[PATCH 0/15 V10] Work to enable SmartMedia/xD support in mtd

[PATCH 0/15 V10] Work to enable SmartMedia/xD support in mtd

[Maxim Levitsky]

Changes:

* Rebased the series against patches from Ben Hutchings.
Now it applies cleanly on top of
git://git.infradead.org/users/dedekind/l2-mtd-2.6.git

* Fixed another blktrans hotplug problem. 
If you remove that card while it is used, then add it while,
the user still holds the block device open, block device wasn't accesable

* Fixed minor memory leak in mtdblock (unrelated to the patchset)

Best regards,
	Maxim Levitsky

[PATCH 0/15 V10] Work to enable SmartMedia/xD support in mtd

[Maxim Levitsky]

Changes:

* Rebased the series against patches from Ben Hutchings.
Now it applies cleanly on top of
git://git.infradead.org/users/dedekind/l2-mtd-2.6.git

* Fixed another blktrans hotplug problem. 
If you remove that card while it is used, then add it while,
the user still holds the block device open, block device wasn't accesable

* Fixed minor memory leak in mtdblock (unrelated to the patchset)

Best regards,
	Maxim Levitsky

[PATCH 01/15] MTD: create unlocked versions of {get,put}_mtd_device

[Maxim Levitsky]

Use these only if you know that you already hold mtd_table_mutex

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/mtdcore.c   |   60 ++++++++++++++++++++++++++++++----------------
 include/linux/mtd/mtd.h |    3 +-
 2 files changed, 41 insertions(+), 22 deletions(-)

diff --git a/drivers/mtd/mtdcore.c b/drivers/mtd/mtdcore.c
index b3b98d1..67669a7 100644
--- a/drivers/mtd/mtdcore.c
+++ b/drivers/mtd/mtdcore.c
@@ -463,27 +463,38 @@ struct mtd_info *get_mtd_device(struct mtd_info *mtd, int num)
 			ret = NULL;
 	}
 
-	if (!ret)
-		goto out_unlock;
-
-	if (!try_module_get(ret->owner))
-		goto out_unlock;
-
-	if (ret->get_device) {
-		err = ret->get_device(ret);
-		if (err)
-			goto out_put;
+	if (!ret) {
+		ret = ERR_PTR(err);
+		goto out;
 	}
 
-	ret->usecount++;
+	err = __get_mtd_device(ret);
+	if (err)
+		ret = ERR_PTR(err);
+out:
 	mutex_unlock(&mtd_table_mutex);
 	return ret;
+}
 
-out_put:
-	module_put(ret->owner);
-out_unlock:
-	mutex_unlock(&mtd_table_mutex);
-	return ERR_PTR(err);
+
+int __get_mtd_device(struct mtd_info *mtd)
+{
+	int err;
+
+	if (!try_module_get(mtd->owner))
+		return -ENODEV;
+
+	if (mtd->get_device) {
+
+		err = mtd->get_device(mtd);
+
+		if (err) {
+			module_put(mtd->owner);
+			return err;
+		}
+	}
+	mtd->usecount++;
+	return 0;
 }
 
 /**
@@ -534,14 +545,19 @@ out_unlock:
 
 void put_mtd_device(struct mtd_info *mtd)
 {
-	int c;
-
 	mutex_lock(&mtd_table_mutex);
-	c = --mtd->usecount;
+	__put_mtd_device(mtd);
+	mutex_unlock(&mtd_table_mutex);
+
+}
+
+void __put_mtd_device(struct mtd_info *mtd)
+{
+	--mtd->usecount;
+	BUG_ON(mtd->usecount < 0);
+
 	if (mtd->put_device)
 		mtd->put_device(mtd);
-	mutex_unlock(&mtd_table_mutex);
-	BUG_ON(c < 0);
 
 	module_put(mtd->owner);
 }
@@ -579,7 +595,9 @@ EXPORT_SYMBOL_GPL(add_mtd_device);
 EXPORT_SYMBOL_GPL(del_mtd_device);
 EXPORT_SYMBOL_GPL(get_mtd_device);
 EXPORT_SYMBOL_GPL(get_mtd_device_nm);
+EXPORT_SYMBOL_GPL(__get_mtd_device);
 EXPORT_SYMBOL_GPL(put_mtd_device);
+EXPORT_SYMBOL_GPL(__put_mtd_device);
 EXPORT_SYMBOL_GPL(register_mtd_user);
 EXPORT_SYMBOL_GPL(unregister_mtd_user);
 EXPORT_SYMBOL_GPL(default_mtd_writev);
diff --git a/include/linux/mtd/mtd.h b/include/linux/mtd/mtd.h
index ba53ecc..11d8e68 100644
--- a/include/linux/mtd/mtd.h
+++ b/include/linux/mtd/mtd.h
@@ -289,8 +289,9 @@ extern int add_mtd_device(struct mtd_info *mtd);
 extern int del_mtd_device (struct mtd_info *mtd);
 
 extern struct mtd_info *get_mtd_device(struct mtd_info *mtd, int num);
+extern int __get_mtd_device(struct mtd_info *mtd);
+extern void __put_mtd_device(struct mtd_info *mtd);
 extern struct mtd_info *get_mtd_device_nm(const char *name);
-
 extern void put_mtd_device(struct mtd_info *mtd);
 
 
-- 
1.6.3.3

[PATCH 01/15] MTD: create unlocked versions of {get,put}_mtd_device

[Maxim Levitsky]

Use these only if you know that you already hold mtd_table_mutex

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/mtdcore.c   |   60 ++++++++++++++++++++++++++++++----------------
 include/linux/mtd/mtd.h |    3 +-
 2 files changed, 41 insertions(+), 22 deletions(-)

diff --git a/drivers/mtd/mtdcore.c b/drivers/mtd/mtdcore.c
index b3b98d1..67669a7 100644
--- a/drivers/mtd/mtdcore.c
+++ b/drivers/mtd/mtdcore.c
@@ -463,27 +463,38 @@ struct mtd_info *get_mtd_device(struct mtd_info *mtd, int num)
 			ret = NULL;
 	}
 
-	if (!ret)
-		goto out_unlock;
-
-	if (!try_module_get(ret->owner))
-		goto out_unlock;
-
-	if (ret->get_device) {
-		err = ret->get_device(ret);
-		if (err)
-			goto out_put;
+	if (!ret) {
+		ret = ERR_PTR(err);
+		goto out;
 	}
 
-	ret->usecount++;
+	err = __get_mtd_device(ret);
+	if (err)
+		ret = ERR_PTR(err);
+out:
 	mutex_unlock(&mtd_table_mutex);
 	return ret;
+}
 
-out_put:
-	module_put(ret->owner);
-out_unlock:
-	mutex_unlock(&mtd_table_mutex);
-	return ERR_PTR(err);
+
+int __get_mtd_device(struct mtd_info *mtd)
+{
+	int err;
+
+	if (!try_module_get(mtd->owner))
+		return -ENODEV;
+
+	if (mtd->get_device) {
+
+		err = mtd->get_device(mtd);
+
+		if (err) {
+			module_put(mtd->owner);
+			return err;
+		}
+	}
+	mtd->usecount++;
+	return 0;
 }
 
 /**
@@ -534,14 +545,19 @@ out_unlock:
 
 void put_mtd_device(struct mtd_info *mtd)
 {
-	int c;
-
 	mutex_lock(&mtd_table_mutex);
-	c = --mtd->usecount;
+	__put_mtd_device(mtd);
+	mutex_unlock(&mtd_table_mutex);
+
+}
+
+void __put_mtd_device(struct mtd_info *mtd)
+{
+	--mtd->usecount;
+	BUG_ON(mtd->usecount < 0);
+
 	if (mtd->put_device)
 		mtd->put_device(mtd);
-	mutex_unlock(&mtd_table_mutex);
-	BUG_ON(c < 0);
 
 	module_put(mtd->owner);
 }
@@ -579,7 +595,9 @@ EXPORT_SYMBOL_GPL(add_mtd_device);
 EXPORT_SYMBOL_GPL(del_mtd_device);
 EXPORT_SYMBOL_GPL(get_mtd_device);
 EXPORT_SYMBOL_GPL(get_mtd_device_nm);
+EXPORT_SYMBOL_GPL(__get_mtd_device);
 EXPORT_SYMBOL_GPL(put_mtd_device);
+EXPORT_SYMBOL_GPL(__put_mtd_device);
 EXPORT_SYMBOL_GPL(register_mtd_user);
 EXPORT_SYMBOL_GPL(unregister_mtd_user);
 EXPORT_SYMBOL_GPL(default_mtd_writev);
diff --git a/include/linux/mtd/mtd.h b/include/linux/mtd/mtd.h
index ba53ecc..11d8e68 100644
--- a/include/linux/mtd/mtd.h
+++ b/include/linux/mtd/mtd.h
@@ -289,8 +289,9 @@ extern int add_mtd_device(struct mtd_info *mtd);
 extern int del_mtd_device (struct mtd_info *mtd);
 
 extern struct mtd_info *get_mtd_device(struct mtd_info *mtd, int num);
+extern int __get_mtd_device(struct mtd_info *mtd);
+extern void __put_mtd_device(struct mtd_info *mtd);
 extern struct mtd_info *get_mtd_device_nm(const char *name);
-
 extern void put_mtd_device(struct mtd_info *mtd);
 
 
-- 
1.6.3.3

[PATCH 02/15] blktrans: remove mtd_blkcore_priv and switch to per device queue and thread

[Maxim Levitsky]

This is the biggest change. To make hotplug possible, and this layer clean,
the mtd_blktrans_dev now contains everything for a single mtd block translation device.
Also removed some very old leftovers

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/mtd_blkdevs.c    |  124 ++++++++++++++++++++----------------------
 include/linux/mtd/blktrans.h |   10 ++--
 2 files changed, 63 insertions(+), 71 deletions(-)

diff --git a/drivers/mtd/mtd_blkdevs.c b/drivers/mtd/mtd_blkdevs.c
index 2f8c202..6a57262 100644
--- a/drivers/mtd/mtd_blkdevs.c
+++ b/drivers/mtd/mtd_blkdevs.c
@@ -14,7 +14,6 @@
 #include <linux/mtd/mtd.h>
 #include <linux/blkdev.h>
 #include <linux/blkpg.h>
-#include <linux/freezer.h>
 #include <linux/spinlock.h>
 #include <linux/hdreg.h>
 #include <linux/init.h>
@@ -26,11 +25,6 @@
 
 static LIST_HEAD(blktrans_majors);
 
-struct mtd_blkcore_priv {
-	struct task_struct *thread;
-	struct request_queue *rq;
-	spinlock_t queue_lock;
-};
 
 static int do_blktrans_request(struct mtd_blktrans_ops *tr,
 			       struct mtd_blktrans_dev *dev,
@@ -61,7 +55,6 @@ static int do_blktrans_request(struct mtd_blktrans_ops *tr,
 				return -EIO;
 		rq_flush_dcache_pages(req);
 		return 0;
-
 	case WRITE:
 		if (!tr->writesect)
 			return -EIO;
@@ -71,7 +64,6 @@ static int do_blktrans_request(struct mtd_blktrans_ops *tr,
 			if (tr->writesect(dev, block, buf))
 				return -EIO;
 		return 0;
-
 	default:
 		printk(KERN_NOTICE "Unknown request %u\n", rq_data_dir(req));
 		return -EIO;
@@ -80,14 +72,13 @@ static int do_blktrans_request(struct mtd_blktrans_ops *tr,
 
 static int mtd_blktrans_thread(void *arg)
 {
-	struct mtd_blktrans_ops *tr = arg;
-	struct request_queue *rq = tr->blkcore_priv->rq;
+	struct mtd_blktrans_dev *dev = arg;
+	struct request_queue *rq = dev->rq;
 	struct request *req = NULL;
 
 	spin_lock_irq(rq->queue_lock);
 
 	while (!kthread_should_stop()) {
-		struct mtd_blktrans_dev *dev;
 		int res;
 
 		if (!req && !(req = blk_fetch_request(rq))) {
@@ -98,13 +89,10 @@ static int mtd_blktrans_thread(void *arg)
 			continue;
 		}
 
-		dev = req->rq_disk->private_data;
-		tr = dev->tr;
-
 		spin_unlock_irq(rq->queue_lock);
 
 		mutex_lock(&dev->lock);
-		res = do_blktrans_request(tr, dev, req);
+		res = do_blktrans_request(dev->tr, dev, req);
 		mutex_unlock(&dev->lock);
 
 		spin_lock_irq(rq->queue_lock);
@@ -123,8 +111,8 @@ static int mtd_blktrans_thread(void *arg)
 
 static void mtd_blktrans_request(struct request_queue *rq)
 {
-	struct mtd_blktrans_ops *tr = rq->queuedata;
-	wake_up_process(tr->blkcore_priv->thread);
+	struct mtd_blktrans_dev *dev = rq->queuedata;
+	wake_up_process(dev->thread);
 }
 
 
@@ -214,6 +202,7 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new)
 	struct mtd_blktrans_dev *d;
 	int last_devnum = -1;
 	struct gendisk *gd;
+	int ret;
 
 	if (mutex_trylock(&mtd_table_mutex)) {
 		mutex_unlock(&mtd_table_mutex);
@@ -239,6 +228,8 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new)
 		}
 		last_devnum = d->devnum;
 	}
+
+	ret = -EBUSY;
 	if (new->devnum == -1)
 		new->devnum = last_devnum+1;
 
@@ -247,7 +238,7 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new)
 	 * with this number. */
 	if (new->devnum > (MINORMASK >> tr->part_bits) ||
 	    (tr->part_bits && new->devnum >= 27 * 26))
-		return -EBUSY;
+		goto error1;
 
 	list_add_tail(&new->list, &tr->devs);
  added:
@@ -255,11 +246,16 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new)
 	if (!tr->writesect)
 		new->readonly = 1;
 
+
+	/* Create gendisk */
+	ret = -ENOMEM;
 	gd = alloc_disk(1 << tr->part_bits);
-	if (!gd) {
-		list_del(&new->list);
-		return -ENOMEM;
-	}
+
+	if (!gd)
+		goto error2;
+
+	new->disk = gd;
+	gd->private_data = new;
 	gd->major = tr->major;
 	gd->first_minor = (new->devnum) << tr->part_bits;
 	gd->fops = &mtd_blktrans_ops;
@@ -277,21 +273,49 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new)
 		snprintf(gd->disk_name, sizeof(gd->disk_name),
 			 "%s%d", tr->name, new->devnum);
 
-	/* 2.5 has capacity in units of 512 bytes while still
-	   having BLOCK_SIZE_BITS set to 10. Just to keep us amused. */
 	set_capacity(gd, (new->size * tr->blksize) >> 9);
 
-	gd->private_data = new;
-	new->blkcore_priv = gd;
-	gd->queue = tr->blkcore_priv->rq;
+
+	/* Create the request queue */
+	spin_lock_init(&new->queue_lock);
+	new->rq = blk_init_queue(mtd_blktrans_request, &new->queue_lock);
+
+	if (!new->rq)
+		goto error3;
+
+	new->rq->queuedata = new;
+	blk_queue_logical_block_size(new->rq, tr->blksize);
+
+	if (tr->discard)
+		queue_flag_set_unlocked(QUEUE_FLAG_DISCARD,
+					new->rq);
+
+	gd->queue = new->rq;
+
+	/* Create processing thread */
+	/* TODO: workqueue ? */
+	new->thread = kthread_run(mtd_blktrans_thread, new,
+			"%s%d", tr->name, new->mtd->index);
+	if (IS_ERR(new->thread)) {
+		ret = PTR_ERR(new->thread);
+		goto error4;
+	}
 	gd->driverfs_dev = &new->mtd->dev;
 
 	if (new->readonly)
 		set_disk_ro(gd, 1);
 
 	add_disk(gd);
-
 	return 0;
+error4:
+	blk_cleanup_queue(new->rq);
+error3:
+	put_disk(new->disk);
+error2:
+	list_del(&new->list);
+error1:
+	kfree(new);
+	return ret;
 }
 
 int del_mtd_blktrans_dev(struct mtd_blktrans_dev *old)
@@ -303,9 +327,13 @@ int del_mtd_blktrans_dev(struct mtd_blktrans_dev *old)
 
 	list_del(&old->list);
 
-	del_gendisk(old->blkcore_priv);
-	put_disk(old->blkcore_priv);
+	/* stop new requests to arrive */
+	del_gendisk(old->disk);
 
+	/* Stop the thread */
+	kthread_stop(old->thread);
+
+	blk_cleanup_queue(old->rq);
 	return 0;
 }
 
@@ -347,9 +375,6 @@ int register_mtd_blktrans(struct mtd_blktrans_ops *tr)
 	if (!blktrans_notifier.list.next)
 		register_mtd_user(&blktrans_notifier);
 
-	tr->blkcore_priv = kzalloc(sizeof(*tr->blkcore_priv), GFP_KERNEL);
-	if (!tr->blkcore_priv)
-		return -ENOMEM;
 
 	mutex_lock(&mtd_table_mutex);
 
@@ -357,39 +382,12 @@ int register_mtd_blktrans(struct mtd_blktrans_ops *tr)
 	if (ret) {
 		printk(KERN_WARNING "Unable to register %s block device on major %d: %d\n",
 		       tr->name, tr->major, ret);
-		kfree(tr->blkcore_priv);
 		mutex_unlock(&mtd_table_mutex);
 		return ret;
 	}
-	spin_lock_init(&tr->blkcore_priv->queue_lock);
-
-	tr->blkcore_priv->rq = blk_init_queue(mtd_blktrans_request, &tr->blkcore_priv->queue_lock);
-	if (!tr->blkcore_priv->rq) {
-		unregister_blkdev(tr->major, tr->name);
-		kfree(tr->blkcore_priv);
-		mutex_unlock(&mtd_table_mutex);
-		return -ENOMEM;
-	}
-
-	tr->blkcore_priv->rq->queuedata = tr;
-	blk_queue_logical_block_size(tr->blkcore_priv->rq, tr->blksize);
-	if (tr->discard)
-		queue_flag_set_unlocked(QUEUE_FLAG_DISCARD,
-					tr->blkcore_priv->rq);
 
 	tr->blkshift = ffs(tr->blksize) - 1;
 
-	tr->blkcore_priv->thread = kthread_run(mtd_blktrans_thread, tr,
-			"%sd", tr->name);
-	if (IS_ERR(tr->blkcore_priv->thread)) {
-		ret = PTR_ERR(tr->blkcore_priv->thread);
-		blk_cleanup_queue(tr->blkcore_priv->rq);
-		unregister_blkdev(tr->major, tr->name);
-		kfree(tr->blkcore_priv);
-		mutex_unlock(&mtd_table_mutex);
-		return ret;
-	}
-
 	INIT_LIST_HEAD(&tr->devs);
 	list_add(&tr->list, &blktrans_majors);
 
@@ -408,8 +406,6 @@ int deregister_mtd_blktrans(struct mtd_blktrans_ops *tr)
 
 	mutex_lock(&mtd_table_mutex);
 
-	/* Clean up the kernel thread */
-	kthread_stop(tr->blkcore_priv->thread);
 
 	/* Remove it from the list of active majors */
 	list_del(&tr->list);
@@ -417,13 +413,9 @@ int deregister_mtd_blktrans(struct mtd_blktrans_ops *tr)
 	list_for_each_entry_safe(dev, next, &tr->devs, list)
 		tr->remove_dev(dev);
 
-	blk_cleanup_queue(tr->blkcore_priv->rq);
 	unregister_blkdev(tr->major, tr->name);
-
 	mutex_unlock(&mtd_table_mutex);
 
-	kfree(tr->blkcore_priv);
-
 	BUG_ON(!list_empty(&tr->devs));
 	return 0;
 }
diff --git a/include/linux/mtd/blktrans.h b/include/linux/mtd/blktrans.h
index 8b4aa05..a4b3928 100644
--- a/include/linux/mtd/blktrans.h
+++ b/include/linux/mtd/blktrans.h
@@ -24,11 +24,13 @@ struct mtd_blktrans_dev {
 	int devnum;
 	unsigned long size;
 	int readonly;
-	void *blkcore_priv; /* gendisk in 2.5, devfs_handle in 2.4 */
+	struct gendisk *disk;
+	struct task_struct *thread;
+	struct request_queue *rq;
+	spinlock_t queue_lock;
+	void *priv;
 };
 
-struct blkcore_priv; /* Differs for 2.4 and 2.5 kernels; private */
-
 struct mtd_blktrans_ops {
 	char *name;
 	int major;
@@ -60,8 +62,6 @@ struct mtd_blktrans_ops {
 	struct list_head devs;
 	struct list_head list;
 	struct module *owner;
-
-	struct mtd_blkcore_priv *blkcore_priv;
 };
 
 extern int register_mtd_blktrans(struct mtd_blktrans_ops *tr);
-- 
1.6.3.3

[PATCH 02/15] blktrans: remove mtd_blkcore_priv and switch to per device queue and thread

[Maxim Levitsky]

This is the biggest change. To make hotplug possible, and this layer clean,
the mtd_blktrans_dev now contains everything for a single mtd block translation device.
Also removed some very old leftovers

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/mtd_blkdevs.c    |  124 ++++++++++++++++++++----------------------
 include/linux/mtd/blktrans.h |   10 ++--
 2 files changed, 63 insertions(+), 71 deletions(-)

diff --git a/drivers/mtd/mtd_blkdevs.c b/drivers/mtd/mtd_blkdevs.c
index 2f8c202..6a57262 100644
--- a/drivers/mtd/mtd_blkdevs.c
+++ b/drivers/mtd/mtd_blkdevs.c
@@ -14,7 +14,6 @@
 #include <linux/mtd/mtd.h>
 #include <linux/blkdev.h>
 #include <linux/blkpg.h>
-#include <linux/freezer.h>
 #include <linux/spinlock.h>
 #include <linux/hdreg.h>
 #include <linux/init.h>
@@ -26,11 +25,6 @@
 
 static LIST_HEAD(blktrans_majors);
 
-struct mtd_blkcore_priv {
-	struct task_struct *thread;
-	struct request_queue *rq;
-	spinlock_t queue_lock;
-};
 
 static int do_blktrans_request(struct mtd_blktrans_ops *tr,
 			       struct mtd_blktrans_dev *dev,
@@ -61,7 +55,6 @@ static int do_blktrans_request(struct mtd_blktrans_ops *tr,
 				return -EIO;
 		rq_flush_dcache_pages(req);
 		return 0;
-
 	case WRITE:
 		if (!tr->writesect)
 			return -EIO;
@@ -71,7 +64,6 @@ static int do_blktrans_request(struct mtd_blktrans_ops *tr,
 			if (tr->writesect(dev, block, buf))
 				return -EIO;
 		return 0;
-
 	default:
 		printk(KERN_NOTICE "Unknown request %u\n", rq_data_dir(req));
 		return -EIO;
@@ -80,14 +72,13 @@ static int do_blktrans_request(struct mtd_blktrans_ops *tr,
 
 static int mtd_blktrans_thread(void *arg)
 {
-	struct mtd_blktrans_ops *tr = arg;
-	struct request_queue *rq = tr->blkcore_priv->rq;
+	struct mtd_blktrans_dev *dev = arg;
+	struct request_queue *rq = dev->rq;
 	struct request *req = NULL;
 
 	spin_lock_irq(rq->queue_lock);
 
 	while (!kthread_should_stop()) {
-		struct mtd_blktrans_dev *dev;
 		int res;
 
 		if (!req && !(req = blk_fetch_request(rq))) {
@@ -98,13 +89,10 @@ static int mtd_blktrans_thread(void *arg)
 			continue;
 		}
 
-		dev = req->rq_disk->private_data;
-		tr = dev->tr;
-
 		spin_unlock_irq(rq->queue_lock);
 
 		mutex_lock(&dev->lock);
-		res = do_blktrans_request(tr, dev, req);
+		res = do_blktrans_request(dev->tr, dev, req);
 		mutex_unlock(&dev->lock);
 
 		spin_lock_irq(rq->queue_lock);
@@ -123,8 +111,8 @@ static int mtd_blktrans_thread(void *arg)
 
 static void mtd_blktrans_request(struct request_queue *rq)
 {
-	struct mtd_blktrans_ops *tr = rq->queuedata;
-	wake_up_process(tr->blkcore_priv->thread);
+	struct mtd_blktrans_dev *dev = rq->queuedata;
+	wake_up_process(dev->thread);
 }
 
 
@@ -214,6 +202,7 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new)
 	struct mtd_blktrans_dev *d;
 	int last_devnum = -1;
 	struct gendisk *gd;
+	int ret;
 
 	if (mutex_trylock(&mtd_table_mutex)) {
 		mutex_unlock(&mtd_table_mutex);
@@ -239,6 +228,8 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new)
 		}
 		last_devnum = d->devnum;
 	}
+
+	ret = -EBUSY;
 	if (new->devnum == -1)
 		new->devnum = last_devnum+1;
 
@@ -247,7 +238,7 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new)
 	 * with this number. */
 	if (new->devnum > (MINORMASK >> tr->part_bits) ||
 	    (tr->part_bits && new->devnum >= 27 * 26))
-		return -EBUSY;
+		goto error1;
 
 	list_add_tail(&new->list, &tr->devs);
  added:
@@ -255,11 +246,16 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new)
 	if (!tr->writesect)
 		new->readonly = 1;
 
+
+	/* Create gendisk */
+	ret = -ENOMEM;
 	gd = alloc_disk(1 << tr->part_bits);
-	if (!gd) {
-		list_del(&new->list);
-		return -ENOMEM;
-	}
+
+	if (!gd)
+		goto error2;
+
+	new->disk = gd;
+	gd->private_data = new;
 	gd->major = tr->major;
 	gd->first_minor = (new->devnum) << tr->part_bits;
 	gd->fops = &mtd_blktrans_ops;
@@ -277,21 +273,49 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new)
 		snprintf(gd->disk_name, sizeof(gd->disk_name),
 			 "%s%d", tr->name, new->devnum);
 
-	/* 2.5 has capacity in units of 512 bytes while still
-	   having BLOCK_SIZE_BITS set to 10. Just to keep us amused. */
 	set_capacity(gd, (new->size * tr->blksize) >> 9);
 
-	gd->private_data = new;
-	new->blkcore_priv = gd;
-	gd->queue = tr->blkcore_priv->rq;
+
+	/* Create the request queue */
+	spin_lock_init(&new->queue_lock);
+	new->rq = blk_init_queue(mtd_blktrans_request, &new->queue_lock);
+
+	if (!new->rq)
+		goto error3;
+
+	new->rq->queuedata = new;
+	blk_queue_logical_block_size(new->rq, tr->blksize);
+
+	if (tr->discard)
+		queue_flag_set_unlocked(QUEUE_FLAG_DISCARD,
+					new->rq);
+
+	gd->queue = new->rq;
+
+	/* Create processing thread */
+	/* TODO: workqueue ? */
+	new->thread = kthread_run(mtd_blktrans_thread, new,
+			"%s%d", tr->name, new->mtd->index);
+	if (IS_ERR(new->thread)) {
+		ret = PTR_ERR(new->thread);
+		goto error4;
+	}
 	gd->driverfs_dev = &new->mtd->dev;
 
 	if (new->readonly)
 		set_disk_ro(gd, 1);
 
 	add_disk(gd);
-
 	return 0;
+error4:
+	blk_cleanup_queue(new->rq);
+error3:
+	put_disk(new->disk);
+error2:
+	list_del(&new->list);
+error1:
+	kfree(new);
+	return ret;
 }
 
 int del_mtd_blktrans_dev(struct mtd_blktrans_dev *old)
@@ -303,9 +327,13 @@ int del_mtd_blktrans_dev(struct mtd_blktrans_dev *old)
 
 	list_del(&old->list);
 
-	del_gendisk(old->blkcore_priv);
-	put_disk(old->blkcore_priv);
+	/* stop new requests to arrive */
+	del_gendisk(old->disk);
 
+	/* Stop the thread */
+	kthread_stop(old->thread);
+
+	blk_cleanup_queue(old->rq);
 	return 0;
 }
 
@@ -347,9 +375,6 @@ int register_mtd_blktrans(struct mtd_blktrans_ops *tr)
 	if (!blktrans_notifier.list.next)
 		register_mtd_user(&blktrans_notifier);
 
-	tr->blkcore_priv = kzalloc(sizeof(*tr->blkcore_priv), GFP_KERNEL);
-	if (!tr->blkcore_priv)
-		return -ENOMEM;
 
 	mutex_lock(&mtd_table_mutex);
 
@@ -357,39 +382,12 @@ int register_mtd_blktrans(struct mtd_blktrans_ops *tr)
 	if (ret) {
 		printk(KERN_WARNING "Unable to register %s block device on major %d: %d\n",
 		       tr->name, tr->major, ret);
-		kfree(tr->blkcore_priv);
 		mutex_unlock(&mtd_table_mutex);
 		return ret;
 	}
-	spin_lock_init(&tr->blkcore_priv->queue_lock);
-
-	tr->blkcore_priv->rq = blk_init_queue(mtd_blktrans_request, &tr->blkcore_priv->queue_lock);
-	if (!tr->blkcore_priv->rq) {
-		unregister_blkdev(tr->major, tr->name);
-		kfree(tr->blkcore_priv);
-		mutex_unlock(&mtd_table_mutex);
-		return -ENOMEM;
-	}
-
-	tr->blkcore_priv->rq->queuedata = tr;
-	blk_queue_logical_block_size(tr->blkcore_priv->rq, tr->blksize);
-	if (tr->discard)
-		queue_flag_set_unlocked(QUEUE_FLAG_DISCARD,
-					tr->blkcore_priv->rq);
 
 	tr->blkshift = ffs(tr->blksize) - 1;
 
-	tr->blkcore_priv->thread = kthread_run(mtd_blktrans_thread, tr,
-			"%sd", tr->name);
-	if (IS_ERR(tr->blkcore_priv->thread)) {
-		ret = PTR_ERR(tr->blkcore_priv->thread);
-		blk_cleanup_queue(tr->blkcore_priv->rq);
-		unregister_blkdev(tr->major, tr->name);
-		kfree(tr->blkcore_priv);
-		mutex_unlock(&mtd_table_mutex);
-		return ret;
-	}
-
 	INIT_LIST_HEAD(&tr->devs);
 	list_add(&tr->list, &blktrans_majors);
 
@@ -408,8 +406,6 @@ int deregister_mtd_blktrans(struct mtd_blktrans_ops *tr)
 
 	mutex_lock(&mtd_table_mutex);
 
-	/* Clean up the kernel thread */
-	kthread_stop(tr->blkcore_priv->thread);
 
 	/* Remove it from the list of active majors */
 	list_del(&tr->list);
@@ -417,13 +413,9 @@ int deregister_mtd_blktrans(struct mtd_blktrans_ops *tr)
 	list_for_each_entry_safe(dev, next, &tr->devs, list)
 		tr->remove_dev(dev);
 
-	blk_cleanup_queue(tr->blkcore_priv->rq);
 	unregister_blkdev(tr->major, tr->name);
-
 	mutex_unlock(&mtd_table_mutex);
 
-	kfree(tr->blkcore_priv);
-
 	BUG_ON(!list_empty(&tr->devs));
 	return 0;
 }
diff --git a/include/linux/mtd/blktrans.h b/include/linux/mtd/blktrans.h
index 8b4aa05..a4b3928 100644
--- a/include/linux/mtd/blktrans.h
+++ b/include/linux/mtd/blktrans.h
@@ -24,11 +24,13 @@ struct mtd_blktrans_dev {
 	int devnum;
 	unsigned long size;
 	int readonly;
-	void *blkcore_priv; /* gendisk in 2.5, devfs_handle in 2.4 */
+	struct gendisk *disk;
+	struct task_struct *thread;
+	struct request_queue *rq;
+	spinlock_t queue_lock;
+	void *priv;
 };
 
-struct blkcore_priv; /* Differs for 2.4 and 2.5 kernels; private */
-
 struct mtd_blktrans_ops {
 	char *name;
 	int major;
@@ -60,8 +62,6 @@ struct mtd_blktrans_ops {
 	struct list_head devs;
 	struct list_head list;
 	struct module *owner;
-
-	struct mtd_blkcore_priv *blkcore_priv;
 };
 
 extern int register_mtd_blktrans(struct mtd_blktrans_ops *tr);
-- 
1.6.3.3

[PATCH 03/15] blktrans: Hotplug fixes

[Maxim Levitsky]

* Add locking where it was missing.
* Don't do a get_mtd_device in blktrans_open because it would lead to a deadlock
  instead do that in add_mtd_blktrans_dev.

* Only free the mtd_blktrans_dev structure when the last user exits.

* Flush request queue on device removal.

* Track users, and call tr->release in del_mtd_blktrans_dev
  Due to that ->open and release aren't called more that once.

Now it is safe to call del_mtd_blktrans_dev while the device is still in use.

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/ftl.c            |    1 -
 drivers/mtd/inftlcore.c      |    1 -
 drivers/mtd/mtd_blkdevs.c    |  202 ++++++++++++++++++++++++++++++-----------
 drivers/mtd/mtdblock.c       |    2 -
 drivers/mtd/mtdblock_ro.c    |    1 -
 drivers/mtd/nftlcore.c       |    1 -
 drivers/mtd/rfd_ftl.c        |    1 -
 drivers/mtd/ssfdc.c          |    1 -
 include/linux/mtd/blktrans.h |    3 +
 9 files changed, 151 insertions(+), 62 deletions(-)

diff --git a/drivers/mtd/ftl.c b/drivers/mtd/ftl.c
index e56d6b4..62da9eb 100644
--- a/drivers/mtd/ftl.c
+++ b/drivers/mtd/ftl.c
@@ -1082,7 +1082,6 @@ static void ftl_remove_dev(struct mtd_blktrans_dev *dev)
 {
 	del_mtd_blktrans_dev(dev);
 	ftl_freepart((partition_t *)dev);
-	kfree(dev);
 }
 
 static struct mtd_blktrans_ops ftl_tr = {
diff --git a/drivers/mtd/inftlcore.c b/drivers/mtd/inftlcore.c
index 8aca552..015a7fe 100755
--- a/drivers/mtd/inftlcore.c
+++ b/drivers/mtd/inftlcore.c
@@ -139,7 +139,6 @@ static void inftl_remove_dev(struct mtd_blktrans_dev *dev)
 
 	kfree(inftl->PUtable);
 	kfree(inftl->VUtable);
-	kfree(inftl);
 }
 
 /*
diff --git a/drivers/mtd/mtd_blkdevs.c b/drivers/mtd/mtd_blkdevs.c
index 6a57262..646cc84 100644
--- a/drivers/mtd/mtd_blkdevs.c
+++ b/drivers/mtd/mtd_blkdevs.c
@@ -24,6 +24,40 @@
 #include "mtdcore.h"
 
 static LIST_HEAD(blktrans_majors);
+static DEFINE_MUTEX(blktrans_ref_mutex);
+
+void blktrans_dev_release(struct kref *kref)
+{
+	struct mtd_blktrans_dev *dev =
+		container_of(kref, struct mtd_blktrans_dev, ref);
+
+	dev->disk->private_data = NULL;
+	put_disk(dev->disk);
+	list_del(&dev->list);
+	kfree(dev);
+}
+
+static struct mtd_blktrans_dev *blktrans_dev_get(struct gendisk *disk)
+{
+	struct mtd_blktrans_dev *dev;
+
+	mutex_lock(&blktrans_ref_mutex);
+	dev = disk->private_data;
+
+	if (!dev)
+		goto unlock;
+	kref_get(&dev->ref);
+unlock:
+	mutex_unlock(&blktrans_ref_mutex);
+	return dev;
+}
+
+void blktrans_dev_put(struct mtd_blktrans_dev *dev)
+{
+	mutex_lock(&blktrans_ref_mutex);
+	kref_put(&dev->ref, blktrans_dev_release);
+	mutex_unlock(&blktrans_ref_mutex);
+}
 
 
 static int do_blktrans_request(struct mtd_blktrans_ops *tr,
@@ -111,81 +145,112 @@ static int mtd_blktrans_thread(void *arg)
 
 static void mtd_blktrans_request(struct request_queue *rq)
 {
-	struct mtd_blktrans_dev *dev = rq->queuedata;
-	wake_up_process(dev->thread);
-}
+	struct mtd_blktrans_dev *dev;
+	struct request *req = NULL;
+
+	dev = rq->queuedata;
 
+	if (!dev)
+		while ((req = blk_fetch_request(rq)) != NULL)
+			__blk_end_request_all(req, -ENODEV);
+	else
+		wake_up_process(dev->thread);
+}
 
 static int blktrans_open(struct block_device *bdev, fmode_t mode)
 {
-	struct mtd_blktrans_dev *dev = bdev->bd_disk->private_data;
-	struct mtd_blktrans_ops *tr = dev->tr;
-	int ret = -ENODEV;
-
-	if (!get_mtd_device(NULL, dev->mtd->index))
-		goto out;
-
-	if (!try_module_get(tr->owner))
-		goto out_tr;
-
-	/* FIXME: Locking. A hot pluggable device can go away
-	   (del_mtd_device can be called for it) without its module
-	   being unloaded. */
-	dev->mtd->usecount++;
-
-	ret = 0;
-	if (tr->open && (ret = tr->open(dev))) {
-		dev->mtd->usecount--;
-		put_mtd_device(dev->mtd);
-	out_tr:
-		module_put(tr->owner);
+	struct mtd_blktrans_dev *dev = blktrans_dev_get(bdev->bd_disk);
+	int ret;
+
+	if (!dev)
+		return -ERESTARTSYS;
+
+	mutex_lock(&dev->lock);
+
+	if (!dev->mtd) {
+		ret = -ENXIO;
+		goto unlock;
 	}
- out:
+
+	ret = !dev->open++ && dev->tr->open ? dev->tr->open(dev) : 0;
+
+	/* Take another reference on the device so it won't go away till
+		last release */
+	if (!ret)
+		kref_get(&dev->ref);
+unlock:
+	mutex_unlock(&dev->lock);
+	blktrans_dev_put(dev);
 	return ret;
 }
 
 static int blktrans_release(struct gendisk *disk, fmode_t mode)
 {
-	struct mtd_blktrans_dev *dev = disk->private_data;
-	struct mtd_blktrans_ops *tr = dev->tr;
-	int ret = 0;
+	struct mtd_blktrans_dev *dev = blktrans_dev_get(disk);
+	int ret = -ENXIO;
 
-	if (tr->release)
-		ret = tr->release(dev);
+	if (!dev)
+		return ret;
 
-	if (!ret) {
-		dev->mtd->usecount--;
-		put_mtd_device(dev->mtd);
-		module_put(tr->owner);
-	}
+	mutex_lock(&dev->lock);
+
+	/* Release one reference, we sure its not the last one here*/
+	kref_put(&dev->ref, blktrans_dev_release);
 
+	if (!dev->mtd)
+		goto unlock;
+
+	ret = !--dev->open && dev->tr->release ? dev->tr->release(dev) : 0;
+unlock:
+	mutex_unlock(&dev->lock);
+	blktrans_dev_put(dev);
 	return ret;
 }
 
 static int blktrans_getgeo(struct block_device *bdev, struct hd_geometry *geo)
 {
-	struct mtd_blktrans_dev *dev = bdev->bd_disk->private_data;
+	struct mtd_blktrans_dev *dev = blktrans_dev_get(bdev->bd_disk);
+	int ret = -ENXIO;
+
+	if (!dev)
+		return ret;
+
+	mutex_lock(&dev->lock);
 
-	if (dev->tr->getgeo)
-		return dev->tr->getgeo(dev, geo);
-	return -ENOTTY;
+	if (!dev->mtd)
+		goto unlock;
+
+	ret = dev->tr->getgeo ? dev->tr->getgeo(dev, geo) : 0;
+unlock:
+	mutex_unlock(&dev->lock);
+	blktrans_dev_put(dev);
+	return ret;
 }
 
 static int blktrans_ioctl(struct block_device *bdev, fmode_t mode,
 			      unsigned int cmd, unsigned long arg)
 {
-	struct mtd_blktrans_dev *dev = bdev->bd_disk->private_data;
-	struct mtd_blktrans_ops *tr = dev->tr;
+	struct mtd_blktrans_dev *dev = blktrans_dev_get(bdev->bd_disk);
+	int ret = -ENXIO;
+
+	if (!dev)
+		return ret;
+
+	mutex_lock(&dev->lock);
+
+	if (!dev->mtd)
+		goto unlock;
 
 	switch (cmd) {
 	case BLKFLSBUF:
-		if (tr->flush)
-			return tr->flush(dev);
-		/* The core code did the work, we had nothing to do. */
-		return 0;
+		ret = dev->tr->flush ? dev->tr->flush(dev) : 0;
 	default:
-		return -ENOTTY;
+		ret = -ENOTTY;
 	}
+unlock:
+	mutex_unlock(&dev->lock);
+	blktrans_dev_put(dev);
+	return ret;
 }
 
 static const struct block_device_operations mtd_blktrans_ops = {
@@ -209,6 +274,7 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new)
 		BUG();
 	}
 
+	mutex_lock(&blktrans_ref_mutex);
 	list_for_each_entry(d, &tr->devs, list) {
 		if (new->devnum == -1) {
 			/* Use first free number */
@@ -220,6 +286,7 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new)
 			}
 		} else if (d->devnum == new->devnum) {
 			/* Required number taken */
+			mutex_unlock(&blktrans_ref_mutex);
 			return -EBUSY;
 		} else if (d->devnum > new->devnum) {
 			/* Required number was free */
@@ -237,16 +304,20 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new)
 	 * minor numbers and that the disk naming code below can cope
 	 * with this number. */
 	if (new->devnum > (MINORMASK >> tr->part_bits) ||
-	    (tr->part_bits && new->devnum >= 27 * 26))
+	    (tr->part_bits && new->devnum >= 27 * 26)) {
+		mutex_unlock(&blktrans_ref_mutex);
 		goto error1;
+	}
 
 	list_add_tail(&new->list, &tr->devs);
  added:
+	mutex_unlock(&blktrans_ref_mutex);
+
 	mutex_init(&new->lock);
+	kref_init(&new->ref);
 	if (!tr->writesect)
 		new->readonly = 1;
 
-
 	/* Create gendisk */
 	ret = -ENOMEM;
 	gd = alloc_disk(1 << tr->part_bits);
@@ -275,7 +346,6 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new)
 
 	set_capacity(gd, (new->size * tr->blksize) >> 9);
 
-
 	/* Create the request queue */
 	spin_lock_init(&new->queue_lock);
 	new->rq = blk_init_queue(mtd_blktrans_request, &new->queue_lock);
@@ -292,6 +362,9 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new)
 
 	gd->queue = new->rq;
 
+	__get_mtd_device(new->mtd);
+	__module_get(tr->owner);
+
 	/* Create processing thread */
 	/* TODO: workqueue ? */
 	new->thread = kthread_run(mtd_blktrans_thread, new,
@@ -308,6 +381,8 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new)
 	add_disk(gd);
 	return 0;
 error4:
+	module_put(tr->owner);
+	__put_mtd_device(new->mtd);
 	blk_cleanup_queue(new->rq);
 error3:
 	put_disk(new->disk);
@@ -320,20 +395,41 @@ error1:
 
 int del_mtd_blktrans_dev(struct mtd_blktrans_dev *old)
 {
+	unsigned long flags;
+
 	if (mutex_trylock(&mtd_table_mutex)) {
 		mutex_unlock(&mtd_table_mutex);
 		BUG();
 	}
 
-	list_del(&old->list);
-
-	/* stop new requests to arrive */
+	/* Stop new requests to arrive */
 	del_gendisk(old->disk);
 
 	/* Stop the thread */
 	kthread_stop(old->thread);
 
+	/* Kill current requests */
+	spin_lock_irqsave(&old->queue_lock, flags);
+	old->rq->queuedata = NULL;
+	blk_start_queue(old->rq);
+	spin_unlock_irqrestore(&old->queue_lock, flags);
 	blk_cleanup_queue(old->rq);
+
+	/* Ask trans driver for release to the mtd device */
+	mutex_lock(&old->lock);
+	if (old->open && old->tr->release) {
+		old->tr->release(old);
+		old->open = 0;
+	}
+
+	__put_mtd_device(old->mtd);
+	module_put(old->tr->owner);
+
+	/* At that point, we don't touch the mtd anymore */
+	old->mtd = NULL;
+
+	mutex_unlock(&old->lock);
+	blktrans_dev_put(old);
 	return 0;
 }
 
@@ -396,7 +492,6 @@ int register_mtd_blktrans(struct mtd_blktrans_ops *tr)
 			tr->add_mtd(tr, mtd);
 
 	mutex_unlock(&mtd_table_mutex);
-
 	return 0;
 }
 
@@ -406,7 +501,6 @@ int deregister_mtd_blktrans(struct mtd_blktrans_ops *tr)
 
 	mutex_lock(&mtd_table_mutex);
 
-
 	/* Remove it from the list of active majors */
 	list_del(&tr->list);
 
diff --git a/drivers/mtd/mtdblock.c b/drivers/mtd/mtdblock.c
index 69f6bf2..8e5da1e 100644
--- a/drivers/mtd/mtdblock.c
+++ b/drivers/mtd/mtdblock.c
@@ -354,9 +354,7 @@ static void mtdblock_add_mtd(struct mtd_blktrans_ops *tr, struct mtd_info *mtd)
 static void mtdblock_remove_dev(struct mtd_blktrans_dev *dev)
 {
 	struct mtdblk_dev *mtdblk = container_of(dev, struct mtdblk_dev, mbd);
-
 	del_mtd_blktrans_dev(dev);
-	kfree(mtdblk);
 }
 
 static struct mtd_blktrans_ops mtdblock_tr = {
diff --git a/drivers/mtd/mtdblock_ro.c b/drivers/mtd/mtdblock_ro.c
index 852165f..54ff288 100644
--- a/drivers/mtd/mtdblock_ro.c
+++ b/drivers/mtd/mtdblock_ro.c
@@ -49,7 +49,6 @@ static void mtdblock_add_mtd(struct mtd_blktrans_ops *tr, struct mtd_info *mtd)
 static void mtdblock_remove_dev(struct mtd_blktrans_dev *dev)
 {
 	del_mtd_blktrans_dev(dev);
-	kfree(dev);
 }
 
 static struct mtd_blktrans_ops mtdblock_tr = {
diff --git a/drivers/mtd/nftlcore.c b/drivers/mtd/nftlcore.c
index 1002e18..a4578bf 100644
--- a/drivers/mtd/nftlcore.c
+++ b/drivers/mtd/nftlcore.c
@@ -126,7 +126,6 @@ static void nftl_remove_dev(struct mtd_blktrans_dev *dev)
 	del_mtd_blktrans_dev(dev);
 	kfree(nftl->ReplUnitTable);
 	kfree(nftl->EUNtable);
-	kfree(nftl);
 }
 
 /*
diff --git a/drivers/mtd/rfd_ftl.c b/drivers/mtd/rfd_ftl.c
index d2aa9c4..63b83c0 100644
--- a/drivers/mtd/rfd_ftl.c
+++ b/drivers/mtd/rfd_ftl.c
@@ -817,7 +817,6 @@ static void rfd_ftl_remove_dev(struct mtd_blktrans_dev *dev)
 	vfree(part->sector_map);
 	kfree(part->header_cache);
 	kfree(part->blocks);
-	kfree(part);
 }
 
 static struct mtd_blktrans_ops rfd_ftl_tr = {
diff --git a/drivers/mtd/ssfdc.c b/drivers/mtd/ssfdc.c
index 3f67e00..81c4ecd 100644
--- a/drivers/mtd/ssfdc.c
+++ b/drivers/mtd/ssfdc.c
@@ -375,7 +375,6 @@ static void ssfdcr_remove_dev(struct mtd_blktrans_dev *dev)
 
 	del_mtd_blktrans_dev(dev);
 	kfree(ssfdc->logic_block_map);
-	kfree(ssfdc);
 }
 
 static int ssfdcr_readsect(struct mtd_blktrans_dev *dev,
diff --git a/include/linux/mtd/blktrans.h b/include/linux/mtd/blktrans.h
index a4b3928..d89b8fb 100644
--- a/include/linux/mtd/blktrans.h
+++ b/include/linux/mtd/blktrans.h
@@ -9,6 +9,7 @@
 #define __MTD_TRANS_H__
 
 #include <linux/mutex.h>
+#include <linux/kref.h>
 
 struct hd_geometry;
 struct mtd_info;
@@ -24,6 +25,8 @@ struct mtd_blktrans_dev {
 	int devnum;
 	unsigned long size;
 	int readonly;
+	int open;
+	struct kref ref;
 	struct gendisk *disk;
 	struct task_struct *thread;
 	struct request_queue *rq;
-- 
1.6.3.3

[PATCH 03/15] blktrans: Hotplug fixes

[Maxim Levitsky]

* Add locking where it was missing.
* Don't do a get_mtd_device in blktrans_open because it would lead to a deadlock
  instead do that in add_mtd_blktrans_dev.

* Only free the mtd_blktrans_dev structure when the last user exits.

* Flush request queue on device removal.

* Track users, and call tr->release in del_mtd_blktrans_dev
  Due to that ->open and release aren't called more that once.

Now it is safe to call del_mtd_blktrans_dev while the device is still in use.

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/ftl.c            |    1 -
 drivers/mtd/inftlcore.c      |    1 -
 drivers/mtd/mtd_blkdevs.c    |  202 ++++++++++++++++++++++++++++++-----------
 drivers/mtd/mtdblock.c       |    2 -
 drivers/mtd/mtdblock_ro.c    |    1 -
 drivers/mtd/nftlcore.c       |    1 -
 drivers/mtd/rfd_ftl.c        |    1 -
 drivers/mtd/ssfdc.c          |    1 -
 include/linux/mtd/blktrans.h |    3 +
 9 files changed, 151 insertions(+), 62 deletions(-)

diff --git a/drivers/mtd/ftl.c b/drivers/mtd/ftl.c
index e56d6b4..62da9eb 100644
--- a/drivers/mtd/ftl.c
+++ b/drivers/mtd/ftl.c
@@ -1082,7 +1082,6 @@ static void ftl_remove_dev(struct mtd_blktrans_dev *dev)
 {
 	del_mtd_blktrans_dev(dev);
 	ftl_freepart((partition_t *)dev);
-	kfree(dev);
 }
 
 static struct mtd_blktrans_ops ftl_tr = {
diff --git a/drivers/mtd/inftlcore.c b/drivers/mtd/inftlcore.c
index 8aca552..015a7fe 100755
--- a/drivers/mtd/inftlcore.c
+++ b/drivers/mtd/inftlcore.c
@@ -139,7 +139,6 @@ static void inftl_remove_dev(struct mtd_blktrans_dev *dev)
 
 	kfree(inftl->PUtable);
 	kfree(inftl->VUtable);
-	kfree(inftl);
 }
 
 /*
diff --git a/drivers/mtd/mtd_blkdevs.c b/drivers/mtd/mtd_blkdevs.c
index 6a57262..646cc84 100644
--- a/drivers/mtd/mtd_blkdevs.c
+++ b/drivers/mtd/mtd_blkdevs.c
@@ -24,6 +24,40 @@
 #include "mtdcore.h"
 
 static LIST_HEAD(blktrans_majors);
+static DEFINE_MUTEX(blktrans_ref_mutex);
+
+void blktrans_dev_release(struct kref *kref)
+{
+	struct mtd_blktrans_dev *dev =
+		container_of(kref, struct mtd_blktrans_dev, ref);
+
+	dev->disk->private_data = NULL;
+	put_disk(dev->disk);
+	list_del(&dev->list);
+	kfree(dev);
+}
+
+static struct mtd_blktrans_dev *blktrans_dev_get(struct gendisk *disk)
+{
+	struct mtd_blktrans_dev *dev;
+
+	mutex_lock(&blktrans_ref_mutex);
+	dev = disk->private_data;
+
+	if (!dev)
+		goto unlock;
+	kref_get(&dev->ref);
+unlock:
+	mutex_unlock(&blktrans_ref_mutex);
+	return dev;
+}
+
+void blktrans_dev_put(struct mtd_blktrans_dev *dev)
+{
+	mutex_lock(&blktrans_ref_mutex);
+	kref_put(&dev->ref, blktrans_dev_release);
+	mutex_unlock(&blktrans_ref_mutex);
+}
 
 
 static int do_blktrans_request(struct mtd_blktrans_ops *tr,
@@ -111,81 +145,112 @@ static int mtd_blktrans_thread(void *arg)
 
 static void mtd_blktrans_request(struct request_queue *rq)
 {
-	struct mtd_blktrans_dev *dev = rq->queuedata;
-	wake_up_process(dev->thread);
-}
+	struct mtd_blktrans_dev *dev;
+	struct request *req = NULL;
+
+	dev = rq->queuedata;
 
+	if (!dev)
+		while ((req = blk_fetch_request(rq)) != NULL)
+			__blk_end_request_all(req, -ENODEV);
+	else
+		wake_up_process(dev->thread);
+}
 
 static int blktrans_open(struct block_device *bdev, fmode_t mode)
 {
-	struct mtd_blktrans_dev *dev = bdev->bd_disk->private_data;
-	struct mtd_blktrans_ops *tr = dev->tr;
-	int ret = -ENODEV;
-
-	if (!get_mtd_device(NULL, dev->mtd->index))
-		goto out;
-
-	if (!try_module_get(tr->owner))
-		goto out_tr;
-
-	/* FIXME: Locking. A hot pluggable device can go away
-	   (del_mtd_device can be called for it) without its module
-	   being unloaded. */
-	dev->mtd->usecount++;
-
-	ret = 0;
-	if (tr->open && (ret = tr->open(dev))) {
-		dev->mtd->usecount--;
-		put_mtd_device(dev->mtd);
-	out_tr:
-		module_put(tr->owner);
+	struct mtd_blktrans_dev *dev = blktrans_dev_get(bdev->bd_disk);
+	int ret;
+
+	if (!dev)
+		return -ERESTARTSYS;
+
+	mutex_lock(&dev->lock);
+
+	if (!dev->mtd) {
+		ret = -ENXIO;
+		goto unlock;
 	}
- out:
+
+	ret = !dev->open++ && dev->tr->open ? dev->tr->open(dev) : 0;
+
+	/* Take another reference on the device so it won't go away till
+		last release */
+	if (!ret)
+		kref_get(&dev->ref);
+unlock:
+	mutex_unlock(&dev->lock);
+	blktrans_dev_put(dev);
 	return ret;
 }
 
 static int blktrans_release(struct gendisk *disk, fmode_t mode)
 {
-	struct mtd_blktrans_dev *dev = disk->private_data;
-	struct mtd_blktrans_ops *tr = dev->tr;
-	int ret = 0;
+	struct mtd_blktrans_dev *dev = blktrans_dev_get(disk);
+	int ret = -ENXIO;
 
-	if (tr->release)
-		ret = tr->release(dev);
+	if (!dev)
+		return ret;
 
-	if (!ret) {
-		dev->mtd->usecount--;
-		put_mtd_device(dev->mtd);
-		module_put(tr->owner);
-	}
+	mutex_lock(&dev->lock);
+
+	/* Release one reference, we sure its not the last one here*/
+	kref_put(&dev->ref, blktrans_dev_release);
 
+	if (!dev->mtd)
+		goto unlock;
+
+	ret = !--dev->open && dev->tr->release ? dev->tr->release(dev) : 0;
+unlock:
+	mutex_unlock(&dev->lock);
+	blktrans_dev_put(dev);
 	return ret;
 }
 
 static int blktrans_getgeo(struct block_device *bdev, struct hd_geometry *geo)
 {
-	struct mtd_blktrans_dev *dev = bdev->bd_disk->private_data;
+	struct mtd_blktrans_dev *dev = blktrans_dev_get(bdev->bd_disk);
+	int ret = -ENXIO;
+
+	if (!dev)
+		return ret;
+
+	mutex_lock(&dev->lock);
 
-	if (dev->tr->getgeo)
-		return dev->tr->getgeo(dev, geo);
-	return -ENOTTY;
+	if (!dev->mtd)
+		goto unlock;
+
+	ret = dev->tr->getgeo ? dev->tr->getgeo(dev, geo) : 0;
+unlock:
+	mutex_unlock(&dev->lock);
+	blktrans_dev_put(dev);
+	return ret;
 }
 
 static int blktrans_ioctl(struct block_device *bdev, fmode_t mode,
 			      unsigned int cmd, unsigned long arg)
 {
-	struct mtd_blktrans_dev *dev = bdev->bd_disk->private_data;
-	struct mtd_blktrans_ops *tr = dev->tr;
+	struct mtd_blktrans_dev *dev = blktrans_dev_get(bdev->bd_disk);
+	int ret = -ENXIO;
+
+	if (!dev)
+		return ret;
+
+	mutex_lock(&dev->lock);
+
+	if (!dev->mtd)
+		goto unlock;
 
 	switch (cmd) {
 	case BLKFLSBUF:
-		if (tr->flush)
-			return tr->flush(dev);
-		/* The core code did the work, we had nothing to do. */
-		return 0;
+		ret = dev->tr->flush ? dev->tr->flush(dev) : 0;
 	default:
-		return -ENOTTY;
+		ret = -ENOTTY;
 	}
+unlock:
+	mutex_unlock(&dev->lock);
+	blktrans_dev_put(dev);
+	return ret;
 }
 
 static const struct block_device_operations mtd_blktrans_ops = {
@@ -209,6 +274,7 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new)
 		BUG();
 	}
 
+	mutex_lock(&blktrans_ref_mutex);
 	list_for_each_entry(d, &tr->devs, list) {
 		if (new->devnum == -1) {
 			/* Use first free number */
@@ -220,6 +286,7 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new)
 			}
 		} else if (d->devnum == new->devnum) {
 			/* Required number taken */
+			mutex_unlock(&blktrans_ref_mutex);
 			return -EBUSY;
 		} else if (d->devnum > new->devnum) {
 			/* Required number was free */
@@ -237,16 +304,20 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new)
 	 * minor numbers and that the disk naming code below can cope
 	 * with this number. */
 	if (new->devnum > (MINORMASK >> tr->part_bits) ||
-	    (tr->part_bits && new->devnum >= 27 * 26))
+	    (tr->part_bits && new->devnum >= 27 * 26)) {
+		mutex_unlock(&blktrans_ref_mutex);
 		goto error1;
+	}
 
 	list_add_tail(&new->list, &tr->devs);
  added:
+	mutex_unlock(&blktrans_ref_mutex);
+
 	mutex_init(&new->lock);
+	kref_init(&new->ref);
 	if (!tr->writesect)
 		new->readonly = 1;
 
-
 	/* Create gendisk */
 	ret = -ENOMEM;
 	gd = alloc_disk(1 << tr->part_bits);
@@ -275,7 +346,6 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new)
 
 	set_capacity(gd, (new->size * tr->blksize) >> 9);
 
-
 	/* Create the request queue */
 	spin_lock_init(&new->queue_lock);
 	new->rq = blk_init_queue(mtd_blktrans_request, &new->queue_lock);
@@ -292,6 +362,9 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new)
 
 	gd->queue = new->rq;
 
+	__get_mtd_device(new->mtd);
+	__module_get(tr->owner);
+
 	/* Create processing thread */
 	/* TODO: workqueue ? */
 	new->thread = kthread_run(mtd_blktrans_thread, new,
@@ -308,6 +381,8 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new)
 	add_disk(gd);
 	return 0;
 error4:
+	module_put(tr->owner);
+	__put_mtd_device(new->mtd);
 	blk_cleanup_queue(new->rq);
 error3:
 	put_disk(new->disk);
@@ -320,20 +395,41 @@ error1:
 
 int del_mtd_blktrans_dev(struct mtd_blktrans_dev *old)
 {
+	unsigned long flags;
+
 	if (mutex_trylock(&mtd_table_mutex)) {
 		mutex_unlock(&mtd_table_mutex);
 		BUG();
 	}
 
-	list_del(&old->list);
-
-	/* stop new requests to arrive */
+	/* Stop new requests to arrive */
 	del_gendisk(old->disk);
 
 	/* Stop the thread */
 	kthread_stop(old->thread);
 
+	/* Kill current requests */
+	spin_lock_irqsave(&old->queue_lock, flags);
+	old->rq->queuedata = NULL;
+	blk_start_queue(old->rq);
+	spin_unlock_irqrestore(&old->queue_lock, flags);
 	blk_cleanup_queue(old->rq);
+
+	/* Ask trans driver for release to the mtd device */
+	mutex_lock(&old->lock);
+	if (old->open && old->tr->release) {
+		old->tr->release(old);
+		old->open = 0;
+	}
+
+	__put_mtd_device(old->mtd);
+	module_put(old->tr->owner);
+
+	/* At that point, we don't touch the mtd anymore */
+	old->mtd = NULL;
+
+	mutex_unlock(&old->lock);
+	blktrans_dev_put(old);
 	return 0;
 }
 
@@ -396,7 +492,6 @@ int register_mtd_blktrans(struct mtd_blktrans_ops *tr)
 			tr->add_mtd(tr, mtd);
 
 	mutex_unlock(&mtd_table_mutex);
-
 	return 0;
 }
 
@@ -406,7 +501,6 @@ int deregister_mtd_blktrans(struct mtd_blktrans_ops *tr)
 
 	mutex_lock(&mtd_table_mutex);
 
-
 	/* Remove it from the list of active majors */
 	list_del(&tr->list);
 
diff --git a/drivers/mtd/mtdblock.c b/drivers/mtd/mtdblock.c
index 69f6bf2..8e5da1e 100644
--- a/drivers/mtd/mtdblock.c
+++ b/drivers/mtd/mtdblock.c
@@ -354,9 +354,7 @@ static void mtdblock_add_mtd(struct mtd_blktrans_ops *tr, struct mtd_info *mtd)
 static void mtdblock_remove_dev(struct mtd_blktrans_dev *dev)
 {
 	struct mtdblk_dev *mtdblk = container_of(dev, struct mtdblk_dev, mbd);
-
 	del_mtd_blktrans_dev(dev);
-	kfree(mtdblk);
 }
 
 static struct mtd_blktrans_ops mtdblock_tr = {
diff --git a/drivers/mtd/mtdblock_ro.c b/drivers/mtd/mtdblock_ro.c
index 852165f..54ff288 100644
--- a/drivers/mtd/mtdblock_ro.c
+++ b/drivers/mtd/mtdblock_ro.c
@@ -49,7 +49,6 @@ static void mtdblock_add_mtd(struct mtd_blktrans_ops *tr, struct mtd_info *mtd)
 static void mtdblock_remove_dev(struct mtd_blktrans_dev *dev)
 {
 	del_mtd_blktrans_dev(dev);
-	kfree(dev);
 }
 
 static struct mtd_blktrans_ops mtdblock_tr = {
diff --git a/drivers/mtd/nftlcore.c b/drivers/mtd/nftlcore.c
index 1002e18..a4578bf 100644
--- a/drivers/mtd/nftlcore.c
+++ b/drivers/mtd/nftlcore.c
@@ -126,7 +126,6 @@ static void nftl_remove_dev(struct mtd_blktrans_dev *dev)
 	del_mtd_blktrans_dev(dev);
 	kfree(nftl->ReplUnitTable);
 	kfree(nftl->EUNtable);
-	kfree(nftl);
 }
 
 /*
diff --git a/drivers/mtd/rfd_ftl.c b/drivers/mtd/rfd_ftl.c
index d2aa9c4..63b83c0 100644
--- a/drivers/mtd/rfd_ftl.c
+++ b/drivers/mtd/rfd_ftl.c
@@ -817,7 +817,6 @@ static void rfd_ftl_remove_dev(struct mtd_blktrans_dev *dev)
 	vfree(part->sector_map);
 	kfree(part->header_cache);
 	kfree(part->blocks);
-	kfree(part);
 }
 
 static struct mtd_blktrans_ops rfd_ftl_tr = {
diff --git a/drivers/mtd/ssfdc.c b/drivers/mtd/ssfdc.c
index 3f67e00..81c4ecd 100644
--- a/drivers/mtd/ssfdc.c
+++ b/drivers/mtd/ssfdc.c
@@ -375,7 +375,6 @@ static void ssfdcr_remove_dev(struct mtd_blktrans_dev *dev)
 
 	del_mtd_blktrans_dev(dev);
 	kfree(ssfdc->logic_block_map);
-	kfree(ssfdc);
 }
 
 static int ssfdcr_readsect(struct mtd_blktrans_dev *dev,
diff --git a/include/linux/mtd/blktrans.h b/include/linux/mtd/blktrans.h
index a4b3928..d89b8fb 100644
--- a/include/linux/mtd/blktrans.h
+++ b/include/linux/mtd/blktrans.h
@@ -9,6 +9,7 @@
 #define __MTD_TRANS_H__
 
 #include <linux/mutex.h>
+#include <linux/kref.h>
 
 struct hd_geometry;
 struct mtd_info;
@@ -24,6 +25,8 @@ struct mtd_blktrans_dev {
 	int devnum;
 	unsigned long size;
 	int readonly;
+	int open;
+	struct kref ref;
 	struct gendisk *disk;
 	struct task_struct *thread;
 	struct request_queue *rq;
-- 
1.6.3.3

[PATCH 04/15] MTD: mtdblock: test return value of add_mtd_blktrans_dev, because if can fail

[Maxim Levitsky]

This prevents a memory leak

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/mtdblock.c    |    3 ++-
 drivers/mtd/mtdblock_ro.c |    3 ++-
 2 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/drivers/mtd/mtdblock.c b/drivers/mtd/mtdblock.c
index 8e5da1e..7ce30a2 100644
--- a/drivers/mtd/mtdblock.c
+++ b/drivers/mtd/mtdblock.c
@@ -348,7 +348,8 @@ static void mtdblock_add_mtd(struct mtd_blktrans_ops *tr, struct mtd_info *mtd)
 	if (!(mtd->flags & MTD_WRITEABLE))
 		dev->mbd.readonly = 1;
 
-	add_mtd_blktrans_dev(&dev->mbd);
+	if (add_mtd_blktrans_dev(&dev->mbd))
+		kfree(dev);
 }
 
 static void mtdblock_remove_dev(struct mtd_blktrans_dev *dev)
diff --git a/drivers/mtd/mtdblock_ro.c b/drivers/mtd/mtdblock_ro.c
index 54ff288..d0d3f79 100644
--- a/drivers/mtd/mtdblock_ro.c
+++ b/drivers/mtd/mtdblock_ro.c
@@ -43,7 +43,8 @@ static void mtdblock_add_mtd(struct mtd_blktrans_ops *tr, struct mtd_info *mtd)
 	dev->tr = tr;
 	dev->readonly = 1;
 
-	add_mtd_blktrans_dev(dev);
+	if (add_mtd_blktrans_dev(dev))
+		kfree(dev);
 }
 
 static void mtdblock_remove_dev(struct mtd_blktrans_dev *dev)
-- 
1.6.3.3

[PATCH 04/15] MTD: mtdblock: test return value of add_mtd_blktrans_dev, because if can fail

[Maxim Levitsky]

This prevents a memory leak

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/mtdblock.c    |    3 ++-
 drivers/mtd/mtdblock_ro.c |    3 ++-
 2 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/drivers/mtd/mtdblock.c b/drivers/mtd/mtdblock.c
index 8e5da1e..7ce30a2 100644
--- a/drivers/mtd/mtdblock.c
+++ b/drivers/mtd/mtdblock.c
@@ -348,7 +348,8 @@ static void mtdblock_add_mtd(struct mtd_blktrans_ops *tr, struct mtd_info *mtd)
 	if (!(mtd->flags & MTD_WRITEABLE))
 		dev->mbd.readonly = 1;
 
-	add_mtd_blktrans_dev(&dev->mbd);
+	if (add_mtd_blktrans_dev(&dev->mbd))
+		kfree(dev);
 }
 
 static void mtdblock_remove_dev(struct mtd_blktrans_dev *dev)
diff --git a/drivers/mtd/mtdblock_ro.c b/drivers/mtd/mtdblock_ro.c
index 54ff288..d0d3f79 100644
--- a/drivers/mtd/mtdblock_ro.c
+++ b/drivers/mtd/mtdblock_ro.c
@@ -43,7 +43,8 @@ static void mtdblock_add_mtd(struct mtd_blktrans_ops *tr, struct mtd_info *mtd)
 	dev->tr = tr;
 	dev->readonly = 1;
 
-	add_mtd_blktrans_dev(dev);
+	if (add_mtd_blktrans_dev(dev))
+		kfree(dev);
 }
 
 static void mtdblock_remove_dev(struct mtd_blktrans_dev *dev)
-- 
1.6.3.3

[PATCH 05/15] MTD: call the remove notifiers before assuming it is in use

[Maxim Levitsky]

Now that mtd block common layer is prepared for proper hotplug support,
enable it here

Now all users of the mtd device have a chance to put the mtd device
when they are notified to do so, and they have to do so to make hotplug work.

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/mtdcore.c |   19 +++++++++++--------
 1 files changed, 11 insertions(+), 8 deletions(-)

diff --git a/drivers/mtd/mtdcore.c b/drivers/mtd/mtdcore.c
index 67669a7..70a7858 100644
--- a/drivers/mtd/mtdcore.c
+++ b/drivers/mtd/mtdcore.c
@@ -350,31 +350,34 @@ fail_locked:
 int del_mtd_device (struct mtd_info *mtd)
 {
 	int ret;
+	struct mtd_notifier *not;
 
 	mutex_lock(&mtd_table_mutex);
 
 	if (idr_find(&mtd_idr, mtd->index) != mtd) {
 		ret = -ENODEV;
-	} else if (mtd->usecount) {
+		goto out_error;
+	}
+
+	/* No need to get a refcount on the module containing
+		the notifier, since we hold the mtd_table_mutex */
+	list_for_each_entry(not, &mtd_notifiers, list)
+		not->remove(mtd);
+
+	if (mtd->usecount) {
 		printk(KERN_NOTICE "Removing MTD device #%d (%s) with use count %d\n",
 		       mtd->index, mtd->name, mtd->usecount);
 		ret = -EBUSY;
 	} else {
-		struct mtd_notifier *not;
-
 		device_unregister(&mtd->dev);
 
-		/* No need to get a refcount on the module containing
-		   the notifier, since we hold the mtd_table_mutex */
-		list_for_each_entry(not, &mtd_notifiers, list)
-			not->remove(mtd);
-
 		idr_remove(&mtd_idr, mtd->index);
 
 		module_put(THIS_MODULE);
 		ret = 0;
 	}
 
+out_error:
 	mutex_unlock(&mtd_table_mutex);
 	return ret;
 }
-- 
1.6.3.3

[PATCH 05/15] MTD: call the remove notifiers before assuming it is in use

[Maxim Levitsky]

Now that mtd block common layer is prepared for proper hotplug support,
enable it here

Now all users of the mtd device have a chance to put the mtd device
when they are notified to do so, and they have to do so to make hotplug work.

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/mtdcore.c |   19 +++++++++++--------
 1 files changed, 11 insertions(+), 8 deletions(-)

diff --git a/drivers/mtd/mtdcore.c b/drivers/mtd/mtdcore.c
index 67669a7..70a7858 100644
--- a/drivers/mtd/mtdcore.c
+++ b/drivers/mtd/mtdcore.c
@@ -350,31 +350,34 @@ fail_locked:
 int del_mtd_device (struct mtd_info *mtd)
 {
 	int ret;
+	struct mtd_notifier *not;
 
 	mutex_lock(&mtd_table_mutex);
 
 	if (idr_find(&mtd_idr, mtd->index) != mtd) {
 		ret = -ENODEV;
-	} else if (mtd->usecount) {
+		goto out_error;
+	}
+
+	/* No need to get a refcount on the module containing
+		the notifier, since we hold the mtd_table_mutex */
+	list_for_each_entry(not, &mtd_notifiers, list)
+		not->remove(mtd);
+
+	if (mtd->usecount) {
 		printk(KERN_NOTICE "Removing MTD device #%d (%s) with use count %d\n",
 		       mtd->index, mtd->name, mtd->usecount);
 		ret = -EBUSY;
 	} else {
-		struct mtd_notifier *not;
-
 		device_unregister(&mtd->dev);
 
-		/* No need to get a refcount on the module containing
-		   the notifier, since we hold the mtd_table_mutex */
-		list_for_each_entry(not, &mtd_notifiers, list)
-			not->remove(mtd);
-
 		idr_remove(&mtd_idr, mtd->index);
 
 		module_put(THIS_MODULE);
 		ret = 0;
 	}
 
+out_error:
 	mutex_unlock(&mtd_table_mutex);
 	return ret;
 }
-- 
1.6.3.3

[PATCH 06/15] blktrans: allow FTL drivers to export sysfs attributes

[Maxim Levitsky]

This patch adds an ability to export sysfs attributes below
the block disk device.

This can be used to pass the udev an information about the FTL
and could include the vendor, serial, version, etc...

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/mtd_blkdevs.c    |    8 ++++++++
 include/linux/mtd/blktrans.h |    2 ++
 2 files changed, 10 insertions(+), 0 deletions(-)

diff --git a/drivers/mtd/mtd_blkdevs.c b/drivers/mtd/mtd_blkdevs.c
index 646cc84..9dd23d6 100644
--- a/drivers/mtd/mtd_blkdevs.c
+++ b/drivers/mtd/mtd_blkdevs.c
@@ -379,6 +379,10 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new)
 		set_disk_ro(gd, 1);
 
 	add_disk(gd);
+
+	if (new->disk_attributes)
+		sysfs_create_group(&disk_to_dev(gd)->kobj,
+					new->disk_attributes);
 	return 0;
 error4:
 	module_put(tr->owner);
@@ -405,6 +409,10 @@ int del_mtd_blktrans_dev(struct mtd_blktrans_dev *old)
 	/* Stop new requests to arrive */
 	del_gendisk(old->disk);
 
+	if (old->disk_attributes)
+		sysfs_remove_group(&disk_to_dev(old->disk)->kobj,
+						old->disk_attributes);
+
 	/* Stop the thread */
 	kthread_stop(old->thread);
 
diff --git a/include/linux/mtd/blktrans.h b/include/linux/mtd/blktrans.h
index d89b8fb..b481ccd 100644
--- a/include/linux/mtd/blktrans.h
+++ b/include/linux/mtd/blktrans.h
@@ -10,6 +10,7 @@
 
 #include <linux/mutex.h>
 #include <linux/kref.h>
+#include <linux/sysfs.h>
 
 struct hd_geometry;
 struct mtd_info;
@@ -28,6 +29,7 @@ struct mtd_blktrans_dev {
 	int open;
 	struct kref ref;
 	struct gendisk *disk;
+	struct attribute_group *disk_attributes;
 	struct task_struct *thread;
 	struct request_queue *rq;
 	spinlock_t queue_lock;
-- 
1.6.3.3

[PATCH 06/15] blktrans: allow FTL drivers to export sysfs attributes

[Maxim Levitsky]

This patch adds an ability to export sysfs attributes below
the block disk device.

This can be used to pass the udev an information about the FTL
and could include the vendor, serial, version, etc...

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/mtd_blkdevs.c    |    8 ++++++++
 include/linux/mtd/blktrans.h |    2 ++
 2 files changed, 10 insertions(+), 0 deletions(-)

diff --git a/drivers/mtd/mtd_blkdevs.c b/drivers/mtd/mtd_blkdevs.c
index 646cc84..9dd23d6 100644
--- a/drivers/mtd/mtd_blkdevs.c
+++ b/drivers/mtd/mtd_blkdevs.c
@@ -379,6 +379,10 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new)
 		set_disk_ro(gd, 1);
 
 	add_disk(gd);
+
+	if (new->disk_attributes)
+		sysfs_create_group(&disk_to_dev(gd)->kobj,
+					new->disk_attributes);
 	return 0;
 error4:
 	module_put(tr->owner);
@@ -405,6 +409,10 @@ int del_mtd_blktrans_dev(struct mtd_blktrans_dev *old)
 	/* Stop new requests to arrive */
 	del_gendisk(old->disk);
 
+	if (old->disk_attributes)
+		sysfs_remove_group(&disk_to_dev(old->disk)->kobj,
+						old->disk_attributes);
+
 	/* Stop the thread */
 	kthread_stop(old->thread);
 
diff --git a/include/linux/mtd/blktrans.h b/include/linux/mtd/blktrans.h
index d89b8fb..b481ccd 100644
--- a/include/linux/mtd/blktrans.h
+++ b/include/linux/mtd/blktrans.h
@@ -10,6 +10,7 @@
 
 #include <linux/mutex.h>
 #include <linux/kref.h>
+#include <linux/sysfs.h>
 
 struct hd_geometry;
 struct mtd_info;
@@ -28,6 +29,7 @@ struct mtd_blktrans_dev {
 	int open;
 	struct kref ref;
 	struct gendisk *disk;
+	struct attribute_group *disk_attributes;
 	struct task_struct *thread;
 	struct request_queue *rq;
 	spinlock_t queue_lock;
-- 
1.6.3.3

[PATCH 07/15] MTD: nand: make suspend work if device is accessed by kernel threads.

[Maxim Levitsky]

Since all userspace threads are frozen at the time the nand_suspend is called,
they aren't inside any nand function.

We don't call try_to_freeze in nand ether. Thus the only user that can be inside
the nand functions is an non freezeable kernel thread.
Thus we can safely wait for it to finish.

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/nand/nand_base.c |    3 ---
 1 files changed, 0 insertions(+), 3 deletions(-)

diff --git a/drivers/mtd/nand/nand_base.c b/drivers/mtd/nand/nand_base.c
index ed62e1e..7442b3a 100644
--- a/drivers/mtd/nand/nand_base.c
+++ b/drivers/mtd/nand/nand_base.c
@@ -773,9 +773,6 @@ nand_get_device(struct nand_chip *chip, struct mtd_info *mtd, int new_state)
 			chip->state = FL_PM_SUSPENDED;
 			spin_unlock(lock);
 			return 0;
-		} else {
-			spin_unlock(lock);
-			return -EAGAIN;
 		}
 	}
 	set_current_state(TASK_UNINTERRUPTIBLE);
-- 
1.6.3.3

[PATCH 07/15] MTD: nand: make suspend work if device is accessed by kernel threads.

[Maxim Levitsky]

Since all userspace threads are frozen at the time the nand_suspend is called,
they aren't inside any nand function.

We don't call try_to_freeze in nand ether. Thus the only user that can be inside
the nand functions is an non freezeable kernel thread.
Thus we can safely wait for it to finish.

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/nand/nand_base.c |    3 ---
 1 files changed, 0 insertions(+), 3 deletions(-)

diff --git a/drivers/mtd/nand/nand_base.c b/drivers/mtd/nand/nand_base.c
index ed62e1e..7442b3a 100644
--- a/drivers/mtd/nand/nand_base.c
+++ b/drivers/mtd/nand/nand_base.c
@@ -773,9 +773,6 @@ nand_get_device(struct nand_chip *chip, struct mtd_info *mtd, int new_state)
 			chip->state = FL_PM_SUSPENDED;
 			spin_unlock(lock);
 			return 0;
-		} else {
-			spin_unlock(lock);
-			return -EAGAIN;
 		}
 	}
 	set_current_state(TASK_UNINTERRUPTIBLE);
-- 
1.6.3.3

[PATCH 08/15] MTD: nand: make MTD_OOB_PLACE work correctly.

[Maxim Levitsky]

MTD_OOB_PLACE is supposed to read/write the raw oob data similiar to the MTD_OOB_RAW
however due to a bug, currently it is not possible to read more data that
is specified by the oob 'free' regions.

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/nand/nand_base.c |    6 +++++-
 1 files changed, 5 insertions(+), 1 deletions(-)

diff --git a/drivers/mtd/nand/nand_base.c b/drivers/mtd/nand/nand_base.c
index 7442b3a..cada4cf 100644
--- a/drivers/mtd/nand/nand_base.c
+++ b/drivers/mtd/nand/nand_base.c
@@ -1420,6 +1420,9 @@ static int nand_do_read_ops(struct mtd_info *mtd, loff_t from,
 	int ret = 0;
 	uint32_t readlen = ops->len;
 	uint32_t oobreadlen = ops->ooblen;
+	uint32_t max_oobsize = ops->mode == MTD_OOB_AUTO ?
+		mtd->oobavail : mtd->oobsize;
+
 	uint8_t *bufpoi, *oob, *buf;
 
 	stats = mtd->ecc_stats;
@@ -1470,10 +1473,11 @@ static int nand_do_read_ops(struct mtd_info *mtd, loff_t from,
 			buf += bytes;
 
 			if (unlikely(oob)) {
+
 				/* Raw mode does data:oob:data:oob */
 				if (ops->mode != MTD_OOB_RAW) {
 					int toread = min(oobreadlen,
-						chip->ecc.layout->oobavail);
+								max_oobsize);
 					if (toread) {
 						oob = nand_transfer_oob(chip,
 							oob, ops, toread);
-- 
1.6.3.3

[PATCH 08/15] MTD: nand: make MTD_OOB_PLACE work correctly.

[Maxim Levitsky]

MTD_OOB_PLACE is supposed to read/write the raw oob data similiar to the MTD_OOB_RAW
however due to a bug, currently it is not possible to read more data that
is specified by the oob 'free' regions.

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/nand/nand_base.c |    6 +++++-
 1 files changed, 5 insertions(+), 1 deletions(-)

diff --git a/drivers/mtd/nand/nand_base.c b/drivers/mtd/nand/nand_base.c
index 7442b3a..cada4cf 100644
--- a/drivers/mtd/nand/nand_base.c
+++ b/drivers/mtd/nand/nand_base.c
@@ -1420,6 +1420,9 @@ static int nand_do_read_ops(struct mtd_info *mtd, loff_t from,
 	int ret = 0;
 	uint32_t readlen = ops->len;
 	uint32_t oobreadlen = ops->ooblen;
+	uint32_t max_oobsize = ops->mode == MTD_OOB_AUTO ?
+		mtd->oobavail : mtd->oobsize;
+
 	uint8_t *bufpoi, *oob, *buf;
 
 	stats = mtd->ecc_stats;
@@ -1470,10 +1473,11 @@ static int nand_do_read_ops(struct mtd_info *mtd, loff_t from,
 			buf += bytes;
 
 			if (unlikely(oob)) {
+
 				/* Raw mode does data:oob:data:oob */
 				if (ops->mode != MTD_OOB_RAW) {
 					int toread = min(oobreadlen,
-						chip->ecc.layout->oobavail);
+								max_oobsize);
 					if (toread) {
 						oob = nand_transfer_oob(chip,
 							oob, ops, toread);
-- 
1.6.3.3

[PATCH 09/15] MTD: nand: cleanup the nand_do_write_ops

[Maxim Levitsky]

nand_do_write_ops have broken in regard to writing several pages,
 each with its own oob.

Although nand_do_write_ops intends to allow such mode, it fails do do so
Probably this was never tested.

Also add missing checks for attemts to write at illegal offsets.

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/nand/nand_base.c |   24 +++++++++++++++++-------
 1 files changed, 17 insertions(+), 7 deletions(-)

diff --git a/drivers/mtd/nand/nand_base.c b/drivers/mtd/nand/nand_base.c
index cada4cf..1386741 100644
--- a/drivers/mtd/nand/nand_base.c
+++ b/drivers/mtd/nand/nand_base.c
@@ -2072,11 +2072,9 @@ static int nand_write_page(struct mtd_info *mtd, struct nand_chip *chip,
  * @oob:	oob data buffer
  * @ops:	oob ops structure
  */
-static uint8_t *nand_fill_oob(struct nand_chip *chip, uint8_t *oob,
-				  struct mtd_oob_ops *ops)
+static uint8_t *nand_fill_oob(struct nand_chip *chip, uint8_t *oob, size_t len,
+						struct mtd_oob_ops *ops)
 {
-	size_t len = ops->ooblen;
-
 	switch(ops->mode) {
 
 	case MTD_OOB_PLACE:
@@ -2131,6 +2129,11 @@ static int nand_do_write_ops(struct mtd_info *mtd, loff_t to,
 	int chipnr, realpage, page, blockmask, column;
 	struct nand_chip *chip = mtd->priv;
 	uint32_t writelen = ops->len;
+
+	uint32_t oobwritelen = ops->ooblen;
+	uint32_t oobmaxlen = ops->mode == MTD_OOB_AUTO ?
+				mtd->oobavail : mtd->oobsize;
+
 	uint8_t *oob = ops->oobbuf;
 	uint8_t *buf = ops->datbuf;
 	int ret, subpage;
@@ -2172,6 +2175,10 @@ static int nand_do_write_ops(struct mtd_info *mtd, loff_t to,
 	if (likely(!oob))
 		memset(chip->oob_poi, 0xff, mtd->oobsize);
 
+	/* Don't allow multipage oob writes with offset */
+	if (ops->ooboffs && (ops->ooboffs + ops->ooblen > oobmaxlen))
+		return -EINVAL;
+
 	while(1) {
 		int bytes = mtd->writesize;
 		int cached = writelen > bytes && page != blockmask;
@@ -2187,8 +2194,11 @@ static int nand_do_write_ops(struct mtd_info *mtd, loff_t to,
 			wbuf = chip->buffers->databuf;
 		}
 
-		if (unlikely(oob))
-			oob = nand_fill_oob(chip, oob, ops);
+		if (unlikely(oob)) {
+			size_t len = min(oobwritelen, oobmaxlen);
+			oob = nand_fill_oob(chip, oob, len, ops);
+			oobwritelen -= len;
+		}
 
 		ret = chip->write_page(mtd, chip, wbuf, page, cached,
 				       (ops->mode == MTD_OOB_RAW));
@@ -2362,7 +2372,7 @@ static int nand_do_write_oob(struct mtd_info *mtd, loff_t to,
 		chip->pagebuf = -1;
 
 	memset(chip->oob_poi, 0xff, mtd->oobsize);
-	nand_fill_oob(chip, ops->oobbuf, ops);
+	nand_fill_oob(chip, ops->oobbuf, ops->ooblen, ops);
 	status = chip->ecc.write_oob(mtd, chip, page & chip->pagemask);
 	memset(chip->oob_poi, 0xff, mtd->oobsize);
 
-- 
1.6.3.3

[PATCH 09/15] MTD: nand: cleanup the nand_do_write_ops

[Maxim Levitsky]

nand_do_write_ops have broken in regard to writing several pages,
 each with its own oob.

Although nand_do_write_ops intends to allow such mode, it fails do do so
Probably this was never tested.

Also add missing checks for attemts to write at illegal offsets.

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/nand/nand_base.c |   24 +++++++++++++++++-------
 1 files changed, 17 insertions(+), 7 deletions(-)

diff --git a/drivers/mtd/nand/nand_base.c b/drivers/mtd/nand/nand_base.c
index cada4cf..1386741 100644
--- a/drivers/mtd/nand/nand_base.c
+++ b/drivers/mtd/nand/nand_base.c
@@ -2072,11 +2072,9 @@ static int nand_write_page(struct mtd_info *mtd, struct nand_chip *chip,
  * @oob:	oob data buffer
  * @ops:	oob ops structure
  */
-static uint8_t *nand_fill_oob(struct nand_chip *chip, uint8_t *oob,
-				  struct mtd_oob_ops *ops)
+static uint8_t *nand_fill_oob(struct nand_chip *chip, uint8_t *oob, size_t len,
+						struct mtd_oob_ops *ops)
 {
-	size_t len = ops->ooblen;
-
 	switch(ops->mode) {
 
 	case MTD_OOB_PLACE:
@@ -2131,6 +2129,11 @@ static int nand_do_write_ops(struct mtd_info *mtd, loff_t to,
 	int chipnr, realpage, page, blockmask, column;
 	struct nand_chip *chip = mtd->priv;
 	uint32_t writelen = ops->len;
+
+	uint32_t oobwritelen = ops->ooblen;
+	uint32_t oobmaxlen = ops->mode == MTD_OOB_AUTO ?
+				mtd->oobavail : mtd->oobsize;
+
 	uint8_t *oob = ops->oobbuf;
 	uint8_t *buf = ops->datbuf;
 	int ret, subpage;
@@ -2172,6 +2175,10 @@ static int nand_do_write_ops(struct mtd_info *mtd, loff_t to,
 	if (likely(!oob))
 		memset(chip->oob_poi, 0xff, mtd->oobsize);
 
+	/* Don't allow multipage oob writes with offset */
+	if (ops->ooboffs && (ops->ooboffs + ops->ooblen > oobmaxlen))
+		return -EINVAL;
+
 	while(1) {
 		int bytes = mtd->writesize;
 		int cached = writelen > bytes && page != blockmask;
@@ -2187,8 +2194,11 @@ static int nand_do_write_ops(struct mtd_info *mtd, loff_t to,
 			wbuf = chip->buffers->databuf;
 		}
 
-		if (unlikely(oob))
-			oob = nand_fill_oob(chip, oob, ops);
+		if (unlikely(oob)) {
+			size_t len = min(oobwritelen, oobmaxlen);
+			oob = nand_fill_oob(chip, oob, len, ops);
+			oobwritelen -= len;
+		}
 
 		ret = chip->write_page(mtd, chip, wbuf, page, cached,
 				       (ops->mode == MTD_OOB_RAW));
@@ -2362,7 +2372,7 @@ static int nand_do_write_oob(struct mtd_info *mtd, loff_t to,
 		chip->pagebuf = -1;
 
 	memset(chip->oob_poi, 0xff, mtd->oobsize);
-	nand_fill_oob(chip, ops->oobbuf, ops);
+	nand_fill_oob(chip, ops->oobbuf, ops->ooblen, ops);
 	status = chip->ecc.write_oob(mtd, chip, page & chip->pagemask);
 	memset(chip->oob_poi, 0xff, mtd->oobsize);
 
-- 
1.6.3.3

[PATCH 10/15] MTD: nand: make reads using MTD_OOB_RAW affect only ECC validation

[Maxim Levitsky]

This changes the behavier of MTD_OOB_RAW. It used to read both OOB and data
to the data buffer, however you would still need to specify the dummy oob buffer.

This is only used in one place, but makes it hard to read data+oob without ECC
test, thus I removed that behavier, and fixed the user.

Now MTD_OOB_RAW behaves just like MTD_OOB_PLACE, but doesn't do ECC validation

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/nand/nand_base.c |   19 +++++++------------
 drivers/mtd/nand/nand_bbt.c  |   26 ++++++++++++++++++++++----
 include/linux/mtd/mtd.h      |    4 +---
 3 files changed, 30 insertions(+), 19 deletions(-)

diff --git a/drivers/mtd/nand/nand_base.c b/drivers/mtd/nand/nand_base.c
index 1386741..51dfea1 100644
--- a/drivers/mtd/nand/nand_base.c
+++ b/drivers/mtd/nand/nand_base.c
@@ -1474,18 +1474,13 @@ static int nand_do_read_ops(struct mtd_info *mtd, loff_t from,
 
 			if (unlikely(oob)) {
 
-				/* Raw mode does data:oob:data:oob */
-				if (ops->mode != MTD_OOB_RAW) {
-					int toread = min(oobreadlen,
-								max_oobsize);
-					if (toread) {
-						oob = nand_transfer_oob(chip,
-							oob, ops, toread);
-						oobreadlen -= toread;
-					}
-				} else
-					buf = nand_transfer_oob(chip,
-						buf, ops, mtd->oobsize);
+				int toread = min(oobreadlen, max_oobsize);
+
+				if (toread) {
+					oob = nand_transfer_oob(chip,
+						oob, ops, toread);
+					oobreadlen -= toread;
+				}
 			}
 
 			if (!(chip->options & NAND_NO_READRDY)) {
diff --git a/drivers/mtd/nand/nand_bbt.c b/drivers/mtd/nand/nand_bbt.c
index 55c23e5..387c45c 100644
--- a/drivers/mtd/nand/nand_bbt.c
+++ b/drivers/mtd/nand/nand_bbt.c
@@ -237,15 +237,33 @@ static int scan_read_raw(struct mtd_info *mtd, uint8_t *buf, loff_t offs,
 			 size_t len)
 {
 	struct mtd_oob_ops ops;
+	int res;
 
 	ops.mode = MTD_OOB_RAW;
 	ops.ooboffs = 0;
 	ops.ooblen = mtd->oobsize;
-	ops.oobbuf = buf;
-	ops.datbuf = buf;
-	ops.len = len;
 
-	return mtd->read_oob(mtd, offs, &ops);
+
+	while (len > 0) {
+		if (len <= mtd->writesize) {
+			ops.oobbuf = buf + len;
+			ops.datbuf = buf;
+			ops.len = len;
+			return mtd->read_oob(mtd, offs, &ops);
+		} else {
+			ops.oobbuf = buf + mtd->writesize;
+			ops.datbuf = buf;
+			ops.len = mtd->writesize;
+			res = mtd->read_oob(mtd, offs, &ops);
+
+			if (res)
+				return res;
+		}
+
+		buf += mtd->oobsize + mtd->writesize;
+		len -= mtd->writesize;
+	}
+	return 0;
 }
 
 /*
diff --git a/include/linux/mtd/mtd.h b/include/linux/mtd/mtd.h
index 11d8e68..5326435 100644
--- a/include/linux/mtd/mtd.h
+++ b/include/linux/mtd/mtd.h
@@ -60,9 +60,7 @@ struct mtd_erase_region_info {
  * MTD_OOB_PLACE:	oob data are placed at the given offset
  * MTD_OOB_AUTO:	oob data are automatically placed at the free areas
  *			which are defined by the ecclayout
- * MTD_OOB_RAW:		mode to read raw data+oob in one chunk. The oob data
- *			is inserted into the data. Thats a raw image of the
- *			flash contents.
+ * MTD_OOB_RAW:		mode to read oob and data without doing ECC checking
  */
 typedef enum {
 	MTD_OOB_PLACE,
-- 
1.6.3.3

[PATCH 10/15] MTD: nand: make reads using MTD_OOB_RAW affect only ECC validation

[Maxim Levitsky]

This changes the behavier of MTD_OOB_RAW. It used to read both OOB and data
to the data buffer, however you would still need to specify the dummy oob buffer.

This is only used in one place, but makes it hard to read data+oob without ECC
test, thus I removed that behavier, and fixed the user.

Now MTD_OOB_RAW behaves just like MTD_OOB_PLACE, but doesn't do ECC validation

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/nand/nand_base.c |   19 +++++++------------
 drivers/mtd/nand/nand_bbt.c  |   26 ++++++++++++++++++++++----
 include/linux/mtd/mtd.h      |    4 +---
 3 files changed, 30 insertions(+), 19 deletions(-)

diff --git a/drivers/mtd/nand/nand_base.c b/drivers/mtd/nand/nand_base.c
index 1386741..51dfea1 100644
--- a/drivers/mtd/nand/nand_base.c
+++ b/drivers/mtd/nand/nand_base.c
@@ -1474,18 +1474,13 @@ static int nand_do_read_ops(struct mtd_info *mtd, loff_t from,
 
 			if (unlikely(oob)) {
 
-				/* Raw mode does data:oob:data:oob */
-				if (ops->mode != MTD_OOB_RAW) {
-					int toread = min(oobreadlen,
-								max_oobsize);
-					if (toread) {
-						oob = nand_transfer_oob(chip,
-							oob, ops, toread);
-						oobreadlen -= toread;
-					}
-				} else
-					buf = nand_transfer_oob(chip,
-						buf, ops, mtd->oobsize);
+				int toread = min(oobreadlen, max_oobsize);
+
+				if (toread) {
+					oob = nand_transfer_oob(chip,
+						oob, ops, toread);
+					oobreadlen -= toread;
+				}
 			}
 
 			if (!(chip->options & NAND_NO_READRDY)) {
diff --git a/drivers/mtd/nand/nand_bbt.c b/drivers/mtd/nand/nand_bbt.c
index 55c23e5..387c45c 100644
--- a/drivers/mtd/nand/nand_bbt.c
+++ b/drivers/mtd/nand/nand_bbt.c
@@ -237,15 +237,33 @@ static int scan_read_raw(struct mtd_info *mtd, uint8_t *buf, loff_t offs,
 			 size_t len)
 {
 	struct mtd_oob_ops ops;
+	int res;
 
 	ops.mode = MTD_OOB_RAW;
 	ops.ooboffs = 0;
 	ops.ooblen = mtd->oobsize;
-	ops.oobbuf = buf;
-	ops.datbuf = buf;
-	ops.len = len;
 
-	return mtd->read_oob(mtd, offs, &ops);
+
+	while (len > 0) {
+		if (len <= mtd->writesize) {
+			ops.oobbuf = buf + len;
+			ops.datbuf = buf;
+			ops.len = len;
+			return mtd->read_oob(mtd, offs, &ops);
+		} else {
+			ops.oobbuf = buf + mtd->writesize;
+			ops.datbuf = buf;
+			ops.len = mtd->writesize;
+			res = mtd->read_oob(mtd, offs, &ops);
+
+			if (res)
+				return res;
+		}
+
+		buf += mtd->oobsize + mtd->writesize;
+		len -= mtd->writesize;
+	}
+	return 0;
 }
 
 /*
diff --git a/include/linux/mtd/mtd.h b/include/linux/mtd/mtd.h
index 11d8e68..5326435 100644
--- a/include/linux/mtd/mtd.h
+++ b/include/linux/mtd/mtd.h
@@ -60,9 +60,7 @@ struct mtd_erase_region_info {
  * MTD_OOB_PLACE:	oob data are placed at the given offset
  * MTD_OOB_AUTO:	oob data are automatically placed at the free areas
  *			which are defined by the ecclayout
- * MTD_OOB_RAW:		mode to read raw data+oob in one chunk. The oob data
- *			is inserted into the data. Thats a raw image of the
- *			flash contents.
+ * MTD_OOB_RAW:		mode to read oob and data without doing ECC checking
  */
 typedef enum {
 	MTD_OOB_PLACE,
-- 
1.6.3.3

Re: [PATCH 10/15] MTD: nand: make reads using MTD_OOB_RAW affect only ECC validation

[Thomas Gleixner]

On Mon, 22 Feb 2010, Maxim Levitsky wrote:

> This changes the behavier of MTD_OOB_RAW. It used to read both OOB and data
> to the data buffer, however you would still need to specify the dummy oob buffer.
> 
> This is only used in one place, but makes it hard to read data+oob without ECC
> test, thus I removed that behavier, and fixed the user.
> 
> Now MTD_OOB_RAW behaves just like MTD_OOB_PLACE, but doesn't do ECC validation

Is this tested against existing user space tools like nanddump ? Can I
still get the raw data from flash ?

Thanks,

	tglx
 

Re: [PATCH 10/15] MTD: nand: make reads using MTD_OOB_RAW affect only ECC validation

[Thomas Gleixner]

On Mon, 22 Feb 2010, Maxim Levitsky wrote:

> This changes the behavier of MTD_OOB_RAW. It used to read both OOB and data
> to the data buffer, however you would still need to specify the dummy oob buffer.
> 
> This is only used in one place, but makes it hard to read data+oob without ECC
> test, thus I removed that behavier, and fixed the user.
> 
> Now MTD_OOB_RAW behaves just like MTD_OOB_PLACE, but doesn't do ECC validation

Is this tested against existing user space tools like nanddump ? Can I
still get the raw data from flash ?

Thanks,

	tglx
 

Re: [PATCH 10/15] MTD: nand: make reads using MTD_OOB_RAW affect only ECC validation

[Maxim Levitsky]

On Mon, 2010-02-22 at 22:20 +0100, Thomas Gleixner wrote: 
> On Mon, 22 Feb 2010, Maxim Levitsky wrote:
> 
> > This changes the behavier of MTD_OOB_RAW. It used to read both OOB and data
> > to the data buffer, however you would still need to specify the dummy oob buffer.
> > 
> > This is only used in one place, but makes it hard to read data+oob without ECC
> > test, thus I removed that behavier, and fixed the user.
> > 
> > Now MTD_OOB_RAW behaves just like MTD_OOB_PLACE, but doesn't do ECC validation
> 
> Is this tested against existing user space tools like nanddump ? Can I
> still get the raw data from flash ?

Thats the point.
Userspace doesn't/can't use that mode.
It is not exposed through mtdchar.
Userspace reads the page, and then reads the oob.

It does use MTD_OOB_RAW, but without data buffer, and this path I don't
change.

The only user of this, is the nand itself, when it reads bad block
table.

I confess that I didn't run test that I ported this code correctly.
But I did logically verified many times that the new code works just
like old one.


Best regards,
Maxim Levitsky

Re: [PATCH 10/15] MTD: nand: make reads using MTD_OOB_RAW affect only ECC validation

[Maxim Levitsky]

On Mon, 2010-02-22 at 22:20 +0100, Thomas Gleixner wrote: 
> On Mon, 22 Feb 2010, Maxim Levitsky wrote:
> 
> > This changes the behavier of MTD_OOB_RAW. It used to read both OOB and data
> > to the data buffer, however you would still need to specify the dummy oob buffer.
> > 
> > This is only used in one place, but makes it hard to read data+oob without ECC
> > test, thus I removed that behavier, and fixed the user.
> > 
> > Now MTD_OOB_RAW behaves just like MTD_OOB_PLACE, but doesn't do ECC validation
> 
> Is this tested against existing user space tools like nanddump ? Can I
> still get the raw data from flash ?

Thats the point.
Userspace doesn't/can't use that mode.
It is not exposed through mtdchar.
Userspace reads the page, and then reads the oob.

It does use MTD_OOB_RAW, but without data buffer, and this path I don't
change.

The only user of this, is the nand itself, when it reads bad block
table.

I confess that I didn't run test that I ported this code correctly.
But I did logically verified many times that the new code works just
like old one.


Best regards,
Maxim Levitsky

Re: [PATCH 10/15] MTD: nand: make reads using MTD_OOB_RAW affect only ECC validation

[Thomas Gleixner]

On Mon, 22 Feb 2010, Maxim Levitsky wrote:
> On Mon, 2010-02-22 at 22:20 +0100, Thomas Gleixner wrote: 
> > On Mon, 22 Feb 2010, Maxim Levitsky wrote:
> > 
> > > This changes the behavier of MTD_OOB_RAW. It used to read both OOB and data
> > > to the data buffer, however you would still need to specify the dummy oob buffer.
> > > 
> > > This is only used in one place, but makes it hard to read data+oob without ECC
> > > test, thus I removed that behavier, and fixed the user.
> > > 
> > > Now MTD_OOB_RAW behaves just like MTD_OOB_PLACE, but doesn't do ECC validation
> > 
> > Is this tested against existing user space tools like nanddump ? Can I
> > still get the raw data from flash ?
> 
> Thats the point.
> Userspace doesn't/can't use that mode.
> It is not exposed through mtdchar.
> Userspace reads the page, and then reads the oob.
> 
> It does use MTD_OOB_RAW, but without data buffer, and this path I don't
> change.
> 
> The only user of this, is the nand itself, when it reads bad block
> table.
> 
> I confess that I didn't run test that I ported this code correctly.
> But I did logically verified many times that the new code works just
> like old one.

Well, then it's about time to run the tests :)

Thanks,

      tglx

Re: [PATCH 10/15] MTD: nand: make reads using MTD_OOB_RAW affect only ECC validation

[Thomas Gleixner]

On Mon, 22 Feb 2010, Maxim Levitsky wrote:
> On Mon, 2010-02-22 at 22:20 +0100, Thomas Gleixner wrote: 
> > On Mon, 22 Feb 2010, Maxim Levitsky wrote:
> > 
> > > This changes the behavier of MTD_OOB_RAW. It used to read both OOB and data
> > > to the data buffer, however you would still need to specify the dummy oob buffer.
> > > 
> > > This is only used in one place, but makes it hard to read data+oob without ECC
> > > test, thus I removed that behavier, and fixed the user.
> > > 
> > > Now MTD_OOB_RAW behaves just like MTD_OOB_PLACE, but doesn't do ECC validation
> > 
> > Is this tested against existing user space tools like nanddump ? Can I
> > still get the raw data from flash ?
> 
> Thats the point.
> Userspace doesn't/can't use that mode.
> It is not exposed through mtdchar.
> Userspace reads the page, and then reads the oob.
> 
> It does use MTD_OOB_RAW, but without data buffer, and this path I don't
> change.
> 
> The only user of this, is the nand itself, when it reads bad block
> table.
> 
> I confess that I didn't run test that I ported this code correctly.
> But I did logically verified many times that the new code works just
> like old one.

Well, then it's about time to run the tests :)

Thanks,

      tglx

Re: [PATCH 10/15] MTD: nand: make reads using MTD_OOB_RAW affect only ECC validation

[Maxim Levitsky]

On Mon, 2010-02-22 at 22:29 +0100, Thomas Gleixner wrote: 
> On Mon, 22 Feb 2010, Maxim Levitsky wrote:
> > On Mon, 2010-02-22 at 22:20 +0100, Thomas Gleixner wrote: 
> > > On Mon, 22 Feb 2010, Maxim Levitsky wrote:
> > > 
> > > > This changes the behavier of MTD_OOB_RAW. It used to read both OOB and data
> > > > to the data buffer, however you would still need to specify the dummy oob buffer.
> > > > 
> > > > This is only used in one place, but makes it hard to read data+oob without ECC
> > > > test, thus I removed that behavier, and fixed the user.
> > > > 
> > > > Now MTD_OOB_RAW behaves just like MTD_OOB_PLACE, but doesn't do ECC validation
> > > 
> > > Is this tested against existing user space tools like nanddump ? Can I
> > > still get the raw data from flash ?
> > 
> > Thats the point.
> > Userspace doesn't/can't use that mode.
> > It is not exposed through mtdchar.
> > Userspace reads the page, and then reads the oob.
> > 
> > It does use MTD_OOB_RAW, but without data buffer, and this path I don't
> > change.
> > 
> > The only user of this, is the nand itself, when it reads bad block
> > table.
> > 
> > I confess that I didn't run test that I ported this code correctly.
> > But I did logically verified many times that the new code works just
> > like old one.
> 
> Well, then it's about time to run the tests :)

Can this be done with nandsim?

Best regards,
Maxim Levitsky

Re: [PATCH 10/15] MTD: nand: make reads using MTD_OOB_RAW affect only ECC validation

[Maxim Levitsky]

On Mon, 2010-02-22 at 22:29 +0100, Thomas Gleixner wrote: 
> On Mon, 22 Feb 2010, Maxim Levitsky wrote:
> > On Mon, 2010-02-22 at 22:20 +0100, Thomas Gleixner wrote: 
> > > On Mon, 22 Feb 2010, Maxim Levitsky wrote:
> > > 
> > > > This changes the behavier of MTD_OOB_RAW. It used to read both OOB and data
> > > > to the data buffer, however you would still need to specify the dummy oob buffer.
> > > > 
> > > > This is only used in one place, but makes it hard to read data+oob without ECC
> > > > test, thus I removed that behavier, and fixed the user.
> > > > 
> > > > Now MTD_OOB_RAW behaves just like MTD_OOB_PLACE, but doesn't do ECC validation
> > > 
> > > Is this tested against existing user space tools like nanddump ? Can I
> > > still get the raw data from flash ?
> > 
> > Thats the point.
> > Userspace doesn't/can't use that mode.
> > It is not exposed through mtdchar.
> > Userspace reads the page, and then reads the oob.
> > 
> > It does use MTD_OOB_RAW, but without data buffer, and this path I don't
> > change.
> > 
> > The only user of this, is the nand itself, when it reads bad block
> > table.
> > 
> > I confess that I didn't run test that I ported this code correctly.
> > But I did logically verified many times that the new code works just
> > like old one.
> 
> Well, then it's about time to run the tests :)

Can this be done with nandsim?

Best regards,
Maxim Levitsky

Re: [PATCH 10/15] MTD: nand: make reads using MTD_OOB_RAW affect only ECC validation

[Maxim Levitsky]

On Mon, 2010-02-22 at 23:34 +0200, Maxim Levitsky wrote: 
> On Mon, 2010-02-22 at 22:29 +0100, Thomas Gleixner wrote: 
> > On Mon, 22 Feb 2010, Maxim Levitsky wrote:
> > > On Mon, 2010-02-22 at 22:20 +0100, Thomas Gleixner wrote: 
> > > > On Mon, 22 Feb 2010, Maxim Levitsky wrote:
> > > > 
> > > > > This changes the behavier of MTD_OOB_RAW. It used to read both OOB and data
> > > > > to the data buffer, however you would still need to specify the dummy oob buffer.
> > > > > 
> > > > > This is only used in one place, but makes it hard to read data+oob without ECC
> > > > > test, thus I removed that behavier, and fixed the user.
> > > > > 
> > > > > Now MTD_OOB_RAW behaves just like MTD_OOB_PLACE, but doesn't do ECC validation
> > > > 
> > > > Is this tested against existing user space tools like nanddump ? Can I
> > > > still get the raw data from flash ?
> > > 
> > > Thats the point.
> > > Userspace doesn't/can't use that mode.
> > > It is not exposed through mtdchar.
> > > Userspace reads the page, and then reads the oob.
> > > 
> > > It does use MTD_OOB_RAW, but without data buffer, and this path I don't
> > > change.
> > > 
> > > The only user of this, is the nand itself, when it reads bad block
> > > table.
> > > 
> > > I confess that I didn't run test that I ported this code correctly.
> > > But I did logically verified many times that the new code works just
> > > like old one.
Note that only this patch is in question, and only . Rest are tested
with my driver, 

While doing some testing, I tried to use ubi, but it appears to try 256
byte writes (subpage write I guess), and it fails _without_ my patches:


[  723.478676] UBI: attached mtd0 to ubi0
[  723.478686] UBI: MTD device name:            "NAND simulator partition 0"
[  723.478701] UBI: MTD device size:            8 MiB
[  723.478711] UBI: number of good PEBs:        1024
[  723.478720] UBI: number of bad PEBs:         0
[  723.478730] UBI: max. allowed volumes:       44
[  723.478739] UBI: wear-leveling threshold:    4096
[  723.478749] UBI: number of internal volumes: 1
[  723.478758] UBI: number of user volumes:     0
[  723.478767] UBI: available PEBs:             1010
[  723.478777] UBI: total number of reserved PEBs: 14
[  723.478787] UBI: number of PEBs reserved for bad PEB handling: 10
[  723.478800] UBI: max/mean erase counter: 0/0
[  723.478809] UBI: image sequence number: 0
[  723.480508] UBI: background thread "ubi_bgt0d" started, PID 4556
[  757.904206] UBI DBG (pid 4580): ubi_cdev_ioctl: create volume
[  757.904231] UBI DBG (pid 4580): ubi_create_volume: search for vacant volume ID
[  757.904246] UBI DBG (pid 4580): ubi_create_volume: create device 0, volume 0, 7756800 bytes, type 3, name aaa
[  757.906376] UBI error: ubi_io_write: error -5 while writing 256 bytes to PEB 0:256, written 0 bytes
[  757.906398] Pid: 4580, comm: ubimkvol Tainted: P           2.6.33-rc8-wl #25
[  757.906412] Call Trace:
[  757.906431]  [<ffffffffa0d8b1f2>] ubi_io_write+0x252/0x270 [ubi]
[  757.906447]  [<ffffffffa0d8b283>] ubi_io_write_vid_hdr+0x73/0x190 [ubi]
[  757.906464]  [<ffffffffa0d89f58>] ubi_eba_write_leb+0x198/0x7b0 [ubi]
[  757.906481]  [<ffffffff81350c90>] ? _raw_spin_unlock+0x30/0x60
[  757.906497]  [<ffffffffa0d892b1>] ? leb_write_unlock+0xa1/0xf0 [ubi]
[  757.906515]  [<ffffffffa0d82483>] ubi_change_vtbl_record+0x103/0x1c0 [ubi]
[  757.906534]  [<ffffffff81234474>] ? device_create_file+0x14/0x20
[  757.906552]  [<ffffffffa0d837ee>] ubi_create_volume+0x4ce/0x710 [ubi]
[  757.906569]  [<ffffffff8103418b>] ? release_console_sem+0x1eb/0x240
[  757.906586]  [<ffffffffa0d86589>] ubi_cdev_ioctl+0x2e9/0xb50 [ubi]
[  757.906603]  [<ffffffff810bf399>] ? __dentry_open+0x149/0x330
[  757.906619]  [<ffffffff810d0658>] vfs_ioctl+0x38/0xd0
[  757.906632]  [<ffffffff810d082a>] do_vfs_ioctl+0x8a/0x5b0
[  757.906645]  [<ffffffff8105e69d>] ? trace_hardirqs_on+0xd/0x10
[  757.906658]  [<ffffffff810d0d9a>] sys_ioctl+0x4a/0x80
[  757.906672]  [<ffffffff81002cab>] system_call_fastpath+0x16/0x1b
[  757.906753] UBI DBG (pid 4580): ubi_dbg_dump_flash: dumping 256 bytes of data from PEB 0, offset 256
[  757.906772] 00000000: 55 42 49 21 01 01 00 05 7f ff ef ff 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  UBI!............................
[  757.906797] 00000020: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 01 00 00 00 00 00 00 00 00 00 00 00 00 65 b3 bd 2d  ............................e..-
[  757.906821] 00000040: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................................
[  757.906846] 00000060: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................................
[  757.906874] 00000080: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................................
[  757.906898] 000000a0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................................
[  757.906920] 000000c0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................................
[  757.906943] 000000e0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................................
[  757.906977] UBI warning: ubi_eba_write_leb: failed to write VID header to LEB 2147479551:0, PEB 0
[  757.907016] UBI: run torture test for PEB 0
[  757.908132] UBI: PEB 0 passed torture test, do not mark it a bad
[  757.908211] UBI: try another PEB
[  757.908227] UBI error: ubi_io_write: error -5 while writing 256 bytes to PEB 0:256, written 0 bytes
[  757.908243] Pid: 4580, comm: ubimkvol Tainted: P           2.6.33-rc8-wl #25
[  757.908256] Call Trace:
[  757.908268]  [<ffffffffa0d8b1f2>] ubi_io_write+0x252/0x270 [ubi]
[  757.908284]  [<ffffffffa0d8b283>] ubi_io_write_vid_hdr+0x73/0x190 [ubi]
[  757.908300]  [<ffffffffa0d89f58>] ubi_eba_write_leb+0x198/0x7b0 [ubi]
[  757.908314]  [<ffffffff81350c90>] ? _raw_spin_unlock+0x30/0x60
[  757.908329]  [<ffffffffa0d892b1>] ? leb_write_unlock+0xa1/0xf0 [ubi]
[  757.908345]  [<ffffffffa0d82483>] ubi_change_vtbl_record+0x103/0x1c0 [ubi]
[  757.908359]  [<ffffffff81234474>] ? device_create_file+0x14/0x20
[  757.908374]  [<ffffffffa0d837ee>] ubi_create_volume+0x4ce/0x710 [ubi]
[  757.908388]  [<ffffffff8103418b>] ? release_console_sem+0x1eb/0x240
[  757.908404]  [<ffffffffa0d86589>] ubi_cdev_ioctl+0x2e9/0xb50 [ubi]
[  757.908418]  [<ffffffff810bf399>] ? __dentry_open+0x149/0x330
[  757.908431]  [<ffffffff810d0658>] vfs_ioctl+0x38/0xd0
[  757.908443]  [<ffffffff810d082a>] do_vfs_ioctl+0x8a/0x5b0
[  757.908455]  [<ffffffff8105e69d>] ? trace_hardirqs_on+0xd/0x10
[  757.908467]  [<ffffffff810d0d9a>] sys_ioctl+0x4a/0x80
[  757.908479]  [<ffffffff81002cab>] system_call_fastpath+0x16/0x1b
[  757.908543] UBI DBG (pid 4580): ubi_dbg_dump_flash: dumping 256 bytes of data from PEB 0, offset 256
[  757.908560] 00000000: 55 42 49 21 01 01 00 05 7f ff ef ff 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  UBI!............................
[  757.908582] 00000020: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 02 00 00 00 00 00 00 00 00 00 00 00 00 d8 79 d1 e3  .............................y..
[  757.908605] 00000040: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................................
[  757.908628] 00000060: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................................
[  757.908650] 00000080: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................................
[  757.908673] 000000a0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................................
[  757.908695] 000000c0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................................
[  757.908718] 000000e0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................................
[  757.908752] UBI warning: ubi_eba_write_leb: failed to write VID header to LEB 2147479551:0, PEB 0
[  757.908788] UBI: run torture test for PEB 0
[  757.909892] UBI: PEB 0 passed torture test, do not mark it a bad
[  757.909971] UBI: try another PEB
[  757.909987] UBI error: ubi_io_write: error -5 while writing 256 bytes to PEB 0:256, written 0 bytes
[  757.910214] Pid: 4580, comm: ubimkvol Tainted: P           2.6.33-rc8-wl #25



What I did so far:


sudo modprobe nandsim
sudo modprobe ubi
sudo ubiformat /dev/mtd0
sudo ubiattach /dev/ubi_ctrl -m0 -d 0
sudo ubimkvol -m -N test /dev/ubi0



Then I suppose I can mount /dev/ubi0 with ubifs, right?
(If there were no errors of course)


Best regards,
Maxim Levitsky

Re: [PATCH 10/15] MTD: nand: make reads using MTD_OOB_RAW affect only ECC validation

[Maxim Levitsky]

On Mon, 2010-02-22 at 23:34 +0200, Maxim Levitsky wrote: 
> On Mon, 2010-02-22 at 22:29 +0100, Thomas Gleixner wrote: 
> > On Mon, 22 Feb 2010, Maxim Levitsky wrote:
> > > On Mon, 2010-02-22 at 22:20 +0100, Thomas Gleixner wrote: 
> > > > On Mon, 22 Feb 2010, Maxim Levitsky wrote:
> > > > 
> > > > > This changes the behavier of MTD_OOB_RAW. It used to read both OOB and data
> > > > > to the data buffer, however you would still need to specify the dummy oob buffer.
> > > > > 
> > > > > This is only used in one place, but makes it hard to read data+oob without ECC
> > > > > test, thus I removed that behavier, and fixed the user.
> > > > > 
> > > > > Now MTD_OOB_RAW behaves just like MTD_OOB_PLACE, but doesn't do ECC validation
> > > > 
> > > > Is this tested against existing user space tools like nanddump ? Can I
> > > > still get the raw data from flash ?
> > > 
> > > Thats the point.
> > > Userspace doesn't/can't use that mode.
> > > It is not exposed through mtdchar.
> > > Userspace reads the page, and then reads the oob.
> > > 
> > > It does use MTD_OOB_RAW, but without data buffer, and this path I don't
> > > change.
> > > 
> > > The only user of this, is the nand itself, when it reads bad block
> > > table.
> > > 
> > > I confess that I didn't run test that I ported this code correctly.
> > > But I did logically verified many times that the new code works just
> > > like old one.
Note that only this patch is in question, and only . Rest are tested
with my driver, 

While doing some testing, I tried to use ubi, but it appears to try 256
byte writes (subpage write I guess), and it fails _without_ my patches:


[  723.478676] UBI: attached mtd0 to ubi0
[  723.478686] UBI: MTD device name:            "NAND simulator partition 0"
[  723.478701] UBI: MTD device size:            8 MiB
[  723.478711] UBI: number of good PEBs:        1024
[  723.478720] UBI: number of bad PEBs:         0
[  723.478730] UBI: max. allowed volumes:       44
[  723.478739] UBI: wear-leveling threshold:    4096
[  723.478749] UBI: number of internal volumes: 1
[  723.478758] UBI: number of user volumes:     0
[  723.478767] UBI: available PEBs:             1010
[  723.478777] UBI: total number of reserved PEBs: 14
[  723.478787] UBI: number of PEBs reserved for bad PEB handling: 10
[  723.478800] UBI: max/mean erase counter: 0/0
[  723.478809] UBI: image sequence number: 0
[  723.480508] UBI: background thread "ubi_bgt0d" started, PID 4556
[  757.904206] UBI DBG (pid 4580): ubi_cdev_ioctl: create volume
[  757.904231] UBI DBG (pid 4580): ubi_create_volume: search for vacant volume ID
[  757.904246] UBI DBG (pid 4580): ubi_create_volume: create device 0, volume 0, 7756800 bytes, type 3, name aaa
[  757.906376] UBI error: ubi_io_write: error -5 while writing 256 bytes to PEB 0:256, written 0 bytes
[  757.906398] Pid: 4580, comm: ubimkvol Tainted: P           2.6.33-rc8-wl #25
[  757.906412] Call Trace:
[  757.906431]  [<ffffffffa0d8b1f2>] ubi_io_write+0x252/0x270 [ubi]
[  757.906447]  [<ffffffffa0d8b283>] ubi_io_write_vid_hdr+0x73/0x190 [ubi]
[  757.906464]  [<ffffffffa0d89f58>] ubi_eba_write_leb+0x198/0x7b0 [ubi]
[  757.906481]  [<ffffffff81350c90>] ? _raw_spin_unlock+0x30/0x60
[  757.906497]  [<ffffffffa0d892b1>] ? leb_write_unlock+0xa1/0xf0 [ubi]
[  757.906515]  [<ffffffffa0d82483>] ubi_change_vtbl_record+0x103/0x1c0 [ubi]
[  757.906534]  [<ffffffff81234474>] ? device_create_file+0x14/0x20
[  757.906552]  [<ffffffffa0d837ee>] ubi_create_volume+0x4ce/0x710 [ubi]
[  757.906569]  [<ffffffff8103418b>] ? release_console_sem+0x1eb/0x240
[  757.906586]  [<ffffffffa0d86589>] ubi_cdev_ioctl+0x2e9/0xb50 [ubi]
[  757.906603]  [<ffffffff810bf399>] ? __dentry_open+0x149/0x330
[  757.906619]  [<ffffffff810d0658>] vfs_ioctl+0x38/0xd0
[  757.906632]  [<ffffffff810d082a>] do_vfs_ioctl+0x8a/0x5b0
[  757.906645]  [<ffffffff8105e69d>] ? trace_hardirqs_on+0xd/0x10
[  757.906658]  [<ffffffff810d0d9a>] sys_ioctl+0x4a/0x80
[  757.906672]  [<ffffffff81002cab>] system_call_fastpath+0x16/0x1b
[  757.906753] UBI DBG (pid 4580): ubi_dbg_dump_flash: dumping 256 bytes of data from PEB 0, offset 256
[  757.906772] 00000000: 55 42 49 21 01 01 00 05 7f ff ef ff 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  UBI!............................
[  757.906797] 00000020: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 01 00 00 00 00 00 00 00 00 00 00 00 00 65 b3 bd 2d  ............................e..-
[  757.906821] 00000040: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................................
[  757.906846] 00000060: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................................
[  757.906874] 00000080: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................................
[  757.906898] 000000a0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................................
[  757.906920] 000000c0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................................
[  757.906943] 000000e0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................................
[  757.906977] UBI warning: ubi_eba_write_leb: failed to write VID header to LEB 2147479551:0, PEB 0
[  757.907016] UBI: run torture test for PEB 0
[  757.908132] UBI: PEB 0 passed torture test, do not mark it a bad
[  757.908211] UBI: try another PEB
[  757.908227] UBI error: ubi_io_write: error -5 while writing 256 bytes to PEB 0:256, written 0 bytes
[  757.908243] Pid: 4580, comm: ubimkvol Tainted: P           2.6.33-rc8-wl #25
[  757.908256] Call Trace:
[  757.908268]  [<ffffffffa0d8b1f2>] ubi_io_write+0x252/0x270 [ubi]
[  757.908284]  [<ffffffffa0d8b283>] ubi_io_write_vid_hdr+0x73/0x190 [ubi]
[  757.908300]  [<ffffffffa0d89f58>] ubi_eba_write_leb+0x198/0x7b0 [ubi]
[  757.908314]  [<ffffffff81350c90>] ? _raw_spin_unlock+0x30/0x60
[  757.908329]  [<ffffffffa0d892b1>] ? leb_write_unlock+0xa1/0xf0 [ubi]
[  757.908345]  [<ffffffffa0d82483>] ubi_change_vtbl_record+0x103/0x1c0 [ubi]
[  757.908359]  [<ffffffff81234474>] ? device_create_file+0x14/0x20
[  757.908374]  [<ffffffffa0d837ee>] ubi_create_volume+0x4ce/0x710 [ubi]
[  757.908388]  [<ffffffff8103418b>] ? release_console_sem+0x1eb/0x240
[  757.908404]  [<ffffffffa0d86589>] ubi_cdev_ioctl+0x2e9/0xb50 [ubi]
[  757.908418]  [<ffffffff810bf399>] ? __dentry_open+0x149/0x330
[  757.908431]  [<ffffffff810d0658>] vfs_ioctl+0x38/0xd0
[  757.908443]  [<ffffffff810d082a>] do_vfs_ioctl+0x8a/0x5b0
[  757.908455]  [<ffffffff8105e69d>] ? trace_hardirqs_on+0xd/0x10
[  757.908467]  [<ffffffff810d0d9a>] sys_ioctl+0x4a/0x80
[  757.908479]  [<ffffffff81002cab>] system_call_fastpath+0x16/0x1b
[  757.908543] UBI DBG (pid 4580): ubi_dbg_dump_flash: dumping 256 bytes of data from PEB 0, offset 256
[  757.908560] 00000000: 55 42 49 21 01 01 00 05 7f ff ef ff 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  UBI!............................
[  757.908582] 00000020: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 02 00 00 00 00 00 00 00 00 00 00 00 00 d8 79 d1 e3  .............................y..
[  757.908605] 00000040: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................................
[  757.908628] 00000060: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................................
[  757.908650] 00000080: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................................
[  757.908673] 000000a0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................................
[  757.908695] 000000c0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................................
[  757.908718] 000000e0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................................
[  757.908752] UBI warning: ubi_eba_write_leb: failed to write VID header to LEB 2147479551:0, PEB 0
[  757.908788] UBI: run torture test for PEB 0
[  757.909892] UBI: PEB 0 passed torture test, do not mark it a bad
[  757.909971] UBI: try another PEB
[  757.909987] UBI error: ubi_io_write: error -5 while writing 256 bytes to PEB 0:256, written 0 bytes
[  757.910214] Pid: 4580, comm: ubimkvol Tainted: P           2.6.33-rc8-wl #25



What I did so far:


sudo modprobe nandsim
sudo modprobe ubi
sudo ubiformat /dev/mtd0
sudo ubiattach /dev/ubi_ctrl -m0 -d 0
sudo ubimkvol -m -N test /dev/ubi0



Then I suppose I can mount /dev/ubi0 with ubifs, right?
(If there were no errors of course)


Best regards,
Maxim Levitsky

Re: [PATCH 10/15] MTD: nand: make reads using MTD_OOB_RAW affect only ECC validation

[Maxim Levitsky]

On Mon, 2010-02-22 at 22:29 +0100, Thomas Gleixner wrote: 
> On Mon, 22 Feb 2010, Maxim Levitsky wrote:
> > On Mon, 2010-02-22 at 22:20 +0100, Thomas Gleixner wrote: 
> > > On Mon, 22 Feb 2010, Maxim Levitsky wrote:
> > > 
> > > > This changes the behavier of MTD_OOB_RAW. It used to read both OOB and data
> > > > to the data buffer, however you would still need to specify the dummy oob buffer.
> > > > 
> > > > This is only used in one place, but makes it hard to read data+oob without ECC
> > > > test, thus I removed that behavier, and fixed the user.
> > > > 
> > > > Now MTD_OOB_RAW behaves just like MTD_OOB_PLACE, but doesn't do ECC validation
> > > 
> > > Is this tested against existing user space tools like nanddump ? Can I
> > > still get the raw data from flash ?
> > 
> > Thats the point.
> > Userspace doesn't/can't use that mode.
> > It is not exposed through mtdchar.
> > Userspace reads the page, and then reads the oob.
> > 
> > It does use MTD_OOB_RAW, but without data buffer, and this path I don't
> > change.
> > 
> > The only user of this, is the nand itself, when it reads bad block
> > table.
> > 
> > I confess that I didn't run test that I ported this code correctly.
> > But I did logically verified many times that the new code works just
> > like old one.
> 
> Well, then it's about time to run the tests :)
> 
> Thanks,

Just to be sure, I tested jffs2 and nanddump.

Works just fine.

Best regards,
Maxim Levitsky

Re: [PATCH 10/15] MTD: nand: make reads using MTD_OOB_RAW affect only ECC validation

[Maxim Levitsky]

On Mon, 2010-02-22 at 22:29 +0100, Thomas Gleixner wrote: 
> On Mon, 22 Feb 2010, Maxim Levitsky wrote:
> > On Mon, 2010-02-22 at 22:20 +0100, Thomas Gleixner wrote: 
> > > On Mon, 22 Feb 2010, Maxim Levitsky wrote:
> > > 
> > > > This changes the behavier of MTD_OOB_RAW. It used to read both OOB and data
> > > > to the data buffer, however you would still need to specify the dummy oob buffer.
> > > > 
> > > > This is only used in one place, but makes it hard to read data+oob without ECC
> > > > test, thus I removed that behavier, and fixed the user.
> > > > 
> > > > Now MTD_OOB_RAW behaves just like MTD_OOB_PLACE, but doesn't do ECC validation
> > > 
> > > Is this tested against existing user space tools like nanddump ? Can I
> > > still get the raw data from flash ?
> > 
> > Thats the point.
> > Userspace doesn't/can't use that mode.
> > It is not exposed through mtdchar.
> > Userspace reads the page, and then reads the oob.
> > 
> > It does use MTD_OOB_RAW, but without data buffer, and this path I don't
> > change.
> > 
> > The only user of this, is the nand itself, when it reads bad block
> > table.
> > 
> > I confess that I didn't run test that I ported this code correctly.
> > But I did logically verified many times that the new code works just
> > like old one.
> 
> Well, then it's about time to run the tests :)
> 
> Thanks,

Just to be sure, I tested jffs2 and nanddump.

Works just fine.

Best regards,
Maxim Levitsky

[PATCH 11/15] MTD: nand: add ->badblockbits to specify the minimum number of bits in bad block byte to consider the block good

[Maxim Levitsky]

This can be used to protect against bitflips in that field, but now mostly
for smartmedia.

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/nand/nand_base.c |   13 +++++++++----
 include/linux/mtd/nand.h     |    1 +
 2 files changed, 10 insertions(+), 4 deletions(-)

diff --git a/drivers/mtd/nand/nand_base.c b/drivers/mtd/nand/nand_base.c
index 51dfea1..ba29a29 100644
--- a/drivers/mtd/nand/nand_base.c
+++ b/drivers/mtd/nand/nand_base.c
@@ -364,14 +364,18 @@ static int nand_block_bad(struct mtd_info *mtd, loff_t ofs, int getchip)
 		bad = cpu_to_le16(chip->read_word(mtd));
 		if (chip->badblockpos & 0x1)
 			bad >>= 8;
-		if ((bad & 0xFF) != 0xff)
-			res = 1;
+		else
+			bad &= 0xFF;
 	} else {
 		chip->cmdfunc(mtd, NAND_CMD_READOOB, chip->badblockpos, page);
-		if (chip->read_byte(mtd) != 0xff)
-			res = 1;
+		bad = chip->read_byte(mtd);
 	}
 
+	if (likely(chip->badblockbits == 8))
+		res = bad != 0xFF;
+	else
+		res = hweight8(bad) < chip->badblockbits;
+
 	if (getchip)
 		nand_release_device(mtd);
 
@@ -2884,6 +2888,7 @@ static struct nand_flash_dev *nand_get_flash_type(struct mtd_info *mtd,
 	/* Set the bad block position */
 	chip->badblockpos = mtd->writesize > 512 ?
 		NAND_LARGE_BADBLOCK_POS : NAND_SMALL_BADBLOCK_POS;
+	chip->badblockbits = 8;
 
 	/* Get chip options, preserve non chip based options */
 	chip->options &= ~NAND_CHIPOPTIONS_MSK;
diff --git a/include/linux/mtd/nand.h b/include/linux/mtd/nand.h
index 48bc2c5..f2d4a1a 100644
--- a/include/linux/mtd/nand.h
+++ b/include/linux/mtd/nand.h
@@ -401,6 +401,7 @@ struct nand_chip {
 	int		subpagesize;
 	uint8_t		cellinfo;
 	int		badblockpos;
+	int		badblockbits;
 
 	flstate_t	state;
 
-- 
1.6.3.3

[PATCH 11/15] MTD: nand: add ->badblockbits to specify the minimum number of bits in bad block byte to consider the block good

[Maxim Levitsky]

This can be used to protect against bitflips in that field, but now mostly
for smartmedia.

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/nand/nand_base.c |   13 +++++++++----
 include/linux/mtd/nand.h     |    1 +
 2 files changed, 10 insertions(+), 4 deletions(-)

diff --git a/drivers/mtd/nand/nand_base.c b/drivers/mtd/nand/nand_base.c
index 51dfea1..ba29a29 100644
--- a/drivers/mtd/nand/nand_base.c
+++ b/drivers/mtd/nand/nand_base.c
@@ -364,14 +364,18 @@ static int nand_block_bad(struct mtd_info *mtd, loff_t ofs, int getchip)
 		bad = cpu_to_le16(chip->read_word(mtd));
 		if (chip->badblockpos & 0x1)
 			bad >>= 8;
-		if ((bad & 0xFF) != 0xff)
-			res = 1;
+		else
+			bad &= 0xFF;
 	} else {
 		chip->cmdfunc(mtd, NAND_CMD_READOOB, chip->badblockpos, page);
-		if (chip->read_byte(mtd) != 0xff)
-			res = 1;
+		bad = chip->read_byte(mtd);
 	}
 
+	if (likely(chip->badblockbits == 8))
+		res = bad != 0xFF;
+	else
+		res = hweight8(bad) < chip->badblockbits;
+
 	if (getchip)
 		nand_release_device(mtd);
 
@@ -2884,6 +2888,7 @@ static struct nand_flash_dev *nand_get_flash_type(struct mtd_info *mtd,
 	/* Set the bad block position */
 	chip->badblockpos = mtd->writesize > 512 ?
 		NAND_LARGE_BADBLOCK_POS : NAND_SMALL_BADBLOCK_POS;
+	chip->badblockbits = 8;
 
 	/* Get chip options, preserve non chip based options */
 	chip->options &= ~NAND_CHIPOPTIONS_MSK;
diff --git a/include/linux/mtd/nand.h b/include/linux/mtd/nand.h
index 48bc2c5..f2d4a1a 100644
--- a/include/linux/mtd/nand.h
+++ b/include/linux/mtd/nand.h
@@ -401,6 +401,7 @@ struct nand_chip {
 	int		subpagesize;
 	uint8_t		cellinfo;
 	int		badblockpos;
+	int		badblockbits;
 
 	flstate_t	state;
 
-- 
1.6.3.3

[PATCH 12/15] MTD: common module for smartmedia/xD support

[Maxim Levitsky]

This small module implements few helpers that are usefull
for nand drivers for SmartMedia/xD card readers.

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/nand/Kconfig     |    9 ++++
 drivers/mtd/nand/Makefile    |    1 +
 drivers/mtd/nand/sm_common.c |  107 ++++++++++++++++++++++++++++++++++++++++++
 drivers/mtd/nand/sm_common.h |   61 ++++++++++++++++++++++++
 4 files changed, 178 insertions(+), 0 deletions(-)
 create mode 100644 drivers/mtd/nand/sm_common.c
 create mode 100644 drivers/mtd/nand/sm_common.h

diff --git a/drivers/mtd/nand/Kconfig b/drivers/mtd/nand/Kconfig
index 22cee48..31273c7 100644
--- a/drivers/mtd/nand/Kconfig
+++ b/drivers/mtd/nand/Kconfig
@@ -18,6 +18,10 @@ config MTD_NAND_VERIFY_WRITE
 	  device thinks the write was successful, a bit could have been
 	  flipped accidentally due to device wear or something else.
 
+config MTD_NAND_SMARTMEDIA
+	boolean
+	default n
+
 config MTD_NAND_ECC_SMC
 	bool "NAND ECC Smart Media byte order"
 	default n
@@ -25,6 +29,11 @@ config MTD_NAND_ECC_SMC
 	  Software ECC according to the Smart Media Specification.
 	  The original Linux implementation had byte 0 and 1 swapped.
 
+config MTD_SM_COMMON
+	select MTD_NAND_SMARTMEDIA
+	tristate
+	default n
+
 config MTD_NAND_MUSEUM_IDS
 	bool "Enable chip ids for obsolete ancient NAND devices"
 	depends on MTD_NAND
diff --git a/drivers/mtd/nand/Makefile b/drivers/mtd/nand/Makefile
index 39b4af2..e699bef 100644
--- a/drivers/mtd/nand/Makefile
+++ b/drivers/mtd/nand/Makefile
@@ -4,6 +4,7 @@
 
 obj-$(CONFIG_MTD_NAND)			+= nand.o nand_ecc.o
 obj-$(CONFIG_MTD_NAND_IDS)		+= nand_ids.o
+obj-$(CONFIG_MTD_SM_COMMON) 		+= sm_common.o
 
 obj-$(CONFIG_MTD_NAND_CAFE)		+= cafe_nand.o
 obj-$(CONFIG_MTD_NAND_SPIA)		+= spia.o
diff --git a/drivers/mtd/nand/sm_common.c b/drivers/mtd/nand/sm_common.c
new file mode 100644
index 0000000..fc4ff71
--- /dev/null
+++ b/drivers/mtd/nand/sm_common.c
@@ -0,0 +1,107 @@
+/*
+ * Copyright (C) 2009 - Maxim Levitsky
+ * Common routines & support for xD format
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include <linux/kernel.h>
+#include <linux/mtd/nand.h>
+#include "sm_common.h"
+
+static struct nand_ecclayout nand_oob_sm = {
+	.eccbytes = 6,
+	.eccpos = {8, 9, 10, 13, 14, 15},
+	.oobfree = {
+		{.offset = 0 , .length = 4}, /* reserved */
+		{.offset = 6 , .length = 2}, /* LBA1 */
+		{.offset = 11, .length = 2}  /* LBA2 */
+	}
+};
+
+/* NOTE: This layout is is not compatabable with SmartMedia, */
+/* because the 256 byte devices have page depenent oob layout */
+/* However it does preserve the bad block markers */
+/* If you use smftl, it will bypass this and work correctly */
+/* If you not, then you break SmartMedia compliance anyway */
+
+static struct nand_ecclayout nand_oob_sm_small = {
+	.eccbytes = 3,
+	.eccpos = {0, 1, 2},
+	.oobfree = {
+		{.offset = 3 , .length = 2}, /* reserved */
+		{.offset = 6 , .length = 2}, /* LBA1 */
+	}
+};
+
+
+static int sm_block_markbad(struct mtd_info *mtd, loff_t ofs)
+{
+	struct mtd_oob_ops ops;
+	struct sm_oob oob;
+	int ret, error = 0;
+
+	memset(&oob, -1, SM_OOB_SIZE);
+	oob.block_status = 0x0F;
+
+	/* As long as this function is called on erase block boundaries
+		it will work correctly for 256 byte nand */
+	ops.mode = MTD_OOB_PLACE;
+	ops.ooboffs = 0;
+	ops.ooblen = mtd->oobsize;
+	ops.oobbuf = (void *)&oob;
+	ops.datbuf = NULL;
+
+
+	ret = mtd->write_oob(mtd, ofs, &ops);
+	if (ret < 0 || ops.oobretlen != SM_OOB_SIZE) {
+		printk(KERN_NOTICE
+			"sm_common: can't mark sector at %i as bad\n",
+								(int)ofs);
+		error = -EIO;
+	} else
+		mtd->ecc_stats.badblocks++;
+
+	return error;
+}
+
+
+int sm_register_device(struct mtd_info *mtd)
+{
+	struct nand_chip *chip = (struct nand_chip *)mtd->priv;
+	int ret;
+
+	chip->options |= NAND_SKIP_BBTSCAN | NAND_SMARTMEDIA;
+
+	/* Scan for card properties */
+	ret = nand_scan_ident(mtd, 1);
+
+	if (ret)
+		return ret;
+
+	/* Bad block marker postion */
+	chip->badblockpos = 0x05;
+	chip->badblockbits = 7;
+	chip->block_markbad = sm_block_markbad;
+
+	/* ECC layout */
+	if (mtd->writesize == SM_SECTOR_SIZE)
+		chip->ecc.layout = &nand_oob_sm;
+	else if (mtd->writesize == SM_SMALL_PAGE)
+		chip->ecc.layout = &nand_oob_sm_small;
+	else
+		return -ENODEV;
+
+	ret = nand_scan_tail(mtd);
+
+	if (ret)
+		return ret;
+
+	return add_mtd_device(mtd);
+}
+EXPORT_SYMBOL_GPL(sm_register_device);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Maxim Levitsky <maximlevitsky@gmail.com>");
+MODULE_DESCRIPTION("Common SmartMedia/xD functions");
diff --git a/drivers/mtd/nand/sm_common.h b/drivers/mtd/nand/sm_common.h
new file mode 100644
index 0000000..2e6b517
--- /dev/null
+++ b/drivers/mtd/nand/sm_common.h
@@ -0,0 +1,61 @@
+/*
+ * Copyright (C) 2009 - Maxim Levitsky
+ * Common routines & support for SmartMedia/xD format
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include <linux/bitops.h>
+#include <linux/mtd/mtd.h>
+
+/* Full oob structure as written on the flash */
+struct sm_oob {
+	u32 reserved;
+	u8  data_status;
+	u8  block_status;
+	u8 lba_copy1[2];
+	u8  ecc2[3];
+	u8 lba_copy2[2];
+	u8  ecc1[3];
+} __attribute__((packed));
+
+
+/* one sector is always 512 bytes, but it can consist of two nand pages */
+#define SM_SECTOR_SIZE		512
+
+/* oob area is also 16 bytes, but might be from two pages */
+#define SM_OOB_SIZE		16
+
+/* This is maximum zone size, and all devices that have more that one zone
+   have this size */
+#define SM_MAX_ZONE_SIZE 	1024
+
+/* support for small page nand */
+#define SM_SMALL_PAGE 		256
+#define SM_SMALL_OOB_SIZE	8
+
+
+extern int sm_register_device(struct mtd_info *mtd);
+
+
+inline int sm_sector_valid(struct sm_oob *oob)
+{
+	return hweight16(oob->data_status) >= 5;
+}
+
+inline int sm_block_valid(struct sm_oob *oob)
+{
+	return hweight16(oob->block_status) >= 7;
+}
+
+inline int sm_block_erased(struct sm_oob *oob)
+{
+	static const u32 erased_pattern[4] = {
+		0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF };
+
+	/* First test for erased block */
+	if (!memcmp(oob, erased_pattern, sizeof(*oob)))
+		return 1;
+	return 0;
+}
-- 
1.6.3.3

[PATCH 12/15] MTD: common module for smartmedia/xD support

[Maxim Levitsky]

This small module implements few helpers that are usefull
for nand drivers for SmartMedia/xD card readers.

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/nand/Kconfig     |    9 ++++
 drivers/mtd/nand/Makefile    |    1 +
 drivers/mtd/nand/sm_common.c |  107 ++++++++++++++++++++++++++++++++++++++++++
 drivers/mtd/nand/sm_common.h |   61 ++++++++++++++++++++++++
 4 files changed, 178 insertions(+), 0 deletions(-)
 create mode 100644 drivers/mtd/nand/sm_common.c
 create mode 100644 drivers/mtd/nand/sm_common.h

diff --git a/drivers/mtd/nand/Kconfig b/drivers/mtd/nand/Kconfig
index 22cee48..31273c7 100644
--- a/drivers/mtd/nand/Kconfig
+++ b/drivers/mtd/nand/Kconfig
@@ -18,6 +18,10 @@ config MTD_NAND_VERIFY_WRITE
 	  device thinks the write was successful, a bit could have been
 	  flipped accidentally due to device wear or something else.
 
+config MTD_NAND_SMARTMEDIA
+	boolean
+	default n
+
 config MTD_NAND_ECC_SMC
 	bool "NAND ECC Smart Media byte order"
 	default n
@@ -25,6 +29,11 @@ config MTD_NAND_ECC_SMC
 	  Software ECC according to the Smart Media Specification.
 	  The original Linux implementation had byte 0 and 1 swapped.
 
+config MTD_SM_COMMON
+	select MTD_NAND_SMARTMEDIA
+	tristate
+	default n
+
 config MTD_NAND_MUSEUM_IDS
 	bool "Enable chip ids for obsolete ancient NAND devices"
 	depends on MTD_NAND
diff --git a/drivers/mtd/nand/Makefile b/drivers/mtd/nand/Makefile
index 39b4af2..e699bef 100644
--- a/drivers/mtd/nand/Makefile
+++ b/drivers/mtd/nand/Makefile
@@ -4,6 +4,7 @@
 
 obj-$(CONFIG_MTD_NAND)			+= nand.o nand_ecc.o
 obj-$(CONFIG_MTD_NAND_IDS)		+= nand_ids.o
+obj-$(CONFIG_MTD_SM_COMMON) 		+= sm_common.o
 
 obj-$(CONFIG_MTD_NAND_CAFE)		+= cafe_nand.o
 obj-$(CONFIG_MTD_NAND_SPIA)		+= spia.o
diff --git a/drivers/mtd/nand/sm_common.c b/drivers/mtd/nand/sm_common.c
new file mode 100644
index 0000000..fc4ff71
--- /dev/null
+++ b/drivers/mtd/nand/sm_common.c
@@ -0,0 +1,107 @@
+/*
+ * Copyright (C) 2009 - Maxim Levitsky
+ * Common routines & support for xD format
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include <linux/kernel.h>
+#include <linux/mtd/nand.h>
+#include "sm_common.h"
+
+static struct nand_ecclayout nand_oob_sm = {
+	.eccbytes = 6,
+	.eccpos = {8, 9, 10, 13, 14, 15},
+	.oobfree = {
+		{.offset = 0 , .length = 4}, /* reserved */
+		{.offset = 6 , .length = 2}, /* LBA1 */
+		{.offset = 11, .length = 2}  /* LBA2 */
+	}
+};
+
+/* NOTE: This layout is is not compatabable with SmartMedia, */
+/* because the 256 byte devices have page depenent oob layout */
+/* However it does preserve the bad block markers */
+/* If you use smftl, it will bypass this and work correctly */
+/* If you not, then you break SmartMedia compliance anyway */
+
+static struct nand_ecclayout nand_oob_sm_small = {
+	.eccbytes = 3,
+	.eccpos = {0, 1, 2},
+	.oobfree = {
+		{.offset = 3 , .length = 2}, /* reserved */
+		{.offset = 6 , .length = 2}, /* LBA1 */
+	}
+};
+
+
+static int sm_block_markbad(struct mtd_info *mtd, loff_t ofs)
+{
+	struct mtd_oob_ops ops;
+	struct sm_oob oob;
+	int ret, error = 0;
+
+	memset(&oob, -1, SM_OOB_SIZE);
+	oob.block_status = 0x0F;
+
+	/* As long as this function is called on erase block boundaries
+		it will work correctly for 256 byte nand */
+	ops.mode = MTD_OOB_PLACE;
+	ops.ooboffs = 0;
+	ops.ooblen = mtd->oobsize;
+	ops.oobbuf = (void *)&oob;
+	ops.datbuf = NULL;
+
+
+	ret = mtd->write_oob(mtd, ofs, &ops);
+	if (ret < 0 || ops.oobretlen != SM_OOB_SIZE) {
+		printk(KERN_NOTICE
+			"sm_common: can't mark sector at %i as bad\n",
+								(int)ofs);
+		error = -EIO;
+	} else
+		mtd->ecc_stats.badblocks++;
+
+	return error;
+}
+
+
+int sm_register_device(struct mtd_info *mtd)
+{
+	struct nand_chip *chip = (struct nand_chip *)mtd->priv;
+	int ret;
+
+	chip->options |= NAND_SKIP_BBTSCAN | NAND_SMARTMEDIA;
+
+	/* Scan for card properties */
+	ret = nand_scan_ident(mtd, 1);
+
+	if (ret)
+		return ret;
+
+	/* Bad block marker postion */
+	chip->badblockpos = 0x05;
+	chip->badblockbits = 7;
+	chip->block_markbad = sm_block_markbad;
+
+	/* ECC layout */
+	if (mtd->writesize == SM_SECTOR_SIZE)
+		chip->ecc.layout = &nand_oob_sm;
+	else if (mtd->writesize == SM_SMALL_PAGE)
+		chip->ecc.layout = &nand_oob_sm_small;
+	else
+		return -ENODEV;
+
+	ret = nand_scan_tail(mtd);
+
+	if (ret)
+		return ret;
+
+	return add_mtd_device(mtd);
+}
+EXPORT_SYMBOL_GPL(sm_register_device);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Maxim Levitsky <maximlevitsky@gmail.com>");
+MODULE_DESCRIPTION("Common SmartMedia/xD functions");
diff --git a/drivers/mtd/nand/sm_common.h b/drivers/mtd/nand/sm_common.h
new file mode 100644
index 0000000..2e6b517
--- /dev/null
+++ b/drivers/mtd/nand/sm_common.h
@@ -0,0 +1,61 @@
+/*
+ * Copyright (C) 2009 - Maxim Levitsky
+ * Common routines & support for SmartMedia/xD format
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include <linux/bitops.h>
+#include <linux/mtd/mtd.h>
+
+/* Full oob structure as written on the flash */
+struct sm_oob {
+	u32 reserved;
+	u8  data_status;
+	u8  block_status;
+	u8 lba_copy1[2];
+	u8  ecc2[3];
+	u8 lba_copy2[2];
+	u8  ecc1[3];
+} __attribute__((packed));
+
+
+/* one sector is always 512 bytes, but it can consist of two nand pages */
+#define SM_SECTOR_SIZE		512
+
+/* oob area is also 16 bytes, but might be from two pages */
+#define SM_OOB_SIZE		16
+
+/* This is maximum zone size, and all devices that have more that one zone
+   have this size */
+#define SM_MAX_ZONE_SIZE 	1024
+
+/* support for small page nand */
+#define SM_SMALL_PAGE 		256
+#define SM_SMALL_OOB_SIZE	8
+
+
+extern int sm_register_device(struct mtd_info *mtd);
+
+
+inline int sm_sector_valid(struct sm_oob *oob)
+{
+	return hweight16(oob->data_status) >= 5;
+}
+
+inline int sm_block_valid(struct sm_oob *oob)
+{
+	return hweight16(oob->block_status) >= 7;
+}
+
+inline int sm_block_erased(struct sm_oob *oob)
+{
+	static const u32 erased_pattern[4] = {
+		0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF };
+
+	/* First test for erased block */
+	if (!memcmp(oob, erased_pattern, sizeof(*oob)))
+		return 1;
+	return 0;
+}
-- 
1.6.3.3

[PATCH 13/15] MTD: add few workarounds to nand system for SmartMedia/xD chips.

[Maxim Levitsky]

* Add an option NAND_SMARTMEDIA that can be set by nand driver
 If set, it will cause separate ID table to be used, which includes
 mask rom devices and new xD cards

* Workaround for wrong write protect status on some xD cards

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/nand/nand_base.c |   32 +++++++++++++++++++++++---------
 drivers/mtd/nand/nand_ids.c  |   39 +++++++++++++++++++++++++++++++++++++++
 include/linux/mtd/nand.h     |   11 +++++++++++
 3 files changed, 73 insertions(+), 9 deletions(-)

diff --git a/drivers/mtd/nand/nand_base.c b/drivers/mtd/nand/nand_base.c
index ba29a29..fd9b00d 100644
--- a/drivers/mtd/nand/nand_base.c
+++ b/drivers/mtd/nand/nand_base.c
@@ -434,9 +434,18 @@ static int nand_default_block_markbad(struct mtd_info *mtd, loff_t ofs)
 static int nand_check_wp(struct mtd_info *mtd)
 {
 	struct nand_chip *chip = mtd->priv;
+	int wp;
+
+	/* broken xD cards report WP despite beeing writable */
+	if (chip->options & NAND_BROKEN_XD)
+		return 0;
+
 	/* Check the WP bit */
 	chip->cmdfunc(mtd, NAND_CMD_STATUS, -1, -1);
-	return (chip->read_byte(mtd) & NAND_STATUS_WP) ? 0 : 1;
+	wp = (chip->read_byte(mtd) & NAND_STATUS_WP) ? 0 : 1;
+
+
+	return wp;
 }
 
 /**
@@ -2769,7 +2778,7 @@ static struct nand_flash_dev *nand_get_flash_type(struct mtd_info *mtd,
 						  int busw, int *maf_id)
 {
 	struct nand_flash_dev *type = NULL;
-	int i, dev_id, maf_idx;
+	int dev_id, maf_idx;
 	int tmp_id, tmp_manf;
 
 	/* Select the device */
@@ -2809,14 +2818,18 @@ static struct nand_flash_dev *nand_get_flash_type(struct mtd_info *mtd,
 	}
 
 	/* Lookup the flash id */
-	for (i = 0; nand_flash_ids[i].name != NULL; i++) {
-		if (dev_id == nand_flash_ids[i].id) {
-			type =  &nand_flash_ids[i];
+#ifdef CONFIG_MTD_NAND_SMARTMEDIA
+	if (chip->options & NAND_SMARTMEDIA)
+		type = nand_smartmedia_flash_ids;
+	else
+#endif
+		type = nand_flash_ids;
+
+	for (; type->name != NULL; type++)
+		if (dev_id == type->id)
 			break;
-		}
-	}
 
-	if (!type)
+	if (!type->name)
 		return ERR_PTR(-ENODEV);
 
 	if (!mtd->name)
@@ -3174,7 +3187,8 @@ int nand_scan_tail(struct mtd_info *mtd)
 
 	/* Fill in remaining MTD driver data */
 	mtd->type = MTD_NANDFLASH;
-	mtd->flags = MTD_CAP_NANDFLASH;
+	mtd->flags = chip->options & NAND_ROM ? MTD_CAP_ROM :
+							MTD_CAP_NANDFLASH;
 	mtd->erase = nand_erase;
 	mtd->point = NULL;
 	mtd->unpoint = NULL;
diff --git a/drivers/mtd/nand/nand_ids.c b/drivers/mtd/nand/nand_ids.c
index 69ee2c9..f9e72d5 100644
--- a/drivers/mtd/nand/nand_ids.c
+++ b/drivers/mtd/nand/nand_ids.c
@@ -127,6 +127,45 @@ struct nand_flash_dev nand_flash_ids[] = {
 	{NULL,}
 };
 
+#ifdef CONFIG_MTD_NAND_SMARTMEDIA
+struct nand_flash_dev nand_smartmedia_flash_ids[] = {
+
+	/* SmartMedia */
+	{"SmartMedia 1MiB 5V",		0x6e, 256, 1, 0x1000, 0},
+	{"SmartMedia 1MiB 3,3V",	0xe8, 256, 1, 0x1000, 0},
+	{"SmartMedia 1MiB 3,3V",	0xec, 256, 1, 0x1000, 0},
+	{"SmartMedia 2MiB 3,3V",	0xea, 256, 2, 0x1000, 0},
+	{"SmartMedia 2MiB 5V",		0x64, 256, 2, 0x1000, 0},
+	{"SmartMedia 2MiB 3,3V ROM",	0x5d, 512, 2, 0x2000, NAND_ROM},
+	{"SmartMedia 4MiB 3,3V",	0xe3, 512, 4, 0x2000, 0},
+	{"SmartMedia 4MiB 3,3/5V",	0xe5, 512, 4, 0x2000, 0},
+	{"SmartMedia 4MiB 5V",		0x6b, 512, 4, 0x2000, 0},
+	{"SmartMedia 4MiB 3,3V ROM",	0xd5, 512, 4, 0x2000, NAND_ROM},
+	{"SmartMedia 8MiB 3,3V",	0xe6, 512, 8, 0x2000, 0},
+	{"SmartMedia 8MiB 3,3V ROM",	0xd6, 512, 8, 0x2000, NAND_ROM},
+
+#define XD_TYPEM	(NAND_NO_AUTOINCR | NAND_BROKEN_XD)
+	/* xD / SmartMedia */
+	{"SmartMedia/xD 16MiB 3,3V",	0x73, 512, 16, 0x4000, 0},
+	{"SmartMedia 16MiB 3,3V ROM",	0x57, 512, 16, 0x4000, NAND_ROM},
+	{"SmartMedia/xD 32MiB 3,3V",	0x75, 512, 32, 0x4000, 0},
+	{"SmartMedia 32MiB 3,3V ROM",	0x58, 512, 32, 0x4000, NAND_ROM},
+	{"SmartMedia/xD 64MiB 3,3V",	0x76, 512, 64, 0x4000, 0},
+	{"SmartMedia 64MiB 3,3V ROM",	0xd9, 512, 64, 0x4000, NAND_ROM},
+	{"SmartMedia/xD 128MiB 3,3V",	0x79, 512, 128, 0x4000, 0},
+	{"SmartMedia 128MiB 3,3V ROM",	0xda, 512, 128, 0x4000, NAND_ROM},
+	{"SmartMedia/xD 256MiB 3,3V",	0x71, 512, 256, 0x4000, XD_TYPEM},
+	{"SmartMedia 256MiB 3,3V ROM",	0x5b, 512, 256, 0x4000, NAND_ROM},
+
+	/* xD only */
+	{"xD 512MiB 3,3V",		0xDC, 512, 512, 0x4000, XD_TYPEM},
+	{"xD 1GiB 3,3V",		0xD3, 512, 1024, 0x4000, XD_TYPEM},
+	{"xD 2GiB 3,3V",		0xD5, 512, 2048, 0x4000, XD_TYPEM},
+	{NULL,}
+};
+EXPORT_SYMBOL(nand_smartmedia_flash_ids);
+#endif
+
 /*
 *	Manufacturer ID list
 */
diff --git a/include/linux/mtd/nand.h b/include/linux/mtd/nand.h
index f2d4a1a..7291725 100644
--- a/include/linux/mtd/nand.h
+++ b/include/linux/mtd/nand.h
@@ -180,6 +180,12 @@ typedef enum {
 /* Chip does not allow subpage writes */
 #define NAND_NO_SUBPAGE_WRITE	0x00000200
 
+/* Device is one of 'new' xD cards that expose fake nand command set */
+#define NAND_BROKEN_XD		0x00000400
+
+/* Device behaves just like nand, but is readonly */
+#define NAND_ROM		0x00000800
+
 /* Options valid for Samsung large page devices */
 #define NAND_SAMSUNG_LP_OPTIONS \
 	(NAND_NO_PADDING | NAND_CACHEPRG | NAND_COPYBACK)
@@ -205,9 +211,13 @@ typedef enum {
 /* This option is defined if the board driver allocates its own buffers
    (e.g. because it needs them DMA-coherent */
 #define NAND_OWN_BUFFERS	0x00040000
+
 /* Chip may not exist, so silence any errors in scan */
 #define NAND_SCAN_SILENT_NODEV	0x00080000
 
+/* controller supports only xD/SmartMedia cards*/
+#define NAND_SMARTMEDIA		0x00100000
+
 /* Options set by nand scan */
 /* Nand scan has allocated controller struct */
 #define NAND_CONTROLLER_ALLOC	0x80000000
@@ -469,6 +479,7 @@ struct nand_manufacturers {
 };
 
 extern struct nand_flash_dev nand_flash_ids[];
+extern struct nand_flash_dev nand_smartmedia_flash_ids[];
 extern struct nand_manufacturers nand_manuf_ids[];
 
 extern int nand_scan_bbt(struct mtd_info *mtd, struct nand_bbt_descr *bd);
-- 
1.6.3.3

[PATCH 13/15] MTD: add few workarounds to nand system for SmartMedia/xD chips.

[Maxim Levitsky]

* Add an option NAND_SMARTMEDIA that can be set by nand driver
 If set, it will cause separate ID table to be used, which includes
 mask rom devices and new xD cards

* Workaround for wrong write protect status on some xD cards

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/nand/nand_base.c |   32 +++++++++++++++++++++++---------
 drivers/mtd/nand/nand_ids.c  |   39 +++++++++++++++++++++++++++++++++++++++
 include/linux/mtd/nand.h     |   11 +++++++++++
 3 files changed, 73 insertions(+), 9 deletions(-)

diff --git a/drivers/mtd/nand/nand_base.c b/drivers/mtd/nand/nand_base.c
index ba29a29..fd9b00d 100644
--- a/drivers/mtd/nand/nand_base.c
+++ b/drivers/mtd/nand/nand_base.c
@@ -434,9 +434,18 @@ static int nand_default_block_markbad(struct mtd_info *mtd, loff_t ofs)
 static int nand_check_wp(struct mtd_info *mtd)
 {
 	struct nand_chip *chip = mtd->priv;
+	int wp;
+
+	/* broken xD cards report WP despite beeing writable */
+	if (chip->options & NAND_BROKEN_XD)
+		return 0;
+
 	/* Check the WP bit */
 	chip->cmdfunc(mtd, NAND_CMD_STATUS, -1, -1);
-	return (chip->read_byte(mtd) & NAND_STATUS_WP) ? 0 : 1;
+	wp = (chip->read_byte(mtd) & NAND_STATUS_WP) ? 0 : 1;
+
+
+	return wp;
 }
 
 /**
@@ -2769,7 +2778,7 @@ static struct nand_flash_dev *nand_get_flash_type(struct mtd_info *mtd,
 						  int busw, int *maf_id)
 {
 	struct nand_flash_dev *type = NULL;
-	int i, dev_id, maf_idx;
+	int dev_id, maf_idx;
 	int tmp_id, tmp_manf;
 
 	/* Select the device */
@@ -2809,14 +2818,18 @@ static struct nand_flash_dev *nand_get_flash_type(struct mtd_info *mtd,
 	}
 
 	/* Lookup the flash id */
-	for (i = 0; nand_flash_ids[i].name != NULL; i++) {
-		if (dev_id == nand_flash_ids[i].id) {
-			type =  &nand_flash_ids[i];
+#ifdef CONFIG_MTD_NAND_SMARTMEDIA
+	if (chip->options & NAND_SMARTMEDIA)
+		type = nand_smartmedia_flash_ids;
+	else
+#endif
+		type = nand_flash_ids;
+
+	for (; type->name != NULL; type++)
+		if (dev_id == type->id)
 			break;
-		}
-	}
 
-	if (!type)
+	if (!type->name)
 		return ERR_PTR(-ENODEV);
 
 	if (!mtd->name)
@@ -3174,7 +3187,8 @@ int nand_scan_tail(struct mtd_info *mtd)
 
 	/* Fill in remaining MTD driver data */
 	mtd->type = MTD_NANDFLASH;
-	mtd->flags = MTD_CAP_NANDFLASH;
+	mtd->flags = chip->options & NAND_ROM ? MTD_CAP_ROM :
+							MTD_CAP_NANDFLASH;
 	mtd->erase = nand_erase;
 	mtd->point = NULL;
 	mtd->unpoint = NULL;
diff --git a/drivers/mtd/nand/nand_ids.c b/drivers/mtd/nand/nand_ids.c
index 69ee2c9..f9e72d5 100644
--- a/drivers/mtd/nand/nand_ids.c
+++ b/drivers/mtd/nand/nand_ids.c
@@ -127,6 +127,45 @@ struct nand_flash_dev nand_flash_ids[] = {
 	{NULL,}
 };
 
+#ifdef CONFIG_MTD_NAND_SMARTMEDIA
+struct nand_flash_dev nand_smartmedia_flash_ids[] = {
+
+	/* SmartMedia */
+	{"SmartMedia 1MiB 5V",		0x6e, 256, 1, 0x1000, 0},
+	{"SmartMedia 1MiB 3,3V",	0xe8, 256, 1, 0x1000, 0},
+	{"SmartMedia 1MiB 3,3V",	0xec, 256, 1, 0x1000, 0},
+	{"SmartMedia 2MiB 3,3V",	0xea, 256, 2, 0x1000, 0},
+	{"SmartMedia 2MiB 5V",		0x64, 256, 2, 0x1000, 0},
+	{"SmartMedia 2MiB 3,3V ROM",	0x5d, 512, 2, 0x2000, NAND_ROM},
+	{"SmartMedia 4MiB 3,3V",	0xe3, 512, 4, 0x2000, 0},
+	{"SmartMedia 4MiB 3,3/5V",	0xe5, 512, 4, 0x2000, 0},
+	{"SmartMedia 4MiB 5V",		0x6b, 512, 4, 0x2000, 0},
+	{"SmartMedia 4MiB 3,3V ROM",	0xd5, 512, 4, 0x2000, NAND_ROM},
+	{"SmartMedia 8MiB 3,3V",	0xe6, 512, 8, 0x2000, 0},
+	{"SmartMedia 8MiB 3,3V ROM",	0xd6, 512, 8, 0x2000, NAND_ROM},
+
+#define XD_TYPEM	(NAND_NO_AUTOINCR | NAND_BROKEN_XD)
+	/* xD / SmartMedia */
+	{"SmartMedia/xD 16MiB 3,3V",	0x73, 512, 16, 0x4000, 0},
+	{"SmartMedia 16MiB 3,3V ROM",	0x57, 512, 16, 0x4000, NAND_ROM},
+	{"SmartMedia/xD 32MiB 3,3V",	0x75, 512, 32, 0x4000, 0},
+	{"SmartMedia 32MiB 3,3V ROM",	0x58, 512, 32, 0x4000, NAND_ROM},
+	{"SmartMedia/xD 64MiB 3,3V",	0x76, 512, 64, 0x4000, 0},
+	{"SmartMedia 64MiB 3,3V ROM",	0xd9, 512, 64, 0x4000, NAND_ROM},
+	{"SmartMedia/xD 128MiB 3,3V",	0x79, 512, 128, 0x4000, 0},
+	{"SmartMedia 128MiB 3,3V ROM",	0xda, 512, 128, 0x4000, NAND_ROM},
+	{"SmartMedia/xD 256MiB 3,3V",	0x71, 512, 256, 0x4000, XD_TYPEM},
+	{"SmartMedia 256MiB 3,3V ROM",	0x5b, 512, 256, 0x4000, NAND_ROM},
+
+	/* xD only */
+	{"xD 512MiB 3,3V",		0xDC, 512, 512, 0x4000, XD_TYPEM},
+	{"xD 1GiB 3,3V",		0xD3, 512, 1024, 0x4000, XD_TYPEM},
+	{"xD 2GiB 3,3V",		0xD5, 512, 2048, 0x4000, XD_TYPEM},
+	{NULL,}
+};
+EXPORT_SYMBOL(nand_smartmedia_flash_ids);
+#endif
+
 /*
 *	Manufacturer ID list
 */
diff --git a/include/linux/mtd/nand.h b/include/linux/mtd/nand.h
index f2d4a1a..7291725 100644
--- a/include/linux/mtd/nand.h
+++ b/include/linux/mtd/nand.h
@@ -180,6 +180,12 @@ typedef enum {
 /* Chip does not allow subpage writes */
 #define NAND_NO_SUBPAGE_WRITE	0x00000200
 
+/* Device is one of 'new' xD cards that expose fake nand command set */
+#define NAND_BROKEN_XD		0x00000400
+
+/* Device behaves just like nand, but is readonly */
+#define NAND_ROM		0x00000800
+
 /* Options valid for Samsung large page devices */
 #define NAND_SAMSUNG_LP_OPTIONS \
 	(NAND_NO_PADDING | NAND_CACHEPRG | NAND_COPYBACK)
@@ -205,9 +211,13 @@ typedef enum {
 /* This option is defined if the board driver allocates its own buffers
    (e.g. because it needs them DMA-coherent */
 #define NAND_OWN_BUFFERS	0x00040000
+
 /* Chip may not exist, so silence any errors in scan */
 #define NAND_SCAN_SILENT_NODEV	0x00080000
 
+/* controller supports only xD/SmartMedia cards*/
+#define NAND_SMARTMEDIA		0x00100000
+
 /* Options set by nand scan */
 /* Nand scan has allocated controller struct */
 #define NAND_CONTROLLER_ALLOC	0x80000000
@@ -469,6 +479,7 @@ struct nand_manufacturers {
 };
 
 extern struct nand_flash_dev nand_flash_ids[];
+extern struct nand_flash_dev nand_smartmedia_flash_ids[];
 extern struct nand_manufacturers nand_manuf_ids[];
 
 extern int nand_scan_bbt(struct mtd_info *mtd, struct nand_bbt_descr *bd);
-- 
1.6.3.3

Re: [PATCH 13/15] MTD: add few workarounds to nand system for SmartMedia/xD chips.

[Thomas Gleixner]

On Mon, 22 Feb 2010, Maxim Levitsky wrote:

> * Add an option NAND_SMARTMEDIA that can be set by nand driver
>  If set, it will cause separate ID table to be used, which includes
>  mask rom devices and new xD cards

Why that option ? We can just extend the existing ids table and be
done. No extra magic needed.
 
Thanks,

	tglx

Re: [PATCH 13/15] MTD: add few workarounds to nand system for SmartMedia/xD chips.

[Thomas Gleixner]

On Mon, 22 Feb 2010, Maxim Levitsky wrote:

> * Add an option NAND_SMARTMEDIA that can be set by nand driver
>  If set, it will cause separate ID table to be used, which includes
>  mask rom devices and new xD cards

Why that option ? We can just extend the existing ids table and be
done. No extra magic needed.
 
Thanks,

	tglx

Re: [PATCH 13/15] MTD: add few workarounds to nand system for SmartMedia/xD chips.

[Maxim Levitsky]

On Mon, 2010-02-22 at 22:25 +0100, Thomas Gleixner wrote: 
> On Mon, 22 Feb 2010, Maxim Levitsky wrote:
> 
> > * Add an option NAND_SMARTMEDIA that can be set by nand driver
> >  If set, it will cause separate ID table to be used, which includes
> >  mask rom devices and new xD cards
> 
> Why that option ? We can just extend the existing ids table and be
> done. No extra magic needed.
>  

Two reasons.

First of all several xD chips (I belive the Type M) have exactly same
IDs like normal nand chips. However they don't report capabilities about
pagesize, blocksize, etc.

I am confident that these cards have an internal FTL and controller, and
just 'emulate' that nand interface.

Also, my card  reports write protect, although, xD cards don't have any
'switch' to make them protected.
Even if there were readonly ROM xD cards (the odds of this are virtually
zero), these won't just expose this in WP bit.

Best regards,
Maxim Levitsky

Re: [PATCH 13/15] MTD: add few workarounds to nand system for SmartMedia/xD chips.

[Maxim Levitsky]

On Mon, 2010-02-22 at 22:25 +0100, Thomas Gleixner wrote: 
> On Mon, 22 Feb 2010, Maxim Levitsky wrote:
> 
> > * Add an option NAND_SMARTMEDIA that can be set by nand driver
> >  If set, it will cause separate ID table to be used, which includes
> >  mask rom devices and new xD cards
> 
> Why that option ? We can just extend the existing ids table and be
> done. No extra magic needed.
>  

Two reasons.

First of all several xD chips (I belive the Type M) have exactly same
IDs like normal nand chips. However they don't report capabilities about
pagesize, blocksize, etc.

I am confident that these cards have an internal FTL and controller, and
just 'emulate' that nand interface.

Also, my card  reports write protect, although, xD cards don't have any
'switch' to make them protected.
Even if there were readonly ROM xD cards (the odds of this are virtually
zero), these won't just expose this in WP bit.

Best regards,
Maxim Levitsky

Re: [PATCH 13/15] MTD: add few workarounds to nand system for SmartMedia/xD chips.

[Thomas Gleixner]

On Mon, 22 Feb 2010, Maxim Levitsky wrote:

> On Mon, 2010-02-22 at 22:25 +0100, Thomas Gleixner wrote: 
> > On Mon, 22 Feb 2010, Maxim Levitsky wrote:
> > 
> > > * Add an option NAND_SMARTMEDIA that can be set by nand driver
> > >  If set, it will cause separate ID table to be used, which includes
> > >  mask rom devices and new xD cards
> > 
> > Why that option ? We can just extend the existing ids table and be
> > done. No extra magic needed.
> >  
> 
> Two reasons.
> 
> First of all several xD chips (I belive the Type M) have exactly same
> IDs like normal nand chips. However they don't report capabilities about
> pagesize, blocksize, etc.

You mean that crap ?

+       /* xD only */
+       {"xD 512MiB 3,3V",              0xDC, 512, 512, 0x4000, XD_TYPEM},
+       {"xD 1GiB 3,3V",                0xD3, 512, 1024, 0x4000, XD_TYPEM},
+       {"xD 2GiB 3,3V",                0xD5, 512, 2048, 0x4000, XD_TYPEM},

Oh well, that's a perfect example of designed by comittee shit.
 
They abuse the 2k chips IDs and define crappy sizes just to fit the
stupid spec.

> I am confident that these cards have an internal FTL and controller, and
> just 'emulate' that nand interface.

Either that or the adapter translates the 512B commands to 2K
commands. I bet on the latter. AFAIK are xD cards just as stupid as
SmartMedia ones, i.e. bare NAND chips in a flat plastic housing with
gold contacts.

> Also, my card  reports write protect, although, xD cards don't have any
> 'switch' to make them protected.

Neither have SmartMedia Cards nor bare NAND chips. WP is a pin on the
interface which is controlled by the adapter hardware.

> Even if there were readonly ROM xD cards (the odds of this are virtually
> zero), these won't just expose this in WP bit.

Hmm, I wonder whether that horror has the WP bit inverted.

Thanks,

	tglx

Re: [PATCH 13/15] MTD: add few workarounds to nand system for SmartMedia/xD chips.

[Thomas Gleixner]

On Mon, 22 Feb 2010, Maxim Levitsky wrote:

> On Mon, 2010-02-22 at 22:25 +0100, Thomas Gleixner wrote: 
> > On Mon, 22 Feb 2010, Maxim Levitsky wrote:
> > 
> > > * Add an option NAND_SMARTMEDIA that can be set by nand driver
> > >  If set, it will cause separate ID table to be used, which includes
> > >  mask rom devices and new xD cards
> > 
> > Why that option ? We can just extend the existing ids table and be
> > done. No extra magic needed.
> >  
> 
> Two reasons.
> 
> First of all several xD chips (I belive the Type M) have exactly same
> IDs like normal nand chips. However they don't report capabilities about
> pagesize, blocksize, etc.

You mean that crap ?

+       /* xD only */
+       {"xD 512MiB 3,3V",              0xDC, 512, 512, 0x4000, XD_TYPEM},
+       {"xD 1GiB 3,3V",                0xD3, 512, 1024, 0x4000, XD_TYPEM},
+       {"xD 2GiB 3,3V",                0xD5, 512, 2048, 0x4000, XD_TYPEM},

Oh well, that's a perfect example of designed by comittee shit.
 
They abuse the 2k chips IDs and define crappy sizes just to fit the
stupid spec.

> I am confident that these cards have an internal FTL and controller, and
> just 'emulate' that nand interface.

Either that or the adapter translates the 512B commands to 2K
commands. I bet on the latter. AFAIK are xD cards just as stupid as
SmartMedia ones, i.e. bare NAND chips in a flat plastic housing with
gold contacts.

> Also, my card  reports write protect, although, xD cards don't have any
> 'switch' to make them protected.

Neither have SmartMedia Cards nor bare NAND chips. WP is a pin on the
interface which is controlled by the adapter hardware.

> Even if there were readonly ROM xD cards (the odds of this are virtually
> zero), these won't just expose this in WP bit.

Hmm, I wonder whether that horror has the WP bit inverted.

Thanks,

	tglx

Re: [PATCH 13/15] MTD: add few workarounds to nand system for SmartMedia/xD chips.

[Maxim Levitsky]

On Mon, 2010-02-22 at 22:53 +0100, Thomas Gleixner wrote: 
> On Mon, 22 Feb 2010, Maxim Levitsky wrote:
> 
> > On Mon, 2010-02-22 at 22:25 +0100, Thomas Gleixner wrote: 
> > > On Mon, 22 Feb 2010, Maxim Levitsky wrote:
> > > 
> > > > * Add an option NAND_SMARTMEDIA that can be set by nand driver
> > > >  If set, it will cause separate ID table to be used, which includes
> > > >  mask rom devices and new xD cards
> > > 
> > > Why that option ? We can just extend the existing ids table and be
> > > done. No extra magic needed.
> > >  
> > 
> > Two reasons.
> > 
> > First of all several xD chips (I belive the Type M) have exactly same
> > IDs like normal nand chips. However they don't report capabilities about
> > pagesize, blocksize, etc.

Sure, I didn't meant the #WP line. Sure it is set by the controller.
I mean WP bit in status report.

Best regards,
Maxim Levitsky

Re: [PATCH 13/15] MTD: add few workarounds to nand system for SmartMedia/xD chips.

[Maxim Levitsky]

On Mon, 2010-02-22 at 22:53 +0100, Thomas Gleixner wrote: 
> On Mon, 22 Feb 2010, Maxim Levitsky wrote:
> 
> > On Mon, 2010-02-22 at 22:25 +0100, Thomas Gleixner wrote: 
> > > On Mon, 22 Feb 2010, Maxim Levitsky wrote:
> > > 
> > > > * Add an option NAND_SMARTMEDIA that can be set by nand driver
> > > >  If set, it will cause separate ID table to be used, which includes
> > > >  mask rom devices and new xD cards
> > > 
> > > Why that option ? We can just extend the existing ids table and be
> > > done. No extra magic needed.
> > >  
> > 
> > Two reasons.
> > 
> > First of all several xD chips (I belive the Type M) have exactly same
> > IDs like normal nand chips. However they don't report capabilities about
> > pagesize, blocksize, etc.

Sure, I didn't meant the #WP line. Sure it is set by the controller.
I mean WP bit in status report.

Best regards,
Maxim Levitsky

[PATCH 14/15] MTD: Add new SmartMedia/xD FTL

[Maxim Levitsky]

This implements new readwrite SmartMedia/xd FTL.

mtd driver must have support proper ECC and badblock vertification
based on oob parts for 512 bytes nand.

Also mtd driver must define read_oob and write_oob, which are used
to read and write both data and oob together.

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/Kconfig  |   21 +
 drivers/mtd/Makefile |    1 +
 drivers/mtd/sm_ftl.c | 1284 ++++++++++++++++++++++++++++++++++++++++++++++++++
 drivers/mtd/sm_ftl.h |   94 ++++
 4 files changed, 1400 insertions(+), 0 deletions(-)
 create mode 100644 drivers/mtd/sm_ftl.c
 create mode 100644 drivers/mtd/sm_ftl.h

diff --git a/drivers/mtd/Kconfig b/drivers/mtd/Kconfig
index ecf90f5..0105ae0 100644
--- a/drivers/mtd/Kconfig
+++ b/drivers/mtd/Kconfig
@@ -304,6 +304,27 @@ config SSFDC
 	  This enables read only access to SmartMedia formatted NAND
 	  flash. You can mount it with FAT file system.
 
+
+config SM_FTL
+	tristate "SmartMedia/xD new translation layer"
+	depends on EXPERIMENTAL && BLOCK
+	select MTD_BLKDEVS
+	help
+	  This enables new and very EXPERMENTAL support for SmartMedia/xD
+	  FTL (Flash tanslation layer)
+	  Write support isn't yet well tested, therefore this code IS likely to
+	  eat your card, so please don't use it together with valuable data.
+	  Use readonly driver (CONFIG_SSFDC) instead.
+
+config SM_FTL_MUSEUM
+	boolean "Additional Support for 1MB and 2MB SmartMedia cards"
+	depends on SM_FTL && MTD_NAND
+	select MTD_NAND_ECC_SMC
+	help
+	  Very old SmartMedia cards need ECC to be caluculated in the FTL
+	  Such cards are very rare, thus enabling this option is mostly useless
+	  Also this support is completely UNTESTED.
+
 config MTD_OOPS
 	tristate "Log panic/oops to an MTD buffer"
 	depends on MTD
diff --git a/drivers/mtd/Makefile b/drivers/mtd/Makefile
index 82d1e4d..d53357b 100644
--- a/drivers/mtd/Makefile
+++ b/drivers/mtd/Makefile
@@ -24,6 +24,7 @@ obj-$(CONFIG_NFTL)		+= nftl.o
 obj-$(CONFIG_INFTL)		+= inftl.o
 obj-$(CONFIG_RFD_FTL)		+= rfd_ftl.o
 obj-$(CONFIG_SSFDC)		+= ssfdc.o
+obj-$(CONFIG_SM_FTL)		+= sm_ftl.o
 obj-$(CONFIG_MTD_OOPS)		+= mtdoops.o
 
 nftl-objs		:= nftlcore.o nftlmount.o
diff --git a/drivers/mtd/sm_ftl.c b/drivers/mtd/sm_ftl.c
new file mode 100644
index 0000000..c2a409f
--- /dev/null
+++ b/drivers/mtd/sm_ftl.c
@@ -0,0 +1,1284 @@
+/*
+ * Copyright (C) 2009 - Maxim Levitsky
+ * SmartMedia/xD translation layer
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/random.h>
+#include <linux/hdreg.h>
+#include <linux/kthread.h>
+#include <linux/freezer.h>
+#include <linux/sysfs.h>
+#include <linux/bitops.h>
+#include "nand/sm_common.h"
+#include "sm_ftl.h"
+
+#ifdef CONFIG_SM_FTL_MUSEUM
+#include <linux/mtd/nand_ecc.h>
+#endif
+
+
+struct workqueue_struct *cache_flush_workqueue;
+
+static int cache_timeout = 1000;
+module_param(cache_timeout, bool, S_IRUGO);
+MODULE_PARM_DESC(cache_timeout,
+	"Timeout (in ms) for cache flush (1000 ms default");
+
+static int debug;
+module_param(debug, int, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(debug, "Debug level (0-2)");
+
+
+/* ------------------- sysfs attributtes ---------------------------------- */
+struct sm_sysfs_attribute {
+	struct device_attribute dev_attr;
+	char *data;
+	int len;
+};
+
+ssize_t sm_attr_show(struct device *dev, struct device_attribute *attr,
+			char *buf)
+{
+	struct sm_sysfs_attribute *sm_attr =
+		container_of(attr, struct sm_sysfs_attribute, dev_attr);
+
+	strncpy(buf, sm_attr->data, sm_attr->len);
+	return sm_attr->len;
+}
+
+
+#define NUM_ATTRIBUTES 1
+#define SM_CIS_VENDOR_OFFSET 0x59
+struct attribute_group *sm_create_sysfs_attributes(struct sm_ftl *ftl)
+{
+	struct attribute_group *attr_group;
+	struct attribute **attributes;
+	struct sm_sysfs_attribute *vendor_attribute;
+
+	int vendor_len = strnlen(ftl->cis_buffer + SM_CIS_VENDOR_OFFSET,
+					SM_SMALL_PAGE - SM_CIS_VENDOR_OFFSET);
+
+	char *vendor = kmalloc(vendor_len, GFP_KERNEL);
+	memcpy(vendor, ftl->cis_buffer + SM_CIS_VENDOR_OFFSET, vendor_len);
+	vendor[vendor_len] = 0;
+
+	/* Initialize sysfs attributes */
+	vendor_attribute =
+		kzalloc(sizeof(struct sm_sysfs_attribute), GFP_KERNEL);
+
+	vendor_attribute->data = vendor;
+	vendor_attribute->len = vendor_len;
+	vendor_attribute->dev_attr.attr.name = "vendor";
+	vendor_attribute->dev_attr.attr.mode = S_IRUGO;
+	vendor_attribute->dev_attr.show = sm_attr_show;
+
+
+	/* Create array of pointers to the attributes */
+	attributes = kzalloc(sizeof(struct attribute *) * (NUM_ATTRIBUTES + 1),
+								GFP_KERNEL);
+	attributes[0] = &vendor_attribute->dev_attr.attr;
+
+	/* Finally create the attribute group */
+	attr_group = kzalloc(sizeof(struct attribute_group), GFP_KERNEL);
+	attr_group->attrs = attributes;
+	return attr_group;
+}
+
+void sm_delete_sysfs_attributes(struct sm_ftl *ftl)
+{
+	struct attribute **attributes = ftl->disk_attributes->attrs;
+	int i;
+
+	for (i = 0; attributes[i] ; i++) {
+
+		struct device_attribute *dev_attr = container_of(attributes[i],
+			struct device_attribute, attr);
+
+		struct sm_sysfs_attribute *sm_attr =
+			container_of(dev_attr,
+				struct sm_sysfs_attribute, dev_attr);
+
+		kfree(sm_attr->data);
+		kfree(sm_attr);
+	}
+
+	kfree(ftl->disk_attributes->attrs);
+	kfree(ftl->disk_attributes);
+}
+
+
+/* ----------------------- oob helpers -------------------------------------- */
+
+static int sm_get_lba(u8 *lba)
+{
+	/* check fixed bits */
+	if ((lba[0] & 0xF8) != 0x10)
+		return -2;
+
+	/* check parity - endianess doesn't matter */
+	if (hweight16(*(u16 *)lba) & 1)
+		return -2;
+
+	return (lba[1] >> 1) | ((lba[0] & 0x07) << 7);
+}
+
+
+/*
+ * Read LBA asscociated with block
+ * returns -1, if block is erased
+ * returns -2 if error happens
+ */
+static int sm_read_lba(struct sm_oob *oob)
+{
+	static const u32 erased_pattern[4] = {
+		0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF };
+
+	u16 lba_test;
+	int lba;
+
+	/* First test for erased block */
+	if (!memcmp(oob, erased_pattern, SM_OOB_SIZE))
+		return -1;
+
+	/* Now check is both copies of the LBA differ too much */
+	lba_test = *(u16 *)oob->lba_copy1 ^ *(u16*)oob->lba_copy2;
+	if (lba_test && !is_power_of_2(lba_test))
+		return -2;
+
+	/* And read it */
+	lba = sm_get_lba(oob->lba_copy1);
+
+	if (lba == -2)
+		lba = sm_get_lba(oob->lba_copy2);
+
+	return lba;
+}
+
+static void sm_write_lba(struct sm_oob *oob, u16 lba)
+{
+	u8 tmp[2];
+
+	WARN_ON(lba >= 1000);
+
+	tmp[0] = 0x10 | ((lba >> 7) & 0x07);
+	tmp[1] = (lba << 1) & 0xFF;
+
+	if (hweight16(*(u16 *)tmp) & 0x01)
+		tmp[1] |= 1;
+
+	oob->lba_copy1[0] = oob->lba_copy2[0] = tmp[0];
+	oob->lba_copy1[1] = oob->lba_copy2[1] = tmp[1];
+}
+
+
+/* Make offset from parts */
+static loff_t sm_mkoffset(struct sm_ftl *ftl, int zone, int block, int boffset)
+{
+	WARN_ON(boffset & (SM_SECTOR_SIZE - 1));
+	WARN_ON(zone < 0 || zone >= ftl->zone_count);
+	WARN_ON(block >= ftl->zone_size);
+	WARN_ON(boffset >= ftl->block_size);
+
+	if (block == -1)
+		return -1;
+
+	return (zone * SM_MAX_ZONE_SIZE + block) * ftl->block_size + boffset;
+}
+
+/* Breaks offset into parts */
+static void sm_break_offset(struct sm_ftl *ftl, loff_t offset,
+					int *zone, int *block, int *boffset)
+{
+	*boffset = do_div(offset, ftl->block_size);
+	*block = do_div(offset, ftl->max_lba);
+	*zone = offset >= ftl->zone_count ? -1 : offset;
+}
+
+/* ---------------------- low level IO ------------------------------------- */
+
+static int sm_correct_sector(u8 *buffer, struct sm_oob *oob)
+{
+#ifdef CONFIG_SM_FTL_MUSEUM
+	u8 ecc[3];
+
+	__nand_calculate_ecc(buffer, SM_SMALL_PAGE, ecc);
+	if (__nand_correct_data(buffer, ecc, oob->ecc1, SM_SMALL_PAGE) < 0)
+		return -EIO;
+
+	buffer += SM_SMALL_PAGE;
+
+	__nand_calculate_ecc(buffer, SM_SMALL_PAGE, ecc);
+	if (__nand_correct_data(buffer, ecc, oob->ecc2, SM_SMALL_PAGE) < 0)
+		return -EIO;
+#endif
+	return 0;
+}
+
+/* Reads a sector + oob*/
+static int sm_read_sector(struct sm_ftl *ftl,
+			int zone, int block, int boffset,
+					u8 *buffer, struct sm_oob *oob)
+{
+	struct mtd_info *mtd = ftl->trans->mtd;
+	struct mtd_oob_ops ops;
+	struct sm_oob tmp_oob;
+	int ret;
+	int try = 0;
+
+	/* FTL can contain -1 entries that are by default filled with bits */
+	if (block == -1) {
+		memset(buffer, 0xFF, SM_SECTOR_SIZE);
+		return 0;
+	}
+
+	/* User might not need the oob, but we do for data vertification */
+	if (!oob)
+		oob = &tmp_oob;
+
+	ops.mode = ftl->smallpagenand ? MTD_OOB_RAW : MTD_OOB_PLACE;
+	ops.ooboffs = 0;
+	ops.ooblen = SM_OOB_SIZE;
+	ops.oobbuf = (void *)oob;
+	ops.len = SM_SECTOR_SIZE;
+	ops.datbuf = buffer;
+
+again:
+	if (try++) {
+		/* Avoid infinite recursion on CIS reads, sm_recheck_media
+			won't help anyway */
+		if (zone == 0 && block == ftl->cis_block && boffset ==
+			ftl->cis_boffset)
+			return ret;
+
+		/* Test if media is stable */
+		if (try == 3 || sm_recheck_media(ftl))
+			return ret;
+	}
+
+	/* Unfortunelly, oob read will _always_ succeed,
+		despite card removal..... */
+	ret = mtd->read_oob(mtd, sm_mkoffset(ftl, zone, block, boffset), &ops);
+
+	/* Test for unknown errors */
+	if (ret != 0 && ret != -EUCLEAN && ret != -EBADMSG) {
+		dbg("read of block %d at zone %d, failed due to error (%d)",
+			block, zone, ret);
+		goto again;
+	}
+
+	/* Do a basic test on the oob, to guard against returned garbage */
+	if (oob->reserved != 0xFFFFFFFF && !is_power_of_2(~oob->reserved))
+		goto again;
+
+	/* This should never happen, unless there is a bug in the mtd driver */
+	WARN_ON(ops.oobretlen != SM_OOB_SIZE);
+	WARN_ON(buffer && ops.retlen != SM_SECTOR_SIZE);
+
+	if (!buffer)
+		return 0;
+
+	/* Test if sector marked as bad */
+	if (!sm_sector_valid(oob)) {
+		dbg("read of block %d at zone %d, failed because it is marked"
+			" as bad" , block, zone);
+		goto again;
+	}
+
+	/* Test ECC*/
+	if (ret == -EBADMSG ||
+		(ftl->smallpagenand && sm_correct_sector(buffer, oob))) {
+
+		dbg("read of block %d at zone %d, failed due to ECC error",
+			block, zone);
+		goto again;
+	}
+
+	return 0;
+}
+
+/* Writes a sector to media */
+static int sm_write_sector(struct sm_ftl *ftl,
+			int zone, int block, int boffset,
+					u8 *buffer, struct sm_oob *oob)
+{
+	struct mtd_oob_ops ops;
+	struct mtd_info *mtd = ftl->trans->mtd;
+	int ret;
+
+	BUG_ON(ftl->readonly);
+
+	if (zone == 0 && (block == ftl->cis_block || block == 0)) {
+		dbg("attempted to write the CIS!");
+		return -EIO;
+	}
+
+	if (ftl->unstable)
+		return -EIO;
+
+	ops.mode = ftl->smallpagenand ? MTD_OOB_RAW : MTD_OOB_PLACE;
+	ops.len = SM_SECTOR_SIZE;
+	ops.datbuf = buffer;
+	ops.ooboffs = 0;
+	ops.ooblen = SM_OOB_SIZE;
+	ops.oobbuf = (void *)oob;
+
+	ret = mtd->write_oob(mtd, sm_mkoffset(ftl, zone, block, boffset), &ops);
+
+	/* Now we assume that hardware will catch write bitflip errors */
+	/* If you are paranoid, use CONFIG_MTD_NAND_VERIFY_WRITE */
+
+	if (ret) {
+		dbg("write to block %d at zone %d, failed with error %d",
+			block, zone, ret);
+
+		sm_recheck_media(ftl);
+		return ret;
+	}
+
+	/* This should never happen, unless there is a bug in the driver */
+	WARN_ON(ops.oobretlen != SM_OOB_SIZE);
+	WARN_ON(buffer && ops.retlen != SM_SECTOR_SIZE);
+
+	return 0;
+}
+
+/* ------------------------ block IO ------------------------------------- */
+
+/* Write a block using data and lba, and invalid sector bitmap */
+static int sm_write_block(struct sm_ftl *ftl, u8 *buf,
+				int zone, int block, int lba,
+						unsigned long invalid_bitmap)
+{
+	struct sm_oob oob;
+	int boffset;
+	int retry = 0;
+
+	/* Initialize the oob with requested values */
+	memset(&oob, 0xFF, SM_OOB_SIZE);
+	sm_write_lba(&oob, lba);
+restart:
+	if (ftl->unstable)
+		return -EIO;
+
+	for (boffset = 0; boffset < ftl->block_size;
+				boffset += SM_SECTOR_SIZE) {
+
+		oob.data_status = 0xFF;
+
+		if (test_bit(boffset / SM_SECTOR_SIZE, &invalid_bitmap)) {
+
+			sm_printk("sector %d of block at LBA %d of zone %d"
+				" coudn't be read, marking it as invalid",
+				boffset / SM_SECTOR_SIZE, lba, zone);
+
+			oob.data_status = 0;
+		}
+
+#ifdef CONFIG_SM_FTL_MUSEUM
+		if (ftl->smallpagenand) {
+			__nand_calculate_ecc(buf + boffset,
+						SM_SMALL_PAGE, oob.ecc1);
+
+			__nand_calculate_ecc(buf + boffset + SM_SMALL_PAGE,
+						SM_SMALL_PAGE, oob.ecc2);
+		}
+#endif
+		if (!sm_write_sector(ftl, zone, block, boffset,
+							buf + boffset, &oob))
+			continue;
+
+		if (!retry) {
+
+			/* If write fails. try to erase the block */
+			/* This is safe, because we never write in blocks
+				that contain valuable data.
+			This is intended to repair block that are marked
+			as erased, but that isn't fully erased*/
+
+			if (sm_erase_block(ftl, zone, block, 0))
+				return -EIO;
+
+			retry = 1;
+			goto restart;
+		} else {
+			sm_mark_block_bad(ftl, zone, block);
+			return -EIO;
+		}
+	}
+	return 0;
+}
+
+
+/* Mark whole block at offset 'offs' as bad. */
+static void sm_mark_block_bad(struct sm_ftl *ftl, int zone, int block)
+{
+	struct sm_oob oob;
+	int boffset;
+
+	memset(&oob, 0xFF, SM_OOB_SIZE);
+	oob.block_status = 0xF0;
+
+	if (ftl->unstable)
+		return;
+
+	if (sm_recheck_media(ftl))
+		return;
+
+	sm_printk("marking block %d of zone %d as bad", block, zone);
+
+	/* We aren't checking the return value, because we don't care */
+	/* This also fails on fake xD cards, but I guess these won't expose
+		any bad blocks till fail completly */
+	for (boffset = 0; boffset < ftl->block_size; boffset += SM_SECTOR_SIZE)
+		sm_write_sector(ftl, zone, block, boffset, NULL, &oob);
+}
+
+/*
+ * Erase a block within a zone
+ * If erase succedes, it updates free block fifo, otherwise marks block as bad
+ */
+static int sm_erase_block(struct sm_ftl *ftl, int zone_num, u16 block,
+								int put_free)
+{
+	struct ftl_zone *zone = &ftl->zones[zone_num];
+	struct mtd_info *mtd = ftl->trans->mtd;
+	struct erase_info erase;
+
+	erase.mtd = mtd;
+	erase.callback = sm_erase_callback;
+	erase.addr = sm_mkoffset(ftl, zone_num, block, 0);
+	erase.len = ftl->block_size;
+	erase.priv = (u_long)ftl;
+
+	if (ftl->unstable)
+		return -EIO;
+
+	BUG_ON(ftl->readonly);
+
+	if (zone_num == 0 && (block == ftl->cis_block || block == 0)) {
+		sm_printk("attempted to erase the CIS!");
+		return -EIO;
+	}
+
+	if (mtd->erase(mtd, &erase)) {
+		sm_printk("erase of block %d in zone %d failed",
+							block, zone_num);
+		goto error;
+	}
+
+	if (erase.state == MTD_ERASE_PENDING)
+		wait_for_completion(&ftl->erase_completion);
+
+	if (erase.state != MTD_ERASE_DONE) {
+		sm_printk("erase of block %d in zone %d failed after wait",
+			block, zone_num);
+		goto error;
+	}
+
+	if (put_free)
+		kfifo_in(&zone->free_sectors,
+			(const unsigned char *)&block, sizeof(block));
+
+	return 0;
+error:
+	sm_mark_block_bad(ftl, zone_num, block);
+	return -EIO;
+}
+
+static void sm_erase_callback(struct erase_info *self)
+{
+	struct sm_ftl *ftl = (struct sm_ftl *)self->priv;
+	complete(&ftl->erase_completion);
+}
+
+/* Throughtly test that block is valid. */
+static int sm_check_block(struct sm_ftl *ftl, int zone, int block)
+{
+	int boffset;
+	struct sm_oob oob;
+	int lbas[] = { -3, 0, 0, 0 };
+	int i = 0;
+	int test_lba;
+
+
+	/* First just check that block doesn't look fishy */
+	/* Only blocks that are valid or are sliced in two parts, are
+		accepted */
+	for (boffset = 0; boffset < ftl->block_size;
+					boffset += SM_SECTOR_SIZE) {
+
+		/* This shoudn't happen anyway */
+		if (sm_read_sector(ftl, zone, block, boffset, NULL, &oob))
+			return -2;
+
+		test_lba = sm_read_lba(&oob);
+
+		if (lbas[i] != test_lba)
+			lbas[++i] = test_lba;
+
+		/* If we found three different LBAs, something is fishy */
+		if (i == 3)
+			return -EIO;
+	}
+
+	/* If the block is sliced (partialy erased usually) erase it */
+	if (i == 2) {
+		sm_erase_block(ftl, zone, block, 1);
+		return 1;
+	}
+
+	return 0;
+}
+
+/* ----------------- media scanning --------------------------------- */
+static const struct chs_entry chs_table[] = {
+	{ 1,    125,  4,  4  },
+	{ 2,    125,  4,  8  },
+	{ 4,    250,  4,  8  },
+	{ 8,    250,  4,  16 },
+	{ 16,   500,  4,  16 },
+	{ 32,   500,  8,  16 },
+	{ 64,   500,  8,  32 },
+	{ 128,  500,  16, 32 },
+	{ 256,  1000, 16, 32 },
+	{ 512,  1015, 32, 63 },
+	{ 1024, 985,  33, 63 },
+	{ 2048, 985,  33, 63 },
+	{ 0 },
+};
+
+
+static const u8 cis_signature[] = {
+	0x01, 0x03, 0xD9, 0x01, 0xFF, 0x18, 0x02, 0xDF, 0x01, 0x20
+};
+/* Find out media parameters.
+ * This ideally has to be based on nand id, but for now device size is enough */
+int sm_get_media_info(struct sm_ftl *ftl, struct mtd_info *mtd)
+{
+	int i;
+	int size_in_megs = mtd->size / (1024 * 1024);
+
+	ftl->readonly = mtd->type == MTD_ROM;
+
+	/* Manual settings for very old devices */
+	ftl->zone_count = 1;
+	ftl->smallpagenand = 0;
+
+	switch (size_in_megs) {
+	case 1:
+		/* 1 Mb flash/rom SmartMedia card (256 byte pages)*/
+		ftl->zone_size = 256;
+		ftl->max_lba = 250;
+		ftl->block_size = 8 * SM_SECTOR_SIZE;
+		ftl->smallpagenand = 1;
+
+		break;
+	case 2:
+		/* 2 Mb flash SmartMedia (256 byte pages)*/
+		if (mtd->writesize == SM_SMALL_PAGE) {
+			ftl->zone_size = 512;
+			ftl->max_lba = 500;
+			ftl->block_size = 8 * SM_SECTOR_SIZE;
+			ftl->smallpagenand = 1;
+		/* 2 Mb rom SmartMedia */
+		} else {
+
+			if (!ftl->readonly)
+				return -ENODEV;
+
+			ftl->zone_size = 256;
+			ftl->max_lba = 250;
+			ftl->block_size = 16 * SM_SECTOR_SIZE;
+		}
+		break;
+	case 4:
+		/* 4 Mb flash/rom SmartMedia device */
+		ftl->zone_size = 512;
+		ftl->max_lba = 500;
+		ftl->block_size = 16 * SM_SECTOR_SIZE;
+		break;
+	case 8:
+		/* 8 Mb flash/rom SmartMedia device */
+		ftl->zone_size = 1024;
+		ftl->max_lba = 1000;
+		ftl->block_size = 16 * SM_SECTOR_SIZE;
+	}
+
+	/* Minimum xD size is 16M. Also, all xD cards have standard zone
+	   sizes. SmartMedia cards exist up to 128 Mb and have same layout*/
+	if (size_in_megs >= 16) {
+		ftl->zone_count = size_in_megs / 16;
+		ftl->zone_size = 1024;
+		ftl->max_lba = 1000;
+		ftl->block_size = 32 * SM_SECTOR_SIZE;
+	}
+
+	/* Test for proper write,erase and oob sizes */
+	if (mtd->erasesize > ftl->block_size)
+		return -ENODEV;
+
+	if (mtd->writesize > SM_SECTOR_SIZE)
+		return -ENODEV;
+
+	if (ftl->smallpagenand && mtd->oobsize < SM_SMALL_OOB_SIZE)
+		return -ENODEV;
+
+	if (!ftl->smallpagenand && mtd->oobsize < SM_OOB_SIZE)
+		return -ENODEV;
+
+	/* We use these functions for IO */
+	if (!mtd->read_oob || !mtd->write_oob)
+		return -ENODEV;
+
+	/* Find geometry information */
+	for (i = 0 ; i < ARRAY_SIZE(chs_table) ; i++) {
+		if (chs_table[i].size == size_in_megs) {
+			ftl->cylinders = chs_table[i].cyl;
+			ftl->heads = chs_table[i].head;
+			ftl->sectors = chs_table[i].sec;
+			return 0;
+		}
+	}
+
+	sm_printk("media has unknown size : %dMiB", size_in_megs);
+	ftl->cylinders = 985;
+	ftl->heads =  33;
+	ftl->sectors = 63;
+	return 0;
+}
+
+/* Validate the CIS */
+static int sm_read_cis(struct sm_ftl *ftl)
+{
+	struct sm_oob oob;
+
+	if (sm_read_sector(ftl,
+		0, ftl->cis_block, ftl->cis_boffset, ftl->cis_buffer, &oob))
+			return -EIO;
+
+	if (!sm_sector_valid(&oob) || !sm_block_valid(&oob))
+		return -EIO;
+
+	if (!memcmp(ftl->cis_buffer + ftl->cis_page_offset,
+			cis_signature, sizeof(cis_signature))) {
+		return 0;
+	}
+
+	return -EIO;
+}
+
+/* Scan the media for the CIS */
+static int sm_find_cis(struct sm_ftl *ftl)
+{
+	struct sm_oob oob;
+	int block, boffset;
+	int block_found = 0;
+	int cis_found = 0;
+
+	/* Search for first valid block */
+	for (block = 0 ; block < ftl->zone_size - ftl->max_lba ; block++) {
+
+		if (sm_read_sector(ftl, 0, block, 0, NULL, &oob))
+			continue;
+
+		if (!sm_block_valid(&oob))
+			continue;
+		block_found = 1;
+		break;
+	}
+
+	if (!block_found)
+		return -EIO;
+
+	/* Search for first valid sector in this block */
+	for (boffset = 0 ; boffset < ftl->block_size;
+						boffset += SM_SECTOR_SIZE) {
+
+		if (sm_read_sector(ftl, 0, block, boffset, NULL, &oob))
+			continue;
+
+		if (!sm_sector_valid(&oob))
+			continue;
+		break;
+	}
+
+	if (boffset == ftl->block_size)
+		return -EIO;
+
+	ftl->cis_block = block;
+	ftl->cis_boffset = boffset;
+	ftl->cis_page_offset = 0;
+
+	cis_found = !sm_read_cis(ftl);
+
+	if (!cis_found) {
+		ftl->cis_page_offset = SM_SMALL_PAGE;
+		cis_found = !sm_read_cis(ftl);
+	}
+
+	if (cis_found) {
+		dbg("CIS block found at offset %x",
+			block * ftl->block_size +
+				boffset + ftl->cis_page_offset);
+		return 0;
+	}
+	return -EIO;
+}
+
+/* Basic test to determine if underlying mtd device if functional */
+static int sm_recheck_media(struct sm_ftl *ftl)
+{
+	if (sm_read_cis(ftl)) {
+
+		if (!ftl->unstable) {
+			sm_printk("media unstable, not allowing writes");
+			ftl->unstable = 1;
+		}
+		return -EIO;
+	}
+	return 0;
+}
+
+/* Initialize a FTL zone */
+static int sm_init_zone(struct sm_ftl *ftl, int zone_num)
+{
+	struct ftl_zone *zone = &ftl->zones[zone_num];
+	struct sm_oob oob;
+	u16 block;
+	int lba;
+	int i = 0;
+
+	dbg("initializing zone %d", zone_num);
+
+	/* Allocate memory for FTL table */
+	zone->lba_to_phys_table = kmalloc(ftl->max_lba * 2, GFP_KERNEL);
+
+	if (!zone->lba_to_phys_table)
+		return -ENOMEM;
+	memset(zone->lba_to_phys_table, -1, ftl->max_lba * 2);
+
+
+	/* Allocate memory for free sectors FIFO */
+	if (kfifo_alloc(&zone->free_sectors, ftl->zone_size * 2, GFP_KERNEL)) {
+		kfree(zone->lba_to_phys_table);
+		return -ENOMEM;
+	}
+
+	/* Now scan the zone */
+	for (block = 0 ; block < ftl->zone_size ; block++) {
+
+		/* Skip blocks till the CIS (including) */
+		if (zone_num == 0 && block <= ftl->cis_block)
+			continue;
+
+		/* Read the oob of first sector */
+		if (sm_read_sector(ftl, zone_num, block, 0, NULL, &oob))
+			return -EIO;
+
+		/* Test to see if block is erased. It is enough to test
+			first sector, because erase happens in one shot */
+		if (sm_block_erased(&oob)) {
+			kfifo_in(&zone->free_sectors,
+				(unsigned char *)&block, 2);
+			continue;
+		}
+
+		/* If block is marked as bad, skip it */
+		/* This assumes we can trust first sector*/
+		/* However the way the block valid status is defined, ensures
+			very low probability of failure here */
+		if (!sm_block_valid(&oob)) {
+			dbg("PH %04d <-> <marked bad>", block);
+			continue;
+		}
+
+
+		lba = sm_read_lba(&oob);
+
+		/* Invalid LBA means that block is damaged. */
+		/* We can try to erase it, or mark it as bad, but
+			lets leave that to recovery application */
+		if (lba == -2 || lba >= ftl->max_lba) {
+			dbg("PH %04d <-> LBA %04d(bad)", block, lba);
+			continue;
+		}
+
+
+		/* If there is no collision,
+			just put the sector in the FTL table */
+		if (zone->lba_to_phys_table[lba] < 0) {
+			dbg_verbose("PH %04d <-> LBA %04d", block, lba);
+			zone->lba_to_phys_table[lba] = block;
+			continue;
+		}
+
+		sm_printk("collision"
+			" of LBA %d between blocks %d and %d in zone %d",
+			lba, zone->lba_to_phys_table[lba], block, zone_num);
+
+		/* Test that this block is valid*/
+		if (sm_check_block(ftl, zone_num, block))
+			continue;
+
+		/* Test now the old block */
+		if (sm_check_block(ftl, zone_num,
+					zone->lba_to_phys_table[lba])) {
+			zone->lba_to_phys_table[lba] = block;
+			continue;
+		}
+
+		/* If both blocks are valid and share same LBA, it means that
+			they hold different versions of same data. It not
+			known which is more recent, thus just erase one of them
+		*/
+		sm_printk("both blocks are valid, erasing the later");
+		sm_erase_block(ftl, zone_num, block, 1);
+	}
+
+	dbg("zone initialized");
+	zone->initialized = 1;
+
+	/* No free sectors, means that the zone is heavily damaged, write won't
+		work, but it can still can be (partially) read */
+	if (!kfifo_len(&zone->free_sectors)) {
+		sm_printk("no free blocks in zone %d", zone_num);
+		return 0;
+	}
+
+	/* Randomize first block we write to */
+	get_random_bytes(&i, 2);
+	i %= (kfifo_len(&zone->free_sectors) / 2);
+
+	while (i--) {
+		kfifo_out(&zone->free_sectors, (unsigned char *)&block, 2);
+		kfifo_in(&zone->free_sectors, (const unsigned char *)&block, 2);
+	}
+	return 0;
+}
+
+/* Get and automaticly initialize an FTL mapping for one zone */
+struct ftl_zone *sm_get_zone(struct sm_ftl *ftl, int zone_num)
+{
+	struct ftl_zone *zone;
+	int error;
+
+	BUG_ON(zone_num >= ftl->zone_count);
+	zone = &ftl->zones[zone_num];
+
+	if (!zone->initialized) {
+		error = sm_init_zone(ftl, zone_num);
+
+		if (error)
+			return ERR_PTR(error);
+	}
+	return zone;
+}
+
+
+/* ----------------- cache handling ------------------------------------------*/
+
+/* Initialize the one block cache */
+void sm_cache_init(struct sm_ftl *ftl)
+{
+	ftl->cache_data_invalid_bitmap = 0xFFFFFFFF;
+	ftl->cache_clean = 1;
+	ftl->cache_zone = -1;
+	ftl->cache_block = -1;
+	/*memset(ftl->cache_data, 0xAA, ftl->block_size);*/
+}
+
+/* Put sector in one block cache */
+void sm_cache_put(struct sm_ftl *ftl, char *buffer, int boffset)
+{
+	memcpy(ftl->cache_data + boffset, buffer, SM_SECTOR_SIZE);
+	clear_bit(boffset / SM_SECTOR_SIZE, &ftl->cache_data_invalid_bitmap);
+	ftl->cache_clean = 0;
+}
+
+/* Read a sector from the cache */
+int sm_cache_get(struct sm_ftl *ftl, char *buffer, int boffset)
+{
+	if (test_bit(boffset / SM_SECTOR_SIZE,
+		&ftl->cache_data_invalid_bitmap))
+			return -1;
+
+	memcpy(buffer, ftl->cache_data + boffset, SM_SECTOR_SIZE);
+	return 0;
+}
+
+/* Write the cache to hardware */
+int sm_cache_flush(struct sm_ftl *ftl)
+{
+	struct ftl_zone *zone;
+
+	int sector_num;
+	u16 write_sector;
+	int zone_num = ftl->cache_zone;
+	int block_num;
+
+	if (ftl->cache_clean)
+		return 0;
+
+	if (ftl->unstable)
+		return -EIO;
+
+	BUG_ON(zone_num < 0);
+	zone = &ftl->zones[zone_num];
+	block_num = zone->lba_to_phys_table[ftl->cache_block];
+
+
+	/* Try to read all unread areas of the cache block*/
+	for_each_bit(sector_num, &ftl->cache_data_invalid_bitmap,
+		ftl->block_size / SM_SECTOR_SIZE) {
+
+		if (!sm_read_sector(ftl,
+			zone_num, block_num, sector_num * SM_SECTOR_SIZE,
+			ftl->cache_data + sector_num * SM_SECTOR_SIZE, NULL))
+				clear_bit(sector_num,
+					&ftl->cache_data_invalid_bitmap);
+	}
+restart:
+
+	if (ftl->unstable)
+		return -EIO;
+	/* No spare blocks */
+	/* We could still continue by erasing the current block,
+		but for such worn out media it doesn't worth the trouble,
+			and the dangers */
+
+	if (!kfifo_len(&zone->free_sectors)) {
+		dbg("no free sectors for write!");
+		return -EIO;
+	}
+
+	kfifo_out(&zone->free_sectors, (unsigned char *)&write_sector, 2);
+
+	if (sm_write_block(ftl, ftl->cache_data, zone_num, write_sector,
+		ftl->cache_block, ftl->cache_data_invalid_bitmap))
+			goto restart;
+
+	/* Update the FTL table */
+	zone->lba_to_phys_table[ftl->cache_block] = write_sector;
+
+	/* Write succesfull, so erase and free the old block */
+	if (block_num > 0)
+		sm_erase_block(ftl, zone_num, block_num, 1);
+
+	sm_cache_init(ftl);
+	return 0;
+}
+
+
+/* flush timer, runs a second after last write */
+static void sm_cache_flush_timer(unsigned long data)
+{
+	struct sm_ftl *ftl = (struct sm_ftl *)data;
+	queue_work(cache_flush_workqueue, &ftl->flush_work);
+}
+
+/* cache flush work, kicked by timer */
+static void sm_cache_flush_work(struct work_struct *work)
+{
+	struct sm_ftl *ftl = container_of(work, struct sm_ftl, flush_work);
+	mutex_lock(&ftl->mutex);
+	sm_cache_flush(ftl);
+	mutex_unlock(&ftl->mutex);
+	return;
+}
+
+/* ---------------- outside interface -------------------------------------- */
+
+/* outside interface: read a sector */
+static int sm_read(struct mtd_blktrans_dev *dev,
+				unsigned long sect_no, char *buf)
+{
+	struct sm_ftl *ftl = dev->priv;
+	struct ftl_zone *zone;
+	int error = 0, in_cache = 0;
+	int zone_num, block, boffset;
+
+	sm_break_offset(ftl, sect_no << 9, &zone_num, &block, &boffset);
+	mutex_lock(&ftl->mutex);
+
+
+	zone = sm_get_zone(ftl, zone_num);
+	if (IS_ERR(zone)) {
+		error = PTR_ERR(zone);
+		goto unlock;
+	}
+
+	/* Have to look at cache first */
+	if (ftl->cache_zone == zone_num && ftl->cache_block == block) {
+		in_cache = 1;
+		if (!sm_cache_get(ftl, buf, boffset))
+			goto unlock;
+	}
+
+	/* Translate the block and return if doesn't exist in the table */
+	block = zone->lba_to_phys_table[block];
+
+	if (block == -1) {
+		memset(buf, 0xFF, SM_SECTOR_SIZE);
+		goto unlock;
+	}
+
+	if (sm_read_sector(ftl, zone_num, block, boffset, buf, NULL)) {
+		error = -EIO;
+		goto unlock;
+	}
+
+	if (in_cache)
+		sm_cache_put(ftl, buf, boffset);
+unlock:
+	mutex_unlock(&ftl->mutex);
+	return error;
+}
+
+/* outside interface: write a sector */
+static int sm_write(struct mtd_blktrans_dev *dev,
+				unsigned long sec_no, char *buf)
+{
+	struct sm_ftl *ftl = dev->priv;
+	struct ftl_zone *zone;
+	int error, zone_num, block, boffset;
+
+	BUG_ON(ftl->readonly);
+	sm_break_offset(ftl, sec_no << 9, &zone_num, &block, &boffset);
+
+	/* No need in flush thread running now */
+	del_timer(&ftl->timer);
+	mutex_lock(&ftl->mutex);
+
+	zone = sm_get_zone(ftl, zone_num);
+	if (IS_ERR(zone)) {
+		error = PTR_ERR(zone);
+		goto unlock;
+	}
+
+	/* If entry is not in cache, flush it */
+	if (ftl->cache_block != block || ftl->cache_zone != zone_num) {
+
+		error = sm_cache_flush(ftl);
+		if (error)
+			goto unlock;
+
+		ftl->cache_block = block;
+		ftl->cache_zone = zone_num;
+	}
+
+	sm_cache_put(ftl, buf, boffset);
+unlock:
+	mod_timer(&ftl->timer, jiffies + msecs_to_jiffies(cache_timeout));
+	mutex_unlock(&ftl->mutex);
+	return error;
+}
+
+/* outside interface: flush everything */
+static int sm_flush(struct mtd_blktrans_dev *dev)
+{
+	struct sm_ftl *ftl = dev->priv;
+	int retval;
+
+	mutex_lock(&ftl->mutex);
+	retval =  sm_cache_flush(ftl);
+	mutex_unlock(&ftl->mutex);
+	return retval;
+}
+
+/* outside interface: device is released */
+static int sm_release(struct mtd_blktrans_dev *dev)
+{
+	struct sm_ftl *ftl = dev->priv;
+
+	mutex_lock(&ftl->mutex);
+	del_timer_sync(&ftl->timer);
+	cancel_work_sync(&ftl->flush_work);
+	sm_cache_flush(ftl);
+	mutex_unlock(&ftl->mutex);
+	return 0;
+}
+
+/* outside interface: get geometry */
+static int sm_getgeo(struct mtd_blktrans_dev *dev, struct hd_geometry *geo)
+{
+	struct sm_ftl *ftl = dev->priv;
+	geo->heads = ftl->heads;
+	geo->sectors = ftl->sectors;
+	geo->cylinders = ftl->cylinders;
+	return 0;
+}
+
+/* external interface: main initialization function */
+static void sm_add_mtd(struct mtd_blktrans_ops *tr, struct mtd_info *mtd)
+{
+	struct mtd_blktrans_dev *trans;
+	struct sm_ftl *ftl;
+
+	/* Allocate & initialize our private structure */
+	ftl = kzalloc(sizeof(struct sm_ftl), GFP_KERNEL);
+	if (!ftl)
+		goto error1;
+
+
+	mutex_init(&ftl->mutex);
+	setup_timer(&ftl->timer, sm_cache_flush_timer, (unsigned long)ftl);
+	INIT_WORK(&ftl->flush_work, sm_cache_flush_work);
+	init_completion(&ftl->erase_completion);
+
+	/* Read media information */
+	if (sm_get_media_info(ftl, mtd)) {
+		dbg("found unsupported mtd device, aborting");
+		goto error2;
+	}
+
+
+	/* Allocate temporary CIS buffer for read retry support */
+	ftl->cis_buffer = kzalloc(SM_SECTOR_SIZE, GFP_KERNEL);
+	if (!ftl->cis_buffer)
+		goto error2;
+
+	/* Allocate zone array, it will be initialized on demand */
+	ftl->zones = kzalloc(sizeof(struct ftl_zone) * ftl->zone_count,
+								GFP_KERNEL);
+	if (!ftl->zones)
+		goto error3;
+
+	/* Allocate the cache*/
+	ftl->cache_data = kzalloc(ftl->block_size, GFP_KERNEL);
+
+	if (!ftl->cache_data)
+		goto error4;
+
+	sm_cache_init(ftl);
+
+
+	/* Allocate upper layer structure and initialize it */
+	trans = kzalloc(sizeof(struct mtd_blktrans_dev), GFP_KERNEL);
+	if (!trans)
+		goto error5;
+
+	ftl->trans = trans;
+	trans->priv = ftl;
+
+	trans->tr = tr;
+	trans->mtd = mtd;
+	trans->devnum = -1;
+	trans->size = (ftl->block_size * ftl->max_lba * ftl->zone_count) >> 9;
+	trans->readonly = ftl->readonly;
+
+	if (sm_find_cis(ftl)) {
+		dbg("CIS not found on mtd device, aborting");
+		goto error6;
+	}
+
+	ftl->disk_attributes = sm_create_sysfs_attributes(ftl);
+	trans->disk_attributes = ftl->disk_attributes;
+
+	sm_printk("Found %d MiB xD/SmartMedia FTL on mtd%d",
+		(int)(mtd->size / (1024 * 1024)), mtd->index);
+
+	dbg("FTL layout:");
+	dbg("%d zone(s), each consists of %d blocks (+%d spares)",
+		ftl->zone_count, ftl->max_lba,
+		ftl->zone_size - ftl->max_lba);
+	dbg("each block consists of %d bytes",
+		ftl->block_size);
+
+
+	/* Register device*/
+	if (add_mtd_blktrans_dev(trans)) {
+		dbg("error in mtdblktrans layer");
+		goto error6;
+	}
+	return;
+error6:
+	kfree(trans);
+error5:
+	kfree(ftl->cache_data);
+error4:
+	kfree(ftl->zones);
+error3:
+	kfree(ftl->cis_buffer);
+error2:
+	kfree(ftl);
+error1:
+	return;
+}
+
+/* main interface: device {surprise,} removal */
+static void sm_remove_dev(struct mtd_blktrans_dev *dev)
+{
+	struct sm_ftl *ftl = dev->priv;
+	int i;
+
+	del_mtd_blktrans_dev(dev);
+	ftl->trans = NULL;
+
+	for (i = 0 ; i < ftl->zone_count; i++) {
+
+		if (!ftl->zones[i].initialized)
+			continue;
+
+		kfree(ftl->zones[i].lba_to_phys_table);
+		kfifo_free(&ftl->zones[i].free_sectors);
+	}
+
+	sm_delete_sysfs_attributes(ftl);
+	kfree(ftl->cis_buffer);
+	kfree(ftl->zones);
+	kfree(ftl->cache_data);
+	kfree(ftl);
+}
+
+static struct mtd_blktrans_ops sm_ftl_ops = {
+	.name		= "smblk",
+	.major		= -1,
+	.part_bits	= SM_FTL_PARTN_BITS,
+	.blksize	= SM_SECTOR_SIZE,
+	.getgeo		= sm_getgeo,
+
+	.add_mtd	= sm_add_mtd,
+	.remove_dev	= sm_remove_dev,
+
+	.readsect	= sm_read,
+	.writesect	= sm_write,
+
+	.flush		= sm_flush,
+	.release	= sm_release,
+
+	.owner		= THIS_MODULE,
+};
+
+static __init int sm_module_init(void)
+{
+	int error = 0;
+	cache_flush_workqueue = create_freezeable_workqueue("smflush");
+
+	if (IS_ERR(cache_flush_workqueue))
+		return PTR_ERR(cache_flush_workqueue);
+
+	error = register_mtd_blktrans(&sm_ftl_ops);
+	if (error)
+		destroy_workqueue(cache_flush_workqueue);
+	return error;
+
+}
+
+static void __exit sm_module_exit(void)
+{
+	destroy_workqueue(cache_flush_workqueue);
+	deregister_mtd_blktrans(&sm_ftl_ops);
+}
+
+module_init(sm_module_init);
+module_exit(sm_module_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Maxim Levitsky <maximlevitsky@gmail.com>");
+MODULE_DESCRIPTION("Smartmedia/xD mtd translation layer");
diff --git a/drivers/mtd/sm_ftl.h b/drivers/mtd/sm_ftl.h
new file mode 100644
index 0000000..519e2e1
--- /dev/null
+++ b/drivers/mtd/sm_ftl.h
@@ -0,0 +1,94 @@
+/*
+ * Copyright (C) 2009 - Maxim Levitsky
+ * SmartMedia/xD translation layer
+ *
+ * Based loosly on ssfdc.c which is
+ *  (c) 2005 Eptar srl
+ *  Author: Claudio Lanconelli <lanconelli.claudio@eptar.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/mtd/blktrans.h>
+#include <linux/kfifo.h>
+#include <linux/sched.h>
+#include <linux/completion.h>
+#include <linux/mtd/mtd.h>
+
+
+
+struct ftl_zone {
+	int initialized;
+	s16 *lba_to_phys_table;		/* LBA to physical table */
+	struct kfifo free_sectors;	/* queue of free sectors */
+};
+
+struct sm_ftl {
+	struct mtd_blktrans_dev *trans;
+
+	struct mutex mutex;		/* protects the structure */
+	struct ftl_zone *zones;		/* FTL tables for each zone */
+
+	/* Media information */
+	int block_size;			/* block size in bytes */
+	int zone_size;			/* zone size in blocks */
+	int zone_count;			/* number of zones */
+	int max_lba;			/* maximum lba in a zone */
+	int smallpagenand;		/* 256 bytes/page nand */
+	int readonly;			/* is FS readonly */
+	int unstable;
+	int cis_block;			/* CIS block location */
+	int cis_boffset;		/* CIS offset in the block */
+	int cis_page_offset;		/* CIS offset in the page */
+	void *cis_buffer;		/* tmp buffer for cis reads */
+
+	/* Cache */
+	int cache_block;		/* block number of cached block */
+	int cache_zone;			/* zone of cached block */
+	unsigned char *cache_data;	/* cached block data */
+	long unsigned int cache_data_invalid_bitmap;
+	int cache_clean;
+	struct work_struct flush_work;
+	struct timer_list timer;
+
+	/* Async erase stuff */
+	struct completion erase_completion;
+
+	/* Geometry stuff */
+	int heads;
+	int sectors;
+	int cylinders;
+
+	struct attribute_group *disk_attributes;
+};
+
+struct chs_entry {
+	unsigned long size;
+	unsigned short cyl;
+	unsigned char head;
+	unsigned char sec;
+};
+
+
+#define SM_FTL_PARTN_BITS	3
+
+#define sm_printk(format, ...) \
+	printk(KERN_WARNING "sm_ftl" ": " format "\n", ## __VA_ARGS__)
+
+#define dbg(format, ...) \
+	if (debug) \
+		printk(KERN_DEBUG "sm_ftl" ": " format "\n", ## __VA_ARGS__)
+
+#define dbg_verbose(format, ...) \
+	if (debug > 1) \
+		printk(KERN_DEBUG "sm_ftl" ": " format "\n", ## __VA_ARGS__)
+
+
+static void sm_erase_callback(struct erase_info *self);
+static int sm_erase_block(struct sm_ftl *ftl, int zone_num, u16 block,
+								int put_free);
+static void sm_mark_block_bad(struct sm_ftl *ftl, int zone_num, int block);
+
+static int sm_recheck_media(struct sm_ftl *ftl);
-- 
1.6.3.3

[PATCH 14/15] MTD: Add new SmartMedia/xD FTL

[Maxim Levitsky]

This implements new readwrite SmartMedia/xd FTL.

mtd driver must have support proper ECC and badblock vertification
based on oob parts for 512 bytes nand.

Also mtd driver must define read_oob and write_oob, which are used
to read and write both data and oob together.

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 drivers/mtd/Kconfig  |   21 +
 drivers/mtd/Makefile |    1 +
 drivers/mtd/sm_ftl.c | 1284 ++++++++++++++++++++++++++++++++++++++++++++++++++
 drivers/mtd/sm_ftl.h |   94 ++++
 4 files changed, 1400 insertions(+), 0 deletions(-)
 create mode 100644 drivers/mtd/sm_ftl.c
 create mode 100644 drivers/mtd/sm_ftl.h

diff --git a/drivers/mtd/Kconfig b/drivers/mtd/Kconfig
index ecf90f5..0105ae0 100644
--- a/drivers/mtd/Kconfig
+++ b/drivers/mtd/Kconfig
@@ -304,6 +304,27 @@ config SSFDC
 	  This enables read only access to SmartMedia formatted NAND
 	  flash. You can mount it with FAT file system.
 
+
+config SM_FTL
+	tristate "SmartMedia/xD new translation layer"
+	depends on EXPERIMENTAL && BLOCK
+	select MTD_BLKDEVS
+	help
+	  This enables new and very EXPERMENTAL support for SmartMedia/xD
+	  FTL (Flash tanslation layer)
+	  Write support isn't yet well tested, therefore this code IS likely to
+	  eat your card, so please don't use it together with valuable data.
+	  Use readonly driver (CONFIG_SSFDC) instead.
+
+config SM_FTL_MUSEUM
+	boolean "Additional Support for 1MB and 2MB SmartMedia cards"
+	depends on SM_FTL && MTD_NAND
+	select MTD_NAND_ECC_SMC
+	help
+	  Very old SmartMedia cards need ECC to be caluculated in the FTL
+	  Such cards are very rare, thus enabling this option is mostly useless
+	  Also this support is completely UNTESTED.
+
 config MTD_OOPS
 	tristate "Log panic/oops to an MTD buffer"
 	depends on MTD
diff --git a/drivers/mtd/Makefile b/drivers/mtd/Makefile
index 82d1e4d..d53357b 100644
--- a/drivers/mtd/Makefile
+++ b/drivers/mtd/Makefile
@@ -24,6 +24,7 @@ obj-$(CONFIG_NFTL)		+= nftl.o
 obj-$(CONFIG_INFTL)		+= inftl.o
 obj-$(CONFIG_RFD_FTL)		+= rfd_ftl.o
 obj-$(CONFIG_SSFDC)		+= ssfdc.o
+obj-$(CONFIG_SM_FTL)		+= sm_ftl.o
 obj-$(CONFIG_MTD_OOPS)		+= mtdoops.o
 
 nftl-objs		:= nftlcore.o nftlmount.o
diff --git a/drivers/mtd/sm_ftl.c b/drivers/mtd/sm_ftl.c
new file mode 100644
index 0000000..c2a409f
--- /dev/null
+++ b/drivers/mtd/sm_ftl.c
@@ -0,0 +1,1284 @@
+/*
+ * Copyright (C) 2009 - Maxim Levitsky
+ * SmartMedia/xD translation layer
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/random.h>
+#include <linux/hdreg.h>
+#include <linux/kthread.h>
+#include <linux/freezer.h>
+#include <linux/sysfs.h>
+#include <linux/bitops.h>
+#include "nand/sm_common.h"
+#include "sm_ftl.h"
+
+#ifdef CONFIG_SM_FTL_MUSEUM
+#include <linux/mtd/nand_ecc.h>
+#endif
+
+
+struct workqueue_struct *cache_flush_workqueue;
+
+static int cache_timeout = 1000;
+module_param(cache_timeout, bool, S_IRUGO);
+MODULE_PARM_DESC(cache_timeout,
+	"Timeout (in ms) for cache flush (1000 ms default");
+
+static int debug;
+module_param(debug, int, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(debug, "Debug level (0-2)");
+
+
+/* ------------------- sysfs attributtes ---------------------------------- */
+struct sm_sysfs_attribute {
+	struct device_attribute dev_attr;
+	char *data;
+	int len;
+};
+
+ssize_t sm_attr_show(struct device *dev, struct device_attribute *attr,
+			char *buf)
+{
+	struct sm_sysfs_attribute *sm_attr =
+		container_of(attr, struct sm_sysfs_attribute, dev_attr);
+
+	strncpy(buf, sm_attr->data, sm_attr->len);
+	return sm_attr->len;
+}
+
+
+#define NUM_ATTRIBUTES 1
+#define SM_CIS_VENDOR_OFFSET 0x59
+struct attribute_group *sm_create_sysfs_attributes(struct sm_ftl *ftl)
+{
+	struct attribute_group *attr_group;
+	struct attribute **attributes;
+	struct sm_sysfs_attribute *vendor_attribute;
+
+	int vendor_len = strnlen(ftl->cis_buffer + SM_CIS_VENDOR_OFFSET,
+					SM_SMALL_PAGE - SM_CIS_VENDOR_OFFSET);
+
+	char *vendor = kmalloc(vendor_len, GFP_KERNEL);
+	memcpy(vendor, ftl->cis_buffer + SM_CIS_VENDOR_OFFSET, vendor_len);
+	vendor[vendor_len] = 0;
+
+	/* Initialize sysfs attributes */
+	vendor_attribute =
+		kzalloc(sizeof(struct sm_sysfs_attribute), GFP_KERNEL);
+
+	vendor_attribute->data = vendor;
+	vendor_attribute->len = vendor_len;
+	vendor_attribute->dev_attr.attr.name = "vendor";
+	vendor_attribute->dev_attr.attr.mode = S_IRUGO;
+	vendor_attribute->dev_attr.show = sm_attr_show;
+
+
+	/* Create array of pointers to the attributes */
+	attributes = kzalloc(sizeof(struct attribute *) * (NUM_ATTRIBUTES + 1),
+								GFP_KERNEL);
+	attributes[0] = &vendor_attribute->dev_attr.attr;
+
+	/* Finally create the attribute group */
+	attr_group = kzalloc(sizeof(struct attribute_group), GFP_KERNEL);
+	attr_group->attrs = attributes;
+	return attr_group;
+}
+
+void sm_delete_sysfs_attributes(struct sm_ftl *ftl)
+{
+	struct attribute **attributes = ftl->disk_attributes->attrs;
+	int i;
+
+	for (i = 0; attributes[i] ; i++) {
+
+		struct device_attribute *dev_attr = container_of(attributes[i],
+			struct device_attribute, attr);
+
+		struct sm_sysfs_attribute *sm_attr =
+			container_of(dev_attr,
+				struct sm_sysfs_attribute, dev_attr);
+
+		kfree(sm_attr->data);
+		kfree(sm_attr);
+	}
+
+	kfree(ftl->disk_attributes->attrs);
+	kfree(ftl->disk_attributes);
+}
+
+
+/* ----------------------- oob helpers -------------------------------------- */
+
+static int sm_get_lba(u8 *lba)
+{
+	/* check fixed bits */
+	if ((lba[0] & 0xF8) != 0x10)
+		return -2;
+
+	/* check parity - endianess doesn't matter */
+	if (hweight16(*(u16 *)lba) & 1)
+		return -2;
+
+	return (lba[1] >> 1) | ((lba[0] & 0x07) << 7);
+}
+
+
+/*
+ * Read LBA asscociated with block
+ * returns -1, if block is erased
+ * returns -2 if error happens
+ */
+static int sm_read_lba(struct sm_oob *oob)
+{
+	static const u32 erased_pattern[4] = {
+		0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF };
+
+	u16 lba_test;
+	int lba;
+
+	/* First test for erased block */
+	if (!memcmp(oob, erased_pattern, SM_OOB_SIZE))
+		return -1;
+
+	/* Now check is both copies of the LBA differ too much */
+	lba_test = *(u16 *)oob->lba_copy1 ^ *(u16*)oob->lba_copy2;
+	if (lba_test && !is_power_of_2(lba_test))
+		return -2;
+
+	/* And read it */
+	lba = sm_get_lba(oob->lba_copy1);
+
+	if (lba == -2)
+		lba = sm_get_lba(oob->lba_copy2);
+
+	return lba;
+}
+
+static void sm_write_lba(struct sm_oob *oob, u16 lba)
+{
+	u8 tmp[2];
+
+	WARN_ON(lba >= 1000);
+
+	tmp[0] = 0x10 | ((lba >> 7) & 0x07);
+	tmp[1] = (lba << 1) & 0xFF;
+
+	if (hweight16(*(u16 *)tmp) & 0x01)
+		tmp[1] |= 1;
+
+	oob->lba_copy1[0] = oob->lba_copy2[0] = tmp[0];
+	oob->lba_copy1[1] = oob->lba_copy2[1] = tmp[1];
+}
+
+
+/* Make offset from parts */
+static loff_t sm_mkoffset(struct sm_ftl *ftl, int zone, int block, int boffset)
+{
+	WARN_ON(boffset & (SM_SECTOR_SIZE - 1));
+	WARN_ON(zone < 0 || zone >= ftl->zone_count);
+	WARN_ON(block >= ftl->zone_size);
+	WARN_ON(boffset >= ftl->block_size);
+
+	if (block == -1)
+		return -1;
+
+	return (zone * SM_MAX_ZONE_SIZE + block) * ftl->block_size + boffset;
+}
+
+/* Breaks offset into parts */
+static void sm_break_offset(struct sm_ftl *ftl, loff_t offset,
+					int *zone, int *block, int *boffset)
+{
+	*boffset = do_div(offset, ftl->block_size);
+	*block = do_div(offset, ftl->max_lba);
+	*zone = offset >= ftl->zone_count ? -1 : offset;
+}
+
+/* ---------------------- low level IO ------------------------------------- */
+
+static int sm_correct_sector(u8 *buffer, struct sm_oob *oob)
+{
+#ifdef CONFIG_SM_FTL_MUSEUM
+	u8 ecc[3];
+
+	__nand_calculate_ecc(buffer, SM_SMALL_PAGE, ecc);
+	if (__nand_correct_data(buffer, ecc, oob->ecc1, SM_SMALL_PAGE) < 0)
+		return -EIO;
+
+	buffer += SM_SMALL_PAGE;
+
+	__nand_calculate_ecc(buffer, SM_SMALL_PAGE, ecc);
+	if (__nand_correct_data(buffer, ecc, oob->ecc2, SM_SMALL_PAGE) < 0)
+		return -EIO;
+#endif
+	return 0;
+}
+
+/* Reads a sector + oob*/
+static int sm_read_sector(struct sm_ftl *ftl,
+			int zone, int block, int boffset,
+					u8 *buffer, struct sm_oob *oob)
+{
+	struct mtd_info *mtd = ftl->trans->mtd;
+	struct mtd_oob_ops ops;
+	struct sm_oob tmp_oob;
+	int ret;
+	int try = 0;
+
+	/* FTL can contain -1 entries that are by default filled with bits */
+	if (block == -1) {
+		memset(buffer, 0xFF, SM_SECTOR_SIZE);
+		return 0;
+	}
+
+	/* User might not need the oob, but we do for data vertification */
+	if (!oob)
+		oob = &tmp_oob;
+
+	ops.mode = ftl->smallpagenand ? MTD_OOB_RAW : MTD_OOB_PLACE;
+	ops.ooboffs = 0;
+	ops.ooblen = SM_OOB_SIZE;
+	ops.oobbuf = (void *)oob;
+	ops.len = SM_SECTOR_SIZE;
+	ops.datbuf = buffer;
+
+again:
+	if (try++) {
+		/* Avoid infinite recursion on CIS reads, sm_recheck_media
+			won't help anyway */
+		if (zone == 0 && block == ftl->cis_block && boffset ==
+			ftl->cis_boffset)
+			return ret;
+
+		/* Test if media is stable */
+		if (try == 3 || sm_recheck_media(ftl))
+			return ret;
+	}
+
+	/* Unfortunelly, oob read will _always_ succeed,
+		despite card removal..... */
+	ret = mtd->read_oob(mtd, sm_mkoffset(ftl, zone, block, boffset), &ops);
+
+	/* Test for unknown errors */
+	if (ret != 0 && ret != -EUCLEAN && ret != -EBADMSG) {
+		dbg("read of block %d at zone %d, failed due to error (%d)",
+			block, zone, ret);
+		goto again;
+	}
+
+	/* Do a basic test on the oob, to guard against returned garbage */
+	if (oob->reserved != 0xFFFFFFFF && !is_power_of_2(~oob->reserved))
+		goto again;
+
+	/* This should never happen, unless there is a bug in the mtd driver */
+	WARN_ON(ops.oobretlen != SM_OOB_SIZE);
+	WARN_ON(buffer && ops.retlen != SM_SECTOR_SIZE);
+
+	if (!buffer)
+		return 0;
+
+	/* Test if sector marked as bad */
+	if (!sm_sector_valid(oob)) {
+		dbg("read of block %d at zone %d, failed because it is marked"
+			" as bad" , block, zone);
+		goto again;
+	}
+
+	/* Test ECC*/
+	if (ret == -EBADMSG ||
+		(ftl->smallpagenand && sm_correct_sector(buffer, oob))) {
+
+		dbg("read of block %d at zone %d, failed due to ECC error",
+			block, zone);
+		goto again;
+	}
+
+	return 0;
+}
+
+/* Writes a sector to media */
+static int sm_write_sector(struct sm_ftl *ftl,
+			int zone, int block, int boffset,
+					u8 *buffer, struct sm_oob *oob)
+{
+	struct mtd_oob_ops ops;
+	struct mtd_info *mtd = ftl->trans->mtd;
+	int ret;
+
+	BUG_ON(ftl->readonly);
+
+	if (zone == 0 && (block == ftl->cis_block || block == 0)) {
+		dbg("attempted to write the CIS!");
+		return -EIO;
+	}
+
+	if (ftl->unstable)
+		return -EIO;
+
+	ops.mode = ftl->smallpagenand ? MTD_OOB_RAW : MTD_OOB_PLACE;
+	ops.len = SM_SECTOR_SIZE;
+	ops.datbuf = buffer;
+	ops.ooboffs = 0;
+	ops.ooblen = SM_OOB_SIZE;
+	ops.oobbuf = (void *)oob;
+
+	ret = mtd->write_oob(mtd, sm_mkoffset(ftl, zone, block, boffset), &ops);
+
+	/* Now we assume that hardware will catch write bitflip errors */
+	/* If you are paranoid, use CONFIG_MTD_NAND_VERIFY_WRITE */
+
+	if (ret) {
+		dbg("write to block %d at zone %d, failed with error %d",
+			block, zone, ret);
+
+		sm_recheck_media(ftl);
+		return ret;
+	}
+
+	/* This should never happen, unless there is a bug in the driver */
+	WARN_ON(ops.oobretlen != SM_OOB_SIZE);
+	WARN_ON(buffer && ops.retlen != SM_SECTOR_SIZE);
+
+	return 0;
+}
+
+/* ------------------------ block IO ------------------------------------- */
+
+/* Write a block using data and lba, and invalid sector bitmap */
+static int sm_write_block(struct sm_ftl *ftl, u8 *buf,
+				int zone, int block, int lba,
+						unsigned long invalid_bitmap)
+{
+	struct sm_oob oob;
+	int boffset;
+	int retry = 0;
+
+	/* Initialize the oob with requested values */
+	memset(&oob, 0xFF, SM_OOB_SIZE);
+	sm_write_lba(&oob, lba);
+restart:
+	if (ftl->unstable)
+		return -EIO;
+
+	for (boffset = 0; boffset < ftl->block_size;
+				boffset += SM_SECTOR_SIZE) {
+
+		oob.data_status = 0xFF;
+
+		if (test_bit(boffset / SM_SECTOR_SIZE, &invalid_bitmap)) {
+
+			sm_printk("sector %d of block at LBA %d of zone %d"
+				" coudn't be read, marking it as invalid",
+				boffset / SM_SECTOR_SIZE, lba, zone);
+
+			oob.data_status = 0;
+		}
+
+#ifdef CONFIG_SM_FTL_MUSEUM
+		if (ftl->smallpagenand) {
+			__nand_calculate_ecc(buf + boffset,
+						SM_SMALL_PAGE, oob.ecc1);
+
+			__nand_calculate_ecc(buf + boffset + SM_SMALL_PAGE,
+						SM_SMALL_PAGE, oob.ecc2);
+		}
+#endif
+		if (!sm_write_sector(ftl, zone, block, boffset,
+							buf + boffset, &oob))
+			continue;
+
+		if (!retry) {
+
+			/* If write fails. try to erase the block */
+			/* This is safe, because we never write in blocks
+				that contain valuable data.
+			This is intended to repair block that are marked
+			as erased, but that isn't fully erased*/
+
+			if (sm_erase_block(ftl, zone, block, 0))
+				return -EIO;
+
+			retry = 1;
+			goto restart;
+		} else {
+			sm_mark_block_bad(ftl, zone, block);
+			return -EIO;
+		}
+	}
+	return 0;
+}
+
+
+/* Mark whole block at offset 'offs' as bad. */
+static void sm_mark_block_bad(struct sm_ftl *ftl, int zone, int block)
+{
+	struct sm_oob oob;
+	int boffset;
+
+	memset(&oob, 0xFF, SM_OOB_SIZE);
+	oob.block_status = 0xF0;
+
+	if (ftl->unstable)
+		return;
+
+	if (sm_recheck_media(ftl))
+		return;
+
+	sm_printk("marking block %d of zone %d as bad", block, zone);
+
+	/* We aren't checking the return value, because we don't care */
+	/* This also fails on fake xD cards, but I guess these won't expose
+		any bad blocks till fail completly */
+	for (boffset = 0; boffset < ftl->block_size; boffset += SM_SECTOR_SIZE)
+		sm_write_sector(ftl, zone, block, boffset, NULL, &oob);
+}
+
+/*
+ * Erase a block within a zone
+ * If erase succedes, it updates free block fifo, otherwise marks block as bad
+ */
+static int sm_erase_block(struct sm_ftl *ftl, int zone_num, u16 block,
+								int put_free)
+{
+	struct ftl_zone *zone = &ftl->zones[zone_num];
+	struct mtd_info *mtd = ftl->trans->mtd;
+	struct erase_info erase;
+
+	erase.mtd = mtd;
+	erase.callback = sm_erase_callback;
+	erase.addr = sm_mkoffset(ftl, zone_num, block, 0);
+	erase.len = ftl->block_size;
+	erase.priv = (u_long)ftl;
+
+	if (ftl->unstable)
+		return -EIO;
+
+	BUG_ON(ftl->readonly);
+
+	if (zone_num == 0 && (block == ftl->cis_block || block == 0)) {
+		sm_printk("attempted to erase the CIS!");
+		return -EIO;
+	}
+
+	if (mtd->erase(mtd, &erase)) {
+		sm_printk("erase of block %d in zone %d failed",
+							block, zone_num);
+		goto error;
+	}
+
+	if (erase.state == MTD_ERASE_PENDING)
+		wait_for_completion(&ftl->erase_completion);
+
+	if (erase.state != MTD_ERASE_DONE) {
+		sm_printk("erase of block %d in zone %d failed after wait",
+			block, zone_num);
+		goto error;
+	}
+
+	if (put_free)
+		kfifo_in(&zone->free_sectors,
+			(const unsigned char *)&block, sizeof(block));
+
+	return 0;
+error:
+	sm_mark_block_bad(ftl, zone_num, block);
+	return -EIO;
+}
+
+static void sm_erase_callback(struct erase_info *self)
+{
+	struct sm_ftl *ftl = (struct sm_ftl *)self->priv;
+	complete(&ftl->erase_completion);
+}
+
+/* Throughtly test that block is valid. */
+static int sm_check_block(struct sm_ftl *ftl, int zone, int block)
+{
+	int boffset;
+	struct sm_oob oob;
+	int lbas[] = { -3, 0, 0, 0 };
+	int i = 0;
+	int test_lba;
+
+
+	/* First just check that block doesn't look fishy */
+	/* Only blocks that are valid or are sliced in two parts, are
+		accepted */
+	for (boffset = 0; boffset < ftl->block_size;
+					boffset += SM_SECTOR_SIZE) {
+
+		/* This shoudn't happen anyway */
+		if (sm_read_sector(ftl, zone, block, boffset, NULL, &oob))
+			return -2;
+
+		test_lba = sm_read_lba(&oob);
+
+		if (lbas[i] != test_lba)
+			lbas[++i] = test_lba;
+
+		/* If we found three different LBAs, something is fishy */
+		if (i == 3)
+			return -EIO;
+	}
+
+	/* If the block is sliced (partialy erased usually) erase it */
+	if (i == 2) {
+		sm_erase_block(ftl, zone, block, 1);
+		return 1;
+	}
+
+	return 0;
+}
+
+/* ----------------- media scanning --------------------------------- */
+static const struct chs_entry chs_table[] = {
+	{ 1,    125,  4,  4  },
+	{ 2,    125,  4,  8  },
+	{ 4,    250,  4,  8  },
+	{ 8,    250,  4,  16 },
+	{ 16,   500,  4,  16 },
+	{ 32,   500,  8,  16 },
+	{ 64,   500,  8,  32 },
+	{ 128,  500,  16, 32 },
+	{ 256,  1000, 16, 32 },
+	{ 512,  1015, 32, 63 },
+	{ 1024, 985,  33, 63 },
+	{ 2048, 985,  33, 63 },
+	{ 0 },
+};
+
+
+static const u8 cis_signature[] = {
+	0x01, 0x03, 0xD9, 0x01, 0xFF, 0x18, 0x02, 0xDF, 0x01, 0x20
+};
+/* Find out media parameters.
+ * This ideally has to be based on nand id, but for now device size is enough */
+int sm_get_media_info(struct sm_ftl *ftl, struct mtd_info *mtd)
+{
+	int i;
+	int size_in_megs = mtd->size / (1024 * 1024);
+
+	ftl->readonly = mtd->type == MTD_ROM;
+
+	/* Manual settings for very old devices */
+	ftl->zone_count = 1;
+	ftl->smallpagenand = 0;
+
+	switch (size_in_megs) {
+	case 1:
+		/* 1 Mb flash/rom SmartMedia card (256 byte pages)*/
+		ftl->zone_size = 256;
+		ftl->max_lba = 250;
+		ftl->block_size = 8 * SM_SECTOR_SIZE;
+		ftl->smallpagenand = 1;
+
+		break;
+	case 2:
+		/* 2 Mb flash SmartMedia (256 byte pages)*/
+		if (mtd->writesize == SM_SMALL_PAGE) {
+			ftl->zone_size = 512;
+			ftl->max_lba = 500;
+			ftl->block_size = 8 * SM_SECTOR_SIZE;
+			ftl->smallpagenand = 1;
+		/* 2 Mb rom SmartMedia */
+		} else {
+
+			if (!ftl->readonly)
+				return -ENODEV;
+
+			ftl->zone_size = 256;
+			ftl->max_lba = 250;
+			ftl->block_size = 16 * SM_SECTOR_SIZE;
+		}
+		break;
+	case 4:
+		/* 4 Mb flash/rom SmartMedia device */
+		ftl->zone_size = 512;
+		ftl->max_lba = 500;
+		ftl->block_size = 16 * SM_SECTOR_SIZE;
+		break;
+	case 8:
+		/* 8 Mb flash/rom SmartMedia device */
+		ftl->zone_size = 1024;
+		ftl->max_lba = 1000;
+		ftl->block_size = 16 * SM_SECTOR_SIZE;
+	}
+
+	/* Minimum xD size is 16M. Also, all xD cards have standard zone
+	   sizes. SmartMedia cards exist up to 128 Mb and have same layout*/
+	if (size_in_megs >= 16) {
+		ftl->zone_count = size_in_megs / 16;
+		ftl->zone_size = 1024;
+		ftl->max_lba = 1000;
+		ftl->block_size = 32 * SM_SECTOR_SIZE;
+	}
+
+	/* Test for proper write,erase and oob sizes */
+	if (mtd->erasesize > ftl->block_size)
+		return -ENODEV;
+
+	if (mtd->writesize > SM_SECTOR_SIZE)
+		return -ENODEV;
+
+	if (ftl->smallpagenand && mtd->oobsize < SM_SMALL_OOB_SIZE)
+		return -ENODEV;
+
+	if (!ftl->smallpagenand && mtd->oobsize < SM_OOB_SIZE)
+		return -ENODEV;
+
+	/* We use these functions for IO */
+	if (!mtd->read_oob || !mtd->write_oob)
+		return -ENODEV;
+
+	/* Find geometry information */
+	for (i = 0 ; i < ARRAY_SIZE(chs_table) ; i++) {
+		if (chs_table[i].size == size_in_megs) {
+			ftl->cylinders = chs_table[i].cyl;
+			ftl->heads = chs_table[i].head;
+			ftl->sectors = chs_table[i].sec;
+			return 0;
+		}
+	}
+
+	sm_printk("media has unknown size : %dMiB", size_in_megs);
+	ftl->cylinders = 985;
+	ftl->heads =  33;
+	ftl->sectors = 63;
+	return 0;
+}
+
+/* Validate the CIS */
+static int sm_read_cis(struct sm_ftl *ftl)
+{
+	struct sm_oob oob;
+
+	if (sm_read_sector(ftl,
+		0, ftl->cis_block, ftl->cis_boffset, ftl->cis_buffer, &oob))
+			return -EIO;
+
+	if (!sm_sector_valid(&oob) || !sm_block_valid(&oob))
+		return -EIO;
+
+	if (!memcmp(ftl->cis_buffer + ftl->cis_page_offset,
+			cis_signature, sizeof(cis_signature))) {
+		return 0;
+	}
+
+	return -EIO;
+}
+
+/* Scan the media for the CIS */
+static int sm_find_cis(struct sm_ftl *ftl)
+{
+	struct sm_oob oob;
+	int block, boffset;
+	int block_found = 0;
+	int cis_found = 0;
+
+	/* Search for first valid block */
+	for (block = 0 ; block < ftl->zone_size - ftl->max_lba ; block++) {
+
+		if (sm_read_sector(ftl, 0, block, 0, NULL, &oob))
+			continue;
+
+		if (!sm_block_valid(&oob))
+			continue;
+		block_found = 1;
+		break;
+	}
+
+	if (!block_found)
+		return -EIO;
+
+	/* Search for first valid sector in this block */
+	for (boffset = 0 ; boffset < ftl->block_size;
+						boffset += SM_SECTOR_SIZE) {
+
+		if (sm_read_sector(ftl, 0, block, boffset, NULL, &oob))
+			continue;
+
+		if (!sm_sector_valid(&oob))
+			continue;
+		break;
+	}
+
+	if (boffset == ftl->block_size)
+		return -EIO;
+
+	ftl->cis_block = block;
+	ftl->cis_boffset = boffset;
+	ftl->cis_page_offset = 0;
+
+	cis_found = !sm_read_cis(ftl);
+
+	if (!cis_found) {
+		ftl->cis_page_offset = SM_SMALL_PAGE;
+		cis_found = !sm_read_cis(ftl);
+	}
+
+	if (cis_found) {
+		dbg("CIS block found at offset %x",
+			block * ftl->block_size +
+				boffset + ftl->cis_page_offset);
+		return 0;
+	}
+	return -EIO;
+}
+
+/* Basic test to determine if underlying mtd device if functional */
+static int sm_recheck_media(struct sm_ftl *ftl)
+{
+	if (sm_read_cis(ftl)) {
+
+		if (!ftl->unstable) {
+			sm_printk("media unstable, not allowing writes");
+			ftl->unstable = 1;
+		}
+		return -EIO;
+	}
+	return 0;
+}
+
+/* Initialize a FTL zone */
+static int sm_init_zone(struct sm_ftl *ftl, int zone_num)
+{
+	struct ftl_zone *zone = &ftl->zones[zone_num];
+	struct sm_oob oob;
+	u16 block;
+	int lba;
+	int i = 0;
+
+	dbg("initializing zone %d", zone_num);
+
+	/* Allocate memory for FTL table */
+	zone->lba_to_phys_table = kmalloc(ftl->max_lba * 2, GFP_KERNEL);
+
+	if (!zone->lba_to_phys_table)
+		return -ENOMEM;
+	memset(zone->lba_to_phys_table, -1, ftl->max_lba * 2);
+
+
+	/* Allocate memory for free sectors FIFO */
+	if (kfifo_alloc(&zone->free_sectors, ftl->zone_size * 2, GFP_KERNEL)) {
+		kfree(zone->lba_to_phys_table);
+		return -ENOMEM;
+	}
+
+	/* Now scan the zone */
+	for (block = 0 ; block < ftl->zone_size ; block++) {
+
+		/* Skip blocks till the CIS (including) */
+		if (zone_num == 0 && block <= ftl->cis_block)
+			continue;
+
+		/* Read the oob of first sector */
+		if (sm_read_sector(ftl, zone_num, block, 0, NULL, &oob))
+			return -EIO;
+
+		/* Test to see if block is erased. It is enough to test
+			first sector, because erase happens in one shot */
+		if (sm_block_erased(&oob)) {
+			kfifo_in(&zone->free_sectors,
+				(unsigned char *)&block, 2);
+			continue;
+		}
+
+		/* If block is marked as bad, skip it */
+		/* This assumes we can trust first sector*/
+		/* However the way the block valid status is defined, ensures
+			very low probability of failure here */
+		if (!sm_block_valid(&oob)) {
+			dbg("PH %04d <-> <marked bad>", block);
+			continue;
+		}
+
+
+		lba = sm_read_lba(&oob);
+
+		/* Invalid LBA means that block is damaged. */
+		/* We can try to erase it, or mark it as bad, but
+			lets leave that to recovery application */
+		if (lba == -2 || lba >= ftl->max_lba) {
+			dbg("PH %04d <-> LBA %04d(bad)", block, lba);
+			continue;
+		}
+
+
+		/* If there is no collision,
+			just put the sector in the FTL table */
+		if (zone->lba_to_phys_table[lba] < 0) {
+			dbg_verbose("PH %04d <-> LBA %04d", block, lba);
+			zone->lba_to_phys_table[lba] = block;
+			continue;
+		}
+
+		sm_printk("collision"
+			" of LBA %d between blocks %d and %d in zone %d",
+			lba, zone->lba_to_phys_table[lba], block, zone_num);
+
+		/* Test that this block is valid*/
+		if (sm_check_block(ftl, zone_num, block))
+			continue;
+
+		/* Test now the old block */
+		if (sm_check_block(ftl, zone_num,
+					zone->lba_to_phys_table[lba])) {
+			zone->lba_to_phys_table[lba] = block;
+			continue;
+		}
+
+		/* If both blocks are valid and share same LBA, it means that
+			they hold different versions of same data. It not
+			known which is more recent, thus just erase one of them
+		*/
+		sm_printk("both blocks are valid, erasing the later");
+		sm_erase_block(ftl, zone_num, block, 1);
+	}
+
+	dbg("zone initialized");
+	zone->initialized = 1;
+
+	/* No free sectors, means that the zone is heavily damaged, write won't
+		work, but it can still can be (partially) read */
+	if (!kfifo_len(&zone->free_sectors)) {
+		sm_printk("no free blocks in zone %d", zone_num);
+		return 0;
+	}
+
+	/* Randomize first block we write to */
+	get_random_bytes(&i, 2);
+	i %= (kfifo_len(&zone->free_sectors) / 2);
+
+	while (i--) {
+		kfifo_out(&zone->free_sectors, (unsigned char *)&block, 2);
+		kfifo_in(&zone->free_sectors, (const unsigned char *)&block, 2);
+	}
+	return 0;
+}
+
+/* Get and automaticly initialize an FTL mapping for one zone */
+struct ftl_zone *sm_get_zone(struct sm_ftl *ftl, int zone_num)
+{
+	struct ftl_zone *zone;
+	int error;
+
+	BUG_ON(zone_num >= ftl->zone_count);
+	zone = &ftl->zones[zone_num];
+
+	if (!zone->initialized) {
+		error = sm_init_zone(ftl, zone_num);
+
+		if (error)
+			return ERR_PTR(error);
+	}
+	return zone;
+}
+
+
+/* ----------------- cache handling ------------------------------------------*/
+
+/* Initialize the one block cache */
+void sm_cache_init(struct sm_ftl *ftl)
+{
+	ftl->cache_data_invalid_bitmap = 0xFFFFFFFF;
+	ftl->cache_clean = 1;
+	ftl->cache_zone = -1;
+	ftl->cache_block = -1;
+	/*memset(ftl->cache_data, 0xAA, ftl->block_size);*/
+}
+
+/* Put sector in one block cache */
+void sm_cache_put(struct sm_ftl *ftl, char *buffer, int boffset)
+{
+	memcpy(ftl->cache_data + boffset, buffer, SM_SECTOR_SIZE);
+	clear_bit(boffset / SM_SECTOR_SIZE, &ftl->cache_data_invalid_bitmap);
+	ftl->cache_clean = 0;
+}
+
+/* Read a sector from the cache */
+int sm_cache_get(struct sm_ftl *ftl, char *buffer, int boffset)
+{
+	if (test_bit(boffset / SM_SECTOR_SIZE,
+		&ftl->cache_data_invalid_bitmap))
+			return -1;
+
+	memcpy(buffer, ftl->cache_data + boffset, SM_SECTOR_SIZE);
+	return 0;
+}
+
+/* Write the cache to hardware */
+int sm_cache_flush(struct sm_ftl *ftl)
+{
+	struct ftl_zone *zone;
+
+	int sector_num;
+	u16 write_sector;
+	int zone_num = ftl->cache_zone;
+	int block_num;
+
+	if (ftl->cache_clean)
+		return 0;
+
+	if (ftl->unstable)
+		return -EIO;
+
+	BUG_ON(zone_num < 0);
+	zone = &ftl->zones[zone_num];
+	block_num = zone->lba_to_phys_table[ftl->cache_block];
+
+
+	/* Try to read all unread areas of the cache block*/
+	for_each_bit(sector_num, &ftl->cache_data_invalid_bitmap,
+		ftl->block_size / SM_SECTOR_SIZE) {
+
+		if (!sm_read_sector(ftl,
+			zone_num, block_num, sector_num * SM_SECTOR_SIZE,
+			ftl->cache_data + sector_num * SM_SECTOR_SIZE, NULL))
+				clear_bit(sector_num,
+					&ftl->cache_data_invalid_bitmap);
+	}
+restart:
+
+	if (ftl->unstable)
+		return -EIO;
+	/* No spare blocks */
+	/* We could still continue by erasing the current block,
+		but for such worn out media it doesn't worth the trouble,
+			and the dangers */
+
+	if (!kfifo_len(&zone->free_sectors)) {
+		dbg("no free sectors for write!");
+		return -EIO;
+	}
+
+	kfifo_out(&zone->free_sectors, (unsigned char *)&write_sector, 2);
+
+	if (sm_write_block(ftl, ftl->cache_data, zone_num, write_sector,
+		ftl->cache_block, ftl->cache_data_invalid_bitmap))
+			goto restart;
+
+	/* Update the FTL table */
+	zone->lba_to_phys_table[ftl->cache_block] = write_sector;
+
+	/* Write succesfull, so erase and free the old block */
+	if (block_num > 0)
+		sm_erase_block(ftl, zone_num, block_num, 1);
+
+	sm_cache_init(ftl);
+	return 0;
+}
+
+
+/* flush timer, runs a second after last write */
+static void sm_cache_flush_timer(unsigned long data)
+{
+	struct sm_ftl *ftl = (struct sm_ftl *)data;
+	queue_work(cache_flush_workqueue, &ftl->flush_work);
+}
+
+/* cache flush work, kicked by timer */
+static void sm_cache_flush_work(struct work_struct *work)
+{
+	struct sm_ftl *ftl = container_of(work, struct sm_ftl, flush_work);
+	mutex_lock(&ftl->mutex);
+	sm_cache_flush(ftl);
+	mutex_unlock(&ftl->mutex);
+	return;
+}
+
+/* ---------------- outside interface -------------------------------------- */
+
+/* outside interface: read a sector */
+static int sm_read(struct mtd_blktrans_dev *dev,
+				unsigned long sect_no, char *buf)
+{
+	struct sm_ftl *ftl = dev->priv;
+	struct ftl_zone *zone;
+	int error = 0, in_cache = 0;
+	int zone_num, block, boffset;
+
+	sm_break_offset(ftl, sect_no << 9, &zone_num, &block, &boffset);
+	mutex_lock(&ftl->mutex);
+
+
+	zone = sm_get_zone(ftl, zone_num);
+	if (IS_ERR(zone)) {
+		error = PTR_ERR(zone);
+		goto unlock;
+	}
+
+	/* Have to look at cache first */
+	if (ftl->cache_zone == zone_num && ftl->cache_block == block) {
+		in_cache = 1;
+		if (!sm_cache_get(ftl, buf, boffset))
+			goto unlock;
+	}
+
+	/* Translate the block and return if doesn't exist in the table */
+	block = zone->lba_to_phys_table[block];
+
+	if (block == -1) {
+		memset(buf, 0xFF, SM_SECTOR_SIZE);
+		goto unlock;
+	}
+
+	if (sm_read_sector(ftl, zone_num, block, boffset, buf, NULL)) {
+		error = -EIO;
+		goto unlock;
+	}
+
+	if (in_cache)
+		sm_cache_put(ftl, buf, boffset);
+unlock:
+	mutex_unlock(&ftl->mutex);
+	return error;
+}
+
+/* outside interface: write a sector */
+static int sm_write(struct mtd_blktrans_dev *dev,
+				unsigned long sec_no, char *buf)
+{
+	struct sm_ftl *ftl = dev->priv;
+	struct ftl_zone *zone;
+	int error, zone_num, block, boffset;
+
+	BUG_ON(ftl->readonly);
+	sm_break_offset(ftl, sec_no << 9, &zone_num, &block, &boffset);
+
+	/* No need in flush thread running now */
+	del_timer(&ftl->timer);
+	mutex_lock(&ftl->mutex);
+
+	zone = sm_get_zone(ftl, zone_num);
+	if (IS_ERR(zone)) {
+		error = PTR_ERR(zone);
+		goto unlock;
+	}
+
+	/* If entry is not in cache, flush it */
+	if (ftl->cache_block != block || ftl->cache_zone != zone_num) {
+
+		error = sm_cache_flush(ftl);
+		if (error)
+			goto unlock;
+
+		ftl->cache_block = block;
+		ftl->cache_zone = zone_num;
+	}
+
+	sm_cache_put(ftl, buf, boffset);
+unlock:
+	mod_timer(&ftl->timer, jiffies + msecs_to_jiffies(cache_timeout));
+	mutex_unlock(&ftl->mutex);
+	return error;
+}
+
+/* outside interface: flush everything */
+static int sm_flush(struct mtd_blktrans_dev *dev)
+{
+	struct sm_ftl *ftl = dev->priv;
+	int retval;
+
+	mutex_lock(&ftl->mutex);
+	retval =  sm_cache_flush(ftl);
+	mutex_unlock(&ftl->mutex);
+	return retval;
+}
+
+/* outside interface: device is released */
+static int sm_release(struct mtd_blktrans_dev *dev)
+{
+	struct sm_ftl *ftl = dev->priv;
+
+	mutex_lock(&ftl->mutex);
+	del_timer_sync(&ftl->timer);
+	cancel_work_sync(&ftl->flush_work);
+	sm_cache_flush(ftl);
+	mutex_unlock(&ftl->mutex);
+	return 0;
+}
+
+/* outside interface: get geometry */
+static int sm_getgeo(struct mtd_blktrans_dev *dev, struct hd_geometry *geo)
+{
+	struct sm_ftl *ftl = dev->priv;
+	geo->heads = ftl->heads;
+	geo->sectors = ftl->sectors;
+	geo->cylinders = ftl->cylinders;
+	return 0;
+}
+
+/* external interface: main initialization function */
+static void sm_add_mtd(struct mtd_blktrans_ops *tr, struct mtd_info *mtd)
+{
+	struct mtd_blktrans_dev *trans;
+	struct sm_ftl *ftl;
+
+	/* Allocate & initialize our private structure */
+	ftl = kzalloc(sizeof(struct sm_ftl), GFP_KERNEL);
+	if (!ftl)
+		goto error1;
+
+
+	mutex_init(&ftl->mutex);
+	setup_timer(&ftl->timer, sm_cache_flush_timer, (unsigned long)ftl);
+	INIT_WORK(&ftl->flush_work, sm_cache_flush_work);
+	init_completion(&ftl->erase_completion);
+
+	/* Read media information */
+	if (sm_get_media_info(ftl, mtd)) {
+		dbg("found unsupported mtd device, aborting");
+		goto error2;
+	}
+
+
+	/* Allocate temporary CIS buffer for read retry support */
+	ftl->cis_buffer = kzalloc(SM_SECTOR_SIZE, GFP_KERNEL);
+	if (!ftl->cis_buffer)
+		goto error2;
+
+	/* Allocate zone array, it will be initialized on demand */
+	ftl->zones = kzalloc(sizeof(struct ftl_zone) * ftl->zone_count,
+								GFP_KERNEL);
+	if (!ftl->zones)
+		goto error3;
+
+	/* Allocate the cache*/
+	ftl->cache_data = kzalloc(ftl->block_size, GFP_KERNEL);
+
+	if (!ftl->cache_data)
+		goto error4;
+
+	sm_cache_init(ftl);
+
+
+	/* Allocate upper layer structure and initialize it */
+	trans = kzalloc(sizeof(struct mtd_blktrans_dev), GFP_KERNEL);
+	if (!trans)
+		goto error5;
+
+	ftl->trans = trans;
+	trans->priv = ftl;
+
+	trans->tr = tr;
+	trans->mtd = mtd;
+	trans->devnum = -1;
+	trans->size = (ftl->block_size * ftl->max_lba * ftl->zone_count) >> 9;
+	trans->readonly = ftl->readonly;
+
+	if (sm_find_cis(ftl)) {
+		dbg("CIS not found on mtd device, aborting");
+		goto error6;
+	}
+
+	ftl->disk_attributes = sm_create_sysfs_attributes(ftl);
+	trans->disk_attributes = ftl->disk_attributes;
+
+	sm_printk("Found %d MiB xD/SmartMedia FTL on mtd%d",
+		(int)(mtd->size / (1024 * 1024)), mtd->index);
+
+	dbg("FTL layout:");
+	dbg("%d zone(s), each consists of %d blocks (+%d spares)",
+		ftl->zone_count, ftl->max_lba,
+		ftl->zone_size - ftl->max_lba);
+	dbg("each block consists of %d bytes",
+		ftl->block_size);
+
+
+	/* Register device*/
+	if (add_mtd_blktrans_dev(trans)) {
+		dbg("error in mtdblktrans layer");
+		goto error6;
+	}
+	return;
+error6:
+	kfree(trans);
+error5:
+	kfree(ftl->cache_data);
+error4:
+	kfree(ftl->zones);
+error3:
+	kfree(ftl->cis_buffer);
+error2:
+	kfree(ftl);
+error1:
+	return;
+}
+
+/* main interface: device {surprise,} removal */
+static void sm_remove_dev(struct mtd_blktrans_dev *dev)
+{
+	struct sm_ftl *ftl = dev->priv;
+	int i;
+
+	del_mtd_blktrans_dev(dev);
+	ftl->trans = NULL;
+
+	for (i = 0 ; i < ftl->zone_count; i++) {
+
+		if (!ftl->zones[i].initialized)
+			continue;
+
+		kfree(ftl->zones[i].lba_to_phys_table);
+		kfifo_free(&ftl->zones[i].free_sectors);
+	}
+
+	sm_delete_sysfs_attributes(ftl);
+	kfree(ftl->cis_buffer);
+	kfree(ftl->zones);
+	kfree(ftl->cache_data);
+	kfree(ftl);
+}
+
+static struct mtd_blktrans_ops sm_ftl_ops = {
+	.name		= "smblk",
+	.major		= -1,
+	.part_bits	= SM_FTL_PARTN_BITS,
+	.blksize	= SM_SECTOR_SIZE,
+	.getgeo		= sm_getgeo,
+
+	.add_mtd	= sm_add_mtd,
+	.remove_dev	= sm_remove_dev,
+
+	.readsect	= sm_read,
+	.writesect	= sm_write,
+
+	.flush		= sm_flush,
+	.release	= sm_release,
+
+	.owner		= THIS_MODULE,
+};
+
+static __init int sm_module_init(void)
+{
+	int error = 0;
+	cache_flush_workqueue = create_freezeable_workqueue("smflush");
+
+	if (IS_ERR(cache_flush_workqueue))
+		return PTR_ERR(cache_flush_workqueue);
+
+	error = register_mtd_blktrans(&sm_ftl_ops);
+	if (error)
+		destroy_workqueue(cache_flush_workqueue);
+	return error;
+
+}
+
+static void __exit sm_module_exit(void)
+{
+	destroy_workqueue(cache_flush_workqueue);
+	deregister_mtd_blktrans(&sm_ftl_ops);
+}
+
+module_init(sm_module_init);
+module_exit(sm_module_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Maxim Levitsky <maximlevitsky@gmail.com>");
+MODULE_DESCRIPTION("Smartmedia/xD mtd translation layer");
diff --git a/drivers/mtd/sm_ftl.h b/drivers/mtd/sm_ftl.h
new file mode 100644
index 0000000..519e2e1
--- /dev/null
+++ b/drivers/mtd/sm_ftl.h
@@ -0,0 +1,94 @@
+/*
+ * Copyright (C) 2009 - Maxim Levitsky
+ * SmartMedia/xD translation layer
+ *
+ * Based loosly on ssfdc.c which is
+ *  (c) 2005 Eptar srl
+ *  Author: Claudio Lanconelli <lanconelli.claudio@eptar.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/mtd/blktrans.h>
+#include <linux/kfifo.h>
+#include <linux/sched.h>
+#include <linux/completion.h>
+#include <linux/mtd/mtd.h>
+
+
+
+struct ftl_zone {
+	int initialized;
+	s16 *lba_to_phys_table;		/* LBA to physical table */
+	struct kfifo free_sectors;	/* queue of free sectors */
+};
+
+struct sm_ftl {
+	struct mtd_blktrans_dev *trans;
+
+	struct mutex mutex;		/* protects the structure */
+	struct ftl_zone *zones;		/* FTL tables for each zone */
+
+	/* Media information */
+	int block_size;			/* block size in bytes */
+	int zone_size;			/* zone size in blocks */
+	int zone_count;			/* number of zones */
+	int max_lba;			/* maximum lba in a zone */
+	int smallpagenand;		/* 256 bytes/page nand */
+	int readonly;			/* is FS readonly */
+	int unstable;
+	int cis_block;			/* CIS block location */
+	int cis_boffset;		/* CIS offset in the block */
+	int cis_page_offset;		/* CIS offset in the page */
+	void *cis_buffer;		/* tmp buffer for cis reads */
+
+	/* Cache */
+	int cache_block;		/* block number of cached block */
+	int cache_zone;			/* zone of cached block */
+	unsigned char *cache_data;	/* cached block data */
+	long unsigned int cache_data_invalid_bitmap;
+	int cache_clean;
+	struct work_struct flush_work;
+	struct timer_list timer;
+
+	/* Async erase stuff */
+	struct completion erase_completion;
+
+	/* Geometry stuff */
+	int heads;
+	int sectors;
+	int cylinders;
+
+	struct attribute_group *disk_attributes;
+};
+
+struct chs_entry {
+	unsigned long size;
+	unsigned short cyl;
+	unsigned char head;
+	unsigned char sec;
+};
+
+
+#define SM_FTL_PARTN_BITS	3
+
+#define sm_printk(format, ...) \
+	printk(KERN_WARNING "sm_ftl" ": " format "\n", ## __VA_ARGS__)
+
+#define dbg(format, ...) \
+	if (debug) \
+		printk(KERN_DEBUG "sm_ftl" ": " format "\n", ## __VA_ARGS__)
+
+#define dbg_verbose(format, ...) \
+	if (debug > 1) \
+		printk(KERN_DEBUG "sm_ftl" ": " format "\n", ## __VA_ARGS__)
+
+
+static void sm_erase_callback(struct erase_info *self);
+static int sm_erase_block(struct sm_ftl *ftl, int zone_num, u16 block,
+								int put_free);
+static void sm_mark_block_bad(struct sm_ftl *ftl, int zone_num, int block);
+
+static int sm_recheck_media(struct sm_ftl *ftl);
-- 
1.6.3.3

[PATCH 15/15] MTD: Add nand driver for Ricoh xD/SmartMedia reader

[Maxim Levitsky]

This adds a driver for Ricoh R5C852 xD card reader.

This reader is a part of larger mulifunction chip
and found at least in R5C832

Driver is complete, but bewere of the fact that some
(probably only type M) xD cards are 'fake' which means that
they have an on board CPU and expose emulated nand command set

These cards don't even store the  oob area on the flash,
but generate it on the fly from something else.

Thus they demand to have proper values written in the oob area,
and therefore only useful with SmartMedia FTL.

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 MAINTAINERS               |    6 +
 drivers/mtd/nand/Kconfig  |   11 +
 drivers/mtd/nand/Makefile |    1 +
 drivers/mtd/nand/r852.c   | 1117 +++++++++++++++++++++++++++++++++++++++++++++
 drivers/mtd/nand/r852.h   |  163 +++++++
 include/linux/pci_ids.h   |    2 +
 6 files changed, 1300 insertions(+), 0 deletions(-)
 create mode 100644 drivers/mtd/nand/r852.c
 create mode 100644 drivers/mtd/nand/r852.h

diff --git a/MAINTAINERS b/MAINTAINERS
index 1b0f053..1723c54 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -4577,6 +4577,12 @@ S:	Maintained
 F:	Documentation/rfkill.txt
 F:	net/rfkill/
 
+RICOH SMARTMEDIA/XD DRIVER
+M:	Maxim Levitsky <maximlevitsky@gmail.com>
+S:	Maintained
+F:	drivers/mtd/nand/r822.c
+F:	drivers/mtd/nand/r822.h
+
 RISCOM8 DRIVER
 S:	Orphan
 F:	Documentation/serial/riscom8.txt
diff --git a/drivers/mtd/nand/Kconfig b/drivers/mtd/nand/Kconfig
index 31273c7..4ab2551 100644
--- a/drivers/mtd/nand/Kconfig
+++ b/drivers/mtd/nand/Kconfig
@@ -107,6 +107,17 @@ config MTD_NAND_OMAP_PREFETCH_DMA
 config MTD_NAND_IDS
 	tristate
 
+config MTD_NAND_RICOH
+	tristate "Ricoh xD card reader"
+	default n
+	select MTD_SM_COMMON
+	help
+	  Enable support for Ricoh R5C852 xD card reader
+	  You also need to enable ether
+	  NAND SSFDC (SmartMedia) read only translation layer' or new
+	  expermental, readwrite
+	  'SmartMedia/xD new translation layer'
+
 config MTD_NAND_AU1550
 	tristate "Au1550/1200 NAND support"
 	depends on SOC_AU1200 || SOC_AU1550
diff --git a/drivers/mtd/nand/Makefile b/drivers/mtd/nand/Makefile
index e699bef..1d5797d 100644
--- a/drivers/mtd/nand/Makefile
+++ b/drivers/mtd/nand/Makefile
@@ -42,5 +42,6 @@ obj-$(CONFIG_MTD_NAND_TXX9NDFMC)	+= txx9ndfmc.o
 obj-$(CONFIG_MTD_NAND_NUC900)		+= nuc900_nand.o
 obj-$(CONFIG_MTD_NAND_NOMADIK)		+= nomadik_nand.o
 obj-$(CONFIG_MTD_NAND_BCM_UMI)		+= bcm_umi_nand.o nand_bcm_umi.o
+obj-$(CONFIG_MTD_NAND_RICOH)		+= r852.o
 
 nand-objs := nand_base.o nand_bbt.o
diff --git a/drivers/mtd/nand/r852.c b/drivers/mtd/nand/r852.c
new file mode 100644
index 0000000..5c718ff
--- /dev/null
+++ b/drivers/mtd/nand/r852.c
@@ -0,0 +1,1117 @@
+/*
+ * Copyright (C) 2009 - Maxim Levitsky
+ * driver for Ricoh xD readers
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/jiffies.h>
+#include <linux/workqueue.h>
+#include <linux/interrupt.h>
+#include <linux/pci_ids.h>
+#include <asm/byteorder.h>
+#include <linux/sched.h>
+#include "sm_common.h"
+#include "r852.h"
+
+
+static int enable_dma = 1;
+module_param(enable_dma, bool, S_IRUGO);
+MODULE_PARM_DESC(enable_dma, "Enable usage of the DMA (default)");
+
+static int debug;
+module_param(debug, int, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(debug, "Debug level (0-2)");
+
+/* read register */
+static inline u8 r852_read_reg(struct r852_device *dev, int address)
+{
+	u8 reg = readb(dev->mmio + address);
+	return reg;
+}
+
+/* write register */
+static inline void r852_write_reg(struct r852_device *dev,
+						int address, u8 value)
+{
+	writeb(value, dev->mmio + address);
+	mmiowb();
+}
+
+
+/* read dword sized register */
+static inline u32 r852_read_reg_dword(struct r852_device *dev, int address)
+{
+	u32 reg = le32_to_cpu(readl(dev->mmio + address));
+	return reg;
+}
+
+/* write dword sized register */
+static inline void r852_write_reg_dword(struct r852_device *dev,
+							int address, u32 value)
+{
+	writel(cpu_to_le32(value), dev->mmio + address);
+	mmiowb();
+}
+
+/* returns pointer to our private structure */
+static inline struct r852_device *r852_get_dev(struct mtd_info *mtd)
+{
+	struct nand_chip *chip = (struct nand_chip *)mtd->priv;
+	return (struct r852_device *)chip->priv;
+}
+
+
+/* check if controller supports dma */
+static void r852_dma_test(struct r852_device *dev)
+{
+	dev->dma_usable = (r852_read_reg(dev, R852_DMA_CAP) &
+		(R852_DMA1 | R852_DMA2)) == (R852_DMA1 | R852_DMA2);
+
+	if (!dev->dma_usable)
+		message("Non dma capable device detected, dma disabled");
+
+	if (!enable_dma) {
+		message("disabling dma on user request");
+		dev->dma_usable = 0;
+	}
+}
+
+/*
+ * Enable dma. Enables ether first or second stage of the DMA,
+ * Expects dev->dma_dir and dev->dma_state be set
+ */
+static void r852_dma_enable(struct r852_device *dev)
+{
+	u8 dma_reg, dma_irq_reg;
+
+	/* Set up dma settings */
+	dma_reg = r852_read_reg_dword(dev, R852_DMA_SETTINGS);
+	dma_reg &= ~(R852_DMA_READ | R852_DMA_INTERNAL | R852_DMA_MEMORY);
+
+	if (dev->dma_dir)
+		dma_reg |= R852_DMA_READ;
+
+	if (dev->dma_state == DMA_INTERNAL)
+		dma_reg |= R852_DMA_INTERNAL;
+	else {
+		dma_reg |= R852_DMA_MEMORY;
+		r852_write_reg_dword(dev, R852_DMA_ADDR,
+			cpu_to_le32(dev->phys_dma_addr));
+	}
+
+	r852_write_reg_dword(dev, R852_DMA_SETTINGS, dma_reg);
+
+	/* Set dma irq */
+	dma_irq_reg = r852_read_reg_dword(dev, R852_DMA_IRQ_ENABLE);
+	r852_write_reg_dword(dev, R852_DMA_IRQ_ENABLE,
+		dma_irq_reg |
+		R852_DMA_IRQ_INTERNAL |
+		R852_DMA_IRQ_ERROR |
+		R852_DMA_IRQ_MEMORY);
+}
+
+/*
+ * Disable dma, called from the interrupt handler, which specifies
+ * success of the operation via 'error' argument
+ */
+static void r852_dma_done(struct r852_device *dev, int error)
+{
+	WARN_ON(dev->dma_stage == 0);
+
+	r852_write_reg_dword(dev, R852_DMA_IRQ_STA,
+			r852_read_reg_dword(dev, R852_DMA_IRQ_STA));
+
+	r852_write_reg_dword(dev, R852_DMA_SETTINGS, 0);
+	r852_write_reg_dword(dev, R852_DMA_IRQ_ENABLE, 0);
+
+	dev->dma_error = error;
+	dev->dma_stage = 0;
+
+	if (dev->phys_dma_addr && dev->phys_dma_addr != dev->phys_bounce_buffer)
+		pci_unmap_single(dev->pci_dev, dev->phys_dma_addr, R852_DMA_LEN,
+			dev->dma_dir ? PCI_DMA_FROMDEVICE : PCI_DMA_TODEVICE);
+	complete(&dev->dma_done);
+}
+
+/*
+ * Wait, till dma is done, which includes both phases of it
+ */
+static int r852_dma_wait(struct r852_device *dev)
+{
+	long timeout = wait_for_completion_timeout(&dev->dma_done,
+				msecs_to_jiffies(1000));
+	if (!timeout) {
+		dbg("timeout waiting for DMA interrupt");
+		return -ETIMEDOUT;
+	}
+
+	return 0;
+}
+
+/*
+ * Read/Write one page using dma. Only pages can be read (512 bytes)
+*/
+static void r852_do_dma(struct r852_device *dev, uint8_t *buf, int do_read)
+{
+	int bounce = 0;
+	unsigned long flags;
+	int error;
+
+	dev->dma_error = 0;
+
+	/* Set dma direction */
+	dev->dma_dir = do_read;
+	dev->dma_stage = 1;
+
+	dbg_verbose("doing dma %s ", do_read ? "read" : "write");
+
+	/* Set intial dma state: for reading first fill on board buffer,
+	  from device, for writes first fill the buffer  from memory*/
+	dev->dma_state = do_read ? DMA_INTERNAL : DMA_MEMORY;
+
+	/* if incoming buffer is not page aligned, we should do bounce */
+	if ((unsigned long)buf & (R852_DMA_LEN-1))
+		bounce = 1;
+
+	if (!bounce) {
+		dev->phys_dma_addr = pci_map_single(dev->pci_dev, (void *)buf,
+			R852_DMA_LEN,
+			(do_read ? PCI_DMA_FROMDEVICE : PCI_DMA_TODEVICE));
+
+		if (dev->phys_dma_addr == DMA_ERROR_CODE)
+			bounce = 1;
+	}
+
+	if (bounce) {
+		dbg_verbose("dma: using bounce buffer");
+		dev->phys_dma_addr = dev->phys_bounce_buffer;
+		if (!do_read)
+			memcpy(dev->bounce_buffer, buf, R852_DMA_LEN);
+	}
+
+	/* Enable DMA */
+	spin_lock_irqsave(&dev->irqlock, flags);
+	r852_dma_enable(dev);
+	spin_unlock_irqrestore(&dev->irqlock, flags);
+
+	/* Wait till complete */
+	error = r852_dma_wait(dev);
+
+	if (error) {
+		r852_dma_done(dev, error);
+		return;
+	}
+
+	if (do_read && bounce)
+		memcpy((void *)buf, dev->bounce_buffer, R852_DMA_LEN);
+}
+
+/*
+ * Program data lines of the nand chip to send data to it
+ */
+void r852_write_buf(struct mtd_info *mtd, const uint8_t *buf, int len)
+{
+	struct r852_device *dev = r852_get_dev(mtd);
+	u32 reg;
+
+	/* Don't allow any access to hardware if we suspect card removal */
+	if (dev->card_unstable)
+		return;
+
+	/* Special case for whole sector read */
+	if (len == R852_DMA_LEN && dev->dma_usable) {
+		r852_do_dma(dev, (uint8_t *)buf, 0);
+		return;
+	}
+
+	/* write DWORD chinks - faster */
+	while (len) {
+		reg = buf[0] | buf[1] << 8 | buf[2] << 16 | buf[3] << 24;
+		r852_write_reg_dword(dev, R852_DATALINE, reg);
+		buf += 4;
+		len -= 4;
+
+	}
+
+	/* write rest */
+	while (len)
+		r852_write_reg(dev, R852_DATALINE, *buf++);
+}
+
+/*
+ * Read data lines of the nand chip to retrieve data
+ */
+void r852_read_buf(struct mtd_info *mtd, uint8_t *buf, int len)
+{
+	struct r852_device *dev = r852_get_dev(mtd);
+	u32 reg;
+
+	if (dev->card_unstable) {
+		/* since we can't signal error here, at least, return
+			predictable buffer */
+		memset(buf, 0, len);
+		return;
+	}
+
+	/* special case for whole sector read */
+	if (len == R852_DMA_LEN && dev->dma_usable) {
+		r852_do_dma(dev, buf, 1);
+		return;
+	}
+
+	/* read in dword sized chunks */
+	while (len >= 4) {
+
+		reg = r852_read_reg_dword(dev, R852_DATALINE);
+		*buf++ = reg & 0xFF;
+		*buf++ = (reg >> 8) & 0xFF;
+		*buf++ = (reg >> 16) & 0xFF;
+		*buf++ = (reg >> 24) & 0xFF;
+		len -= 4;
+	}
+
+	/* read the reset by bytes */
+	while (len--)
+		*buf++ = r852_read_reg(dev, R852_DATALINE);
+}
+
+/*
+ * Read one byte from nand chip
+ */
+static uint8_t r852_read_byte(struct mtd_info *mtd)
+{
+	struct r852_device *dev = r852_get_dev(mtd);
+
+	/* Same problem as in r852_read_buf.... */
+	if (dev->card_unstable)
+		return 0;
+
+	return r852_read_reg(dev, R852_DATALINE);
+}
+
+
+/*
+ * Readback the buffer to verify it
+ */
+int r852_verify_buf(struct mtd_info *mtd, const uint8_t *buf, int len)
+{
+	struct r852_device *dev = r852_get_dev(mtd);
+
+	/* We can't be sure about anything here... */
+	if (dev->card_unstable)
+		return -1;
+
+	/* This will never happen, unless you wired up a nand chip
+		with > 512 bytes page size to the reader */
+	if (len > SM_SECTOR_SIZE)
+		return 0;
+
+	r852_read_buf(mtd, dev->tmp_buffer, len);
+	return memcmp(buf, dev->tmp_buffer, len);
+}
+
+/*
+ * Control several chip lines & send commands
+ */
+void r852_cmdctl(struct mtd_info *mtd, int dat, unsigned int ctrl)
+{
+	struct r852_device *dev = r852_get_dev(mtd);
+
+	if (dev->card_unstable)
+		return;
+
+	if (ctrl & NAND_CTRL_CHANGE) {
+
+		dev->ctlreg &= ~(R852_CTL_DATA | R852_CTL_COMMAND |
+				 R852_CTL_ON | R852_CTL_CARDENABLE);
+
+		if (ctrl & NAND_ALE)
+			dev->ctlreg |= R852_CTL_DATA;
+
+		if (ctrl & NAND_CLE)
+			dev->ctlreg |= R852_CTL_COMMAND;
+
+		if (ctrl & NAND_NCE)
+			dev->ctlreg |= (R852_CTL_CARDENABLE | R852_CTL_ON);
+		else
+			dev->ctlreg &= ~R852_CTL_WRITE;
+
+		/* when write is stareted, enable write access */
+		if (dat == NAND_CMD_ERASE1)
+			dev->ctlreg |= R852_CTL_WRITE;
+
+		r852_write_reg(dev, R852_CTL, dev->ctlreg);
+	}
+
+	 /* HACK: NAND_CMD_SEQIN is called without NAND_CTRL_CHANGE, but we need
+		to set write mode */
+	if (dat == NAND_CMD_SEQIN && (dev->ctlreg & R852_CTL_COMMAND)) {
+		dev->ctlreg |= R852_CTL_WRITE;
+		r852_write_reg(dev, R852_CTL, dev->ctlreg);
+	}
+
+	if (dat != NAND_CMD_NONE)
+		r852_write_reg(dev, R852_DATALINE, dat);
+}
+
+/*
+ * Wait till card is ready.
+ * based on nand_wait, but returns errors on DMA error
+ */
+int r852_wait(struct mtd_info *mtd, struct nand_chip *chip)
+{
+	struct r852_device *dev = (struct r852_device *)chip->priv;
+
+	unsigned long timeout;
+	int status;
+
+	timeout = jiffies + (chip->state == FL_ERASING ?
+		msecs_to_jiffies(400) : msecs_to_jiffies(20));
+
+	while (time_before(jiffies, timeout))
+		if (chip->dev_ready(mtd))
+			break;
+
+	chip->cmdfunc(mtd, NAND_CMD_STATUS, -1, -1);
+	status = (int)chip->read_byte(mtd);
+
+	/* Unfortunelly, no way to send detailed error status... */
+	if (dev->dma_error) {
+		status |= NAND_STATUS_FAIL;
+		dev->dma_error = 0;
+	}
+	return status;
+}
+
+/*
+ * Check if card is ready
+ */
+
+int r852_ready(struct mtd_info *mtd)
+{
+	struct r852_device *dev = r852_get_dev(mtd);
+	return !(r852_read_reg(dev, R852_CARD_STA) & R852_CARD_STA_BUSY);
+}
+
+
+/*
+ * Set ECC engine mode
+*/
+
+void r852_ecc_hwctl(struct mtd_info *mtd, int mode)
+{
+	struct r852_device *dev = r852_get_dev(mtd);
+
+	if (dev->card_unstable)
+		return;
+
+	switch (mode) {
+	case NAND_ECC_READ:
+	case NAND_ECC_WRITE:
+		/* enable ecc generation/check*/
+		dev->ctlreg |= R852_CTL_ECC_ENABLE;
+
+		/* flush ecc buffer */
+		r852_write_reg(dev, R852_CTL,
+			dev->ctlreg | R852_CTL_ECC_ACCESS);
+
+		r852_read_reg_dword(dev, R852_DATALINE);
+		r852_write_reg(dev, R852_CTL, dev->ctlreg);
+		return;
+
+	case NAND_ECC_READSYN:
+		/* disable ecc generation */
+		dev->ctlreg &= ~R852_CTL_ECC_ENABLE;
+		r852_write_reg(dev, R852_CTL, dev->ctlreg);
+	}
+}
+
+/*
+ * Calculate ECC, only used for writes
+ */
+
+int r852_ecc_calculate(struct mtd_info *mtd, const uint8_t *dat,
+							uint8_t *ecc_code)
+{
+	struct r852_device *dev = r852_get_dev(mtd);
+	struct sm_oob *oob = (struct sm_oob *)ecc_code;
+	u32 ecc1, ecc2;
+
+	if (dev->card_unstable)
+		return 0;
+
+	dev->ctlreg &= ~R852_CTL_ECC_ENABLE;
+	r852_write_reg(dev, R852_CTL, dev->ctlreg | R852_CTL_ECC_ACCESS);
+
+	ecc1 = r852_read_reg_dword(dev, R852_DATALINE);
+	ecc2 = r852_read_reg_dword(dev, R852_DATALINE);
+
+	oob->ecc1[0] = (ecc1) & 0xFF;
+	oob->ecc1[1] = (ecc1 >> 8) & 0xFF;
+	oob->ecc1[2] = (ecc1 >> 16) & 0xFF;
+
+	oob->ecc2[0] = (ecc2) & 0xFF;
+	oob->ecc2[1] = (ecc2 >> 8) & 0xFF;
+	oob->ecc2[2] = (ecc2 >> 16) & 0xFF;
+
+	r852_write_reg(dev, R852_CTL, dev->ctlreg);
+	return 0;
+}
+
+/*
+ * Correct the data using ECC, hw did almost everything for us
+ */
+
+int r852_ecc_correct(struct mtd_info *mtd, uint8_t *dat,
+				uint8_t *read_ecc, uint8_t *calc_ecc)
+{
+	u16 ecc_reg;
+	u8 ecc_status, err_byte;
+	int i, error = 0;
+
+	struct r852_device *dev = r852_get_dev(mtd);
+
+	if (dev->card_unstable)
+		return 0;
+
+	r852_write_reg(dev, R852_CTL, dev->ctlreg | R852_CTL_ECC_ACCESS);
+	ecc_reg = r852_read_reg_dword(dev, R852_DATALINE);
+	r852_write_reg(dev, R852_CTL, dev->ctlreg);
+
+	for (i = 0 ; i <= 1 ; i++) {
+
+		ecc_status = (ecc_reg >> 8) & 0xFF;
+
+		/* ecc uncorrectable error */
+		if (ecc_status & R852_ECC_FAIL) {
+			dbg("ecc: unrecoverable error, in half %d", i);
+			error = -1;
+			goto exit;
+		}
+
+		/* correctable error */
+		if (ecc_status & R852_ECC_CORRECTABLE) {
+
+			err_byte = ecc_reg & 0xFF;
+			dbg("ecc: recoverable error, "
+				"in half %d, byte %d, bit %d", i,
+				err_byte, ecc_status & R852_ECC_ERR_BIT_MSK);
+
+			dat[err_byte] ^=
+				1 << (ecc_status & R852_ECC_ERR_BIT_MSK);
+			error++;
+		}
+
+		dat += 256;
+		ecc_reg >>= 16;
+	}
+exit:
+	return error;
+}
+
+/*
+ * This is copy of nand_read_oob_std
+ * nand_read_oob_syndrome assumes we can send column address - we can't
+ */
+static int r852_read_oob(struct mtd_info *mtd, struct nand_chip *chip,
+			     int page, int sndcmd)
+{
+	if (sndcmd) {
+		chip->cmdfunc(mtd, NAND_CMD_READOOB, 0, page);
+		sndcmd = 0;
+	}
+	chip->read_buf(mtd, chip->oob_poi, mtd->oobsize);
+	return sndcmd;
+}
+
+/*
+ * Start the nand engine
+ */
+
+void r852_engine_enable(struct r852_device *dev)
+{
+	if (r852_read_reg_dword(dev, R852_HW) & R852_HW_UNKNOWN) {
+		r852_write_reg(dev, R852_CTL, R852_CTL_RESET | R852_CTL_ON);
+		r852_write_reg_dword(dev, R852_HW, R852_HW_ENABLED);
+	} else {
+		r852_write_reg_dword(dev, R852_HW, R852_HW_ENABLED);
+		r852_write_reg(dev, R852_CTL, R852_CTL_RESET | R852_CTL_ON);
+	}
+	msleep(300);
+	r852_write_reg(dev, R852_CTL, 0);
+}
+
+
+/*
+ * Stop the nand engine
+ */
+
+void r852_engine_disable(struct r852_device *dev)
+{
+	r852_write_reg_dword(dev, R852_HW, 0);
+	r852_write_reg(dev, R852_CTL, R852_CTL_RESET);
+}
+
+/*
+ * Test if card is present
+ */
+
+void r852_card_update_present(struct r852_device *dev)
+{
+	unsigned long flags;
+	u8 reg;
+
+	spin_lock_irqsave(&dev->irqlock, flags);
+	reg = r852_read_reg(dev, R852_CARD_STA);
+	dev->card_detected = !!(reg & R852_CARD_STA_PRESENT);
+	spin_unlock_irqrestore(&dev->irqlock, flags);
+}
+
+/*
+ * Update card detection IRQ state according to current card state
+ * which is read in r852_card_update_present
+ */
+void r852_update_card_detect(struct r852_device *dev)
+{
+	int card_detect_reg = r852_read_reg(dev, R852_CARD_IRQ_ENABLE);
+
+	card_detect_reg &= ~(R852_CARD_IRQ_REMOVE | R852_CARD_IRQ_INSERT);
+	card_detect_reg |= R852_CARD_IRQ_GENABLE;
+
+	card_detect_reg |= dev->card_detected ?
+		R852_CARD_IRQ_REMOVE : R852_CARD_IRQ_INSERT;
+
+	r852_write_reg(dev, R852_CARD_IRQ_ENABLE, card_detect_reg);
+}
+
+ssize_t r852_media_type_show(struct device *sys_dev,
+		struct device_attribute *attr, char *buf)
+{
+	struct mtd_info *mtd = container_of(sys_dev, struct mtd_info, dev);
+	struct r852_device *dev = r852_get_dev(mtd);
+	char *data = dev->sm ? "smartmedia" : "xd";
+
+	strcpy(buf, data);
+	return strlen(data);
+}
+
+DEVICE_ATTR(media_type, S_IRUGO, r852_media_type_show, NULL);
+
+
+/* Detect properties of card in slot */
+void r852_update_media_status(struct r852_device *dev)
+{
+	u8 reg;
+	unsigned long flags;
+	int readonly;
+
+	spin_lock_irqsave(&dev->irqlock, flags);
+	if (!dev->card_detected) {
+		message("card removed");
+		spin_unlock_irqrestore(&dev->irqlock, flags);
+		return ;
+	}
+
+	readonly  = r852_read_reg(dev, R852_CARD_STA) & R852_CARD_STA_RO;
+	reg = r852_read_reg(dev, R852_DMA_CAP);
+	dev->sm = (reg & (R852_DMA1 | R852_DMA2)) && (reg & R852_SMBIT);
+
+	message("detected %s %s card in slot",
+		dev->sm ? "SmartMedia" : "xD",
+		readonly ? "readonly" : "writeable");
+
+	dev->readonly = readonly;
+	spin_unlock_irqrestore(&dev->irqlock, flags);
+}
+
+/*
+ * Register the nand device
+ * Called when the card is detected
+ */
+int r852_register_nand_device(struct r852_device *dev)
+{
+	dev->mtd = kzalloc(sizeof(struct mtd_info), GFP_KERNEL);
+
+	if (!dev->mtd)
+		goto error1;
+
+	WARN_ON(dev->card_registred);
+
+	dev->mtd->owner = THIS_MODULE;
+	dev->mtd->priv = dev->chip;
+	dev->mtd->dev.parent = &dev->pci_dev->dev;
+
+	if (dev->readonly)
+		dev->chip->options |= NAND_ROM;
+
+	r852_engine_enable(dev);
+
+	if (sm_register_device(dev->mtd))
+		goto error2;
+
+	device_create_file(&dev->mtd->dev, &dev_attr_media_type);
+	dev->card_registred = 1;
+	return 0;
+error2:
+	kfree(dev->mtd);
+error1:
+	/* Force card redetect */
+	dev->card_detected = 0;
+	return -1;
+}
+
+/*
+ * Unregister the card
+ */
+
+void r852_unregister_nand_device(struct r852_device *dev)
+{
+	if (!dev->card_registred)
+		return;
+
+	device_remove_file(&dev->mtd->dev, &dev_attr_media_type);
+	nand_release(dev->mtd);
+	r852_engine_disable(dev);
+	dev->card_registred = 0;
+	kfree(dev->mtd);
+	dev->mtd = NULL;
+}
+
+/* Card state updater */
+void r852_card_detect_work(struct work_struct *work)
+{
+	struct r852_device *dev =
+		container_of(work, struct r852_device, card_detect_work.work);
+
+	r852_update_card_detect(dev);
+	dev->card_unstable = 0;
+
+	/* false alarm */
+	if (dev->card_detected == dev->card_registred)
+		goto exit;
+
+	/* Read media properties */
+	r852_update_media_status(dev);
+
+	/* Register the card */
+	if (dev->card_detected)
+		r852_register_nand_device(dev);
+	else
+		r852_unregister_nand_device(dev);
+exit:
+	/* Update detection logic */
+	r852_update_card_detect(dev);
+}
+
+/* Ack + disable IRQ generation */
+static void r852_disable_irqs(struct r852_device *dev)
+{
+	u8 reg;
+	reg = r852_read_reg(dev, R852_CARD_IRQ_ENABLE);
+	r852_write_reg(dev, R852_CARD_IRQ_ENABLE, reg & ~R852_CARD_IRQ_MASK);
+
+	reg = r852_read_reg_dword(dev, R852_DMA_IRQ_ENABLE);
+	r852_write_reg_dword(dev, R852_DMA_IRQ_ENABLE,
+					reg & ~R852_DMA_IRQ_MASK);
+
+	r852_write_reg(dev, R852_CARD_IRQ_STA, R852_CARD_IRQ_MASK);
+	r852_write_reg_dword(dev, R852_DMA_IRQ_STA, R852_DMA_IRQ_MASK);
+}
+
+/* Interrupt handler */
+static irqreturn_t r852_irq(int irq, void *data)
+{
+	struct r852_device *dev = (struct r852_device *)data;
+
+	u8 card_status, dma_status;
+	unsigned long flags;
+	irqreturn_t ret = IRQ_NONE;
+
+	spin_lock_irqsave(&dev->irqlock, flags);
+
+	/* We can recieve shared interrupt while pci is suspended
+		in that case reads will return 0xFFFFFFFF.... */
+	if (dev->insuspend)
+		goto out;
+
+	/* handle card detection interrupts first */
+	card_status = r852_read_reg(dev, R852_CARD_IRQ_STA);
+	r852_write_reg(dev, R852_CARD_IRQ_STA, card_status);
+
+	if (card_status & (R852_CARD_IRQ_INSERT|R852_CARD_IRQ_REMOVE)) {
+
+		ret = IRQ_HANDLED;
+		dev->card_detected = !!(card_status & R852_CARD_IRQ_INSERT);
+
+		/* we shouldn't recieve any interrupts if we wait for card
+			to settle */
+		WARN_ON(dev->card_unstable);
+
+		/* disable irqs while card is unstable */
+		/* this will timeout DMA if active, but better that garbage */
+		r852_disable_irqs(dev);
+
+		if (dev->card_unstable)
+			goto out;
+
+		/* let, card state to settle a bit, and then do the work */
+		dev->card_unstable = 1;
+		queue_delayed_work(dev->card_workqueue,
+			&dev->card_detect_work, msecs_to_jiffies(100));
+		goto out;
+	}
+
+
+	/* Handle dma interrupts */
+	dma_status = r852_read_reg_dword(dev, R852_DMA_IRQ_STA);
+	r852_write_reg_dword(dev, R852_DMA_IRQ_STA, dma_status);
+
+	if (dma_status & R852_DMA_IRQ_MASK) {
+
+		ret = IRQ_HANDLED;
+
+		if (dma_status & R852_DMA_IRQ_ERROR) {
+			dbg("recieved dma error IRQ");
+			r852_dma_done(dev, -EIO);
+			goto out;
+		}
+
+		/* recieved DMA interrupt out of nowhere? */
+		WARN_ON_ONCE(dev->dma_stage == 0);
+
+		if (dev->dma_stage == 0)
+			goto out;
+
+		/* done device access */
+		if (dev->dma_state == DMA_INTERNAL &&
+				(dma_status & R852_DMA_IRQ_INTERNAL)) {
+
+			dev->dma_state = DMA_MEMORY;
+			dev->dma_stage++;
+		}
+
+		/* done memory DMA */
+		if (dev->dma_state == DMA_MEMORY &&
+				(dma_status & R852_DMA_IRQ_MEMORY)) {
+			dev->dma_state = DMA_INTERNAL;
+			dev->dma_stage++;
+		}
+
+		/* Enable 2nd half of dma dance */
+		if (dev->dma_stage == 2)
+			r852_dma_enable(dev);
+
+		/* Operation done */
+		if (dev->dma_stage == 3)
+			r852_dma_done(dev, 0);
+		goto out;
+	}
+
+	/* Handle unknown interrupts */
+	if (dma_status)
+		dbg("bad dma IRQ status = %x", dma_status);
+
+	if (card_status & ~R852_CARD_STA_CD)
+		dbg("strange card status = %x", card_status);
+
+out:
+	spin_unlock_irqrestore(&dev->irqlock, flags);
+	return ret;
+}
+
+int  r852_probe(struct pci_dev *pci_dev, const struct pci_device_id *id)
+{
+	int error;
+	struct nand_chip *chip;
+	struct r852_device *dev;
+
+	/* pci initialization */
+	error = pci_enable_device(pci_dev);
+
+	if (error)
+		goto error1;
+
+	pci_set_master(pci_dev);
+
+	error = pci_set_dma_mask(pci_dev, DMA_32BIT_MASK);
+	if (error)
+		goto error2;
+
+	error = pci_request_regions(pci_dev, DRV_NAME);
+
+	if (error)
+		goto error3;
+
+	error = -ENOMEM;
+
+	/* init nand chip, but register it only on card insert */
+	chip = kzalloc(sizeof(struct nand_chip), GFP_KERNEL);
+
+	if (!chip)
+		goto error4;
+
+	/* commands */
+	chip->cmd_ctrl = r852_cmdctl;
+	chip->waitfunc = r852_wait;
+	chip->dev_ready = r852_ready;
+
+	/* I/O */
+	chip->read_byte = r852_read_byte;
+	chip->read_buf = r852_read_buf;
+	chip->write_buf = r852_write_buf;
+	chip->verify_buf = r852_verify_buf;
+
+	/* ecc */
+	chip->ecc.mode = NAND_ECC_HW_SYNDROME;
+	chip->ecc.size = R852_DMA_LEN;
+	chip->ecc.bytes = SM_OOB_SIZE;
+	chip->ecc.hwctl = r852_ecc_hwctl;
+	chip->ecc.calculate = r852_ecc_calculate;
+	chip->ecc.correct = r852_ecc_correct;
+
+	/* TODO: hack */
+	chip->ecc.read_oob = r852_read_oob;
+
+	/* init our device structure */
+	dev = kzalloc(sizeof(struct r852_device), GFP_KERNEL);
+
+	if (!dev)
+		goto error5;
+
+	chip->priv = dev;
+	dev->chip = chip;
+	dev->pci_dev = pci_dev;
+	pci_set_drvdata(pci_dev, dev);
+
+	dev->bounce_buffer = pci_alloc_consistent(pci_dev, R852_DMA_LEN,
+		&dev->phys_bounce_buffer);
+
+	if (!dev->bounce_buffer)
+		goto error6;
+
+
+	error = -ENODEV;
+	dev->mmio = pci_ioremap_bar(pci_dev, 0);
+
+	if (!dev->mmio)
+		goto error7;
+
+	error = -ENOMEM;
+	dev->tmp_buffer = kzalloc(SM_SECTOR_SIZE, GFP_KERNEL);
+
+	if (!dev->tmp_buffer)
+		goto error8;
+
+	init_completion(&dev->dma_done);
+
+	dev->card_workqueue = create_freezeable_workqueue(DRV_NAME);
+
+	if (!dev->card_workqueue)
+		goto error9;
+
+	INIT_DELAYED_WORK(&dev->card_detect_work, r852_card_detect_work);
+
+	/* shutdown everything - precation */
+	r852_engine_disable(dev);
+	r852_disable_irqs(dev);
+
+	r852_dma_test(dev);
+
+	/*register irq handler*/
+	error = -ENODEV;
+	if (request_irq(pci_dev->irq, &r852_irq, IRQF_SHARED,
+			  DRV_NAME, dev))
+		goto error10;
+
+	dev->irq = pci_dev->irq;
+	spin_lock_init(&dev->irqlock);
+
+	/* kick initial present test */
+	dev->card_detected = 0;
+	r852_card_update_present(dev);
+	queue_delayed_work(dev->card_workqueue,
+		&dev->card_detect_work, 0);
+
+
+	printk(KERN_NOTICE DRV_NAME ": driver loaded succesfully\n");
+	return 0;
+
+error10:
+	destroy_workqueue(dev->card_workqueue);
+error9:
+	kfree(dev->tmp_buffer);
+error8:
+	pci_iounmap(pci_dev, dev->mmio);
+error7:
+	pci_free_consistent(pci_dev, R852_DMA_LEN,
+		dev->bounce_buffer, dev->phys_bounce_buffer);
+error6:
+	kfree(dev);
+error5:
+	kfree(chip);
+error4:
+	pci_release_regions(pci_dev);
+error3:
+error2:
+	pci_disable_device(pci_dev);
+error1:
+	return error;
+}
+
+void r852_remove(struct pci_dev *pci_dev)
+{
+	struct r852_device *dev = pci_get_drvdata(pci_dev);
+
+	/* Stop detect workqueue -
+		we are going to unregister the device anyway*/
+	cancel_delayed_work_sync(&dev->card_detect_work);
+	destroy_workqueue(dev->card_workqueue);
+
+	/* Unregister the device, this might make more IO */
+	r852_unregister_nand_device(dev);
+
+	/* Stop interrupts */
+	r852_disable_irqs(dev);
+	synchronize_irq(dev->irq);
+	free_irq(dev->irq, dev);
+
+	/* Cleanup */
+	kfree(dev->tmp_buffer);
+	pci_iounmap(pci_dev, dev->mmio);
+	pci_free_consistent(pci_dev, R852_DMA_LEN,
+		dev->bounce_buffer, dev->phys_bounce_buffer);
+
+	kfree(dev->chip);
+	kfree(dev);
+
+	/* Shutdown the PCI device */
+	pci_release_regions(pci_dev);
+	pci_disable_device(pci_dev);
+}
+
+void r852_shutdown(struct pci_dev *pci_dev)
+{
+	struct r852_device *dev = pci_get_drvdata(pci_dev);
+
+	cancel_delayed_work_sync(&dev->card_detect_work);
+	r852_disable_irqs(dev);
+	synchronize_irq(dev->irq);
+	pci_disable_device(pci_dev);
+}
+
+int r852_suspend(struct device *device)
+{
+	struct r852_device *dev = pci_get_drvdata(to_pci_dev(device));
+	unsigned long flags;
+
+	if (dev->ctlreg & R852_CTL_CARDENABLE)
+		return -EBUSY;
+
+	/* First make sure the detect work is gone */
+	cancel_delayed_work_sync(&dev->card_detect_work);
+
+	/* Turn off the interrupts and stop the device */
+	r852_disable_irqs(dev);
+	r852_engine_disable(dev);
+
+	spin_lock_irqsave(&dev->irqlock, flags);
+	dev->insuspend = 1;
+	spin_unlock_irqrestore(&dev->irqlock, flags);
+
+	/* At that point, even if interrupt handler is running, it will quit */
+	/* So wait for this to happen explictly */
+	synchronize_irq(dev->irq);
+
+	/* If card was pulled off just during the suspend, which is very
+		unlikely, we will remove it on resume, it too late now
+		anyway... */
+	dev->card_unstable = 0;
+
+	pci_save_state(to_pci_dev(device));
+	return pci_prepare_to_sleep(to_pci_dev(device));
+}
+
+int r852_resume(struct device *device)
+{
+	struct r852_device *dev = pci_get_drvdata(to_pci_dev(device));
+	unsigned long flags;
+
+	/* Turn on the hardware */
+	pci_back_from_sleep(to_pci_dev(device));
+	pci_restore_state(to_pci_dev(device));
+
+	r852_disable_irqs(dev);
+	r852_card_update_present(dev);
+	r852_engine_disable(dev);
+
+
+	/* Now its safe for IRQ to run */
+	spin_lock_irqsave(&dev->irqlock, flags);
+	dev->insuspend = 0;
+	spin_unlock_irqrestore(&dev->irqlock, flags);
+
+
+	/* If card status changed, just do the work */
+	if (dev->card_detected != dev->card_registred) {
+		dbg("card was %s during low power state",
+			dev->card_detected ? "added" : "removed");
+
+		queue_delayed_work(dev->card_workqueue,
+		&dev->card_detect_work, 1000);
+		return 0;
+	}
+
+	/* Otherwise, initialize the card */
+	if (dev->card_registred) {
+		r852_engine_enable(dev);
+		dev->chip->select_chip(dev->mtd, 0);
+		dev->chip->cmdfunc(dev->mtd, NAND_CMD_RESET, -1, -1);
+		dev->chip->select_chip(dev->mtd, -1);
+	}
+
+	/* Program card detection IRQ */
+	r852_update_card_detect(dev);
+	return 0;
+}
+
+static const struct pci_device_id r852_pci_id_tbl[] = {
+
+	{ PCI_VDEVICE(RICOH, PCI_DEVICE_ID_RICOH_R5C852), },
+	{ },
+};
+
+MODULE_DEVICE_TABLE(pci, r852_pci_id_tbl);
+
+SIMPLE_DEV_PM_OPS(r852_pm_ops, r852_suspend, r852_resume);
+
+
+static struct pci_driver r852_pci_driver = {
+	.name		= DRV_NAME,
+	.id_table	= r852_pci_id_tbl,
+	.probe		= r852_probe,
+	.remove		= r852_remove,
+	.shutdown	= r852_shutdown,
+	.driver.pm	= &r852_pm_ops,
+};
+
+static __init int r852_module_init(void)
+{
+	return pci_register_driver(&r852_pci_driver);
+}
+
+static void __exit r852_module_exit(void)
+{
+	pci_unregister_driver(&r852_pci_driver);
+}
+
+module_init(r852_module_init);
+module_exit(r852_module_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Maxim Levitsky <maximlevitsky@gmail.com>");
+MODULE_DESCRIPTION("Ricoh 85xx xD/smartmedia card reader driver");
diff --git a/drivers/mtd/nand/r852.h b/drivers/mtd/nand/r852.h
new file mode 100644
index 0000000..e05e795
--- /dev/null
+++ b/drivers/mtd/nand/r852.h
@@ -0,0 +1,163 @@
+/*
+ * Copyright (C) 2009 - Maxim Levitsky
+ * driver for Ricoh xD readers
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/pci.h>
+#include <linux/completion.h>
+#include <linux/workqueue.h>
+#include <linux/mtd/nand.h>
+#include <linux/spinlock.h>
+
+
+/* nand interface + ecc
+   byte write/read does one cycle on nand data lines.
+   dword write/read does 4 cycles
+   if R852_CTL_ECC_ACCESS is set in R852_CTL, then dword read reads
+   results of ecc correction, if DMA read was done before.
+   If write was done two dword reads read generated ecc checksums
+*/
+#define	R852_DATALINE		0x00
+
+/* control register */
+#define R852_CTL		0x04
+#define R852_CTL_COMMAND 	0x01	/* send command (#CLE)*/
+#define R852_CTL_DATA		0x02	/* read/write data (#ALE)*/
+#define R852_CTL_ON		0x04	/* only seem to controls the hd led, */
+					/* but has to be set on start...*/
+#define R852_CTL_RESET		0x08	/* unknown, set only on start once*/
+#define R852_CTL_CARDENABLE	0x10	/* probably (#CE) - always set*/
+#define R852_CTL_ECC_ENABLE	0x20	/* enable ecc engine */
+#define R852_CTL_ECC_ACCESS	0x40	/* read/write ecc via reg #0*/
+#define R852_CTL_WRITE		0x80	/* set when performing writes (#WP) */
+
+/* card detection status */
+#define R852_CARD_STA		0x05
+
+#define R852_CARD_STA_CD	0x01	/* state of #CD line, same as 0x04 */
+#define R852_CARD_STA_RO	0x02	/* card is readonly */
+#define R852_CARD_STA_PRESENT	0x04	/* card is present (#CD) */
+#define R852_CARD_STA_ABSENT	0x08	/* card is absent */
+#define R852_CARD_STA_BUSY	0x80	/* card is busy - (#R/B) */
+
+/* card detection irq status & enable*/
+#define R852_CARD_IRQ_STA	0x06	/* IRQ status */
+#define R852_CARD_IRQ_ENABLE	0x07	/* IRQ enable */
+
+#define R852_CARD_IRQ_CD	0x01	/* fire when #CD lights, same as 0x04*/
+#define R852_CARD_IRQ_REMOVE	0x04	/* detect card removal */
+#define R852_CARD_IRQ_INSERT	0x08	/* detect card insert */
+#define R852_CARD_IRQ_UNK1	0x10	/* unknown */
+#define R852_CARD_IRQ_GENABLE	0x80	/* general enable */
+#define R852_CARD_IRQ_MASK	0x1D
+
+
+
+/* hardware enable */
+#define R852_HW			0x08
+#define R852_HW_ENABLED		0x01	/* hw enabled */
+#define R852_HW_UNKNOWN		0x80
+
+
+/* dma capabilities */
+#define R852_DMA_CAP		0x09
+#define R852_SMBIT		0x20	/* if set with bit #6 or bit #7, then */
+					/* hw is smartmedia */
+#define R852_DMA1		0x40	/* if set w/bit #7, dma is supported */
+#define R852_DMA2		0x80	/* if set w/bit #6, dma is supported */
+
+
+/* physical DMA address - 32 bit value*/
+#define R852_DMA_ADDR		0x0C
+
+
+/* dma settings */
+#define R852_DMA_SETTINGS	0x10
+#define R852_DMA_MEMORY		0x01	/* (memory <-> internal hw buffer) */
+#define R852_DMA_READ		0x02	/* 0 = write, 1 = read */
+#define R852_DMA_INTERNAL	0x04	/* (internal hw buffer <-> card) */
+
+/* dma IRQ status */
+#define R852_DMA_IRQ_STA		0x14
+
+/* dma IRQ enable */
+#define R852_DMA_IRQ_ENABLE	0x18
+
+#define R852_DMA_IRQ_MEMORY	0x01	/* (memory <-> internal hw buffer) */
+#define R852_DMA_IRQ_ERROR	0x02	/* error did happen */
+#define R852_DMA_IRQ_INTERNAL	0x04	/* (internal hw buffer <-> card) */
+#define R852_DMA_IRQ_MASK	0x07	/* mask of all IRQ bits */
+
+
+/* ECC syndrome format - read from reg #0 will return two copies of these for
+   each half of the page.
+   first byte is error byte location, and second, bit location + flags */
+#define R852_ECC_ERR_BIT_MSK	0x07	/* error bit location */
+#define R852_ECC_CORRECT		0x10	/* no errors - (guessed) */
+#define R852_ECC_CORRECTABLE	0x20	/* correctable error exist */
+#define R852_ECC_FAIL		0x40	/* non correctable error detected */
+
+#define R852_DMA_LEN		512
+
+#define DMA_INTERNAL	0
+#define DMA_MEMORY	1
+
+struct r852_device {
+	void __iomem *mmio;		/* mmio */
+	struct mtd_info *mtd;		/* mtd backpointer */
+	struct nand_chip *chip;		/* nand chip backpointer */
+	struct pci_dev *pci_dev;	/* pci backpointer */
+
+	/* dma area */
+	dma_addr_t phys_dma_addr;	/* bus address of buffer*/
+	struct completion dma_done;	/* data transfer done */
+
+	dma_addr_t phys_bounce_buffer;	/* bus address of bounce buffer */
+	u8 *bounce_buffer;		/* virtual address of bounce buffer */
+
+	int dma_dir;			/* 1 = read, 0 = write */
+	int dma_stage;			/* 0 - idle, 1 - first step,
+					   2 - second step */
+
+	int dma_state;			/* 0 = internal, 1 = memory */
+	int dma_error;			/* dma errors */
+	int dma_usable;			/* is it possible to use dma */
+
+	/* card status area */
+	struct delayed_work card_detect_work;
+	struct workqueue_struct *card_workqueue;
+	int card_registred;		/* card registered with mtd */
+	int card_detected;		/* card detected in slot */
+	int card_unstable;		/* whenever the card is inserted,
+					   is not known yet */
+	int readonly;			/* card is readonly */
+	int sm;				/* Is card smartmedia */
+
+	/* interrupt handling */
+	spinlock_t irqlock;		/* IRQ protecting lock */
+	int irq;			/* irq num */
+	int insuspend;			/* device is suspended */
+
+	/* misc */
+	void *tmp_buffer;		/* temporary buffer */
+	u8 ctlreg;			/* cached contents of control reg */
+};
+
+#define DRV_NAME "r852"
+
+
+#define dbg(format, ...) \
+	if (debug) \
+		printk(KERN_DEBUG DRV_NAME ": " format "\n", ## __VA_ARGS__)
+
+#define dbg_verbose(format, ...) \
+	if (debug > 1) \
+		printk(KERN_DEBUG DRV_NAME ": " format "\n", ## __VA_ARGS__)
+
+
+#define message(format, ...) \
+	printk(KERN_INFO DRV_NAME ": " format "\n", ## __VA_ARGS__)
diff --git a/include/linux/pci_ids.h b/include/linux/pci_ids.h
index cca8a04..cef49f2 100644
--- a/include/linux/pci_ids.h
+++ b/include/linux/pci_ids.h
@@ -1523,6 +1523,8 @@
 #define PCI_DEVICE_ID_RICOH_R5C822	0x0822
 #define PCI_DEVICE_ID_RICOH_R5C832	0x0832
 #define PCI_DEVICE_ID_RICOH_R5C843	0x0843
+#define PCI_DEVICE_ID_RICOH_R5C852	0x0852
+
 
 #define PCI_VENDOR_ID_DLINK		0x1186
 #define PCI_DEVICE_ID_DLINK_DGE510T	0x4c00
-- 
1.6.3.3

[PATCH 15/15] MTD: Add nand driver for Ricoh xD/SmartMedia reader

[Maxim Levitsky]

This adds a driver for Ricoh R5C852 xD card reader.

This reader is a part of larger mulifunction chip
and found at least in R5C832

Driver is complete, but bewere of the fact that some
(probably only type M) xD cards are 'fake' which means that
they have an on board CPU and expose emulated nand command set

These cards don't even store the  oob area on the flash,
but generate it on the fly from something else.

Thus they demand to have proper values written in the oob area,
and therefore only useful with SmartMedia FTL.

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
---
 MAINTAINERS               |    6 +
 drivers/mtd/nand/Kconfig  |   11 +
 drivers/mtd/nand/Makefile |    1 +
 drivers/mtd/nand/r852.c   | 1117 +++++++++++++++++++++++++++++++++++++++++++++
 drivers/mtd/nand/r852.h   |  163 +++++++
 include/linux/pci_ids.h   |    2 +
 6 files changed, 1300 insertions(+), 0 deletions(-)
 create mode 100644 drivers/mtd/nand/r852.c
 create mode 100644 drivers/mtd/nand/r852.h

diff --git a/MAINTAINERS b/MAINTAINERS
index 1b0f053..1723c54 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -4577,6 +4577,12 @@ S:	Maintained
 F:	Documentation/rfkill.txt
 F:	net/rfkill/
 
+RICOH SMARTMEDIA/XD DRIVER
+M:	Maxim Levitsky <maximlevitsky@gmail.com>
+S:	Maintained
+F:	drivers/mtd/nand/r822.c
+F:	drivers/mtd/nand/r822.h
+
 RISCOM8 DRIVER
 S:	Orphan
 F:	Documentation/serial/riscom8.txt
diff --git a/drivers/mtd/nand/Kconfig b/drivers/mtd/nand/Kconfig
index 31273c7..4ab2551 100644
--- a/drivers/mtd/nand/Kconfig
+++ b/drivers/mtd/nand/Kconfig
@@ -107,6 +107,17 @@ config MTD_NAND_OMAP_PREFETCH_DMA
 config MTD_NAND_IDS
 	tristate
 
+config MTD_NAND_RICOH
+	tristate "Ricoh xD card reader"
+	default n
+	select MTD_SM_COMMON
+	help
+	  Enable support for Ricoh R5C852 xD card reader
+	  You also need to enable ether
+	  NAND SSFDC (SmartMedia) read only translation layer' or new
+	  expermental, readwrite
+	  'SmartMedia/xD new translation layer'
+
 config MTD_NAND_AU1550
 	tristate "Au1550/1200 NAND support"
 	depends on SOC_AU1200 || SOC_AU1550
diff --git a/drivers/mtd/nand/Makefile b/drivers/mtd/nand/Makefile
index e699bef..1d5797d 100644
--- a/drivers/mtd/nand/Makefile
+++ b/drivers/mtd/nand/Makefile
@@ -42,5 +42,6 @@ obj-$(CONFIG_MTD_NAND_TXX9NDFMC)	+= txx9ndfmc.o
 obj-$(CONFIG_MTD_NAND_NUC900)		+= nuc900_nand.o
 obj-$(CONFIG_MTD_NAND_NOMADIK)		+= nomadik_nand.o
 obj-$(CONFIG_MTD_NAND_BCM_UMI)		+= bcm_umi_nand.o nand_bcm_umi.o
+obj-$(CONFIG_MTD_NAND_RICOH)		+= r852.o
 
 nand-objs := nand_base.o nand_bbt.o
diff --git a/drivers/mtd/nand/r852.c b/drivers/mtd/nand/r852.c
new file mode 100644
index 0000000..5c718ff
--- /dev/null
+++ b/drivers/mtd/nand/r852.c
@@ -0,0 +1,1117 @@
+/*
+ * Copyright (C) 2009 - Maxim Levitsky
+ * driver for Ricoh xD readers
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/jiffies.h>
+#include <linux/workqueue.h>
+#include <linux/interrupt.h>
+#include <linux/pci_ids.h>
+#include <asm/byteorder.h>
+#include <linux/sched.h>
+#include "sm_common.h"
+#include "r852.h"
+
+
+static int enable_dma = 1;
+module_param(enable_dma, bool, S_IRUGO);
+MODULE_PARM_DESC(enable_dma, "Enable usage of the DMA (default)");
+
+static int debug;
+module_param(debug, int, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(debug, "Debug level (0-2)");
+
+/* read register */
+static inline u8 r852_read_reg(struct r852_device *dev, int address)
+{
+	u8 reg = readb(dev->mmio + address);
+	return reg;
+}
+
+/* write register */
+static inline void r852_write_reg(struct r852_device *dev,
+						int address, u8 value)
+{
+	writeb(value, dev->mmio + address);
+	mmiowb();
+}
+
+
+/* read dword sized register */
+static inline u32 r852_read_reg_dword(struct r852_device *dev, int address)
+{
+	u32 reg = le32_to_cpu(readl(dev->mmio + address));
+	return reg;
+}
+
+/* write dword sized register */
+static inline void r852_write_reg_dword(struct r852_device *dev,
+							int address, u32 value)
+{
+	writel(cpu_to_le32(value), dev->mmio + address);
+	mmiowb();
+}
+
+/* returns pointer to our private structure */
+static inline struct r852_device *r852_get_dev(struct mtd_info *mtd)
+{
+	struct nand_chip *chip = (struct nand_chip *)mtd->priv;
+	return (struct r852_device *)chip->priv;
+}
+
+
+/* check if controller supports dma */
+static void r852_dma_test(struct r852_device *dev)
+{
+	dev->dma_usable = (r852_read_reg(dev, R852_DMA_CAP) &
+		(R852_DMA1 | R852_DMA2)) == (R852_DMA1 | R852_DMA2);
+
+	if (!dev->dma_usable)
+		message("Non dma capable device detected, dma disabled");
+
+	if (!enable_dma) {
+		message("disabling dma on user request");
+		dev->dma_usable = 0;
+	}
+}
+
+/*
+ * Enable dma. Enables ether first or second stage of the DMA,
+ * Expects dev->dma_dir and dev->dma_state be set
+ */
+static void r852_dma_enable(struct r852_device *dev)
+{
+	u8 dma_reg, dma_irq_reg;
+
+	/* Set up dma settings */
+	dma_reg = r852_read_reg_dword(dev, R852_DMA_SETTINGS);
+	dma_reg &= ~(R852_DMA_READ | R852_DMA_INTERNAL | R852_DMA_MEMORY);
+
+	if (dev->dma_dir)
+		dma_reg |= R852_DMA_READ;
+
+	if (dev->dma_state == DMA_INTERNAL)
+		dma_reg |= R852_DMA_INTERNAL;
+	else {
+		dma_reg |= R852_DMA_MEMORY;
+		r852_write_reg_dword(dev, R852_DMA_ADDR,
+			cpu_to_le32(dev->phys_dma_addr));
+	}
+
+	r852_write_reg_dword(dev, R852_DMA_SETTINGS, dma_reg);
+
+	/* Set dma irq */
+	dma_irq_reg = r852_read_reg_dword(dev, R852_DMA_IRQ_ENABLE);
+	r852_write_reg_dword(dev, R852_DMA_IRQ_ENABLE,
+		dma_irq_reg |
+		R852_DMA_IRQ_INTERNAL |
+		R852_DMA_IRQ_ERROR |
+		R852_DMA_IRQ_MEMORY);
+}
+
+/*
+ * Disable dma, called from the interrupt handler, which specifies
+ * success of the operation via 'error' argument
+ */
+static void r852_dma_done(struct r852_device *dev, int error)
+{
+	WARN_ON(dev->dma_stage == 0);
+
+	r852_write_reg_dword(dev, R852_DMA_IRQ_STA,
+			r852_read_reg_dword(dev, R852_DMA_IRQ_STA));
+
+	r852_write_reg_dword(dev, R852_DMA_SETTINGS, 0);
+	r852_write_reg_dword(dev, R852_DMA_IRQ_ENABLE, 0);
+
+	dev->dma_error = error;
+	dev->dma_stage = 0;
+
+	if (dev->phys_dma_addr && dev->phys_dma_addr != dev->phys_bounce_buffer)
+		pci_unmap_single(dev->pci_dev, dev->phys_dma_addr, R852_DMA_LEN,
+			dev->dma_dir ? PCI_DMA_FROMDEVICE : PCI_DMA_TODEVICE);
+	complete(&dev->dma_done);
+}
+
+/*
+ * Wait, till dma is done, which includes both phases of it
+ */
+static int r852_dma_wait(struct r852_device *dev)
+{
+	long timeout = wait_for_completion_timeout(&dev->dma_done,
+				msecs_to_jiffies(1000));
+	if (!timeout) {
+		dbg("timeout waiting for DMA interrupt");
+		return -ETIMEDOUT;
+	}
+
+	return 0;
+}
+
+/*
+ * Read/Write one page using dma. Only pages can be read (512 bytes)
+*/
+static void r852_do_dma(struct r852_device *dev, uint8_t *buf, int do_read)
+{
+	int bounce = 0;
+	unsigned long flags;
+	int error;
+
+	dev->dma_error = 0;
+
+	/* Set dma direction */
+	dev->dma_dir = do_read;
+	dev->dma_stage = 1;
+
+	dbg_verbose("doing dma %s ", do_read ? "read" : "write");
+
+	/* Set intial dma state: for reading first fill on board buffer,
+	  from device, for writes first fill the buffer  from memory*/
+	dev->dma_state = do_read ? DMA_INTERNAL : DMA_MEMORY;
+
+	/* if incoming buffer is not page aligned, we should do bounce */
+	if ((unsigned long)buf & (R852_DMA_LEN-1))
+		bounce = 1;
+
+	if (!bounce) {
+		dev->phys_dma_addr = pci_map_single(dev->pci_dev, (void *)buf,
+			R852_DMA_LEN,
+			(do_read ? PCI_DMA_FROMDEVICE : PCI_DMA_TODEVICE));
+
+		if (dev->phys_dma_addr == DMA_ERROR_CODE)
+			bounce = 1;
+	}
+
+	if (bounce) {
+		dbg_verbose("dma: using bounce buffer");
+		dev->phys_dma_addr = dev->phys_bounce_buffer;
+		if (!do_read)
+			memcpy(dev->bounce_buffer, buf, R852_DMA_LEN);
+	}
+
+	/* Enable DMA */
+	spin_lock_irqsave(&dev->irqlock, flags);
+	r852_dma_enable(dev);
+	spin_unlock_irqrestore(&dev->irqlock, flags);
+
+	/* Wait till complete */
+	error = r852_dma_wait(dev);
+
+	if (error) {
+		r852_dma_done(dev, error);
+		return;
+	}
+
+	if (do_read && bounce)
+		memcpy((void *)buf, dev->bounce_buffer, R852_DMA_LEN);
+}
+
+/*
+ * Program data lines of the nand chip to send data to it
+ */
+void r852_write_buf(struct mtd_info *mtd, const uint8_t *buf, int len)
+{
+	struct r852_device *dev = r852_get_dev(mtd);
+	u32 reg;
+
+	/* Don't allow any access to hardware if we suspect card removal */
+	if (dev->card_unstable)
+		return;
+
+	/* Special case for whole sector read */
+	if (len == R852_DMA_LEN && dev->dma_usable) {
+		r852_do_dma(dev, (uint8_t *)buf, 0);
+		return;
+	}
+
+	/* write DWORD chinks - faster */
+	while (len) {
+		reg = buf[0] | buf[1] << 8 | buf[2] << 16 | buf[3] << 24;
+		r852_write_reg_dword(dev, R852_DATALINE, reg);
+		buf += 4;
+		len -= 4;
+
+	}
+
+	/* write rest */
+	while (len)
+		r852_write_reg(dev, R852_DATALINE, *buf++);
+}
+
+/*
+ * Read data lines of the nand chip to retrieve data
+ */
+void r852_read_buf(struct mtd_info *mtd, uint8_t *buf, int len)
+{
+	struct r852_device *dev = r852_get_dev(mtd);
+	u32 reg;
+
+	if (dev->card_unstable) {
+		/* since we can't signal error here, at least, return
+			predictable buffer */
+		memset(buf, 0, len);
+		return;
+	}
+
+	/* special case for whole sector read */
+	if (len == R852_DMA_LEN && dev->dma_usable) {
+		r852_do_dma(dev, buf, 1);
+		return;
+	}
+
+	/* read in dword sized chunks */
+	while (len >= 4) {
+
+		reg = r852_read_reg_dword(dev, R852_DATALINE);
+		*buf++ = reg & 0xFF;
+		*buf++ = (reg >> 8) & 0xFF;
+		*buf++ = (reg >> 16) & 0xFF;
+		*buf++ = (reg >> 24) & 0xFF;
+		len -= 4;
+	}
+
+	/* read the reset by bytes */
+	while (len--)
+		*buf++ = r852_read_reg(dev, R852_DATALINE);
+}
+
+/*
+ * Read one byte from nand chip
+ */
+static uint8_t r852_read_byte(struct mtd_info *mtd)
+{
+	struct r852_device *dev = r852_get_dev(mtd);
+
+	/* Same problem as in r852_read_buf.... */
+	if (dev->card_unstable)
+		return 0;
+
+	return r852_read_reg(dev, R852_DATALINE);
+}
+
+
+/*
+ * Readback the buffer to verify it
+ */
+int r852_verify_buf(struct mtd_info *mtd, const uint8_t *buf, int len)
+{
+	struct r852_device *dev = r852_get_dev(mtd);
+
+	/* We can't be sure about anything here... */
+	if (dev->card_unstable)
+		return -1;
+
+	/* This will never happen, unless you wired up a nand chip
+		with > 512 bytes page size to the reader */
+	if (len > SM_SECTOR_SIZE)
+		return 0;
+
+	r852_read_buf(mtd, dev->tmp_buffer, len);
+	return memcmp(buf, dev->tmp_buffer, len);
+}
+
+/*
+ * Control several chip lines & send commands
+ */
+void r852_cmdctl(struct mtd_info *mtd, int dat, unsigned int ctrl)
+{
+	struct r852_device *dev = r852_get_dev(mtd);
+
+	if (dev->card_unstable)
+		return;
+
+	if (ctrl & NAND_CTRL_CHANGE) {
+
+		dev->ctlreg &= ~(R852_CTL_DATA | R852_CTL_COMMAND |
+				 R852_CTL_ON | R852_CTL_CARDENABLE);
+
+		if (ctrl & NAND_ALE)
+			dev->ctlreg |= R852_CTL_DATA;
+
+		if (ctrl & NAND_CLE)
+			dev->ctlreg |= R852_CTL_COMMAND;
+
+		if (ctrl & NAND_NCE)
+			dev->ctlreg |= (R852_CTL_CARDENABLE | R852_CTL_ON);
+		else
+			dev->ctlreg &= ~R852_CTL_WRITE;
+
+		/* when write is stareted, enable write access */
+		if (dat == NAND_CMD_ERASE1)
+			dev->ctlreg |= R852_CTL_WRITE;
+
+		r852_write_reg(dev, R852_CTL, dev->ctlreg);
+	}
+
+	 /* HACK: NAND_CMD_SEQIN is called without NAND_CTRL_CHANGE, but we need
+		to set write mode */
+	if (dat == NAND_CMD_SEQIN && (dev->ctlreg & R852_CTL_COMMAND)) {
+		dev->ctlreg |= R852_CTL_WRITE;
+		r852_write_reg(dev, R852_CTL, dev->ctlreg);
+	}
+
+	if (dat != NAND_CMD_NONE)
+		r852_write_reg(dev, R852_DATALINE, dat);
+}
+
+/*
+ * Wait till card is ready.
+ * based on nand_wait, but returns errors on DMA error
+ */
+int r852_wait(struct mtd_info *mtd, struct nand_chip *chip)
+{
+	struct r852_device *dev = (struct r852_device *)chip->priv;
+
+	unsigned long timeout;
+	int status;
+
+	timeout = jiffies + (chip->state == FL_ERASING ?
+		msecs_to_jiffies(400) : msecs_to_jiffies(20));
+
+	while (time_before(jiffies, timeout))
+		if (chip->dev_ready(mtd))
+			break;
+
+	chip->cmdfunc(mtd, NAND_CMD_STATUS, -1, -1);
+	status = (int)chip->read_byte(mtd);
+
+	/* Unfortunelly, no way to send detailed error status... */
+	if (dev->dma_error) {
+		status |= NAND_STATUS_FAIL;
+		dev->dma_error = 0;
+	}
+	return status;
+}
+
+/*
+ * Check if card is ready
+ */
+
+int r852_ready(struct mtd_info *mtd)
+{
+	struct r852_device *dev = r852_get_dev(mtd);
+	return !(r852_read_reg(dev, R852_CARD_STA) & R852_CARD_STA_BUSY);
+}
+
+
+/*
+ * Set ECC engine mode
+*/
+
+void r852_ecc_hwctl(struct mtd_info *mtd, int mode)
+{
+	struct r852_device *dev = r852_get_dev(mtd);
+
+	if (dev->card_unstable)
+		return;
+
+	switch (mode) {
+	case NAND_ECC_READ:
+	case NAND_ECC_WRITE:
+		/* enable ecc generation/check*/
+		dev->ctlreg |= R852_CTL_ECC_ENABLE;
+
+		/* flush ecc buffer */
+		r852_write_reg(dev, R852_CTL,
+			dev->ctlreg | R852_CTL_ECC_ACCESS);
+
+		r852_read_reg_dword(dev, R852_DATALINE);
+		r852_write_reg(dev, R852_CTL, dev->ctlreg);
+		return;
+
+	case NAND_ECC_READSYN:
+		/* disable ecc generation */
+		dev->ctlreg &= ~R852_CTL_ECC_ENABLE;
+		r852_write_reg(dev, R852_CTL, dev->ctlreg);
+	}
+}
+
+/*
+ * Calculate ECC, only used for writes
+ */
+
+int r852_ecc_calculate(struct mtd_info *mtd, const uint8_t *dat,
+							uint8_t *ecc_code)
+{
+	struct r852_device *dev = r852_get_dev(mtd);
+	struct sm_oob *oob = (struct sm_oob *)ecc_code;
+	u32 ecc1, ecc2;
+
+	if (dev->card_unstable)
+		return 0;
+
+	dev->ctlreg &= ~R852_CTL_ECC_ENABLE;
+	r852_write_reg(dev, R852_CTL, dev->ctlreg | R852_CTL_ECC_ACCESS);
+
+	ecc1 = r852_read_reg_dword(dev, R852_DATALINE);
+	ecc2 = r852_read_reg_dword(dev, R852_DATALINE);
+
+	oob->ecc1[0] = (ecc1) & 0xFF;
+	oob->ecc1[1] = (ecc1 >> 8) & 0xFF;
+	oob->ecc1[2] = (ecc1 >> 16) & 0xFF;
+
+	oob->ecc2[0] = (ecc2) & 0xFF;
+	oob->ecc2[1] = (ecc2 >> 8) & 0xFF;
+	oob->ecc2[2] = (ecc2 >> 16) & 0xFF;
+
+	r852_write_reg(dev, R852_CTL, dev->ctlreg);
+	return 0;
+}
+
+/*
+ * Correct the data using ECC, hw did almost everything for us
+ */
+
+int r852_ecc_correct(struct mtd_info *mtd, uint8_t *dat,
+				uint8_t *read_ecc, uint8_t *calc_ecc)
+{
+	u16 ecc_reg;
+	u8 ecc_status, err_byte;
+	int i, error = 0;
+
+	struct r852_device *dev = r852_get_dev(mtd);
+
+	if (dev->card_unstable)
+		return 0;
+
+	r852_write_reg(dev, R852_CTL, dev->ctlreg | R852_CTL_ECC_ACCESS);
+	ecc_reg = r852_read_reg_dword(dev, R852_DATALINE);
+	r852_write_reg(dev, R852_CTL, dev->ctlreg);
+
+	for (i = 0 ; i <= 1 ; i++) {
+
+		ecc_status = (ecc_reg >> 8) & 0xFF;
+
+		/* ecc uncorrectable error */
+		if (ecc_status & R852_ECC_FAIL) {
+			dbg("ecc: unrecoverable error, in half %d", i);
+			error = -1;
+			goto exit;
+		}
+
+		/* correctable error */
+		if (ecc_status & R852_ECC_CORRECTABLE) {
+
+			err_byte = ecc_reg & 0xFF;
+			dbg("ecc: recoverable error, "
+				"in half %d, byte %d, bit %d", i,
+				err_byte, ecc_status & R852_ECC_ERR_BIT_MSK);
+
+			dat[err_byte] ^=
+				1 << (ecc_status & R852_ECC_ERR_BIT_MSK);
+			error++;
+		}
+
+		dat += 256;
+		ecc_reg >>= 16;
+	}
+exit:
+	return error;
+}
+
+/*
+ * This is copy of nand_read_oob_std
+ * nand_read_oob_syndrome assumes we can send column address - we can't
+ */
+static int r852_read_oob(struct mtd_info *mtd, struct nand_chip *chip,
+			     int page, int sndcmd)
+{
+	if (sndcmd) {
+		chip->cmdfunc(mtd, NAND_CMD_READOOB, 0, page);
+		sndcmd = 0;
+	}
+	chip->read_buf(mtd, chip->oob_poi, mtd->oobsize);
+	return sndcmd;
+}
+
+/*
+ * Start the nand engine
+ */
+
+void r852_engine_enable(struct r852_device *dev)
+{
+	if (r852_read_reg_dword(dev, R852_HW) & R852_HW_UNKNOWN) {
+		r852_write_reg(dev, R852_CTL, R852_CTL_RESET | R852_CTL_ON);
+		r852_write_reg_dword(dev, R852_HW, R852_HW_ENABLED);
+	} else {
+		r852_write_reg_dword(dev, R852_HW, R852_HW_ENABLED);
+		r852_write_reg(dev, R852_CTL, R852_CTL_RESET | R852_CTL_ON);
+	}
+	msleep(300);
+	r852_write_reg(dev, R852_CTL, 0);
+}
+
+
+/*
+ * Stop the nand engine
+ */
+
+void r852_engine_disable(struct r852_device *dev)
+{
+	r852_write_reg_dword(dev, R852_HW, 0);
+	r852_write_reg(dev, R852_CTL, R852_CTL_RESET);
+}
+
+/*
+ * Test if card is present
+ */
+
+void r852_card_update_present(struct r852_device *dev)
+{
+	unsigned long flags;
+	u8 reg;
+
+	spin_lock_irqsave(&dev->irqlock, flags);
+	reg = r852_read_reg(dev, R852_CARD_STA);
+	dev->card_detected = !!(reg & R852_CARD_STA_PRESENT);
+	spin_unlock_irqrestore(&dev->irqlock, flags);
+}
+
+/*
+ * Update card detection IRQ state according to current card state
+ * which is read in r852_card_update_present
+ */
+void r852_update_card_detect(struct r852_device *dev)
+{
+	int card_detect_reg = r852_read_reg(dev, R852_CARD_IRQ_ENABLE);
+
+	card_detect_reg &= ~(R852_CARD_IRQ_REMOVE | R852_CARD_IRQ_INSERT);
+	card_detect_reg |= R852_CARD_IRQ_GENABLE;
+
+	card_detect_reg |= dev->card_detected ?
+		R852_CARD_IRQ_REMOVE : R852_CARD_IRQ_INSERT;
+
+	r852_write_reg(dev, R852_CARD_IRQ_ENABLE, card_detect_reg);
+}
+
+ssize_t r852_media_type_show(struct device *sys_dev,
+		struct device_attribute *attr, char *buf)
+{
+	struct mtd_info *mtd = container_of(sys_dev, struct mtd_info, dev);
+	struct r852_device *dev = r852_get_dev(mtd);
+	char *data = dev->sm ? "smartmedia" : "xd";
+
+	strcpy(buf, data);
+	return strlen(data);
+}
+
+DEVICE_ATTR(media_type, S_IRUGO, r852_media_type_show, NULL);
+
+
+/* Detect properties of card in slot */
+void r852_update_media_status(struct r852_device *dev)
+{
+	u8 reg;
+	unsigned long flags;
+	int readonly;
+
+	spin_lock_irqsave(&dev->irqlock, flags);
+	if (!dev->card_detected) {
+		message("card removed");
+		spin_unlock_irqrestore(&dev->irqlock, flags);
+		return ;
+	}
+
+	readonly  = r852_read_reg(dev, R852_CARD_STA) & R852_CARD_STA_RO;
+	reg = r852_read_reg(dev, R852_DMA_CAP);
+	dev->sm = (reg & (R852_DMA1 | R852_DMA2)) && (reg & R852_SMBIT);
+
+	message("detected %s %s card in slot",
+		dev->sm ? "SmartMedia" : "xD",
+		readonly ? "readonly" : "writeable");
+
+	dev->readonly = readonly;
+	spin_unlock_irqrestore(&dev->irqlock, flags);
+}
+
+/*
+ * Register the nand device
+ * Called when the card is detected
+ */
+int r852_register_nand_device(struct r852_device *dev)
+{
+	dev->mtd = kzalloc(sizeof(struct mtd_info), GFP_KERNEL);
+
+	if (!dev->mtd)
+		goto error1;
+
+	WARN_ON(dev->card_registred);
+
+	dev->mtd->owner = THIS_MODULE;
+	dev->mtd->priv = dev->chip;
+	dev->mtd->dev.parent = &dev->pci_dev->dev;
+
+	if (dev->readonly)
+		dev->chip->options |= NAND_ROM;
+
+	r852_engine_enable(dev);
+
+	if (sm_register_device(dev->mtd))
+		goto error2;
+
+	device_create_file(&dev->mtd->dev, &dev_attr_media_type);
+	dev->card_registred = 1;
+	return 0;
+error2:
+	kfree(dev->mtd);
+error1:
+	/* Force card redetect */
+	dev->card_detected = 0;
+	return -1;
+}
+
+/*
+ * Unregister the card
+ */
+
+void r852_unregister_nand_device(struct r852_device *dev)
+{
+	if (!dev->card_registred)
+		return;
+
+	device_remove_file(&dev->mtd->dev, &dev_attr_media_type);
+	nand_release(dev->mtd);
+	r852_engine_disable(dev);
+	dev->card_registred = 0;
+	kfree(dev->mtd);
+	dev->mtd = NULL;
+}
+
+/* Card state updater */
+void r852_card_detect_work(struct work_struct *work)
+{
+	struct r852_device *dev =
+		container_of(work, struct r852_device, card_detect_work.work);
+
+	r852_update_card_detect(dev);
+	dev->card_unstable = 0;
+
+	/* false alarm */
+	if (dev->card_detected == dev->card_registred)
+		goto exit;
+
+	/* Read media properties */
+	r852_update_media_status(dev);
+
+	/* Register the card */
+	if (dev->card_detected)
+		r852_register_nand_device(dev);
+	else
+		r852_unregister_nand_device(dev);
+exit:
+	/* Update detection logic */
+	r852_update_card_detect(dev);
+}
+
+/* Ack + disable IRQ generation */
+static void r852_disable_irqs(struct r852_device *dev)
+{
+	u8 reg;
+	reg = r852_read_reg(dev, R852_CARD_IRQ_ENABLE);
+	r852_write_reg(dev, R852_CARD_IRQ_ENABLE, reg & ~R852_CARD_IRQ_MASK);
+
+	reg = r852_read_reg_dword(dev, R852_DMA_IRQ_ENABLE);
+	r852_write_reg_dword(dev, R852_DMA_IRQ_ENABLE,
+					reg & ~R852_DMA_IRQ_MASK);
+
+	r852_write_reg(dev, R852_CARD_IRQ_STA, R852_CARD_IRQ_MASK);
+	r852_write_reg_dword(dev, R852_DMA_IRQ_STA, R852_DMA_IRQ_MASK);
+}
+
+/* Interrupt handler */
+static irqreturn_t r852_irq(int irq, void *data)
+{
+	struct r852_device *dev = (struct r852_device *)data;
+
+	u8 card_status, dma_status;
+	unsigned long flags;
+	irqreturn_t ret = IRQ_NONE;
+
+	spin_lock_irqsave(&dev->irqlock, flags);
+
+	/* We can recieve shared interrupt while pci is suspended
+		in that case reads will return 0xFFFFFFFF.... */
+	if (dev->insuspend)
+		goto out;
+
+	/* handle card detection interrupts first */
+	card_status = r852_read_reg(dev, R852_CARD_IRQ_STA);
+	r852_write_reg(dev, R852_CARD_IRQ_STA, card_status);
+
+	if (card_status & (R852_CARD_IRQ_INSERT|R852_CARD_IRQ_REMOVE)) {
+
+		ret = IRQ_HANDLED;
+		dev->card_detected = !!(card_status & R852_CARD_IRQ_INSERT);
+
+		/* we shouldn't recieve any interrupts if we wait for card
+			to settle */
+		WARN_ON(dev->card_unstable);
+
+		/* disable irqs while card is unstable */
+		/* this will timeout DMA if active, but better that garbage */
+		r852_disable_irqs(dev);
+
+		if (dev->card_unstable)
+			goto out;
+
+		/* let, card state to settle a bit, and then do the work */
+		dev->card_unstable = 1;
+		queue_delayed_work(dev->card_workqueue,
+			&dev->card_detect_work, msecs_to_jiffies(100));
+		goto out;
+	}
+
+
+	/* Handle dma interrupts */
+	dma_status = r852_read_reg_dword(dev, R852_DMA_IRQ_STA);
+	r852_write_reg_dword(dev, R852_DMA_IRQ_STA, dma_status);
+
+	if (dma_status & R852_DMA_IRQ_MASK) {
+
+		ret = IRQ_HANDLED;
+
+		if (dma_status & R852_DMA_IRQ_ERROR) {
+			dbg("recieved dma error IRQ");
+			r852_dma_done(dev, -EIO);
+			goto out;
+		}
+
+		/* recieved DMA interrupt out of nowhere? */
+		WARN_ON_ONCE(dev->dma_stage == 0);
+
+		if (dev->dma_stage == 0)
+			goto out;
+
+		/* done device access */
+		if (dev->dma_state == DMA_INTERNAL &&
+				(dma_status & R852_DMA_IRQ_INTERNAL)) {
+
+			dev->dma_state = DMA_MEMORY;
+			dev->dma_stage++;
+		}
+
+		/* done memory DMA */
+		if (dev->dma_state == DMA_MEMORY &&
+				(dma_status & R852_DMA_IRQ_MEMORY)) {
+			dev->dma_state = DMA_INTERNAL;
+			dev->dma_stage++;
+		}
+
+		/* Enable 2nd half of dma dance */
+		if (dev->dma_stage == 2)
+			r852_dma_enable(dev);
+
+		/* Operation done */
+		if (dev->dma_stage == 3)
+			r852_dma_done(dev, 0);
+		goto out;
+	}
+
+	/* Handle unknown interrupts */
+	if (dma_status)
+		dbg("bad dma IRQ status = %x", dma_status);
+
+	if (card_status & ~R852_CARD_STA_CD)
+		dbg("strange card status = %x", card_status);
+
+out:
+	spin_unlock_irqrestore(&dev->irqlock, flags);
+	return ret;
+}
+
+int  r852_probe(struct pci_dev *pci_dev, const struct pci_device_id *id)
+{
+	int error;
+	struct nand_chip *chip;
+	struct r852_device *dev;
+
+	/* pci initialization */
+	error = pci_enable_device(pci_dev);
+
+	if (error)
+		goto error1;
+
+	pci_set_master(pci_dev);
+
+	error = pci_set_dma_mask(pci_dev, DMA_32BIT_MASK);
+	if (error)
+		goto error2;
+
+	error = pci_request_regions(pci_dev, DRV_NAME);
+
+	if (error)
+		goto error3;
+
+	error = -ENOMEM;
+
+	/* init nand chip, but register it only on card insert */
+	chip = kzalloc(sizeof(struct nand_chip), GFP_KERNEL);
+
+	if (!chip)
+		goto error4;
+
+	/* commands */
+	chip->cmd_ctrl = r852_cmdctl;
+	chip->waitfunc = r852_wait;
+	chip->dev_ready = r852_ready;
+
+	/* I/O */
+	chip->read_byte = r852_read_byte;
+	chip->read_buf = r852_read_buf;
+	chip->write_buf = r852_write_buf;
+	chip->verify_buf = r852_verify_buf;
+
+	/* ecc */
+	chip->ecc.mode = NAND_ECC_HW_SYNDROME;
+	chip->ecc.size = R852_DMA_LEN;
+	chip->ecc.bytes = SM_OOB_SIZE;
+	chip->ecc.hwctl = r852_ecc_hwctl;
+	chip->ecc.calculate = r852_ecc_calculate;
+	chip->ecc.correct = r852_ecc_correct;
+
+	/* TODO: hack */
+	chip->ecc.read_oob = r852_read_oob;
+
+	/* init our device structure */
+	dev = kzalloc(sizeof(struct r852_device), GFP_KERNEL);
+
+	if (!dev)
+		goto error5;
+
+	chip->priv = dev;
+	dev->chip = chip;
+	dev->pci_dev = pci_dev;
+	pci_set_drvdata(pci_dev, dev);
+
+	dev->bounce_buffer = pci_alloc_consistent(pci_dev, R852_DMA_LEN,
+		&dev->phys_bounce_buffer);
+
+	if (!dev->bounce_buffer)
+		goto error6;
+
+
+	error = -ENODEV;
+	dev->mmio = pci_ioremap_bar(pci_dev, 0);
+
+	if (!dev->mmio)
+		goto error7;
+
+	error = -ENOMEM;
+	dev->tmp_buffer = kzalloc(SM_SECTOR_SIZE, GFP_KERNEL);
+
+	if (!dev->tmp_buffer)
+		goto error8;
+
+	init_completion(&dev->dma_done);
+
+	dev->card_workqueue = create_freezeable_workqueue(DRV_NAME);
+
+	if (!dev->card_workqueue)
+		goto error9;
+
+	INIT_DELAYED_WORK(&dev->card_detect_work, r852_card_detect_work);
+
+	/* shutdown everything - precation */
+	r852_engine_disable(dev);
+	r852_disable_irqs(dev);
+
+	r852_dma_test(dev);
+
+	/*register irq handler*/
+	error = -ENODEV;
+	if (request_irq(pci_dev->irq, &r852_irq, IRQF_SHARED,
+			  DRV_NAME, dev))
+		goto error10;
+
+	dev->irq = pci_dev->irq;
+	spin_lock_init(&dev->irqlock);
+
+	/* kick initial present test */
+	dev->card_detected = 0;
+	r852_card_update_present(dev);
+	queue_delayed_work(dev->card_workqueue,
+		&dev->card_detect_work, 0);
+
+
+	printk(KERN_NOTICE DRV_NAME ": driver loaded succesfully\n");
+	return 0;
+
+error10:
+	destroy_workqueue(dev->card_workqueue);
+error9:
+	kfree(dev->tmp_buffer);
+error8:
+	pci_iounmap(pci_dev, dev->mmio);
+error7:
+	pci_free_consistent(pci_dev, R852_DMA_LEN,
+		dev->bounce_buffer, dev->phys_bounce_buffer);
+error6:
+	kfree(dev);
+error5:
+	kfree(chip);
+error4:
+	pci_release_regions(pci_dev);
+error3:
+error2:
+	pci_disable_device(pci_dev);
+error1:
+	return error;
+}
+
+void r852_remove(struct pci_dev *pci_dev)
+{
+	struct r852_device *dev = pci_get_drvdata(pci_dev);
+
+	/* Stop detect workqueue -
+		we are going to unregister the device anyway*/
+	cancel_delayed_work_sync(&dev->card_detect_work);
+	destroy_workqueue(dev->card_workqueue);
+
+	/* Unregister the device, this might make more IO */
+	r852_unregister_nand_device(dev);
+
+	/* Stop interrupts */
+	r852_disable_irqs(dev);
+	synchronize_irq(dev->irq);
+	free_irq(dev->irq, dev);
+
+	/* Cleanup */
+	kfree(dev->tmp_buffer);
+	pci_iounmap(pci_dev, dev->mmio);
+	pci_free_consistent(pci_dev, R852_DMA_LEN,
+		dev->bounce_buffer, dev->phys_bounce_buffer);
+
+	kfree(dev->chip);
+	kfree(dev);
+
+	/* Shutdown the PCI device */
+	pci_release_regions(pci_dev);
+	pci_disable_device(pci_dev);
+}
+
+void r852_shutdown(struct pci_dev *pci_dev)
+{
+	struct r852_device *dev = pci_get_drvdata(pci_dev);
+
+	cancel_delayed_work_sync(&dev->card_detect_work);
+	r852_disable_irqs(dev);
+	synchronize_irq(dev->irq);
+	pci_disable_device(pci_dev);
+}
+
+int r852_suspend(struct device *device)
+{
+	struct r852_device *dev = pci_get_drvdata(to_pci_dev(device));
+	unsigned long flags;
+
+	if (dev->ctlreg & R852_CTL_CARDENABLE)
+		return -EBUSY;
+
+	/* First make sure the detect work is gone */
+	cancel_delayed_work_sync(&dev->card_detect_work);
+
+	/* Turn off the interrupts and stop the device */
+	r852_disable_irqs(dev);
+	r852_engine_disable(dev);
+
+	spin_lock_irqsave(&dev->irqlock, flags);
+	dev->insuspend = 1;
+	spin_unlock_irqrestore(&dev->irqlock, flags);
+
+	/* At that point, even if interrupt handler is running, it will quit */
+	/* So wait for this to happen explictly */
+	synchronize_irq(dev->irq);
+
+	/* If card was pulled off just during the suspend, which is very
+		unlikely, we will remove it on resume, it too late now
+		anyway... */
+	dev->card_unstable = 0;
+
+	pci_save_state(to_pci_dev(device));
+	return pci_prepare_to_sleep(to_pci_dev(device));
+}
+
+int r852_resume(struct device *device)
+{
+	struct r852_device *dev = pci_get_drvdata(to_pci_dev(device));
+	unsigned long flags;
+
+	/* Turn on the hardware */
+	pci_back_from_sleep(to_pci_dev(device));
+	pci_restore_state(to_pci_dev(device));
+
+	r852_disable_irqs(dev);
+	r852_card_update_present(dev);
+	r852_engine_disable(dev);
+
+
+	/* Now its safe for IRQ to run */
+	spin_lock_irqsave(&dev->irqlock, flags);
+	dev->insuspend = 0;
+	spin_unlock_irqrestore(&dev->irqlock, flags);
+
+
+	/* If card status changed, just do the work */
+	if (dev->card_detected != dev->card_registred) {
+		dbg("card was %s during low power state",
+			dev->card_detected ? "added" : "removed");
+
+		queue_delayed_work(dev->card_workqueue,
+		&dev->card_detect_work, 1000);
+		return 0;
+	}
+
+	/* Otherwise, initialize the card */
+	if (dev->card_registred) {
+		r852_engine_enable(dev);
+		dev->chip->select_chip(dev->mtd, 0);
+		dev->chip->cmdfunc(dev->mtd, NAND_CMD_RESET, -1, -1);
+		dev->chip->select_chip(dev->mtd, -1);
+	}
+
+	/* Program card detection IRQ */
+	r852_update_card_detect(dev);
+	return 0;
+}
+
+static const struct pci_device_id r852_pci_id_tbl[] = {
+
+	{ PCI_VDEVICE(RICOH, PCI_DEVICE_ID_RICOH_R5C852), },
+	{ },
+};
+
+MODULE_DEVICE_TABLE(pci, r852_pci_id_tbl);
+
+SIMPLE_DEV_PM_OPS(r852_pm_ops, r852_suspend, r852_resume);
+
+
+static struct pci_driver r852_pci_driver = {
+	.name		= DRV_NAME,
+	.id_table	= r852_pci_id_tbl,
+	.probe		= r852_probe,
+	.remove		= r852_remove,
+	.shutdown	= r852_shutdown,
+	.driver.pm	= &r852_pm_ops,
+};
+
+static __init int r852_module_init(void)
+{
+	return pci_register_driver(&r852_pci_driver);
+}
+
+static void __exit r852_module_exit(void)
+{
+	pci_unregister_driver(&r852_pci_driver);
+}
+
+module_init(r852_module_init);
+module_exit(r852_module_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Maxim Levitsky <maximlevitsky@gmail.com>");
+MODULE_DESCRIPTION("Ricoh 85xx xD/smartmedia card reader driver");
diff --git a/drivers/mtd/nand/r852.h b/drivers/mtd/nand/r852.h
new file mode 100644
index 0000000..e05e795
--- /dev/null
+++ b/drivers/mtd/nand/r852.h
@@ -0,0 +1,163 @@
+/*
+ * Copyright (C) 2009 - Maxim Levitsky
+ * driver for Ricoh xD readers
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/pci.h>
+#include <linux/completion.h>
+#include <linux/workqueue.h>
+#include <linux/mtd/nand.h>
+#include <linux/spinlock.h>
+
+
+/* nand interface + ecc
+   byte write/read does one cycle on nand data lines.
+   dword write/read does 4 cycles
+   if R852_CTL_ECC_ACCESS is set in R852_CTL, then dword read reads
+   results of ecc correction, if DMA read was done before.
+   If write was done two dword reads read generated ecc checksums
+*/
+#define	R852_DATALINE		0x00
+
+/* control register */
+#define R852_CTL		0x04
+#define R852_CTL_COMMAND 	0x01	/* send command (#CLE)*/
+#define R852_CTL_DATA		0x02	/* read/write data (#ALE)*/
+#define R852_CTL_ON		0x04	/* only seem to controls the hd led, */
+					/* but has to be set on start...*/
+#define R852_CTL_RESET		0x08	/* unknown, set only on start once*/
+#define R852_CTL_CARDENABLE	0x10	/* probably (#CE) - always set*/
+#define R852_CTL_ECC_ENABLE	0x20	/* enable ecc engine */
+#define R852_CTL_ECC_ACCESS	0x40	/* read/write ecc via reg #0*/
+#define R852_CTL_WRITE		0x80	/* set when performing writes (#WP) */
+
+/* card detection status */
+#define R852_CARD_STA		0x05
+
+#define R852_CARD_STA_CD	0x01	/* state of #CD line, same as 0x04 */
+#define R852_CARD_STA_RO	0x02	/* card is readonly */
+#define R852_CARD_STA_PRESENT	0x04	/* card is present (#CD) */
+#define R852_CARD_STA_ABSENT	0x08	/* card is absent */
+#define R852_CARD_STA_BUSY	0x80	/* card is busy - (#R/B) */
+
+/* card detection irq status & enable*/
+#define R852_CARD_IRQ_STA	0x06	/* IRQ status */
+#define R852_CARD_IRQ_ENABLE	0x07	/* IRQ enable */
+
+#define R852_CARD_IRQ_CD	0x01	/* fire when #CD lights, same as 0x04*/
+#define R852_CARD_IRQ_REMOVE	0x04	/* detect card removal */
+#define R852_CARD_IRQ_INSERT	0x08	/* detect card insert */
+#define R852_CARD_IRQ_UNK1	0x10	/* unknown */
+#define R852_CARD_IRQ_GENABLE	0x80	/* general enable */
+#define R852_CARD_IRQ_MASK	0x1D
+
+
+
+/* hardware enable */
+#define R852_HW			0x08
+#define R852_HW_ENABLED		0x01	/* hw enabled */
+#define R852_HW_UNKNOWN		0x80
+
+
+/* dma capabilities */
+#define R852_DMA_CAP		0x09
+#define R852_SMBIT		0x20	/* if set with bit #6 or bit #7, then */
+					/* hw is smartmedia */
+#define R852_DMA1		0x40	/* if set w/bit #7, dma is supported */
+#define R852_DMA2		0x80	/* if set w/bit #6, dma is supported */
+
+
+/* physical DMA address - 32 bit value*/
+#define R852_DMA_ADDR		0x0C
+
+
+/* dma settings */
+#define R852_DMA_SETTINGS	0x10
+#define R852_DMA_MEMORY		0x01	/* (memory <-> internal hw buffer) */
+#define R852_DMA_READ		0x02	/* 0 = write, 1 = read */
+#define R852_DMA_INTERNAL	0x04	/* (internal hw buffer <-> card) */
+
+/* dma IRQ status */
+#define R852_DMA_IRQ_STA		0x14
+
+/* dma IRQ enable */
+#define R852_DMA_IRQ_ENABLE	0x18
+
+#define R852_DMA_IRQ_MEMORY	0x01	/* (memory <-> internal hw buffer) */
+#define R852_DMA_IRQ_ERROR	0x02	/* error did happen */
+#define R852_DMA_IRQ_INTERNAL	0x04	/* (internal hw buffer <-> card) */
+#define R852_DMA_IRQ_MASK	0x07	/* mask of all IRQ bits */
+
+
+/* ECC syndrome format - read from reg #0 will return two copies of these for
+   each half of the page.
+   first byte is error byte location, and second, bit location + flags */
+#define R852_ECC_ERR_BIT_MSK	0x07	/* error bit location */
+#define R852_ECC_CORRECT		0x10	/* no errors - (guessed) */
+#define R852_ECC_CORRECTABLE	0x20	/* correctable error exist */
+#define R852_ECC_FAIL		0x40	/* non correctable error detected */
+
+#define R852_DMA_LEN		512
+
+#define DMA_INTERNAL	0
+#define DMA_MEMORY	1
+
+struct r852_device {
+	void __iomem *mmio;		/* mmio */
+	struct mtd_info *mtd;		/* mtd backpointer */
+	struct nand_chip *chip;		/* nand chip backpointer */
+	struct pci_dev *pci_dev;	/* pci backpointer */
+
+	/* dma area */
+	dma_addr_t phys_dma_addr;	/* bus address of buffer*/
+	struct completion dma_done;	/* data transfer done */
+
+	dma_addr_t phys_bounce_buffer;	/* bus address of bounce buffer */
+	u8 *bounce_buffer;		/* virtual address of bounce buffer */
+
+	int dma_dir;			/* 1 = read, 0 = write */
+	int dma_stage;			/* 0 - idle, 1 - first step,
+					   2 - second step */
+
+	int dma_state;			/* 0 = internal, 1 = memory */
+	int dma_error;			/* dma errors */
+	int dma_usable;			/* is it possible to use dma */
+
+	/* card status area */
+	struct delayed_work card_detect_work;
+	struct workqueue_struct *card_workqueue;
+	int card_registred;		/* card registered with mtd */
+	int card_detected;		/* card detected in slot */
+	int card_unstable;		/* whenever the card is inserted,
+					   is not known yet */
+	int readonly;			/* card is readonly */
+	int sm;				/* Is card smartmedia */
+
+	/* interrupt handling */
+	spinlock_t irqlock;		/* IRQ protecting lock */
+	int irq;			/* irq num */
+	int insuspend;			/* device is suspended */
+
+	/* misc */
+	void *tmp_buffer;		/* temporary buffer */
+	u8 ctlreg;			/* cached contents of control reg */
+};
+
+#define DRV_NAME "r852"
+
+
+#define dbg(format, ...) \
+	if (debug) \
+		printk(KERN_DEBUG DRV_NAME ": " format "\n", ## __VA_ARGS__)
+
+#define dbg_verbose(format, ...) \
+	if (debug > 1) \
+		printk(KERN_DEBUG DRV_NAME ": " format "\n", ## __VA_ARGS__)
+
+
+#define message(format, ...) \
+	printk(KERN_INFO DRV_NAME ": " format "\n", ## __VA_ARGS__)
diff --git a/include/linux/pci_ids.h b/include/linux/pci_ids.h
index cca8a04..cef49f2 100644
--- a/include/linux/pci_ids.h
+++ b/include/linux/pci_ids.h
@@ -1523,6 +1523,8 @@
 #define PCI_DEVICE_ID_RICOH_R5C822	0x0822
 #define PCI_DEVICE_ID_RICOH_R5C832	0x0832
 #define PCI_DEVICE_ID_RICOH_R5C843	0x0843
+#define PCI_DEVICE_ID_RICOH_R5C852	0x0852
+
 
 #define PCI_VENDOR_ID_DLINK		0x1186
 #define PCI_DEVICE_ID_DLINK_DGE510T	0x4c00
-- 
1.6.3.3