torture)
 {
 	struct ubi_work *wl_wrk;
 
@@ -502,14 +519,12 @@ static int schedule_erase(struct ubi_device *ubi, struct ubi_wl_entry *e,
 	dbg_wl("schedule erasure of PEB %d, EC %d, torture %d",
 	       e->pnum, e->ec, torture);
 
-	wl_wrk = kmalloc(sizeof(struct ubi_work), GFP_NOFS);
+	wl_wrk = ubi_alloc_erase_work(ubi, e, torture);
 	if (!wl_wrk)
 		return -ENOMEM;
 
 	wl_wrk->func = &erase_worker;
 	wl_wrk->e = e;
-	wl_wrk->vol_id = vol_id;
-	wl_wrk->lnum = lnum;
 	wl_wrk->torture = torture;
 
 	ubi_schedule_work(ubi, wl_wrk);
@@ -522,13 +537,11 @@ static int __erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk);
  * do_sync_erase - run the erase worker synchronously.
  * @ubi: UBI device description object
  * @e: the WL entry of the physical eraseblock to erase
- * @vol_id: the volume ID that last used this PEB
- * @lnum: the last used logical eraseblock number for the PEB
  * @torture: if the physical eraseblock has to be tortured
  *
  */
 static int do_sync_erase(struct ubi_device *ubi, struct ubi_wl_entry *e,
-			 int vol_id, int lnum, int torture)
+			 int torture)
 {
 	struct ubi_work *wl_wrk;
 
@@ -539,8 +552,6 @@ static int do_sync_erase(struct ubi_device *ubi, struct ubi_wl_entry *e,
 		return -ENOMEM;
 
 	wl_wrk->e = e;
-	wl_wrk->vol_id = vol_id;
-	wl_wrk->lnum = lnum;
 	wl_wrk->torture = torture;
 
 	return __erase_worker(ubi, wl_wrk);
@@ -775,7 +786,7 @@ static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk,
 	ubi->move_to_put = ubi->wl_scheduled = 0;
 	spin_unlock(&ubi->wl_lock);
 
-	err = do_sync_erase(ubi, e1, vol_id, lnum, 0);
+	err = do_sync_erase(ubi, e1, 0);
 	if (err) {
 		if (e2)
 			wl_entry_destroy(ubi, e2);
@@ -789,7 +800,7 @@ static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk,
 		 */
 		dbg_wl("PEB %d (LEB %d:%d) was put meanwhile, erase",
 		       e2->pnum, vol_id, lnum);
-		err = do_sync_erase(ubi, e2, vol_id, lnum, 0);
+		err = do_sync_erase(ubi, e2, 0);
 		if (err)
 			goto out_ro;
 	}
@@ -834,7 +845,7 @@ out_not_moved:
 	if (dst_leb_clean) {
 		ensure_wear_leveling(ubi);
 	} else {
-		err = do_sync_erase(ubi, e2, vol_id, lnum, torture);
+		err = do_sync_erase(ubi, e2, torture);
 		if (err)
 			goto out_ro;
 	}
@@ -955,12 +966,9 @@ static int __erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk)
 {
 	struct ubi_wl_entry *e = wl_wrk->e;
 	int pnum = e->pnum;
-	int vol_id = wl_wrk->vol_id;
-	int lnum = wl_wrk->lnum;
 	int err, available_consumed = 0;
 
-	dbg_wl("erase PEB %d EC %d LEB %d:%d",
-	       pnum, e->ec, wl_wrk->vol_id, wl_wrk->lnum);
+	dbg_wl("erase PEB %d EC %d", pnum, e->ec);
 
 	err = sync_erase(ubi, e, wl_wrk->torture);
 	if (!err) {
@@ -987,7 +995,7 @@ static int __erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk)
 		int err1;
 
 		/* Re-schedule the LEB for erasure */
-		err1 = schedule_erase(ubi, e, vol_id, lnum, 0);
+		err1 = schedule_erase(ubi, e, 0);
 		if (err1) {
 			wl_entry_destroy(ubi, e);
 			err = err1;
@@ -1085,8 +1093,6 @@ static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk,
 /**
  * ubi_wl_put_peb - return a PEB to the wear-leveling sub-system.
  * @ubi: UBI device description object
- * @vol_id: the volume ID that last used this PEB
- * @lnum: the last used logical eraseblock number for the PEB
  * @pnum: physical eraseblock to return
  * @torture: if this physical eraseblock has to be tortured
  *
@@ -1095,8 +1101,7 @@ static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk,
  * occurred to this @pnum and it has to be tested. This function returns zero
  * in case of success, and a negative error code in case of failure.
  */
-int ubi_wl_put_peb(struct ubi_device *ubi, int vol_id, int lnum,
-		   int pnum, int torture)
+int ubi_wl_put_peb(struct ubi_device *ubi, int pnum, int torture)
 {
 	int err;
 	struct ubi_wl_entry *e;
@@ -1166,7 +1171,7 @@ retry:
 	}
 	spin_unlock(&ubi->wl_lock);
 
-	err = schedule_erase(ubi, e, vol_id, lnum, torture);
+	err = schedule_erase(ubi, e, torture);
 	if (err) {
 		spin_lock(&ubi->wl_lock);
 		wl_tree_add(e, &ubi->used);
@@ -1317,7 +1322,7 @@ int ubi_wl_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
 		e->pnum = aeb->pnum;
 		e->ec = aeb->ec;
 		ubi->lookuptbl[e->pnum] = e;
-		if (schedule_erase(ubi, e, aeb->vol_id, aeb->lnum, 0)) {
+		if (schedule_erase(ubi, e, 0)) {
 			wl_entry_destroy(ubi, e);
 			goto out_free;
 		}
-- 
2.7.3

[PATCH 09/13] ubi: Remove usless debug info from wear_leveling_worker()

[Richard Weinberger]

Signed-off-by: Richard Weinberger <richard@nod.at>
---
 drivers/mtd/ubi/wl.c | 27 +++++++--------------------
 1 file changed, 7 insertions(+), 20 deletions(-)

diff --git a/drivers/mtd/ubi/wl.c b/drivers/mtd/ubi/wl.c
index 7d42739..bf4e6b2 100644
--- a/drivers/mtd/ubi/wl.c
+++ b/drivers/mtd/ubi/wl.c
@@ -573,7 +573,6 @@ static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk,
 				int shutdown)
 {
 	int err, scrubbing = 0, torture = 0, protect = 0, erroneous = 0;
-	int vol_id = -1, lnum = -1;
 #ifdef CONFIG_MTD_UBI_FASTMAP
 	int anchor = wrk->anchor;
 #endif
@@ -715,9 +714,6 @@ static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk,
 		goto out_error;
 	}
 
-	vol_id = be32_to_cpu(vid_hdr->vol_id);
-	lnum = be32_to_cpu(vid_hdr->lnum);
-
 	err = ubi_eba_copy_leb(ubi, e1->pnum, e2->pnum, vid_hdr);
 	if (err) {
 		if (err == MOVE_CANCEL_RACE) {
@@ -773,8 +769,8 @@ static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk,
 
 	/* The PEB has been successfully moved */
 	if (scrubbing)
-		ubi_msg(ubi, "scrubbed PEB %d (LEB %d:%d), data moved to PEB %d",
-			e1->pnum, vol_id, lnum, e2->pnum);
+		ubi_msg(ubi, "scrubbed PEB %d, data moved to PEB %d",
+			e1->pnum, e2->pnum);
 	ubi_free_vid_hdr(ubi, vid_hdr);
 
 	spin_lock(&ubi->wl_lock);
@@ -798,8 +794,8 @@ static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk,
 		 * Well, the target PEB was put meanwhile, schedule it for
 		 * erasure.
 		 */
-		dbg_wl("PEB %d (LEB %d:%d) was put meanwhile, erase",
-		       e2->pnum, vol_id, lnum);
+		dbg_wl("PEB %d was put meanwhile, erase",
+		       e2->pnum);
 		err = do_sync_erase(ubi, e2, 0);
 		if (err)
 			goto out_ro;
@@ -815,12 +811,7 @@ static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk,
 	 * have been changed, schedule it for erasure.
 	 */
 out_not_moved:
-	if (vol_id != -1)
-		dbg_wl("cancel moving PEB %d (LEB %d:%d) to PEB %d (%d)",
-		       e1->pnum, vol_id, lnum, e2->pnum, err);
-	else
-		dbg_wl("cancel moving PEB %d to PEB %d (%d)",
-		       e1->pnum, e2->pnum, err);
+	dbg_wl("cancel moving PEB %d to PEB %d (%d)", e1->pnum, e2->pnum, err);
 	spin_lock(&ubi->wl_lock);
 	if (protect)
 		prot_queue_add(ubi, e1);
@@ -854,12 +845,8 @@ out_not_moved:
 	return 0;
 
 out_error:
-	if (vol_id != -1)
-		ubi_err(ubi, "error %d while moving PEB %d to PEB %d",
-			err, e1->pnum, e2->pnum);
-	else
-		ubi_err(ubi, "error %d while moving PEB %d (LEB %d:%d) to PEB %d",
-			err, e1->pnum, vol_id, lnum, e2->pnum);
+	ubi_err(ubi, "error %d while moving PEB %d to PEB %d", err, e1->pnum,
+		e2->pnum);
 	spin_lock(&ubi->wl_lock);
 	ubi->move_from = ubi->move_to = NULL;
 	ubi->move_to_put = ubi->wl_scheduled = 0;
-- 
2.7.3

[PATCH 10/13] ubi: SLC mode

[Richard Weinberger]

Implements a software SLC mode emulation such that we can write
MLC chips in SLC mode.
As only lower pages are written a power cut will not corrupt data.
This patch was implemented by Boris Brezillon and me.

TODO: some self checks are disabled because vmalloc()ing 2MiB+ (MLC)
will OOM the kernel soon.

Signed-off-by: Richard Weinberger <richard@nod.at>
---
 drivers/mtd/ubi/build.c |   8 +-
 drivers/mtd/ubi/io.c    | 426 ++++++++++++++++++++++++++++++++++++------------
 drivers/mtd/ubi/ubi.h   |  10 ++
 3 files changed, 340 insertions(+), 104 deletions(-)

diff --git a/drivers/mtd/ubi/build.c b/drivers/mtd/ubi/build.c
index 1448f05..99e31ed 100644
--- a/drivers/mtd/ubi/build.c
+++ b/drivers/mtd/ubi/build.c
@@ -650,7 +650,10 @@ static int io_init(struct ubi_device *ubi, int max_beb_per1024)
 	 * physical eraseblocks maximum.
 	 */
 
-	ubi->peb_size   = ubi->mtd->erasesize;
+	ubi->consolidated_peb_size = ubi->mtd->erasesize;
+	ubi->peb_size   = ubi->consolidated_peb_size /
+			  mtd_pairing_groups_per_eb(ubi->mtd);
+	ubi->lebs_per_consolidated_peb = mtd_pairing_groups_per_eb(ubi->mtd);
 	ubi->peb_count  = mtd_div_by_eb(ubi->mtd->size, ubi->mtd);
 	ubi->flash_size = ubi->mtd->size;
 
@@ -969,7 +972,7 @@ int ubi_attach_mtd_dev(struct mtd_info *mtd, int ubi_num,
 		goto out_free;
 
 	err = -ENOMEM;
-	ubi->peb_buf = vmalloc(ubi->peb_size);
+	ubi->peb_buf = vmalloc(ubi->consolidated_peb_size);
 	if (!ubi->peb_buf)
 		goto out_free;
 
@@ -1023,6 +1026,7 @@ int ubi_attach_mtd_dev(struct mtd_info *mtd, int ubi_num,
 		ubi->image_seq);
 	ubi_msg(ubi, "available PEBs: %d, total reserved PEBs: %d, PEBs reserved for bad PEB handling: %d",
 		ubi->avail_pebs, ubi->rsvd_pebs, ubi->beb_rsvd_pebs);
+	ubi_msg(ubi, "LEBs per PEB: %d", ubi->lebs_per_consolidated_peb);
 
 	/*
 	 * The below lock makes sure we do not race with 'ubi_thread()' which
diff --git a/drivers/mtd/ubi/io.c b/drivers/mtd/ubi/io.c
index 10cf3b5..604434a 100644
--- a/drivers/mtd/ubi/io.c
+++ b/drivers/mtd/ubi/io.c
@@ -99,10 +99,71 @@ static int self_check_peb_vid_hdr(const struct ubi_device *ubi, int pnum);
 static int self_check_vid_hdr(const struct ubi_device *ubi, int pnum,
 			      const struct ubi_vid_hdr *vid_hdr);
 static int self_check_write(struct ubi_device *ubi, const void *buf, int pnum,
-			    int offset, int len);
+			    int offset, int len, bool raw);
 
 /**
- * ubi_io_read - read data from a physical eraseblock.
+ * ubi_io_mtd_read - mtd_read wrapper handling MLC specificities.
+ * @ubi: UBI device description object
+ * @buf: buffer where to store the read data
+ * @pnum: physical eraseblock number to read from
+ * @offset: offset within the physical eraseblock from where to read
+ * @len: how many bytes to read
+ * @read: number of bytes successfully read from the underlying MTD device
+ *
+ * This function reads data from offset @offset of physical eraseblock @pnum
+ * and stores the read data in the @buf buffer. The following return codes are
+ * possible:
+ *
+ * o %0 if all the requested data were successfully read;
+ * o other negative error codes in case of other errors.
+ */
+static int ubi_io_mtd_read(const struct ubi_device *ubi, void *buf, int pnum,
+			   int offset, int len, size_t *read, bool raw)
+{
+	loff_t addr = (loff_t)pnum * ubi->consolidated_peb_size;
+	int wunitoffs, chunklen, err = 0, end = offset + len;
+	struct mtd_pairing_info info;
+
+	/*
+	 * Call mtd_read() directly if we're doing a raw read of interacting
+	 * with an SLC chip.
+	 */
+	if (raw || mtd_pairing_groups_per_eb(ubi->mtd) == 1)
+		return mtd_read(ubi->mtd, addr + offset, len, read, buf);
+
+	wunitoffs = offset % ubi->mtd->writesize;
+	info.pair = offset / ubi->mtd->writesize;
+	info.group = 0;
+	*read = 0;
+
+	while (offset < end) {
+		int realoffs;
+		size_t chunkread = 0;
+
+		chunklen = min_t(int, ubi->mtd->writesize - wunitoffs,
+				 end - offset);
+		realoffs = mtd_pairing_info_to_wunit(ubi->mtd, &info);
+		realoffs *= ubi->mtd->writesize;
+		realoffs += wunitoffs;
+		err = mtd_read(ubi->mtd, addr + realoffs, chunklen,
+			       &chunkread, buf);
+		*read += chunkread;
+		if (err && !mtd_is_bitflip(err))
+			return err;
+
+		offset += chunklen;
+		buf += chunklen;
+		info.pair++;
+
+		if (wunitoffs)
+			wunitoffs = 0;
+	}
+
+	return err;
+}
+
+/**
+ * __ubi_io_read - read data from a physical eraseblock.
  * @ubi: UBI device description object
  * @buf: buffer where to store the read data
  * @pnum: physical eraseblock number to read from
@@ -123,17 +184,17 @@ static int self_check_write(struct ubi_device *ubi, const void *buf, int pnum,
  * o %-EIO if some I/O error occurred;
  * o other negative error codes in case of other errors.
  */
-int ubi_io_read(const struct ubi_device *ubi, void *buf, int pnum, int offset,
-		int len)
+static int __ubi_io_read(const struct ubi_device *ubi, void *buf, int pnum,
+			 int offset, int len, bool raw)
 {
+	int peb_size = raw ? ubi->consolidated_peb_size : ubi->peb_size;
 	int err, retries = 0;
 	size_t read;
-	loff_t addr;
 
 	dbg_io("read %d bytes from PEB %d:%d", len, pnum, offset);
 
 	ubi_assert(pnum >= 0 && pnum < ubi->peb_count);
-	ubi_assert(offset >= 0 && offset + len <= ubi->peb_size);
+	ubi_assert(offset >= 0 && offset + len <= peb_size);
 	ubi_assert(len > 0);
 
 	err = self_check_not_bad(ubi, pnum);
@@ -162,9 +223,8 @@ int ubi_io_read(const struct ubi_device *ubi, void *buf, int pnum, int offset,
 	 */
 	*((uint8_t *)buf) ^= 0xFF;
 
-	addr = (loff_t)pnum * ubi->peb_size + offset;
 retry:
-	err = mtd_read(ubi->mtd, addr, len, &read, buf);
+	err = ubi_io_mtd_read(ubi, buf, pnum, offset, len, &read, raw);
 	if (err) {
 		const char *errstr = mtd_is_eccerr(err) ? " (ECC error)" : "";
 
@@ -215,6 +275,54 @@ retry:
 	return err;
 }
 
+int ubi_io_read(const struct ubi_device *ubi, void *buf, int pnum, int offset,
+		int len)
+{
+	return __ubi_io_read(ubi, buf, pnum, offset, len, false);
+}
+
+int ubi_io_raw_read(const struct ubi_device *ubi, void *buf, int pnum,
+		    int offset, int len)
+{
+	return __ubi_io_read(ubi, buf, pnum, offset, len, true);
+}
+
+int ubi_io_mtd_write(struct ubi_device *ubi, const void *buf, int pnum, int offset,
+		     int len, size_t *written, bool raw)
+{
+	loff_t addr = (loff_t)pnum * ubi->consolidated_peb_size;
+	int chunklen, err = 0, end = offset + len;
+	struct mtd_pairing_info info;
+
+	if (raw || mtd_pairing_groups_per_eb(ubi->mtd) == 1)
+		return mtd_write(ubi->mtd, addr + offset, len, written, buf);
+
+	info.pair = offset / ubi->mtd->writesize;
+	info.group = 0;
+	*written = 0;
+
+	while (offset < end) {
+		int realoffs;
+		size_t chunkwritten = 0;
+
+		chunklen = min_t(int, ubi->mtd->writesize,
+				 end - offset);
+		realoffs = mtd_pairing_info_to_wunit(ubi->mtd, &info);
+		realoffs *= ubi->mtd->writesize;
+		err = mtd_write(ubi->mtd, addr + realoffs, chunklen,
+				&chunkwritten, buf);
+		*written += chunkwritten;
+		if (err && !mtd_is_bitflip(err))
+			return err;
+
+		offset += chunklen;
+		buf += chunklen;
+		info.pair++;
+	}
+
+	return err;
+}
+
 /**
  * ubi_io_write - write data to a physical eraseblock.
  * @ubi: UBI device description object
@@ -232,17 +340,18 @@ retry:
  * Note, in case of an error, it is possible that something was still written
  * to the flash media, but may be some garbage.
  */
-int ubi_io_write(struct ubi_device *ubi, const void *buf, int pnum, int offset,
-		 int len)
+static int __ubi_io_write(struct ubi_device *ubi, const void *buf, int pnum,
+			  int offset, int len, bool raw)
 {
+	int peb_size = raw ? ubi->consolidated_peb_size : ubi->peb_size;
+	int rawoffs = offset, rawlen = len;
 	int err;
 	size_t written;
-	loff_t addr;
 
 	dbg_io("write %d bytes to PEB %d:%d", len, pnum, offset);
 
 	ubi_assert(pnum >= 0 && pnum < ubi->peb_count);
-	ubi_assert(offset >= 0 && offset + len <= ubi->peb_size);
+	ubi_assert(offset >= 0 && offset + len <= peb_size);
 	ubi_assert(offset % ubi->hdrs_min_io_size == 0);
 	ubi_assert(len > 0 && len % ubi->hdrs_min_io_size == 0);
 
@@ -255,8 +364,23 @@ int ubi_io_write(struct ubi_device *ubi, const void *buf, int pnum, int offset,
 	if (err)
 		return err;
 
+	if (!raw && mtd_pairing_groups_per_eb(ubi->mtd) > 1) {
+		struct mtd_pairing_info info;
+
+		info.pair = offset / ubi->mtd->writesize;
+		info.group = 0;
+		rawoffs = mtd_pairing_info_to_wunit(ubi->mtd, &info);
+		rawoffs *= ubi->mtd->writesize;
+
+		info.pair = (offset + len) / ubi->mtd->writesize;
+		info.group = 0;
+		rawlen = mtd_pairing_info_to_wunit(ubi->mtd, &info);
+		rawlen *= ubi->mtd->writesize;
+		rawlen -= rawoffs;
+	}
+
 	/* The area we are writing to has to contain all 0xFF bytes */
-	err = ubi_self_check_all_ff(ubi, pnum, offset, len);
+	err = ubi_self_check_all_ff(ubi, pnum, rawoffs, rawlen);
 	if (err)
 		return err;
 
@@ -280,8 +404,7 @@ int ubi_io_write(struct ubi_device *ubi, const void *buf, int pnum, int offset,
 		return -EIO;
 	}
 
-	addr = (loff_t)pnum * ubi->peb_size + offset;
-	err = mtd_write(ubi->mtd, addr, len, &written, buf);
+	err = ubi_io_mtd_write(ubi, buf, pnum, offset, len, &written, raw);
 	if (err) {
 		ubi_err(ubi, "error %d while writing %d bytes to PEB %d:%d, written %zd bytes",
 			err, len, pnum, offset, written);
@@ -291,7 +414,7 @@ int ubi_io_write(struct ubi_device *ubi, const void *buf, int pnum, int offset,
 		ubi_assert(written == len);
 
 	if (!err) {
-		err = self_check_write(ubi, buf, pnum, offset, len);
+		err = self_check_write(ubi, buf, pnum, offset, len, raw);
 		if (err)
 			return err;
 
@@ -299,15 +422,30 @@ int ubi_io_write(struct ubi_device *ubi, const void *buf, int pnum, int offset,
 		 * Since we always write sequentially, the rest of the PEB has
 		 * to contain only 0xFF bytes.
 		 */
-		offset += len;
-		len = ubi->peb_size - offset;
-		if (len)
-			err = ubi_self_check_all_ff(ubi, pnum, offset, len);
+		rawoffs += rawlen;
+		rawlen = peb_size - rawoffs;
+		if (rawlen)
+			err = ubi_self_check_all_ff(ubi, pnum, rawoffs,
+						    rawlen);
 	}
 
 	return err;
 }
 
+int ubi_io_write(struct ubi_device *ubi, const void *buf, int pnum, int offset,
+		 int len)
+{
+	ubi_assert(!ubi->consolidated || !ubi->consolidated[pnum]);
+
+	return __ubi_io_write(ubi, buf, pnum, offset, len, false);
+}
+
+int ubi_io_raw_write(struct ubi_device *ubi, const void *buf, int pnum,
+		     int offset, int len)
+{
+	return __ubi_io_write(ubi, buf, pnum, offset, len, true);
+}
+
 /**
  * erase_callback - MTD erasure call-back.
  * @ei: MTD erase information object.
@@ -348,8 +486,8 @@ retry:
 	memset(&ei, 0, sizeof(struct erase_info));
 
 	ei.mtd      = ubi->mtd;
-	ei.addr     = (loff_t)pnum * ubi->peb_size;
-	ei.len      = ubi->peb_size;
+	ei.addr     = (loff_t)pnum * ubi->consolidated_peb_size;
+	ei.len      = ubi->consolidated_peb_size;
 	ei.callback = erase_callback;
 	ei.priv     = (unsigned long)&wq;
 
@@ -385,7 +523,7 @@ retry:
 		return -EIO;
 	}
 
-	err = ubi_self_check_all_ff(ubi, pnum, 0, ubi->peb_size);
+	err = ubi_self_check_all_ff(ubi, pnum, 0, ubi->consolidated_peb_size);
 	if (err)
 		return err;
 
@@ -424,11 +562,13 @@ static int torture_peb(struct ubi_device *ubi, int pnum)
 			goto out;
 
 		/* Make sure the PEB contains only 0xFF bytes */
-		err = ubi_io_read(ubi, ubi->peb_buf, pnum, 0, ubi->peb_size);
+		err = ubi_io_raw_read(ubi, ubi->peb_buf, pnum, 0,
+				      ubi->consolidated_peb_size);
 		if (err)
 			goto out;
 
-		err = ubi_check_pattern(ubi->peb_buf, 0xFF, ubi->peb_size);
+		err = ubi_check_pattern(ubi->peb_buf, 0xFF,
+					ubi->consolidated_peb_size);
 		if (err == 0) {
 			ubi_err(ubi, "erased PEB %d, but a non-0xFF byte found",
 				pnum);
@@ -437,18 +577,20 @@ static int torture_peb(struct ubi_device *ubi, int pnum)
 		}
 
 		/* Write a pattern and check it */
-		memset(ubi->peb_buf, patterns[i], ubi->peb_size);
-		err = ubi_io_write(ubi, ubi->peb_buf, pnum, 0, ubi->peb_size);
+		memset(ubi->peb_buf, patterns[i], ubi->consolidated_peb_size);
+		err = ubi_io_raw_write(ubi, ubi->peb_buf, pnum, 0,
+				       ubi->consolidated_peb_size);
 		if (err)
 			goto out;
 
 		memset(ubi->peb_buf, ~patterns[i], ubi->peb_size);
-		err = ubi_io_read(ubi, ubi->peb_buf, pnum, 0, ubi->peb_size);
+		err = ubi_io_raw_read(ubi, ubi->peb_buf, pnum, 0,
+				      ubi->consolidated_peb_size);
 		if (err)
 			goto out;
 
 		err = ubi_check_pattern(ubi->peb_buf, patterns[i],
-					ubi->peb_size);
+					ubi->consolidated_peb_size);
 		if (err == 0) {
 			ubi_err(ubi, "pattern %x checking failed for PEB %d",
 				patterns[i], pnum);
@@ -519,7 +661,7 @@ static int nor_erase_prepare(struct ubi_device *ubi, int pnum)
 	 * invalid VID header, in which case UBI will treat this PEB as
 	 * corrupted and will try to preserve it, and print scary warnings.
 	 */
-	addr = (loff_t)pnum * ubi->peb_size;
+	addr = (loff_t)pnum * ubi->consolidated_peb_size;
 	err = ubi_io_read_ec_hdr(ubi, pnum, &ec_hdr, 0);
 	if (err != UBI_IO_BAD_HDR_EBADMSG && err != UBI_IO_BAD_HDR &&
 	    err != UBI_IO_FF){
@@ -545,7 +687,7 @@ error:
 	 * return an error.
 	 */
 	ubi_err(ubi, "cannot invalidate PEB %d, write returned %d", pnum, err);
-	ubi_dump_flash(ubi, pnum, 0, ubi->peb_size);
+	ubi_dump_flash(ubi, pnum, 0, ubi->consolidated_peb_size);
 	return -EIO;
 }
 
@@ -616,7 +758,9 @@ int ubi_io_is_bad(const struct ubi_device *ubi, int pnum)
 	if (ubi->bad_allowed) {
 		int ret;
 
-		ret = mtd_block_isbad(mtd, (loff_t)pnum * ubi->peb_size);
+		ret = mtd_block_isbad(mtd,
+				      (loff_t)pnum *
+				      ubi->consolidated_peb_size);
 		if (ret < 0)
 			ubi_err(ubi, "error %d while checking if PEB %d is bad",
 				ret, pnum);
@@ -651,7 +795,8 @@ int ubi_io_mark_bad(const struct ubi_device *ubi, int pnum)
 	if (!ubi->bad_allowed)
 		return 0;
 
-	err = mtd_block_markbad(mtd, (loff_t)pnum * ubi->peb_size);
+	err = mtd_block_markbad(mtd,
+				(loff_t)pnum * ubi->consolidated_peb_size);
 	if (err)
 		ubi_err(ubi, "cannot mark PEB %d bad, error %d", pnum, err);
 	return err;
@@ -992,92 +1137,142 @@ bad:
 }
 
 /**
- * ubi_io_read_vid_hdr - read and check a volume identifier header.
+ * ubi_io_read_vid_hdrs - read and check a volume identifier header.
  * @ubi: UBI device description object
  * @pnum: physical eraseblock number to read from
- * @vid_hdr: &struct ubi_vid_hdr object where to store the read volume
- * identifier header
+ * @vid_hdrs: &struct ubi_vid_hdr pointer where to store the read volume
+ * identifier headers
+ * @max: maximum number of headers that can be retrieved
  * @verbose: be verbose if the header is corrupted or wasn't found
  *
- * This function reads the volume identifier header from physical eraseblock
- * @pnum and stores it in @vid_hdr. It also checks CRC checksum of the read
+ * This function reads the volume identifier headers from physical eraseblock
+ * @pnum and stores it in @vid_hdrs. It also checks CRC checksum of the read
  * volume identifier header. The error codes are the same as in
  * 'ubi_io_read_ec_hdr()'.
  *
  * Note, the implementation of this function is also very similar to
  * 'ubi_io_read_ec_hdr()', so refer commentaries in 'ubi_io_read_ec_hdr()'.
  */
-int ubi_io_read_vid_hdr(struct ubi_device *ubi, int pnum,
-			struct ubi_vid_hdr *vid_hdr, int verbose)
+int ubi_io_read_vid_hdrs(struct ubi_device *ubi, int pnum,
+			 struct ubi_vid_hdr *vid_hdrs, int *num,
+			 int verbose)
 {
-	int err, read_err;
+	int err = 0, ret, read_err, i, max = *num;
 	uint32_t crc, magic, hdr_crc;
 	void *p;
 
-	dbg_io("read VID header from PEB %d", pnum);
+	dbg_io("read VID headers from PEB %d", pnum);
 	ubi_assert(pnum >= 0 &&  pnum < ubi->peb_count);
 
-	p = (char *)vid_hdr - ubi->vid_hdr_shift;
+	p = (char *)vid_hdrs - ubi->vid_hdr_shift;
 	read_err = ubi_io_read(ubi, p, pnum, ubi->vid_hdr_aloffset,
 			  ubi->vid_hdr_alsize);
-	if (read_err && read_err != UBI_IO_BITFLIPS && !mtd_is_eccerr(read_err))
+	if (read_err && read_err != UBI_IO_BITFLIPS &&
+	    !mtd_is_eccerr(read_err))
 		return read_err;
 
-	magic = be32_to_cpu(vid_hdr->magic);
-	if (magic != UBI_VID_HDR_MAGIC) {
-		if (mtd_is_eccerr(read_err))
-			return UBI_IO_BAD_HDR_EBADMSG;
+	for (i = 0; i < max; i++) {
+		struct ubi_vid_hdr *vid_hdr = vid_hdrs + i;
 
-		if (ubi_check_pattern(vid_hdr, 0xFF, UBI_VID_HDR_SIZE)) {
-			if (verbose)
-				ubi_warn(ubi, "no VID header found at PEB %d, only 0xFF bytes",
-					 pnum);
-			dbg_bld("no VID header found at PEB %d, only 0xFF bytes",
-				pnum);
+		magic = be32_to_cpu(vid_hdr->magic);
+		if (magic != UBI_VID_HDR_MAGIC) {
+			if (mtd_is_eccerr(read_err)) {
+				err = UBI_IO_BAD_HDR_EBADMSG;
+				break;
+			}
+
+			if (ubi_check_pattern(vid_hdr, 0xFF, UBI_VID_HDR_SIZE)) {
+				if (verbose && !i)
+					ubi_warn(ubi, "no VID header found at PEB %d, only 0xFF bytes",
+						 pnum);
+				dbg_bld("no VID header found at PEB %d, only 0xFF bytes",
+					pnum);
+				if (!read_err)
+					err = UBI_IO_FF;
+				else
+					err = UBI_IO_FF_BITFLIPS;
+
+				/*
+				 * Propagate UBI_IO_FF/_BITFLIPS only for the first VID header,
+				 * otherwise UBI will treat the whole PEB as empty.
+				 */
+				if (i)
+					err = 0;
+
+				break;
+			}
+
+			if (verbose && !i) {
+				ubi_warn(ubi, "bad magic number at PEB %d: %08x instead of %08x",
+						pnum, magic, UBI_VID_HDR_MAGIC);
+				ubi_dump_vid_hdr(vid_hdr);
+			}
+			dbg_bld("bad magic number at PEB %d: %08x instead of %08x",
+					pnum, magic, UBI_VID_HDR_MAGIC);
+			err = UBI_IO_BAD_HDR;
+			break;
+		}
+
+		crc = crc32(UBI_CRC32_INIT, vid_hdr, UBI_VID_HDR_SIZE_CRC);
+		hdr_crc = be32_to_cpu(vid_hdr->hdr_crc);
+
+		if (hdr_crc != crc) {
+			if (verbose && !i) {
+				ubi_warn(ubi, "bad CRC at PEB %d, calculated %#08x, read %#08x",
+						pnum, crc, hdr_crc);
+				ubi_dump_vid_hdr(vid_hdr);
+			}
+			dbg_bld("bad CRC at PEB %d, calculated %#08x, read %#08x",
+					pnum, crc, hdr_crc);
 			if (!read_err)
-				return UBI_IO_FF;
+				err = UBI_IO_BAD_HDR;
 			else
-				return UBI_IO_FF_BITFLIPS;
+				err = UBI_IO_BAD_HDR_EBADMSG;
+			break;
 		}
 
-		if (verbose) {
-			ubi_warn(ubi, "bad magic number at PEB %d: %08x instead of %08x",
-				 pnum, magic, UBI_VID_HDR_MAGIC);
-			ubi_dump_vid_hdr(vid_hdr);
+		ret = validate_vid_hdr(ubi, vid_hdr);
+		if (ret && !i) {
+			ubi_err(ubi, "validation failed for PEB %d", pnum);
+			err = -EINVAL;
+			break;
 		}
-		dbg_bld("bad magic number at PEB %d: %08x instead of %08x",
-			pnum, magic, UBI_VID_HDR_MAGIC);
-		return UBI_IO_BAD_HDR;
+
+		if (read_err)
+			err |= UBI_IO_BITFLIPS;
 	}
 
-	crc = crc32(UBI_CRC32_INIT, vid_hdr, UBI_VID_HDR_SIZE_CRC);
-	hdr_crc = be32_to_cpu(vid_hdr->hdr_crc);
+	*num = i;
 
-	if (hdr_crc != crc) {
-		if (verbose) {
-			ubi_warn(ubi, "bad CRC at PEB %d, calculated %#08x, read %#08x",
-				 pnum, crc, hdr_crc);
-			ubi_dump_vid_hdr(vid_hdr);
-		}
-		dbg_bld("bad CRC at PEB %d, calculated %#08x, read %#08x",
-			pnum, crc, hdr_crc);
-		if (!read_err)
-			return UBI_IO_BAD_HDR;
-		else
-			return UBI_IO_BAD_HDR_EBADMSG;
-	}
+	return err;
+}
 
-	err = validate_vid_hdr(ubi, vid_hdr);
-	if (err) {
-		ubi_err(ubi, "validation failed for PEB %d", pnum);
-		return -EINVAL;
-	}
+/**
+ * ubi_io_read_vid_hdr - read and check a volume identifier header.
+ * @ubi: UBI device description object
+ * @pnum: physical eraseblock number to read from
+ * @vid_hdr: &struct ubi_vid_hdr object where to store the read volume
+ * identifier header
+ * @verbose: be verbose if the header is corrupted or wasn't found
+ *
+ * This function reads the volume identifier header from physical eraseblock
+ * @pnum and stores it in @vid_hdr. It also checks CRC checksum of the read
+ * volume identifier header. The error codes are the same as in
+ * 'ubi_io_read_ec_hdr()'.
+ *
+ * Note, the implementation of this function is also very similar to
+ * 'ubi_io_read_ec_hdr()', so refer commentaries in 'ubi_io_read_ec_hdr()'.
+ */
+int ubi_io_read_vid_hdr(struct ubi_device *ubi, int pnum,
+			struct ubi_vid_hdr *vid_hdr, int verbose)
+{
+	int nhdrs = 1;
 
-	return read_err ? UBI_IO_BITFLIPS : 0;
+	return ubi_io_read_vid_hdrs(ubi, pnum, vid_hdr, &nhdrs, verbose);
 }
 
 /**
- * ubi_io_write_vid_hdr - write a volume identifier header.
+ * ubi_io_write_vid_hdrs - write a volume identifier header.
  * @ubi: UBI device description object
  * @pnum: the physical eraseblock number to write to
  * @vid_hdr: the volume identifier header to write
@@ -1091,10 +1286,10 @@ int ubi_io_read_vid_hdr(struct ubi_device *ubi, int pnum,
  * case of failure. If %-EIO is returned, the physical eraseblock probably went
  * bad.
  */
-int ubi_io_write_vid_hdr(struct ubi_device *ubi, int pnum,
-			 struct ubi_vid_hdr *vid_hdr)
+int ubi_io_write_vid_hdrs(struct ubi_device *ubi, int pnum,
+			  struct ubi_vid_hdr *vid_hdrs, int num)
 {
-	int err;
+	int err, i;
 	uint32_t crc;
 	void *p;
 
@@ -1105,25 +1300,49 @@ int ubi_io_write_vid_hdr(struct ubi_device *ubi, int pnum,
 	if (err)
 		return err;
 
-	vid_hdr->magic = cpu_to_be32(UBI_VID_HDR_MAGIC);
-	vid_hdr->version = UBI_VERSION;
-	crc = crc32(UBI_CRC32_INIT, vid_hdr, UBI_VID_HDR_SIZE_CRC);
-	vid_hdr->hdr_crc = cpu_to_be32(crc);
+	for (i = 0; i < num; i++) {
+		vid_hdrs[i].magic = cpu_to_be32(UBI_VID_HDR_MAGIC);
+		vid_hdrs[i].version = UBI_VERSION;
+		crc = crc32(UBI_CRC32_INIT, &vid_hdrs[i],
+			    UBI_VID_HDR_SIZE_CRC);
+		vid_hdrs[i].hdr_crc = cpu_to_be32(crc);
 
-	err = self_check_vid_hdr(ubi, pnum, vid_hdr);
-	if (err)
-		return err;
+		err = self_check_vid_hdr(ubi, pnum, &vid_hdrs[i]);
+		if (err)
+			return err;
+	}
 
 	if (ubi_dbg_power_cut(ubi, POWER_CUT_VID_WRITE))
 		return -EROFS;
 
-	p = (char *)vid_hdr - ubi->vid_hdr_shift;
+	p = (char *)vid_hdrs - ubi->vid_hdr_shift;
 	err = ubi_io_write(ubi, p, pnum, ubi->vid_hdr_aloffset,
 			   ubi->vid_hdr_alsize);
 	return err;
 }
 
 /**
+ * ubi_io_write_vid_hdr - write a volume identifier header.
+ * @ubi: UBI device description object
+ * @pnum: the physical eraseblock number to write to
+ * @vid_hdr: the volume identifier header to write
+ *
+ * This function writes the volume identifier header described by @vid_hdr to
+ * physical eraseblock @pnum. This function automatically fills the
+ * @vid_hdr->magic and the @vid_hdr->version fields, as well as calculates
+ * header CRC checksum and stores it at vid_hdr->hdr_crc.
+ *
+ * This function returns zero in case of success and a negative error code in
+ * case of failure. If %-EIO is returned, the physical eraseblock probably went
+ * bad.
+ */
+int ubi_io_write_vid_hdr(struct ubi_device *ubi, int pnum,
+			 struct ubi_vid_hdr *vid_hdr)
+{
+	return ubi_io_write_vid_hdrs(ubi, pnum, vid_hdr, 1);
+}
+
+/**
  * self_check_not_bad - ensure that a physical eraseblock is not bad.
  * @ubi: UBI device description object
  * @pnum: physical eraseblock number to check
@@ -1331,23 +1550,23 @@ exit:
  * match and a negative error code if not or in case of failure.
  */
 static int self_check_write(struct ubi_device *ubi, const void *buf, int pnum,
-			    int offset, int len)
+			    int offset, int len, bool raw)
 {
 	int err, i;
-	size_t read;
 	void *buf1;
-	loff_t addr = (loff_t)pnum * ubi->peb_size + offset;
+
+	return 0;
 
 	if (!ubi_dbg_chk_io(ubi))
 		return 0;
 
-	buf1 = __vmalloc(len, GFP_NOFS, PAGE_KERNEL);
+	buf1 = __vmalloc(0x1f8000, GFP_NOFS, PAGE_KERNEL);
 	if (!buf1) {
 		ubi_err(ubi, "cannot allocate memory to check writes");
 		return 0;
 	}
 
-	err = mtd_read(ubi->mtd, addr, len, &read, buf1);
+	err = __ubi_io_read(ubi, buf1, pnum, offset, len, raw);
 	if (err && !mtd_is_bitflip(err))
 		goto out_free;
 
@@ -1400,7 +1619,10 @@ int ubi_self_check_all_ff(struct ubi_device *ubi, int pnum, int offset, int len)
 	size_t read;
 	int err;
 	void *buf;
-	loff_t addr = (loff_t)pnum * ubi->peb_size + offset;
+	loff_t addr = (loff_t)pnum * ubi->consolidated_peb_size + offset;
+
+	//XXX
+	return 0;
 
 	if (!ubi_dbg_chk_io(ubi))
 		return 0;
diff --git a/drivers/mtd/ubi/ubi.h b/drivers/mtd/ubi/ubi.h
index c901cc0..aab984f 100644
--- a/drivers/mtd/ubi/ubi.h
+++ b/drivers/mtd/ubi/ubi.h
@@ -602,6 +602,7 @@ struct ubi_device {
 	long long flash_size;
 	int peb_count;
 	int peb_size;
+	int consolidated_peb_size;
 	int bad_peb_count;
 	int good_peb_count;
 	int corr_peb_count;
@@ -879,8 +880,12 @@ bool ubi_work_join_one(struct ubi_device *ubi);
 /* io.c */
 int ubi_io_read(const struct ubi_device *ubi, void *buf, int pnum, int offset,
 		int len);
+int ubi_io_raw_read(const struct ubi_device *ubi, void *buf, int pnum,
+		    int offset, int len);
 int ubi_io_write(struct ubi_device *ubi, const void *buf, int pnum, int offset,
 		 int len);
+int ubi_io_raw_write(struct ubi_device *ubi, const void *buf, int pnum,
+		     int offset, int len);
 int ubi_io_sync_erase(struct ubi_device *ubi, int pnum, int torture);
 int ubi_io_is_bad(const struct ubi_device *ubi, int pnum);
 int ubi_io_mark_bad(const struct ubi_device *ubi, int pnum);
@@ -890,8 +895,13 @@ int ubi_io_write_ec_hdr(struct ubi_device *ubi, int pnum,
 			struct ubi_ec_hdr *ec_hdr);
 int ubi_io_read_vid_hdr(struct ubi_device *ubi, int pnum,
 			struct ubi_vid_hdr *vid_hdr, int verbose);
+int ubi_io_read_vid_hdrs(struct ubi_device *ubi, int pnum,
+			 struct ubi_vid_hdr *vid_hdrs, int *num,
+			 int verbose);
 int ubi_io_write_vid_hdr(struct ubi_device *ubi, int pnum,
 			 struct ubi_vid_hdr *vid_hdr);
+int ubi_io_write_vid_hdrs(struct ubi_device *ubi, int pnum,
+			  struct ubi_vid_hdr *vid_hdrs, int num);
 
 /* build.c */
 int ubi_attach_mtd_dev(struct mtd_info *mtd, int ubi_num,
-- 
2.7.3

[PATCH 11/13] ubi: LEB consolidation

[Richard Weinberger]

Implements LEB consolidation for MLC NAND. By default data is written
in SLC mode. As soon UBI runs out of space two fully written PEBs will
get merged to produce free space.
The merge operation takes two PEB written in SLC mode and writes back
all data into one PEB in MLC mode. The operation works almost as atomic
LEB change and is power cut safe.
This patch was implemented by Boris Brezillon and me.

TODO:
- Reconstruct “full”-states upon attach
- Make all paths generic for pairings > 2 too to support TLC NAND in future
- Cut the new on-flash layout into stone and raise UBI version (by having feature flags)

Signed-off-by: Richard Weinberger <richard@nod.at>
---
 drivers/mtd/ubi/Kconfig       |   4 +
 drivers/mtd/ubi/Makefile      |   1 +
 drivers/mtd/ubi/attach.c      | 407 ++++++++++++++++++++--------------
 drivers/mtd/ubi/build.c       |  18 +-
 drivers/mtd/ubi/cdev.c        |  12 +-
 drivers/mtd/ubi/consolidate.c | 499 ++++++++++++++++++++++++++++++++++++++++++
 drivers/mtd/ubi/debug.c       |  16 +-
 drivers/mtd/ubi/debug.h       |   2 +-
 drivers/mtd/ubi/eba.c         | 411 ++++++++++++++++++++++++++++------
 drivers/mtd/ubi/fastmap-wl.c  |  19 +-
 drivers/mtd/ubi/fastmap.c     | 261 +++++++++++++++-------
 drivers/mtd/ubi/kapi.c        |  14 +-
 drivers/mtd/ubi/ubi-media.h   |  18 +-
 drivers/mtd/ubi/ubi.h         | 162 +++++++++++---
 drivers/mtd/ubi/upd.c         |   2 +-
 drivers/mtd/ubi/vmt.c         | 104 +++++----
 drivers/mtd/ubi/vtbl.c        |  87 +++++---
 drivers/mtd/ubi/wl.c          | 187 +++++++++++-----
 18 files changed, 1718 insertions(+), 506 deletions(-)
 create mode 100644 drivers/mtd/ubi/consolidate.c

diff --git a/drivers/mtd/ubi/Kconfig b/drivers/mtd/ubi/Kconfig
index f0855ce..a6f7d3b 100644
--- a/drivers/mtd/ubi/Kconfig
+++ b/drivers/mtd/ubi/Kconfig
@@ -103,4 +103,8 @@ config MTD_UBI_BLOCK
 
 	   If in doubt, say "N".
 
+config MTD_UBI_CONSOLIDATE
+	bool "LEB consolidation support"
+	default n
+
 endif # MTD_UBI
diff --git a/drivers/mtd/ubi/Makefile b/drivers/mtd/ubi/Makefile
index e9d4b1d..86a77c5 100644
--- a/drivers/mtd/ubi/Makefile
+++ b/drivers/mtd/ubi/Makefile
@@ -4,5 +4,6 @@ ubi-y += vtbl.o vmt.o upd.o build.o cdev.o kapi.o eba.o io.o wl.o attach.o
 ubi-y += work.o misc.o debug.o
 ubi-$(CONFIG_MTD_UBI_FASTMAP) += fastmap.o
 ubi-$(CONFIG_MTD_UBI_BLOCK) += block.o
+ubi-$(CONFIG_MTD_UBI_CONSOLIDATE) += consolidate.o
 
 obj-$(CONFIG_MTD_UBI_GLUEBI) += gluebi.o
diff --git a/drivers/mtd/ubi/attach.c b/drivers/mtd/ubi/attach.c
index c1aaf03..d1bd34c 100644
--- a/drivers/mtd/ubi/attach.c
+++ b/drivers/mtd/ubi/attach.c
@@ -116,10 +116,14 @@ static struct ubi_vid_hdr *vidh;
  * returns zero in case of success and a negative error code in case of
  * failure.
  */
-static int add_to_list(struct ubi_attach_info *ai, int pnum, int vol_id,
-		       int lnum, int ec, int to_head, struct list_head *list)
+static void add_peb_to_list(struct ubi_attach_info *ai,
+			    struct ubi_ainf_peb *aeb, int to_head,
+			    struct list_head *list)
 {
-	struct ubi_ainf_peb *aeb;
+	int pnum = aeb->pnum, ec = aeb->ec;
+
+	if (!list_empty(&aeb->list))
+		list_del(&aeb->list);
 
 	if (list == &ai->free) {
 		dbg_bld("add to free: PEB %d, EC %d", pnum, ec);
@@ -131,18 +135,28 @@ static int add_to_list(struct ubi_attach_info *ai, int pnum, int vol_id,
 	} else
 		BUG();
 
-	aeb = kmem_cache_alloc(ai->aeb_slab_cache, GFP_KERNEL);
+	if (to_head)
+		list_add(&aeb->list, list);
+	else
+		list_add_tail(&aeb->list, list);
+
+}
+
+static int add_to_list(struct ubi_attach_info *ai, int pnum, int ec,
+		       int to_head, struct list_head *list)
+{
+	struct ubi_ainf_peb *aeb;
+
+	aeb = kmem_cache_zalloc(ai->apeb_slab_cache, GFP_KERNEL);
 	if (!aeb)
 		return -ENOMEM;
 
+	INIT_LIST_HEAD(&aeb->list);
 	aeb->pnum = pnum;
-	aeb->vol_id = vol_id;
-	aeb->lnum = lnum;
 	aeb->ec = ec;
-	if (to_head)
-		list_add(&aeb->u.list, list);
-	else
-		list_add_tail(&aeb->u.list, list);
+
+	add_peb_to_list(ai, aeb, to_head, list);
+
 	return 0;
 }
 
@@ -163,14 +177,15 @@ static int add_corrupted(struct ubi_attach_info *ai, int pnum, int ec)
 
 	dbg_bld("add to corrupted: PEB %d, EC %d", pnum, ec);
 
-	aeb = kmem_cache_alloc(ai->aeb_slab_cache, GFP_KERNEL);
+	aeb = kmem_cache_alloc(ai->apeb_slab_cache, GFP_KERNEL);
 	if (!aeb)
 		return -ENOMEM;
 
+	INIT_LIST_HEAD(&aeb->list);
 	ai->corr_peb_count += 1;
 	aeb->pnum = pnum;
 	aeb->ec = ec;
-	list_add(&aeb->u.list, &ai->corr);
+	list_add(&aeb->list, &ai->corr);
 	return 0;
 }
 
@@ -321,8 +336,8 @@ static struct ubi_ainf_volume *add_volume(struct ubi_attach_info *ai,
  *     o bit 2 is cleared: the older LEB is not corrupted;
  *     o bit 2 is set: the older LEB is corrupted.
  */
-int ubi_compare_lebs(struct ubi_device *ubi, const struct ubi_ainf_peb *aeb,
-			int pnum, const struct ubi_vid_hdr *vid_hdr)
+int ubi_compare_lebs(struct ubi_device *ubi, const struct ubi_ainf_leb *aeb,
+		     int pnum, const struct ubi_vid_hdr *vid_hdr)
 {
 	int len, err, second_is_newer, bitflips = 0, corrupted = 0;
 	uint32_t data_crc, crc;
@@ -362,6 +377,8 @@ int ubi_compare_lebs(struct ubi_device *ubi, const struct ubi_ainf_peb *aeb,
 			return 1;
 		}
 	} else {
+		int nvidh = ubi->lebs_per_cpeb;
+
 		if (!aeb->copy_flag) {
 			/* It is not a copy, so it is newer */
 			dbg_bld("first PEB %d is newer, copy_flag is unset",
@@ -373,8 +390,8 @@ int ubi_compare_lebs(struct ubi_device *ubi, const struct ubi_ainf_peb *aeb,
 		if (!vh)
 			return -ENOMEM;
 
-		pnum = aeb->pnum;
-		err = ubi_io_read_vid_hdr(ubi, pnum, vh, 0);
+		pnum = aeb->peb->pnum;
+		err = ubi_io_read_vid_hdrs(ubi, pnum, vh, &nvidh, 0);
 		if (err) {
 			if (err == UBI_IO_BITFLIPS)
 				bitflips = 1;
@@ -388,7 +405,8 @@ int ubi_compare_lebs(struct ubi_device *ubi, const struct ubi_ainf_peb *aeb,
 			}
 		}
 
-		vid_hdr = vh;
+		ubi_assert(aeb->peb_pos < nvidh);
+		vid_hdr = &vh[aeb->peb_pos];
 	}
 
 	/* Read the data of the copy and check the CRC */
@@ -446,18 +464,21 @@ out_free_vidh:
  * to be picked, while the older one has to be dropped. This function returns
  * zero in case of success and a negative error code in case of failure.
  */
-int ubi_add_to_av(struct ubi_device *ubi, struct ubi_attach_info *ai, int pnum,
-		  int ec, const struct ubi_vid_hdr *vid_hdr, int bitflips)
+int ubi_add_to_av(struct ubi_device *ubi, struct ubi_attach_info *ai,
+		  struct ubi_ainf_peb *peb, const struct ubi_vid_hdr *vid_hdr,
+		  int peb_pos, int bitflips, bool full)
 {
-	int err, vol_id, lnum;
+	int err, vol_id, lnum, pnum, ec;
 	unsigned long long sqnum;
 	struct ubi_ainf_volume *av;
-	struct ubi_ainf_peb *aeb;
+	struct ubi_ainf_leb *leb;
 	struct rb_node **p, *parent = NULL;
 
 	vol_id = be32_to_cpu(vid_hdr->vol_id);
 	lnum = be32_to_cpu(vid_hdr->lnum);
 	sqnum = be64_to_cpu(vid_hdr->sqnum);
+	pnum = peb->pnum;
+	ec = peb->ec;
 
 	dbg_bld("PEB %d, LEB %d:%d, EC %d, sqnum %llu, bitflips %d",
 		pnum, vol_id, lnum, ec, sqnum, bitflips);
@@ -478,9 +499,9 @@ int ubi_add_to_av(struct ubi_device *ubi, struct ubi_attach_info *ai, int pnum,
 		int cmp_res;
 
 		parent = *p;
-		aeb = rb_entry(parent, struct ubi_ainf_peb, u.rb);
-		if (lnum != aeb->lnum) {
-			if (lnum < aeb->lnum)
+		leb = rb_entry(parent, struct ubi_ainf_leb, rb);
+		if (lnum != leb->desc.lnum) {
+			if (lnum < leb->desc.lnum)
 				p = &(*p)->rb_left;
 			else
 				p = &(*p)->rb_right;
@@ -493,7 +514,7 @@ int ubi_add_to_av(struct ubi_device *ubi, struct ubi_attach_info *ai, int pnum,
 		 */
 
 		dbg_bld("this LEB already exists: PEB %d, sqnum %llu, EC %d",
-			aeb->pnum, aeb->sqnum, aeb->ec);
+			leb->peb->pnum, leb->sqnum, leb->peb->ec);
 
 		/*
 		 * Make sure that the logical eraseblocks have different
@@ -508,10 +529,10 @@ int ubi_add_to_av(struct ubi_device *ubi, struct ubi_attach_info *ai, int pnum,
 		 * images, but refuse attaching old images with duplicated
 		 * logical eraseblocks because there was an unclean reboot.
 		 */
-		if (aeb->sqnum == sqnum && sqnum != 0) {
+		if (leb->sqnum == sqnum && sqnum != 0) {
 			ubi_err(ubi, "two LEBs with same sequence number %llu",
 				sqnum);
-			ubi_dump_aeb(aeb, 0);
+			ubi_dump_aeb(leb, 0);
 			ubi_dump_vid_hdr(vid_hdr);
 			return -EINVAL;
 		}
@@ -520,7 +541,7 @@ int ubi_add_to_av(struct ubi_device *ubi, struct ubi_attach_info *ai, int pnum,
 		 * Now we have to drop the older one and preserve the newer
 		 * one.
 		 */
-		cmp_res = ubi_compare_lebs(ubi, aeb, pnum, vid_hdr);
+		cmp_res = ubi_compare_lebs(ubi, leb, pnum, vid_hdr);
 		if (cmp_res < 0)
 			return cmp_res;
 
@@ -533,19 +554,16 @@ int ubi_add_to_av(struct ubi_device *ubi, struct ubi_attach_info *ai, int pnum,
 			if (err)
 				return err;
 
-			err = add_to_list(ai, aeb->pnum, aeb->vol_id,
-					  aeb->lnum, aeb->ec, cmp_res & 4,
-					  &ai->erase);
-			if (err)
-				return err;
+			if (--leb->peb->refcount <= 0)
+				add_peb_to_list(ai, leb->peb, cmp_res & 4,
+						&ai->erase);
 
-			aeb->ec = ec;
-			aeb->pnum = pnum;
-			aeb->vol_id = vol_id;
-			aeb->lnum = lnum;
-			aeb->scrub = ((cmp_res & 2) || bitflips);
-			aeb->copy_flag = vid_hdr->copy_flag;
-			aeb->sqnum = sqnum;
+			leb->peb_pos = peb_pos;
+			leb->peb = peb;
+			peb->scrub = ((cmp_res & 2) || bitflips || peb->scrub);
+			leb->copy_flag = vid_hdr->copy_flag;
+			leb->sqnum = sqnum;
+			leb->full = full;
 
 			if (av->highest_lnum == lnum)
 				av->last_data_size =
@@ -557,8 +575,11 @@ int ubi_add_to_av(struct ubi_device *ubi, struct ubi_attach_info *ai, int pnum,
 			 * This logical eraseblock is older than the one found
 			 * previously.
 			 */
-			return add_to_list(ai, pnum, vol_id, lnum, ec,
-					   cmp_res & 4, &ai->erase);
+			if (--peb->refcount <= 0)
+				add_peb_to_list(ai, peb, cmp_res & 4,
+						&ai->erase);
+
+			return 0;
 		}
 	}
 
@@ -571,17 +592,18 @@ int ubi_add_to_av(struct ubi_device *ubi, struct ubi_attach_info *ai, int pnum,
 	if (err)
 		return err;
 
-	aeb = kmem_cache_alloc(ai->aeb_slab_cache, GFP_KERNEL);
-	if (!aeb)
+	leb = kmem_cache_alloc(ai->aleb_slab_cache, GFP_KERNEL);
+	if (!leb)
 		return -ENOMEM;
 
-	aeb->ec = ec;
-	aeb->pnum = pnum;
-	aeb->vol_id = vol_id;
-	aeb->lnum = lnum;
-	aeb->scrub = bitflips;
-	aeb->copy_flag = vid_hdr->copy_flag;
-	aeb->sqnum = sqnum;
+	leb->peb = peb;
+	leb->peb_pos = peb_pos;
+	leb->desc.vol_id = vol_id;
+	leb->desc.lnum = lnum;
+	peb->scrub = (bitflips || peb->scrub);
+	leb->copy_flag = vid_hdr->copy_flag;
+	leb->sqnum = sqnum;
+	leb->full = full;
 
 	if (av->highest_lnum <= lnum) {
 		av->highest_lnum = lnum;
@@ -589,8 +611,8 @@ int ubi_add_to_av(struct ubi_device *ubi, struct ubi_attach_info *ai, int pnum,
 	}
 
 	av->leb_count += 1;
-	rb_link_node(&aeb->u.rb, parent, p);
-	rb_insert_color(&aeb->u.rb, &av->root);
+	rb_link_node(&leb->rb, parent, p);
+	rb_insert_color(&leb->rb, &av->root);
 	return 0;
 }
 
@@ -631,14 +653,17 @@ struct ubi_ainf_volume *ubi_find_av(const struct ubi_attach_info *ai,
 void ubi_remove_av(struct ubi_attach_info *ai, struct ubi_ainf_volume *av)
 {
 	struct rb_node *rb;
-	struct ubi_ainf_peb *aeb;
+	struct ubi_ainf_leb *aeb;
 
 	dbg_bld("remove attaching information about volume %d", av->vol_id);
 
 	while ((rb = rb_first(&av->root))) {
-		aeb = rb_entry(rb, struct ubi_ainf_peb, u.rb);
-		rb_erase(&aeb->u.rb, &av->root);
-		list_add_tail(&aeb->u.list, &ai->erase);
+		aeb = rb_entry(rb, struct ubi_ainf_leb, rb);
+		rb_erase(&aeb->rb, &av->root);
+		if (--aeb->peb->refcount <= 0)
+			list_move(&aeb->peb->list, &ai->erase);
+
+		kmem_cache_free(ai->aleb_slab_cache, aeb);
 	}
 
 	rb_erase(&av->rb, &ai->volumes);
@@ -713,8 +738,8 @@ struct ubi_ainf_peb *ubi_early_get_peb(struct ubi_device *ubi,
 	struct ubi_ainf_peb *aeb, *tmp_aeb;
 
 	if (!list_empty(&ai->free)) {
-		aeb = list_entry(ai->free.next, struct ubi_ainf_peb, u.list);
-		list_del(&aeb->u.list);
+		aeb = list_entry(ai->free.next, struct ubi_ainf_peb, list);
+		list_del(&aeb->list);
 		dbg_bld("return free PEB %d, EC %d", aeb->pnum, aeb->ec);
 		return aeb;
 	}
@@ -725,7 +750,7 @@ struct ubi_ainf_peb *ubi_early_get_peb(struct ubi_device *ubi,
 	 * so forth. We don't want to take care about bad eraseblocks here -
 	 * they'll be handled later.
 	 */
-	list_for_each_entry_safe(aeb, tmp_aeb, &ai->erase, u.list) {
+	list_for_each_entry_safe(aeb, tmp_aeb, &ai->erase, list) {
 		if (aeb->ec == UBI_UNKNOWN)
 			aeb->ec = ai->mean_ec;
 
@@ -734,7 +759,7 @@ struct ubi_ainf_peb *ubi_early_get_peb(struct ubi_device *ubi,
 			continue;
 
 		aeb->ec += 1;
-		list_del(&aeb->u.list);
+		list_del(&aeb->list);
 		dbg_bld("return PEB %d, EC %d", aeb->pnum, aeb->ec);
 		return aeb;
 	}
@@ -820,7 +845,9 @@ static int scan_peb(struct ubi_device *ubi, struct ubi_attach_info *ai,
 		    int pnum, int *vid, unsigned long long *sqnum)
 {
 	long long uninitialized_var(ec);
-	int err, bitflips = 0, vol_id = -1, ec_err = 0;
+	int err, bitflips = 0, vol_id = -1, ec_err = 0, nvidh, i;
+	struct ubi_ainf_peb *aeb;
+	bool full = false;
 
 	dbg_bld("scan PEB %d", pnum);
 
@@ -844,12 +871,10 @@ static int scan_peb(struct ubi_device *ubi, struct ubi_attach_info *ai,
 		break;
 	case UBI_IO_FF:
 		ai->empty_peb_count += 1;
-		return add_to_list(ai, pnum, UBI_UNKNOWN, UBI_UNKNOWN,
-				   UBI_UNKNOWN, 0, &ai->erase);
+		return add_to_list(ai, pnum, UBI_UNKNOWN, 0, &ai->erase);
 	case UBI_IO_FF_BITFLIPS:
 		ai->empty_peb_count += 1;
-		return add_to_list(ai, pnum, UBI_UNKNOWN, UBI_UNKNOWN,
-				   UBI_UNKNOWN, 1, &ai->erase);
+		return add_to_list(ai, pnum, UBI_UNKNOWN, 1, &ai->erase);
 	case UBI_IO_BAD_HDR_EBADMSG:
 	case UBI_IO_BAD_HDR:
 		/*
@@ -915,8 +940,8 @@ static int scan_peb(struct ubi_device *ubi, struct ubi_attach_info *ai,
 	}
 
 	/* OK, we've done with the EC header, let's look at the VID header */
-
-	err = ubi_io_read_vid_hdr(ubi, pnum, vidh, 0);
+	nvidh = ubi->lebs_per_cpeb;
+	err = ubi_io_read_vid_hdrs(ubi, pnum, vidh, &nvidh, 0);
 	if (err < 0)
 		return err;
 	switch (err) {
@@ -960,8 +985,7 @@ static int scan_peb(struct ubi_device *ubi, struct ubi_attach_info *ai,
 			return err;
 		else if (!err)
 			/* This corruption is caused by a power cut */
-			err = add_to_list(ai, pnum, UBI_UNKNOWN,
-					  UBI_UNKNOWN, ec, 1, &ai->erase);
+			err = add_to_list(ai, pnum, ec, 1, &ai->erase);
 		else
 			/* This is an unexpected corruption */
 			err = add_corrupted(ai, pnum, ec);
@@ -969,23 +993,20 @@ static int scan_peb(struct ubi_device *ubi, struct ubi_attach_info *ai,
 			return err;
 		goto adjust_mean_ec;
 	case UBI_IO_FF_BITFLIPS:
-		err = add_to_list(ai, pnum, UBI_UNKNOWN, UBI_UNKNOWN,
-				  ec, 1, &ai->erase);
+		err = add_to_list(ai, pnum, ec, 1, &ai->erase);
 		if (err)
 			return err;
 		goto adjust_mean_ec;
 	case UBI_IO_FF:
 		if (ec_err || bitflips)
-			err = add_to_list(ai, pnum, UBI_UNKNOWN,
-					  UBI_UNKNOWN, ec, 1, &ai->erase);
+			err = add_to_list(ai, pnum, ec, 1, &ai->erase);
 		else
-			err = add_to_list(ai, pnum, UBI_UNKNOWN,
-					  UBI_UNKNOWN, ec, 0, &ai->free);
+			err = add_to_list(ai, pnum, ec, 0, &ai->free);
 		if (err)
 			return err;
 		goto adjust_mean_ec;
 	default:
-		ubi_err(ubi, "'ubi_io_read_vid_hdr()' returned unknown code %d",
+		ubi_err(ubi, "'ubi_io_read_vid_hdrs()' returned unknown code %d",
 			err);
 		return -EINVAL;
 	}
@@ -1006,8 +1027,7 @@ static int scan_peb(struct ubi_device *ubi, struct ubi_attach_info *ai,
 				ubi_msg(ubi, "\"delete\" compatible internal volume %d:%d found, will remove it",
 					vol_id, lnum);
 			}
-			err = add_to_list(ai, pnum, vol_id, lnum,
-					  ec, 1, &ai->erase);
+			err = add_to_list(ai, pnum, ec, 1, &ai->erase);
 			if (err)
 				return err;
 			return 0;
@@ -1021,8 +1041,7 @@ static int scan_peb(struct ubi_device *ubi, struct ubi_attach_info *ai,
 		case UBI_COMPAT_PRESERVE:
 			ubi_msg(ubi, "\"preserve\" compatible internal volume %d:%d found",
 				vol_id, lnum);
-			err = add_to_list(ai, pnum, vol_id, lnum,
-					  ec, 0, &ai->alien);
+			err = add_to_list(ai, pnum, ec, 0, &ai->alien);
 			if (err)
 				return err;
 			return 0;
@@ -1037,9 +1056,30 @@ static int scan_peb(struct ubi_device *ubi, struct ubi_attach_info *ai,
 	if (ec_err)
 		ubi_warn(ubi, "valid VID header but corrupted EC header at PEB %d",
 			 pnum);
-	err = ubi_add_to_av(ubi, ai, pnum, ec, vidh, bitflips);
-	if (err)
-		return err;
+
+	if (nvidh == 1) {
+		err = ubi_io_read(ubi, ech, pnum,
+				  ubi->peb_size - ubi->hdrs_min_io_size,
+				  ubi->hdrs_min_io_size);
+		if (!err && !ubi_check_pattern(ech, 0xff, ubi->hdrs_min_io_size))
+			full = true;
+	}
+
+	aeb = kmem_cache_zalloc(ai->apeb_slab_cache, GFP_KERNEL);
+	if (!aeb)
+		return -ENOMEM;
+
+	aeb->consolidated = nvidh > 1;
+	aeb->refcount = nvidh;
+	aeb->pnum = pnum;
+	aeb->ec = ec;
+	list_add_tail(&aeb->list, &ai->used);
+
+	for (i = 0; i < nvidh; i++) {
+		err = ubi_add_to_av(ubi, ai, aeb, &vidh[i], i, bitflips, full);
+		if (err)
+			return err;
+	}
 
 adjust_mean_ec:
 	if (!ec_err) {
@@ -1082,7 +1122,7 @@ static int late_analysis(struct ubi_device *ubi, struct ubi_attach_info *ai)
 		ubi_err(ubi, "%d PEBs are corrupted and preserved",
 			ai->corr_peb_count);
 		pr_err("Corrupted PEBs are:");
-		list_for_each_entry(aeb, &ai->corr, u.list)
+		list_for_each_entry(aeb, &ai->corr, list)
 			pr_cont(" %d", aeb->pnum);
 		pr_cont("\n");
 
@@ -1136,7 +1176,7 @@ static int late_analysis(struct ubi_device *ubi, struct ubi_attach_info *ai)
  */
 static void destroy_av(struct ubi_attach_info *ai, struct ubi_ainf_volume *av)
 {
-	struct ubi_ainf_peb *aeb;
+	struct ubi_ainf_leb *aeb;
 	struct rb_node *this = av->root.rb_node;
 
 	while (this) {
@@ -1145,16 +1185,16 @@ static void destroy_av(struct ubi_attach_info *ai, struct ubi_ainf_volume *av)
 		else if (this->rb_right)
 			this = this->rb_right;
 		else {
-			aeb = rb_entry(this, struct ubi_ainf_peb, u.rb);
+			aeb = rb_entry(this, struct ubi_ainf_leb, rb);
 			this = rb_parent(this);
 			if (this) {
-				if (this->rb_left == &aeb->u.rb)
+				if (this->rb_left == &aeb->rb)
 					this->rb_left = NULL;
 				else
 					this->rb_right = NULL;
 			}
 
-			kmem_cache_free(ai->aeb_slab_cache, aeb);
+			kmem_cache_free(ai->aleb_slab_cache, aeb);
 		}
 	}
 	kfree(av);
@@ -1170,23 +1210,6 @@ static void destroy_ai(struct ubi_attach_info *ai)
 	struct ubi_ainf_volume *av;
 	struct rb_node *rb;
 
-	list_for_each_entry_safe(aeb, aeb_tmp, &ai->alien, u.list) {
-		list_del(&aeb->u.list);
-		kmem_cache_free(ai->aeb_slab_cache, aeb);
-	}
-	list_for_each_entry_safe(aeb, aeb_tmp, &ai->erase, u.list) {
-		list_del(&aeb->u.list);
-		kmem_cache_free(ai->aeb_slab_cache, aeb);
-	}
-	list_for_each_entry_safe(aeb, aeb_tmp, &ai->corr, u.list) {
-		list_del(&aeb->u.list);
-		kmem_cache_free(ai->aeb_slab_cache, aeb);
-	}
-	list_for_each_entry_safe(aeb, aeb_tmp, &ai->free, u.list) {
-		list_del(&aeb->u.list);
-		kmem_cache_free(ai->aeb_slab_cache, aeb);
-	}
-
 	/* Destroy the volume RB-tree */
 	rb = ai->volumes.rb_node;
 	while (rb) {
@@ -1209,7 +1232,29 @@ static void destroy_ai(struct ubi_attach_info *ai)
 		}
 	}
 
-	kmem_cache_destroy(ai->aeb_slab_cache);
+	list_for_each_entry_safe(aeb, aeb_tmp, &ai->alien, list) {
+		list_del(&aeb->list);
+		kmem_cache_free(ai->apeb_slab_cache, aeb);
+	}
+	list_for_each_entry_safe(aeb, aeb_tmp, &ai->erase, list) {
+		list_del(&aeb->list);
+		kmem_cache_free(ai->apeb_slab_cache, aeb);
+	}
+	list_for_each_entry_safe(aeb, aeb_tmp, &ai->corr, list) {
+		list_del(&aeb->list);
+		kmem_cache_free(ai->apeb_slab_cache, aeb);
+	}
+	list_for_each_entry_safe(aeb, aeb_tmp, &ai->free, list) {
+		list_del(&aeb->list);
+		kmem_cache_free(ai->apeb_slab_cache, aeb);
+	}
+	list_for_each_entry_safe(aeb, aeb_tmp, &ai->used, list) {
+		list_del(&aeb->list);
+		kmem_cache_free(ai->apeb_slab_cache, aeb);
+	}
+
+	kmem_cache_destroy(ai->apeb_slab_cache);
+	kmem_cache_destroy(ai->aleb_slab_cache);
 	kfree(ai);
 }
 
@@ -1227,8 +1272,6 @@ static int scan_all(struct ubi_device *ubi, struct ubi_attach_info *ai,
 		    int start)
 {
 	int err, pnum;
-	struct rb_node *rb1, *rb2;
-	struct ubi_ainf_volume *av;
 	struct ubi_ainf_peb *aeb;
 
 	err = -ENOMEM;
@@ -1264,22 +1307,20 @@ static int scan_all(struct ubi_device *ubi, struct ubi_attach_info *ai,
 	 * In case of unknown erase counter we use the mean erase counter
 	 * value.
 	 */
-	ubi_rb_for_each_entry(rb1, av, &ai->volumes, rb) {
-		ubi_rb_for_each_entry(rb2, aeb, &av->root, u.rb)
-			if (aeb->ec == UBI_UNKNOWN)
-				aeb->ec = ai->mean_ec;
-	}
 
-	list_for_each_entry(aeb, &ai->free, u.list) {
+	list_for_each_entry(aeb, &ai->erase, list)
+		if (aeb->ec == UBI_UNKNOWN)
+			aeb->ec = ai->mean_ec;
+
+	list_for_each_entry(aeb, &ai->free, list)
 		if (aeb->ec == UBI_UNKNOWN)
 			aeb->ec = ai->mean_ec;
-	}
 
-	list_for_each_entry(aeb, &ai->corr, u.list)
+	list_for_each_entry(aeb, &ai->corr, list)
 		if (aeb->ec == UBI_UNKNOWN)
 			aeb->ec = ai->mean_ec;
 
-	list_for_each_entry(aeb, &ai->erase, u.list)
+	list_for_each_entry(aeb, &ai->erase, list)
 		if (aeb->ec == UBI_UNKNOWN)
 			aeb->ec = ai->mean_ec;
 
@@ -1311,16 +1352,28 @@ static struct ubi_attach_info *alloc_ai(void)
 	INIT_LIST_HEAD(&ai->free);
 	INIT_LIST_HEAD(&ai->erase);
 	INIT_LIST_HEAD(&ai->alien);
+	INIT_LIST_HEAD(&ai->used);
 	ai->volumes = RB_ROOT;
-	ai->aeb_slab_cache = kmem_cache_create("ubi_aeb_slab_cache",
+	ai->apeb_slab_cache = kmem_cache_create("ubi_apeb_slab_cache",
 					       sizeof(struct ubi_ainf_peb),
 					       0, 0, NULL);
-	if (!ai->aeb_slab_cache) {
-		kfree(ai);
-		ai = NULL;
-	}
+	if (!ai->apeb_slab_cache)
+		goto err_free_ai;
+
+	ai->aleb_slab_cache = kmem_cache_create("ubi_aleb_slab_cache",
+					       sizeof(struct ubi_ainf_leb),
+					       0, 0, NULL);
+	if (!ai->aleb_slab_cache)
+		goto err_destroy_apeb_cache;
 
 	return ai;
+
+err_destroy_apeb_cache:
+	kmem_cache_destroy(ai->apeb_slab_cache);
+err_free_ai:
+	kfree(ai);
+
+	return NULL;
 }
 
 #ifdef CONFIG_MTD_UBI_FASTMAP
@@ -1451,10 +1504,14 @@ int ubi_attach(struct ubi_device *ubi, int force_scan)
 	if (err)
 		goto out_vtbl;
 
-	err = ubi_eba_init(ubi, ai);
+	err = ubi_conso_init(ubi);
 	if (err)
 		goto out_wl;
 
+	err = ubi_eba_init(ubi, ai);
+	if (err)
+		goto out_conso;
+
 #ifdef CONFIG_MTD_UBI_FASTMAP
 	if (ubi->fm && ubi_dbg_chk_fastmap(ubi)) {
 		struct ubi_attach_info *scan_ai;
@@ -1482,6 +1539,8 @@ int ubi_attach(struct ubi_device *ubi, int force_scan)
 	destroy_ai(ai);
 	return 0;
 
+out_conso:
+	ubi_conso_close(ubi);
 out_wl:
 	ubi_wl_close(ubi);
 out_vtbl:
@@ -1505,7 +1564,8 @@ static int self_check_ai(struct ubi_device *ubi, struct ubi_attach_info *ai)
 	int pnum, err, vols_found = 0;
 	struct rb_node *rb1, *rb2;
 	struct ubi_ainf_volume *av;
-	struct ubi_ainf_peb *aeb, *last_aeb;
+	struct ubi_ainf_peb *peb;
+	struct ubi_ainf_leb *leb, *last_leb;
 	uint8_t *buf;
 
 	if (!ubi_dbg_chk_gen(ubi))
@@ -1556,38 +1616,39 @@ static int self_check_ai(struct ubi_device *ubi, struct ubi_attach_info *ai)
 			goto bad_av;
 		}
 
-		last_aeb = NULL;
-		ubi_rb_for_each_entry(rb2, aeb, &av->root, u.rb) {
+		last_leb = NULL;
+		ubi_rb_for_each_entry(rb2, leb, &av->root, rb) {
 			cond_resched();
 
-			last_aeb = aeb;
+			last_leb = leb;
 			leb_count += 1;
+			peb = leb->peb;
 
-			if (aeb->pnum < 0 || aeb->ec < 0) {
+			if (peb->pnum < 0 || peb->ec < 0) {
 				ubi_err(ubi, "negative values");
 				goto bad_aeb;
 			}
 
-			if (aeb->ec < ai->min_ec) {
+			if (peb->ec < ai->min_ec) {
 				ubi_err(ubi, "bad ai->min_ec (%d), %d found",
-					ai->min_ec, aeb->ec);
+					ai->min_ec, peb->ec);
 				goto bad_aeb;
 			}
 
-			if (aeb->ec > ai->max_ec) {
+			if (peb->ec > ai->max_ec) {
 				ubi_err(ubi, "bad ai->max_ec (%d), %d found",
-					ai->max_ec, aeb->ec);
+					ai->max_ec, peb->ec);
 				goto bad_aeb;
 			}
 
-			if (aeb->pnum >= ubi->peb_count) {
+			if (peb->pnum >= ubi->peb_count) {
 				ubi_err(ubi, "too high PEB number %d, total PEBs %d",
-					aeb->pnum, ubi->peb_count);
+					peb->pnum, ubi->peb_count);
 				goto bad_aeb;
 			}
 
 			if (av->vol_type == UBI_STATIC_VOLUME) {
-				if (aeb->lnum >= av->used_ebs) {
+				if (leb->desc.lnum >= av->used_ebs) {
 					ubi_err(ubi, "bad lnum or used_ebs");
 					goto bad_aeb;
 				}
@@ -1598,7 +1659,7 @@ static int self_check_ai(struct ubi_device *ubi, struct ubi_attach_info *ai)
 				}
 			}
 
-			if (aeb->lnum > av->highest_lnum) {
+			if (leb->desc.lnum > av->highest_lnum) {
 				ubi_err(ubi, "incorrect highest_lnum or lnum");
 				goto bad_aeb;
 			}
@@ -1610,12 +1671,12 @@ static int self_check_ai(struct ubi_device *ubi, struct ubi_attach_info *ai)
 			goto bad_av;
 		}
 
-		if (!last_aeb)
+		if (!last_leb)
 			continue;
 
-		aeb = last_aeb;
+		leb = last_leb;
 
-		if (aeb->lnum != av->highest_lnum) {
+		if (leb->desc.lnum != av->highest_lnum) {
 			ubi_err(ubi, "bad highest_lnum");
 			goto bad_aeb;
 		}
@@ -1629,15 +1690,19 @@ static int self_check_ai(struct ubi_device *ubi, struct ubi_attach_info *ai)
 
 	/* Check that attaching information is correct */
 	ubi_rb_for_each_entry(rb1, av, &ai->volumes, rb) {
-		last_aeb = NULL;
-		ubi_rb_for_each_entry(rb2, aeb, &av->root, u.rb) {
+		struct ubi_vid_hdr *vh;
+
+		last_leb = NULL;
+		ubi_rb_for_each_entry(rb2, leb, &av->root, rb) {
 			int vol_type;
+			int nvidh = ubi->lebs_per_cpeb;
 
 			cond_resched();
 
-			last_aeb = aeb;
+			last_leb = leb;
+			peb = leb->peb;
 
-			err = ubi_io_read_vid_hdr(ubi, aeb->pnum, vidh, 1);
+			err = ubi_io_read_vid_hdrs(ubi, peb->pnum, vidh, &nvidh, 1);
 			if (err && err != UBI_IO_BITFLIPS) {
 				ubi_err(ubi, "VID header is not OK (%d)",
 					err);
@@ -1646,53 +1711,56 @@ static int self_check_ai(struct ubi_device *ubi, struct ubi_attach_info *ai)
 				return err;
 			}
 
-			vol_type = vidh->vol_type == UBI_VID_DYNAMIC ?
+			ubi_assert(leb->peb_pos < nvidh);
+			vh = &vidh[leb->peb_pos];
+
+			vol_type = vh->vol_type == UBI_VID_DYNAMIC ?
 				   UBI_DYNAMIC_VOLUME : UBI_STATIC_VOLUME;
 			if (av->vol_type != vol_type) {
 				ubi_err(ubi, "bad vol_type");
 				goto bad_vid_hdr;
 			}
 
-			if (aeb->sqnum != be64_to_cpu(vidh->sqnum)) {
-				ubi_err(ubi, "bad sqnum %llu", aeb->sqnum);
+			if (leb->sqnum != be64_to_cpu(vh->sqnum)) {
+				ubi_err(ubi, "bad sqnum %llu", leb->sqnum);
 				goto bad_vid_hdr;
 			}
 
-			if (av->vol_id != be32_to_cpu(vidh->vol_id)) {
+			if (av->vol_id != be32_to_cpu(vh->vol_id)) {
 				ubi_err(ubi, "bad vol_id %d", av->vol_id);
 				goto bad_vid_hdr;
 			}
 
-			if (av->compat != vidh->compat) {
-				ubi_err(ubi, "bad compat %d", vidh->compat);
+			if (av->compat != vh->compat) {
+				ubi_err(ubi, "bad compat %d", vh->compat);
 				goto bad_vid_hdr;
 			}
 
-			if (aeb->lnum != be32_to_cpu(vidh->lnum)) {
-				ubi_err(ubi, "bad lnum %d", aeb->lnum);
+			if (leb->desc.lnum != be32_to_cpu(vh->lnum)) {
+				ubi_err(ubi, "bad lnum %d", leb->desc.lnum);
 				goto bad_vid_hdr;
 			}
 
-			if (av->used_ebs != be32_to_cpu(vidh->used_ebs)) {
+			if (av->used_ebs != be32_to_cpu(vh->used_ebs)) {
 				ubi_err(ubi, "bad used_ebs %d", av->used_ebs);
 				goto bad_vid_hdr;
 			}
 
-			if (av->data_pad != be32_to_cpu(vidh->data_pad)) {
+			if (av->data_pad != be32_to_cpu(vh->data_pad)) {
 				ubi_err(ubi, "bad data_pad %d", av->data_pad);
 				goto bad_vid_hdr;
 			}
 		}
 
-		if (!last_aeb)
+		if (!last_leb)
 			continue;
 
-		if (av->highest_lnum != be32_to_cpu(vidh->lnum)) {
+		if (av->highest_lnum != be32_to_cpu(vh->lnum)) {
 			ubi_err(ubi, "bad highest_lnum %d", av->highest_lnum);
 			goto bad_vid_hdr;
 		}
 
-		if (av->last_data_size != be32_to_cpu(vidh->data_size)) {
+		if (av->last_data_size != be32_to_cpu(vh->data_size)) {
 			ubi_err(ubi, "bad last_data_size %d",
 				av->last_data_size);
 			goto bad_vid_hdr;
@@ -1716,21 +1784,20 @@ static int self_check_ai(struct ubi_device *ubi, struct ubi_attach_info *ai)
 			buf[pnum] = 1;
 	}
 
-	ubi_rb_for_each_entry(rb1, av, &ai->volumes, rb)
-		ubi_rb_for_each_entry(rb2, aeb, &av->root, u.rb)
-			buf[aeb->pnum] = 1;
+	list_for_each_entry(peb, &ai->used, list)
+		buf[peb->pnum] = 1;
 
-	list_for_each_entry(aeb, &ai->free, u.list)
-		buf[aeb->pnum] = 1;
+	list_for_each_entry(peb, &ai->free, list)
+		buf[peb->pnum] = 1;
 
-	list_for_each_entry(aeb, &ai->corr, u.list)
-		buf[aeb->pnum] = 1;
+	list_for_each_entry(peb, &ai->corr, list)
+		buf[peb->pnum] = 1;
 
-	list_for_each_entry(aeb, &ai->erase, u.list)
-		buf[aeb->pnum] = 1;
+	list_for_each_entry(peb, &ai->erase, list)
+		buf[peb->pnum] = 1;
 
-	list_for_each_entry(aeb, &ai->alien, u.list)
-		buf[aeb->pnum] = 1;
+	list_for_each_entry(peb, &ai->alien, list)
+		buf[peb->pnum] = 1;
 
 	err = 0;
 	for (pnum = 0; pnum < ubi->peb_count; pnum++)
@@ -1745,8 +1812,8 @@ static int self_check_ai(struct ubi_device *ubi, struct ubi_attach_info *ai)
 	return 0;
 
 bad_aeb:
-	ubi_err(ubi, "bad attaching information about LEB %d", aeb->lnum);
-	ubi_dump_aeb(aeb, 0);
+	ubi_err(ubi, "bad attaching information about LEB %d", leb->desc.lnum);
+	ubi_dump_aeb(leb, 0);
 	ubi_dump_av(av);
 	goto out;
 
diff --git a/drivers/mtd/ubi/build.c b/drivers/mtd/ubi/build.c
index 99e31ed..306d71f 100644
--- a/drivers/mtd/ubi/build.c
+++ b/drivers/mtd/ubi/build.c
@@ -368,7 +368,9 @@ static ssize_t dev_attribute_show(struct device *dev,
 	if (attr == &dev_eraseblock_size)
 		ret = sprintf(buf, "%d\n", ubi->leb_size);
 	else if (attr == &dev_avail_eraseblocks)
-		ret = sprintf(buf, "%d\n", ubi->avail_pebs);
+		ret = sprintf(buf, "%d\n",
+			      ubi->avail_pebs *
+			      ubi->lebs_per_cpeb);
 	else if (attr == &dev_total_eraseblocks)
 		ret = sprintf(buf, "%d\n", ubi->good_peb_count);
 	else if (attr == &dev_volumes_count)
@@ -653,7 +655,7 @@ static int io_init(struct ubi_device *ubi, int max_beb_per1024)
 	ubi->consolidated_peb_size = ubi->mtd->erasesize;
 	ubi->peb_size   = ubi->consolidated_peb_size /
 			  mtd_pairing_groups_per_eb(ubi->mtd);
-	ubi->lebs_per_consolidated_peb = mtd_pairing_groups_per_eb(ubi->mtd);
+	ubi->lebs_per_cpeb = mtd_pairing_groups_per_eb(ubi->mtd);
 	ubi->peb_count  = mtd_div_by_eb(ubi->mtd->size, ubi->mtd);
 	ubi->flash_size = ubi->mtd->size;
 
@@ -797,7 +799,7 @@ static int autoresize(struct ubi_device *ubi, int vol_id)
 {
 	struct ubi_volume_desc desc;
 	struct ubi_volume *vol = ubi->volumes[vol_id];
-	int err, old_reserved_pebs = vol->reserved_pebs;
+	int err, old_reserved_lebs = vol->reserved_lebs;
 
 	if (ubi->ro_mode) {
 		ubi_warn(ubi, "skip auto-resize because of R/O mode");
@@ -824,9 +826,11 @@ static int autoresize(struct ubi_device *ubi, int vol_id)
 			ubi_err(ubi, "cannot clean auto-resize flag for volume %d",
 				vol_id);
 	} else {
+		int avail_lebs = ubi->avail_pebs *
+				 ubi->lebs_per_cpeb;
+
 		desc.vol = vol;
-		err = ubi_resize_volume(&desc,
-					old_reserved_pebs + ubi->avail_pebs);
+		err = ubi_resize_volume(&desc, old_reserved_lebs + avail_lebs);
 		if (err)
 			ubi_err(ubi, "cannot auto-resize volume %d",
 				vol_id);
@@ -836,7 +840,7 @@ static int autoresize(struct ubi_device *ubi, int vol_id)
 		return err;
 
 	ubi_msg(ubi, "volume %d (\"%s\") re-sized from %d to %d LEBs",
-		vol_id, vol->name, old_reserved_pebs, vol->reserved_pebs);
+		vol_id, vol->name, old_reserved_lebs, vol->reserved_lebs);
 	return 0;
 }
 
@@ -1026,7 +1030,7 @@ int ubi_attach_mtd_dev(struct mtd_info *mtd, int ubi_num,
 		ubi->image_seq);
 	ubi_msg(ubi, "available PEBs: %d, total reserved PEBs: %d, PEBs reserved for bad PEB handling: %d",
 		ubi->avail_pebs, ubi->rsvd_pebs, ubi->beb_rsvd_pebs);
-	ubi_msg(ubi, "LEBs per PEB: %d", ubi->lebs_per_consolidated_peb);
+	ubi_msg(ubi, "LEBs per PEB: %d", ubi->lebs_per_cpeb);
 
 	/*
 	 * The below lock makes sure we do not race with 'ubi_thread()' which
diff --git a/drivers/mtd/ubi/cdev.c b/drivers/mtd/ubi/cdev.c
index 833c0a82..230232f 100644
--- a/drivers/mtd/ubi/cdev.c
+++ b/drivers/mtd/ubi/cdev.c
@@ -415,7 +415,7 @@ static long vol_cdev_ioctl(struct file *file, unsigned int cmd,
 			break;
 		}
 
-		rsvd_bytes = (long long)vol->reserved_pebs *
+		rsvd_bytes = (long long)vol->reserved_lebs *
 					ubi->leb_size-vol->data_pad;
 		if (bytes < 0 || bytes > rsvd_bytes) {
 			err = -EINVAL;
@@ -454,7 +454,7 @@ static long vol_cdev_ioctl(struct file *file, unsigned int cmd,
 
 		/* Validate the request */
 		err = -EINVAL;
-		if (req.lnum < 0 || req.lnum >= vol->reserved_pebs ||
+		if (req.lnum < 0 || req.lnum >= vol->reserved_lebs ||
 		    req.bytes < 0 || req.bytes > vol->usable_leb_size)
 			break;
 
@@ -485,7 +485,7 @@ static long vol_cdev_ioctl(struct file *file, unsigned int cmd,
 			break;
 		}
 
-		if (lnum < 0 || lnum >= vol->reserved_pebs) {
+		if (lnum < 0 || lnum >= vol->reserved_lebs) {
 			err = -EINVAL;
 			break;
 		}
@@ -909,7 +909,7 @@ static long ubi_cdev_ioctl(struct file *file, unsigned int cmd,
 	/* Re-size volume command */
 	case UBI_IOCRSVOL:
 	{
-		int pebs;
+		int lebs;
 		struct ubi_rsvol_req req;
 
 		dbg_gen("re-size volume");
@@ -929,11 +929,11 @@ static long ubi_cdev_ioctl(struct file *file, unsigned int cmd,
 			break;
 		}
 
-		pebs = div_u64(req.bytes + desc->vol->usable_leb_size - 1,
+		lebs = div_u64(req.bytes + desc->vol->usable_leb_size - 1,
 			       desc->vol->usable_leb_size);
 
 		mutex_lock(&ubi->device_mutex);
-		err = ubi_resize_volume(desc, pebs);
+		err = ubi_resize_volume(desc, lebs);
 		mutex_unlock(&ubi->device_mutex);
 		ubi_close_volume(desc);
 		break;
diff --git a/drivers/mtd/ubi/consolidate.c b/drivers/mtd/ubi/consolidate.c
new file mode 100644
index 0000000..de1479d
--- /dev/null
+++ b/drivers/mtd/ubi/consolidate.c
@@ -0,0 +1,499 @@
+#include <linux/slab.h>
+#include <linux/crc32.h>
+#include "ubi.h"
+
+static void consolidation_unlock(struct ubi_device *ubi,
+				 struct ubi_leb_desc *clebs)
+{
+	int i;
+
+	for (i = 0; i < ubi->lebs_per_cpeb; i++)
+		ubi_eba_leb_write_unlock(ubi, clebs[i].vol_id, clebs[i].lnum);
+}
+
+static int find_consolidable_lebs(struct ubi_device *ubi,
+				  struct ubi_leb_desc *clebs,
+				  struct ubi_volume **vols)
+{
+	struct ubi_full_leb *fleb;
+	LIST_HEAD(found);
+	int i, err = 0;
+
+	spin_lock(&ubi->full_lock);
+	if (ubi->full_count < ubi->lebs_per_cpeb)
+		err = -EAGAIN;
+	spin_unlock(&ubi->full_lock);
+	if (err)
+		return err;
+
+	for (i = 0; i < ubi->lebs_per_cpeb;) {
+		spin_lock(&ubi->full_lock);
+		fleb = list_first_entry_or_null(&ubi->full,
+						struct ubi_full_leb, node);
+		spin_unlock(&ubi->full_lock);
+
+		if (!fleb) {
+			err = -EAGAIN;
+			goto err;
+		} else {
+			list_del_init(&fleb->node);
+			list_add_tail(&fleb->node, &found);
+			ubi->full_count--;
+		}
+
+		clebs[i] = fleb->desc;
+
+		err = ubi_eba_leb_write_lock_nested(ubi, clebs[i].vol_id,
+						    clebs[i].lnum, i);
+		if (err) {
+			spin_lock(&ubi->full_lock);
+			list_del(&fleb->node);
+			list_add_tail(&fleb->node, &ubi->full);
+			ubi->full_count++;
+			spin_unlock(&ubi->full_lock);
+			goto err;
+		}
+
+		spin_lock(&ubi->volumes_lock);
+		vols[i] = ubi->volumes[vol_id2idx(ubi, clebs[i].vol_id)];
+		spin_unlock(&ubi->volumes_lock);
+		/* volume vanished under us */
+		//TODO clarify/document when/why this can happen
+		if (!vols[i]) {
+			ubi_assert(0);
+			ubi_eba_leb_write_unlock(ubi, clebs[i].vol_id, clebs[i].lnum);
+			spin_lock(&ubi->full_lock);
+			list_del_init(&fleb->node);
+			kfree(fleb);
+			spin_unlock(&ubi->full_lock);
+			continue;
+		}
+
+		i++;
+	}
+
+	while(!list_empty(&found)) {
+		fleb = list_first_entry(&found, struct ubi_full_leb, node);
+		list_del(&fleb->node);
+		kfree(fleb);
+	}
+
+	if (i < ubi->lebs_per_cpeb - 1) {
+		return -EAGAIN;
+	}
+
+	return 0;
+
+err:
+	while(!list_empty(&found)) {
+		spin_lock(&ubi->full_lock);
+		fleb = list_first_entry(&found, struct ubi_full_leb, node);
+		list_del(&fleb->node);
+		list_add_tail(&fleb->node, &ubi->full);
+		ubi->full_count++;
+		spin_unlock(&ubi->full_lock);
+		ubi_eba_leb_write_unlock(ubi, fleb->desc.vol_id, fleb->desc.lnum);
+	}
+
+	return err;
+}
+
+static int consolidate_lebs(struct ubi_device *ubi)
+{
+	int i, pnum, offset = ubi->leb_start, err = 0;
+	struct ubi_vid_hdr *vid_hdrs;
+	struct ubi_leb_desc *clebs = NULL, *new_clebs = NULL;
+	struct ubi_volume **vols = NULL;
+	int *opnums = NULL;
+
+	if (!ubi_conso_consolidation_needed(ubi))
+		return 0;
+
+	vols = kzalloc(sizeof(*vols) * ubi->lebs_per_cpeb, GFP_KERNEL);
+	if (!vols)
+		return -ENOMEM;
+
+	opnums = kzalloc(sizeof(*opnums) * ubi->lebs_per_cpeb, GFP_KERNEL);
+	if (!opnums) {
+		err = -ENOMEM;
+		goto err_free_mem;
+	}
+
+	clebs = kzalloc(sizeof(*clebs) * ubi->lebs_per_cpeb, GFP_KERNEL);
+	if (!clebs) {
+		err = -ENOMEM;
+		goto err_free_mem;
+	}
+
+	new_clebs = kzalloc(sizeof(*clebs) * ubi->lebs_per_cpeb, GFP_KERNEL);
+	if (!new_clebs) {
+		err = -ENOMEM;
+		goto err_free_mem;
+	}
+
+	err = find_consolidable_lebs(ubi, clebs, vols);
+	if (err)
+		goto err_free_mem;
+
+	memcpy(new_clebs, clebs, sizeof(*clebs) * ubi->lebs_per_cpeb);
+
+	mutex_lock(&ubi->buf_mutex);
+
+	pnum = ubi_wl_get_peb(ubi, true);
+	if (pnum < 0) {
+		err = pnum;
+		//TODO cleanup exit path
+		mutex_unlock(&ubi->buf_mutex);
+		up_read(&ubi->fm_eba_sem);
+		goto err_unlock_lebs;
+	}
+
+	memset(ubi->peb_buf, 0, ubi->peb_size);
+	vid_hdrs = ubi->peb_buf + ubi->vid_hdr_aloffset + ubi->vid_hdr_shift;
+
+	for (i = 0; i < ubi->lebs_per_cpeb; i++) {
+		int vol_id = clebs[i].vol_id, lnum = clebs[i].lnum, lpos = clebs[i].lpos;
+		void *buf = ubi->peb_buf + offset;
+		struct ubi_volume *vol = vols[i];
+		int spnum;
+		int data_size;
+		u32 crc;
+		bool raw;
+
+		spnum = vol->eba_tbl[lnum];
+
+		/* we raced against leb unmap */
+		if (spnum == UBI_LEB_UNMAPPED) {
+			//TODO: should be fixed now and no longer trigger.
+			ubi_assert(0);
+			err = 0;
+			goto err_unlock_fm_eba;
+		}
+
+		if (ubi->consolidated[spnum]) {
+			ubi_assert(ubi_conso_invalidate_leb(ubi, spnum, vol_id, lnum) == true);
+			raw = true;
+		} else {
+			ubi_assert(!lpos);
+			raw = false;
+		}
+
+		ubi_assert(offset + ubi->leb_size < ubi->consolidated_peb_size);
+
+		if (!raw)
+			err = ubi_io_read(ubi, buf, spnum, ubi->leb_start, ubi->leb_size);
+		else
+			err = ubi_io_raw_read(ubi, buf, spnum, ubi->leb_start + (lpos * ubi->leb_size), ubi->leb_size);
+
+		if (err && err != UBI_IO_BITFLIPS)
+			goto err_unlock_fm_eba;
+
+		if (vol->vol_type == UBI_DYNAMIC_VOLUME) {
+			data_size = ubi->leb_size - vol->data_pad;
+			vid_hdrs[i].vol_type = UBI_VID_DYNAMIC;
+		} else {
+			int nvidh = ubi->lebs_per_cpeb;
+			struct ubi_vid_hdr *vh;
+
+			vh = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
+			if (!vh) {
+				err = -ENOMEM;
+				goto err_unlock_fm_eba;
+			}
+
+			err = ubi_io_read_vid_hdrs(ubi, spnum, vh, &nvidh, 0);
+			if (err && err != UBI_IO_BITFLIPS) {
+				ubi_free_vid_hdr(ubi, vh);
+				goto err_unlock_fm_eba;
+			}
+
+			ubi_free_vid_hdr(ubi, vh);
+
+			data_size = be32_to_cpu(vh[lpos].data_size);
+			vid_hdrs[i].vol_type = UBI_VID_STATIC;
+			vid_hdrs[i].used_ebs = cpu_to_be32(vol->used_ebs);
+		}
+
+		vid_hdrs[i].data_pad = cpu_to_be32(vol->data_pad);
+		vid_hdrs[i].sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
+		vid_hdrs[i].vol_id = cpu_to_be32(vol_id);
+		vid_hdrs[i].lnum = cpu_to_be32(lnum);
+		vid_hdrs[i].compat = ubi_get_compat(ubi, vol_id);
+		vid_hdrs[i].data_size = cpu_to_be32(data_size);
+		vid_hdrs[i].copy_flag = 1;
+		crc = crc32(UBI_CRC32_INIT, buf, data_size);
+		vid_hdrs[i].data_crc = cpu_to_be32(crc);
+		offset += ubi->leb_size;
+
+		new_clebs[i].lpos = i;
+	}
+
+	/*
+	 * Pad remaining pages with zeros to prevent problem on some MLC chip
+	 * that expect the whole block to be programmed in order to work
+	 * reliably (some Hynix chips are impacted).
+	 */
+	memset(ubi->peb_buf + offset, 0, ubi->consolidated_peb_size - offset);
+
+	err = ubi_io_write_vid_hdrs(ubi, pnum, vid_hdrs, ubi->lebs_per_cpeb);
+	if (err) {
+		ubi_warn(ubi, "failed to write VID headers to PEB %d",
+			 pnum);
+		goto err_unlock_lebs;
+	}
+
+	err = ubi_io_raw_write(ubi, ubi->peb_buf + ubi->leb_start,
+			       pnum, ubi->leb_start,
+			       ubi->consolidated_peb_size - ubi->leb_start);
+	if (err) {
+		ubi_warn(ubi, "failed to write %d bytes of data to PEB %d",
+			 ubi->consolidated_peb_size - ubi->leb_start, pnum);
+		goto err_unlock_fm_eba;
+	}
+
+	for (i = 0; i < ubi->lebs_per_cpeb; i++) {
+		struct ubi_volume *vol = vols[i];
+		int lnum = clebs[i].lnum;
+
+		opnums[i] = vol->eba_tbl[lnum];
+
+		vol->eba_tbl[lnum] = pnum;
+	}
+	ubi->consolidated[pnum] = new_clebs;
+
+	up_read(&ubi->fm_eba_sem);
+	mutex_unlock(&ubi->buf_mutex);
+	consolidation_unlock(ubi, clebs);
+
+	for (i = 0; i < ubi->lebs_per_cpeb; i++) {
+		//TODO set torture if needed
+		ubi_wl_put_peb(ubi, opnums[i], 0);
+	}
+
+	kfree(clebs);
+	kfree(opnums);
+	kfree(vols);
+
+	return 0;
+
+err_unlock_fm_eba:
+	mutex_unlock(&ubi->buf_mutex);
+	up_read(&ubi->fm_eba_sem);
+
+	for (i = 0; i < ubi->lebs_per_cpeb; i++)
+		ubi_coso_add_full_leb(ubi, clebs[i].vol_id, clebs[i].lnum, clebs[i].lpos);
+
+	ubi_wl_put_peb(ubi, pnum, 0);
+err_unlock_lebs:
+	consolidation_unlock(ubi, clebs);
+err_free_mem:
+	kfree(new_clebs);
+	kfree(clebs);
+	kfree(opnums);
+	kfree(vols);
+
+	return err;
+}
+
+static int consolidation_worker(struct ubi_device *ubi,
+				struct ubi_work *wrk,
+				int shutdown)
+{
+	int ret;
+
+	if (shutdown)
+		return 0;
+
+	ret = consolidate_lebs(ubi);
+	if (ret == -EAGAIN)
+		ret = 0;
+
+	ubi->conso_scheduled = 0;
+	smp_wmb();
+
+	if (ubi_conso_consolidation_needed(ubi))
+		ubi_conso_schedule(ubi);
+
+	return ret;
+}
+
+static bool consolidation_possible(struct ubi_device *ubi)
+{
+	if (ubi->lebs_per_cpeb < 2)
+		return false;
+
+	if (ubi->full_count < ubi->lebs_per_cpeb)
+		return false;
+
+	return true;
+}
+
+bool ubi_conso_consolidation_needed(struct ubi_device *ubi)
+{
+	if (!consolidation_possible(ubi))
+		return false;
+
+	return ubi->free_count - ubi->beb_rsvd_pebs <=
+	       ubi->consolidation_threshold;
+}
+
+void ubi_conso_schedule(struct ubi_device *ubi)
+{
+	struct ubi_work *wrk;
+
+	if (ubi->conso_scheduled)
+		return;
+
+	wrk = ubi_alloc_work(ubi);
+	if (wrk) {
+		ubi->conso_scheduled = 1;
+		smp_wmb();
+
+		wrk->func = &consolidation_worker;
+		INIT_LIST_HEAD(&wrk->list);
+		ubi_schedule_work(ubi, wrk);
+	} else
+		BUG();
+}
+
+void ubi_eba_consolidate(struct ubi_device *ubi)
+{
+	if (consolidation_possible(ubi) && ubi->consolidation_pnum >= 0)
+		ubi_conso_schedule(ubi);
+}
+
+void ubi_conso_remove_full_leb(struct ubi_device *ubi, int vol_id, int lnum)
+{
+	struct ubi_full_leb *fleb;
+
+	spin_lock(&ubi->full_lock);
+	list_for_each_entry(fleb, &ubi->full, node) {
+		if (fleb->desc.lnum == lnum && fleb->desc.vol_id == vol_id) {
+			ubi->full_count--;
+			list_del(&fleb->node);
+			kfree(fleb);
+			break;
+		}
+	}
+	spin_unlock(&ubi->full_lock);
+}
+
+struct ubi_leb_desc *
+ubi_conso_get_consolidated(struct ubi_device *ubi, int pnum)
+{
+	if (ubi->consolidated)
+		return ubi->consolidated[pnum];
+
+	return NULL;
+}
+
+int ubi_coso_add_full_leb(struct ubi_device *ubi, int vol_id, int lnum, int lpos)
+{
+	struct ubi_full_leb *fleb;
+
+	/*
+	 * We don't track full LEBs if we don't need to (which is the case
+	 * when UBI does not need or does not support LEB consolidation).
+	 */
+	if (!ubi->consolidated)
+		return 0;
+
+	fleb = kzalloc(sizeof(*fleb), GFP_KERNEL);
+	if (!fleb)
+		return -ENOMEM;
+
+	fleb->desc.vol_id = vol_id;
+	fleb->desc.lnum = lnum;
+	fleb->desc.lpos = lpos;
+
+	spin_lock(&ubi->full_lock);
+	list_add_tail(&fleb->node, &ubi->full);
+	ubi->full_count++;
+	spin_unlock(&ubi->full_lock);
+
+	return 0;
+}
+
+bool ubi_conso_invalidate_leb(struct ubi_device *ubi, int pnum,
+				   int vol_id, int lnum)
+{
+	struct ubi_leb_desc *clebs = NULL;
+
+	if (!ubi->consolidated)
+		return true;
+
+	clebs = ubi->consolidated[pnum];
+	if (!clebs)
+		return true;
+
+	//TODO: make this generic again
+	BUG_ON(ubi->lebs_per_cpeb > 2);
+
+	if (clebs[0].lnum == lnum && clebs[0].vol_id == vol_id) {
+		clebs[0].lnum = -1;
+		clebs[0].vol_id = -1;
+
+		if (clebs[1].lnum > -1 && clebs[1].vol_id > -1) {
+			ubi_coso_add_full_leb(ubi, clebs[1].vol_id, clebs[1].lnum, clebs[1].lpos);
+
+			return false;
+		}
+	} else if (clebs[1].lnum == lnum && clebs[1].vol_id == vol_id) {
+		clebs[1].lnum = -1;
+		clebs[1].vol_id = -1;
+
+		if (clebs[0].lnum > -1 && clebs[0].vol_id > -1) {
+			ubi_coso_add_full_leb(ubi, clebs[0].vol_id, clebs[0].lnum, clebs[0].lpos);
+
+			return false;
+		}
+	} else
+		ubi_assert(0);
+
+	ubi->consolidated[pnum] = NULL;
+	kfree(clebs);
+
+	return true;
+}
+
+int ubi_conso_init(struct ubi_device *ubi)
+{
+	spin_lock_init(&ubi->full_lock);
+	INIT_LIST_HEAD(&ubi->full);
+	ubi->full_count = 0;
+	ubi->consolidation_threshold = (ubi->avail_pebs + ubi->rsvd_pebs) / 3;
+
+	if (ubi->consolidation_threshold < ubi->lebs_per_cpeb)
+		ubi->consolidation_threshold = ubi->lebs_per_cpeb;
+
+	if (ubi->lebs_per_cpeb == 1)
+		return 0;
+
+	if (ubi->avail_pebs < UBI_CONSO_RESERVED_PEBS) {
+		ubi_err(ubi, "no enough physical eraseblocks (%d, need %d)",
+			ubi->avail_pebs, UBI_CONSO_RESERVED_PEBS);
+		if (ubi->corr_peb_count)
+			ubi_err(ubi, "%d PEBs are corrupted and not used",
+				ubi->corr_peb_count);
+		return -ENOSPC;
+	}
+
+	ubi->avail_pebs -= UBI_CONSO_RESERVED_PEBS;
+	ubi->rsvd_pebs += UBI_CONSO_RESERVED_PEBS;
+
+	return 0;
+}
+
+void ubi_conso_close(struct ubi_device *ubi)
+{
+	struct ubi_full_leb *fleb;
+
+	while(!list_empty(&ubi->full)) {
+		fleb = list_first_entry(&ubi->full, struct ubi_full_leb, node);
+		list_del(&fleb->node);
+		kfree(fleb);
+		ubi->full_count--;
+	}
+
+	ubi_assert(ubi->full_count == 0);
+}
diff --git a/drivers/mtd/ubi/debug.c b/drivers/mtd/ubi/debug.c
index ed23009..6178fa1 100644
--- a/drivers/mtd/ubi/debug.c
+++ b/drivers/mtd/ubi/debug.c
@@ -36,7 +36,7 @@ void ubi_dump_flash(struct ubi_device *ubi, int pnum, int offset, int len)
 	int err;
 	size_t read;
 	void *buf;
-	loff_t addr = (loff_t)pnum * ubi->peb_size + offset;
+	loff_t addr = (loff_t)pnum * ubi->consolidated_peb_size + offset;
 
 	buf = vmalloc(len);
 	if (!buf)
@@ -108,7 +108,7 @@ void ubi_dump_vol_info(const struct ubi_volume *vol)
 {
 	pr_err("Volume information dump:\n");
 	pr_err("\tvol_id          %d\n", vol->vol_id);
-	pr_err("\treserved_pebs   %d\n", vol->reserved_pebs);
+	pr_err("\treserved_lebs   %d\n", vol->reserved_lebs);
 	pr_err("\talignment       %d\n", vol->alignment);
 	pr_err("\tdata_pad        %d\n", vol->data_pad);
 	pr_err("\tvol_type        %d\n", vol->vol_type);
@@ -140,7 +140,7 @@ void ubi_dump_vtbl_record(const struct ubi_vtbl_record *r, int idx)
 	int name_len = be16_to_cpu(r->name_len);
 
 	pr_err("Volume table record %d dump:\n", idx);
-	pr_err("\treserved_pebs   %d\n", be32_to_cpu(r->reserved_pebs));
+	pr_err("\treserved_pebs   %d\n", be32_to_cpu(r->reserved_lebs));
 	pr_err("\talignment       %d\n", be32_to_cpu(r->alignment));
 	pr_err("\tdata_pad        %d\n", be32_to_cpu(r->data_pad));
 	pr_err("\tvol_type        %d\n", (int)r->vol_type);
@@ -185,14 +185,14 @@ void ubi_dump_av(const struct ubi_ainf_volume *av)
  * @aeb: the object to dump
  * @type: object type: 0 - not corrupted, 1 - corrupted
  */
-void ubi_dump_aeb(const struct ubi_ainf_peb *aeb, int type)
+void ubi_dump_aeb(const struct ubi_ainf_leb *aeb, int type)
 {
 	pr_err("eraseblock attaching information dump:\n");
-	pr_err("\tec       %d\n", aeb->ec);
-	pr_err("\tpnum     %d\n", aeb->pnum);
+	pr_err("\tec       %d\n", aeb->peb->ec);
+	pr_err("\tpnum     %d\n", aeb->peb->pnum);
 	if (type == 0) {
-		pr_err("\tlnum     %d\n", aeb->lnum);
-		pr_err("\tscrub    %d\n", aeb->scrub);
+		pr_err("\tlnum     %d\n", aeb->desc.lnum);
+		pr_err("\tscrub    %d\n", aeb->peb->scrub);
 		pr_err("\tsqnum    %llu\n", aeb->sqnum);
 	}
 }
diff --git a/drivers/mtd/ubi/debug.h b/drivers/mtd/ubi/debug.h
index 47c447d..c3ed0d5 100644
--- a/drivers/mtd/ubi/debug.h
+++ b/drivers/mtd/ubi/debug.h
@@ -56,7 +56,7 @@ void ubi_dump_vid_hdr(const struct ubi_vid_hdr *vid_hdr);
 void ubi_dump_vol_info(const struct ubi_volume *vol);
 void ubi_dump_vtbl_record(const struct ubi_vtbl_record *r, int idx);
 void ubi_dump_av(const struct ubi_ainf_volume *av);
-void ubi_dump_aeb(const struct ubi_ainf_peb *aeb, int type);
+void ubi_dump_aeb(const struct ubi_ainf_leb *aeb, int type);
 void ubi_dump_mkvol_req(const struct ubi_mkvol_req *req);
 int ubi_self_check_all_ff(struct ubi_device *ubi, int pnum, int offset,
 			  int len);
diff --git a/drivers/mtd/ubi/eba.c b/drivers/mtd/ubi/eba.c
index 229be7c..953091e 100644
--- a/drivers/mtd/ubi/eba.c
+++ b/drivers/mtd/ubi/eba.c
@@ -69,21 +69,6 @@ unsigned long long ubi_next_sqnum(struct ubi_device *ubi)
 }
 
 /**
- * ubi_get_compat - get compatibility flags of a volume.
- * @ubi: UBI device description object
- * @vol_id: volume ID
- *
- * This function returns compatibility flags for an internal volume. User
- * volumes have no compatibility flags, so %0 is returned.
- */
-static int ubi_get_compat(const struct ubi_device *ubi, int vol_id)
-{
-	if (vol_id == UBI_LAYOUT_VOLUME_ID)
-		return UBI_LAYOUT_VOLUME_COMPAT;
-	return 0;
-}
-
-/**
  * ltree_lookup - look up the lock tree.
  * @ubi: UBI device description object
  * @vol_id: volume ID
@@ -256,6 +241,31 @@ static int leb_write_lock(struct ubi_device *ubi, int vol_id, int lnum)
 }
 
 /**
+ * ubi_eba_leb_write_lock_nested - lock logical eraseblock for writing, allow nesting.
+ * @ubi: UBI device description object
+ * @vol_id: volume ID
+ * @lnum: logical eraseblock number
+ * @level: nesting level
+ *
+ * This function locks a logical eraseblock for consolidation.
+ * Returns zero in case of success and a negative error code in case
+ * of failure.
+ */
+int ubi_eba_leb_write_lock_nested(struct ubi_device *ubi, int vol_id, int lnum,
+				  int level)
+{
+	struct ubi_ltree_entry *le;
+
+	le = ltree_add_entry(ubi, vol_id, lnum);
+	if (IS_ERR(le))
+		return PTR_ERR(le);
+
+	down_write_nested(&le->mutex, level);
+
+	return 0;
+}
+
+/**
  * leb_write_lock - lock logical eraseblock for writing.
  * @ubi: UBI device description object
  * @vol_id: volume ID
@@ -295,7 +305,7 @@ static int leb_write_trylock(struct ubi_device *ubi, int vol_id, int lnum)
  * @vol_id: volume ID
  * @lnum: logical eraseblock number
  */
-static void leb_write_unlock(struct ubi_device *ubi, int vol_id, int lnum)
+void ubi_eba_leb_write_unlock(struct ubi_device *ubi, int vol_id, int lnum)
 {
 	struct ubi_ltree_entry *le;
 
@@ -311,6 +321,7 @@ static void leb_write_unlock(struct ubi_device *ubi, int vol_id, int lnum)
 	spin_unlock(&ubi->ltree_lock);
 }
 
+
 /**
  * ubi_eba_unmap_leb - un-map logical eraseblock.
  * @ubi: UBI device description object
@@ -325,6 +336,7 @@ int ubi_eba_unmap_leb(struct ubi_device *ubi, struct ubi_volume *vol,
 		      int lnum)
 {
 	int err, pnum, vol_id = vol->vol_id;
+	bool release_peb = false;
 
 	if (ubi->ro_mode)
 		return -EROFS;
@@ -342,11 +354,15 @@ int ubi_eba_unmap_leb(struct ubi_device *ubi, struct ubi_volume *vol,
 
 	down_read(&ubi->fm_eba_sem);
 	vol->eba_tbl[lnum] = UBI_LEB_UNMAPPED;
+	release_peb = ubi_conso_invalidate_leb(ubi, pnum, vol_id, lnum);
 	up_read(&ubi->fm_eba_sem);
-	err = ubi_wl_put_peb(ubi, pnum, 0);
+	ubi_conso_remove_full_leb(ubi, vol_id, lnum);
 
 out_unlock:
-	leb_write_unlock(ubi, vol_id, lnum);
+	ubi_eba_leb_write_unlock(ubi, vol_id, lnum);
+	if (release_peb)
+		err = ubi_wl_put_peb(ubi, pnum, 0);
+
 	return err;
 }
 
@@ -372,9 +388,10 @@ out_unlock:
 int ubi_eba_read_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
 		     void *buf, int offset, int len, int check)
 {
-	int err, pnum, scrub = 0, vol_id = vol->vol_id;
+	int err, pnum, scrub = 0, vol_id = vol->vol_id, loffs = 0, lpos = 0;
 	struct ubi_vid_hdr *vid_hdr;
 	uint32_t uninitialized_var(crc);
+	struct ubi_leb_desc *clebs;
 
 	err = leb_read_lock(ubi, vol_id, lnum);
 	if (err)
@@ -401,15 +418,31 @@ int ubi_eba_read_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
 	if (vol->vol_type == UBI_DYNAMIC_VOLUME)
 		check = 0;
 
+	clebs = ubi_conso_get_consolidated(ubi, pnum);
+	if (clebs) {
+		for (; lpos < ubi->lebs_per_cpeb; lpos++) {
+			if (clebs[lpos].vol_id == vol->vol_id &&
+			    clebs[lpos].lnum == lnum)
+				break;
+		}
+
+		if (lpos == ubi->lebs_per_cpeb)
+			return -EINVAL;
+
+		loffs = ubi->leb_start + (lpos * ubi->leb_size);
+	}
+
 retry:
 	if (check) {
+		int nvidh = ubi->lebs_per_cpeb;
+
 		vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
 		if (!vid_hdr) {
 			err = -ENOMEM;
 			goto out_unlock;
 		}
 
-		err = ubi_io_read_vid_hdr(ubi, pnum, vid_hdr, 1);
+		err = ubi_io_read_vid_hdrs(ubi, pnum, vid_hdr, &nvidh, 1);
 		if (err && err != UBI_IO_BITFLIPS) {
 			if (err > 0) {
 				/*
@@ -451,14 +484,18 @@ retry:
 		} else if (err == UBI_IO_BITFLIPS)
 			scrub = 1;
 
-		ubi_assert(lnum < be32_to_cpu(vid_hdr->used_ebs));
-		ubi_assert(len == be32_to_cpu(vid_hdr->data_size));
+		ubi_assert(lnum < be32_to_cpu(vid_hdr[lpos].used_ebs));
+		ubi_assert(len == be32_to_cpu(vid_hdr[lpos].data_size));
 
-		crc = be32_to_cpu(vid_hdr->data_crc);
+		crc = be32_to_cpu(vid_hdr[lpos].data_crc);
 		ubi_free_vid_hdr(ubi, vid_hdr);
 	}
 
-	err = ubi_io_read_data(ubi, buf, pnum, offset, len);
+	if (!clebs)
+		err = ubi_io_read_data(ubi, buf, pnum, offset, len);
+	else
+		err = ubi_io_raw_read(ubi, buf, pnum, offset + loffs, len);
+
 	if (err) {
 		if (err == UBI_IO_BITFLIPS)
 			scrub = 1;
@@ -581,7 +618,7 @@ static int recover_peb(struct ubi_device *ubi, int pnum, int vol_id, int lnum,
 		return -ENOMEM;
 
 retry:
-	new_pnum = ubi_wl_get_peb(ubi);
+	new_pnum = ubi_wl_get_peb(ubi, false);
 	if (new_pnum < 0) {
 		ubi_free_vid_hdr(ubi, vid_hdr);
 		up_read(&ubi->fm_eba_sem);
@@ -633,7 +670,7 @@ retry:
 
 	vol->eba_tbl[lnum] = new_pnum;
 	up_read(&ubi->fm_eba_sem);
-	ubi_wl_put_peb(ubi, vol_id, 1);
+	ubi_wl_put_peb(ubi, pnum, 1);
 
 	ubi_msg(ubi, "data was successfully recovered");
 	return 0;
@@ -679,16 +716,24 @@ int ubi_eba_write_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
 {
 	int err, pnum, tries = 0, vol_id = vol->vol_id;
 	struct ubi_vid_hdr *vid_hdr;
+	struct ubi_leb_desc *clebs;
+	bool full;
 
 	if (ubi->ro_mode)
 		return -EROFS;
 
+	full = (offset + len > ubi->leb_size - ubi->min_io_size);
+
 	err = leb_write_lock(ubi, vol_id, lnum);
 	if (err)
 		return err;
 
 	pnum = vol->eba_tbl[lnum];
 	if (pnum >= 0) {
+		clebs = ubi_conso_get_consolidated(ubi, pnum);
+		/* TODO: handle the write on consolidated PEB case */
+		BUG_ON(clebs);
+
 		dbg_eba("write %d bytes at offset %d of LEB %d:%d, PEB %d",
 			len, offset, vol_id, lnum, pnum);
 
@@ -701,7 +746,22 @@ int ubi_eba_write_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
 			if (err)
 				ubi_ro_mode(ubi);
 		}
-		leb_write_unlock(ubi, vol_id, lnum);
+
+		if (full) {
+			int ret;
+
+			ret = ubi_coso_add_full_leb(ubi, vol_id, lnum, 0);
+			if (ret)
+				ubi_warn(ubi,
+					 "failed to add LEB %d:%d to the full LEB list",
+					 vol_id, lnum);
+		}
+
+		ubi_eba_leb_write_unlock(ubi, vol_id, lnum);
+
+		if (full && !err && ubi_conso_consolidation_needed(ubi))
+			ubi_conso_schedule(ubi);
+
 		return err;
 	}
 
@@ -711,7 +771,7 @@ int ubi_eba_write_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
 	 */
 	vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
 	if (!vid_hdr) {
-		leb_write_unlock(ubi, vol_id, lnum);
+		ubi_eba_leb_write_unlock(ubi, vol_id, lnum);
 		return -ENOMEM;
 	}
 
@@ -723,10 +783,10 @@ int ubi_eba_write_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
 	vid_hdr->data_pad = cpu_to_be32(vol->data_pad);
 
 retry:
-	pnum = ubi_wl_get_peb(ubi);
+	pnum = ubi_wl_get_peb(ubi, false);
 	if (pnum < 0) {
 		ubi_free_vid_hdr(ubi, vid_hdr);
-		leb_write_unlock(ubi, vol_id, lnum);
+		ubi_eba_leb_write_unlock(ubi, vol_id, lnum);
 		up_read(&ubi->fm_eba_sem);
 		return pnum;
 	}
@@ -755,14 +815,26 @@ retry:
 	vol->eba_tbl[lnum] = pnum;
 	up_read(&ubi->fm_eba_sem);
 
-	leb_write_unlock(ubi, vol_id, lnum);
+	if (full) {
+		err = ubi_coso_add_full_leb(ubi, vol_id, lnum, 0);
+		if (err)
+			ubi_warn(ubi,
+				 "failed to add LEB %d:%d to the full LEB list",
+				 vol_id, lnum);
+	}
+
+	ubi_eba_leb_write_unlock(ubi, vol_id, lnum);
 	ubi_free_vid_hdr(ubi, vid_hdr);
+
+	if (full && ubi_conso_consolidation_needed(ubi))
+		ubi_conso_schedule(ubi);
+
 	return 0;
 
 write_error:
 	if (err != -EIO || !ubi->bad_allowed) {
 		ubi_ro_mode(ubi);
-		leb_write_unlock(ubi, vol_id, lnum);
+		ubi_eba_leb_write_unlock(ubi, vol_id, lnum);
 		ubi_free_vid_hdr(ubi, vid_hdr);
 		return err;
 	}
@@ -775,7 +847,7 @@ write_error:
 	err = ubi_wl_put_peb(ubi, pnum, 1);
 	if (err || ++tries > UBI_IO_RETRIES) {
 		ubi_ro_mode(ubi);
-		leb_write_unlock(ubi, vol_id, lnum);
+		ubi_eba_leb_write_unlock(ubi, vol_id, lnum);
 		ubi_free_vid_hdr(ubi, vid_hdr);
 		return err;
 	}
@@ -846,10 +918,10 @@ int ubi_eba_write_leb_st(struct ubi_device *ubi, struct ubi_volume *vol,
 	vid_hdr->data_crc = cpu_to_be32(crc);
 
 retry:
-	pnum = ubi_wl_get_peb(ubi);
+	pnum = ubi_wl_get_peb(ubi, false);
 	if (pnum < 0) {
 		ubi_free_vid_hdr(ubi, vid_hdr);
-		leb_write_unlock(ubi, vol_id, lnum);
+		ubi_eba_leb_write_unlock(ubi, vol_id, lnum);
 		up_read(&ubi->fm_eba_sem);
 		return pnum;
 	}
@@ -875,10 +947,20 @@ retry:
 
 	ubi_assert(vol->eba_tbl[lnum] < 0);
 	vol->eba_tbl[lnum] = pnum;
+	vol->used_ebs = used_ebs; //XXX
 	up_read(&ubi->fm_eba_sem);
 
-	leb_write_unlock(ubi, vol_id, lnum);
+	err = ubi_coso_add_full_leb(ubi, vol_id, lnum, 0);
+	if (err)
+		ubi_warn(ubi, "failed to add LEB %d:%d to the full LEB list",
+			 vol_id, lnum);
+
+	ubi_eba_leb_write_unlock(ubi, vol_id, lnum);
 	ubi_free_vid_hdr(ubi, vid_hdr);
+
+	if (ubi_conso_consolidation_needed(ubi))
+		ubi_conso_schedule(ubi);
+
 	return 0;
 
 write_error:
@@ -889,7 +971,7 @@ write_error:
 		 * mode just in case.
 		 */
 		ubi_ro_mode(ubi);
-		leb_write_unlock(ubi, vol_id, lnum);
+		ubi_eba_leb_write_unlock(ubi, vol_id, lnum);
 		ubi_free_vid_hdr(ubi, vid_hdr);
 		return err;
 	}
@@ -897,7 +979,7 @@ write_error:
 	err = ubi_wl_put_peb(ubi, pnum, 1);
 	if (err || ++tries > UBI_IO_RETRIES) {
 		ubi_ro_mode(ubi);
-		leb_write_unlock(ubi, vol_id, lnum);
+		ubi_eba_leb_write_unlock(ubi, vol_id, lnum);
 		ubi_free_vid_hdr(ubi, vid_hdr);
 		return err;
 	}
@@ -926,7 +1008,9 @@ int ubi_eba_atomic_leb_change(struct ubi_device *ubi, struct ubi_volume *vol,
 {
 	int err, pnum, old_pnum, tries = 0, vol_id = vol->vol_id;
 	struct ubi_vid_hdr *vid_hdr;
+	bool release_peb = false;
 	uint32_t crc;
+	bool full;
 
 	if (ubi->ro_mode)
 		return -EROFS;
@@ -942,6 +1026,8 @@ int ubi_eba_atomic_leb_change(struct ubi_device *ubi, struct ubi_volume *vol,
 		return ubi_eba_write_leb(ubi, vol, lnum, NULL, 0, 0);
 	}
 
+	full = (len > ubi->leb_size - ubi->min_io_size);
+
 	vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
 	if (!vid_hdr)
 		return -ENOMEM;
@@ -963,7 +1049,7 @@ int ubi_eba_atomic_leb_change(struct ubi_device *ubi, struct ubi_volume *vol,
 	vid_hdr->data_crc = cpu_to_be32(crc);
 
 retry:
-	pnum = ubi_wl_get_peb(ubi);
+	pnum = ubi_wl_get_peb(ubi, false);
 	if (pnum < 0) {
 		err = pnum;
 		up_read(&ubi->fm_eba_sem);
@@ -991,18 +1077,33 @@ retry:
 
 	old_pnum = vol->eba_tbl[lnum];
 	vol->eba_tbl[lnum] = pnum;
+	if (old_pnum >= 0)
+		release_peb = ubi_conso_invalidate_leb(ubi, old_pnum, vol_id, lnum);
 	up_read(&ubi->fm_eba_sem);
+	ubi_conso_remove_full_leb(ubi, vol_id, lnum);
+	if (full) {
+		int ret;
+
+		ret = ubi_coso_add_full_leb(ubi, vol_id, lnum, 0);
+		if (ret)
+			ubi_warn(ubi,
+				"failed to add LEB %d:%d to the full LEB list",
+				vol_id, lnum);
+	}
 
-	if (old_pnum >= 0) {
+out_leb_unlock:
+	ubi_eba_leb_write_unlock(ubi, vol_id, lnum);
+	if (release_peb) {
 		err = ubi_wl_put_peb(ubi, old_pnum, 0);
 		if (err)
 			goto out_leb_unlock;
 	}
-
-out_leb_unlock:
-	leb_write_unlock(ubi, vol_id, lnum);
 out_mutex:
 	ubi_free_vid_hdr(ubi, vid_hdr);
+
+	if (full && !err && ubi_conso_consolidation_needed(ubi))
+		ubi_conso_schedule(ubi);
+
 	return err;
 
 write_error:
@@ -1224,7 +1325,178 @@ int ubi_eba_copy_leb(struct ubi_device *ubi, int from, int to,
 out_unlock_buf:
 	mutex_unlock(&ubi->buf_mutex);
 out_unlock_leb:
-	leb_write_unlock(ubi, vol_id, lnum);
+	ubi_eba_leb_write_unlock(ubi, vol_id, lnum);
+	return err;
+}
+
+
+/**
+ * ubi_eba_copy_lebs - copy consolidated logical eraseblocks.
+ *
+ * Works like ubi_eba_copy_leb but on consolidated LEB.
+ * It is less complicated as a PEB containing consolidated LEBs
+ * has only full LEBs and we don't have to do a lot of space
+ * calucation.
+ * TODO: clean this function up, more clean error handling, etc...
+ */
+int ubi_eba_copy_lebs(struct ubi_device *ubi, int from, int to,
+		     struct ubi_vid_hdr *vid_hdr, int nvidh)
+{
+	int err, i;
+	int *vol_id = NULL, *lnum = NULL;
+	struct ubi_volume **vol = NULL;
+	uint32_t crc;
+
+	vol_id = kmalloc(nvidh * sizeof(*vol_id), GFP_NOFS);
+	lnum = kmalloc(nvidh * sizeof(*lnum), GFP_NOFS);
+	vol = kmalloc(nvidh * sizeof(*vol), GFP_NOFS);
+
+	if (!vol_id || !lnum || !vol) {
+		kfree(vol_id);
+		kfree(lnum);
+		kfree(vol);
+		return -ENOMEM;
+	}
+
+	dbg_wl("copy LEBs of PEB %d to PEB %d", from, to);
+
+	spin_lock(&ubi->volumes_lock);
+
+	for (i = 0; i < nvidh; i++) {
+		vol_id[i] = be32_to_cpu(vid_hdr[i].vol_id);
+		lnum[i] = be32_to_cpu(vid_hdr[i].lnum);
+		vol[i] = ubi->volumes[vol_id2idx(ubi, vol_id[i])];
+	}
+
+	/*
+	 * Note, we may race with volume deletion, which means that the volume
+	 * this logical eraseblock belongs to might be being deleted. Since the
+	 * volume deletion un-maps all the volume's logical eraseblocks, it will
+	 * be locked in 'ubi_wl_put_peb()' and wait for the WL worker to finish.
+	 */
+	spin_unlock(&ubi->volumes_lock);
+
+	for (i = 0; i < nvidh; i++) {
+		if (!vol[i]) {
+			/* No need to do further work, cancel */
+			ubi_msg(ubi, "volume %d is being removed, cancel", vol_id[i]);
+			kfree(vol_id);
+			kfree(lnum);
+			kfree(vol);
+			return MOVE_CANCEL_RACE;
+		}
+	}
+
+	/*
+	 * We do not want anybody to write to this logical eraseblock while we
+	 * are moving it, so lock it.
+	 *
+	 * Note, we are using non-waiting locking here, because we cannot sleep
+	 * on the LEB, since it may cause deadlocks. Indeed, imagine a task is
+	 * unmapping the LEB which is mapped to the PEB we are going to move
+	 * (@from). This task locks the LEB and goes sleep in the
+	 * 'ubi_wl_put_peb()' function on the @ubi->move_mutex. In turn, we are
+	 * holding @ubi->move_mutex and go sleep on the LEB lock. So, if the
+	 * LEB is already locked, we just do not move it and return
+	 * %MOVE_RETRY. Note, we do not return %MOVE_CANCEL_RACE here because
+	 * we do not know the reasons of the contention - it may be just a
+	 * normal I/O on this LEB, so we want to re-try.
+	 */
+
+	for (i = 0; i < nvidh; i++) {
+		err = leb_write_trylock(ubi, vol_id[i], lnum[i]);
+		if (err) {
+			int j;
+
+			for (j = 0; j < i; j++)
+				ubi_eba_leb_write_unlock(ubi, vol_id[j], lnum[j]);
+
+			kfree(vol_id);
+			kfree(lnum);
+			kfree(vol);
+			return MOVE_RETRY;
+		}
+	}
+	for (i = 0; i < nvidh; i++) {
+		/*
+		 * The LEB might have been put meanwhile, and the task which put it is
+		 * probably waiting on @ubi->move_mutex. No need to continue the work,
+		 * cancel it.
+		 */
+		if (vol[i]->eba_tbl[lnum[i]] != from) {
+			ubi_msg(ubi, "LEB %d:%d is no longer mapped to PEB %d, mapped to PEB %d, cancel",
+			       vol_id[i], lnum[i], from, vol[i]->eba_tbl[lnum[i]]);
+			err = MOVE_CANCEL_RACE;
+			goto out_unlock_leb;
+		}
+	}
+
+	/*
+	 * OK, now the LEB is locked and we can safely start moving it. Since
+	 * this function utilizes the @ubi->peb_buf buffer which is shared
+	 * with some other functions - we lock the buffer by taking the
+	 * @ubi->buf_mutex.
+	 */
+	mutex_lock(&ubi->buf_mutex);
+	dbg_wl("read %d bytes of data", ubi->consolidated_peb_size - ubi->leb_start);
+	err = ubi_io_raw_read(ubi, ubi->peb_buf, from, ubi->leb_start, ubi->consolidated_peb_size - ubi->leb_start);
+	if (err && err != UBI_IO_BITFLIPS) {
+		ubi_warn(ubi, "error %d while reading data from PEB %d",
+			 err, from);
+		err = MOVE_SOURCE_RD_ERR;
+		goto out_unlock_buf;
+	}
+
+	cond_resched();
+	for (i = 0; i < nvidh; i++) {
+		//TODO: we could skip crc calucation as consolidated LEB _always_ hav copy_flag=1 and hence also a valid crc...
+		crc = crc32(UBI_CRC32_INIT, ubi->peb_buf + ubi->leb_start + (i * ubi->leb_size), be32_to_cpu(vid_hdr[i].data_size));
+		vid_hdr[i].copy_flag = 1;
+		vid_hdr[i].data_crc = cpu_to_be32(crc);
+		vid_hdr[i].sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
+
+		cond_resched();
+	}
+
+
+	err = ubi_io_write_vid_hdrs(ubi, to, vid_hdr, nvidh);
+	if (err) {
+		if (err == -EIO)
+			err = MOVE_TARGET_WR_ERR;
+		goto out_unlock_buf;
+	}
+
+	cond_resched();
+
+	err = ubi_io_raw_write(ubi, ubi->peb_buf, to, ubi->leb_start, ubi->consolidated_peb_size - ubi->leb_start);
+	if (err) {
+		if (err == -EIO)
+			err = MOVE_TARGET_WR_ERR;
+		goto out_unlock_buf;
+	}
+
+	cond_resched();
+
+	down_read(&ubi->fm_eba_sem);
+	for (i = 0; i < nvidh; i++) {
+		ubi_assert(vol[i]->eba_tbl[lnum[i]] == from);
+		vol[i]->eba_tbl[lnum[i]] = to;
+	}
+
+	ubi->consolidated[to] = ubi->consolidated[from];
+	ubi->consolidated[from] = NULL;
+
+	up_read(&ubi->fm_eba_sem);
+
+out_unlock_buf:
+	mutex_unlock(&ubi->buf_mutex);
+out_unlock_leb:
+	for (i = 0; i < nvidh; i++)
+		ubi_eba_leb_write_unlock(ubi, vol_id[i], lnum[i]);
+	kfree(vol_id);
+	kfree(lnum);
+	kfree(vol);
+
 	return err;
 }
 
@@ -1286,7 +1558,7 @@ int self_check_eba(struct ubi_device *ubi, struct ubi_attach_info *ai_fastmap,
 	int **scan_eba, **fm_eba;
 	struct ubi_ainf_volume *av;
 	struct ubi_volume *vol;
-	struct ubi_ainf_peb *aeb;
+	struct ubi_ainf_leb *aeb;
 	struct rb_node *rb;
 
 	num_volumes = ubi->vtbl_slots + UBI_INT_VOL_COUNT;
@@ -1306,38 +1578,38 @@ int self_check_eba(struct ubi_device *ubi, struct ubi_attach_info *ai_fastmap,
 		if (!vol)
 			continue;
 
-		scan_eba[i] = kmalloc(vol->reserved_pebs * sizeof(**scan_eba),
+		scan_eba[i] = kmalloc(vol->reserved_lebs * sizeof(**scan_eba),
 				      GFP_KERNEL);
 		if (!scan_eba[i]) {
 			ret = -ENOMEM;
 			goto out_free;
 		}
 
-		fm_eba[i] = kmalloc(vol->reserved_pebs * sizeof(**fm_eba),
+		fm_eba[i] = kmalloc(vol->reserved_lebs * sizeof(**fm_eba),
 				    GFP_KERNEL);
 		if (!fm_eba[i]) {
 			ret = -ENOMEM;
 			goto out_free;
 		}
 
-		for (j = 0; j < vol->reserved_pebs; j++)
+		for (j = 0; j < vol->reserved_lebs; j++)
 			scan_eba[i][j] = fm_eba[i][j] = UBI_LEB_UNMAPPED;
 
 		av = ubi_find_av(ai_scan, idx2vol_id(ubi, i));
 		if (!av)
 			continue;
 
-		ubi_rb_for_each_entry(rb, aeb, &av->root, u.rb)
-			scan_eba[i][aeb->lnum] = aeb->pnum;
+		ubi_rb_for_each_entry(rb, aeb, &av->root, rb)
+			scan_eba[i][aeb->desc.lnum] = aeb->peb->pnum;
 
 		av = ubi_find_av(ai_fastmap, idx2vol_id(ubi, i));
 		if (!av)
 			continue;
 
-		ubi_rb_for_each_entry(rb, aeb, &av->root, u.rb)
-			fm_eba[i][aeb->lnum] = aeb->pnum;
+		ubi_rb_for_each_entry(rb, aeb, &av->root, rb)
+			fm_eba[i][aeb->desc.lnum] = aeb->peb->pnum;
 
-		for (j = 0; j < vol->reserved_pebs; j++) {
+		for (j = 0; j < vol->reserved_lebs; j++) {
 			if (scan_eba[i][j] != fm_eba[i][j]) {
 				if (scan_eba[i][j] == UBI_LEB_UNMAPPED ||
 					fm_eba[i][j] == UBI_LEB_UNMAPPED)
@@ -1378,8 +1650,9 @@ int ubi_eba_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
 	int i, j, err, num_volumes;
 	struct ubi_ainf_volume *av;
 	struct ubi_volume *vol;
-	struct ubi_ainf_peb *aeb;
+	struct ubi_ainf_leb *aeb;
 	struct rb_node *rb;
+	int eba_rsvd = EBA_RESERVED_PEBS;
 
 	dbg_eba("initialize EBA sub-system");
 
@@ -1396,43 +1669,48 @@ int ubi_eba_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
 
 		cond_resched();
 
-		vol->eba_tbl = kmalloc(vol->reserved_pebs * sizeof(int),
+		vol->eba_tbl = kmalloc(vol->reserved_lebs * sizeof(int),
 				       GFP_KERNEL);
 		if (!vol->eba_tbl) {
 			err = -ENOMEM;
 			goto out_free;
 		}
 
-		for (j = 0; j < vol->reserved_pebs; j++)
+		for (j = 0; j < vol->reserved_lebs; j++)
 			vol->eba_tbl[j] = UBI_LEB_UNMAPPED;
 
 		av = ubi_find_av(ai, idx2vol_id(ubi, i));
 		if (!av)
 			continue;
 
-		ubi_rb_for_each_entry(rb, aeb, &av->root, u.rb) {
-			if (aeb->lnum >= vol->reserved_pebs)
+		ubi_rb_for_each_entry(rb, aeb, &av->root, rb) {
+			if (aeb->desc.lnum >= vol->reserved_lebs) {
 				/*
 				 * This may happen in case of an unclean reboot
 				 * during re-size.
 				 */
-				ubi_move_aeb_to_list(av, aeb, &ai->erase);
-			else
-				vol->eba_tbl[aeb->lnum] = aeb->pnum;
+				if (--aeb->peb->refcount <= 0)
+					list_move_tail(&aeb->peb->list, &ai->erase);
+			} else {
+				vol->eba_tbl[aeb->desc.lnum] = aeb->peb->pnum;
+				if (aeb->full)
+					ubi_coso_add_full_leb(ubi, vol->vol_id,
+							      aeb->desc.lnum, 0);
+			}
 		}
 	}
 
-	if (ubi->avail_pebs < EBA_RESERVED_PEBS) {
+	if (ubi->avail_pebs < eba_rsvd) {
 		ubi_err(ubi, "no enough physical eraseblocks (%d, need %d)",
-			ubi->avail_pebs, EBA_RESERVED_PEBS);
+			ubi->avail_pebs, eba_rsvd);
 		if (ubi->corr_peb_count)
 			ubi_err(ubi, "%d PEBs are corrupted and not used",
 				ubi->corr_peb_count);
 		err = -ENOSPC;
 		goto out_free;
 	}
-	ubi->avail_pebs -= EBA_RESERVED_PEBS;
-	ubi->rsvd_pebs += EBA_RESERVED_PEBS;
+	ubi->avail_pebs -= eba_rsvd;
+	ubi->rsvd_pebs += eba_rsvd;
 
 	if (ubi->bad_allowed) {
 		ubi_calculate_reserved(ubi);
@@ -1448,6 +1726,9 @@ int ubi_eba_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
 		ubi->rsvd_pebs  += ubi->beb_rsvd_pebs;
 	}
 
+	if (ubi->lebs_per_cpeb > 1)
+		ubi_conso_schedule(ubi);
+
 	dbg_eba("EBA sub-system is initialized");
 	return 0;
 
diff --git a/drivers/mtd/ubi/fastmap-wl.c b/drivers/mtd/ubi/fastmap-wl.c
index cafa7b0..ec593fc 100644
--- a/drivers/mtd/ubi/fastmap-wl.c
+++ b/drivers/mtd/ubi/fastmap-wl.c
@@ -180,6 +180,8 @@ void ubi_refill_pools(struct ubi_device *ubi)
  * disabled. Returns zero in case of success and a negative error code in case
  * of failure.
  */
+
+#error call ubi_eba_consolidate
 static int produce_free_peb(struct ubi_device *ubi)
 {
 	while (!ubi->free.rb_node) {
@@ -201,7 +203,7 @@ static int produce_free_peb(struct ubi_device *ubi)
  * negative error code in case of failure.
  * Returns with ubi->fm_eba_sem held in read mode!
  */
-int ubi_wl_get_peb(struct ubi_device *ubi)
+int ubi_wl_get_peb(struct ubi_device *ubi, bool nested)
 {
 	int ret, retried = 0;
 	struct ubi_fm_pool *pool = &ubi->fm_pool;
@@ -211,6 +213,15 @@ again:
 	down_read(&ubi->fm_eba_sem);
 	spin_lock(&ubi->wl_lock);
 
+	if (nested) {
+		if (pool->used == pool->size) {
+			ret = -ENOSPC;
+			goto out_unlock;
+		}
+
+		goto out_get_peb;
+	}
+
 	/* We check here also for the WL pool because at this point we can
 	 * refill the WL pool synchronous. */
 	if (pool->used == pool->size || wl_pool->used == wl_pool->size) {
@@ -243,9 +254,11 @@ again:
 		goto again;
 	}
 
+out_get_peb:
 	ubi_assert(pool->used < pool->size);
 	ret = pool->pebs[pool->used++];
 	prot_queue_add(ubi, ubi->lookuptbl[ret]);
+out_unlock:
 	spin_unlock(&ubi->wl_lock);
 out:
 	return ret;
@@ -291,7 +304,7 @@ int ubi_ensure_anchor_pebs(struct ubi_device *ubi)
 	ubi->wl_scheduled = 1;
 	spin_unlock(&ubi->wl_lock);
 
-	wrk = kmalloc(sizeof(struct ubi_work), GFP_NOFS);
+	wrk = ubi_alloc_work(ubi);
 	if (!wrk) {
 		spin_lock(&ubi->wl_lock);
 		ubi->wl_scheduled = 0;
@@ -342,7 +355,7 @@ int ubi_wl_put_fm_peb(struct ubi_device *ubi, struct ubi_wl_entry *fm_e,
 	spin_unlock(&ubi->wl_lock);
 
 	vol_id = lnum ? UBI_FM_DATA_VOLUME_ID : UBI_FM_SB_VOLUME_ID;
-	return schedule_erase(ubi, e, torture);
+	return schedule_erase(ubi, e, torture, false);
 }
 
 /**
diff --git a/drivers/mtd/ubi/fastmap.c b/drivers/mtd/ubi/fastmap.c
index c878313..61f8fc6 100644
--- a/drivers/mtd/ubi/fastmap.c
+++ b/drivers/mtd/ubi/fastmap.c
@@ -143,15 +143,13 @@ static int add_aeb(struct ubi_attach_info *ai, struct list_head *list,
 {
 	struct ubi_ainf_peb *aeb;
 
-	aeb = kmem_cache_alloc(ai->aeb_slab_cache, GFP_KERNEL);
+	aeb = kmem_cache_alloc(ai->apeb_slab_cache, GFP_KERNEL);
 	if (!aeb)
 		return -ENOMEM;
 
 	aeb->pnum = pnum;
 	aeb->ec = ec;
-	aeb->lnum = -1;
 	aeb->scrub = scrub;
-	aeb->copy_flag = aeb->sqnum = 0;
 
 	ai->ec_sum += aeb->ec;
 	ai->ec_count++;
@@ -162,7 +160,7 @@ static int add_aeb(struct ubi_attach_info *ai, struct list_head *list,
 	if (ai->min_ec > aeb->ec)
 		ai->min_ec = aeb->ec;
 
-	list_add_tail(&aeb->u.list, list);
+	list_add_tail(&aeb->list, list);
 
 	return 0;
 }
@@ -229,19 +227,19 @@ out:
  * @av: target scan volume
  */
 static void assign_aeb_to_av(struct ubi_attach_info *ai,
-			     struct ubi_ainf_peb *aeb,
+			     struct ubi_ainf_leb *aeb,
 			     struct ubi_ainf_volume *av)
 {
-	struct ubi_ainf_peb *tmp_aeb;
+	struct ubi_ainf_leb *tmp_aeb;
 	struct rb_node **p = &ai->volumes.rb_node, *parent = NULL;
 
 	p = &av->root.rb_node;
 	while (*p) {
 		parent = *p;
 
-		tmp_aeb = rb_entry(parent, struct ubi_ainf_peb, u.rb);
-		if (aeb->lnum != tmp_aeb->lnum) {
-			if (aeb->lnum < tmp_aeb->lnum)
+		tmp_aeb = rb_entry(parent, struct ubi_ainf_leb, rb);
+		if (aeb->desc.lnum != tmp_aeb->desc.lnum) {
+			if (aeb->desc.lnum < tmp_aeb->desc.lnum)
 				p = &(*p)->rb_left;
 			else
 				p = &(*p)->rb_right;
@@ -251,11 +249,10 @@ static void assign_aeb_to_av(struct ubi_attach_info *ai,
 			break;
 	}
 
-	list_del(&aeb->u.list);
 	av->leb_count++;
 
-	rb_link_node(&aeb->u.rb, parent, p);
-	rb_insert_color(&aeb->u.rb, &av->root);
+	rb_link_node(&aeb->rb, parent, p);
+	rb_insert_color(&aeb->rb, &av->root);
 }
 
 /**
@@ -270,18 +267,18 @@ static void assign_aeb_to_av(struct ubi_attach_info *ai,
  */
 static int update_vol(struct ubi_device *ubi, struct ubi_attach_info *ai,
 		      struct ubi_ainf_volume *av, struct ubi_vid_hdr *new_vh,
-		      struct ubi_ainf_peb *new_aeb)
+		      struct ubi_ainf_peb *new_peb, int peb_pos, bool full)
 {
 	struct rb_node **p = &av->root.rb_node, *parent = NULL;
-	struct ubi_ainf_peb *aeb, *victim;
-	int cmp_res;
+	struct ubi_ainf_leb *aeb;
+	int cmp_res, lnum = be32_to_cpu(new_vh->lnum);
 
 	while (*p) {
 		parent = *p;
-		aeb = rb_entry(parent, struct ubi_ainf_peb, u.rb);
+		aeb = rb_entry(parent, struct ubi_ainf_leb, rb);
 
-		if (be32_to_cpu(new_vh->lnum) != aeb->lnum) {
-			if (be32_to_cpu(new_vh->lnum) < aeb->lnum)
+		if (be32_to_cpu(new_vh->lnum) != aeb->desc.lnum) {
+			if (be32_to_cpu(new_vh->lnum) < aeb->desc.lnum)
 				p = &(*p)->rb_left;
 			else
 				p = &(*p)->rb_right;
@@ -293,51 +290,58 @@ static int update_vol(struct ubi_device *ubi, struct ubi_attach_info *ai,
 		 * because of a volume change (creation, deletion, ..).
 		 * Then a PEB can be within the persistent EBA and the pool.
 		 */
-		if (aeb->pnum == new_aeb->pnum) {
-			ubi_assert(aeb->lnum == new_aeb->lnum);
-			kmem_cache_free(ai->aeb_slab_cache, new_aeb);
+		if (aeb->peb->pnum == new_peb->pnum) {
+			ubi_assert(aeb->desc.lnum == lnum);
+			kmem_cache_free(ai->apeb_slab_cache, new_peb);
 
 			return 0;
 		}
 
-		cmp_res = ubi_compare_lebs(ubi, aeb, new_aeb->pnum, new_vh);
+		cmp_res = ubi_compare_lebs(ubi, aeb, new_peb->pnum, new_vh);
 		if (cmp_res < 0)
 			return cmp_res;
 
 		/* new_aeb is newer */
 		if (cmp_res & 1) {
-			victim = kmem_cache_alloc(ai->aeb_slab_cache,
-				GFP_KERNEL);
-			if (!victim)
-				return -ENOMEM;
-
-			victim->ec = aeb->ec;
-			victim->pnum = aeb->pnum;
-			list_add_tail(&victim->u.list, &ai->erase);
+			if (--aeb->peb->refcount <= 0)
+				list_move(&aeb->peb->list, &ai->erase);
 
 			if (av->highest_lnum == be32_to_cpu(new_vh->lnum))
 				av->last_data_size =
 					be32_to_cpu(new_vh->data_size);
 
 			dbg_bld("vol %i: AEB %i's PEB %i is the newer",
-				av->vol_id, aeb->lnum, new_aeb->pnum);
+				av->vol_id, aeb->desc.lnum,
+				new_peb->pnum);
 
-			aeb->ec = new_aeb->ec;
-			aeb->pnum = new_aeb->pnum;
+			aeb->peb_pos = peb_pos;
+			aeb->peb = new_peb;
 			aeb->copy_flag = new_vh->copy_flag;
-			aeb->scrub = new_aeb->scrub;
-			kmem_cache_free(ai->aeb_slab_cache, new_aeb);
+			aeb->sqnum = be64_to_cpu(new_vh->sqnum);
+			aeb->full = full;
 
 		/* new_aeb is older */
 		} else {
 			dbg_bld("vol %i: AEB %i's PEB %i is old, dropping it",
-				av->vol_id, aeb->lnum, new_aeb->pnum);
-			list_add_tail(&new_aeb->u.list, &ai->erase);
+				av->vol_id, aeb->desc.lnum, new_peb->pnum);
+			if (--aeb->peb->refcount <= 0)
+				list_move(&aeb->peb->list, &ai->erase);
 		}
 
 		return 0;
 	}
 	/* This LEB is new, let's add it to the volume */
+	aeb = kmem_cache_alloc(ai->aleb_slab_cache, GFP_KERNEL);
+	if (!aeb)
+		return -ENOMEM;
+
+	aeb->peb = new_peb;
+	aeb->peb_pos = peb_pos;
+	aeb->copy_flag = new_vh->copy_flag;
+	aeb->full = full;
+	aeb->sqnum = be64_to_cpu(new_vh->sqnum);
+	aeb->desc.lnum = lnum;
+	aeb->desc.vol_id = be32_to_cpu(new_vh->vol_id);
 
 	if (av->highest_lnum <= be32_to_cpu(new_vh->lnum)) {
 		av->highest_lnum = be32_to_cpu(new_vh->lnum);
@@ -349,8 +353,8 @@ static int update_vol(struct ubi_device *ubi, struct ubi_attach_info *ai,
 
 	av->leb_count++;
 
-	rb_link_node(&new_aeb->u.rb, parent, p);
-	rb_insert_color(&new_aeb->u.rb, &av->root);
+	rb_link_node(&aeb->rb, parent, p);
+	rb_insert_color(&aeb->rb, &av->root);
 
 	return 0;
 }
@@ -366,7 +370,8 @@ static int update_vol(struct ubi_device *ubi, struct ubi_attach_info *ai,
  */
 static int process_pool_aeb(struct ubi_device *ubi, struct ubi_attach_info *ai,
 			    struct ubi_vid_hdr *new_vh,
-			    struct ubi_ainf_peb *new_aeb)
+			    struct ubi_ainf_peb *new_aeb, int peb_pos,
+			    bool full)
 {
 	struct ubi_ainf_volume *av, *tmp_av = NULL;
 	struct rb_node **p = &ai->volumes.rb_node, *parent = NULL;
@@ -374,7 +379,7 @@ static int process_pool_aeb(struct ubi_device *ubi, struct ubi_attach_info *ai,
 
 	if (be32_to_cpu(new_vh->vol_id) == UBI_FM_SB_VOLUME_ID ||
 		be32_to_cpu(new_vh->vol_id) == UBI_FM_DATA_VOLUME_ID) {
-		kmem_cache_free(ai->aeb_slab_cache, new_aeb);
+		kmem_cache_free(ai->apeb_slab_cache, new_aeb);
 
 		return 0;
 	}
@@ -398,13 +403,13 @@ static int process_pool_aeb(struct ubi_device *ubi, struct ubi_attach_info *ai,
 		av = tmp_av;
 	else {
 		ubi_err(ubi, "orphaned volume in fastmap pool!");
-		kmem_cache_free(ai->aeb_slab_cache, new_aeb);
+		kmem_cache_free(ai->apeb_slab_cache, new_aeb);
 		return UBI_BAD_FASTMAP;
 	}
 
 	ubi_assert(be32_to_cpu(new_vh->vol_id) == av->vol_id);
 
-	return update_vol(ubi, ai, av, new_vh, new_aeb);
+	return update_vol(ubi, ai, av, new_vh, new_aeb, peb_pos, full);
 }
 
 /**
@@ -419,18 +424,20 @@ static void unmap_peb(struct ubi_attach_info *ai, int pnum)
 {
 	struct ubi_ainf_volume *av;
 	struct rb_node *node, *node2;
-	struct ubi_ainf_peb *aeb;
+	struct ubi_ainf_leb *aeb;
 
 	for (node = rb_first(&ai->volumes); node; node = rb_next(node)) {
 		av = rb_entry(node, struct ubi_ainf_volume, rb);
 
 		for (node2 = rb_first(&av->root); node2;
 		     node2 = rb_next(node2)) {
-			aeb = rb_entry(node2, struct ubi_ainf_peb, u.rb);
-			if (aeb->pnum == pnum) {
-				rb_erase(&aeb->u.rb, &av->root);
+			aeb = rb_entry(node2, struct ubi_ainf_leb, rb);
+			if (aeb->peb->pnum == pnum) {
+				rb_erase(&aeb->rb, &av->root);
 				av->leb_count--;
-				kmem_cache_free(ai->aeb_slab_cache, aeb);
+				if (--aeb->peb->refcount <= 0)
+					list_move(&aeb->peb->list, &ai->erase);
+				kmem_cache_free(ai->apeb_slab_cache, aeb);
 				return;
 			}
 		}
@@ -456,7 +463,7 @@ static int scan_pool(struct ubi_device *ubi, struct ubi_attach_info *ai,
 	struct ubi_vid_hdr *vh;
 	struct ubi_ec_hdr *ech;
 	struct ubi_ainf_peb *new_aeb;
-	int i, pnum, err, ret = 0;
+	int i, pnum, err, ret = 0, nvid;
 
 	ech = kzalloc(ubi->ec_hdr_alsize, GFP_KERNEL);
 	if (!ech)
@@ -508,7 +515,9 @@ static int scan_pool(struct ubi_device *ubi, struct ubi_attach_info *ai,
 			goto out;
 		}
 
-		err = ubi_io_read_vid_hdr(ubi, pnum, vh, 0);
+		/* TODO: support consolidate PEBs */
+		nvid = ubi->lebs_per_cpeb;
+		err = ubi_io_read_vid_hdrs(ubi, pnum, vh, &nvid, 0);
 		if (err == UBI_IO_FF || err == UBI_IO_FF_BITFLIPS) {
 			unsigned long long ec = be64_to_cpu(ech->ec);
 			unmap_peb(ai, pnum);
@@ -519,12 +528,15 @@ static int scan_pool(struct ubi_device *ubi, struct ubi_attach_info *ai,
 				add_aeb(ai, free, pnum, ec, 0);
 			continue;
 		} else if (err == 0 || err == UBI_IO_BITFLIPS) {
+			bool full = false;
+			int peb_pos;
+
 			dbg_bld("Found non empty PEB:%i in pool", pnum);
 
 			if (err == UBI_IO_BITFLIPS)
 				scrub = 1;
 
-			new_aeb = kmem_cache_alloc(ai->aeb_slab_cache,
+			new_aeb = kmem_cache_alloc(ai->apeb_slab_cache,
 						   GFP_KERNEL);
 			if (!new_aeb) {
 				ret = -ENOMEM;
@@ -533,18 +545,33 @@ static int scan_pool(struct ubi_device *ubi, struct ubi_attach_info *ai,
 
 			new_aeb->ec = be64_to_cpu(ech->ec);
 			new_aeb->pnum = pnum;
+			new_aeb->refcount = 1;
+/*
 			new_aeb->lnum = be32_to_cpu(vh->lnum);
 			new_aeb->sqnum = be64_to_cpu(vh->sqnum);
 			new_aeb->copy_flag = vh->copy_flag;
+*/
 			new_aeb->scrub = scrub;
+			if (nvid == 1) {
+				err = ubi_io_read(ubi, ech, pnum,
+						  ubi->peb_size - ubi->hdrs_min_io_size,
+						  ubi->hdrs_min_io_size);
+				if (!err && !ubi_check_pattern(ech, 0xff, ubi->hdrs_min_io_size))
+					full = true;
+			}
 
-			if (*max_sqnum < new_aeb->sqnum)
-				*max_sqnum = new_aeb->sqnum;
+			new_aeb->consolidated = nvid > 1;
 
-			err = process_pool_aeb(ubi, ai, vh, new_aeb);
-			if (err) {
-				ret = err > 0 ? UBI_BAD_FASTMAP : err;
-				goto out;
+			for (peb_pos = 0; peb_pos < nvid; peb_pos++) {
+				if (*max_sqnum < be64_to_cpu(vh[peb_pos].sqnum))
+					*max_sqnum = be64_to_cpu(vh[peb_pos].sqnum);
+
+				err = process_pool_aeb(ubi, ai, &vh[peb_pos], new_aeb,
+						       peb_pos, full);
+				if (err) {
+					ret = err > 0 ? UBI_BAD_FASTMAP : err;
+					goto out;
+				}
 			}
 		} else {
 			/* We are paranoid and fall back to scanning mode */
@@ -568,19 +595,16 @@ out:
 static int count_fastmap_pebs(struct ubi_attach_info *ai)
 {
 	struct ubi_ainf_peb *aeb;
-	struct ubi_ainf_volume *av;
-	struct rb_node *rb1, *rb2;
 	int n = 0;
 
-	list_for_each_entry(aeb, &ai->erase, u.list)
+	list_for_each_entry(aeb, &ai->erase, list)
 		n++;
 
-	list_for_each_entry(aeb, &ai->free, u.list)
+	list_for_each_entry(aeb, &ai->free, list)
 		n++;
 
-	 ubi_rb_for_each_entry(rb1, av, &ai->volumes, rb)
-		ubi_rb_for_each_entry(rb2, aeb, &av->root, u.rb)
-			n++;
+	list_for_each_entry(aeb, &ai->used, list)
+		n++;
 
 	return n;
 }
@@ -731,6 +755,9 @@ static int ubi_attach_fastmap(struct ubi_device *ubi,
 
 	/* Iterate over all volumes and read their EBA table */
 	for (i = 0; i < be32_to_cpu(fmhdr->vol_count); i++) {
+		struct ubi_fm_consolidated_leb *fclebs = NULL;
+		int nconso = 0;
+
 		fmvhdr = (struct ubi_fm_volhdr *)(fm_raw + fm_pos);
 		fm_pos += sizeof(*fmvhdr);
 		if (fm_pos >= fm_size)
@@ -760,6 +787,18 @@ static int ubi_attach_fastmap(struct ubi_device *ubi,
 		if (ai->highest_vol_id < be32_to_cpu(fmvhdr->vol_id))
 			ai->highest_vol_id = be32_to_cpu(fmvhdr->vol_id);
 
+		nconso = be32_to_cpu(fmvhdr->consolidated_ebs);
+		if (nconso) {
+			struct ubi_fm_consolidated *fmconso = NULL;
+
+			fmconso = fm_raw + fm_pos;
+			if (be32_to_cpu(fmvhdr->magic) != UBI_FM_CONSO_MAGIC)
+				goto fail_bad;
+
+			fclebs = fmconso->lebs;
+			fm_pos += sizeof(*fmconso) + (nconso * sizeof(*fclebs));
+		}
+
 		fm_eba = (struct ubi_fm_eba *)(fm_raw + fm_pos);
 		fm_pos += sizeof(*fm_eba);
 		fm_pos += (sizeof(__be32) * be32_to_cpu(fm_eba->reserved_pebs));
@@ -774,12 +813,14 @@ static int ubi_attach_fastmap(struct ubi_device *ubi,
 
 		for (j = 0; j < be32_to_cpu(fm_eba->reserved_pebs); j++) {
 			int pnum = be32_to_cpu(fm_eba->pnum[j]);
+			struct ubi_ainf_leb *leb;
+			int k;
 
 			if (pnum < 0)
 				continue;
 
 			aeb = NULL;
-			list_for_each_entry(tmp_aeb, &used, u.list) {
+			list_for_each_entry(tmp_aeb, &used, list) {
 				if (tmp_aeb->pnum == pnum) {
 					aeb = tmp_aeb;
 					break;
@@ -791,18 +832,34 @@ static int ubi_attach_fastmap(struct ubi_device *ubi,
 				goto fail_bad;
 			}
 
-			aeb->lnum = j;
+			leb = kmem_cache_alloc(ai->aleb_slab_cache, GFP_KERNEL);
+			if (!leb)
+				goto fail_bad;
+
+			leb->desc.lnum = j;
+			leb->desc.vol_id = av->vol_id;
+			leb->peb_pos = 0;
+			for (k = 0; k < nconso; k++) {
+				if (be32_to_cpu(fclebs[k].lnum) !=
+				    leb->desc.lnum)
+					continue;
 
-			if (av->highest_lnum <= aeb->lnum)
-				av->highest_lnum = aeb->lnum;
+				leb->peb_pos = be32_to_cpu(fclebs[k].peb_pos);
+				aeb->consolidated = true;
+			}
+			leb->peb = aeb;
 
-			assign_aeb_to_av(ai, aeb, av);
+			if (av->highest_lnum <= leb->desc.lnum)
+				av->highest_lnum = leb->desc.lnum;
+
+			assign_aeb_to_av(ai, leb, av);
 
 			dbg_bld("inserting PEB:%i (LEB %i) to vol %i",
-				aeb->pnum, aeb->lnum, av->vol_id);
+				aeb->pnum, leb->desc.lnum, av->vol_id);
 		}
 	}
 
+
 	ret = scan_pool(ubi, ai, fmpl->pebs, pool_size, &max_sqnum, &free);
 	if (ret)
 		goto fail;
@@ -814,11 +871,11 @@ static int ubi_attach_fastmap(struct ubi_device *ubi,
 	if (max_sqnum > ai->max_sqnum)
 		ai->max_sqnum = max_sqnum;
 
-	list_for_each_entry_safe(tmp_aeb, _tmp_aeb, &free, u.list)
-		list_move_tail(&tmp_aeb->u.list, &ai->free);
+	list_for_each_entry_safe(tmp_aeb, _tmp_aeb, &free, list)
+		list_move_tail(&tmp_aeb->list, &ai->free);
 
-	list_for_each_entry_safe(tmp_aeb, _tmp_aeb, &used, u.list)
-		list_move_tail(&tmp_aeb->u.list, &ai->erase);
+	list_for_each_entry_safe(tmp_aeb, _tmp_aeb, &used, list)
+		list_move_tail(&tmp_aeb->list, &ai->erase);
 
 	ubi_assert(list_empty(&free));
 
@@ -837,13 +894,13 @@ static int ubi_attach_fastmap(struct ubi_device *ubi,
 fail_bad:
 	ret = UBI_BAD_FASTMAP;
 fail:
-	list_for_each_entry_safe(tmp_aeb, _tmp_aeb, &used, u.list) {
-		list_del(&tmp_aeb->u.list);
-		kmem_cache_free(ai->aeb_slab_cache, tmp_aeb);
+	list_for_each_entry_safe(tmp_aeb, _tmp_aeb, &used, list) {
+		list_del(&tmp_aeb->list);
+		kmem_cache_free(ai->apeb_slab_cache, tmp_aeb);
 	}
-	list_for_each_entry_safe(tmp_aeb, _tmp_aeb, &free, u.list) {
-		list_del(&tmp_aeb->u.list);
-		kmem_cache_free(ai->aeb_slab_cache, tmp_aeb);
+	list_for_each_entry_safe(tmp_aeb, _tmp_aeb, &free, list) {
+		list_del(&tmp_aeb->list);
+		kmem_cache_free(ai->apeb_slab_cache, tmp_aeb);
 	}
 
 	return ret;
@@ -1242,6 +1299,8 @@ static int ubi_write_fastmap(struct ubi_device *ubi,
 	fmh->erase_peb_count = cpu_to_be32(erase_peb_count);
 
 	for (i = 0; i < UBI_MAX_VOLUMES + UBI_INT_VOL_COUNT; i++) {
+		int nconso = 0;
+
 		vol = ubi->volumes[i];
 
 		if (!vol)
@@ -1263,11 +1322,49 @@ static int ubi_write_fastmap(struct ubi_device *ubi,
 		ubi_assert(vol->vol_type == UBI_DYNAMIC_VOLUME ||
 			vol->vol_type == UBI_STATIC_VOLUME);
 
+		if (ubi->consolidated) {
+			struct ubi_fm_consolidated *fconso;
+			struct ubi_fm_consolidated_leb *fcleb;
+
+			fconso = (struct ubi_fm_consolidated *)(fm_raw + fm_pos);
+			for (j = 0; j < vol->reserved_lebs; j++) {
+				struct ubi_leb_desc *cleb;
+				int k;
+
+				cleb = ubi->consolidated[vol->eba_tbl[j]];
+				if (!cleb)
+					continue;
+
+				fcleb = &fconso->lebs[nconso];
+				fcleb->lnum = cpu_to_be32(j);
+				for (k = 0; k < ubi->lebs_per_cpeb;
+				     k++) {
+					if (cleb[k].vol_id != vol->vol_id ||
+					    cleb[k].lnum != j)
+						continue;
+
+					fcleb->peb_pos = cpu_to_be32(k);
+					break;
+				}
+
+				if (k > ubi->lebs_per_cpeb) {
+					ret = -EAGAIN;
+					goto out_kfree;
+				}
+
+				nconso++;
+			}
+
+			if (nconso)
+				fm_pos += sizeof(*fconso) +
+					  (nconso * sizeof(*fcleb));
+		}
+
 		feba = (struct ubi_fm_eba *)(fm_raw + fm_pos);
-		fm_pos += sizeof(*feba) + (sizeof(__be32) * vol->reserved_pebs);
+		fm_pos += sizeof(*feba) + (sizeof(__be32) * vol->reserved_lebs);
 		ubi_assert(fm_pos <= ubi->fm_size);
 
-		for (j = 0; j < vol->reserved_pebs; j++)
+		for (j = 0; j < vol->reserved_lebs; j++)
 			feba->pnum[j] = cpu_to_be32(vol->eba_tbl[j]);
 
 		feba->reserved_pebs = cpu_to_be32(j);
diff --git a/drivers/mtd/ubi/kapi.c b/drivers/mtd/ubi/kapi.c
index dc315c2..c880897 100644
--- a/drivers/mtd/ubi/kapi.c
+++ b/drivers/mtd/ubi/kapi.c
@@ -82,7 +82,7 @@ void ubi_do_get_volume_info(struct ubi_device *ubi, struct ubi_volume *vol,
 {
 	vi->vol_id = vol->vol_id;
 	vi->ubi_num = ubi->ubi_num;
-	vi->size = vol->reserved_pebs;
+	vi->size = vol->reserved_lebs;
 	vi->used_bytes = vol->used_bytes;
 	vi->vol_type = vol->vol_type;
 	vi->corrupted = vol->corrupted;
@@ -532,7 +532,7 @@ int ubi_leb_write(struct ubi_volume_desc *desc, int lnum, const void *buf,
 	if (desc->mode == UBI_READONLY || vol->vol_type == UBI_STATIC_VOLUME)
 		return -EROFS;
 
-	if (lnum < 0 || lnum >= vol->reserved_pebs || offset < 0 || len < 0 ||
+	if (lnum < 0 || lnum >= vol->reserved_lebs || offset < 0 || len < 0 ||
 	    offset + len > vol->usable_leb_size ||
 	    offset & (ubi->min_io_size - 1) || len & (ubi->min_io_size - 1))
 		return -EINVAL;
@@ -577,7 +577,7 @@ int ubi_leb_change(struct ubi_volume_desc *desc, int lnum, const void *buf,
 	if (desc->mode == UBI_READONLY || vol->vol_type == UBI_STATIC_VOLUME)
 		return -EROFS;
 
-	if (lnum < 0 || lnum >= vol->reserved_pebs || len < 0 ||
+	if (lnum < 0 || lnum >= vol->reserved_lebs || len < 0 ||
 	    len > vol->usable_leb_size || len & (ubi->min_io_size - 1))
 		return -EINVAL;
 
@@ -614,7 +614,7 @@ int ubi_leb_erase(struct ubi_volume_desc *desc, int lnum)
 	if (desc->mode == UBI_READONLY || vol->vol_type == UBI_STATIC_VOLUME)
 		return -EROFS;
 
-	if (lnum < 0 || lnum >= vol->reserved_pebs)
+	if (lnum < 0 || lnum >= vol->reserved_lebs)
 		return -EINVAL;
 
 	if (vol->upd_marker)
@@ -674,7 +674,7 @@ int ubi_leb_unmap(struct ubi_volume_desc *desc, int lnum)
 	if (desc->mode == UBI_READONLY || vol->vol_type == UBI_STATIC_VOLUME)
 		return -EROFS;
 
-	if (lnum < 0 || lnum >= vol->reserved_pebs)
+	if (lnum < 0 || lnum >= vol->reserved_lebs)
 		return -EINVAL;
 
 	if (vol->upd_marker)
@@ -710,7 +710,7 @@ int ubi_leb_map(struct ubi_volume_desc *desc, int lnum)
 	if (desc->mode == UBI_READONLY || vol->vol_type == UBI_STATIC_VOLUME)
 		return -EROFS;
 
-	if (lnum < 0 || lnum >= vol->reserved_pebs)
+	if (lnum < 0 || lnum >= vol->reserved_lebs)
 		return -EINVAL;
 
 	if (vol->upd_marker)
@@ -745,7 +745,7 @@ int ubi_is_mapped(struct ubi_volume_desc *desc, int lnum)
 
 	dbg_gen("test LEB %d:%d", vol->vol_id, lnum);
 
-	if (lnum < 0 || lnum >= vol->reserved_pebs)
+	if (lnum < 0 || lnum >= vol->reserved_lebs)
 		return -EINVAL;
 
 	if (vol->upd_marker)
diff --git a/drivers/mtd/ubi/ubi-media.h b/drivers/mtd/ubi/ubi-media.h
index 22ed3f6..500d7da 100644
--- a/drivers/mtd/ubi/ubi-media.h
+++ b/drivers/mtd/ubi/ubi-media.h
@@ -363,7 +363,7 @@ struct ubi_vid_hdr {
  * Empty records contain all zeroes and the CRC checksum of those zeroes.
  */
 struct ubi_vtbl_record {
-	__be32  reserved_pebs;
+	__be32  reserved_lebs;
 	__be32  alignment;
 	__be32  data_pad;
 	__u8    vol_type;
@@ -388,6 +388,7 @@ struct ubi_vtbl_record {
 #define UBI_FM_VHDR_MAGIC	0xFA370ED1
 #define UBI_FM_POOL_MAGIC	0x67AF4D08
 #define UBI_FM_EBA_MAGIC	0xf0c040a8
+#define UBI_FM_CONSO_MAGIC	0xc025011d
 
 /* A fastmap supber block can be located between PEB 0 and
  * UBI_FM_MAX_START */
@@ -444,7 +445,7 @@ struct ubi_fm_hdr {
 	__be32 bad_peb_count;
 	__be32 erase_peb_count;
 	__be32 vol_count;
-	__u8 padding[4];
+	__be32 consolidated_count;
 } __packed;
 
 /* struct ubi_fm_hdr is followed by two struct ubi_fm_scan_pool */
@@ -494,7 +495,8 @@ struct ubi_fm_volhdr {
 	__be32 data_pad;
 	__be32 used_ebs;
 	__be32 last_eb_bytes;
-	__u8 padding2[8];
+	__be32 consolidated_ebs;
+	__u8 padding2[4];
 } __packed;
 
 /* struct ubi_fm_volhdr is followed by one struct ubi_fm_eba records */
@@ -510,4 +512,14 @@ struct ubi_fm_eba {
 	__be32 reserved_pebs;
 	__be32 pnum[0];
 } __packed;
+
+struct ubi_fm_consolidated_leb {
+	__be32 lnum;
+	__be32 peb_pos;
+} __packed;
+
+struct ubi_fm_consolidated {
+	__be32 magic;
+	struct ubi_fm_consolidated_leb lebs[0];
+} __packed;
 #endif /* !__UBI_MEDIA_H__ */
diff --git a/drivers/mtd/ubi/ubi.h b/drivers/mtd/ubi/ubi.h
index aab984f..47e5219 100644
--- a/drivers/mtd/ubi/ubi.h
+++ b/drivers/mtd/ubi/ubi.h
@@ -90,6 +90,12 @@ void ubi_err(const struct ubi_device *ubi, const char *fmt, ...);
 /* The volume ID/LEB number/erase counter is unknown */
 #define UBI_UNKNOWN -1
 
+#ifdef CONFIG_MTD_UBI_CONSOLIDATE
+/* Number of PEBs reserved for consolidation */
+#define UBI_CONSO_RESERVED_PEBS 1
+#else
+#define UBI_CONSO_RESERVED_PEBS 0
+#endif
 /*
  * The UBI debugfs directory name pattern and maximum name length (3 for "ubi"
  * + 2 for the number plus 1 for the trailing zero byte.
@@ -169,6 +175,20 @@ enum {
 };
 
 /**
+ * struct ubi_leb_desc - UBI logical eraseblock description.
+ * @vol_id: volume ID of the locked logical eraseblock
+ * @lnum: locked logical eraseblock number
+ * @lpos: n'th LEB within this PEB, starting at 0
+ *
+ * This data structure is used in describe a logical eraseblock.
+ */
+struct ubi_leb_desc {
+	int vol_id;
+	int lnum;
+	int lpos;
+};
+
+/**
  * struct ubi_wl_entry - wear-leveling entry.
  * @u.rb: link in the corresponding (free/used) RB-tree
  * @u.list: link in the protection queue
@@ -210,6 +230,11 @@ struct ubi_ltree_entry {
 	struct rw_semaphore mutex;
 };
 
+struct ubi_full_leb {
+	struct list_head node;
+	struct ubi_leb_desc desc;
+};
+
 /**
  * struct ubi_rename_entry - volume re-name description data structure.
  * @new_name_len: new volume name length
@@ -269,6 +294,18 @@ struct ubi_fm_pool {
 };
 
 /**
+ * struct ubi_consolidable_leb - UBI consolidable LEB.
+ * @list: links RB-tree nodes
+ * @vol_id: volume ID
+ * @lnum: locked logical eraseblock number
+ */
+struct ubi_consolidable_leb {
+	struct list_head list;
+	int vol_id;
+	int lnum;
+};
+
+/**
  * struct ubi_volume - UBI volume description data structure.
  * @dev: device object to make use of the the Linux device model
  * @cdev: character device object to create character device
@@ -329,7 +366,7 @@ struct ubi_volume {
 	int exclusive;
 	int metaonly;
 
-	int reserved_pebs;
+	int reserved_lebs;
 	int vol_type;
 	int usable_leb_size;
 	int used_ebs;
@@ -561,6 +598,15 @@ struct ubi_device {
 	spinlock_t ltree_lock;
 	struct rb_root ltree;
 
+	int lebs_per_cpeb;
+	struct ubi_leb_desc **consolidated;
+	spinlock_t full_lock;
+	struct list_head full;
+	int full_count;
+	int consolidation_threshold;
+	int consolidation_pnum;
+	struct list_head consolidable;
+
 	/* Fastmap stuff */
 	int fm_disabled;
 	struct ubi_fastmap_layout *fm;
@@ -587,6 +633,7 @@ struct ubi_device {
 	struct mutex work_mutex;
 	struct ubi_work *cur_work;
 	int wl_scheduled;
+	int conso_scheduled;
 	struct ubi_wl_entry **lookuptbl;
 	struct ubi_wl_entry *move_from;
 	struct ubi_wl_entry *move_to;
@@ -648,17 +695,22 @@ struct ubi_device {
  * volume, the @vol_id and @lnum fields are initialized to %UBI_UNKNOWN.
  */
 struct ubi_ainf_peb {
+	struct list_head list;
 	int ec;
 	int pnum;
-	int vol_id;
-	int lnum;
+	int refcount;
 	unsigned int scrub:1;
-	unsigned int copy_flag:1;
+	unsigned int consolidated:1;
+};
+
+struct ubi_ainf_leb {
+	struct rb_node rb;
+	struct ubi_leb_desc desc;
 	unsigned long long sqnum;
-	union {
-		struct rb_node rb;
-		struct list_head list;
-	} u;
+	unsigned int copy_flag:1;
+	unsigned int full:1;
+	int peb_pos;
+	struct ubi_ainf_peb *peb;
 };
 
 /**
@@ -685,6 +737,7 @@ struct ubi_ainf_peb {
 struct ubi_ainf_volume {
 	int vol_id;
 	int highest_lnum;
+	unsigned long long int highest_sqnum;
 	int leb_count;
 	int vol_type;
 	int used_ebs;
@@ -731,6 +784,7 @@ struct ubi_attach_info {
 	struct list_head free;
 	struct list_head erase;
 	struct list_head alien;
+	struct list_head used;
 	int corr_peb_count;
 	int empty_peb_count;
 	int alien_peb_count;
@@ -745,7 +799,8 @@ struct ubi_attach_info {
 	int mean_ec;
 	uint64_t ec_sum;
 	int ec_count;
-	struct kmem_cache *aeb_slab_cache;
+	struct kmem_cache *apeb_slab_cache;
+	struct kmem_cache *aleb_slab_cache;
 };
 
 /**
@@ -790,8 +845,9 @@ extern struct mutex ubi_devices_mutex;
 extern struct blocking_notifier_head ubi_notifiers;
 
 /* attach.c */
-int ubi_add_to_av(struct ubi_device *ubi, struct ubi_attach_info *ai, int pnum,
-		  int ec, const struct ubi_vid_hdr *vid_hdr, int bitflips);
+int ubi_add_to_av(struct ubi_device *ubi, struct ubi_attach_info *ai,
+		  struct ubi_ainf_peb *peb, const struct ubi_vid_hdr *vid_hdr,
+		  int peb_pos, int bitflips, bool full);
 struct ubi_ainf_volume *ubi_find_av(const struct ubi_attach_info *ai,
 				    int vol_id);
 void ubi_remove_av(struct ubi_attach_info *ai, struct ubi_ainf_volume *av);
@@ -849,13 +905,56 @@ int ubi_eba_atomic_leb_change(struct ubi_device *ubi, struct ubi_volume *vol,
 			      int lnum, const void *buf, int len);
 int ubi_eba_copy_leb(struct ubi_device *ubi, int from, int to,
 		     struct ubi_vid_hdr *vid_hdr);
+int ubi_eba_copy_lebs(struct ubi_device *ubi, int from, int to,
+		     struct ubi_vid_hdr *vid_hdr, int nvidh);
 int ubi_eba_init(struct ubi_device *ubi, struct ubi_attach_info *ai);
 unsigned long long ubi_next_sqnum(struct ubi_device *ubi);
+int ubi_eba_leb_write_lock_nested(struct ubi_device *ubi, int vol_id, int lnum,
+				  int level);
+void ubi_eba_leb_write_unlock(struct ubi_device *ubi, int vol_id, int lnum);
 int self_check_eba(struct ubi_device *ubi, struct ubi_attach_info *ai_fastmap,
 		   struct ubi_attach_info *ai_scan);
 
+/* consolidate.c */
+#ifdef CONFIG_MTD_UBI_CONSOLIDATE
+bool ubi_conso_consolidation_needed(struct ubi_device *ubi);
+void ubi_conso_schedule(struct ubi_device *ubi);
+void ubi_eba_consolidate(struct ubi_device *ubi);
+void ubi_conso_remove_full_leb(struct ubi_device *ubi, int vol_id, int lnum);
+struct ubi_leb_desc *ubi_conso_get_consolidated(struct ubi_device *ubi,
+						int pnum);
+bool ubi_conso_invalidate_leb(struct ubi_device *ubi, int pnum,
+			      int vol_id, int lnum);
+int ubi_coso_add_full_leb(struct ubi_device *ubi, int vol_id, int lnum, int lpos);
+int ubi_conso_init(struct ubi_device *ubi);
+void ubi_conso_close(struct ubi_device *ubi);
+#else
+static inline bool ubi_conso_consolidation_needed(struct ubi_device *ubi)
+{
+	return false;
+}
+static inline void ubi_conso_schedule(struct ubi_device *ubi) {}
+static inline void ubi_eba_consolidate(struct ubi_device *ubi) {}
+static inline void ubi_conso_remove_full_leb(struct ubi_device *ubi, int vol_id, int lnum) {}
+static inline struct ubi_leb_desc *ubi_conso_get_consolidated(struct ubi_device *ubi, int pnum)
+{
+	return NULL;
+}
+static inline bool ubi_conso_invalidate_leb(struct ubi_device *ubi, int pnum, int vol_id, int lnum)
+{
+	return true;
+}
+static inline int ubi_coso_add_full_leb(struct ubi_device *ubi, int vol_id, int lnum, int lpos)
+{
+	return 0;
+}
+static inline int ubi_conso_init(struct ubi_device *ubi) { return 0; }
+static inline void ubi_conso_close(struct ubi_device *ubi) {}
+#endif
+
 /* wl.c */
-int ubi_wl_get_peb(struct ubi_device *ubi);
+int ubi_wl_get_peb(struct ubi_device *ubi, bool producing);
+int ubi_wl_flush(struct ubi_device *ubi);
 int ubi_wl_put_peb(struct ubi_device *ubi, int pnum, int torture);
 int ubi_wl_scrub_peb(struct ubi_device *ubi, int pnum);
 int ubi_wl_init(struct ubi_device *ubi, struct ubi_attach_info *ai);
@@ -876,7 +975,9 @@ void ubi_work_close(struct ubi_device *ubi, int error);
 struct ubi_work *ubi_alloc_work(struct ubi_device *ubi);
 int ubi_work_flush(struct ubi_device *ubi);
 bool ubi_work_join_one(struct ubi_device *ubi);
-
+struct ubi_work *ubi_alloc_erase_work(struct ubi_device *ubi,
+				      struct ubi_wl_entry *e,
+				      int torture);
 /* io.c */
 int ubi_io_read(const struct ubi_device *ubi, void *buf, int pnum, int offset,
 		int len);
@@ -923,8 +1024,8 @@ void ubi_do_get_device_info(struct ubi_device *ubi, struct ubi_device_info *di);
 void ubi_do_get_volume_info(struct ubi_device *ubi, struct ubi_volume *vol,
 			    struct ubi_volume_info *vi);
 /* scan.c */
-int ubi_compare_lebs(struct ubi_device *ubi, const struct ubi_ainf_peb *aeb,
-		      int pnum, const struct ubi_vid_hdr *vid_hdr);
+int ubi_compare_lebs(struct ubi_device *ubi, const struct ubi_ainf_leb *aeb,
+		     int pnum, const struct ubi_vid_hdr *vid_hdr);
 
 /* fastmap.c */
 #ifdef CONFIG_MTD_UBI_FASTMAP
@@ -955,6 +1056,22 @@ static inline int ubiblock_remove(struct ubi_volume_info *vi)
 }
 #endif
 
+/**
+ * ubi_get_compat - get compatibility flags of a volume.
+ * @ubi: UBI device description object
+ * @vol_id: volume ID
+ *
+ * This function returns compatibility flags for an internal volume. User
+ * volumes have no compatibility flags, so %0 is returned.
+ */
+static inline int ubi_get_compat(const struct ubi_device *ubi, int vol_id)
+{
+	if (vol_id == UBI_LAYOUT_VOLUME_ID)
+		return UBI_LAYOUT_VOLUME_COMPAT;
+	return 0;
+}
+
+
 /*
  * ubi_for_each_free_peb - walk the UBI free RB tree.
  * @ubi: UBI device description object
@@ -1006,21 +1123,6 @@ static inline int ubiblock_remove(struct ubi_volume_info *vi)
 	     rb = rb_next(rb),                                               \
 	     pos = (rb ? container_of(rb, typeof(*pos), member) : NULL))
 
-/*
- * ubi_move_aeb_to_list - move a PEB from the volume tree to a list.
- *
- * @av: volume attaching information
- * @aeb: attaching eraseblock information
- * @list: the list to move to
- */
-static inline void ubi_move_aeb_to_list(struct ubi_ainf_volume *av,
-					 struct ubi_ainf_peb *aeb,
-					 struct list_head *list)
-{
-		rb_erase(&aeb->u.rb, &av->root);
-		list_add_tail(&aeb->u.list, list);
-}
-
 /**
  * ubi_zalloc_vid_hdr - allocate a volume identifier header object.
  * @ubi: UBI device description object
diff --git a/drivers/mtd/ubi/upd.c b/drivers/mtd/ubi/upd.c
index ffaface..cc21f64 100644
--- a/drivers/mtd/ubi/upd.c
+++ b/drivers/mtd/ubi/upd.c
@@ -142,7 +142,7 @@ int ubi_start_update(struct ubi_device *ubi, struct ubi_volume *vol,
 		return err;
 
 	/* Before updating - wipe out the volume */
-	for (i = 0; i < vol->reserved_pebs; i++) {
+	for (i = 0; i < vol->reserved_lebs; i++) {
 		err = ubi_eba_unmap_leb(ubi, vol, i);
 		if (err)
 			return err;
diff --git a/drivers/mtd/ubi/vmt.c b/drivers/mtd/ubi/vmt.c
index 8a2e081..736acb2 100644
--- a/drivers/mtd/ubi/vmt.c
+++ b/drivers/mtd/ubi/vmt.c
@@ -86,7 +86,7 @@ static ssize_t vol_attribute_show(struct device *dev,
 	spin_unlock(&ubi->volumes_lock);
 
 	if (attr == &attr_vol_reserved_ebs)
-		ret = sprintf(buf, "%d\n", vol->reserved_pebs);
+		ret = sprintf(buf, "%d\n", vol->reserved_lebs);
 	else if (attr == &attr_vol_type) {
 		const char *tp;
 
@@ -158,6 +158,7 @@ int ubi_create_volume(struct ubi_device *ubi, struct ubi_mkvol_req *req)
 	int i, err, vol_id = req->vol_id, do_free = 1;
 	struct ubi_volume *vol;
 	struct ubi_vtbl_record vtbl_rec;
+	int rsvd_pebs = 0;
 	dev_t dev;
 
 	if (ubi->ro_mode)
@@ -208,11 +209,13 @@ int ubi_create_volume(struct ubi_device *ubi, struct ubi_mkvol_req *req)
 
 	/* Calculate how many eraseblocks are requested */
 	vol->usable_leb_size = ubi->leb_size - ubi->leb_size % req->alignment;
-	vol->reserved_pebs = div_u64(req->bytes + vol->usable_leb_size - 1,
+	vol->reserved_lebs = div_u64(req->bytes + vol->usable_leb_size - 1,
 				     vol->usable_leb_size);
 
 	/* Reserve physical eraseblocks */
-	if (vol->reserved_pebs > ubi->avail_pebs) {
+	rsvd_pebs = DIV_ROUND_UP(vol->reserved_lebs,
+				 ubi->lebs_per_cpeb);
+	if (rsvd_pebs > ubi->avail_pebs) {
 		ubi_err(ubi, "not enough PEBs, only %d available",
 			ubi->avail_pebs);
 		if (ubi->corr_peb_count)
@@ -221,8 +224,8 @@ int ubi_create_volume(struct ubi_device *ubi, struct ubi_mkvol_req *req)
 		err = -ENOSPC;
 		goto out_unlock;
 	}
-	ubi->avail_pebs -= vol->reserved_pebs;
-	ubi->rsvd_pebs += vol->reserved_pebs;
+	ubi->avail_pebs -= rsvd_pebs;
+	ubi->rsvd_pebs += rsvd_pebs;
 	spin_unlock(&ubi->volumes_lock);
 
 	vol->vol_id    = vol_id;
@@ -241,17 +244,31 @@ int ubi_create_volume(struct ubi_device *ubi, struct ubi_mkvol_req *req)
 	if (err)
 		goto out_acc;
 
-	vol->eba_tbl = kmalloc(vol->reserved_pebs * sizeof(int), GFP_KERNEL);
+	vol->eba_tbl = kmalloc(vol->reserved_lebs * sizeof(int), GFP_KERNEL);
 	if (!vol->eba_tbl) {
 		err = -ENOMEM;
 		goto out_acc;
 	}
 
-	for (i = 0; i < vol->reserved_pebs; i++)
+#ifdef CONFIG_UBI_EXTENDED_PEB
+	/*
+	 * TODO: check in the underlying MTD device has a page pairing scheme
+	 * requiring the consolidated bitmap creation.
+	 */
+	vol->consolidated = kzalloc(DIV_ROUND_UP(vol->reserved_lebs,
+						 BITS_PER_LONG),
+				    GFP_KERNEL);
+	if (!vol->consolidated) {
+		err = -ENOMEM;
+		goto out_mapping;
+	}
+#endif
+
+	for (i = 0; i < vol->reserved_lebs; i++)
 		vol->eba_tbl[i] = UBI_LEB_UNMAPPED;
 
 	if (vol->vol_type == UBI_DYNAMIC_VOLUME) {
-		vol->used_ebs = vol->reserved_pebs;
+		vol->used_ebs = vol->reserved_lebs;
 		vol->last_eb_bytes = vol->usable_leb_size;
 		vol->used_bytes =
 			(long long)vol->used_ebs * vol->usable_leb_size;
@@ -290,7 +307,7 @@ int ubi_create_volume(struct ubi_device *ubi, struct ubi_mkvol_req *req)
 
 	/* Fill volume table record */
 	memset(&vtbl_rec, 0, sizeof(struct ubi_vtbl_record));
-	vtbl_rec.reserved_pebs = cpu_to_be32(vol->reserved_pebs);
+	vtbl_rec.reserved_lebs = cpu_to_be32(vol->reserved_lebs);
 	vtbl_rec.alignment     = cpu_to_be32(vol->alignment);
 	vtbl_rec.data_pad      = cpu_to_be32(vol->data_pad);
 	vtbl_rec.name_len      = cpu_to_be16(vol->name_len);
@@ -328,12 +345,16 @@ out_sysfs:
 out_cdev:
 	cdev_del(&vol->cdev);
 out_mapping:
-	if (do_free)
+	if (do_free) {
 		kfree(vol->eba_tbl);
+#ifdef CONFIG_UBI_EXTENDED_PEB
+		kfree(vol->consolidated);
+#endif
+	}
 out_acc:
 	spin_lock(&ubi->volumes_lock);
-	ubi->rsvd_pebs -= vol->reserved_pebs;
-	ubi->avail_pebs += vol->reserved_pebs;
+	ubi->rsvd_pebs -= rsvd_pebs;
+	ubi->avail_pebs += rsvd_pebs;
 out_unlock:
 	spin_unlock(&ubi->volumes_lock);
 	if (do_free)
@@ -358,7 +379,7 @@ int ubi_remove_volume(struct ubi_volume_desc *desc, int no_vtbl)
 {
 	struct ubi_volume *vol = desc->vol;
 	struct ubi_device *ubi = vol->ubi;
-	int i, err, vol_id = vol->vol_id, reserved_pebs = vol->reserved_pebs;
+	int i, err, vol_id = vol->vol_id, reserved_pebs;
 
 	dbg_gen("remove device %d, volume %d", ubi->ubi_num, vol_id);
 	ubi_assert(desc->mode == UBI_EXCLUSIVE);
@@ -385,7 +406,7 @@ int ubi_remove_volume(struct ubi_volume_desc *desc, int no_vtbl)
 			goto out_err;
 	}
 
-	for (i = 0; i < vol->reserved_pebs; i++) {
+	for (i = 0; i < vol->reserved_lebs; i++) {
 		err = ubi_eba_unmap_leb(ubi, vol, i);
 		if (err)
 			goto out_err;
@@ -394,6 +415,8 @@ int ubi_remove_volume(struct ubi_volume_desc *desc, int no_vtbl)
 	cdev_del(&vol->cdev);
 	device_unregister(&vol->dev);
 
+	reserved_pebs = DIV_ROUND_UP(vol->reserved_lebs,
+				     ubi->lebs_per_cpeb);
 	spin_lock(&ubi->volumes_lock);
 	ubi->rsvd_pebs -= reserved_pebs;
 	ubi->avail_pebs += reserved_pebs;
@@ -425,9 +448,9 @@ out_unlock:
  * negative error code in case of failure. The caller has to have the
  * @ubi->device_mutex locked.
  */
-int ubi_resize_volume(struct ubi_volume_desc *desc, int reserved_pebs)
+int ubi_resize_volume(struct ubi_volume_desc *desc, int reserved_lebs)
 {
-	int i, err, pebs, *new_mapping;
+	int i, err, lebs, pebs, *new_mapping;
 	struct ubi_volume *vol = desc->vol;
 	struct ubi_device *ubi = vol->ubi;
 	struct ubi_vtbl_record vtbl_rec;
@@ -437,24 +460,24 @@ int ubi_resize_volume(struct ubi_volume_desc *desc, int reserved_pebs)
 		return -EROFS;
 
 	dbg_gen("re-size device %d, volume %d to from %d to %d PEBs",
-		ubi->ubi_num, vol_id, vol->reserved_pebs, reserved_pebs);
+		ubi->ubi_num, vol_id, vol->reserved_lebs, reserved_lebs);
 
 	if (vol->vol_type == UBI_STATIC_VOLUME &&
-	    reserved_pebs < vol->used_ebs) {
+	    reserved_lebs < vol->used_ebs) {
 		ubi_err(ubi, "too small size %d, %d LEBs contain data",
-			reserved_pebs, vol->used_ebs);
+			reserved_lebs, vol->used_ebs);
 		return -EINVAL;
 	}
 
 	/* If the size is the same, we have nothing to do */
-	if (reserved_pebs == vol->reserved_pebs)
+	if (reserved_lebs == vol->reserved_lebs)
 		return 0;
 
-	new_mapping = kmalloc(reserved_pebs * sizeof(int), GFP_KERNEL);
+	new_mapping = kmalloc(reserved_lebs * sizeof(int), GFP_KERNEL);
 	if (!new_mapping)
 		return -ENOMEM;
 
-	for (i = 0; i < reserved_pebs; i++)
+	for (i = 0; i < reserved_lebs; i++)
 		new_mapping[i] = UBI_LEB_UNMAPPED;
 
 	spin_lock(&ubi->volumes_lock);
@@ -466,8 +489,12 @@ int ubi_resize_volume(struct ubi_volume_desc *desc, int reserved_pebs)
 	spin_unlock(&ubi->volumes_lock);
 
 	/* Reserve physical eraseblocks */
-	pebs = reserved_pebs - vol->reserved_pebs;
-	if (pebs > 0) {
+	lebs = reserved_lebs - vol->reserved_lebs;
+	pebs = DIV_ROUND_UP(reserved_lebs,
+			    ubi->lebs_per_cpeb) -
+	       DIV_ROUND_UP(vol->reserved_lebs,
+			    ubi->lebs_per_cpeb);
+	if (lebs > 0) {
 		spin_lock(&ubi->volumes_lock);
 		if (pebs > ubi->avail_pebs) {
 			ubi_err(ubi, "not enough PEBs: requested %d, available %d",
@@ -481,7 +508,7 @@ int ubi_resize_volume(struct ubi_volume_desc *desc, int reserved_pebs)
 		}
 		ubi->avail_pebs -= pebs;
 		ubi->rsvd_pebs += pebs;
-		for (i = 0; i < vol->reserved_pebs; i++)
+		for (i = 0; i < vol->reserved_lebs; i++)
 			new_mapping[i] = vol->eba_tbl[i];
 		kfree(vol->eba_tbl);
 		vol->eba_tbl = new_mapping;
@@ -490,31 +517,32 @@ int ubi_resize_volume(struct ubi_volume_desc *desc, int reserved_pebs)
 
 	/* Change volume table record */
 	vtbl_rec = ubi->vtbl[vol_id];
-	vtbl_rec.reserved_pebs = cpu_to_be32(reserved_pebs);
+	vtbl_rec.reserved_lebs = cpu_to_be32(reserved_lebs);
 	err = ubi_change_vtbl_record(ubi, vol_id, &vtbl_rec);
 	if (err)
 		goto out_acc;
 
-	if (pebs < 0) {
-		for (i = 0; i < -pebs; i++) {
-			err = ubi_eba_unmap_leb(ubi, vol, reserved_pebs + i);
+	if (lebs < 0) {
+		for (i = 0; i < -lebs; i++) {
+			err = ubi_eba_unmap_leb(ubi, vol, reserved_lebs + i);
 			if (err)
 				goto out_acc;
 		}
+
 		spin_lock(&ubi->volumes_lock);
 		ubi->rsvd_pebs += pebs;
 		ubi->avail_pebs -= pebs;
 		ubi_update_reserved(ubi);
-		for (i = 0; i < reserved_pebs; i++)
+		for (i = 0; i < reserved_lebs; i++)
 			new_mapping[i] = vol->eba_tbl[i];
 		kfree(vol->eba_tbl);
 		vol->eba_tbl = new_mapping;
 		spin_unlock(&ubi->volumes_lock);
 	}
 
-	vol->reserved_pebs = reserved_pebs;
+	vol->reserved_lebs = reserved_lebs;
 	if (vol->vol_type == UBI_DYNAMIC_VOLUME) {
-		vol->used_ebs = reserved_pebs;
+		vol->used_ebs = reserved_lebs;
 		vol->last_eb_bytes = vol->usable_leb_size;
 		vol->used_bytes =
 			(long long)vol->used_ebs * vol->usable_leb_size;
@@ -525,7 +553,7 @@ int ubi_resize_volume(struct ubi_volume_desc *desc, int reserved_pebs)
 	return err;
 
 out_acc:
-	if (pebs > 0) {
+	if (lebs > 0) {
 		spin_lock(&ubi->volumes_lock);
 		ubi->rsvd_pebs -= pebs;
 		ubi->avail_pebs += pebs;
@@ -653,7 +681,7 @@ static int self_check_volume(struct ubi_device *ubi, int vol_id)
 	const char *name;
 
 	spin_lock(&ubi->volumes_lock);
-	reserved_pebs = be32_to_cpu(ubi->vtbl[vol_id].reserved_pebs);
+	reserved_pebs = be32_to_cpu(ubi->vtbl[vol_id].reserved_lebs);
 	vol = ubi->volumes[idx];
 
 	if (!vol) {
@@ -665,7 +693,7 @@ static int self_check_volume(struct ubi_device *ubi, int vol_id)
 		return 0;
 	}
 
-	if (vol->reserved_pebs < 0 || vol->alignment < 0 || vol->data_pad < 0 ||
+	if (vol->reserved_lebs < 0 || vol->alignment < 0 || vol->data_pad < 0 ||
 	    vol->name_len < 0) {
 		ubi_err(ubi, "negative values");
 		goto fail;
@@ -698,7 +726,7 @@ static int self_check_volume(struct ubi_device *ubi, int vol_id)
 		goto fail;
 	}
 
-	if (vol->reserved_pebs > ubi->good_peb_count) {
+	if (vol->reserved_lebs > ubi->good_peb_count) {
 		ubi_err(ubi, "too large reserved_pebs");
 		goto fail;
 	}
@@ -727,7 +755,7 @@ static int self_check_volume(struct ubi_device *ubi, int vol_id)
 			ubi_err(ubi, "corrupted dynamic volume");
 			goto fail;
 		}
-		if (vol->used_ebs != vol->reserved_pebs) {
+		if (vol->used_ebs != vol->reserved_lebs) {
 			ubi_err(ubi, "bad used_ebs");
 			goto fail;
 		}
@@ -740,7 +768,7 @@ static int self_check_volume(struct ubi_device *ubi, int vol_id)
 			goto fail;
 		}
 	} else {
-		if (vol->used_ebs < 0 || vol->used_ebs > vol->reserved_pebs) {
+		if (vol->used_ebs < 0 || vol->used_ebs > vol->reserved_lebs) {
 			ubi_err(ubi, "bad used_ebs");
 			goto fail;
 		}
diff --git a/drivers/mtd/ubi/vtbl.c b/drivers/mtd/ubi/vtbl.c
index d85c197..fd40fe5 100644
--- a/drivers/mtd/ubi/vtbl.c
+++ b/drivers/mtd/ubi/vtbl.c
@@ -170,7 +170,7 @@ int ubi_vtbl_rename_volumes(struct ubi_device *ubi,
 static int vtbl_check(const struct ubi_device *ubi,
 		      const struct ubi_vtbl_record *vtbl)
 {
-	int i, n, reserved_pebs, alignment, data_pad, vol_type, name_len;
+	int i, n, reserved_lebs, alignment, data_pad, vol_type, name_len;
 	int upd_marker, err;
 	uint32_t crc;
 	const char *name;
@@ -178,7 +178,7 @@ static int vtbl_check(const struct ubi_device *ubi,
 	for (i = 0; i < ubi->vtbl_slots; i++) {
 		cond_resched();
 
-		reserved_pebs = be32_to_cpu(vtbl[i].reserved_pebs);
+		reserved_lebs = be32_to_cpu(vtbl[i].reserved_lebs);
 		alignment = be32_to_cpu(vtbl[i].alignment);
 		data_pad = be32_to_cpu(vtbl[i].data_pad);
 		upd_marker = vtbl[i].upd_marker;
@@ -194,7 +194,7 @@ static int vtbl_check(const struct ubi_device *ubi,
 			return 1;
 		}
 
-		if (reserved_pebs == 0) {
+		if (reserved_lebs == 0) {
 			if (memcmp(&vtbl[i], &empty_vtbl_record,
 						UBI_VTBL_RECORD_SIZE)) {
 				err = 2;
@@ -203,7 +203,7 @@ static int vtbl_check(const struct ubi_device *ubi,
 			continue;
 		}
 
-		if (reserved_pebs < 0 || alignment < 0 || data_pad < 0 ||
+		if (reserved_lebs < 0 || alignment < 0 || data_pad < 0 ||
 		    name_len < 0) {
 			err = 3;
 			goto bad;
@@ -237,9 +237,10 @@ static int vtbl_check(const struct ubi_device *ubi,
 			goto bad;
 		}
 
-		if (reserved_pebs > ubi->good_peb_count) {
-			ubi_err(ubi, "too large reserved_pebs %d, good PEBs %d",
-				reserved_pebs, ubi->good_peb_count);
+		reserved_lebs = DIV_ROUND_UP(reserved_lebs, ubi->lebs_per_cpeb);
+		if (reserved_lebs > ubi->good_peb_count) {
+			ubi_err(ubi, "too large reserved_lebs %d, good PEBs %d",
+				reserved_lebs, ubi->good_peb_count);
 			err = 9;
 			goto bad;
 		}
@@ -333,12 +334,20 @@ retry:
 	if (err)
 		goto write_error;
 
+	if (ubi->vtbl_size < ubi->leb_size) { //XXX
+		err = ubi_io_write_data(ubi, vtbl, new_aeb->pnum,
+					ubi->leb_size - ubi->min_io_size,
+					ubi->min_io_size);
+	}
+
+	if (err)
+		goto write_error;
 	/*
 	 * And add it to the attaching information. Don't delete the old version
 	 * of this LEB as it will be deleted and freed in 'ubi_add_to_av()'.
 	 */
-	err = ubi_add_to_av(ubi, ai, new_aeb->pnum, new_aeb->ec, vid_hdr, 0);
-	kmem_cache_free(ai->aeb_slab_cache, new_aeb);
+	list_add_tail(&new_aeb->list, &ai->used);
+	err = ubi_add_to_av(ubi, ai, new_aeb, vid_hdr, 0, 0, true);
 	ubi_free_vid_hdr(ubi, vid_hdr);
 	return err;
 
@@ -348,10 +357,10 @@ write_error:
 		 * Probably this physical eraseblock went bad, try to pick
 		 * another one.
 		 */
-		list_add(&new_aeb->u.list, &ai->erase);
+		list_add(&new_aeb->list, &ai->erase);
 		goto retry;
 	}
-	kmem_cache_free(ai->aeb_slab_cache, new_aeb);
+	kmem_cache_free(ai->apeb_slab_cache, new_aeb);
 out_free:
 	ubi_free_vid_hdr(ubi, vid_hdr);
 	return err;
@@ -374,7 +383,7 @@ static struct ubi_vtbl_record *process_lvol(struct ubi_device *ubi,
 {
 	int err;
 	struct rb_node *rb;
-	struct ubi_ainf_peb *aeb;
+	struct ubi_ainf_leb *aeb;
 	struct ubi_vtbl_record *leb[UBI_LAYOUT_VOLUME_EBS] = { NULL, NULL };
 	int leb_corrupted[UBI_LAYOUT_VOLUME_EBS] = {1, 1};
 
@@ -406,15 +415,22 @@ static struct ubi_vtbl_record *process_lvol(struct ubi_device *ubi,
 	dbg_gen("check layout volume");
 
 	/* Read both LEB 0 and LEB 1 into memory */
-	ubi_rb_for_each_entry(rb, aeb, &av->root, u.rb) {
-		leb[aeb->lnum] = vzalloc(ubi->vtbl_size);
-		if (!leb[aeb->lnum]) {
+	ubi_rb_for_each_entry(rb, aeb, &av->root, rb) {
+		leb[aeb->desc.lnum] = vzalloc(ubi->vtbl_size);
+		if (!leb[aeb->desc.lnum]) {
 			err = -ENOMEM;
 			goto out_free;
 		}
 
-		err = ubi_io_read_data(ubi, leb[aeb->lnum], aeb->pnum, 0,
-				       ubi->vtbl_size);
+		if (!aeb->peb->consolidated) {
+			err = ubi_io_read_data(ubi, leb[aeb->desc.lnum],
+				aeb->peb->pnum, 0, ubi->vtbl_size);
+		} else {
+			err = ubi_io_raw_read(ubi, leb[aeb->desc.lnum],
+				aeb->peb->pnum, ubi->leb_start + (aeb->peb_pos * ubi->leb_size),
+				ubi->vtbl_size);
+		}
+
 		if (err == UBI_IO_BITFLIPS || mtd_is_eccerr(err))
 			/*
 			 * Scrub the PEB later. Note, -EBADMSG indicates an
@@ -426,7 +442,7 @@ static struct ubi_vtbl_record *process_lvol(struct ubi_device *ubi,
 			 * aeb->scrub will be cleared in
 			 * 'ubi_add_to_av()'.
 			 */
-			aeb->scrub = 1;
+			aeb->peb->scrub = 1;
 		else if (err)
 			goto out_free;
 	}
@@ -531,21 +547,21 @@ static int init_volumes(struct ubi_device *ubi,
 			const struct ubi_attach_info *ai,
 			const struct ubi_vtbl_record *vtbl)
 {
-	int i, reserved_pebs = 0;
+	int i, reserved_lebs = 0;
 	struct ubi_ainf_volume *av;
 	struct ubi_volume *vol;
 
 	for (i = 0; i < ubi->vtbl_slots; i++) {
 		cond_resched();
 
-		if (be32_to_cpu(vtbl[i].reserved_pebs) == 0)
+		if (be32_to_cpu(vtbl[i].reserved_lebs) == 0)
 			continue; /* Empty record */
 
 		vol = kzalloc(sizeof(struct ubi_volume), GFP_KERNEL);
 		if (!vol)
 			return -ENOMEM;
 
-		vol->reserved_pebs = be32_to_cpu(vtbl[i].reserved_pebs);
+		vol->reserved_lebs = be32_to_cpu(vtbl[i].reserved_lebs);
 		vol->alignment = be32_to_cpu(vtbl[i].alignment);
 		vol->data_pad = be32_to_cpu(vtbl[i].data_pad);
 		vol->upd_marker = vtbl[i].upd_marker;
@@ -573,14 +589,14 @@ static int init_volumes(struct ubi_device *ubi,
 		ubi->volumes[i] = vol;
 		ubi->vol_count += 1;
 		vol->ubi = ubi;
-		reserved_pebs += vol->reserved_pebs;
+		reserved_lebs += vol->reserved_lebs;
 
 		/*
 		 * In case of dynamic volume UBI knows nothing about how many
 		 * data is stored there. So assume the whole volume is used.
 		 */
 		if (vol->vol_type == UBI_DYNAMIC_VOLUME) {
-			vol->used_ebs = vol->reserved_pebs;
+			vol->used_ebs = vol->reserved_lebs;
 			vol->last_eb_bytes = vol->usable_leb_size;
 			vol->used_bytes =
 				(long long)vol->used_ebs * vol->usable_leb_size;
@@ -624,14 +640,14 @@ static int init_volumes(struct ubi_device *ubi,
 	if (!vol)
 		return -ENOMEM;
 
-	vol->reserved_pebs = UBI_LAYOUT_VOLUME_EBS;
+	vol->reserved_lebs = UBI_LAYOUT_VOLUME_EBS;
 	vol->alignment = UBI_LAYOUT_VOLUME_ALIGN;
 	vol->vol_type = UBI_DYNAMIC_VOLUME;
 	vol->name_len = sizeof(UBI_LAYOUT_VOLUME_NAME) - 1;
 	memcpy(vol->name, UBI_LAYOUT_VOLUME_NAME, vol->name_len + 1);
 	vol->usable_leb_size = ubi->leb_size;
-	vol->used_ebs = vol->reserved_pebs;
-	vol->last_eb_bytes = vol->reserved_pebs;
+	vol->used_ebs = vol->reserved_lebs;
+	vol->last_eb_bytes = vol->reserved_lebs;
 	vol->used_bytes =
 		(long long)vol->used_ebs * (ubi->leb_size - vol->data_pad);
 	vol->vol_id = UBI_LAYOUT_VOLUME_ID;
@@ -639,20 +655,21 @@ static int init_volumes(struct ubi_device *ubi,
 
 	ubi_assert(!ubi->volumes[i]);
 	ubi->volumes[vol_id2idx(ubi, vol->vol_id)] = vol;
-	reserved_pebs += vol->reserved_pebs;
+	reserved_lebs += vol->reserved_lebs;
 	ubi->vol_count += 1;
 	vol->ubi = ubi;
 
-	if (reserved_pebs > ubi->avail_pebs) {
+	reserved_lebs = DIV_ROUND_UP(reserved_lebs, ubi->lebs_per_cpeb);
+	if (reserved_lebs > ubi->avail_pebs) {
 		ubi_err(ubi, "not enough PEBs, required %d, available %d",
-			reserved_pebs, ubi->avail_pebs);
+			reserved_lebs, ubi->avail_pebs);
 		if (ubi->corr_peb_count)
 			ubi_err(ubi, "%d PEBs are corrupted and not used",
 				ubi->corr_peb_count);
 		return -ENOSPC;
 	}
-	ubi->rsvd_pebs += reserved_pebs;
-	ubi->avail_pebs -= reserved_pebs;
+	ubi->rsvd_pebs += reserved_lebs;
+	ubi->avail_pebs -= reserved_lebs;
 
 	return 0;
 }
@@ -670,11 +687,11 @@ static int check_av(const struct ubi_volume *vol,
 {
 	int err;
 
-	if (av->highest_lnum >= vol->reserved_pebs) {
+	if (av->highest_lnum >= vol->reserved_lebs) {
 		err = 1;
 		goto bad;
 	}
-	if (av->leb_count > vol->reserved_pebs) {
+	if (av->leb_count > vol->reserved_lebs) {
 		err = 2;
 		goto bad;
 	}
@@ -682,7 +699,7 @@ static int check_av(const struct ubi_volume *vol,
 		err = 3;
 		goto bad;
 	}
-	if (av->used_ebs > vol->reserved_pebs) {
+	if (av->used_ebs > vol->reserved_lebs) {
 		err = 4;
 		goto bad;
 	}
@@ -740,7 +757,7 @@ static int check_attaching_info(const struct ubi_device *ubi,
 			continue;
 		}
 
-		if (vol->reserved_pebs == 0) {
+		if (vol->reserved_lebs == 0) {
 			ubi_assert(i < ubi->vtbl_slots);
 
 			if (!av)
diff --git a/drivers/mtd/ubi/wl.c b/drivers/mtd/ubi/wl.c
index bf4e6b2..ed1031d 100644
--- a/drivers/mtd/ubi/wl.c
+++ b/drivers/mtd/ubi/wl.c
@@ -302,6 +302,11 @@ static struct ubi_wl_entry *find_mean_wl_entry(struct ubi_device *ubi,
 {
 	struct ubi_wl_entry *e, *first, *last;
 
+	ubi_assert(root->rb_node);
+
+	if (!root->rb_node)
+		return NULL;
+
 	first = rb_entry(rb_first(root), struct ubi_wl_entry, u.rb);
 	last = rb_entry(rb_last(root), struct ubi_wl_entry, u.rb);
 
@@ -481,9 +486,8 @@ repeat:
 static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk,
 			int shutdown);
 
-static struct ubi_work *ubi_alloc_erase_work(struct ubi_device *ubi,
-					     struct ubi_wl_entry *e,
-					     int torture)
+struct ubi_work *ubi_alloc_erase_work(struct ubi_device *ubi,
+				      struct ubi_wl_entry *e, int torture)
 {
 	struct ubi_work *wl_wrk;
 
@@ -510,11 +514,12 @@ static struct ubi_work *ubi_alloc_erase_work(struct ubi_device *ubi,
  * failure.
  */
 static int schedule_erase(struct ubi_device *ubi, struct ubi_wl_entry *e,
-			  int torture)
+			  int torture, bool nested)
 {
 	struct ubi_work *wl_wrk;
 
 	ubi_assert(e);
+	ubi_assert(!ubi->consolidated || !ubi->consolidated[e->pnum]);
 
 	dbg_wl("schedule erasure of PEB %d, EC %d, torture %d",
 	       e->pnum, e->ec, torture);
@@ -551,6 +556,7 @@ static int do_sync_erase(struct ubi_device *ubi, struct ubi_wl_entry *e,
 	if (!wl_wrk)
 		return -ENOMEM;
 
+	INIT_LIST_HEAD(&wl_wrk->list);
 	wl_wrk->e = e;
 	wl_wrk->torture = torture;
 
@@ -579,6 +585,7 @@ static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk,
 	struct ubi_wl_entry *e1, *e2;
 	struct ubi_vid_hdr *vid_hdr;
 	int dst_leb_clean = 0;
+	int nvidh = ubi->lebs_per_cpeb;
 
 	if (shutdown)
 		return 0;
@@ -680,7 +687,7 @@ static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk,
 	 * which is being moved was unmapped.
 	 */
 
-	err = ubi_io_read_vid_hdr(ubi, e1->pnum, vid_hdr, 0);
+	err = ubi_io_read_vid_hdrs(ubi, e1->pnum, vid_hdr, &nvidh, 0);
 	if (err && err != UBI_IO_BITFLIPS) {
 		dst_leb_clean = 1;
 		if (err == UBI_IO_FF) {
@@ -714,7 +721,11 @@ static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk,
 		goto out_error;
 	}
 
-	err = ubi_eba_copy_leb(ubi, e1->pnum, e2->pnum, vid_hdr);
+	if (ubi->consolidated && ubi->consolidated[e1->pnum])
+		err = ubi_eba_copy_lebs(ubi, e1->pnum, e2->pnum, vid_hdr, nvidh);
+	else
+		err = ubi_eba_copy_leb(ubi, e1->pnum, e2->pnum, vid_hdr);
+
 	if (err) {
 		if (err == MOVE_CANCEL_RACE) {
 			/*
@@ -917,7 +928,7 @@ static int ensure_wear_leveling(struct ubi_device *ubi)
 	ubi->wl_scheduled = 1;
 	spin_unlock(&ubi->wl_lock);
 
-	wrk = kmalloc(sizeof(struct ubi_work), GFP_NOFS);
+	wrk = ubi_alloc_work(ubi);
 	if (!wrk) {
 		err = -ENOMEM;
 		goto out_cancel;
@@ -982,7 +993,7 @@ static int __erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk)
 		int err1;
 
 		/* Re-schedule the LEB for erasure */
-		err1 = schedule_erase(ubi, e, 0);
+		err1 = schedule_erase(ubi, e, 0, true);
 		if (err1) {
 			wl_entry_destroy(ubi, e);
 			err = err1;
@@ -1092,10 +1103,12 @@ int ubi_wl_put_peb(struct ubi_device *ubi, int pnum, int torture)
 {
 	int err;
 	struct ubi_wl_entry *e;
+	struct ubi_work *wrk;
 
 	dbg_wl("PEB %d", pnum);
 	ubi_assert(pnum >= 0);
 	ubi_assert(pnum < ubi->peb_count);
+	ubi_assert(!ubi->consolidated || !ubi->consolidated[pnum]);
 
 	down_read(&ubi->fm_protect);
 
@@ -1158,15 +1171,20 @@ retry:
 	}
 	spin_unlock(&ubi->wl_lock);
 
-	err = schedule_erase(ubi, e, torture);
-	if (err) {
+	wrk = ubi_alloc_erase_work(ubi, e, torture);
+	if (!wrk) {
 		spin_lock(&ubi->wl_lock);
 		wl_tree_add(e, &ubi->used);
 		spin_unlock(&ubi->wl_lock);
 	}
-
 	up_read(&ubi->fm_protect);
-	return err;
+
+	if (!wrk)
+		return -ENOMEM;
+
+	ubi_schedule_work(ubi, wrk);
+
+	return 0;
 }
 
 /**
@@ -1277,8 +1295,10 @@ int ubi_wl_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
 	int err, i, reserved_pebs, found_pebs = 0;
 	struct rb_node *rb1, *rb2;
 	struct ubi_ainf_volume *av;
-	struct ubi_ainf_peb *aeb, *tmp;
+	struct ubi_ainf_leb *leb;
+	struct ubi_ainf_peb *peb, *tmp;
 	struct ubi_wl_entry *e;
+	struct ubi_leb_desc *clebs;
 
 	ubi->used = ubi->erroneous = ubi->free = ubi->scrub = RB_ROOT;
 	spin_lock_init(&ubi->wl_lock);
@@ -1294,22 +1314,34 @@ int ubi_wl_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
 	if (!ubi->lookuptbl)
 		return err;
 
+	if (ubi->lebs_per_cpeb > 1) {
+		ubi->consolidated = kzalloc(ubi->peb_count * sizeof(void *),
+					    GFP_KERNEL);
+		if (!ubi->consolidated) {
+			kfree(ubi->lookuptbl);
+			return err;
+		}
+	}
+
 	for (i = 0; i < UBI_PROT_QUEUE_LEN; i++)
 		INIT_LIST_HEAD(&ubi->pq[i]);
 	ubi->pq_head = 0;
 
 	ubi->free_count = 0;
-	list_for_each_entry_safe(aeb, tmp, &ai->erase, u.list) {
+	list_for_each_entry_safe(peb, tmp, &ai->erase, list) {
 		cond_resched();
 
+		if (ubi->lookuptbl[peb->pnum])
+			continue;
+
 		e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL);
 		if (!e)
 			goto out_free;
 
-		e->pnum = aeb->pnum;
-		e->ec = aeb->ec;
+		e->pnum = peb->pnum;
+		e->ec = peb->ec;
 		ubi->lookuptbl[e->pnum] = e;
-		if (schedule_erase(ubi, e, 0)) {
+		if (schedule_erase(ubi, e, 0, false)) {
 			wl_entry_destroy(ubi, e);
 			goto out_free;
 		}
@@ -1317,15 +1349,18 @@ int ubi_wl_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
 		found_pebs++;
 	}
 
-	list_for_each_entry(aeb, &ai->free, u.list) {
+	list_for_each_entry(peb, &ai->free, list) {
 		cond_resched();
 
+		if (ubi->lookuptbl[peb->pnum])
+			continue;
+
 		e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL);
 		if (!e)
 			goto out_free;
 
-		e->pnum = aeb->pnum;
-		e->ec = aeb->ec;
+		e->pnum = peb->pnum;
+		e->ec = peb->ec;
 		ubi_assert(e->ec >= 0);
 
 		wl_tree_add(e, &ubi->free);
@@ -1336,29 +1371,54 @@ int ubi_wl_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
 		found_pebs++;
 	}
 
-	ubi_rb_for_each_entry(rb1, av, &ai->volumes, rb) {
-		ubi_rb_for_each_entry(rb2, aeb, &av->root, u.rb) {
-			cond_resched();
+	list_for_each_entry(peb, &ai->used, list) {
+		e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL);
+		if (!e)
+			goto out_free;
 
-			e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL);
-			if (!e)
-				goto out_free;
+		e->pnum = peb->pnum;
+		e->ec = peb->ec;
+		ubi->lookuptbl[e->pnum] = e;
 
-			e->pnum = aeb->pnum;
-			e->ec = aeb->ec;
-			ubi->lookuptbl[e->pnum] = e;
+		if (!peb->scrub) {
+			dbg_wl("add PEB %d EC %d to the used tree",
+			       e->pnum, e->ec);
+			wl_tree_add(e, &ubi->used);
+		} else {
+			dbg_wl("add PEB %d EC %d to the scrub tree",
+			       e->pnum, e->ec);
+			wl_tree_add(e, &ubi->scrub);
+		}
+
+		if (peb->consolidated) {
+			int i;
+
+			clebs = kmalloc(sizeof(*clebs) *
+					ubi->lebs_per_cpeb,
+					GFP_KERNEL);
+			if (!clebs)
+				goto out_free;
 
-			if (!aeb->scrub) {
-				dbg_wl("add PEB %d EC %d to the used tree",
-				       e->pnum, e->ec);
-				wl_tree_add(e, &ubi->used);
-			} else {
-				dbg_wl("add PEB %d EC %d to the scrub tree",
-				       e->pnum, e->ec);
-				wl_tree_add(e, &ubi->scrub);
+			for (i = 0; i < ubi->lebs_per_cpeb; i++) {
+				clebs[i].lnum = -1;
+				clebs[i].vol_id = -1;
 			}
 
-			found_pebs++;
+			ubi->consolidated[peb->pnum] = clebs;
+		}
+
+		found_pebs++;
+	}
+
+	ubi_rb_for_each_entry(rb1, av, &ai->volumes, rb) {
+		ubi_rb_for_each_entry(rb2, leb, &av->root, rb) {
+			cond_resched();
+
+			if (ubi->lebs_per_cpeb > 1) {
+				clebs = ubi->consolidated[leb->peb->pnum];
+				if (clebs)
+					clebs[leb->peb_pos] = leb->desc;
+			}
 		}
 	}
 
@@ -1403,6 +1463,7 @@ out_free:
 	tree_destroy(ubi, &ubi->used);
 	tree_destroy(ubi, &ubi->free);
 	tree_destroy(ubi, &ubi->scrub);
+	kfree(ubi->consolidated);
 	kfree(ubi->lookuptbl);
 	return err;
 }
@@ -1439,6 +1500,7 @@ void ubi_wl_close(struct ubi_device *ubi)
 	tree_destroy(ubi, &ubi->free);
 	tree_destroy(ubi, &ubi->scrub);
 	kfree(ubi->lookuptbl);
+	kfree(ubi->consolidated);
 }
 
 /**
@@ -1536,11 +1598,24 @@ static int self_check_in_pq(const struct ubi_device *ubi,
 	dump_stack();
 	return -EINVAL;
 }
+
+static bool enough_free_pebs(struct ubi_device *ubi)
+{
+	/*
+	 * Hold back one PEB for the producing case,
+	 * currently only for consolidation.
+	 */
+	return ubi->free_count > UBI_CONSO_RESERVED_PEBS;
+}
+
 #ifndef CONFIG_MTD_UBI_FASTMAP
 static struct ubi_wl_entry *get_peb_for_wl(struct ubi_device *ubi)
 {
 	struct ubi_wl_entry *e;
 
+	if (!enough_free_pebs(ubi))
+		return NULL;
+
 	e = find_wl_entry(ubi, &ubi->free, WL_FREE_MAX_DIFF);
 	self_check_in_wl_tree(ubi, e, &ubi->free);
 	ubi->free_count--;
@@ -1563,9 +1638,11 @@ static int produce_free_peb(struct ubi_device *ubi)
 {
 	ubi_assert(spin_is_locked(&ubi->wl_lock));
 
-	while (!ubi->free.rb_node) {
+	while (!enough_free_pebs(ubi)) {
 		spin_unlock(&ubi->wl_lock);
 
+		ubi_eba_consolidate(ubi);
+
 		dbg_wl("do one work synchronously");
 		if (!ubi_work_join_one(ubi)) {
 			spin_lock(&ubi->wl_lock);
@@ -1582,41 +1659,51 @@ static int produce_free_peb(struct ubi_device *ubi)
 /**
  * ubi_wl_get_peb - get a physical eraseblock.
  * @ubi: UBI device description object
+ * @producing: true if this function is being called from a context
+ * which is trying to produce more free PEBs but needs a new one to
+ * achieve that. i.e. consolidatation work.
  *
  * This function returns a physical eraseblock in case of success and a
  * negative error code in case of failure.
  * Returns with ubi->fm_eba_sem held in read mode!
  */
-int ubi_wl_get_peb(struct ubi_device *ubi)
+int ubi_wl_get_peb(struct ubi_device *ubi, bool producing)
 {
-	int err;
+	int err = 0;
 	struct ubi_wl_entry *e;
 
 retry:
 	down_read(&ubi->fm_eba_sem);
 	spin_lock(&ubi->wl_lock);
-	if (!ubi->free.rb_node) {
-		if (ubi->works_count == 0) {
-			ubi_err(ubi, "no free eraseblocks");
-			ubi_assert(list_empty(&ubi->works));
-			spin_unlock(&ubi->wl_lock);
-			return -ENOSPC;
-		}
 
+	if (!enough_free_pebs(ubi) && !producing) {
 		err = produce_free_peb(ubi);
 		if (err < 0) {
+			ubi_err(ubi, "unable to produce free eraseblocks: %i", err);
 			spin_unlock(&ubi->wl_lock);
 			return err;
 		}
 		spin_unlock(&ubi->wl_lock);
 		up_read(&ubi->fm_eba_sem);
 		goto retry;
-
 	}
+	else if (!ubi->free_count && producing) {
+		ubi_err(ubi, "no free eraseblocks in producing case");
+		ubi_assert(0);
+		spin_unlock(&ubi->wl_lock);
+		return -ENOSPC;
+	}
+
 	e = wl_get_wle(ubi);
-	prot_queue_add(ubi, e);
+	if (e)
+		prot_queue_add(ubi, e);
+	else
+		err = -ENOSPC;
 	spin_unlock(&ubi->wl_lock);
 
+	if (err)
+		return err;
+
 	err = ubi_self_check_all_ff(ubi, e->pnum, ubi->vid_hdr_aloffset,
 				    ubi->peb_size - ubi->vid_hdr_aloffset);
 	if (err) {
-- 
2.7.3

[PATCH 12/13] ubi: Add debugfs knob to force LEB consolidation

[Richard Weinberger]

From: David Gstir <david@sigma-star.at>

Normally UBI only consolidates LEB when free space is running low. This debugfs
knob changes this behaviour to always consolidate LEBs when possible.

Signed-off-by: David Gstir <david@sigma-star.at>
Signed-off-by: Richard Weinberger <richard@nod.at>
---
 drivers/mtd/ubi/consolidate.c |  3 +++
 drivers/mtd/ubi/debug.c       | 10 ++++++++++
 drivers/mtd/ubi/debug.h       |  5 +++++
 drivers/mtd/ubi/ubi.h         |  4 ++++
 4 files changed, 22 insertions(+)

diff --git a/drivers/mtd/ubi/consolidate.c b/drivers/mtd/ubi/consolidate.c
index de1479d..7313be0 100644
--- a/drivers/mtd/ubi/consolidate.c
+++ b/drivers/mtd/ubi/consolidate.c
@@ -333,6 +333,9 @@ bool ubi_conso_consolidation_needed(struct ubi_device *ubi)
 	if (!consolidation_possible(ubi))
 		return false;
 
+	if (ubi_dbg_force_leb_consolidation(ubi))
+		return true;
+
 	return ubi->free_count - ubi->beb_rsvd_pebs <=
 	       ubi->consolidation_threshold;
 }
diff --git a/drivers/mtd/ubi/debug.c b/drivers/mtd/ubi/debug.c
index 6178fa1..14b0ce5 100644
--- a/drivers/mtd/ubi/debug.c
+++ b/drivers/mtd/ubi/debug.c
@@ -281,6 +281,8 @@ static ssize_t dfs_file_read(struct file *file, char __user *user_buf,
 		val = d->emulate_bitflips;
 	else if (dent == d->dfs_emulate_io_failures)
 		val = d->emulate_io_failures;
+	else if (dent == d->dfs_force_leb_consolidation)
+		val = d->force_leb_consolidation;
 	else if (dent == d->dfs_emulate_power_cut) {
 		snprintf(buf, sizeof(buf), "%u\n", d->emulate_power_cut);
 		count = simple_read_from_buffer(user_buf, count, ppos,
@@ -374,6 +376,8 @@ static ssize_t dfs_file_write(struct file *file, const char __user *user_buf,
 		d->emulate_bitflips = val;
 	else if (dent == d->dfs_emulate_io_failures)
 		d->emulate_io_failures = val;
+	else if (dent == d->dfs_force_leb_consolidation)
+		d->force_leb_consolidation = val;
 	else
 		count = -EINVAL;
 
@@ -480,6 +484,12 @@ int ubi_debugfs_init_dev(struct ubi_device *ubi)
 		goto out_remove;
 	d->dfs_power_cut_max = dent;
 
+	fname = "force_leb_consolidation";
+	dent = debugfs_create_file(fname, S_IWUSR, d->dfs_dir, (void *)ubi_num,
+				   &dfs_fops);
+	if (IS_ERR_OR_NULL(dent))
+		goto out_remove;
+	d->dfs_force_leb_consolidation = dent;
 	return 0;
 
 out_remove:
diff --git a/drivers/mtd/ubi/debug.h b/drivers/mtd/ubi/debug.h
index c3ed0d5..c9bace4 100644
--- a/drivers/mtd/ubi/debug.h
+++ b/drivers/mtd/ubi/debug.h
@@ -121,6 +121,11 @@ static inline int ubi_dbg_chk_fastmap(const struct ubi_device *ubi)
 	return ubi->dbg.chk_fastmap;
 }
 
+static inline int ubi_dbg_force_leb_consolidation(struct ubi_device *ubi)
+{
+	return ubi->dbg.force_leb_consolidation;
+}
+
 static inline void ubi_enable_dbg_chk_fastmap(struct ubi_device *ubi)
 {
 	ubi->dbg.chk_fastmap = 1;
diff --git a/drivers/mtd/ubi/ubi.h b/drivers/mtd/ubi/ubi.h
index 47e5219..15c8ff6 100644
--- a/drivers/mtd/ubi/ubi.h
+++ b/drivers/mtd/ubi/ubi.h
@@ -414,6 +414,7 @@ struct ubi_wl_entry;
  * @emulate_bitflips: emulate bit-flips for testing purposes
  * @emulate_io_failures: emulate write/erase failures for testing purposes
  * @emulate_power_cut: emulate power cut for testing purposes
+ * @force_leb_consolidation: always consolidate LEBs
  * @power_cut_counter: count down for writes left until emulated power cut
  * @power_cut_min: minimum number of writes before emulating a power cut
  * @power_cut_max: maximum number of writes until emulating a power cut
@@ -424,6 +425,7 @@ struct ubi_wl_entry;
  * @dfs_chk_fastmap: debugfs knob to enable UBI fastmap extra checks
  * @dfs_emulate_bitflips: debugfs knob to emulate bit-flips
  * @dfs_emulate_io_failures: debugfs knob to emulate write/erase failures
+ * @dfs_force_leb_consolidation: debugfs knob to enable constant LEB conso.
  * @dfs_emulate_power_cut: debugfs knob to emulate power cuts
  * @dfs_power_cut_min: debugfs knob for minimum writes before power cut
  * @dfs_power_cut_max: debugfs knob for maximum writes until power cut
@@ -435,6 +437,7 @@ struct ubi_debug_info {
 	unsigned int emulate_bitflips:1;
 	unsigned int emulate_io_failures:1;
 	unsigned int emulate_power_cut:2;
+	unsigned int force_leb_consolidation:1;
 	unsigned int power_cut_counter;
 	unsigned int power_cut_min;
 	unsigned int power_cut_max;
@@ -445,6 +448,7 @@ struct ubi_debug_info {
 	struct dentry *dfs_chk_fastmap;
 	struct dentry *dfs_emulate_bitflips;
 	struct dentry *dfs_emulate_io_failures;
+	struct dentry *dfs_force_leb_consolidation;
 	struct dentry *dfs_emulate_power_cut;
 	struct dentry *dfs_power_cut_min;
 	struct dentry *dfs_power_cut_max;
-- 
2.7.3

[PATCH 13/13] ubi: Add debugfs knob to trigger LEB consolidation

[Richard Weinberger]

...useful in combination with force_leb_consolidation
to test the worker.

Signed-off-by: Richard Weinberger <richard@nod.at>
---
 drivers/mtd/ubi/consolidate.c | 14 ++++++++++++++
 drivers/mtd/ubi/debug.c       | 12 ++++++++++++
 drivers/mtd/ubi/ubi.h         |  5 +++++
 drivers/mtd/ubi/work.c        | 42 +++++++++++++++++++++++++++++-------------
 4 files changed, 60 insertions(+), 13 deletions(-)

diff --git a/drivers/mtd/ubi/consolidate.c b/drivers/mtd/ubi/consolidate.c
index 7313be0..61b9903 100644
--- a/drivers/mtd/ubi/consolidate.c
+++ b/drivers/mtd/ubi/consolidate.c
@@ -359,6 +359,20 @@ void ubi_conso_schedule(struct ubi_device *ubi)
 		BUG();
 }
 
+int ubi_conso_sync(struct ubi_device *ubi)
+{
+	int ret = -ENOMEM;
+
+	struct ubi_work *wrk = ubi_alloc_work(ubi);
+
+	if (wrk) {
+		wrk->func = &consolidation_worker;
+		ret = ubi_schedule_work_sync(ubi, wrk);
+	}
+
+	return ret;
+}
+
 void ubi_eba_consolidate(struct ubi_device *ubi)
 {
 	if (consolidation_possible(ubi) && ubi->consolidation_pnum >= 0)
diff --git a/drivers/mtd/ubi/debug.c b/drivers/mtd/ubi/debug.c
index 14b0ce5..98812bc 100644
--- a/drivers/mtd/ubi/debug.c
+++ b/drivers/mtd/ubi/debug.c
@@ -378,6 +378,11 @@ static ssize_t dfs_file_write(struct file *file, const char __user *user_buf,
 		d->emulate_io_failures = val;
 	else if (dent == d->dfs_force_leb_consolidation)
 		d->force_leb_consolidation = val;
+	else if (dent == d->dfs_trigger_leb_consolidation) {
+		val = ubi_conso_sync(ubi);
+		if (val < 0)
+			count = val;
+	}
 	else
 		count = -EINVAL;
 
@@ -490,6 +495,13 @@ int ubi_debugfs_init_dev(struct ubi_device *ubi)
 	if (IS_ERR_OR_NULL(dent))
 		goto out_remove;
 	d->dfs_force_leb_consolidation = dent;
+
+	fname = "trigger_leb_consolidation";
+	dent = debugfs_create_file(fname, S_IWUSR, d->dfs_dir, (void *)ubi_num,
+				   &dfs_fops);
+	if (IS_ERR_OR_NULL(dent))
+		goto out_remove;
+	d->dfs_trigger_leb_consolidation = dent;
 	return 0;
 
 out_remove:
diff --git a/drivers/mtd/ubi/ubi.h b/drivers/mtd/ubi/ubi.h
index 15c8ff6..f022e9a 100644
--- a/drivers/mtd/ubi/ubi.h
+++ b/drivers/mtd/ubi/ubi.h
@@ -449,6 +449,7 @@ struct ubi_debug_info {
 	struct dentry *dfs_emulate_bitflips;
 	struct dentry *dfs_emulate_io_failures;
 	struct dentry *dfs_force_leb_consolidation;
+	struct dentry *dfs_trigger_leb_consolidation;
 	struct dentry *dfs_emulate_power_cut;
 	struct dentry *dfs_power_cut_min;
 	struct dentry *dfs_power_cut_max;
@@ -932,6 +933,7 @@ bool ubi_conso_invalidate_leb(struct ubi_device *ubi, int pnum,
 int ubi_coso_add_full_leb(struct ubi_device *ubi, int vol_id, int lnum, int lpos);
 int ubi_conso_init(struct ubi_device *ubi);
 void ubi_conso_close(struct ubi_device *ubi);
+int ubi_conso_sync(struct ubi_device *ubi);
 #else
 static inline bool ubi_conso_consolidation_needed(struct ubi_device *ubi)
 {
@@ -954,6 +956,7 @@ static inline int ubi_coso_add_full_leb(struct ubi_device *ubi, int vol_id, int
 }
 static inline int ubi_conso_init(struct ubi_device *ubi) { return 0; }
 static inline void ubi_conso_close(struct ubi_device *ubi) {}
+static inline int ubi_conso_sync(struct ubi_device *ubi) { return 0; }
 #endif
 
 /* wl.c */
@@ -982,6 +985,8 @@ bool ubi_work_join_one(struct ubi_device *ubi);
 struct ubi_work *ubi_alloc_erase_work(struct ubi_device *ubi,
 				      struct ubi_wl_entry *e,
 				      int torture);
+int ubi_schedule_work_sync(struct ubi_device *ubi, struct ubi_work *wrk);
+
 /* io.c */
 int ubi_io_read(const struct ubi_device *ubi, void *buf, int pnum, int offset,
 		int len);
diff --git a/drivers/mtd/ubi/work.c b/drivers/mtd/ubi/work.c
index d2f8ba8..cd1c38b 100644
--- a/drivers/mtd/ubi/work.c
+++ b/drivers/mtd/ubi/work.c
@@ -18,6 +18,18 @@ static bool work_suspended(struct ubi_device *ubi)
 	return ubi->thread_suspended || !ubi->thread_enabled;
 }
 
+/**
+ * destroy_work - destroy an UBI work.
+ * @ref: kref object
+ *
+ * This function is called by kref upon the last reference is gone.
+ */
+static void destroy_work(struct kref *ref)
+{
+	struct ubi_work *wrk = container_of(ref, struct ubi_work, ref);
+
+	kfree(wrk);
+}
 
 /**
  * ubi_schedule_work - schedule a work.
@@ -42,6 +54,23 @@ void ubi_schedule_work(struct ubi_device *ubi, struct ubi_work *wrk)
 	mutex_unlock(&ubi->work_mutex);
 }
 
+int ubi_schedule_work_sync(struct ubi_device *ubi, struct ubi_work *wrk)
+{
+	int ret;
+
+	kref_get(&wrk->ref);
+
+	ubi_schedule_work(ubi, wrk);
+	wait_for_completion(&wrk->comp);
+
+	spin_lock(&ubi->wl_lock);
+	ret = wrk->ret;
+	kref_put(&wrk->ref, destroy_work);
+	spin_unlock(&ubi->wl_lock);
+
+	return ret;
+}
+
 struct ubi_work *ubi_alloc_work(struct ubi_device *ubi)
 {
 	struct ubi_work *wrk;
@@ -57,19 +86,6 @@ struct ubi_work *ubi_alloc_work(struct ubi_device *ubi)
 	return wrk;
 }
 
-/**
- * destroy_work - destroy an UBI work.
- * @ref: kref object
- *
- * This function is called by kref upon the last reference is gone.
- */
-static void destroy_work(struct kref *ref)
-{
-	struct ubi_work *wrk = container_of(ref, struct ubi_work, ref);
-
-	kfree(wrk);
-}
-
 static void shutdown_work(struct ubi_device *ubi, int error)
 {
 	struct ubi_work *wrk;
-- 
2.7.3

Re: [PATCH 04/13] ubi: Kill ubi->alc_mutex

[Richard Weinberger]

Am 30.05.2016 um 14:04 schrieb Richard Weinberger:
> With our new work model work can never run in parallel.
> So, the serialization mutex is in vain
> 
> Signed-off-by: Richard Weinberger <richard@nod.at>

Please ignore this patch. It shouldn't be part of this series.
I've dropped it already.

Thanks,
//richard

Re: [RFC] UBI: MLC Support v0

[Boris Brezillon]

On Mon, 30 May 2016 14:04:21 +0200
Richard Weinberger <richard@nod.at> wrote:

> as announced in [1] I'd like to submit the very first patch series to
> have MLC support in UBI (and therefore also UBIFS).
> The code has not changed much since the announcement but now the patches
> are in a more reviewable shape. Still not mainline ready but better.
> I'm sending this now as I'm on vacation until June 15th and hope that
> we get some reviews on the concept.
> The series consists of three parts:
> 	a) Generic UBI rework to make the MLC implementation possible.
> 	b) SLC emulation mode to access only lower pages on MLC NAND.
> 	c) LEB consolidation to exploit MLC NAND's higher pages.
> 
> [1] http://lists.infradead.org/pipermail/linux-mtd/2016-April/067322.html

Artem, I know you're quite busy, but that would be great to have your
feedback. Not asking for a detailed review, but at least some comments
on the general approach.

Bean, I'd like to have your feedback too. You proposed a solution to
address the same problem a while ago, and your feedback on this
solution would be valuable.

Thanks,

Boris

> 
> Thanks,
> //richard
> 
> ---
> 
> David Gstir (1):
>       ubi: Add debugfs knob to force LEB consolidation
> 
> Richard Weinberger (12):
>       ubi: Undo "UBI: modify ubi_wl_flush function to clear work queue for a lnum"
>       ubi: Rework UBI worker
>       ubi: auto re-size after UBI thread is ready
>       ubi: Kill ubi->alc_mutex
>       ubi: Get rid of __schedule_ubi_work()
>       ubi: Remove tst_disable_bgt debugfs knob
>       ubi: Move work related functions to work.c
>       ubi: Remove lnum and vol_id from erase work
>       ubi: Remove usless debug info from wear_leveling_worker()
>       ubi: SLC mode
>       ubi: LEB consolidation
>       ubi: Add debugfs knob to trigger LEB consolidation
> 
>  drivers/mtd/ubi/Kconfig       |   4 +
>  drivers/mtd/ubi/Makefile      |   3 +-
>  drivers/mtd/ubi/attach.c      | 407 ++++++++++++++++++--------------
>  drivers/mtd/ubi/build.c       |  41 ++--
>  drivers/mtd/ubi/cdev.c        |  14 +-
>  drivers/mtd/ubi/consolidate.c | 516 ++++++++++++++++++++++++++++++++++++++++
>  drivers/mtd/ubi/debug.c       |  49 ++--
>  drivers/mtd/ubi/debug.h       |  19 +-
>  drivers/mtd/ubi/eba.c         | 429 ++++++++++++++++++++++++++++------
>  drivers/mtd/ubi/fastmap-wl.c  |  33 ++-
>  drivers/mtd/ubi/fastmap.c     | 265 ++++++++++++++-------
>  drivers/mtd/ubi/io.c          | 426 +++++++++++++++++++++++++--------
>  drivers/mtd/ubi/kapi.c        |  43 +---
>  drivers/mtd/ubi/ubi-media.h   |  18 +-
>  drivers/mtd/ubi/ubi.h         | 221 ++++++++++++++----
>  drivers/mtd/ubi/upd.c         |   7 +-
>  drivers/mtd/ubi/vmt.c         | 106 +++++----
>  drivers/mtd/ubi/vtbl.c        |  87 ++++---
>  drivers/mtd/ubi/wl.c          | 531 +++++++++++++++---------------------------
>  drivers/mtd/ubi/work.c        | 332 ++++++++++++++++++++++++++
>  include/linux/mtd/ubi.h       |   1 -
>  21 files changed, 2554 insertions(+), 998 deletions(-)
> 
> ______________________________________________________
> Linux MTD discussion mailing list
> http://lists.infradead.org/mailman/listinfo/linux-mtd/



-- 
Boris Brezillon, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

Re: [PATCH 11/13] ubi: LEB consolidation

[Boris Brezillon]

On Mon, 30 May 2016 14:04:32 +0200
Richard Weinberger <richard@nod.at> wrote:

> Implements LEB consolidation for MLC NAND. By default data is written
> in SLC mode. As soon UBI runs out of space two fully written PEBs will
> get merged to produce free space.
> The merge operation takes two PEB written in SLC mode and writes back
> all data into one PEB in MLC mode. The operation works almost as atomic
> LEB change and is power cut safe.
> This patch was implemented by Boris Brezillon and me.
> 
> TODO:
> - Reconstruct “full”-states upon attach
> - Make all paths generic for pairings > 2 too to support TLC NAND in future

Hm, I thought it was already the case. Could you point places that are
assuming ->lebs_per_consolidated_peb is <= 2?

> - Cut the new on-flash layout into stone and raise UBI version (by having feature flags)
> 
> Signed-off-by: Richard Weinberger <richard@nod.at>
> ---
>  drivers/mtd/ubi/Kconfig       |   4 +
>  drivers/mtd/ubi/Makefile      |   1 +
>  drivers/mtd/ubi/attach.c      | 407 ++++++++++++++++++++--------------
>  drivers/mtd/ubi/build.c       |  18 +-
>  drivers/mtd/ubi/cdev.c        |  12 +-
>  drivers/mtd/ubi/consolidate.c | 499 ++++++++++++++++++++++++++++++++++++++++++
>  drivers/mtd/ubi/debug.c       |  16 +-
>  drivers/mtd/ubi/debug.h       |   2 +-
>  drivers/mtd/ubi/eba.c         | 411 ++++++++++++++++++++++++++++------
>  drivers/mtd/ubi/fastmap-wl.c  |  19 +-
>  drivers/mtd/ubi/fastmap.c     | 261 +++++++++++++++-------
>  drivers/mtd/ubi/kapi.c        |  14 +-
>  drivers/mtd/ubi/ubi-media.h   |  18 +-
>  drivers/mtd/ubi/ubi.h         | 162 +++++++++++---
>  drivers/mtd/ubi/upd.c         |   2 +-
>  drivers/mtd/ubi/vmt.c         | 104 +++++----
>  drivers/mtd/ubi/vtbl.c        |  87 +++++---
>  drivers/mtd/ubi/wl.c          | 187 +++++++++++-----
>  18 files changed, 1718 insertions(+), 506 deletions(-)
>  create mode 100644 drivers/mtd/ubi/consolidate.c
> 
> diff --git a/drivers/mtd/ubi/Kconfig b/drivers/mtd/ubi/Kconfig
> index f0855ce..a6f7d3b 100644
> --- a/drivers/mtd/ubi/Kconfig
> +++ b/drivers/mtd/ubi/Kconfig
> @@ -103,4 +103,8 @@ config MTD_UBI_BLOCK
>  
>  	   If in doubt, say "N".
>  
> +config MTD_UBI_CONSOLIDATE
> +	bool "LEB consolidation support"
> +	default n
> +
>  endif # MTD_UBI
> diff --git a/drivers/mtd/ubi/Makefile b/drivers/mtd/ubi/Makefile
> index e9d4b1d..86a77c5 100644
> --- a/drivers/mtd/ubi/Makefile
> +++ b/drivers/mtd/ubi/Makefile
> @@ -4,5 +4,6 @@ ubi-y += vtbl.o vmt.o upd.o build.o cdev.o kapi.o eba.o io.o wl.o attach.o
>  ubi-y += work.o misc.o debug.o
>  ubi-$(CONFIG_MTD_UBI_FASTMAP) += fastmap.o
>  ubi-$(CONFIG_MTD_UBI_BLOCK) += block.o
> +ubi-$(CONFIG_MTD_UBI_CONSOLIDATE) += consolidate.o
>  
>  obj-$(CONFIG_MTD_UBI_GLUEBI) += gluebi.o
> diff --git a/drivers/mtd/ubi/attach.c b/drivers/mtd/ubi/attach.c
> index c1aaf03..d1bd34c 100644
> --- a/drivers/mtd/ubi/attach.c
> +++ b/drivers/mtd/ubi/attach.c
> @@ -116,10 +116,14 @@ static struct ubi_vid_hdr *vidh;
>   * returns zero in case of success and a negative error code in case of
>   * failure.
>   */
> -static int add_to_list(struct ubi_attach_info *ai, int pnum, int vol_id,
> -		       int lnum, int ec, int to_head, struct list_head *list)
> +static void add_peb_to_list(struct ubi_attach_info *ai,
> +			    struct ubi_ainf_peb *aeb, int to_head,
> +			    struct list_head *list)
>  {
> -	struct ubi_ainf_peb *aeb;
> +	int pnum = aeb->pnum, ec = aeb->ec;
> +
> +	if (!list_empty(&aeb->list))
> +		list_del(&aeb->list);
>  
>  	if (list == &ai->free) {
>  		dbg_bld("add to free: PEB %d, EC %d", pnum, ec);
> @@ -131,18 +135,28 @@ static int add_to_list(struct ubi_attach_info *ai, int pnum, int vol_id,
>  	} else
>  		BUG();
>  
> -	aeb = kmem_cache_alloc(ai->aeb_slab_cache, GFP_KERNEL);
> +	if (to_head)
> +		list_add(&aeb->list, list);
> +	else
> +		list_add_tail(&aeb->list, list);
> +
> +}
> +
> +static int add_to_list(struct ubi_attach_info *ai, int pnum, int ec,
> +		       int to_head, struct list_head *list)
> +{
> +	struct ubi_ainf_peb *aeb;
> +
> +	aeb = kmem_cache_zalloc(ai->apeb_slab_cache, GFP_KERNEL);
>  	if (!aeb)
>  		return -ENOMEM;
>  
> +	INIT_LIST_HEAD(&aeb->list);
>  	aeb->pnum = pnum;
> -	aeb->vol_id = vol_id;
> -	aeb->lnum = lnum;
>  	aeb->ec = ec;
> -	if (to_head)
> -		list_add(&aeb->u.list, list);
> -	else
> -		list_add_tail(&aeb->u.list, list);
> +
> +	add_peb_to_list(ai, aeb, to_head, list);
> +
>  	return 0;
>  }
>  
> @@ -163,14 +177,15 @@ static int add_corrupted(struct ubi_attach_info *ai, int pnum, int ec)
>  
>  	dbg_bld("add to corrupted: PEB %d, EC %d", pnum, ec);
>  
> -	aeb = kmem_cache_alloc(ai->aeb_slab_cache, GFP_KERNEL);
> +	aeb = kmem_cache_alloc(ai->apeb_slab_cache, GFP_KERNEL);
>  	if (!aeb)
>  		return -ENOMEM;
>  
> +	INIT_LIST_HEAD(&aeb->list);
>  	ai->corr_peb_count += 1;
>  	aeb->pnum = pnum;
>  	aeb->ec = ec;
> -	list_add(&aeb->u.list, &ai->corr);
> +	list_add(&aeb->list, &ai->corr);
>  	return 0;
>  }
>  
> @@ -321,8 +336,8 @@ static struct ubi_ainf_volume *add_volume(struct ubi_attach_info *ai,
>   *     o bit 2 is cleared: the older LEB is not corrupted;
>   *     o bit 2 is set: the older LEB is corrupted.
>   */
> -int ubi_compare_lebs(struct ubi_device *ubi, const struct ubi_ainf_peb *aeb,
> -			int pnum, const struct ubi_vid_hdr *vid_hdr)
> +int ubi_compare_lebs(struct ubi_device *ubi, const struct ubi_ainf_leb *aeb,
> +		     int pnum, const struct ubi_vid_hdr *vid_hdr)
>  {
>  	int len, err, second_is_newer, bitflips = 0, corrupted = 0;
>  	uint32_t data_crc, crc;
> @@ -362,6 +377,8 @@ int ubi_compare_lebs(struct ubi_device *ubi, const struct ubi_ainf_peb *aeb,
>  			return 1;
>  		}
>  	} else {
> +		int nvidh = ubi->lebs_per_cpeb;
> +
>  		if (!aeb->copy_flag) {
>  			/* It is not a copy, so it is newer */
>  			dbg_bld("first PEB %d is newer, copy_flag is unset",
> @@ -373,8 +390,8 @@ int ubi_compare_lebs(struct ubi_device *ubi, const struct ubi_ainf_peb *aeb,
>  		if (!vh)
>  			return -ENOMEM;
>  
> -		pnum = aeb->pnum;
> -		err = ubi_io_read_vid_hdr(ubi, pnum, vh, 0);
> +		pnum = aeb->peb->pnum;
> +		err = ubi_io_read_vid_hdrs(ubi, pnum, vh, &nvidh, 0);
>  		if (err) {
>  			if (err == UBI_IO_BITFLIPS)
>  				bitflips = 1;
> @@ -388,7 +405,8 @@ int ubi_compare_lebs(struct ubi_device *ubi, const struct ubi_ainf_peb *aeb,
>  			}
>  		}
>  
> -		vid_hdr = vh;
> +		ubi_assert(aeb->peb_pos < nvidh);
> +		vid_hdr = &vh[aeb->peb_pos];
>  	}
>  
>  	/* Read the data of the copy and check the CRC */
> @@ -446,18 +464,21 @@ out_free_vidh:
>   * to be picked, while the older one has to be dropped. This function returns
>   * zero in case of success and a negative error code in case of failure.
>   */
> -int ubi_add_to_av(struct ubi_device *ubi, struct ubi_attach_info *ai, int pnum,
> -		  int ec, const struct ubi_vid_hdr *vid_hdr, int bitflips)
> +int ubi_add_to_av(struct ubi_device *ubi, struct ubi_attach_info *ai,
> +		  struct ubi_ainf_peb *peb, const struct ubi_vid_hdr *vid_hdr,
> +		  int peb_pos, int bitflips, bool full)
>  {
> -	int err, vol_id, lnum;
> +	int err, vol_id, lnum, pnum, ec;
>  	unsigned long long sqnum;
>  	struct ubi_ainf_volume *av;
> -	struct ubi_ainf_peb *aeb;
> +	struct ubi_ainf_leb *leb;
>  	struct rb_node **p, *parent = NULL;
>  
>  	vol_id = be32_to_cpu(vid_hdr->vol_id);
>  	lnum = be32_to_cpu(vid_hdr->lnum);
>  	sqnum = be64_to_cpu(vid_hdr->sqnum);
> +	pnum = peb->pnum;
> +	ec = peb->ec;
>  
>  	dbg_bld("PEB %d, LEB %d:%d, EC %d, sqnum %llu, bitflips %d",
>  		pnum, vol_id, lnum, ec, sqnum, bitflips);
> @@ -478,9 +499,9 @@ int ubi_add_to_av(struct ubi_device *ubi, struct ubi_attach_info *ai, int pnum,
>  		int cmp_res;
>  
>  		parent = *p;
> -		aeb = rb_entry(parent, struct ubi_ainf_peb, u.rb);
> -		if (lnum != aeb->lnum) {
> -			if (lnum < aeb->lnum)
> +		leb = rb_entry(parent, struct ubi_ainf_leb, rb);
> +		if (lnum != leb->desc.lnum) {
> +			if (lnum < leb->desc.lnum)
>  				p = &(*p)->rb_left;
>  			else
>  				p = &(*p)->rb_right;
> @@ -493,7 +514,7 @@ int ubi_add_to_av(struct ubi_device *ubi, struct ubi_attach_info *ai, int pnum,
>  		 */
>  
>  		dbg_bld("this LEB already exists: PEB %d, sqnum %llu, EC %d",
> -			aeb->pnum, aeb->sqnum, aeb->ec);
> +			leb->peb->pnum, leb->sqnum, leb->peb->ec);
>  
>  		/*
>  		 * Make sure that the logical eraseblocks have different
> @@ -508,10 +529,10 @@ int ubi_add_to_av(struct ubi_device *ubi, struct ubi_attach_info *ai, int pnum,
>  		 * images, but refuse attaching old images with duplicated
>  		 * logical eraseblocks because there was an unclean reboot.
>  		 */
> -		if (aeb->sqnum == sqnum && sqnum != 0) {
> +		if (leb->sqnum == sqnum && sqnum != 0) {
>  			ubi_err(ubi, "two LEBs with same sequence number %llu",
>  				sqnum);
> -			ubi_dump_aeb(aeb, 0);
> +			ubi_dump_aeb(leb, 0);
>  			ubi_dump_vid_hdr(vid_hdr);
>  			return -EINVAL;
>  		}
> @@ -520,7 +541,7 @@ int ubi_add_to_av(struct ubi_device *ubi, struct ubi_attach_info *ai, int pnum,
>  		 * Now we have to drop the older one and preserve the newer
>  		 * one.
>  		 */
> -		cmp_res = ubi_compare_lebs(ubi, aeb, pnum, vid_hdr);
> +		cmp_res = ubi_compare_lebs(ubi, leb, pnum, vid_hdr);
>  		if (cmp_res < 0)
>  			return cmp_res;
>  
> @@ -533,19 +554,16 @@ int ubi_add_to_av(struct ubi_device *ubi, struct ubi_attach_info *ai, int pnum,
>  			if (err)
>  				return err;
>  
> -			err = add_to_list(ai, aeb->pnum, aeb->vol_id,
> -					  aeb->lnum, aeb->ec, cmp_res & 4,
> -					  &ai->erase);
> -			if (err)
> -				return err;
> +			if (--leb->peb->refcount <= 0)
> +				add_peb_to_list(ai, leb->peb, cmp_res & 4,
> +						&ai->erase);
>  
> -			aeb->ec = ec;
> -			aeb->pnum = pnum;
> -			aeb->vol_id = vol_id;
> -			aeb->lnum = lnum;
> -			aeb->scrub = ((cmp_res & 2) || bitflips);
> -			aeb->copy_flag = vid_hdr->copy_flag;
> -			aeb->sqnum = sqnum;
> +			leb->peb_pos = peb_pos;
> +			leb->peb = peb;
> +			peb->scrub = ((cmp_res & 2) || bitflips || peb->scrub);
> +			leb->copy_flag = vid_hdr->copy_flag;
> +			leb->sqnum = sqnum;
> +			leb->full = full;
>  
>  			if (av->highest_lnum == lnum)
>  				av->last_data_size =
> @@ -557,8 +575,11 @@ int ubi_add_to_av(struct ubi_device *ubi, struct ubi_attach_info *ai, int pnum,
>  			 * This logical eraseblock is older than the one found
>  			 * previously.
>  			 */
> -			return add_to_list(ai, pnum, vol_id, lnum, ec,
> -					   cmp_res & 4, &ai->erase);
> +			if (--peb->refcount <= 0)
> +				add_peb_to_list(ai, peb, cmp_res & 4,
> +						&ai->erase);
> +
> +			return 0;
>  		}
>  	}
>  
> @@ -571,17 +592,18 @@ int ubi_add_to_av(struct ubi_device *ubi, struct ubi_attach_info *ai, int pnum,
>  	if (err)
>  		return err;
>  
> -	aeb = kmem_cache_alloc(ai->aeb_slab_cache, GFP_KERNEL);
> -	if (!aeb)
> +	leb = kmem_cache_alloc(ai->aleb_slab_cache, GFP_KERNEL);
> +	if (!leb)
>  		return -ENOMEM;
>  
> -	aeb->ec = ec;
> -	aeb->pnum = pnum;
> -	aeb->vol_id = vol_id;
> -	aeb->lnum = lnum;
> -	aeb->scrub = bitflips;
> -	aeb->copy_flag = vid_hdr->copy_flag;
> -	aeb->sqnum = sqnum;
> +	leb->peb = peb;
> +	leb->peb_pos = peb_pos;
> +	leb->desc.vol_id = vol_id;
> +	leb->desc.lnum = lnum;
> +	peb->scrub = (bitflips || peb->scrub);
> +	leb->copy_flag = vid_hdr->copy_flag;
> +	leb->sqnum = sqnum;
> +	leb->full = full;
>  
>  	if (av->highest_lnum <= lnum) {
>  		av->highest_lnum = lnum;
> @@ -589,8 +611,8 @@ int ubi_add_to_av(struct ubi_device *ubi, struct ubi_attach_info *ai, int pnum,
>  	}
>  
>  	av->leb_count += 1;
> -	rb_link_node(&aeb->u.rb, parent, p);
> -	rb_insert_color(&aeb->u.rb, &av->root);
> +	rb_link_node(&leb->rb, parent, p);
> +	rb_insert_color(&leb->rb, &av->root);
>  	return 0;
>  }
>  
> @@ -631,14 +653,17 @@ struct ubi_ainf_volume *ubi_find_av(const struct ubi_attach_info *ai,
>  void ubi_remove_av(struct ubi_attach_info *ai, struct ubi_ainf_volume *av)
>  {
>  	struct rb_node *rb;
> -	struct ubi_ainf_peb *aeb;
> +	struct ubi_ainf_leb *aeb;
>  
>  	dbg_bld("remove attaching information about volume %d", av->vol_id);
>  
>  	while ((rb = rb_first(&av->root))) {
> -		aeb = rb_entry(rb, struct ubi_ainf_peb, u.rb);
> -		rb_erase(&aeb->u.rb, &av->root);
> -		list_add_tail(&aeb->u.list, &ai->erase);
> +		aeb = rb_entry(rb, struct ubi_ainf_leb, rb);
> +		rb_erase(&aeb->rb, &av->root);
> +		if (--aeb->peb->refcount <= 0)
> +			list_move(&aeb->peb->list, &ai->erase);
> +
> +		kmem_cache_free(ai->aleb_slab_cache, aeb);
>  	}
>  
>  	rb_erase(&av->rb, &ai->volumes);
> @@ -713,8 +738,8 @@ struct ubi_ainf_peb *ubi_early_get_peb(struct ubi_device *ubi,
>  	struct ubi_ainf_peb *aeb, *tmp_aeb;
>  
>  	if (!list_empty(&ai->free)) {
> -		aeb = list_entry(ai->free.next, struct ubi_ainf_peb, u.list);
> -		list_del(&aeb->u.list);
> +		aeb = list_entry(ai->free.next, struct ubi_ainf_peb, list);
> +		list_del(&aeb->list);
>  		dbg_bld("return free PEB %d, EC %d", aeb->pnum, aeb->ec);
>  		return aeb;
>  	}
> @@ -725,7 +750,7 @@ struct ubi_ainf_peb *ubi_early_get_peb(struct ubi_device *ubi,
>  	 * so forth. We don't want to take care about bad eraseblocks here -
>  	 * they'll be handled later.
>  	 */
> -	list_for_each_entry_safe(aeb, tmp_aeb, &ai->erase, u.list) {
> +	list_for_each_entry_safe(aeb, tmp_aeb, &ai->erase, list) {
>  		if (aeb->ec == UBI_UNKNOWN)
>  			aeb->ec = ai->mean_ec;
>  
> @@ -734,7 +759,7 @@ struct ubi_ainf_peb *ubi_early_get_peb(struct ubi_device *ubi,
>  			continue;
>  
>  		aeb->ec += 1;
> -		list_del(&aeb->u.list);
> +		list_del(&aeb->list);
>  		dbg_bld("return PEB %d, EC %d", aeb->pnum, aeb->ec);
>  		return aeb;
>  	}
> @@ -820,7 +845,9 @@ static int scan_peb(struct ubi_device *ubi, struct ubi_attach_info *ai,
>  		    int pnum, int *vid, unsigned long long *sqnum)
>  {
>  	long long uninitialized_var(ec);
> -	int err, bitflips = 0, vol_id = -1, ec_err = 0;
> +	int err, bitflips = 0, vol_id = -1, ec_err = 0, nvidh, i;
> +	struct ubi_ainf_peb *aeb;
> +	bool full = false;
>  
>  	dbg_bld("scan PEB %d", pnum);
>  
> @@ -844,12 +871,10 @@ static int scan_peb(struct ubi_device *ubi, struct ubi_attach_info *ai,
>  		break;
>  	case UBI_IO_FF:
>  		ai->empty_peb_count += 1;
> -		return add_to_list(ai, pnum, UBI_UNKNOWN, UBI_UNKNOWN,
> -				   UBI_UNKNOWN, 0, &ai->erase);
> +		return add_to_list(ai, pnum, UBI_UNKNOWN, 0, &ai->erase);
>  	case UBI_IO_FF_BITFLIPS:
>  		ai->empty_peb_count += 1;
> -		return add_to_list(ai, pnum, UBI_UNKNOWN, UBI_UNKNOWN,
> -				   UBI_UNKNOWN, 1, &ai->erase);
> +		return add_to_list(ai, pnum, UBI_UNKNOWN, 1, &ai->erase);
>  	case UBI_IO_BAD_HDR_EBADMSG:
>  	case UBI_IO_BAD_HDR:
>  		/*
> @@ -915,8 +940,8 @@ static int scan_peb(struct ubi_device *ubi, struct ubi_attach_info *ai,
>  	}
>  
>  	/* OK, we've done with the EC header, let's look at the VID header */
> -
> -	err = ubi_io_read_vid_hdr(ubi, pnum, vidh, 0);
> +	nvidh = ubi->lebs_per_cpeb;
> +	err = ubi_io_read_vid_hdrs(ubi, pnum, vidh, &nvidh, 0);
>  	if (err < 0)
>  		return err;
>  	switch (err) {
> @@ -960,8 +985,7 @@ static int scan_peb(struct ubi_device *ubi, struct ubi_attach_info *ai,
>  			return err;
>  		else if (!err)
>  			/* This corruption is caused by a power cut */
> -			err = add_to_list(ai, pnum, UBI_UNKNOWN,
> -					  UBI_UNKNOWN, ec, 1, &ai->erase);
> +			err = add_to_list(ai, pnum, ec, 1, &ai->erase);
>  		else
>  			/* This is an unexpected corruption */
>  			err = add_corrupted(ai, pnum, ec);
> @@ -969,23 +993,20 @@ static int scan_peb(struct ubi_device *ubi, struct ubi_attach_info *ai,
>  			return err;
>  		goto adjust_mean_ec;
>  	case UBI_IO_FF_BITFLIPS:
> -		err = add_to_list(ai, pnum, UBI_UNKNOWN, UBI_UNKNOWN,
> -				  ec, 1, &ai->erase);
> +		err = add_to_list(ai, pnum, ec, 1, &ai->erase);
>  		if (err)
>  			return err;
>  		goto adjust_mean_ec;
>  	case UBI_IO_FF:
>  		if (ec_err || bitflips)
> -			err = add_to_list(ai, pnum, UBI_UNKNOWN,
> -					  UBI_UNKNOWN, ec, 1, &ai->erase);
> +			err = add_to_list(ai, pnum, ec, 1, &ai->erase);
>  		else
> -			err = add_to_list(ai, pnum, UBI_UNKNOWN,
> -					  UBI_UNKNOWN, ec, 0, &ai->free);
> +			err = add_to_list(ai, pnum, ec, 0, &ai->free);
>  		if (err)
>  			return err;
>  		goto adjust_mean_ec;
>  	default:
> -		ubi_err(ubi, "'ubi_io_read_vid_hdr()' returned unknown code %d",
> +		ubi_err(ubi, "'ubi_io_read_vid_hdrs()' returned unknown code %d",
>  			err);
>  		return -EINVAL;
>  	}
> @@ -1006,8 +1027,7 @@ static int scan_peb(struct ubi_device *ubi, struct ubi_attach_info *ai,
>  				ubi_msg(ubi, "\"delete\" compatible internal volume %d:%d found, will remove it",
>  					vol_id, lnum);
>  			}
> -			err = add_to_list(ai, pnum, vol_id, lnum,
> -					  ec, 1, &ai->erase);
> +			err = add_to_list(ai, pnum, ec, 1, &ai->erase);
>  			if (err)
>  				return err;
>  			return 0;
> @@ -1021,8 +1041,7 @@ static int scan_peb(struct ubi_device *ubi, struct ubi_attach_info *ai,
>  		case UBI_COMPAT_PRESERVE:
>  			ubi_msg(ubi, "\"preserve\" compatible internal volume %d:%d found",
>  				vol_id, lnum);
> -			err = add_to_list(ai, pnum, vol_id, lnum,
> -					  ec, 0, &ai->alien);
> +			err = add_to_list(ai, pnum, ec, 0, &ai->alien);
>  			if (err)
>  				return err;
>  			return 0;
> @@ -1037,9 +1056,30 @@ static int scan_peb(struct ubi_device *ubi, struct ubi_attach_info *ai,
>  	if (ec_err)
>  		ubi_warn(ubi, "valid VID header but corrupted EC header at PEB %d",
>  			 pnum);
> -	err = ubi_add_to_av(ubi, ai, pnum, ec, vidh, bitflips);
> -	if (err)
> -		return err;
> +
> +	if (nvidh == 1) {
> +		err = ubi_io_read(ubi, ech, pnum,
> +				  ubi->peb_size - ubi->hdrs_min_io_size,
> +				  ubi->hdrs_min_io_size);
> +		if (!err && !ubi_check_pattern(ech, 0xff, ubi->hdrs_min_io_size))
> +			full = true;
> +	}
> +
> +	aeb = kmem_cache_zalloc(ai->apeb_slab_cache, GFP_KERNEL);
> +	if (!aeb)
> +		return -ENOMEM;
> +
> +	aeb->consolidated = nvidh > 1;
> +	aeb->refcount = nvidh;
> +	aeb->pnum = pnum;
> +	aeb->ec = ec;
> +	list_add_tail(&aeb->list, &ai->used);
> +
> +	for (i = 0; i < nvidh; i++) {
> +		err = ubi_add_to_av(ubi, ai, aeb, &vidh[i], i, bitflips, full);
> +		if (err)
> +			return err;
> +	}
>  
>  adjust_mean_ec:
>  	if (!ec_err) {
> @@ -1082,7 +1122,7 @@ static int late_analysis(struct ubi_device *ubi, struct ubi_attach_info *ai)
>  		ubi_err(ubi, "%d PEBs are corrupted and preserved",
>  			ai->corr_peb_count);
>  		pr_err("Corrupted PEBs are:");
> -		list_for_each_entry(aeb, &ai->corr, u.list)
> +		list_for_each_entry(aeb, &ai->corr, list)
>  			pr_cont(" %d", aeb->pnum);
>  		pr_cont("\n");
>  
> @@ -1136,7 +1176,7 @@ static int late_analysis(struct ubi_device *ubi, struct ubi_attach_info *ai)
>   */
>  static void destroy_av(struct ubi_attach_info *ai, struct ubi_ainf_volume *av)
>  {
> -	struct ubi_ainf_peb *aeb;
> +	struct ubi_ainf_leb *aeb;
>  	struct rb_node *this = av->root.rb_node;
>  
>  	while (this) {
> @@ -1145,16 +1185,16 @@ static void destroy_av(struct ubi_attach_info *ai, struct ubi_ainf_volume *av)
>  		else if (this->rb_right)
>  			this = this->rb_right;
>  		else {
> -			aeb = rb_entry(this, struct ubi_ainf_peb, u.rb);
> +			aeb = rb_entry(this, struct ubi_ainf_leb, rb);
>  			this = rb_parent(this);
>  			if (this) {
> -				if (this->rb_left == &aeb->u.rb)
> +				if (this->rb_left == &aeb->rb)
>  					this->rb_left = NULL;
>  				else
>  					this->rb_right = NULL;
>  			}
>  
> -			kmem_cache_free(ai->aeb_slab_cache, aeb);
> +			kmem_cache_free(ai->aleb_slab_cache, aeb);
>  		}
>  	}
>  	kfree(av);
> @@ -1170,23 +1210,6 @@ static void destroy_ai(struct ubi_attach_info *ai)
>  	struct ubi_ainf_volume *av;
>  	struct rb_node *rb;
>  
> -	list_for_each_entry_safe(aeb, aeb_tmp, &ai->alien, u.list) {
> -		list_del(&aeb->u.list);
> -		kmem_cache_free(ai->aeb_slab_cache, aeb);
> -	}
> -	list_for_each_entry_safe(aeb, aeb_tmp, &ai->erase, u.list) {
> -		list_del(&aeb->u.list);
> -		kmem_cache_free(ai->aeb_slab_cache, aeb);
> -	}
> -	list_for_each_entry_safe(aeb, aeb_tmp, &ai->corr, u.list) {
> -		list_del(&aeb->u.list);
> -		kmem_cache_free(ai->aeb_slab_cache, aeb);
> -	}
> -	list_for_each_entry_safe(aeb, aeb_tmp, &ai->free, u.list) {
> -		list_del(&aeb->u.list);
> -		kmem_cache_free(ai->aeb_slab_cache, aeb);
> -	}
> -
>  	/* Destroy the volume RB-tree */
>  	rb = ai->volumes.rb_node;
>  	while (rb) {
> @@ -1209,7 +1232,29 @@ static void destroy_ai(struct ubi_attach_info *ai)
>  		}
>  	}
>  
> -	kmem_cache_destroy(ai->aeb_slab_cache);
> +	list_for_each_entry_safe(aeb, aeb_tmp, &ai->alien, list) {
> +		list_del(&aeb->list);
> +		kmem_cache_free(ai->apeb_slab_cache, aeb);
> +	}
> +	list_for_each_entry_safe(aeb, aeb_tmp, &ai->erase, list) {
> +		list_del(&aeb->list);
> +		kmem_cache_free(ai->apeb_slab_cache, aeb);
> +	}
> +	list_for_each_entry_safe(aeb, aeb_tmp, &ai->corr, list) {
> +		list_del(&aeb->list);
> +		kmem_cache_free(ai->apeb_slab_cache, aeb);
> +	}
> +	list_for_each_entry_safe(aeb, aeb_tmp, &ai->free, list) {
> +		list_del(&aeb->list);
> +		kmem_cache_free(ai->apeb_slab_cache, aeb);
> +	}
> +	list_for_each_entry_safe(aeb, aeb_tmp, &ai->used, list) {
> +		list_del(&aeb->list);
> +		kmem_cache_free(ai->apeb_slab_cache, aeb);
> +	}
> +
> +	kmem_cache_destroy(ai->apeb_slab_cache);
> +	kmem_cache_destroy(ai->aleb_slab_cache);
>  	kfree(ai);
>  }
>  
> @@ -1227,8 +1272,6 @@ static int scan_all(struct ubi_device *ubi, struct ubi_attach_info *ai,
>  		    int start)
>  {
>  	int err, pnum;
> -	struct rb_node *rb1, *rb2;
> -	struct ubi_ainf_volume *av;
>  	struct ubi_ainf_peb *aeb;
>  
>  	err = -ENOMEM;
> @@ -1264,22 +1307,20 @@ static int scan_all(struct ubi_device *ubi, struct ubi_attach_info *ai,
>  	 * In case of unknown erase counter we use the mean erase counter
>  	 * value.
>  	 */
> -	ubi_rb_for_each_entry(rb1, av, &ai->volumes, rb) {
> -		ubi_rb_for_each_entry(rb2, aeb, &av->root, u.rb)
> -			if (aeb->ec == UBI_UNKNOWN)
> -				aeb->ec = ai->mean_ec;
> -	}
>  
> -	list_for_each_entry(aeb, &ai->free, u.list) {
> +	list_for_each_entry(aeb, &ai->erase, list)
> +		if (aeb->ec == UBI_UNKNOWN)
> +			aeb->ec = ai->mean_ec;
> +
> +	list_for_each_entry(aeb, &ai->free, list)
>  		if (aeb->ec == UBI_UNKNOWN)
>  			aeb->ec = ai->mean_ec;
> -	}
>  
> -	list_for_each_entry(aeb, &ai->corr, u.list)
> +	list_for_each_entry(aeb, &ai->corr, list)
>  		if (aeb->ec == UBI_UNKNOWN)
>  			aeb->ec = ai->mean_ec;
>  
> -	list_for_each_entry(aeb, &ai->erase, u.list)
> +	list_for_each_entry(aeb, &ai->erase, list)
>  		if (aeb->ec == UBI_UNKNOWN)
>  			aeb->ec = ai->mean_ec;
>  
> @@ -1311,16 +1352,28 @@ static struct ubi_attach_info *alloc_ai(void)
>  	INIT_LIST_HEAD(&ai->free);
>  	INIT_LIST_HEAD(&ai->erase);
>  	INIT_LIST_HEAD(&ai->alien);
> +	INIT_LIST_HEAD(&ai->used);
>  	ai->volumes = RB_ROOT;
> -	ai->aeb_slab_cache = kmem_cache_create("ubi_aeb_slab_cache",
> +	ai->apeb_slab_cache = kmem_cache_create("ubi_apeb_slab_cache",
>  					       sizeof(struct ubi_ainf_peb),
>  					       0, 0, NULL);
> -	if (!ai->aeb_slab_cache) {
> -		kfree(ai);
> -		ai = NULL;
> -	}
> +	if (!ai->apeb_slab_cache)
> +		goto err_free_ai;
> +
> +	ai->aleb_slab_cache = kmem_cache_create("ubi_aleb_slab_cache",
> +					       sizeof(struct ubi_ainf_leb),
> +					       0, 0, NULL);
> +	if (!ai->aleb_slab_cache)
> +		goto err_destroy_apeb_cache;
>  
>  	return ai;
> +
> +err_destroy_apeb_cache:
> +	kmem_cache_destroy(ai->apeb_slab_cache);
> +err_free_ai:
> +	kfree(ai);
> +
> +	return NULL;
>  }
>  
>  #ifdef CONFIG_MTD_UBI_FASTMAP
> @@ -1451,10 +1504,14 @@ int ubi_attach(struct ubi_device *ubi, int force_scan)
>  	if (err)
>  		goto out_vtbl;
>  
> -	err = ubi_eba_init(ubi, ai);
> +	err = ubi_conso_init(ubi);
>  	if (err)
>  		goto out_wl;
>  
> +	err = ubi_eba_init(ubi, ai);
> +	if (err)
> +		goto out_conso;
> +
>  #ifdef CONFIG_MTD_UBI_FASTMAP
>  	if (ubi->fm && ubi_dbg_chk_fastmap(ubi)) {
>  		struct ubi_attach_info *scan_ai;
> @@ -1482,6 +1539,8 @@ int ubi_attach(struct ubi_device *ubi, int force_scan)
>  	destroy_ai(ai);
>  	return 0;
>  
> +out_conso:
> +	ubi_conso_close(ubi);
>  out_wl:
>  	ubi_wl_close(ubi);
>  out_vtbl:
> @@ -1505,7 +1564,8 @@ static int self_check_ai(struct ubi_device *ubi, struct ubi_attach_info *ai)
>  	int pnum, err, vols_found = 0;
>  	struct rb_node *rb1, *rb2;
>  	struct ubi_ainf_volume *av;
> -	struct ubi_ainf_peb *aeb, *last_aeb;
> +	struct ubi_ainf_peb *peb;
> +	struct ubi_ainf_leb *leb, *last_leb;
>  	uint8_t *buf;
>  
>  	if (!ubi_dbg_chk_gen(ubi))
> @@ -1556,38 +1616,39 @@ static int self_check_ai(struct ubi_device *ubi, struct ubi_attach_info *ai)
>  			goto bad_av;
>  		}
>  
> -		last_aeb = NULL;
> -		ubi_rb_for_each_entry(rb2, aeb, &av->root, u.rb) {
> +		last_leb = NULL;
> +		ubi_rb_for_each_entry(rb2, leb, &av->root, rb) {
>  			cond_resched();
>  
> -			last_aeb = aeb;
> +			last_leb = leb;
>  			leb_count += 1;
> +			peb = leb->peb;
>  
> -			if (aeb->pnum < 0 || aeb->ec < 0) {
> +			if (peb->pnum < 0 || peb->ec < 0) {
>  				ubi_err(ubi, "negative values");
>  				goto bad_aeb;
>  			}
>  
> -			if (aeb->ec < ai->min_ec) {
> +			if (peb->ec < ai->min_ec) {
>  				ubi_err(ubi, "bad ai->min_ec (%d), %d found",
> -					ai->min_ec, aeb->ec);
> +					ai->min_ec, peb->ec);
>  				goto bad_aeb;
>  			}
>  
> -			if (aeb->ec > ai->max_ec) {
> +			if (peb->ec > ai->max_ec) {
>  				ubi_err(ubi, "bad ai->max_ec (%d), %d found",
> -					ai->max_ec, aeb->ec);
> +					ai->max_ec, peb->ec);
>  				goto bad_aeb;
>  			}
>  
> -			if (aeb->pnum >= ubi->peb_count) {
> +			if (peb->pnum >= ubi->peb_count) {
>  				ubi_err(ubi, "too high PEB number %d, total PEBs %d",
> -					aeb->pnum, ubi->peb_count);
> +					peb->pnum, ubi->peb_count);
>  				goto bad_aeb;
>  			}
>  
>  			if (av->vol_type == UBI_STATIC_VOLUME) {
> -				if (aeb->lnum >= av->used_ebs) {
> +				if (leb->desc.lnum >= av->used_ebs) {
>  					ubi_err(ubi, "bad lnum or used_ebs");
>  					goto bad_aeb;
>  				}
> @@ -1598,7 +1659,7 @@ static int self_check_ai(struct ubi_device *ubi, struct ubi_attach_info *ai)
>  				}
>  			}
>  
> -			if (aeb->lnum > av->highest_lnum) {
> +			if (leb->desc.lnum > av->highest_lnum) {
>  				ubi_err(ubi, "incorrect highest_lnum or lnum");
>  				goto bad_aeb;
>  			}
> @@ -1610,12 +1671,12 @@ static int self_check_ai(struct ubi_device *ubi, struct ubi_attach_info *ai)
>  			goto bad_av;
>  		}
>  
> -		if (!last_aeb)
> +		if (!last_leb)
>  			continue;
>  
> -		aeb = last_aeb;
> +		leb = last_leb;
>  
> -		if (aeb->lnum != av->highest_lnum) {
> +		if (leb->desc.lnum != av->highest_lnum) {
>  			ubi_err(ubi, "bad highest_lnum");
>  			goto bad_aeb;
>  		}
> @@ -1629,15 +1690,19 @@ static int self_check_ai(struct ubi_device *ubi, struct ubi_attach_info *ai)
>  
>  	/* Check that attaching information is correct */
>  	ubi_rb_for_each_entry(rb1, av, &ai->volumes, rb) {
> -		last_aeb = NULL;
> -		ubi_rb_for_each_entry(rb2, aeb, &av->root, u.rb) {
> +		struct ubi_vid_hdr *vh;
> +
> +		last_leb = NULL;
> +		ubi_rb_for_each_entry(rb2, leb, &av->root, rb) {
>  			int vol_type;
> +			int nvidh = ubi->lebs_per_cpeb;
>  
>  			cond_resched();
>  
> -			last_aeb = aeb;
> +			last_leb = leb;
> +			peb = leb->peb;
>  
> -			err = ubi_io_read_vid_hdr(ubi, aeb->pnum, vidh, 1);
> +			err = ubi_io_read_vid_hdrs(ubi, peb->pnum, vidh, &nvidh, 1);
>  			if (err && err != UBI_IO_BITFLIPS) {
>  				ubi_err(ubi, "VID header is not OK (%d)",
>  					err);
> @@ -1646,53 +1711,56 @@ static int self_check_ai(struct ubi_device *ubi, struct ubi_attach_info *ai)
>  				return err;
>  			}
>  
> -			vol_type = vidh->vol_type == UBI_VID_DYNAMIC ?
> +			ubi_assert(leb->peb_pos < nvidh);
> +			vh = &vidh[leb->peb_pos];
> +
> +			vol_type = vh->vol_type == UBI_VID_DYNAMIC ?
>  				   UBI_DYNAMIC_VOLUME : UBI_STATIC_VOLUME;
>  			if (av->vol_type != vol_type) {
>  				ubi_err(ubi, "bad vol_type");
>  				goto bad_vid_hdr;
>  			}
>  
> -			if (aeb->sqnum != be64_to_cpu(vidh->sqnum)) {
> -				ubi_err(ubi, "bad sqnum %llu", aeb->sqnum);
> +			if (leb->sqnum != be64_to_cpu(vh->sqnum)) {
> +				ubi_err(ubi, "bad sqnum %llu", leb->sqnum);
>  				goto bad_vid_hdr;
>  			}
>  
> -			if (av->vol_id != be32_to_cpu(vidh->vol_id)) {
> +			if (av->vol_id != be32_to_cpu(vh->vol_id)) {
>  				ubi_err(ubi, "bad vol_id %d", av->vol_id);
>  				goto bad_vid_hdr;
>  			}
>  
> -			if (av->compat != vidh->compat) {
> -				ubi_err(ubi, "bad compat %d", vidh->compat);
> +			if (av->compat != vh->compat) {
> +				ubi_err(ubi, "bad compat %d", vh->compat);
>  				goto bad_vid_hdr;
>  			}
>  
> -			if (aeb->lnum != be32_to_cpu(vidh->lnum)) {
> -				ubi_err(ubi, "bad lnum %d", aeb->lnum);
> +			if (leb->desc.lnum != be32_to_cpu(vh->lnum)) {
> +				ubi_err(ubi, "bad lnum %d", leb->desc.lnum);
>  				goto bad_vid_hdr;
>  			}
>  
> -			if (av->used_ebs != be32_to_cpu(vidh->used_ebs)) {
> +			if (av->used_ebs != be32_to_cpu(vh->used_ebs)) {
>  				ubi_err(ubi, "bad used_ebs %d", av->used_ebs);
>  				goto bad_vid_hdr;
>  			}
>  
> -			if (av->data_pad != be32_to_cpu(vidh->data_pad)) {
> +			if (av->data_pad != be32_to_cpu(vh->data_pad)) {
>  				ubi_err(ubi, "bad data_pad %d", av->data_pad);
>  				goto bad_vid_hdr;
>  			}
>  		}
>  
> -		if (!last_aeb)
> +		if (!last_leb)
>  			continue;
>  
> -		if (av->highest_lnum != be32_to_cpu(vidh->lnum)) {
> +		if (av->highest_lnum != be32_to_cpu(vh->lnum)) {
>  			ubi_err(ubi, "bad highest_lnum %d", av->highest_lnum);
>  			goto bad_vid_hdr;
>  		}
>  
> -		if (av->last_data_size != be32_to_cpu(vidh->data_size)) {
> +		if (av->last_data_size != be32_to_cpu(vh->data_size)) {
>  			ubi_err(ubi, "bad last_data_size %d",
>  				av->last_data_size);
>  			goto bad_vid_hdr;
> @@ -1716,21 +1784,20 @@ static int self_check_ai(struct ubi_device *ubi, struct ubi_attach_info *ai)
>  			buf[pnum] = 1;
>  	}
>  
> -	ubi_rb_for_each_entry(rb1, av, &ai->volumes, rb)
> -		ubi_rb_for_each_entry(rb2, aeb, &av->root, u.rb)
> -			buf[aeb->pnum] = 1;
> +	list_for_each_entry(peb, &ai->used, list)
> +		buf[peb->pnum] = 1;
>  
> -	list_for_each_entry(aeb, &ai->free, u.list)
> -		buf[aeb->pnum] = 1;
> +	list_for_each_entry(peb, &ai->free, list)
> +		buf[peb->pnum] = 1;
>  
> -	list_for_each_entry(aeb, &ai->corr, u.list)
> -		buf[aeb->pnum] = 1;
> +	list_for_each_entry(peb, &ai->corr, list)
> +		buf[peb->pnum] = 1;
>  
> -	list_for_each_entry(aeb, &ai->erase, u.list)
> -		buf[aeb->pnum] = 1;
> +	list_for_each_entry(peb, &ai->erase, list)
> +		buf[peb->pnum] = 1;
>  
> -	list_for_each_entry(aeb, &ai->alien, u.list)
> -		buf[aeb->pnum] = 1;
> +	list_for_each_entry(peb, &ai->alien, list)
> +		buf[peb->pnum] = 1;
>  
>  	err = 0;
>  	for (pnum = 0; pnum < ubi->peb_count; pnum++)
> @@ -1745,8 +1812,8 @@ static int self_check_ai(struct ubi_device *ubi, struct ubi_attach_info *ai)
>  	return 0;
>  
>  bad_aeb:
> -	ubi_err(ubi, "bad attaching information about LEB %d", aeb->lnum);
> -	ubi_dump_aeb(aeb, 0);
> +	ubi_err(ubi, "bad attaching information about LEB %d", leb->desc.lnum);
> +	ubi_dump_aeb(leb, 0);
>  	ubi_dump_av(av);
>  	goto out;
>  
> diff --git a/drivers/mtd/ubi/build.c b/drivers/mtd/ubi/build.c
> index 99e31ed..306d71f 100644
> --- a/drivers/mtd/ubi/build.c
> +++ b/drivers/mtd/ubi/build.c
> @@ -368,7 +368,9 @@ static ssize_t dev_attribute_show(struct device *dev,
>  	if (attr == &dev_eraseblock_size)
>  		ret = sprintf(buf, "%d\n", ubi->leb_size);
>  	else if (attr == &dev_avail_eraseblocks)
> -		ret = sprintf(buf, "%d\n", ubi->avail_pebs);
> +		ret = sprintf(buf, "%d\n",
> +			      ubi->avail_pebs *
> +			      ubi->lebs_per_cpeb);
>  	else if (attr == &dev_total_eraseblocks)
>  		ret = sprintf(buf, "%d\n", ubi->good_peb_count);
>  	else if (attr == &dev_volumes_count)
> @@ -653,7 +655,7 @@ static int io_init(struct ubi_device *ubi, int max_beb_per1024)
>  	ubi->consolidated_peb_size = ubi->mtd->erasesize;
>  	ubi->peb_size   = ubi->consolidated_peb_size /
>  			  mtd_pairing_groups_per_eb(ubi->mtd);
> -	ubi->lebs_per_consolidated_peb = mtd_pairing_groups_per_eb(ubi->mtd);
> +	ubi->lebs_per_cpeb = mtd_pairing_groups_per_eb(ubi->mtd);
>  	ubi->peb_count  = mtd_div_by_eb(ubi->mtd->size, ubi->mtd);
>  	ubi->flash_size = ubi->mtd->size;
>  
> @@ -797,7 +799,7 @@ static int autoresize(struct ubi_device *ubi, int vol_id)
>  {
>  	struct ubi_volume_desc desc;
>  	struct ubi_volume *vol = ubi->volumes[vol_id];
> -	int err, old_reserved_pebs = vol->reserved_pebs;
> +	int err, old_reserved_lebs = vol->reserved_lebs;
>  
>  	if (ubi->ro_mode) {
>  		ubi_warn(ubi, "skip auto-resize because of R/O mode");
> @@ -824,9 +826,11 @@ static int autoresize(struct ubi_device *ubi, int vol_id)
>  			ubi_err(ubi, "cannot clean auto-resize flag for volume %d",
>  				vol_id);
>  	} else {
> +		int avail_lebs = ubi->avail_pebs *
> +				 ubi->lebs_per_cpeb;
> +
>  		desc.vol = vol;
> -		err = ubi_resize_volume(&desc,
> -					old_reserved_pebs + ubi->avail_pebs);
> +		err = ubi_resize_volume(&desc, old_reserved_lebs + avail_lebs);
>  		if (err)
>  			ubi_err(ubi, "cannot auto-resize volume %d",
>  				vol_id);
> @@ -836,7 +840,7 @@ static int autoresize(struct ubi_device *ubi, int vol_id)
>  		return err;
>  
>  	ubi_msg(ubi, "volume %d (\"%s\") re-sized from %d to %d LEBs",
> -		vol_id, vol->name, old_reserved_pebs, vol->reserved_pebs);
> +		vol_id, vol->name, old_reserved_lebs, vol->reserved_lebs);
>  	return 0;
>  }
>  
> @@ -1026,7 +1030,7 @@ int ubi_attach_mtd_dev(struct mtd_info *mtd, int ubi_num,
>  		ubi->image_seq);
>  	ubi_msg(ubi, "available PEBs: %d, total reserved PEBs: %d, PEBs reserved for bad PEB handling: %d",
>  		ubi->avail_pebs, ubi->rsvd_pebs, ubi->beb_rsvd_pebs);
> -	ubi_msg(ubi, "LEBs per PEB: %d", ubi->lebs_per_consolidated_peb);
> +	ubi_msg(ubi, "LEBs per PEB: %d", ubi->lebs_per_cpeb);
>  
>  	/*
>  	 * The below lock makes sure we do not race with 'ubi_thread()' which
> diff --git a/drivers/mtd/ubi/cdev.c b/drivers/mtd/ubi/cdev.c
> index 833c0a82..230232f 100644
> --- a/drivers/mtd/ubi/cdev.c
> +++ b/drivers/mtd/ubi/cdev.c
> @@ -415,7 +415,7 @@ static long vol_cdev_ioctl(struct file *file, unsigned int cmd,
>  			break;
>  		}
>  
> -		rsvd_bytes = (long long)vol->reserved_pebs *
> +		rsvd_bytes = (long long)vol->reserved_lebs *
>  					ubi->leb_size-vol->data_pad;
>  		if (bytes < 0 || bytes > rsvd_bytes) {
>  			err = -EINVAL;
> @@ -454,7 +454,7 @@ static long vol_cdev_ioctl(struct file *file, unsigned int cmd,
>  
>  		/* Validate the request */
>  		err = -EINVAL;
> -		if (req.lnum < 0 || req.lnum >= vol->reserved_pebs ||
> +		if (req.lnum < 0 || req.lnum >= vol->reserved_lebs ||
>  		    req.bytes < 0 || req.bytes > vol->usable_leb_size)
>  			break;
>  
> @@ -485,7 +485,7 @@ static long vol_cdev_ioctl(struct file *file, unsigned int cmd,
>  			break;
>  		}
>  
> -		if (lnum < 0 || lnum >= vol->reserved_pebs) {
> +		if (lnum < 0 || lnum >= vol->reserved_lebs) {
>  			err = -EINVAL;
>  			break;
>  		}
> @@ -909,7 +909,7 @@ static long ubi_cdev_ioctl(struct file *file, unsigned int cmd,
>  	/* Re-size volume command */
>  	case UBI_IOCRSVOL:
>  	{
> -		int pebs;
> +		int lebs;
>  		struct ubi_rsvol_req req;
>  
>  		dbg_gen("re-size volume");
> @@ -929,11 +929,11 @@ static long ubi_cdev_ioctl(struct file *file, unsigned int cmd,
>  			break;
>  		}
>  
> -		pebs = div_u64(req.bytes + desc->vol->usable_leb_size - 1,
> +		lebs = div_u64(req.bytes + desc->vol->usable_leb_size - 1,
>  			       desc->vol->usable_leb_size);
>  
>  		mutex_lock(&ubi->device_mutex);
> -		err = ubi_resize_volume(desc, pebs);
> +		err = ubi_resize_volume(desc, lebs);
>  		mutex_unlock(&ubi->device_mutex);
>  		ubi_close_volume(desc);
>  		break;
> diff --git a/drivers/mtd/ubi/consolidate.c b/drivers/mtd/ubi/consolidate.c
> new file mode 100644
> index 0000000..de1479d
> --- /dev/null
> +++ b/drivers/mtd/ubi/consolidate.c
> @@ -0,0 +1,499 @@
> +#include <linux/slab.h>
> +#include <linux/crc32.h>
> +#include "ubi.h"
> +
> +static void consolidation_unlock(struct ubi_device *ubi,
> +				 struct ubi_leb_desc *clebs)
> +{
> +	int i;
> +
> +	for (i = 0; i < ubi->lebs_per_cpeb; i++)
> +		ubi_eba_leb_write_unlock(ubi, clebs[i].vol_id, clebs[i].lnum);
> +}
> +
> +static int find_consolidable_lebs(struct ubi_device *ubi,
> +				  struct ubi_leb_desc *clebs,
> +				  struct ubi_volume **vols)
> +{
> +	struct ubi_full_leb *fleb;
> +	LIST_HEAD(found);
> +	int i, err = 0;
> +
> +	spin_lock(&ubi->full_lock);
> +	if (ubi->full_count < ubi->lebs_per_cpeb)
> +		err = -EAGAIN;
> +	spin_unlock(&ubi->full_lock);
> +	if (err)
> +		return err;
> +
> +	for (i = 0; i < ubi->lebs_per_cpeb;) {
> +		spin_lock(&ubi->full_lock);
> +		fleb = list_first_entry_or_null(&ubi->full,
> +						struct ubi_full_leb, node);
> +		spin_unlock(&ubi->full_lock);
> +
> +		if (!fleb) {
> +			err = -EAGAIN;
> +			goto err;
> +		} else {
> +			list_del_init(&fleb->node);
> +			list_add_tail(&fleb->node, &found);
> +			ubi->full_count--;
> +		}
> +
> +		clebs[i] = fleb->desc;
> +
> +		err = ubi_eba_leb_write_lock_nested(ubi, clebs[i].vol_id,
> +						    clebs[i].lnum, i);
> +		if (err) {
> +			spin_lock(&ubi->full_lock);
> +			list_del(&fleb->node);
> +			list_add_tail(&fleb->node, &ubi->full);
> +			ubi->full_count++;
> +			spin_unlock(&ubi->full_lock);
> +			goto err;
> +		}
> +
> +		spin_lock(&ubi->volumes_lock);
> +		vols[i] = ubi->volumes[vol_id2idx(ubi, clebs[i].vol_id)];
> +		spin_unlock(&ubi->volumes_lock);
> +		/* volume vanished under us */
> +		//TODO clarify/document when/why this can happen
> +		if (!vols[i]) {
> +			ubi_assert(0);
> +			ubi_eba_leb_write_unlock(ubi, clebs[i].vol_id, clebs[i].lnum);
> +			spin_lock(&ubi->full_lock);
> +			list_del_init(&fleb->node);
> +			kfree(fleb);
> +			spin_unlock(&ubi->full_lock);
> +			continue;
> +		}
> +
> +		i++;
> +	}
> +
> +	while(!list_empty(&found)) {
> +		fleb = list_first_entry(&found, struct ubi_full_leb, node);
> +		list_del(&fleb->node);
> +		kfree(fleb);
> +	}
> +
> +	if (i < ubi->lebs_per_cpeb - 1) {
> +		return -EAGAIN;
> +	}
> +
> +	return 0;
> +
> +err:
> +	while(!list_empty(&found)) {
> +		spin_lock(&ubi->full_lock);
> +		fleb = list_first_entry(&found, struct ubi_full_leb, node);
> +		list_del(&fleb->node);
> +		list_add_tail(&fleb->node, &ubi->full);
> +		ubi->full_count++;
> +		spin_unlock(&ubi->full_lock);
> +		ubi_eba_leb_write_unlock(ubi, fleb->desc.vol_id, fleb->desc.lnum);
> +	}
> +
> +	return err;
> +}
> +
> +static int consolidate_lebs(struct ubi_device *ubi)
> +{
> +	int i, pnum, offset = ubi->leb_start, err = 0;
> +	struct ubi_vid_hdr *vid_hdrs;
> +	struct ubi_leb_desc *clebs = NULL, *new_clebs = NULL;
> +	struct ubi_volume **vols = NULL;
> +	int *opnums = NULL;
> +
> +	if (!ubi_conso_consolidation_needed(ubi))
> +		return 0;
> +
> +	vols = kzalloc(sizeof(*vols) * ubi->lebs_per_cpeb, GFP_KERNEL);
> +	if (!vols)
> +		return -ENOMEM;
> +
> +	opnums = kzalloc(sizeof(*opnums) * ubi->lebs_per_cpeb, GFP_KERNEL);
> +	if (!opnums) {
> +		err = -ENOMEM;
> +		goto err_free_mem;
> +	}
> +
> +	clebs = kzalloc(sizeof(*clebs) * ubi->lebs_per_cpeb, GFP_KERNEL);
> +	if (!clebs) {
> +		err = -ENOMEM;
> +		goto err_free_mem;
> +	}
> +
> +	new_clebs = kzalloc(sizeof(*clebs) * ubi->lebs_per_cpeb, GFP_KERNEL);
> +	if (!new_clebs) {
> +		err = -ENOMEM;
> +		goto err_free_mem;
> +	}
> +
> +	err = find_consolidable_lebs(ubi, clebs, vols);
> +	if (err)
> +		goto err_free_mem;
> +
> +	memcpy(new_clebs, clebs, sizeof(*clebs) * ubi->lebs_per_cpeb);
> +
> +	mutex_lock(&ubi->buf_mutex);
> +
> +	pnum = ubi_wl_get_peb(ubi, true);
> +	if (pnum < 0) {
> +		err = pnum;
> +		//TODO cleanup exit path
> +		mutex_unlock(&ubi->buf_mutex);
> +		up_read(&ubi->fm_eba_sem);
> +		goto err_unlock_lebs;
> +	}
> +
> +	memset(ubi->peb_buf, 0, ubi->peb_size);
> +	vid_hdrs = ubi->peb_buf + ubi->vid_hdr_aloffset + ubi->vid_hdr_shift;
> +
> +	for (i = 0; i < ubi->lebs_per_cpeb; i++) {
> +		int vol_id = clebs[i].vol_id, lnum = clebs[i].lnum, lpos = clebs[i].lpos;
> +		void *buf = ubi->peb_buf + offset;
> +		struct ubi_volume *vol = vols[i];
> +		int spnum;
> +		int data_size;
> +		u32 crc;
> +		bool raw;
> +
> +		spnum = vol->eba_tbl[lnum];
> +
> +		/* we raced against leb unmap */
> +		if (spnum == UBI_LEB_UNMAPPED) {
> +			//TODO: should be fixed now and no longer trigger.
> +			ubi_assert(0);
> +			err = 0;
> +			goto err_unlock_fm_eba;
> +		}
> +
> +		if (ubi->consolidated[spnum]) {
> +			ubi_assert(ubi_conso_invalidate_leb(ubi, spnum, vol_id, lnum) == true);
> +			raw = true;
> +		} else {
> +			ubi_assert(!lpos);
> +			raw = false;
> +		}
> +
> +		ubi_assert(offset + ubi->leb_size < ubi->consolidated_peb_size);
> +
> +		if (!raw)
> +			err = ubi_io_read(ubi, buf, spnum, ubi->leb_start, ubi->leb_size);
> +		else
> +			err = ubi_io_raw_read(ubi, buf, spnum, ubi->leb_start + (lpos * ubi->leb_size), ubi->leb_size);
> +
> +		if (err && err != UBI_IO_BITFLIPS)
> +			goto err_unlock_fm_eba;
> +
> +		if (vol->vol_type == UBI_DYNAMIC_VOLUME) {
> +			data_size = ubi->leb_size - vol->data_pad;
> +			vid_hdrs[i].vol_type = UBI_VID_DYNAMIC;
> +		} else {
> +			int nvidh = ubi->lebs_per_cpeb;
> +			struct ubi_vid_hdr *vh;
> +
> +			vh = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
> +			if (!vh) {
> +				err = -ENOMEM;
> +				goto err_unlock_fm_eba;
> +			}
> +
> +			err = ubi_io_read_vid_hdrs(ubi, spnum, vh, &nvidh, 0);
> +			if (err && err != UBI_IO_BITFLIPS) {
> +				ubi_free_vid_hdr(ubi, vh);
> +				goto err_unlock_fm_eba;
> +			}
> +
> +			ubi_free_vid_hdr(ubi, vh);
> +
> +			data_size = be32_to_cpu(vh[lpos].data_size);
> +			vid_hdrs[i].vol_type = UBI_VID_STATIC;
> +			vid_hdrs[i].used_ebs = cpu_to_be32(vol->used_ebs);
> +		}
> +
> +		vid_hdrs[i].data_pad = cpu_to_be32(vol->data_pad);
> +		vid_hdrs[i].sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
> +		vid_hdrs[i].vol_id = cpu_to_be32(vol_id);
> +		vid_hdrs[i].lnum = cpu_to_be32(lnum);
> +		vid_hdrs[i].compat = ubi_get_compat(ubi, vol_id);
> +		vid_hdrs[i].data_size = cpu_to_be32(data_size);
> +		vid_hdrs[i].copy_flag = 1;
> +		crc = crc32(UBI_CRC32_INIT, buf, data_size);
> +		vid_hdrs[i].data_crc = cpu_to_be32(crc);
> +		offset += ubi->leb_size;
> +
> +		new_clebs[i].lpos = i;
> +	}
> +
> +	/*
> +	 * Pad remaining pages with zeros to prevent problem on some MLC chip
> +	 * that expect the whole block to be programmed in order to work
> +	 * reliably (some Hynix chips are impacted).
> +	 */
> +	memset(ubi->peb_buf + offset, 0, ubi->consolidated_peb_size - offset);
> +
> +	err = ubi_io_write_vid_hdrs(ubi, pnum, vid_hdrs, ubi->lebs_per_cpeb);
> +	if (err) {
> +		ubi_warn(ubi, "failed to write VID headers to PEB %d",
> +			 pnum);
> +		goto err_unlock_lebs;
> +	}
> +
> +	err = ubi_io_raw_write(ubi, ubi->peb_buf + ubi->leb_start,
> +			       pnum, ubi->leb_start,
> +			       ubi->consolidated_peb_size - ubi->leb_start);
> +	if (err) {
> +		ubi_warn(ubi, "failed to write %d bytes of data to PEB %d",
> +			 ubi->consolidated_peb_size - ubi->leb_start, pnum);
> +		goto err_unlock_fm_eba;
> +	}
> +
> +	for (i = 0; i < ubi->lebs_per_cpeb; i++) {
> +		struct ubi_volume *vol = vols[i];
> +		int lnum = clebs[i].lnum;
> +
> +		opnums[i] = vol->eba_tbl[lnum];
> +
> +		vol->eba_tbl[lnum] = pnum;
> +	}
> +	ubi->consolidated[pnum] = new_clebs;
> +
> +	up_read(&ubi->fm_eba_sem);
> +	mutex_unlock(&ubi->buf_mutex);
> +	consolidation_unlock(ubi, clebs);
> +
> +	for (i = 0; i < ubi->lebs_per_cpeb; i++) {
> +		//TODO set torture if needed
> +		ubi_wl_put_peb(ubi, opnums[i], 0);
> +	}
> +
> +	kfree(clebs);
> +	kfree(opnums);
> +	kfree(vols);
> +
> +	return 0;
> +
> +err_unlock_fm_eba:
> +	mutex_unlock(&ubi->buf_mutex);
> +	up_read(&ubi->fm_eba_sem);
> +
> +	for (i = 0; i < ubi->lebs_per_cpeb; i++)
> +		ubi_coso_add_full_leb(ubi, clebs[i].vol_id, clebs[i].lnum, clebs[i].lpos);
> +
> +	ubi_wl_put_peb(ubi, pnum, 0);
> +err_unlock_lebs:
> +	consolidation_unlock(ubi, clebs);
> +err_free_mem:
> +	kfree(new_clebs);
> +	kfree(clebs);
> +	kfree(opnums);
> +	kfree(vols);
> +
> +	return err;
> +}
> +
> +static int consolidation_worker(struct ubi_device *ubi,
> +				struct ubi_work *wrk,
> +				int shutdown)
> +{
> +	int ret;
> +
> +	if (shutdown)
> +		return 0;
> +
> +	ret = consolidate_lebs(ubi);
> +	if (ret == -EAGAIN)
> +		ret = 0;
> +
> +	ubi->conso_scheduled = 0;
> +	smp_wmb();
> +
> +	if (ubi_conso_consolidation_needed(ubi))
> +		ubi_conso_schedule(ubi);
> +
> +	return ret;
> +}
> +
> +static bool consolidation_possible(struct ubi_device *ubi)
> +{
> +	if (ubi->lebs_per_cpeb < 2)
> +		return false;
> +
> +	if (ubi->full_count < ubi->lebs_per_cpeb)
> +		return false;
> +
> +	return true;
> +}
> +
> +bool ubi_conso_consolidation_needed(struct ubi_device *ubi)
> +{
> +	if (!consolidation_possible(ubi))
> +		return false;
> +
> +	return ubi->free_count - ubi->beb_rsvd_pebs <=
> +	       ubi->consolidation_threshold;
> +}
> +
> +void ubi_conso_schedule(struct ubi_device *ubi)
> +{
> +	struct ubi_work *wrk;
> +
> +	if (ubi->conso_scheduled)
> +		return;
> +
> +	wrk = ubi_alloc_work(ubi);
> +	if (wrk) {
> +		ubi->conso_scheduled = 1;
> +		smp_wmb();
> +
> +		wrk->func = &consolidation_worker;
> +		INIT_LIST_HEAD(&wrk->list);
> +		ubi_schedule_work(ubi, wrk);
> +	} else
> +		BUG();
> +}
> +
> +void ubi_eba_consolidate(struct ubi_device *ubi)
> +{
> +	if (consolidation_possible(ubi) && ubi->consolidation_pnum >= 0)
> +		ubi_conso_schedule(ubi);
> +}
> +
> +void ubi_conso_remove_full_leb(struct ubi_device *ubi, int vol_id, int lnum)
> +{
> +	struct ubi_full_leb *fleb;
> +
> +	spin_lock(&ubi->full_lock);
> +	list_for_each_entry(fleb, &ubi->full, node) {
> +		if (fleb->desc.lnum == lnum && fleb->desc.vol_id == vol_id) {
> +			ubi->full_count--;
> +			list_del(&fleb->node);
> +			kfree(fleb);
> +			break;
> +		}
> +	}
> +	spin_unlock(&ubi->full_lock);
> +}
> +
> +struct ubi_leb_desc *
> +ubi_conso_get_consolidated(struct ubi_device *ubi, int pnum)
> +{
> +	if (ubi->consolidated)
> +		return ubi->consolidated[pnum];
> +
> +	return NULL;
> +}
> +
> +int ubi_coso_add_full_leb(struct ubi_device *ubi, int vol_id, int lnum, int lpos)
> +{
> +	struct ubi_full_leb *fleb;
> +
> +	/*
> +	 * We don't track full LEBs if we don't need to (which is the case
> +	 * when UBI does not need or does not support LEB consolidation).
> +	 */
> +	if (!ubi->consolidated)
> +		return 0;
> +
> +	fleb = kzalloc(sizeof(*fleb), GFP_KERNEL);
> +	if (!fleb)
> +		return -ENOMEM;
> +
> +	fleb->desc.vol_id = vol_id;
> +	fleb->desc.lnum = lnum;
> +	fleb->desc.lpos = lpos;
> +
> +	spin_lock(&ubi->full_lock);
> +	list_add_tail(&fleb->node, &ubi->full);
> +	ubi->full_count++;
> +	spin_unlock(&ubi->full_lock);
> +
> +	return 0;
> +}
> +
> +bool ubi_conso_invalidate_leb(struct ubi_device *ubi, int pnum,
> +				   int vol_id, int lnum)
> +{
> +	struct ubi_leb_desc *clebs = NULL;
> +
> +	if (!ubi->consolidated)
> +		return true;
> +
> +	clebs = ubi->consolidated[pnum];
> +	if (!clebs)
> +		return true;
> +
> +	//TODO: make this generic again
> +	BUG_ON(ubi->lebs_per_cpeb > 2);
> +
> +	if (clebs[0].lnum == lnum && clebs[0].vol_id == vol_id) {
> +		clebs[0].lnum = -1;
> +		clebs[0].vol_id = -1;
> +
> +		if (clebs[1].lnum > -1 && clebs[1].vol_id > -1) {
> +			ubi_coso_add_full_leb(ubi, clebs[1].vol_id, clebs[1].lnum, clebs[1].lpos);
> +
> +			return false;
> +		}
> +	} else if (clebs[1].lnum == lnum && clebs[1].vol_id == vol_id) {
> +		clebs[1].lnum = -1;
> +		clebs[1].vol_id = -1;
> +
> +		if (clebs[0].lnum > -1 && clebs[0].vol_id > -1) {
> +			ubi_coso_add_full_leb(ubi, clebs[0].vol_id, clebs[0].lnum, clebs[0].lpos);
> +
> +			return false;
> +		}
> +	} else
> +		ubi_assert(0);
> +
> +	ubi->consolidated[pnum] = NULL;
> +	kfree(clebs);
> +
> +	return true;
> +}
> +
> +int ubi_conso_init(struct ubi_device *ubi)
> +{
> +	spin_lock_init(&ubi->full_lock);
> +	INIT_LIST_HEAD(&ubi->full);
> +	ubi->full_count = 0;
> +	ubi->consolidation_threshold = (ubi->avail_pebs + ubi->rsvd_pebs) / 3;
> +
> +	if (ubi->consolidation_threshold < ubi->lebs_per_cpeb)
> +		ubi->consolidation_threshold = ubi->lebs_per_cpeb;
> +
> +	if (ubi->lebs_per_cpeb == 1)
> +		return 0;
> +
> +	if (ubi->avail_pebs < UBI_CONSO_RESERVED_PEBS) {
> +		ubi_err(ubi, "no enough physical eraseblocks (%d, need %d)",
> +			ubi->avail_pebs, UBI_CONSO_RESERVED_PEBS);
> +		if (ubi->corr_peb_count)
> +			ubi_err(ubi, "%d PEBs are corrupted and not used",
> +				ubi->corr_peb_count);
> +		return -ENOSPC;
> +	}
> +
> +	ubi->avail_pebs -= UBI_CONSO_RESERVED_PEBS;
> +	ubi->rsvd_pebs += UBI_CONSO_RESERVED_PEBS;
> +
> +	return 0;
> +}
> +
> +void ubi_conso_close(struct ubi_device *ubi)
> +{
> +	struct ubi_full_leb *fleb;
> +
> +	while(!list_empty(&ubi->full)) {
> +		fleb = list_first_entry(&ubi->full, struct ubi_full_leb, node);
> +		list_del(&fleb->node);
> +		kfree(fleb);
> +		ubi->full_count--;
> +	}
> +
> +	ubi_assert(ubi->full_count == 0);
> +}
> diff --git a/drivers/mtd/ubi/debug.c b/drivers/mtd/ubi/debug.c
> index ed23009..6178fa1 100644
> --- a/drivers/mtd/ubi/debug.c
> +++ b/drivers/mtd/ubi/debug.c
> @@ -36,7 +36,7 @@ void ubi_dump_flash(struct ubi_device *ubi, int pnum, int offset, int len)
>  	int err;
>  	size_t read;
>  	void *buf;
> -	loff_t addr = (loff_t)pnum * ubi->peb_size + offset;
> +	loff_t addr = (loff_t)pnum * ubi->consolidated_peb_size + offset;
>  
>  	buf = vmalloc(len);
>  	if (!buf)
> @@ -108,7 +108,7 @@ void ubi_dump_vol_info(const struct ubi_volume *vol)
>  {
>  	pr_err("Volume information dump:\n");
>  	pr_err("\tvol_id          %d\n", vol->vol_id);
> -	pr_err("\treserved_pebs   %d\n", vol->reserved_pebs);
> +	pr_err("\treserved_lebs   %d\n", vol->reserved_lebs);
>  	pr_err("\talignment       %d\n", vol->alignment);
>  	pr_err("\tdata_pad        %d\n", vol->data_pad);
>  	pr_err("\tvol_type        %d\n", vol->vol_type);
> @@ -140,7 +140,7 @@ void ubi_dump_vtbl_record(const struct ubi_vtbl_record *r, int idx)
>  	int name_len = be16_to_cpu(r->name_len);
>  
>  	pr_err("Volume table record %d dump:\n", idx);
> -	pr_err("\treserved_pebs   %d\n", be32_to_cpu(r->reserved_pebs));
> +	pr_err("\treserved_pebs   %d\n", be32_to_cpu(r->reserved_lebs));
>  	pr_err("\talignment       %d\n", be32_to_cpu(r->alignment));
>  	pr_err("\tdata_pad        %d\n", be32_to_cpu(r->data_pad));
>  	pr_err("\tvol_type        %d\n", (int)r->vol_type);
> @@ -185,14 +185,14 @@ void ubi_dump_av(const struct ubi_ainf_volume *av)
>   * @aeb: the object to dump
>   * @type: object type: 0 - not corrupted, 1 - corrupted
>   */
> -void ubi_dump_aeb(const struct ubi_ainf_peb *aeb, int type)
> +void ubi_dump_aeb(const struct ubi_ainf_leb *aeb, int type)
>  {
>  	pr_err("eraseblock attaching information dump:\n");
> -	pr_err("\tec       %d\n", aeb->ec);
> -	pr_err("\tpnum     %d\n", aeb->pnum);
> +	pr_err("\tec       %d\n", aeb->peb->ec);
> +	pr_err("\tpnum     %d\n", aeb->peb->pnum);
>  	if (type == 0) {
> -		pr_err("\tlnum     %d\n", aeb->lnum);
> -		pr_err("\tscrub    %d\n", aeb->scrub);
> +		pr_err("\tlnum     %d\n", aeb->desc.lnum);
> +		pr_err("\tscrub    %d\n", aeb->peb->scrub);
>  		pr_err("\tsqnum    %llu\n", aeb->sqnum);
>  	}
>  }
> diff --git a/drivers/mtd/ubi/debug.h b/drivers/mtd/ubi/debug.h
> index 47c447d..c3ed0d5 100644
> --- a/drivers/mtd/ubi/debug.h
> +++ b/drivers/mtd/ubi/debug.h
> @@ -56,7 +56,7 @@ void ubi_dump_vid_hdr(const struct ubi_vid_hdr *vid_hdr);
>  void ubi_dump_vol_info(const struct ubi_volume *vol);
>  void ubi_dump_vtbl_record(const struct ubi_vtbl_record *r, int idx);
>  void ubi_dump_av(const struct ubi_ainf_volume *av);
> -void ubi_dump_aeb(const struct ubi_ainf_peb *aeb, int type);
> +void ubi_dump_aeb(const struct ubi_ainf_leb *aeb, int type);
>  void ubi_dump_mkvol_req(const struct ubi_mkvol_req *req);
>  int ubi_self_check_all_ff(struct ubi_device *ubi, int pnum, int offset,
>  			  int len);
> diff --git a/drivers/mtd/ubi/eba.c b/drivers/mtd/ubi/eba.c
> index 229be7c..953091e 100644
> --- a/drivers/mtd/ubi/eba.c
> +++ b/drivers/mtd/ubi/eba.c
> @@ -69,21 +69,6 @@ unsigned long long ubi_next_sqnum(struct ubi_device *ubi)
>  }
>  
>  /**
> - * ubi_get_compat - get compatibility flags of a volume.
> - * @ubi: UBI device description object
> - * @vol_id: volume ID
> - *
> - * This function returns compatibility flags for an internal volume. User
> - * volumes have no compatibility flags, so %0 is returned.
> - */
> -static int ubi_get_compat(const struct ubi_device *ubi, int vol_id)
> -{
> -	if (vol_id == UBI_LAYOUT_VOLUME_ID)
> -		return UBI_LAYOUT_VOLUME_COMPAT;
> -	return 0;
> -}
> -
> -/**
>   * ltree_lookup - look up the lock tree.
>   * @ubi: UBI device description object
>   * @vol_id: volume ID
> @@ -256,6 +241,31 @@ static int leb_write_lock(struct ubi_device *ubi, int vol_id, int lnum)
>  }
>  
>  /**
> + * ubi_eba_leb_write_lock_nested - lock logical eraseblock for writing, allow nesting.
> + * @ubi: UBI device description object
> + * @vol_id: volume ID
> + * @lnum: logical eraseblock number
> + * @level: nesting level
> + *
> + * This function locks a logical eraseblock for consolidation.
> + * Returns zero in case of success and a negative error code in case
> + * of failure.
> + */
> +int ubi_eba_leb_write_lock_nested(struct ubi_device *ubi, int vol_id, int lnum,
> +				  int level)
> +{
> +	struct ubi_ltree_entry *le;
> +
> +	le = ltree_add_entry(ubi, vol_id, lnum);
> +	if (IS_ERR(le))
> +		return PTR_ERR(le);
> +
> +	down_write_nested(&le->mutex, level);
> +
> +	return 0;
> +}
> +
> +/**
>   * leb_write_lock - lock logical eraseblock for writing.
>   * @ubi: UBI device description object
>   * @vol_id: volume ID
> @@ -295,7 +305,7 @@ static int leb_write_trylock(struct ubi_device *ubi, int vol_id, int lnum)
>   * @vol_id: volume ID
>   * @lnum: logical eraseblock number
>   */
> -static void leb_write_unlock(struct ubi_device *ubi, int vol_id, int lnum)
> +void ubi_eba_leb_write_unlock(struct ubi_device *ubi, int vol_id, int lnum)
>  {
>  	struct ubi_ltree_entry *le;
>  
> @@ -311,6 +321,7 @@ static void leb_write_unlock(struct ubi_device *ubi, int vol_id, int lnum)
>  	spin_unlock(&ubi->ltree_lock);
>  }
>  
> +
>  /**
>   * ubi_eba_unmap_leb - un-map logical eraseblock.
>   * @ubi: UBI device description object
> @@ -325,6 +336,7 @@ int ubi_eba_unmap_leb(struct ubi_device *ubi, struct ubi_volume *vol,
>  		      int lnum)
>  {
>  	int err, pnum, vol_id = vol->vol_id;
> +	bool release_peb = false;
>  
>  	if (ubi->ro_mode)
>  		return -EROFS;
> @@ -342,11 +354,15 @@ int ubi_eba_unmap_leb(struct ubi_device *ubi, struct ubi_volume *vol,
>  
>  	down_read(&ubi->fm_eba_sem);
>  	vol->eba_tbl[lnum] = UBI_LEB_UNMAPPED;
> +	release_peb = ubi_conso_invalidate_leb(ubi, pnum, vol_id, lnum);
>  	up_read(&ubi->fm_eba_sem);
> -	err = ubi_wl_put_peb(ubi, pnum, 0);
> +	ubi_conso_remove_full_leb(ubi, vol_id, lnum);
>  
>  out_unlock:
> -	leb_write_unlock(ubi, vol_id, lnum);
> +	ubi_eba_leb_write_unlock(ubi, vol_id, lnum);
> +	if (release_peb)
> +		err = ubi_wl_put_peb(ubi, pnum, 0);
> +
>  	return err;
>  }
>  
> @@ -372,9 +388,10 @@ out_unlock:
>  int ubi_eba_read_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
>  		     void *buf, int offset, int len, int check)
>  {
> -	int err, pnum, scrub = 0, vol_id = vol->vol_id;
> +	int err, pnum, scrub = 0, vol_id = vol->vol_id, loffs = 0, lpos = 0;
>  	struct ubi_vid_hdr *vid_hdr;
>  	uint32_t uninitialized_var(crc);
> +	struct ubi_leb_desc *clebs;
>  
>  	err = leb_read_lock(ubi, vol_id, lnum);
>  	if (err)
> @@ -401,15 +418,31 @@ int ubi_eba_read_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
>  	if (vol->vol_type == UBI_DYNAMIC_VOLUME)
>  		check = 0;
>  
> +	clebs = ubi_conso_get_consolidated(ubi, pnum);
> +	if (clebs) {
> +		for (; lpos < ubi->lebs_per_cpeb; lpos++) {
> +			if (clebs[lpos].vol_id == vol->vol_id &&
> +			    clebs[lpos].lnum == lnum)
> +				break;
> +		}
> +
> +		if (lpos == ubi->lebs_per_cpeb)
> +			return -EINVAL;
> +
> +		loffs = ubi->leb_start + (lpos * ubi->leb_size);
> +	}
> +
>  retry:
>  	if (check) {
> +		int nvidh = ubi->lebs_per_cpeb;
> +
>  		vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
>  		if (!vid_hdr) {
>  			err = -ENOMEM;
>  			goto out_unlock;
>  		}
>  
> -		err = ubi_io_read_vid_hdr(ubi, pnum, vid_hdr, 1);
> +		err = ubi_io_read_vid_hdrs(ubi, pnum, vid_hdr, &nvidh, 1);
>  		if (err && err != UBI_IO_BITFLIPS) {
>  			if (err > 0) {
>  				/*
> @@ -451,14 +484,18 @@ retry:
>  		} else if (err == UBI_IO_BITFLIPS)
>  			scrub = 1;
>  
> -		ubi_assert(lnum < be32_to_cpu(vid_hdr->used_ebs));
> -		ubi_assert(len == be32_to_cpu(vid_hdr->data_size));
> +		ubi_assert(lnum < be32_to_cpu(vid_hdr[lpos].used_ebs));
> +		ubi_assert(len == be32_to_cpu(vid_hdr[lpos].data_size));
>  
> -		crc = be32_to_cpu(vid_hdr->data_crc);
> +		crc = be32_to_cpu(vid_hdr[lpos].data_crc);
>  		ubi_free_vid_hdr(ubi, vid_hdr);
>  	}
>  
> -	err = ubi_io_read_data(ubi, buf, pnum, offset, len);
> +	if (!clebs)
> +		err = ubi_io_read_data(ubi, buf, pnum, offset, len);
> +	else
> +		err = ubi_io_raw_read(ubi, buf, pnum, offset + loffs, len);
> +
>  	if (err) {
>  		if (err == UBI_IO_BITFLIPS)
>  			scrub = 1;
> @@ -581,7 +618,7 @@ static int recover_peb(struct ubi_device *ubi, int pnum, int vol_id, int lnum,
>  		return -ENOMEM;
>  
>  retry:
> -	new_pnum = ubi_wl_get_peb(ubi);
> +	new_pnum = ubi_wl_get_peb(ubi, false);
>  	if (new_pnum < 0) {
>  		ubi_free_vid_hdr(ubi, vid_hdr);
>  		up_read(&ubi->fm_eba_sem);
> @@ -633,7 +670,7 @@ retry:
>  
>  	vol->eba_tbl[lnum] = new_pnum;
>  	up_read(&ubi->fm_eba_sem);
> -	ubi_wl_put_peb(ubi, vol_id, 1);
> +	ubi_wl_put_peb(ubi, pnum, 1);
>  
>  	ubi_msg(ubi, "data was successfully recovered");
>  	return 0;
> @@ -679,16 +716,24 @@ int ubi_eba_write_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
>  {
>  	int err, pnum, tries = 0, vol_id = vol->vol_id;
>  	struct ubi_vid_hdr *vid_hdr;
> +	struct ubi_leb_desc *clebs;
> +	bool full;
>  
>  	if (ubi->ro_mode)
>  		return -EROFS;
>  
> +	full = (offset + len > ubi->leb_size - ubi->min_io_size);
> +
>  	err = leb_write_lock(ubi, vol_id, lnum);
>  	if (err)
>  		return err;
>  
>  	pnum = vol->eba_tbl[lnum];
>  	if (pnum >= 0) {
> +		clebs = ubi_conso_get_consolidated(ubi, pnum);
> +		/* TODO: handle the write on consolidated PEB case */
> +		BUG_ON(clebs);
> +
>  		dbg_eba("write %d bytes at offset %d of LEB %d:%d, PEB %d",
>  			len, offset, vol_id, lnum, pnum);
>  
> @@ -701,7 +746,22 @@ int ubi_eba_write_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
>  			if (err)
>  				ubi_ro_mode(ubi);
>  		}
> -		leb_write_unlock(ubi, vol_id, lnum);
> +
> +		if (full) {
> +			int ret;
> +
> +			ret = ubi_coso_add_full_leb(ubi, vol_id, lnum, 0);
> +			if (ret)
> +				ubi_warn(ubi,
> +					 "failed to add LEB %d:%d to the full LEB list",
> +					 vol_id, lnum);
> +		}
> +
> +		ubi_eba_leb_write_unlock(ubi, vol_id, lnum);
> +
> +		if (full && !err && ubi_conso_consolidation_needed(ubi))
> +			ubi_conso_schedule(ubi);
> +
>  		return err;
>  	}
>  
> @@ -711,7 +771,7 @@ int ubi_eba_write_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
>  	 */
>  	vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
>  	if (!vid_hdr) {
> -		leb_write_unlock(ubi, vol_id, lnum);
> +		ubi_eba_leb_write_unlock(ubi, vol_id, lnum);
>  		return -ENOMEM;
>  	}
>  
> @@ -723,10 +783,10 @@ int ubi_eba_write_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
>  	vid_hdr->data_pad = cpu_to_be32(vol->data_pad);
>  
>  retry:
> -	pnum = ubi_wl_get_peb(ubi);
> +	pnum = ubi_wl_get_peb(ubi, false);
>  	if (pnum < 0) {
>  		ubi_free_vid_hdr(ubi, vid_hdr);
> -		leb_write_unlock(ubi, vol_id, lnum);
> +		ubi_eba_leb_write_unlock(ubi, vol_id, lnum);
>  		up_read(&ubi->fm_eba_sem);
>  		return pnum;
>  	}
> @@ -755,14 +815,26 @@ retry:
>  	vol->eba_tbl[lnum] = pnum;
>  	up_read(&ubi->fm_eba_sem);
>  
> -	leb_write_unlock(ubi, vol_id, lnum);
> +	if (full) {
> +		err = ubi_coso_add_full_leb(ubi, vol_id, lnum, 0);
> +		if (err)
> +			ubi_warn(ubi,
> +				 "failed to add LEB %d:%d to the full LEB list",
> +				 vol_id, lnum);
> +	}
> +
> +	ubi_eba_leb_write_unlock(ubi, vol_id, lnum);
>  	ubi_free_vid_hdr(ubi, vid_hdr);
> +
> +	if (full && ubi_conso_consolidation_needed(ubi))
> +		ubi_conso_schedule(ubi);
> +
>  	return 0;
>  
>  write_error:
>  	if (err != -EIO || !ubi->bad_allowed) {
>  		ubi_ro_mode(ubi);
> -		leb_write_unlock(ubi, vol_id, lnum);
> +		ubi_eba_leb_write_unlock(ubi, vol_id, lnum);
>  		ubi_free_vid_hdr(ubi, vid_hdr);
>  		return err;
>  	}
> @@ -775,7 +847,7 @@ write_error:
>  	err = ubi_wl_put_peb(ubi, pnum, 1);
>  	if (err || ++tries > UBI_IO_RETRIES) {
>  		ubi_ro_mode(ubi);
> -		leb_write_unlock(ubi, vol_id, lnum);
> +		ubi_eba_leb_write_unlock(ubi, vol_id, lnum);
>  		ubi_free_vid_hdr(ubi, vid_hdr);
>  		return err;
>  	}
> @@ -846,10 +918,10 @@ int ubi_eba_write_leb_st(struct ubi_device *ubi, struct ubi_volume *vol,
>  	vid_hdr->data_crc = cpu_to_be32(crc);
>  
>  retry:
> -	pnum = ubi_wl_get_peb(ubi);
> +	pnum = ubi_wl_get_peb(ubi, false);
>  	if (pnum < 0) {
>  		ubi_free_vid_hdr(ubi, vid_hdr);
> -		leb_write_unlock(ubi, vol_id, lnum);
> +		ubi_eba_leb_write_unlock(ubi, vol_id, lnum);
>  		up_read(&ubi->fm_eba_sem);
>  		return pnum;
>  	}
> @@ -875,10 +947,20 @@ retry:
>  
>  	ubi_assert(vol->eba_tbl[lnum] < 0);
>  	vol->eba_tbl[lnum] = pnum;
> +	vol->used_ebs = used_ebs; //XXX
>  	up_read(&ubi->fm_eba_sem);
>  
> -	leb_write_unlock(ubi, vol_id, lnum);
> +	err = ubi_coso_add_full_leb(ubi, vol_id, lnum, 0);
> +	if (err)
> +		ubi_warn(ubi, "failed to add LEB %d:%d to the full LEB list",
> +			 vol_id, lnum);
> +
> +	ubi_eba_leb_write_unlock(ubi, vol_id, lnum);
>  	ubi_free_vid_hdr(ubi, vid_hdr);
> +
> +	if (ubi_conso_consolidation_needed(ubi))
> +		ubi_conso_schedule(ubi);
> +
>  	return 0;
>  
>  write_error:
> @@ -889,7 +971,7 @@ write_error:
>  		 * mode just in case.
>  		 */
>  		ubi_ro_mode(ubi);
> -		leb_write_unlock(ubi, vol_id, lnum);
> +		ubi_eba_leb_write_unlock(ubi, vol_id, lnum);
>  		ubi_free_vid_hdr(ubi, vid_hdr);
>  		return err;
>  	}
> @@ -897,7 +979,7 @@ write_error:
>  	err = ubi_wl_put_peb(ubi, pnum, 1);
>  	if (err || ++tries > UBI_IO_RETRIES) {
>  		ubi_ro_mode(ubi);
> -		leb_write_unlock(ubi, vol_id, lnum);
> +		ubi_eba_leb_write_unlock(ubi, vol_id, lnum);
>  		ubi_free_vid_hdr(ubi, vid_hdr);
>  		return err;
>  	}
> @@ -926,7 +1008,9 @@ int ubi_eba_atomic_leb_change(struct ubi_device *ubi, struct ubi_volume *vol,
>  {
>  	int err, pnum, old_pnum, tries = 0, vol_id = vol->vol_id;
>  	struct ubi_vid_hdr *vid_hdr;
> +	bool release_peb = false;
>  	uint32_t crc;
> +	bool full;
>  
>  	if (ubi->ro_mode)
>  		return -EROFS;
> @@ -942,6 +1026,8 @@ int ubi_eba_atomic_leb_change(struct ubi_device *ubi, struct ubi_volume *vol,
>  		return ubi_eba_write_leb(ubi, vol, lnum, NULL, 0, 0);
>  	}
>  
> +	full = (len > ubi->leb_size - ubi->min_io_size);
> +
>  	vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
>  	if (!vid_hdr)
>  		return -ENOMEM;
> @@ -963,7 +1049,7 @@ int ubi_eba_atomic_leb_change(struct ubi_device *ubi, struct ubi_volume *vol,
>  	vid_hdr->data_crc = cpu_to_be32(crc);
>  
>  retry:
> -	pnum = ubi_wl_get_peb(ubi);
> +	pnum = ubi_wl_get_peb(ubi, false);
>  	if (pnum < 0) {
>  		err = pnum;
>  		up_read(&ubi->fm_eba_sem);
> @@ -991,18 +1077,33 @@ retry:
>  
>  	old_pnum = vol->eba_tbl[lnum];
>  	vol->eba_tbl[lnum] = pnum;
> +	if (old_pnum >= 0)
> +		release_peb = ubi_conso_invalidate_leb(ubi, old_pnum, vol_id, lnum);
>  	up_read(&ubi->fm_eba_sem);
> +	ubi_conso_remove_full_leb(ubi, vol_id, lnum);
> +	if (full) {
> +		int ret;
> +
> +		ret = ubi_coso_add_full_leb(ubi, vol_id, lnum, 0);
> +		if (ret)
> +			ubi_warn(ubi,
> +				"failed to add LEB %d:%d to the full LEB list",
> +				vol_id, lnum);
> +	}
>  
> -	if (old_pnum >= 0) {
> +out_leb_unlock:
> +	ubi_eba_leb_write_unlock(ubi, vol_id, lnum);
> +	if (release_peb) {
>  		err = ubi_wl_put_peb(ubi, old_pnum, 0);
>  		if (err)
>  			goto out_leb_unlock;
>  	}
> -
> -out_leb_unlock:
> -	leb_write_unlock(ubi, vol_id, lnum);
>  out_mutex:
>  	ubi_free_vid_hdr(ubi, vid_hdr);
> +
> +	if (full && !err && ubi_conso_consolidation_needed(ubi))
> +		ubi_conso_schedule(ubi);
> +
>  	return err;
>  
>  write_error:
> @@ -1224,7 +1325,178 @@ int ubi_eba_copy_leb(struct ubi_device *ubi, int from, int to,
>  out_unlock_buf:
>  	mutex_unlock(&ubi->buf_mutex);
>  out_unlock_leb:
> -	leb_write_unlock(ubi, vol_id, lnum);
> +	ubi_eba_leb_write_unlock(ubi, vol_id, lnum);
> +	return err;
> +}
> +
> +
> +/**
> + * ubi_eba_copy_lebs - copy consolidated logical eraseblocks.
> + *
> + * Works like ubi_eba_copy_leb but on consolidated LEB.
> + * It is less complicated as a PEB containing consolidated LEBs
> + * has only full LEBs and we don't have to do a lot of space
> + * calucation.
> + * TODO: clean this function up, more clean error handling, etc...
> + */
> +int ubi_eba_copy_lebs(struct ubi_device *ubi, int from, int to,
> +		     struct ubi_vid_hdr *vid_hdr, int nvidh)
> +{
> +	int err, i;
> +	int *vol_id = NULL, *lnum = NULL;
> +	struct ubi_volume **vol = NULL;
> +	uint32_t crc;
> +
> +	vol_id = kmalloc(nvidh * sizeof(*vol_id), GFP_NOFS);
> +	lnum = kmalloc(nvidh * sizeof(*lnum), GFP_NOFS);
> +	vol = kmalloc(nvidh * sizeof(*vol), GFP_NOFS);
> +
> +	if (!vol_id || !lnum || !vol) {
> +		kfree(vol_id);
> +		kfree(lnum);
> +		kfree(vol);
> +		return -ENOMEM;
> +	}
> +
> +	dbg_wl("copy LEBs of PEB %d to PEB %d", from, to);
> +
> +	spin_lock(&ubi->volumes_lock);
> +
> +	for (i = 0; i < nvidh; i++) {
> +		vol_id[i] = be32_to_cpu(vid_hdr[i].vol_id);
> +		lnum[i] = be32_to_cpu(vid_hdr[i].lnum);
> +		vol[i] = ubi->volumes[vol_id2idx(ubi, vol_id[i])];
> +	}
> +
> +	/*
> +	 * Note, we may race with volume deletion, which means that the volume
> +	 * this logical eraseblock belongs to might be being deleted. Since the
> +	 * volume deletion un-maps all the volume's logical eraseblocks, it will
> +	 * be locked in 'ubi_wl_put_peb()' and wait for the WL worker to finish.
> +	 */
> +	spin_unlock(&ubi->volumes_lock);
> +
> +	for (i = 0; i < nvidh; i++) {
> +		if (!vol[i]) {
> +			/* No need to do further work, cancel */
> +			ubi_msg(ubi, "volume %d is being removed, cancel", vol_id[i]);
> +			kfree(vol_id);
> +			kfree(lnum);
> +			kfree(vol);
> +			return MOVE_CANCEL_RACE;
> +		}
> +	}
> +
> +	/*
> +	 * We do not want anybody to write to this logical eraseblock while we
> +	 * are moving it, so lock it.
> +	 *
> +	 * Note, we are using non-waiting locking here, because we cannot sleep
> +	 * on the LEB, since it may cause deadlocks. Indeed, imagine a task is
> +	 * unmapping the LEB which is mapped to the PEB we are going to move
> +	 * (@from). This task locks the LEB and goes sleep in the
> +	 * 'ubi_wl_put_peb()' function on the @ubi->move_mutex. In turn, we are
> +	 * holding @ubi->move_mutex and go sleep on the LEB lock. So, if the
> +	 * LEB is already locked, we just do not move it and return
> +	 * %MOVE_RETRY. Note, we do not return %MOVE_CANCEL_RACE here because
> +	 * we do not know the reasons of the contention - it may be just a
> +	 * normal I/O on this LEB, so we want to re-try.
> +	 */
> +
> +	for (i = 0; i < nvidh; i++) {
> +		err = leb_write_trylock(ubi, vol_id[i], lnum[i]);
> +		if (err) {
> +			int j;
> +
> +			for (j = 0; j < i; j++)
> +				ubi_eba_leb_write_unlock(ubi, vol_id[j], lnum[j]);
> +
> +			kfree(vol_id);
> +			kfree(lnum);
> +			kfree(vol);
> +			return MOVE_RETRY;
> +		}
> +	}
> +	for (i = 0; i < nvidh; i++) {
> +		/*
> +		 * The LEB might have been put meanwhile, and the task which put it is
> +		 * probably waiting on @ubi->move_mutex. No need to continue the work,
> +		 * cancel it.
> +		 */
> +		if (vol[i]->eba_tbl[lnum[i]] != from) {
> +			ubi_msg(ubi, "LEB %d:%d is no longer mapped to PEB %d, mapped to PEB %d, cancel",
> +			       vol_id[i], lnum[i], from, vol[i]->eba_tbl[lnum[i]]);
> +			err = MOVE_CANCEL_RACE;
> +			goto out_unlock_leb;
> +		}
> +	}
> +
> +	/*
> +	 * OK, now the LEB is locked and we can safely start moving it. Since
> +	 * this function utilizes the @ubi->peb_buf buffer which is shared
> +	 * with some other functions - we lock the buffer by taking the
> +	 * @ubi->buf_mutex.
> +	 */
> +	mutex_lock(&ubi->buf_mutex);
> +	dbg_wl("read %d bytes of data", ubi->consolidated_peb_size - ubi->leb_start);
> +	err = ubi_io_raw_read(ubi, ubi->peb_buf, from, ubi->leb_start, ubi->consolidated_peb_size - ubi->leb_start);
> +	if (err && err != UBI_IO_BITFLIPS) {
> +		ubi_warn(ubi, "error %d while reading data from PEB %d",
> +			 err, from);
> +		err = MOVE_SOURCE_RD_ERR;
> +		goto out_unlock_buf;
> +	}
> +
> +	cond_resched();
> +	for (i = 0; i < nvidh; i++) {
> +		//TODO: we could skip crc calucation as consolidated LEB _always_ hav copy_flag=1 and hence also a valid crc...
> +		crc = crc32(UBI_CRC32_INIT, ubi->peb_buf + ubi->leb_start + (i * ubi->leb_size), be32_to_cpu(vid_hdr[i].data_size));
> +		vid_hdr[i].copy_flag = 1;
> +		vid_hdr[i].data_crc = cpu_to_be32(crc);
> +		vid_hdr[i].sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
> +
> +		cond_resched();
> +	}
> +
> +
> +	err = ubi_io_write_vid_hdrs(ubi, to, vid_hdr, nvidh);
> +	if (err) {
> +		if (err == -EIO)
> +			err = MOVE_TARGET_WR_ERR;
> +		goto out_unlock_buf;
> +	}
> +
> +	cond_resched();
> +
> +	err = ubi_io_raw_write(ubi, ubi->peb_buf, to, ubi->leb_start, ubi->consolidated_peb_size - ubi->leb_start);
> +	if (err) {
> +		if (err == -EIO)
> +			err = MOVE_TARGET_WR_ERR;
> +		goto out_unlock_buf;
> +	}
> +
> +	cond_resched();
> +
> +	down_read(&ubi->fm_eba_sem);
> +	for (i = 0; i < nvidh; i++) {
> +		ubi_assert(vol[i]->eba_tbl[lnum[i]] == from);
> +		vol[i]->eba_tbl[lnum[i]] = to;
> +	}
> +
> +	ubi->consolidated[to] = ubi->consolidated[from];
> +	ubi->consolidated[from] = NULL;
> +
> +	up_read(&ubi->fm_eba_sem);
> +
> +out_unlock_buf:
> +	mutex_unlock(&ubi->buf_mutex);
> +out_unlock_leb:
> +	for (i = 0; i < nvidh; i++)
> +		ubi_eba_leb_write_unlock(ubi, vol_id[i], lnum[i]);
> +	kfree(vol_id);
> +	kfree(lnum);
> +	kfree(vol);
> +
>  	return err;
>  }
>  
> @@ -1286,7 +1558,7 @@ int self_check_eba(struct ubi_device *ubi, struct ubi_attach_info *ai_fastmap,
>  	int **scan_eba, **fm_eba;
>  	struct ubi_ainf_volume *av;
>  	struct ubi_volume *vol;
> -	struct ubi_ainf_peb *aeb;
> +	struct ubi_ainf_leb *aeb;
>  	struct rb_node *rb;
>  
>  	num_volumes = ubi->vtbl_slots + UBI_INT_VOL_COUNT;
> @@ -1306,38 +1578,38 @@ int self_check_eba(struct ubi_device *ubi, struct ubi_attach_info *ai_fastmap,
>  		if (!vol)
>  			continue;
>  
> -		scan_eba[i] = kmalloc(vol->reserved_pebs * sizeof(**scan_eba),
> +		scan_eba[i] = kmalloc(vol->reserved_lebs * sizeof(**scan_eba),
>  				      GFP_KERNEL);
>  		if (!scan_eba[i]) {
>  			ret = -ENOMEM;
>  			goto out_free;
>  		}
>  
> -		fm_eba[i] = kmalloc(vol->reserved_pebs * sizeof(**fm_eba),
> +		fm_eba[i] = kmalloc(vol->reserved_lebs * sizeof(**fm_eba),
>  				    GFP_KERNEL);
>  		if (!fm_eba[i]) {
>  			ret = -ENOMEM;
>  			goto out_free;
>  		}
>  
> -		for (j = 0; j < vol->reserved_pebs; j++)
> +		for (j = 0; j < vol->reserved_lebs; j++)
>  			scan_eba[i][j] = fm_eba[i][j] = UBI_LEB_UNMAPPED;
>  
>  		av = ubi_find_av(ai_scan, idx2vol_id(ubi, i));
>  		if (!av)
>  			continue;
>  
> -		ubi_rb_for_each_entry(rb, aeb, &av->root, u.rb)
> -			scan_eba[i][aeb->lnum] = aeb->pnum;
> +		ubi_rb_for_each_entry(rb, aeb, &av->root, rb)
> +			scan_eba[i][aeb->desc.lnum] = aeb->peb->pnum;
>  
>  		av = ubi_find_av(ai_fastmap, idx2vol_id(ubi, i));
>  		if (!av)
>  			continue;
>  
> -		ubi_rb_for_each_entry(rb, aeb, &av->root, u.rb)
> -			fm_eba[i][aeb->lnum] = aeb->pnum;
> +		ubi_rb_for_each_entry(rb, aeb, &av->root, rb)
> +			fm_eba[i][aeb->desc.lnum] = aeb->peb->pnum;
>  
> -		for (j = 0; j < vol->reserved_pebs; j++) {
> +		for (j = 0; j < vol->reserved_lebs; j++) {
>  			if (scan_eba[i][j] != fm_eba[i][j]) {
>  				if (scan_eba[i][j] == UBI_LEB_UNMAPPED ||
>  					fm_eba[i][j] == UBI_LEB_UNMAPPED)
> @@ -1378,8 +1650,9 @@ int ubi_eba_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
>  	int i, j, err, num_volumes;
>  	struct ubi_ainf_volume *av;
>  	struct ubi_volume *vol;
> -	struct ubi_ainf_peb *aeb;
> +	struct ubi_ainf_leb *aeb;
>  	struct rb_node *rb;
> +	int eba_rsvd = EBA_RESERVED_PEBS;
>  
>  	dbg_eba("initialize EBA sub-system");
>  
> @@ -1396,43 +1669,48 @@ int ubi_eba_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
>  
>  		cond_resched();
>  
> -		vol->eba_tbl = kmalloc(vol->reserved_pebs * sizeof(int),
> +		vol->eba_tbl = kmalloc(vol->reserved_lebs * sizeof(int),
>  				       GFP_KERNEL);
>  		if (!vol->eba_tbl) {
>  			err = -ENOMEM;
>  			goto out_free;
>  		}
>  
> -		for (j = 0; j < vol->reserved_pebs; j++)
> +		for (j = 0; j < vol->reserved_lebs; j++)
>  			vol->eba_tbl[j] = UBI_LEB_UNMAPPED;
>  
>  		av = ubi_find_av(ai, idx2vol_id(ubi, i));
>  		if (!av)
>  			continue;
>  
> -		ubi_rb_for_each_entry(rb, aeb, &av->root, u.rb) {
> -			if (aeb->lnum >= vol->reserved_pebs)
> +		ubi_rb_for_each_entry(rb, aeb, &av->root, rb) {
> +			if (aeb->desc.lnum >= vol->reserved_lebs) {
>  				/*
>  				 * This may happen in case of an unclean reboot
>  				 * during re-size.
>  				 */
> -				ubi_move_aeb_to_list(av, aeb, &ai->erase);
> -			else
> -				vol->eba_tbl[aeb->lnum] = aeb->pnum;
> +				if (--aeb->peb->refcount <= 0)
> +					list_move_tail(&aeb->peb->list, &ai->erase);
> +			} else {
> +				vol->eba_tbl[aeb->desc.lnum] = aeb->peb->pnum;
> +				if (aeb->full)
> +					ubi_coso_add_full_leb(ubi, vol->vol_id,
> +							      aeb->desc.lnum, 0);
> +			}
>  		}
>  	}
>  
> -	if (ubi->avail_pebs < EBA_RESERVED_PEBS) {
> +	if (ubi->avail_pebs < eba_rsvd) {
>  		ubi_err(ubi, "no enough physical eraseblocks (%d, need %d)",
> -			ubi->avail_pebs, EBA_RESERVED_PEBS);
> +			ubi->avail_pebs, eba_rsvd);
>  		if (ubi->corr_peb_count)
>  			ubi_err(ubi, "%d PEBs are corrupted and not used",
>  				ubi->corr_peb_count);
>  		err = -ENOSPC;
>  		goto out_free;
>  	}
> -	ubi->avail_pebs -= EBA_RESERVED_PEBS;
> -	ubi->rsvd_pebs += EBA_RESERVED_PEBS;
> +	ubi->avail_pebs -= eba_rsvd;
> +	ubi->rsvd_pebs += eba_rsvd;
>  
>  	if (ubi->bad_allowed) {
>  		ubi_calculate_reserved(ubi);
> @@ -1448,6 +1726,9 @@ int ubi_eba_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
>  		ubi->rsvd_pebs  += ubi->beb_rsvd_pebs;
>  	}
>  
> +	if (ubi->lebs_per_cpeb > 1)
> +		ubi_conso_schedule(ubi);
> +
>  	dbg_eba("EBA sub-system is initialized");
>  	return 0;
>  
> diff --git a/drivers/mtd/ubi/fastmap-wl.c b/drivers/mtd/ubi/fastmap-wl.c
> index cafa7b0..ec593fc 100644
> --- a/drivers/mtd/ubi/fastmap-wl.c
> +++ b/drivers/mtd/ubi/fastmap-wl.c
> @@ -180,6 +180,8 @@ void ubi_refill_pools(struct ubi_device *ubi)
>   * disabled. Returns zero in case of success and a negative error code in case
>   * of failure.
>   */
> +
> +#error call ubi_eba_consolidate
>  static int produce_free_peb(struct ubi_device *ubi)
>  {
>  	while (!ubi->free.rb_node) {
> @@ -201,7 +203,7 @@ static int produce_free_peb(struct ubi_device *ubi)
>   * negative error code in case of failure.
>   * Returns with ubi->fm_eba_sem held in read mode!
>   */
> -int ubi_wl_get_peb(struct ubi_device *ubi)
> +int ubi_wl_get_peb(struct ubi_device *ubi, bool nested)
>  {
>  	int ret, retried = 0;
>  	struct ubi_fm_pool *pool = &ubi->fm_pool;
> @@ -211,6 +213,15 @@ again:
>  	down_read(&ubi->fm_eba_sem);
>  	spin_lock(&ubi->wl_lock);
>  
> +	if (nested) {
> +		if (pool->used == pool->size) {
> +			ret = -ENOSPC;
> +			goto out_unlock;
> +		}
> +
> +		goto out_get_peb;
> +	}
> +
>  	/* We check here also for the WL pool because at this point we can
>  	 * refill the WL pool synchronous. */
>  	if (pool->used == pool->size || wl_pool->used == wl_pool->size) {
> @@ -243,9 +254,11 @@ again:
>  		goto again;
>  	}
>  
> +out_get_peb:
>  	ubi_assert(pool->used < pool->size);
>  	ret = pool->pebs[pool->used++];
>  	prot_queue_add(ubi, ubi->lookuptbl[ret]);
> +out_unlock:
>  	spin_unlock(&ubi->wl_lock);
>  out:
>  	return ret;
> @@ -291,7 +304,7 @@ int ubi_ensure_anchor_pebs(struct ubi_device *ubi)
>  	ubi->wl_scheduled = 1;
>  	spin_unlock(&ubi->wl_lock);
>  
> -	wrk = kmalloc(sizeof(struct ubi_work), GFP_NOFS);
> +	wrk = ubi_alloc_work(ubi);
>  	if (!wrk) {
>  		spin_lock(&ubi->wl_lock);
>  		ubi->wl_scheduled = 0;
> @@ -342,7 +355,7 @@ int ubi_wl_put_fm_peb(struct ubi_device *ubi, struct ubi_wl_entry *fm_e,
>  	spin_unlock(&ubi->wl_lock);
>  
>  	vol_id = lnum ? UBI_FM_DATA_VOLUME_ID : UBI_FM_SB_VOLUME_ID;
> -	return schedule_erase(ubi, e, torture);
> +	return schedule_erase(ubi, e, torture, false);
>  }
>  
>  /**
> diff --git a/drivers/mtd/ubi/fastmap.c b/drivers/mtd/ubi/fastmap.c
> index c878313..61f8fc6 100644
> --- a/drivers/mtd/ubi/fastmap.c
> +++ b/drivers/mtd/ubi/fastmap.c
> @@ -143,15 +143,13 @@ static int add_aeb(struct ubi_attach_info *ai, struct list_head *list,
>  {
>  	struct ubi_ainf_peb *aeb;
>  
> -	aeb = kmem_cache_alloc(ai->aeb_slab_cache, GFP_KERNEL);
> +	aeb = kmem_cache_alloc(ai->apeb_slab_cache, GFP_KERNEL);
>  	if (!aeb)
>  		return -ENOMEM;
>  
>  	aeb->pnum = pnum;
>  	aeb->ec = ec;
> -	aeb->lnum = -1;
>  	aeb->scrub = scrub;
> -	aeb->copy_flag = aeb->sqnum = 0;
>  
>  	ai->ec_sum += aeb->ec;
>  	ai->ec_count++;
> @@ -162,7 +160,7 @@ static int add_aeb(struct ubi_attach_info *ai, struct list_head *list,
>  	if (ai->min_ec > aeb->ec)
>  		ai->min_ec = aeb->ec;
>  
> -	list_add_tail(&aeb->u.list, list);
> +	list_add_tail(&aeb->list, list);
>  
>  	return 0;
>  }
> @@ -229,19 +227,19 @@ out:
>   * @av: target scan volume
>   */
>  static void assign_aeb_to_av(struct ubi_attach_info *ai,
> -			     struct ubi_ainf_peb *aeb,
> +			     struct ubi_ainf_leb *aeb,
>  			     struct ubi_ainf_volume *av)
>  {
> -	struct ubi_ainf_peb *tmp_aeb;
> +	struct ubi_ainf_leb *tmp_aeb;
>  	struct rb_node **p = &ai->volumes.rb_node, *parent = NULL;
>  
>  	p = &av->root.rb_node;
>  	while (*p) {
>  		parent = *p;
>  
> -		tmp_aeb = rb_entry(parent, struct ubi_ainf_peb, u.rb);
> -		if (aeb->lnum != tmp_aeb->lnum) {
> -			if (aeb->lnum < tmp_aeb->lnum)
> +		tmp_aeb = rb_entry(parent, struct ubi_ainf_leb, rb);
> +		if (aeb->desc.lnum != tmp_aeb->desc.lnum) {
> +			if (aeb->desc.lnum < tmp_aeb->desc.lnum)
>  				p = &(*p)->rb_left;
>  			else
>  				p = &(*p)->rb_right;
> @@ -251,11 +249,10 @@ static void assign_aeb_to_av(struct ubi_attach_info *ai,
>  			break;
>  	}
>  
> -	list_del(&aeb->u.list);
>  	av->leb_count++;
>  
> -	rb_link_node(&aeb->u.rb, parent, p);
> -	rb_insert_color(&aeb->u.rb, &av->root);
> +	rb_link_node(&aeb->rb, parent, p);
> +	rb_insert_color(&aeb->rb, &av->root);
>  }
>  
>  /**
> @@ -270,18 +267,18 @@ static void assign_aeb_to_av(struct ubi_attach_info *ai,
>   */
>  static int update_vol(struct ubi_device *ubi, struct ubi_attach_info *ai,
>  		      struct ubi_ainf_volume *av, struct ubi_vid_hdr *new_vh,
> -		      struct ubi_ainf_peb *new_aeb)
> +		      struct ubi_ainf_peb *new_peb, int peb_pos, bool full)
>  {
>  	struct rb_node **p = &av->root.rb_node, *parent = NULL;
> -	struct ubi_ainf_peb *aeb, *victim;
> -	int cmp_res;
> +	struct ubi_ainf_leb *aeb;
> +	int cmp_res, lnum = be32_to_cpu(new_vh->lnum);
>  
>  	while (*p) {
>  		parent = *p;
> -		aeb = rb_entry(parent, struct ubi_ainf_peb, u.rb);
> +		aeb = rb_entry(parent, struct ubi_ainf_leb, rb);
>  
> -		if (be32_to_cpu(new_vh->lnum) != aeb->lnum) {
> -			if (be32_to_cpu(new_vh->lnum) < aeb->lnum)
> +		if (be32_to_cpu(new_vh->lnum) != aeb->desc.lnum) {
> +			if (be32_to_cpu(new_vh->lnum) < aeb->desc.lnum)
>  				p = &(*p)->rb_left;
>  			else
>  				p = &(*p)->rb_right;
> @@ -293,51 +290,58 @@ static int update_vol(struct ubi_device *ubi, struct ubi_attach_info *ai,
>  		 * because of a volume change (creation, deletion, ..).
>  		 * Then a PEB can be within the persistent EBA and the pool.
>  		 */
> -		if (aeb->pnum == new_aeb->pnum) {
> -			ubi_assert(aeb->lnum == new_aeb->lnum);
> -			kmem_cache_free(ai->aeb_slab_cache, new_aeb);
> +		if (aeb->peb->pnum == new_peb->pnum) {
> +			ubi_assert(aeb->desc.lnum == lnum);
> +			kmem_cache_free(ai->apeb_slab_cache, new_peb);
>  
>  			return 0;
>  		}
>  
> -		cmp_res = ubi_compare_lebs(ubi, aeb, new_aeb->pnum, new_vh);
> +		cmp_res = ubi_compare_lebs(ubi, aeb, new_peb->pnum, new_vh);
>  		if (cmp_res < 0)
>  			return cmp_res;
>  
>  		/* new_aeb is newer */
>  		if (cmp_res & 1) {
> -			victim = kmem_cache_alloc(ai->aeb_slab_cache,
> -				GFP_KERNEL);
> -			if (!victim)
> -				return -ENOMEM;
> -
> -			victim->ec = aeb->ec;
> -			victim->pnum = aeb->pnum;
> -			list_add_tail(&victim->u.list, &ai->erase);
> +			if (--aeb->peb->refcount <= 0)
> +				list_move(&aeb->peb->list, &ai->erase);
>  
>  			if (av->highest_lnum == be32_to_cpu(new_vh->lnum))
>  				av->last_data_size =
>  					be32_to_cpu(new_vh->data_size);
>  
>  			dbg_bld("vol %i: AEB %i's PEB %i is the newer",
> -				av->vol_id, aeb->lnum, new_aeb->pnum);
> +				av->vol_id, aeb->desc.lnum,
> +				new_peb->pnum);
>  
> -			aeb->ec = new_aeb->ec;
> -			aeb->pnum = new_aeb->pnum;
> +			aeb->peb_pos = peb_pos;
> +			aeb->peb = new_peb;
>  			aeb->copy_flag = new_vh->copy_flag;
> -			aeb->scrub = new_aeb->scrub;
> -			kmem_cache_free(ai->aeb_slab_cache, new_aeb);
> +			aeb->sqnum = be64_to_cpu(new_vh->sqnum);
> +			aeb->full = full;
>  
>  		/* new_aeb is older */
>  		} else {
>  			dbg_bld("vol %i: AEB %i's PEB %i is old, dropping it",
> -				av->vol_id, aeb->lnum, new_aeb->pnum);
> -			list_add_tail(&new_aeb->u.list, &ai->erase);
> +				av->vol_id, aeb->desc.lnum, new_peb->pnum);
> +			if (--aeb->peb->refcount <= 0)
> +				list_move(&aeb->peb->list, &ai->erase);
>  		}
>  
>  		return 0;
>  	}
>  	/* This LEB is new, let's add it to the volume */
> +	aeb = kmem_cache_alloc(ai->aleb_slab_cache, GFP_KERNEL);
> +	if (!aeb)
> +		return -ENOMEM;
> +
> +	aeb->peb = new_peb;
> +	aeb->peb_pos = peb_pos;
> +	aeb->copy_flag = new_vh->copy_flag;
> +	aeb->full = full;
> +	aeb->sqnum = be64_to_cpu(new_vh->sqnum);
> +	aeb->desc.lnum = lnum;
> +	aeb->desc.vol_id = be32_to_cpu(new_vh->vol_id);
>  
>  	if (av->highest_lnum <= be32_to_cpu(new_vh->lnum)) {
>  		av->highest_lnum = be32_to_cpu(new_vh->lnum);
> @@ -349,8 +353,8 @@ static int update_vol(struct ubi_device *ubi, struct ubi_attach_info *ai,
>  
>  	av->leb_count++;
>  
> -	rb_link_node(&new_aeb->u.rb, parent, p);
> -	rb_insert_color(&new_aeb->u.rb, &av->root);
> +	rb_link_node(&aeb->rb, parent, p);
> +	rb_insert_color(&aeb->rb, &av->root);
>  
>  	return 0;
>  }
> @@ -366,7 +370,8 @@ static int update_vol(struct ubi_device *ubi, struct ubi_attach_info *ai,
>   */
>  static int process_pool_aeb(struct ubi_device *ubi, struct ubi_attach_info *ai,
>  			    struct ubi_vid_hdr *new_vh,
> -			    struct ubi_ainf_peb *new_aeb)
> +			    struct ubi_ainf_peb *new_aeb, int peb_pos,
> +			    bool full)
>  {
>  	struct ubi_ainf_volume *av, *tmp_av = NULL;
>  	struct rb_node **p = &ai->volumes.rb_node, *parent = NULL;
> @@ -374,7 +379,7 @@ static int process_pool_aeb(struct ubi_device *ubi, struct ubi_attach_info *ai,
>  
>  	if (be32_to_cpu(new_vh->vol_id) == UBI_FM_SB_VOLUME_ID ||
>  		be32_to_cpu(new_vh->vol_id) == UBI_FM_DATA_VOLUME_ID) {
> -		kmem_cache_free(ai->aeb_slab_cache, new_aeb);
> +		kmem_cache_free(ai->apeb_slab_cache, new_aeb);
>  
>  		return 0;
>  	}
> @@ -398,13 +403,13 @@ static int process_pool_aeb(struct ubi_device *ubi, struct ubi_attach_info *ai,
>  		av = tmp_av;
>  	else {
>  		ubi_err(ubi, "orphaned volume in fastmap pool!");
> -		kmem_cache_free(ai->aeb_slab_cache, new_aeb);
> +		kmem_cache_free(ai->apeb_slab_cache, new_aeb);
>  		return UBI_BAD_FASTMAP;
>  	}
>  
>  	ubi_assert(be32_to_cpu(new_vh->vol_id) == av->vol_id);
>  
> -	return update_vol(ubi, ai, av, new_vh, new_aeb);
> +	return update_vol(ubi, ai, av, new_vh, new_aeb, peb_pos, full);
>  }
>  
>  /**
> @@ -419,18 +424,20 @@ static void unmap_peb(struct ubi_attach_info *ai, int pnum)
>  {
>  	struct ubi_ainf_volume *av;
>  	struct rb_node *node, *node2;
> -	struct ubi_ainf_peb *aeb;
> +	struct ubi_ainf_leb *aeb;
>  
>  	for (node = rb_first(&ai->volumes); node; node = rb_next(node)) {
>  		av = rb_entry(node, struct ubi_ainf_volume, rb);
>  
>  		for (node2 = rb_first(&av->root); node2;
>  		     node2 = rb_next(node2)) {
> -			aeb = rb_entry(node2, struct ubi_ainf_peb, u.rb);
> -			if (aeb->pnum == pnum) {
> -				rb_erase(&aeb->u.rb, &av->root);
> +			aeb = rb_entry(node2, struct ubi_ainf_leb, rb);
> +			if (aeb->peb->pnum == pnum) {
> +				rb_erase(&aeb->rb, &av->root);
>  				av->leb_count--;
> -				kmem_cache_free(ai->aeb_slab_cache, aeb);
> +				if (--aeb->peb->refcount <= 0)
> +					list_move(&aeb->peb->list, &ai->erase);
> +				kmem_cache_free(ai->apeb_slab_cache, aeb);
>  				return;
>  			}
>  		}
> @@ -456,7 +463,7 @@ static int scan_pool(struct ubi_device *ubi, struct ubi_attach_info *ai,
>  	struct ubi_vid_hdr *vh;
>  	struct ubi_ec_hdr *ech;
>  	struct ubi_ainf_peb *new_aeb;
> -	int i, pnum, err, ret = 0;
> +	int i, pnum, err, ret = 0, nvid;
>  
>  	ech = kzalloc(ubi->ec_hdr_alsize, GFP_KERNEL);
>  	if (!ech)
> @@ -508,7 +515,9 @@ static int scan_pool(struct ubi_device *ubi, struct ubi_attach_info *ai,
>  			goto out;
>  		}
>  
> -		err = ubi_io_read_vid_hdr(ubi, pnum, vh, 0);
> +		/* TODO: support consolidate PEBs */
> +		nvid = ubi->lebs_per_cpeb;
> +		err = ubi_io_read_vid_hdrs(ubi, pnum, vh, &nvid, 0);
>  		if (err == UBI_IO_FF || err == UBI_IO_FF_BITFLIPS) {
>  			unsigned long long ec = be64_to_cpu(ech->ec);
>  			unmap_peb(ai, pnum);
> @@ -519,12 +528,15 @@ static int scan_pool(struct ubi_device *ubi, struct ubi_attach_info *ai,
>  				add_aeb(ai, free, pnum, ec, 0);
>  			continue;
>  		} else if (err == 0 || err == UBI_IO_BITFLIPS) {
> +			bool full = false;
> +			int peb_pos;
> +
>  			dbg_bld("Found non empty PEB:%i in pool", pnum);
>  
>  			if (err == UBI_IO_BITFLIPS)
>  				scrub = 1;
>  
> -			new_aeb = kmem_cache_alloc(ai->aeb_slab_cache,
> +			new_aeb = kmem_cache_alloc(ai->apeb_slab_cache,
>  						   GFP_KERNEL);
>  			if (!new_aeb) {
>  				ret = -ENOMEM;
> @@ -533,18 +545,33 @@ static int scan_pool(struct ubi_device *ubi, struct ubi_attach_info *ai,
>  
>  			new_aeb->ec = be64_to_cpu(ech->ec);
>  			new_aeb->pnum = pnum;
> +			new_aeb->refcount = 1;
> +/*
>  			new_aeb->lnum = be32_to_cpu(vh->lnum);
>  			new_aeb->sqnum = be64_to_cpu(vh->sqnum);
>  			new_aeb->copy_flag = vh->copy_flag;
> +*/
>  			new_aeb->scrub = scrub;
> +			if (nvid == 1) {
> +				err = ubi_io_read(ubi, ech, pnum,
> +						  ubi->peb_size - ubi->hdrs_min_io_size,
> +						  ubi->hdrs_min_io_size);
> +				if (!err && !ubi_check_pattern(ech, 0xff, ubi->hdrs_min_io_size))
> +					full = true;
> +			}
>  
> -			if (*max_sqnum < new_aeb->sqnum)
> -				*max_sqnum = new_aeb->sqnum;
> +			new_aeb->consolidated = nvid > 1;
>  
> -			err = process_pool_aeb(ubi, ai, vh, new_aeb);
> -			if (err) {
> -				ret = err > 0 ? UBI_BAD_FASTMAP : err;
> -				goto out;
> +			for (peb_pos = 0; peb_pos < nvid; peb_pos++) {
> +				if (*max_sqnum < be64_to_cpu(vh[peb_pos].sqnum))
> +					*max_sqnum = be64_to_cpu(vh[peb_pos].sqnum);
> +
> +				err = process_pool_aeb(ubi, ai, &vh[peb_pos], new_aeb,
> +						       peb_pos, full);
> +				if (err) {
> +					ret = err > 0 ? UBI_BAD_FASTMAP : err;
> +					goto out;
> +				}
>  			}
>  		} else {
>  			/* We are paranoid and fall back to scanning mode */
> @@ -568,19 +595,16 @@ out:
>  static int count_fastmap_pebs(struct ubi_attach_info *ai)
>  {
>  	struct ubi_ainf_peb *aeb;
> -	struct ubi_ainf_volume *av;
> -	struct rb_node *rb1, *rb2;
>  	int n = 0;
>  
> -	list_for_each_entry(aeb, &ai->erase, u.list)
> +	list_for_each_entry(aeb, &ai->erase, list)
>  		n++;
>  
> -	list_for_each_entry(aeb, &ai->free, u.list)
> +	list_for_each_entry(aeb, &ai->free, list)
>  		n++;
>  
> -	 ubi_rb_for_each_entry(rb1, av, &ai->volumes, rb)
> -		ubi_rb_for_each_entry(rb2, aeb, &av->root, u.rb)
> -			n++;
> +	list_for_each_entry(aeb, &ai->used, list)
> +		n++;
>  
>  	return n;
>  }
> @@ -731,6 +755,9 @@ static int ubi_attach_fastmap(struct ubi_device *ubi,
>  
>  	/* Iterate over all volumes and read their EBA table */
>  	for (i = 0; i < be32_to_cpu(fmhdr->vol_count); i++) {
> +		struct ubi_fm_consolidated_leb *fclebs = NULL;
> +		int nconso = 0;
> +
>  		fmvhdr = (struct ubi_fm_volhdr *)(fm_raw + fm_pos);
>  		fm_pos += sizeof(*fmvhdr);
>  		if (fm_pos >= fm_size)
> @@ -760,6 +787,18 @@ static int ubi_attach_fastmap(struct ubi_device *ubi,
>  		if (ai->highest_vol_id < be32_to_cpu(fmvhdr->vol_id))
>  			ai->highest_vol_id = be32_to_cpu(fmvhdr->vol_id);
>  
> +		nconso = be32_to_cpu(fmvhdr->consolidated_ebs);
> +		if (nconso) {
> +			struct ubi_fm_consolidated *fmconso = NULL;
> +
> +			fmconso = fm_raw + fm_pos;
> +			if (be32_to_cpu(fmvhdr->magic) != UBI_FM_CONSO_MAGIC)
> +				goto fail_bad;
> +
> +			fclebs = fmconso->lebs;
> +			fm_pos += sizeof(*fmconso) + (nconso * sizeof(*fclebs));
> +		}
> +
>  		fm_eba = (struct ubi_fm_eba *)(fm_raw + fm_pos);
>  		fm_pos += sizeof(*fm_eba);
>  		fm_pos += (sizeof(__be32) * be32_to_cpu(fm_eba->reserved_pebs));
> @@ -774,12 +813,14 @@ static int ubi_attach_fastmap(struct ubi_device *ubi,
>  
>  		for (j = 0; j < be32_to_cpu(fm_eba->reserved_pebs); j++) {
>  			int pnum = be32_to_cpu(fm_eba->pnum[j]);
> +			struct ubi_ainf_leb *leb;
> +			int k;
>  
>  			if (pnum < 0)
>  				continue;
>  
>  			aeb = NULL;
> -			list_for_each_entry(tmp_aeb, &used, u.list) {
> +			list_for_each_entry(tmp_aeb, &used, list) {
>  				if (tmp_aeb->pnum == pnum) {
>  					aeb = tmp_aeb;
>  					break;
> @@ -791,18 +832,34 @@ static int ubi_attach_fastmap(struct ubi_device *ubi,
>  				goto fail_bad;
>  			}
>  
> -			aeb->lnum = j;
> +			leb = kmem_cache_alloc(ai->aleb_slab_cache, GFP_KERNEL);
> +			if (!leb)
> +				goto fail_bad;
> +
> +			leb->desc.lnum = j;
> +			leb->desc.vol_id = av->vol_id;
> +			leb->peb_pos = 0;
> +			for (k = 0; k < nconso; k++) {
> +				if (be32_to_cpu(fclebs[k].lnum) !=
> +				    leb->desc.lnum)
> +					continue;
>  
> -			if (av->highest_lnum <= aeb->lnum)
> -				av->highest_lnum = aeb->lnum;
> +				leb->peb_pos = be32_to_cpu(fclebs[k].peb_pos);
> +				aeb->consolidated = true;
> +			}
> +			leb->peb = aeb;
>  
> -			assign_aeb_to_av(ai, aeb, av);
> +			if (av->highest_lnum <= leb->desc.lnum)
> +				av->highest_lnum = leb->desc.lnum;
> +
> +			assign_aeb_to_av(ai, leb, av);
>  
>  			dbg_bld("inserting PEB:%i (LEB %i) to vol %i",
> -				aeb->pnum, aeb->lnum, av->vol_id);
> +				aeb->pnum, leb->desc.lnum, av->vol_id);
>  		}
>  	}
>  
> +
>  	ret = scan_pool(ubi, ai, fmpl->pebs, pool_size, &max_sqnum, &free);
>  	if (ret)
>  		goto fail;
> @@ -814,11 +871,11 @@ static int ubi_attach_fastmap(struct ubi_device *ubi,
>  	if (max_sqnum > ai->max_sqnum)
>  		ai->max_sqnum = max_sqnum;
>  
> -	list_for_each_entry_safe(tmp_aeb, _tmp_aeb, &free, u.list)
> -		list_move_tail(&tmp_aeb->u.list, &ai->free);
> +	list_for_each_entry_safe(tmp_aeb, _tmp_aeb, &free, list)
> +		list_move_tail(&tmp_aeb->list, &ai->free);
>  
> -	list_for_each_entry_safe(tmp_aeb, _tmp_aeb, &used, u.list)
> -		list_move_tail(&tmp_aeb->u.list, &ai->erase);
> +	list_for_each_entry_safe(tmp_aeb, _tmp_aeb, &used, list)
> +		list_move_tail(&tmp_aeb->list, &ai->erase);
>  
>  	ubi_assert(list_empty(&free));
>  
> @@ -837,13 +894,13 @@ static int ubi_attach_fastmap(struct ubi_device *ubi,
>  fail_bad:
>  	ret = UBI_BAD_FASTMAP;
>  fail:
> -	list_for_each_entry_safe(tmp_aeb, _tmp_aeb, &used, u.list) {
> -		list_del(&tmp_aeb->u.list);
> -		kmem_cache_free(ai->aeb_slab_cache, tmp_aeb);
> +	list_for_each_entry_safe(tmp_aeb, _tmp_aeb, &used, list) {
> +		list_del(&tmp_aeb->list);
> +		kmem_cache_free(ai->apeb_slab_cache, tmp_aeb);
>  	}
> -	list_for_each_entry_safe(tmp_aeb, _tmp_aeb, &free, u.list) {
> -		list_del(&tmp_aeb->u.list);
> -		kmem_cache_free(ai->aeb_slab_cache, tmp_aeb);
> +	list_for_each_entry_safe(tmp_aeb, _tmp_aeb, &free, list) {
> +		list_del(&tmp_aeb->list);
> +		kmem_cache_free(ai->apeb_slab_cache, tmp_aeb);
>  	}
>  
>  	return ret;
> @@ -1242,6 +1299,8 @@ static int ubi_write_fastmap(struct ubi_device *ubi,
>  	fmh->erase_peb_count = cpu_to_be32(erase_peb_count);
>  
>  	for (i = 0; i < UBI_MAX_VOLUMES + UBI_INT_VOL_COUNT; i++) {
> +		int nconso = 0;
> +
>  		vol = ubi->volumes[i];
>  
>  		if (!vol)
> @@ -1263,11 +1322,49 @@ static int ubi_write_fastmap(struct ubi_device *ubi,
>  		ubi_assert(vol->vol_type == UBI_DYNAMIC_VOLUME ||
>  			vol->vol_type == UBI_STATIC_VOLUME);
>  
> +		if (ubi->consolidated) {
> +			struct ubi_fm_consolidated *fconso;
> +			struct ubi_fm_consolidated_leb *fcleb;
> +
> +			fconso = (struct ubi_fm_consolidated *)(fm_raw + fm_pos);
> +			for (j = 0; j < vol->reserved_lebs; j++) {
> +				struct ubi_leb_desc *cleb;
> +				int k;
> +
> +				cleb = ubi->consolidated[vol->eba_tbl[j]];
> +				if (!cleb)
> +					continue;
> +
> +				fcleb = &fconso->lebs[nconso];
> +				fcleb->lnum = cpu_to_be32(j);
> +				for (k = 0; k < ubi->lebs_per_cpeb;
> +				     k++) {
> +					if (cleb[k].vol_id != vol->vol_id ||
> +					    cleb[k].lnum != j)
> +						continue;
> +
> +					fcleb->peb_pos = cpu_to_be32(k);
> +					break;
> +				}
> +
> +				if (k > ubi->lebs_per_cpeb) {
> +					ret = -EAGAIN;
> +					goto out_kfree;
> +				}
> +
> +				nconso++;
> +			}
> +
> +			if (nconso)
> +				fm_pos += sizeof(*fconso) +
> +					  (nconso * sizeof(*fcleb));
> +		}
> +
>  		feba = (struct ubi_fm_eba *)(fm_raw + fm_pos);
> -		fm_pos += sizeof(*feba) + (sizeof(__be32) * vol->reserved_pebs);
> +		fm_pos += sizeof(*feba) + (sizeof(__be32) * vol->reserved_lebs);
>  		ubi_assert(fm_pos <= ubi->fm_size);
>  
> -		for (j = 0; j < vol->reserved_pebs; j++)
> +		for (j = 0; j < vol->reserved_lebs; j++)
>  			feba->pnum[j] = cpu_to_be32(vol->eba_tbl[j]);
>  
>  		feba->reserved_pebs = cpu_to_be32(j);
> diff --git a/drivers/mtd/ubi/kapi.c b/drivers/mtd/ubi/kapi.c
> index dc315c2..c880897 100644
> --- a/drivers/mtd/ubi/kapi.c
> +++ b/drivers/mtd/ubi/kapi.c
> @@ -82,7 +82,7 @@ void ubi_do_get_volume_info(struct ubi_device *ubi, struct ubi_volume *vol,
>  {
>  	vi->vol_id = vol->vol_id;
>  	vi->ubi_num = ubi->ubi_num;
> -	vi->size = vol->reserved_pebs;
> +	vi->size = vol->reserved_lebs;
>  	vi->used_bytes = vol->used_bytes;
>  	vi->vol_type = vol->vol_type;
>  	vi->corrupted = vol->corrupted;
> @@ -532,7 +532,7 @@ int ubi_leb_write(struct ubi_volume_desc *desc, int lnum, const void *buf,
>  	if (desc->mode == UBI_READONLY || vol->vol_type == UBI_STATIC_VOLUME)
>  		return -EROFS;
>  
> -	if (lnum < 0 || lnum >= vol->reserved_pebs || offset < 0 || len < 0 ||
> +	if (lnum < 0 || lnum >= vol->reserved_lebs || offset < 0 || len < 0 ||
>  	    offset + len > vol->usable_leb_size ||
>  	    offset & (ubi->min_io_size - 1) || len & (ubi->min_io_size - 1))
>  		return -EINVAL;
> @@ -577,7 +577,7 @@ int ubi_leb_change(struct ubi_volume_desc *desc, int lnum, const void *buf,
>  	if (desc->mode == UBI_READONLY || vol->vol_type == UBI_STATIC_VOLUME)
>  		return -EROFS;
>  
> -	if (lnum < 0 || lnum >= vol->reserved_pebs || len < 0 ||
> +	if (lnum < 0 || lnum >= vol->reserved_lebs || len < 0 ||
>  	    len > vol->usable_leb_size || len & (ubi->min_io_size - 1))
>  		return -EINVAL;
>  
> @@ -614,7 +614,7 @@ int ubi_leb_erase(struct ubi_volume_desc *desc, int lnum)
>  	if (desc->mode == UBI_READONLY || vol->vol_type == UBI_STATIC_VOLUME)
>  		return -EROFS;
>  
> -	if (lnum < 0 || lnum >= vol->reserved_pebs)
> +	if (lnum < 0 || lnum >= vol->reserved_lebs)
>  		return -EINVAL;
>  
>  	if (vol->upd_marker)
> @@ -674,7 +674,7 @@ int ubi_leb_unmap(struct ubi_volume_desc *desc, int lnum)
>  	if (desc->mode == UBI_READONLY || vol->vol_type == UBI_STATIC_VOLUME)
>  		return -EROFS;
>  
> -	if (lnum < 0 || lnum >= vol->reserved_pebs)
> +	if (lnum < 0 || lnum >= vol->reserved_lebs)
>  		return -EINVAL;
>  
>  	if (vol->upd_marker)
> @@ -710,7 +710,7 @@ int ubi_leb_map(struct ubi_volume_desc *desc, int lnum)
>  	if (desc->mode == UBI_READONLY || vol->vol_type == UBI_STATIC_VOLUME)
>  		return -EROFS;
>  
> -	if (lnum < 0 || lnum >= vol->reserved_pebs)
> +	if (lnum < 0 || lnum >= vol->reserved_lebs)
>  		return -EINVAL;
>  
>  	if (vol->upd_marker)
> @@ -745,7 +745,7 @@ int ubi_is_mapped(struct ubi_volume_desc *desc, int lnum)
>  
>  	dbg_gen("test LEB %d:%d", vol->vol_id, lnum);
>  
> -	if (lnum < 0 || lnum >= vol->reserved_pebs)
> +	if (lnum < 0 || lnum >= vol->reserved_lebs)
>  		return -EINVAL;
>  
>  	if (vol->upd_marker)
> diff --git a/drivers/mtd/ubi/ubi-media.h b/drivers/mtd/ubi/ubi-media.h
> index 22ed3f6..500d7da 100644
> --- a/drivers/mtd/ubi/ubi-media.h
> +++ b/drivers/mtd/ubi/ubi-media.h
> @@ -363,7 +363,7 @@ struct ubi_vid_hdr {
>   * Empty records contain all zeroes and the CRC checksum of those zeroes.
>   */
>  struct ubi_vtbl_record {
> -	__be32  reserved_pebs;
> +	__be32  reserved_lebs;
>  	__be32  alignment;
>  	__be32  data_pad;
>  	__u8    vol_type;
> @@ -388,6 +388,7 @@ struct ubi_vtbl_record {
>  #define UBI_FM_VHDR_MAGIC	0xFA370ED1
>  #define UBI_FM_POOL_MAGIC	0x67AF4D08
>  #define UBI_FM_EBA_MAGIC	0xf0c040a8
> +#define UBI_FM_CONSO_MAGIC	0xc025011d
>  
>  /* A fastmap supber block can be located between PEB 0 and
>   * UBI_FM_MAX_START */
> @@ -444,7 +445,7 @@ struct ubi_fm_hdr {
>  	__be32 bad_peb_count;
>  	__be32 erase_peb_count;
>  	__be32 vol_count;
> -	__u8 padding[4];
> +	__be32 consolidated_count;
>  } __packed;
>  
>  /* struct ubi_fm_hdr is followed by two struct ubi_fm_scan_pool */
> @@ -494,7 +495,8 @@ struct ubi_fm_volhdr {
>  	__be32 data_pad;
>  	__be32 used_ebs;
>  	__be32 last_eb_bytes;
> -	__u8 padding2[8];
> +	__be32 consolidated_ebs;
> +	__u8 padding2[4];
>  } __packed;
>  
>  /* struct ubi_fm_volhdr is followed by one struct ubi_fm_eba records */
> @@ -510,4 +512,14 @@ struct ubi_fm_eba {
>  	__be32 reserved_pebs;
>  	__be32 pnum[0];
>  } __packed;
> +
> +struct ubi_fm_consolidated_leb {
> +	__be32 lnum;
> +	__be32 peb_pos;
> +} __packed;
> +
> +struct ubi_fm_consolidated {
> +	__be32 magic;
> +	struct ubi_fm_consolidated_leb lebs[0];
> +} __packed;
>  #endif /* !__UBI_MEDIA_H__ */
> diff --git a/drivers/mtd/ubi/ubi.h b/drivers/mtd/ubi/ubi.h
> index aab984f..47e5219 100644
> --- a/drivers/mtd/ubi/ubi.h
> +++ b/drivers/mtd/ubi/ubi.h
> @@ -90,6 +90,12 @@ void ubi_err(const struct ubi_device *ubi, const char *fmt, ...);
>  /* The volume ID/LEB number/erase counter is unknown */
>  #define UBI_UNKNOWN -1
>  
> +#ifdef CONFIG_MTD_UBI_CONSOLIDATE
> +/* Number of PEBs reserved for consolidation */
> +#define UBI_CONSO_RESERVED_PEBS 1
> +#else
> +#define UBI_CONSO_RESERVED_PEBS 0
> +#endif
>  /*
>   * The UBI debugfs directory name pattern and maximum name length (3 for "ubi"
>   * + 2 for the number plus 1 for the trailing zero byte.
> @@ -169,6 +175,20 @@ enum {
>  };
>  
>  /**
> + * struct ubi_leb_desc - UBI logical eraseblock description.
> + * @vol_id: volume ID of the locked logical eraseblock
> + * @lnum: locked logical eraseblock number
> + * @lpos: n'th LEB within this PEB, starting at 0
> + *
> + * This data structure is used in describe a logical eraseblock.
> + */
> +struct ubi_leb_desc {
> +	int vol_id;
> +	int lnum;
> +	int lpos;
> +};
> +
> +/**
>   * struct ubi_wl_entry - wear-leveling entry.
>   * @u.rb: link in the corresponding (free/used) RB-tree
>   * @u.list: link in the protection queue
> @@ -210,6 +230,11 @@ struct ubi_ltree_entry {
>  	struct rw_semaphore mutex;
>  };
>  
> +struct ubi_full_leb {
> +	struct list_head node;
> +	struct ubi_leb_desc desc;
> +};
> +
>  /**
>   * struct ubi_rename_entry - volume re-name description data structure.
>   * @new_name_len: new volume name length
> @@ -269,6 +294,18 @@ struct ubi_fm_pool {
>  };
>  
>  /**
> + * struct ubi_consolidable_leb - UBI consolidable LEB.
> + * @list: links RB-tree nodes
> + * @vol_id: volume ID
> + * @lnum: locked logical eraseblock number
> + */
> +struct ubi_consolidable_leb {
> +	struct list_head list;
> +	int vol_id;
> +	int lnum;
> +};
> +
> +/**
>   * struct ubi_volume - UBI volume description data structure.
>   * @dev: device object to make use of the the Linux device model
>   * @cdev: character device object to create character device
> @@ -329,7 +366,7 @@ struct ubi_volume {
>  	int exclusive;
>  	int metaonly;
>  
> -	int reserved_pebs;
> +	int reserved_lebs;
>  	int vol_type;
>  	int usable_leb_size;
>  	int used_ebs;
> @@ -561,6 +598,15 @@ struct ubi_device {
>  	spinlock_t ltree_lock;
>  	struct rb_root ltree;
>  
> +	int lebs_per_cpeb;
> +	struct ubi_leb_desc **consolidated;
> +	spinlock_t full_lock;
> +	struct list_head full;
> +	int full_count;
> +	int consolidation_threshold;
> +	int consolidation_pnum;
> +	struct list_head consolidable;
> +
>  	/* Fastmap stuff */
>  	int fm_disabled;
>  	struct ubi_fastmap_layout *fm;
> @@ -587,6 +633,7 @@ struct ubi_device {
>  	struct mutex work_mutex;
>  	struct ubi_work *cur_work;
>  	int wl_scheduled;
> +	int conso_scheduled;
>  	struct ubi_wl_entry **lookuptbl;
>  	struct ubi_wl_entry *move_from;
>  	struct ubi_wl_entry *move_to;
> @@ -648,17 +695,22 @@ struct ubi_device {
>   * volume, the @vol_id and @lnum fields are initialized to %UBI_UNKNOWN.
>   */
>  struct ubi_ainf_peb {
> +	struct list_head list;
>  	int ec;
>  	int pnum;
> -	int vol_id;
> -	int lnum;
> +	int refcount;
>  	unsigned int scrub:1;
> -	unsigned int copy_flag:1;
> +	unsigned int consolidated:1;
> +};
> +
> +struct ubi_ainf_leb {
> +	struct rb_node rb;
> +	struct ubi_leb_desc desc;
>  	unsigned long long sqnum;
> -	union {
> -		struct rb_node rb;
> -		struct list_head list;
> -	} u;
> +	unsigned int copy_flag:1;
> +	unsigned int full:1;
> +	int peb_pos;
> +	struct ubi_ainf_peb *peb;
>  };
>  
>  /**
> @@ -685,6 +737,7 @@ struct ubi_ainf_peb {
>  struct ubi_ainf_volume {
>  	int vol_id;
>  	int highest_lnum;
> +	unsigned long long int highest_sqnum;
>  	int leb_count;
>  	int vol_type;
>  	int used_ebs;
> @@ -731,6 +784,7 @@ struct ubi_attach_info {
>  	struct list_head free;
>  	struct list_head erase;
>  	struct list_head alien;
> +	struct list_head used;
>  	int corr_peb_count;
>  	int empty_peb_count;
>  	int alien_peb_count;
> @@ -745,7 +799,8 @@ struct ubi_attach_info {
>  	int mean_ec;
>  	uint64_t ec_sum;
>  	int ec_count;
> -	struct kmem_cache *aeb_slab_cache;
> +	struct kmem_cache *apeb_slab_cache;
> +	struct kmem_cache *aleb_slab_cache;
>  };
>  
>  /**
> @@ -790,8 +845,9 @@ extern struct mutex ubi_devices_mutex;
>  extern struct blocking_notifier_head ubi_notifiers;
>  
>  /* attach.c */
> -int ubi_add_to_av(struct ubi_device *ubi, struct ubi_attach_info *ai, int pnum,
> -		  int ec, const struct ubi_vid_hdr *vid_hdr, int bitflips);
> +int ubi_add_to_av(struct ubi_device *ubi, struct ubi_attach_info *ai,
> +		  struct ubi_ainf_peb *peb, const struct ubi_vid_hdr *vid_hdr,
> +		  int peb_pos, int bitflips, bool full);
>  struct ubi_ainf_volume *ubi_find_av(const struct ubi_attach_info *ai,
>  				    int vol_id);
>  void ubi_remove_av(struct ubi_attach_info *ai, struct ubi_ainf_volume *av);
> @@ -849,13 +905,56 @@ int ubi_eba_atomic_leb_change(struct ubi_device *ubi, struct ubi_volume *vol,
>  			      int lnum, const void *buf, int len);
>  int ubi_eba_copy_leb(struct ubi_device *ubi, int from, int to,
>  		     struct ubi_vid_hdr *vid_hdr);
> +int ubi_eba_copy_lebs(struct ubi_device *ubi, int from, int to,
> +		     struct ubi_vid_hdr *vid_hdr, int nvidh);
>  int ubi_eba_init(struct ubi_device *ubi, struct ubi_attach_info *ai);
>  unsigned long long ubi_next_sqnum(struct ubi_device *ubi);
> +int ubi_eba_leb_write_lock_nested(struct ubi_device *ubi, int vol_id, int lnum,
> +				  int level);
> +void ubi_eba_leb_write_unlock(struct ubi_device *ubi, int vol_id, int lnum);
>  int self_check_eba(struct ubi_device *ubi, struct ubi_attach_info *ai_fastmap,
>  		   struct ubi_attach_info *ai_scan);
>  
> +/* consolidate.c */
> +#ifdef CONFIG_MTD_UBI_CONSOLIDATE
> +bool ubi_conso_consolidation_needed(struct ubi_device *ubi);
> +void ubi_conso_schedule(struct ubi_device *ubi);
> +void ubi_eba_consolidate(struct ubi_device *ubi);
> +void ubi_conso_remove_full_leb(struct ubi_device *ubi, int vol_id, int lnum);
> +struct ubi_leb_desc *ubi_conso_get_consolidated(struct ubi_device *ubi,
> +						int pnum);
> +bool ubi_conso_invalidate_leb(struct ubi_device *ubi, int pnum,
> +			      int vol_id, int lnum);
> +int ubi_coso_add_full_leb(struct ubi_device *ubi, int vol_id, int lnum, int lpos);
> +int ubi_conso_init(struct ubi_device *ubi);
> +void ubi_conso_close(struct ubi_device *ubi);
> +#else
> +static inline bool ubi_conso_consolidation_needed(struct ubi_device *ubi)
> +{
> +	return false;
> +}
> +static inline void ubi_conso_schedule(struct ubi_device *ubi) {}
> +static inline void ubi_eba_consolidate(struct ubi_device *ubi) {}
> +static inline void ubi_conso_remove_full_leb(struct ubi_device *ubi, int vol_id, int lnum) {}
> +static inline struct ubi_leb_desc *ubi_conso_get_consolidated(struct ubi_device *ubi, int pnum)
> +{
> +	return NULL;
> +}
> +static inline bool ubi_conso_invalidate_leb(struct ubi_device *ubi, int pnum, int vol_id, int lnum)
> +{
> +	return true;
> +}
> +static inline int ubi_coso_add_full_leb(struct ubi_device *ubi, int vol_id, int lnum, int lpos)
> +{
> +	return 0;
> +}
> +static inline int ubi_conso_init(struct ubi_device *ubi) { return 0; }
> +static inline void ubi_conso_close(struct ubi_device *ubi) {}
> +#endif
> +
>  /* wl.c */
> -int ubi_wl_get_peb(struct ubi_device *ubi);
> +int ubi_wl_get_peb(struct ubi_device *ubi, bool producing);
> +int ubi_wl_flush(struct ubi_device *ubi);
>  int ubi_wl_put_peb(struct ubi_device *ubi, int pnum, int torture);
>  int ubi_wl_scrub_peb(struct ubi_device *ubi, int pnum);
>  int ubi_wl_init(struct ubi_device *ubi, struct ubi_attach_info *ai);
> @@ -876,7 +975,9 @@ void ubi_work_close(struct ubi_device *ubi, int error);
>  struct ubi_work *ubi_alloc_work(struct ubi_device *ubi);
>  int ubi_work_flush(struct ubi_device *ubi);
>  bool ubi_work_join_one(struct ubi_device *ubi);
> -
> +struct ubi_work *ubi_alloc_erase_work(struct ubi_device *ubi,
> +				      struct ubi_wl_entry *e,
> +				      int torture);
>  /* io.c */
>  int ubi_io_read(const struct ubi_device *ubi, void *buf, int pnum, int offset,
>  		int len);
> @@ -923,8 +1024,8 @@ void ubi_do_get_device_info(struct ubi_device *ubi, struct ubi_device_info *di);
>  void ubi_do_get_volume_info(struct ubi_device *ubi, struct ubi_volume *vol,
>  			    struct ubi_volume_info *vi);
>  /* scan.c */
> -int ubi_compare_lebs(struct ubi_device *ubi, const struct ubi_ainf_peb *aeb,
> -		      int pnum, const struct ubi_vid_hdr *vid_hdr);
> +int ubi_compare_lebs(struct ubi_device *ubi, const struct ubi_ainf_leb *aeb,
> +		     int pnum, const struct ubi_vid_hdr *vid_hdr);
>  
>  /* fastmap.c */
>  #ifdef CONFIG_MTD_UBI_FASTMAP
> @@ -955,6 +1056,22 @@ static inline int ubiblock_remove(struct ubi_volume_info *vi)
>  }
>  #endif
>  
> +/**
> + * ubi_get_compat - get compatibility flags of a volume.
> + * @ubi: UBI device description object
> + * @vol_id: volume ID
> + *
> + * This function returns compatibility flags for an internal volume. User
> + * volumes have no compatibility flags, so %0 is returned.
> + */
> +static inline int ubi_get_compat(const struct ubi_device *ubi, int vol_id)
> +{
> +	if (vol_id == UBI_LAYOUT_VOLUME_ID)
> +		return UBI_LAYOUT_VOLUME_COMPAT;
> +	return 0;
> +}
> +
> +
>  /*
>   * ubi_for_each_free_peb - walk the UBI free RB tree.
>   * @ubi: UBI device description object
> @@ -1006,21 +1123,6 @@ static inline int ubiblock_remove(struct ubi_volume_info *vi)
>  	     rb = rb_next(rb),                                               \
>  	     pos = (rb ? container_of(rb, typeof(*pos), member) : NULL))
>  
> -/*
> - * ubi_move_aeb_to_list - move a PEB from the volume tree to a list.
> - *
> - * @av: volume attaching information
> - * @aeb: attaching eraseblock information
> - * @list: the list to move to
> - */
> -static inline void ubi_move_aeb_to_list(struct ubi_ainf_volume *av,
> -					 struct ubi_ainf_peb *aeb,
> -					 struct list_head *list)
> -{
> -		rb_erase(&aeb->u.rb, &av->root);
> -		list_add_tail(&aeb->u.list, list);
> -}
> -
>  /**
>   * ubi_zalloc_vid_hdr - allocate a volume identifier header object.
>   * @ubi: UBI device description object
> diff --git a/drivers/mtd/ubi/upd.c b/drivers/mtd/ubi/upd.c
> index ffaface..cc21f64 100644
> --- a/drivers/mtd/ubi/upd.c
> +++ b/drivers/mtd/ubi/upd.c
> @@ -142,7 +142,7 @@ int ubi_start_update(struct ubi_device *ubi, struct ubi_volume *vol,
>  		return err;
>  
>  	/* Before updating - wipe out the volume */
> -	for (i = 0; i < vol->reserved_pebs; i++) {
> +	for (i = 0; i < vol->reserved_lebs; i++) {
>  		err = ubi_eba_unmap_leb(ubi, vol, i);
>  		if (err)
>  			return err;
> diff --git a/drivers/mtd/ubi/vmt.c b/drivers/mtd/ubi/vmt.c
> index 8a2e081..736acb2 100644
> --- a/drivers/mtd/ubi/vmt.c
> +++ b/drivers/mtd/ubi/vmt.c
> @@ -86,7 +86,7 @@ static ssize_t vol_attribute_show(struct device *dev,
>  	spin_unlock(&ubi->volumes_lock);
>  
>  	if (attr == &attr_vol_reserved_ebs)
> -		ret = sprintf(buf, "%d\n", vol->reserved_pebs);
> +		ret = sprintf(buf, "%d\n", vol->reserved_lebs);
>  	else if (attr == &attr_vol_type) {
>  		const char *tp;
>  
> @@ -158,6 +158,7 @@ int ubi_create_volume(struct ubi_device *ubi, struct ubi_mkvol_req *req)
>  	int i, err, vol_id = req->vol_id, do_free = 1;
>  	struct ubi_volume *vol;
>  	struct ubi_vtbl_record vtbl_rec;
> +	int rsvd_pebs = 0;
>  	dev_t dev;
>  
>  	if (ubi->ro_mode)
> @@ -208,11 +209,13 @@ int ubi_create_volume(struct ubi_device *ubi, struct ubi_mkvol_req *req)
>  
>  	/* Calculate how many eraseblocks are requested */
>  	vol->usable_leb_size = ubi->leb_size - ubi->leb_size % req->alignment;
> -	vol->reserved_pebs = div_u64(req->bytes + vol->usable_leb_size - 1,
> +	vol->reserved_lebs = div_u64(req->bytes + vol->usable_leb_size - 1,
>  				     vol->usable_leb_size);
>  
>  	/* Reserve physical eraseblocks */
> -	if (vol->reserved_pebs > ubi->avail_pebs) {
> +	rsvd_pebs = DIV_ROUND_UP(vol->reserved_lebs,
> +				 ubi->lebs_per_cpeb);
> +	if (rsvd_pebs > ubi->avail_pebs) {
>  		ubi_err(ubi, "not enough PEBs, only %d available",
>  			ubi->avail_pebs);
>  		if (ubi->corr_peb_count)
> @@ -221,8 +224,8 @@ int ubi_create_volume(struct ubi_device *ubi, struct ubi_mkvol_req *req)
>  		err = -ENOSPC;
>  		goto out_unlock;
>  	}
> -	ubi->avail_pebs -= vol->reserved_pebs;
> -	ubi->rsvd_pebs += vol->reserved_pebs;
> +	ubi->avail_pebs -= rsvd_pebs;
> +	ubi->rsvd_pebs += rsvd_pebs;
>  	spin_unlock(&ubi->volumes_lock);
>  
>  	vol->vol_id    = vol_id;
> @@ -241,17 +244,31 @@ int ubi_create_volume(struct ubi_device *ubi, struct ubi_mkvol_req *req)
>  	if (err)
>  		goto out_acc;
>  
> -	vol->eba_tbl = kmalloc(vol->reserved_pebs * sizeof(int), GFP_KERNEL);
> +	vol->eba_tbl = kmalloc(vol->reserved_lebs * sizeof(int), GFP_KERNEL);
>  	if (!vol->eba_tbl) {
>  		err = -ENOMEM;
>  		goto out_acc;
>  	}
>  
> -	for (i = 0; i < vol->reserved_pebs; i++)
> +#ifdef CONFIG_UBI_EXTENDED_PEB
> +	/*
> +	 * TODO: check in the underlying MTD device has a page pairing scheme
> +	 * requiring the consolidated bitmap creation.
> +	 */
> +	vol->consolidated = kzalloc(DIV_ROUND_UP(vol->reserved_lebs,
> +						 BITS_PER_LONG),
> +				    GFP_KERNEL);
> +	if (!vol->consolidated) {
> +		err = -ENOMEM;
> +		goto out_mapping;
> +	}
> +#endif
> +
> +	for (i = 0; i < vol->reserved_lebs; i++)
>  		vol->eba_tbl[i] = UBI_LEB_UNMAPPED;
>  
>  	if (vol->vol_type == UBI_DYNAMIC_VOLUME) {
> -		vol->used_ebs = vol->reserved_pebs;
> +		vol->used_ebs = vol->reserved_lebs;
>  		vol->last_eb_bytes = vol->usable_leb_size;
>  		vol->used_bytes =
>  			(long long)vol->used_ebs * vol->usable_leb_size;
> @@ -290,7 +307,7 @@ int ubi_create_volume(struct ubi_device *ubi, struct ubi_mkvol_req *req)
>  
>  	/* Fill volume table record */
>  	memset(&vtbl_rec, 0, sizeof(struct ubi_vtbl_record));
> -	vtbl_rec.reserved_pebs = cpu_to_be32(vol->reserved_pebs);
> +	vtbl_rec.reserved_lebs = cpu_to_be32(vol->reserved_lebs);
>  	vtbl_rec.alignment     = cpu_to_be32(vol->alignment);
>  	vtbl_rec.data_pad      = cpu_to_be32(vol->data_pad);
>  	vtbl_rec.name_len      = cpu_to_be16(vol->name_len);
> @@ -328,12 +345,16 @@ out_sysfs:
>  out_cdev:
>  	cdev_del(&vol->cdev);
>  out_mapping:
> -	if (do_free)
> +	if (do_free) {
>  		kfree(vol->eba_tbl);
> +#ifdef CONFIG_UBI_EXTENDED_PEB
> +		kfree(vol->consolidated);
> +#endif
> +	}
>  out_acc:
>  	spin_lock(&ubi->volumes_lock);
> -	ubi->rsvd_pebs -= vol->reserved_pebs;
> -	ubi->avail_pebs += vol->reserved_pebs;
> +	ubi->rsvd_pebs -= rsvd_pebs;
> +	ubi->avail_pebs += rsvd_pebs;
>  out_unlock:
>  	spin_unlock(&ubi->volumes_lock);
>  	if (do_free)
> @@ -358,7 +379,7 @@ int ubi_remove_volume(struct ubi_volume_desc *desc, int no_vtbl)
>  {
>  	struct ubi_volume *vol = desc->vol;
>  	struct ubi_device *ubi = vol->ubi;
> -	int i, err, vol_id = vol->vol_id, reserved_pebs = vol->reserved_pebs;
> +	int i, err, vol_id = vol->vol_id, reserved_pebs;
>  
>  	dbg_gen("remove device %d, volume %d", ubi->ubi_num, vol_id);
>  	ubi_assert(desc->mode == UBI_EXCLUSIVE);
> @@ -385,7 +406,7 @@ int ubi_remove_volume(struct ubi_volume_desc *desc, int no_vtbl)
>  			goto out_err;
>  	}
>  
> -	for (i = 0; i < vol->reserved_pebs; i++) {
> +	for (i = 0; i < vol->reserved_lebs; i++) {
>  		err = ubi_eba_unmap_leb(ubi, vol, i);
>  		if (err)
>  			goto out_err;
> @@ -394,6 +415,8 @@ int ubi_remove_volume(struct ubi_volume_desc *desc, int no_vtbl)
>  	cdev_del(&vol->cdev);
>  	device_unregister(&vol->dev);
>  
> +	reserved_pebs = DIV_ROUND_UP(vol->reserved_lebs,
> +				     ubi->lebs_per_cpeb);
>  	spin_lock(&ubi->volumes_lock);
>  	ubi->rsvd_pebs -= reserved_pebs;
>  	ubi->avail_pebs += reserved_pebs;
> @@ -425,9 +448,9 @@ out_unlock:
>   * negative error code in case of failure. The caller has to have the
>   * @ubi->device_mutex locked.
>   */
> -int ubi_resize_volume(struct ubi_volume_desc *desc, int reserved_pebs)
> +int ubi_resize_volume(struct ubi_volume_desc *desc, int reserved_lebs)
>  {
> -	int i, err, pebs, *new_mapping;
> +	int i, err, lebs, pebs, *new_mapping;
>  	struct ubi_volume *vol = desc->vol;
>  	struct ubi_device *ubi = vol->ubi;
>  	struct ubi_vtbl_record vtbl_rec;
> @@ -437,24 +460,24 @@ int ubi_resize_volume(struct ubi_volume_desc *desc, int reserved_pebs)
>  		return -EROFS;
>  
>  	dbg_gen("re-size device %d, volume %d to from %d to %d PEBs",
> -		ubi->ubi_num, vol_id, vol->reserved_pebs, reserved_pebs);
> +		ubi->ubi_num, vol_id, vol->reserved_lebs, reserved_lebs);
>  
>  	if (vol->vol_type == UBI_STATIC_VOLUME &&
> -	    reserved_pebs < vol->used_ebs) {
> +	    reserved_lebs < vol->used_ebs) {
>  		ubi_err(ubi, "too small size %d, %d LEBs contain data",
> -			reserved_pebs, vol->used_ebs);
> +			reserved_lebs, vol->used_ebs);
>  		return -EINVAL;
>  	}
>  
>  	/* If the size is the same, we have nothing to do */
> -	if (reserved_pebs == vol->reserved_pebs)
> +	if (reserved_lebs == vol->reserved_lebs)
>  		return 0;
>  
> -	new_mapping = kmalloc(reserved_pebs * sizeof(int), GFP_KERNEL);
> +	new_mapping = kmalloc(reserved_lebs * sizeof(int), GFP_KERNEL);
>  	if (!new_mapping)
>  		return -ENOMEM;
>  
> -	for (i = 0; i < reserved_pebs; i++)
> +	for (i = 0; i < reserved_lebs; i++)
>  		new_mapping[i] = UBI_LEB_UNMAPPED;
>  
>  	spin_lock(&ubi->volumes_lock);
> @@ -466,8 +489,12 @@ int ubi_resize_volume(struct ubi_volume_desc *desc, int reserved_pebs)
>  	spin_unlock(&ubi->volumes_lock);
>  
>  	/* Reserve physical eraseblocks */
> -	pebs = reserved_pebs - vol->reserved_pebs;
> -	if (pebs > 0) {
> +	lebs = reserved_lebs - vol->reserved_lebs;
> +	pebs = DIV_ROUND_UP(reserved_lebs,
> +			    ubi->lebs_per_cpeb) -
> +	       DIV_ROUND_UP(vol->reserved_lebs,
> +			    ubi->lebs_per_cpeb);
> +	if (lebs > 0) {
>  		spin_lock(&ubi->volumes_lock);
>  		if (pebs > ubi->avail_pebs) {
>  			ubi_err(ubi, "not enough PEBs: requested %d, available %d",
> @@ -481,7 +508,7 @@ int ubi_resize_volume(struct ubi_volume_desc *desc, int reserved_pebs)
>  		}
>  		ubi->avail_pebs -= pebs;
>  		ubi->rsvd_pebs += pebs;
> -		for (i = 0; i < vol->reserved_pebs; i++)
> +		for (i = 0; i < vol->reserved_lebs; i++)
>  			new_mapping[i] = vol->eba_tbl[i];
>  		kfree(vol->eba_tbl);
>  		vol->eba_tbl = new_mapping;
> @@ -490,31 +517,32 @@ int ubi_resize_volume(struct ubi_volume_desc *desc, int reserved_pebs)
>  
>  	/* Change volume table record */
>  	vtbl_rec = ubi->vtbl[vol_id];
> -	vtbl_rec.reserved_pebs = cpu_to_be32(reserved_pebs);
> +	vtbl_rec.reserved_lebs = cpu_to_be32(reserved_lebs);
>  	err = ubi_change_vtbl_record(ubi, vol_id, &vtbl_rec);
>  	if (err)
>  		goto out_acc;
>  
> -	if (pebs < 0) {
> -		for (i = 0; i < -pebs; i++) {
> -			err = ubi_eba_unmap_leb(ubi, vol, reserved_pebs + i);
> +	if (lebs < 0) {
> +		for (i = 0; i < -lebs; i++) {
> +			err = ubi_eba_unmap_leb(ubi, vol, reserved_lebs + i);
>  			if (err)
>  				goto out_acc;
>  		}
> +
>  		spin_lock(&ubi->volumes_lock);
>  		ubi->rsvd_pebs += pebs;
>  		ubi->avail_pebs -= pebs;
>  		ubi_update_reserved(ubi);
> -		for (i = 0; i < reserved_pebs; i++)
> +		for (i = 0; i < reserved_lebs; i++)
>  			new_mapping[i] = vol->eba_tbl[i];
>  		kfree(vol->eba_tbl);
>  		vol->eba_tbl = new_mapping;
>  		spin_unlock(&ubi->volumes_lock);
>  	}
>  
> -	vol->reserved_pebs = reserved_pebs;
> +	vol->reserved_lebs = reserved_lebs;
>  	if (vol->vol_type == UBI_DYNAMIC_VOLUME) {
> -		vol->used_ebs = reserved_pebs;
> +		vol->used_ebs = reserved_lebs;
>  		vol->last_eb_bytes = vol->usable_leb_size;
>  		vol->used_bytes =
>  			(long long)vol->used_ebs * vol->usable_leb_size;
> @@ -525,7 +553,7 @@ int ubi_resize_volume(struct ubi_volume_desc *desc, int reserved_pebs)
>  	return err;
>  
>  out_acc:
> -	if (pebs > 0) {
> +	if (lebs > 0) {
>  		spin_lock(&ubi->volumes_lock);
>  		ubi->rsvd_pebs -= pebs;
>  		ubi->avail_pebs += pebs;
> @@ -653,7 +681,7 @@ static int self_check_volume(struct ubi_device *ubi, int vol_id)
>  	const char *name;
>  
>  	spin_lock(&ubi->volumes_lock);
> -	reserved_pebs = be32_to_cpu(ubi->vtbl[vol_id].reserved_pebs);
> +	reserved_pebs = be32_to_cpu(ubi->vtbl[vol_id].reserved_lebs);
>  	vol = ubi->volumes[idx];
>  
>  	if (!vol) {
> @@ -665,7 +693,7 @@ static int self_check_volume(struct ubi_device *ubi, int vol_id)
>  		return 0;
>  	}
>  
> -	if (vol->reserved_pebs < 0 || vol->alignment < 0 || vol->data_pad < 0 ||
> +	if (vol->reserved_lebs < 0 || vol->alignment < 0 || vol->data_pad < 0 ||
>  	    vol->name_len < 0) {
>  		ubi_err(ubi, "negative values");
>  		goto fail;
> @@ -698,7 +726,7 @@ static int self_check_volume(struct ubi_device *ubi, int vol_id)
>  		goto fail;
>  	}
>  
> -	if (vol->reserved_pebs > ubi->good_peb_count) {
> +	if (vol->reserved_lebs > ubi->good_peb_count) {
>  		ubi_err(ubi, "too large reserved_pebs");
>  		goto fail;
>  	}
> @@ -727,7 +755,7 @@ static int self_check_volume(struct ubi_device *ubi, int vol_id)
>  			ubi_err(ubi, "corrupted dynamic volume");
>  			goto fail;
>  		}
> -		if (vol->used_ebs != vol->reserved_pebs) {
> +		if (vol->used_ebs != vol->reserved_lebs) {
>  			ubi_err(ubi, "bad used_ebs");
>  			goto fail;
>  		}
> @@ -740,7 +768,7 @@ static int self_check_volume(struct ubi_device *ubi, int vol_id)
>  			goto fail;
>  		}
>  	} else {
> -		if (vol->used_ebs < 0 || vol->used_ebs > vol->reserved_pebs) {
> +		if (vol->used_ebs < 0 || vol->used_ebs > vol->reserved_lebs) {
>  			ubi_err(ubi, "bad used_ebs");
>  			goto fail;
>  		}
> diff --git a/drivers/mtd/ubi/vtbl.c b/drivers/mtd/ubi/vtbl.c
> index d85c197..fd40fe5 100644
> --- a/drivers/mtd/ubi/vtbl.c
> +++ b/drivers/mtd/ubi/vtbl.c
> @@ -170,7 +170,7 @@ int ubi_vtbl_rename_volumes(struct ubi_device *ubi,
>  static int vtbl_check(const struct ubi_device *ubi,
>  		      const struct ubi_vtbl_record *vtbl)
>  {
> -	int i, n, reserved_pebs, alignment, data_pad, vol_type, name_len;
> +	int i, n, reserved_lebs, alignment, data_pad, vol_type, name_len;
>  	int upd_marker, err;
>  	uint32_t crc;
>  	const char *name;
> @@ -178,7 +178,7 @@ static int vtbl_check(const struct ubi_device *ubi,
>  	for (i = 0; i < ubi->vtbl_slots; i++) {
>  		cond_resched();
>  
> -		reserved_pebs = be32_to_cpu(vtbl[i].reserved_pebs);
> +		reserved_lebs = be32_to_cpu(vtbl[i].reserved_lebs);
>  		alignment = be32_to_cpu(vtbl[i].alignment);
>  		data_pad = be32_to_cpu(vtbl[i].data_pad);
>  		upd_marker = vtbl[i].upd_marker;
> @@ -194,7 +194,7 @@ static int vtbl_check(const struct ubi_device *ubi,
>  			return 1;
>  		}
>  
> -		if (reserved_pebs == 0) {
> +		if (reserved_lebs == 0) {
>  			if (memcmp(&vtbl[i], &empty_vtbl_record,
>  						UBI_VTBL_RECORD_SIZE)) {
>  				err = 2;
> @@ -203,7 +203,7 @@ static int vtbl_check(const struct ubi_device *ubi,
>  			continue;
>  		}
>  
> -		if (reserved_pebs < 0 || alignment < 0 || data_pad < 0 ||
> +		if (reserved_lebs < 0 || alignment < 0 || data_pad < 0 ||
>  		    name_len < 0) {
>  			err = 3;
>  			goto bad;
> @@ -237,9 +237,10 @@ static int vtbl_check(const struct ubi_device *ubi,
>  			goto bad;
>  		}
>  
> -		if (reserved_pebs > ubi->good_peb_count) {
> -			ubi_err(ubi, "too large reserved_pebs %d, good PEBs %d",
> -				reserved_pebs, ubi->good_peb_count);
> +		reserved_lebs = DIV_ROUND_UP(reserved_lebs, ubi->lebs_per_cpeb);
> +		if (reserved_lebs > ubi->good_peb_count) {
> +			ubi_err(ubi, "too large reserved_lebs %d, good PEBs %d",
> +				reserved_lebs, ubi->good_peb_count);
>  			err = 9;
>  			goto bad;
>  		}
> @@ -333,12 +334,20 @@ retry:
>  	if (err)
>  		goto write_error;
>  
> +	if (ubi->vtbl_size < ubi->leb_size) { //XXX
> +		err = ubi_io_write_data(ubi, vtbl, new_aeb->pnum,
> +					ubi->leb_size - ubi->min_io_size,
> +					ubi->min_io_size);
> +	}
> +
> +	if (err)
> +		goto write_error;
>  	/*
>  	 * And add it to the attaching information. Don't delete the old version
>  	 * of this LEB as it will be deleted and freed in 'ubi_add_to_av()'.
>  	 */
> -	err = ubi_add_to_av(ubi, ai, new_aeb->pnum, new_aeb->ec, vid_hdr, 0);
> -	kmem_cache_free(ai->aeb_slab_cache, new_aeb);
> +	list_add_tail(&new_aeb->list, &ai->used);
> +	err = ubi_add_to_av(ubi, ai, new_aeb, vid_hdr, 0, 0, true);
>  	ubi_free_vid_hdr(ubi, vid_hdr);
>  	return err;
>  
> @@ -348,10 +357,10 @@ write_error:
>  		 * Probably this physical eraseblock went bad, try to pick
>  		 * another one.
>  		 */
> -		list_add(&new_aeb->u.list, &ai->erase);
> +		list_add(&new_aeb->list, &ai->erase);
>  		goto retry;
>  	}
> -	kmem_cache_free(ai->aeb_slab_cache, new_aeb);
> +	kmem_cache_free(ai->apeb_slab_cache, new_aeb);
>  out_free:
>  	ubi_free_vid_hdr(ubi, vid_hdr);
>  	return err;
> @@ -374,7 +383,7 @@ static struct ubi_vtbl_record *process_lvol(struct ubi_device *ubi,
>  {
>  	int err;
>  	struct rb_node *rb;
> -	struct ubi_ainf_peb *aeb;
> +	struct ubi_ainf_leb *aeb;
>  	struct ubi_vtbl_record *leb[UBI_LAYOUT_VOLUME_EBS] = { NULL, NULL };
>  	int leb_corrupted[UBI_LAYOUT_VOLUME_EBS] = {1, 1};
>  
> @@ -406,15 +415,22 @@ static struct ubi_vtbl_record *process_lvol(struct ubi_device *ubi,
>  	dbg_gen("check layout volume");
>  
>  	/* Read both LEB 0 and LEB 1 into memory */
> -	ubi_rb_for_each_entry(rb, aeb, &av->root, u.rb) {
> -		leb[aeb->lnum] = vzalloc(ubi->vtbl_size);
> -		if (!leb[aeb->lnum]) {
> +	ubi_rb_for_each_entry(rb, aeb, &av->root, rb) {
> +		leb[aeb->desc.lnum] = vzalloc(ubi->vtbl_size);
> +		if (!leb[aeb->desc.lnum]) {
>  			err = -ENOMEM;
>  			goto out_free;
>  		}
>  
> -		err = ubi_io_read_data(ubi, leb[aeb->lnum], aeb->pnum, 0,
> -				       ubi->vtbl_size);
> +		if (!aeb->peb->consolidated) {
> +			err = ubi_io_read_data(ubi, leb[aeb->desc.lnum],
> +				aeb->peb->pnum, 0, ubi->vtbl_size);
> +		} else {
> +			err = ubi_io_raw_read(ubi, leb[aeb->desc.lnum],
> +				aeb->peb->pnum, ubi->leb_start + (aeb->peb_pos * ubi->leb_size),
> +				ubi->vtbl_size);
> +		}
> +
>  		if (err == UBI_IO_BITFLIPS || mtd_is_eccerr(err))
>  			/*
>  			 * Scrub the PEB later. Note, -EBADMSG indicates an
> @@ -426,7 +442,7 @@ static struct ubi_vtbl_record *process_lvol(struct ubi_device *ubi,
>  			 * aeb->scrub will be cleared in
>  			 * 'ubi_add_to_av()'.
>  			 */
> -			aeb->scrub = 1;
> +			aeb->peb->scrub = 1;
>  		else if (err)
>  			goto out_free;
>  	}
> @@ -531,21 +547,21 @@ static int init_volumes(struct ubi_device *ubi,
>  			const struct ubi_attach_info *ai,
>  			const struct ubi_vtbl_record *vtbl)
>  {
> -	int i, reserved_pebs = 0;
> +	int i, reserved_lebs = 0;
>  	struct ubi_ainf_volume *av;
>  	struct ubi_volume *vol;
>  
>  	for (i = 0; i < ubi->vtbl_slots; i++) {
>  		cond_resched();
>  
> -		if (be32_to_cpu(vtbl[i].reserved_pebs) == 0)
> +		if (be32_to_cpu(vtbl[i].reserved_lebs) == 0)
>  			continue; /* Empty record */
>  
>  		vol = kzalloc(sizeof(struct ubi_volume), GFP_KERNEL);
>  		if (!vol)
>  			return -ENOMEM;
>  
> -		vol->reserved_pebs = be32_to_cpu(vtbl[i].reserved_pebs);
> +		vol->reserved_lebs = be32_to_cpu(vtbl[i].reserved_lebs);
>  		vol->alignment = be32_to_cpu(vtbl[i].alignment);
>  		vol->data_pad = be32_to_cpu(vtbl[i].data_pad);
>  		vol->upd_marker = vtbl[i].upd_marker;
> @@ -573,14 +589,14 @@ static int init_volumes(struct ubi_device *ubi,
>  		ubi->volumes[i] = vol;
>  		ubi->vol_count += 1;
>  		vol->ubi = ubi;
> -		reserved_pebs += vol->reserved_pebs;
> +		reserved_lebs += vol->reserved_lebs;
>  
>  		/*
>  		 * In case of dynamic volume UBI knows nothing about how many
>  		 * data is stored there. So assume the whole volume is used.
>  		 */
>  		if (vol->vol_type == UBI_DYNAMIC_VOLUME) {
> -			vol->used_ebs = vol->reserved_pebs;
> +			vol->used_ebs = vol->reserved_lebs;
>  			vol->last_eb_bytes = vol->usable_leb_size;
>  			vol->used_bytes =
>  				(long long)vol->used_ebs * vol->usable_leb_size;
> @@ -624,14 +640,14 @@ static int init_volumes(struct ubi_device *ubi,
>  	if (!vol)
>  		return -ENOMEM;
>  
> -	vol->reserved_pebs = UBI_LAYOUT_VOLUME_EBS;
> +	vol->reserved_lebs = UBI_LAYOUT_VOLUME_EBS;
>  	vol->alignment = UBI_LAYOUT_VOLUME_ALIGN;
>  	vol->vol_type = UBI_DYNAMIC_VOLUME;
>  	vol->name_len = sizeof(UBI_LAYOUT_VOLUME_NAME) - 1;
>  	memcpy(vol->name, UBI_LAYOUT_VOLUME_NAME, vol->name_len + 1);
>  	vol->usable_leb_size = ubi->leb_size;
> -	vol->used_ebs = vol->reserved_pebs;
> -	vol->last_eb_bytes = vol->reserved_pebs;
> +	vol->used_ebs = vol->reserved_lebs;
> +	vol->last_eb_bytes = vol->reserved_lebs;
>  	vol->used_bytes =
>  		(long long)vol->used_ebs * (ubi->leb_size - vol->data_pad);
>  	vol->vol_id = UBI_LAYOUT_VOLUME_ID;
> @@ -639,20 +655,21 @@ static int init_volumes(struct ubi_device *ubi,
>  
>  	ubi_assert(!ubi->volumes[i]);
>  	ubi->volumes[vol_id2idx(ubi, vol->vol_id)] = vol;
> -	reserved_pebs += vol->reserved_pebs;
> +	reserved_lebs += vol->reserved_lebs;
>  	ubi->vol_count += 1;
>  	vol->ubi = ubi;
>  
> -	if (reserved_pebs > ubi->avail_pebs) {
> +	reserved_lebs = DIV_ROUND_UP(reserved_lebs, ubi->lebs_per_cpeb);
> +	if (reserved_lebs > ubi->avail_pebs) {
>  		ubi_err(ubi, "not enough PEBs, required %d, available %d",
> -			reserved_pebs, ubi->avail_pebs);
> +			reserved_lebs, ubi->avail_pebs);
>  		if (ubi->corr_peb_count)
>  			ubi_err(ubi, "%d PEBs are corrupted and not used",
>  				ubi->corr_peb_count);
>  		return -ENOSPC;
>  	}
> -	ubi->rsvd_pebs += reserved_pebs;
> -	ubi->avail_pebs -= reserved_pebs;
> +	ubi->rsvd_pebs += reserved_lebs;
> +	ubi->avail_pebs -= reserved_lebs;
>  
>  	return 0;
>  }
> @@ -670,11 +687,11 @@ static int check_av(const struct ubi_volume *vol,
>  {
>  	int err;
>  
> -	if (av->highest_lnum >= vol->reserved_pebs) {
> +	if (av->highest_lnum >= vol->reserved_lebs) {
>  		err = 1;
>  		goto bad;
>  	}
> -	if (av->leb_count > vol->reserved_pebs) {
> +	if (av->leb_count > vol->reserved_lebs) {
>  		err = 2;
>  		goto bad;
>  	}
> @@ -682,7 +699,7 @@ static int check_av(const struct ubi_volume *vol,
>  		err = 3;
>  		goto bad;
>  	}
> -	if (av->used_ebs > vol->reserved_pebs) {
> +	if (av->used_ebs > vol->reserved_lebs) {
>  		err = 4;
>  		goto bad;
>  	}
> @@ -740,7 +757,7 @@ static int check_attaching_info(const struct ubi_device *ubi,
>  			continue;
>  		}
>  
> -		if (vol->reserved_pebs == 0) {
> +		if (vol->reserved_lebs == 0) {
>  			ubi_assert(i < ubi->vtbl_slots);
>  
>  			if (!av)
> diff --git a/drivers/mtd/ubi/wl.c b/drivers/mtd/ubi/wl.c
> index bf4e6b2..ed1031d 100644
> --- a/drivers/mtd/ubi/wl.c
> +++ b/drivers/mtd/ubi/wl.c
> @@ -302,6 +302,11 @@ static struct ubi_wl_entry *find_mean_wl_entry(struct ubi_device *ubi,
>  {
>  	struct ubi_wl_entry *e, *first, *last;
>  
> +	ubi_assert(root->rb_node);
> +
> +	if (!root->rb_node)
> +		return NULL;
> +
>  	first = rb_entry(rb_first(root), struct ubi_wl_entry, u.rb);
>  	last = rb_entry(rb_last(root), struct ubi_wl_entry, u.rb);
>  
> @@ -481,9 +486,8 @@ repeat:
>  static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk,
>  			int shutdown);
>  
> -static struct ubi_work *ubi_alloc_erase_work(struct ubi_device *ubi,
> -					     struct ubi_wl_entry *e,
> -					     int torture)
> +struct ubi_work *ubi_alloc_erase_work(struct ubi_device *ubi,
> +				      struct ubi_wl_entry *e, int torture)
>  {
>  	struct ubi_work *wl_wrk;
>  
> @@ -510,11 +514,12 @@ static struct ubi_work *ubi_alloc_erase_work(struct ubi_device *ubi,
>   * failure.
>   */
>  static int schedule_erase(struct ubi_device *ubi, struct ubi_wl_entry *e,
> -			  int torture)
> +			  int torture, bool nested)
>  {
>  	struct ubi_work *wl_wrk;
>  
>  	ubi_assert(e);
> +	ubi_assert(!ubi->consolidated || !ubi->consolidated[e->pnum]);
>  
>  	dbg_wl("schedule erasure of PEB %d, EC %d, torture %d",
>  	       e->pnum, e->ec, torture);
> @@ -551,6 +556,7 @@ static int do_sync_erase(struct ubi_device *ubi, struct ubi_wl_entry *e,
>  	if (!wl_wrk)
>  		return -ENOMEM;
>  
> +	INIT_LIST_HEAD(&wl_wrk->list);
>  	wl_wrk->e = e;
>  	wl_wrk->torture = torture;
>  
> @@ -579,6 +585,7 @@ static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk,
>  	struct ubi_wl_entry *e1, *e2;
>  	struct ubi_vid_hdr *vid_hdr;
>  	int dst_leb_clean = 0;
> +	int nvidh = ubi->lebs_per_cpeb;
>  
>  	if (shutdown)
>  		return 0;
> @@ -680,7 +687,7 @@ static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk,
>  	 * which is being moved was unmapped.
>  	 */
>  
> -	err = ubi_io_read_vid_hdr(ubi, e1->pnum, vid_hdr, 0);
> +	err = ubi_io_read_vid_hdrs(ubi, e1->pnum, vid_hdr, &nvidh, 0);
>  	if (err && err != UBI_IO_BITFLIPS) {
>  		dst_leb_clean = 1;
>  		if (err == UBI_IO_FF) {
> @@ -714,7 +721,11 @@ static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk,
>  		goto out_error;
>  	}
>  
> -	err = ubi_eba_copy_leb(ubi, e1->pnum, e2->pnum, vid_hdr);
> +	if (ubi->consolidated && ubi->consolidated[e1->pnum])
> +		err = ubi_eba_copy_lebs(ubi, e1->pnum, e2->pnum, vid_hdr, nvidh);
> +	else
> +		err = ubi_eba_copy_leb(ubi, e1->pnum, e2->pnum, vid_hdr);
> +
>  	if (err) {
>  		if (err == MOVE_CANCEL_RACE) {
>  			/*
> @@ -917,7 +928,7 @@ static int ensure_wear_leveling(struct ubi_device *ubi)
>  	ubi->wl_scheduled = 1;
>  	spin_unlock(&ubi->wl_lock);
>  
> -	wrk = kmalloc(sizeof(struct ubi_work), GFP_NOFS);
> +	wrk = ubi_alloc_work(ubi);
>  	if (!wrk) {
>  		err = -ENOMEM;
>  		goto out_cancel;
> @@ -982,7 +993,7 @@ static int __erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk)
>  		int err1;
>  
>  		/* Re-schedule the LEB for erasure */
> -		err1 = schedule_erase(ubi, e, 0);
> +		err1 = schedule_erase(ubi, e, 0, true);
>  		if (err1) {
>  			wl_entry_destroy(ubi, e);
>  			err = err1;
> @@ -1092,10 +1103,12 @@ int ubi_wl_put_peb(struct ubi_device *ubi, int pnum, int torture)
>  {
>  	int err;
>  	struct ubi_wl_entry *e;
> +	struct ubi_work *wrk;
>  
>  	dbg_wl("PEB %d", pnum);
>  	ubi_assert(pnum >= 0);
>  	ubi_assert(pnum < ubi->peb_count);
> +	ubi_assert(!ubi->consolidated || !ubi->consolidated[pnum]);
>  
>  	down_read(&ubi->fm_protect);
>  
> @@ -1158,15 +1171,20 @@ retry:
>  	}
>  	spin_unlock(&ubi->wl_lock);
>  
> -	err = schedule_erase(ubi, e, torture);
> -	if (err) {
> +	wrk = ubi_alloc_erase_work(ubi, e, torture);
> +	if (!wrk) {
>  		spin_lock(&ubi->wl_lock);
>  		wl_tree_add(e, &ubi->used);
>  		spin_unlock(&ubi->wl_lock);
>  	}
> -
>  	up_read(&ubi->fm_protect);
> -	return err;
> +
> +	if (!wrk)
> +		return -ENOMEM;
> +
> +	ubi_schedule_work(ubi, wrk);
> +
> +	return 0;
>  }
>  
>  /**
> @@ -1277,8 +1295,10 @@ int ubi_wl_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
>  	int err, i, reserved_pebs, found_pebs = 0;
>  	struct rb_node *rb1, *rb2;
>  	struct ubi_ainf_volume *av;
> -	struct ubi_ainf_peb *aeb, *tmp;
> +	struct ubi_ainf_leb *leb;
> +	struct ubi_ainf_peb *peb, *tmp;
>  	struct ubi_wl_entry *e;
> +	struct ubi_leb_desc *clebs;
>  
>  	ubi->used = ubi->erroneous = ubi->free = ubi->scrub = RB_ROOT;
>  	spin_lock_init(&ubi->wl_lock);
> @@ -1294,22 +1314,34 @@ int ubi_wl_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
>  	if (!ubi->lookuptbl)
>  		return err;
>  
> +	if (ubi->lebs_per_cpeb > 1) {
> +		ubi->consolidated = kzalloc(ubi->peb_count * sizeof(void *),
> +					    GFP_KERNEL);
> +		if (!ubi->consolidated) {
> +			kfree(ubi->lookuptbl);
> +			return err;
> +		}
> +	}
> +
>  	for (i = 0; i < UBI_PROT_QUEUE_LEN; i++)
>  		INIT_LIST_HEAD(&ubi->pq[i]);
>  	ubi->pq_head = 0;
>  
>  	ubi->free_count = 0;
> -	list_for_each_entry_safe(aeb, tmp, &ai->erase, u.list) {
> +	list_for_each_entry_safe(peb, tmp, &ai->erase, list) {
>  		cond_resched();
>  
> +		if (ubi->lookuptbl[peb->pnum])
> +			continue;
> +
>  		e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL);
>  		if (!e)
>  			goto out_free;
>  
> -		e->pnum = aeb->pnum;
> -		e->ec = aeb->ec;
> +		e->pnum = peb->pnum;
> +		e->ec = peb->ec;
>  		ubi->lookuptbl[e->pnum] = e;
> -		if (schedule_erase(ubi, e, 0)) {
> +		if (schedule_erase(ubi, e, 0, false)) {
>  			wl_entry_destroy(ubi, e);
>  			goto out_free;
>  		}
> @@ -1317,15 +1349,18 @@ int ubi_wl_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
>  		found_pebs++;
>  	}
>  
> -	list_for_each_entry(aeb, &ai->free, u.list) {
> +	list_for_each_entry(peb, &ai->free, list) {
>  		cond_resched();
>  
> +		if (ubi->lookuptbl[peb->pnum])
> +			continue;
> +
>  		e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL);
>  		if (!e)
>  			goto out_free;
>  
> -		e->pnum = aeb->pnum;
> -		e->ec = aeb->ec;
> +		e->pnum = peb->pnum;
> +		e->ec = peb->ec;
>  		ubi_assert(e->ec >= 0);
>  
>  		wl_tree_add(e, &ubi->free);
> @@ -1336,29 +1371,54 @@ int ubi_wl_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
>  		found_pebs++;
>  	}
>  
> -	ubi_rb_for_each_entry(rb1, av, &ai->volumes, rb) {
> -		ubi_rb_for_each_entry(rb2, aeb, &av->root, u.rb) {
> -			cond_resched();
> +	list_for_each_entry(peb, &ai->used, list) {
> +		e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL);
> +		if (!e)
> +			goto out_free;
>  
> -			e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL);
> -			if (!e)
> -				goto out_free;
> +		e->pnum = peb->pnum;
> +		e->ec = peb->ec;
> +		ubi->lookuptbl[e->pnum] = e;
>  
> -			e->pnum = aeb->pnum;
> -			e->ec = aeb->ec;
> -			ubi->lookuptbl[e->pnum] = e;
> +		if (!peb->scrub) {
> +			dbg_wl("add PEB %d EC %d to the used tree",
> +			       e->pnum, e->ec);
> +			wl_tree_add(e, &ubi->used);
> +		} else {
> +			dbg_wl("add PEB %d EC %d to the scrub tree",
> +			       e->pnum, e->ec);
> +			wl_tree_add(e, &ubi->scrub);
> +		}
> +
> +		if (peb->consolidated) {
> +			int i;
> +
> +			clebs = kmalloc(sizeof(*clebs) *
> +					ubi->lebs_per_cpeb,
> +					GFP_KERNEL);
> +			if (!clebs)
> +				goto out_free;
>  
> -			if (!aeb->scrub) {
> -				dbg_wl("add PEB %d EC %d to the used tree",
> -				       e->pnum, e->ec);
> -				wl_tree_add(e, &ubi->used);
> -			} else {
> -				dbg_wl("add PEB %d EC %d to the scrub tree",
> -				       e->pnum, e->ec);
> -				wl_tree_add(e, &ubi->scrub);
> +			for (i = 0; i < ubi->lebs_per_cpeb; i++) {
> +				clebs[i].lnum = -1;
> +				clebs[i].vol_id = -1;
>  			}
>  
> -			found_pebs++;
> +			ubi->consolidated[peb->pnum] = clebs;
> +		}
> +
> +		found_pebs++;
> +	}
> +
> +	ubi_rb_for_each_entry(rb1, av, &ai->volumes, rb) {
> +		ubi_rb_for_each_entry(rb2, leb, &av->root, rb) {
> +			cond_resched();
> +
> +			if (ubi->lebs_per_cpeb > 1) {
> +				clebs = ubi->consolidated[leb->peb->pnum];
> +				if (clebs)
> +					clebs[leb->peb_pos] = leb->desc;
> +			}
>  		}
>  	}
>  
> @@ -1403,6 +1463,7 @@ out_free:
>  	tree_destroy(ubi, &ubi->used);
>  	tree_destroy(ubi, &ubi->free);
>  	tree_destroy(ubi, &ubi->scrub);
> +	kfree(ubi->consolidated);
>  	kfree(ubi->lookuptbl);
>  	return err;
>  }
> @@ -1439,6 +1500,7 @@ void ubi_wl_close(struct ubi_device *ubi)
>  	tree_destroy(ubi, &ubi->free);
>  	tree_destroy(ubi, &ubi->scrub);
>  	kfree(ubi->lookuptbl);
> +	kfree(ubi->consolidated);
>  }
>  
>  /**
> @@ -1536,11 +1598,24 @@ static int self_check_in_pq(const struct ubi_device *ubi,
>  	dump_stack();
>  	return -EINVAL;
>  }
> +
> +static bool enough_free_pebs(struct ubi_device *ubi)
> +{
> +	/*
> +	 * Hold back one PEB for the producing case,
> +	 * currently only for consolidation.
> +	 */
> +	return ubi->free_count > UBI_CONSO_RESERVED_PEBS;
> +}
> +
>  #ifndef CONFIG_MTD_UBI_FASTMAP
>  static struct ubi_wl_entry *get_peb_for_wl(struct ubi_device *ubi)
>  {
>  	struct ubi_wl_entry *e;
>  
> +	if (!enough_free_pebs(ubi))
> +		return NULL;
> +
>  	e = find_wl_entry(ubi, &ubi->free, WL_FREE_MAX_DIFF);
>  	self_check_in_wl_tree(ubi, e, &ubi->free);
>  	ubi->free_count--;
> @@ -1563,9 +1638,11 @@ static int produce_free_peb(struct ubi_device *ubi)
>  {
>  	ubi_assert(spin_is_locked(&ubi->wl_lock));
>  
> -	while (!ubi->free.rb_node) {
> +	while (!enough_free_pebs(ubi)) {
>  		spin_unlock(&ubi->wl_lock);
>  
> +		ubi_eba_consolidate(ubi);
> +
>  		dbg_wl("do one work synchronously");
>  		if (!ubi_work_join_one(ubi)) {
>  			spin_lock(&ubi->wl_lock);
> @@ -1582,41 +1659,51 @@ static int produce_free_peb(struct ubi_device *ubi)
>  /**
>   * ubi_wl_get_peb - get a physical eraseblock.
>   * @ubi: UBI device description object
> + * @producing: true if this function is being called from a context
> + * which is trying to produce more free PEBs but needs a new one to
> + * achieve that. i.e. consolidatation work.
>   *
>   * This function returns a physical eraseblock in case of success and a
>   * negative error code in case of failure.
>   * Returns with ubi->fm_eba_sem held in read mode!
>   */
> -int ubi_wl_get_peb(struct ubi_device *ubi)
> +int ubi_wl_get_peb(struct ubi_device *ubi, bool producing)
>  {
> -	int err;
> +	int err = 0;
>  	struct ubi_wl_entry *e;
>  
>  retry:
>  	down_read(&ubi->fm_eba_sem);
>  	spin_lock(&ubi->wl_lock);
> -	if (!ubi->free.rb_node) {
> -		if (ubi->works_count == 0) {
> -			ubi_err(ubi, "no free eraseblocks");
> -			ubi_assert(list_empty(&ubi->works));
> -			spin_unlock(&ubi->wl_lock);
> -			return -ENOSPC;
> -		}
>  
> +	if (!enough_free_pebs(ubi) && !producing) {
>  		err = produce_free_peb(ubi);
>  		if (err < 0) {
> +			ubi_err(ubi, "unable to produce free eraseblocks: %i", err);
>  			spin_unlock(&ubi->wl_lock);
>  			return err;
>  		}
>  		spin_unlock(&ubi->wl_lock);
>  		up_read(&ubi->fm_eba_sem);
>  		goto retry;
> -
>  	}
> +	else if (!ubi->free_count && producing) {
> +		ubi_err(ubi, "no free eraseblocks in producing case");
> +		ubi_assert(0);
> +		spin_unlock(&ubi->wl_lock);
> +		return -ENOSPC;
> +	}
> +
>  	e = wl_get_wle(ubi);
> -	prot_queue_add(ubi, e);
> +	if (e)
> +		prot_queue_add(ubi, e);
> +	else
> +		err = -ENOSPC;
>  	spin_unlock(&ubi->wl_lock);
>  
> +	if (err)
> +		return err;
> +
>  	err = ubi_self_check_all_ff(ubi, e->pnum, ubi->vid_hdr_aloffset,
>  				    ubi->peb_size - ubi->vid_hdr_aloffset);
>  	if (err) {



-- 
Boris Brezillon, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

RE: [RFC] UBI: MLC Support v0

[Bean Huo (beanhuo)]

> On Mon, 30 May 2016 14:04:21 +0200
> Richard Weinberger <richard@nod.at> wrote:
> 
> > as announced in [1] I'd like to submit the very first patch series to
> > have MLC support in UBI (and therefore also UBIFS).
> > The code has not changed much since the announcement but now the
> > patches are in a more reviewable shape. Still not mainline ready but better.
> > I'm sending this now as I'm on vacation until June 15th and hope that
> > we get some reviews on the concept.
> > The series consists of three parts:
> > 	a) Generic UBI rework to make the MLC implementation possible.
> > 	b) SLC emulation mode to access only lower pages on MLC NAND.
> > 	c) LEB consolidation to exploit MLC NAND's higher pages.
> >
> > [1]
> > http://lists.infradead.org/pipermail/linux-mtd/2016-April/067322.html
> 
> Artem, I know you're quite busy, but that would be great to have your
> feedback. Not asking for a detailed review, but at least some comments on
> the general approach.
> 
> Bean, I'd like to have your feedback too. You proposed a solution to address
> the same problem a while ago, and your feedback on this solution would be
> valuable.
> 
Boris 
Sorry later reply for this. this month, I am busy for my immigration.
Now I am working in Munich, today I am just having a f2f meeting with Richard.
To be honest, I really interested in this solution. from this week, I will have more free time
And do some testing and review. Thanks. 

> Thanks,
> 
> Boris
> 
> >
> > Thanks,
> > //richard
> >
> > ---
> >
> > David Gstir (1):
> >       ubi: Add debugfs knob to force LEB consolidation
> >
> > Richard Weinberger (12):
> >       ubi: Undo "UBI: modify ubi_wl_flush function to clear work queue for a
> lnum"
> >       ubi: Rework UBI worker
> >       ubi: auto re-size after UBI thread is ready
> >       ubi: Kill ubi->alc_mutex
> >       ubi: Get rid of __schedule_ubi_work()
> >       ubi: Remove tst_disable_bgt debugfs knob
> >       ubi: Move work related functions to work.c
> >       ubi: Remove lnum and vol_id from erase work
> >       ubi: Remove usless debug info from wear_leveling_worker()
> >       ubi: SLC mode
> >       ubi: LEB consolidation
> >       ubi: Add debugfs knob to trigger LEB consolidation
> >
> >  drivers/mtd/ubi/Kconfig       |   4 +
> >  drivers/mtd/ubi/Makefile      |   3 +-
> >  drivers/mtd/ubi/attach.c      | 407 ++++++++++++++++++--------------
> >  drivers/mtd/ubi/build.c       |  41 ++--
> >  drivers/mtd/ubi/cdev.c        |  14 +-
> >  drivers/mtd/ubi/consolidate.c | 516
> ++++++++++++++++++++++++++++++++++++++++
> >  drivers/mtd/ubi/debug.c       |  49 ++--
> >  drivers/mtd/ubi/debug.h       |  19 +-
> >  drivers/mtd/ubi/eba.c         | 429 ++++++++++++++++++++++++++++------
> >  drivers/mtd/ubi/fastmap-wl.c  |  33 ++-
> >  drivers/mtd/ubi/fastmap.c     | 265 ++++++++++++++-------
> >  drivers/mtd/ubi/io.c          | 426 +++++++++++++++++++++++++--------
> >  drivers/mtd/ubi/kapi.c        |  43 +---
> >  drivers/mtd/ubi/ubi-media.h   |  18 +-
> >  drivers/mtd/ubi/ubi.h         | 221 ++++++++++++++----
> >  drivers/mtd/ubi/upd.c         |   7 +-
> >  drivers/mtd/ubi/vmt.c         | 106 +++++----
> >  drivers/mtd/ubi/vtbl.c        |  87 ++++---
> >  drivers/mtd/ubi/wl.c          | 531 +++++++++++++++---------------------------
> >  drivers/mtd/ubi/work.c        | 332 ++++++++++++++++++++++++++
> >  include/linux/mtd/ubi.h       |   1 -
> >  21 files changed, 2554 insertions(+), 998 deletions(-)
> >
> > ______________________________________________________
> > Linux MTD discussion mailing list
> > http://lists.infradead.org/mailman/listinfo/linux-mtd/
> 
> 
> 
> --
> Boris Brezillon, Free Electrons
> Embedded Linux and Kernel engineering
> http://free-electrons.com