Message-Id: <1314202540-5562-1-git-send-email-david.wagner@free-electrons.com>
Date:	Wed, 24 Aug 2011 18:15:40 +0200
From:	david.wagner@...e-electrons.com
To:	linux-mtd <linux-mtd@...ts.infradead.org>
Cc:	Artem Bityutskiy <dedekind1@...il.com>,
	linux-embedded <linux-embedded@...r.kernel.org>,
	lkml <linux-kernel@...r.kernel.org>,
	Tim Bird <tim.bird@...sony.com>,
	David Woodhouse <dwmw2@...radead.org>,
	David Wagner <david.wagner@...e-electrons.com>
Subject: [PATCHv4] UBI: new module ubiblk: block layer on top of UBI

From: David Wagner <david.wagner@...e-electrons.com>

ubiblk is a read-only block layer on top of UBI.  It presents UBI volumes as
read-only block devices (named ubiblk_X_Y, where X is the UBI device number
and Y the Volume ID).

It is used by putting a block filesystem image on a UBI volume, creating the
corresponding ubiblk device and then mounting it.

It uses the UBI API to register to UBI notifications and to read from the
volumes.  It also creates a ubiblk_ctrl device node that simply receives ioctl
from a userspace tool for creating/removing ubiblk devices.

Some code is taken from mtd_blkdevs and gluebi.  Some code for the ioctl part is
also inspired from ubi's core.

Advantages of ubiblk over gluebi+mtdblock_ro:

 * Simpler architecture

 * The numbering of devices is much easier with ubiblk than with
   gluebi+mtdblock_ro. With gluebi+mtdblock_ro, you get one additional MTD
   device for each UBI volume, so the number of MTD devices grows quite a lot
   and is a bit difficult to understand. For example, mtdblock[0-4] might be
   your real MTD partitions, while mtdblock[5-9] might be your UBI volumes.
   It also means that if a new real MTD partition is added, the number of all
   the MTD devices exposing UBI volumes will be incremented by one, which is a
   bit confusing/annoying.
   As well, if you add an UBI volume, the mtdblock devices that are emulated on
   top of volumes that come after this new one will have their ID incremented.

 * ubiblk devices are created on a 'on-demand' basis, which avoids wasting
   resources.

 * The performance appears to be slightly better with ubiblk than
   gluebi+mtdblock_ro, according to our benchmarks (see
   http://elinux.org/Flash_Filesystem_Benchmarks_2.6.39)

TODO:
 * the modules keeps a table of the devices which length is the maximum number
   of UBI volumes.  There should be a better solution (linked list or, as
   Christoph Hellwig suggests, a radix tree (idr)).

Signed-off-by: David Wagner <david.wagner@...e-electrons.com>
---

	Hi,

changes since v3
~~~~~~~~~~~~~~~~

 a renamed *thread to *req_task
 b kerneldoc comments
 c fix ubiblk-user.h integration in Kbuild
 d use dev_* instead of pr_*
 e fix a bug: it was possible to remove a opened device
 f use linked lists instead of static table
   even if idr provides an address space for minor numbers, we found no way to
   use that feature: when we need to get the device from the idr structure, we
   only have a ubi_volume_info at hand, so we don't know the minor number.  A
   solution might be, instead of getting a minor number from idr, to decide of
   one (for instance, ubi_num * UBI_MAX_VOLUMES + vol_id, but if they were more
   than one UBI device, that would make the idr structure grow needlessly.
   Plus, there are few chances that the number of ubiblk devices be high, so
   it's probably not worth the pain.
   However, if we have time, we may be interessed to see how it could benefit
   UBI.
 g checkpatch'd (it still complains about an assignment in an if ()), is it
   really bad ?


about e), I have a question: it was possible to unmount a mounted filesystem
because ot this.  The fix we found was to return -EBUSY if the ubiblk device is
opened. But is it correct to assume that a mounted filesystem will always hold
the device opened ?  If not, what is the correct way to impeach the removal of a
ubiblk device when it is "mounted" ?
(see ubiblk_remove(), if (dev->desc))

	other questions:
is it ok not to return the error value from the notify function ? I found now
way to return an error value w/o stopping the notifications from being emitted.

there are error messages printed when a ubiblk_remove() is attempted on a
non-existing device - that happens a lot when a UBI device is detached ; should
the messages be debug-only (or even removed) ?


	Best regards,
	David.

 Documentation/ioctl/ioctl-number.txt |    1 +
 drivers/mtd/ubi/Kconfig              |   16 +
 drivers/mtd/ubi/Makefile             |    1 +
 drivers/mtd/ubi/ubiblk.c             |  697 ++++++++++++++++++++++++++++++++++
 include/mtd/Kbuild                   |    1 +
 include/mtd/ubiblk-user.h            |   47 +++
 6 files changed, 763 insertions(+), 0 deletions(-)
 create mode 100644 drivers/mtd/ubi/ubiblk.c
 create mode 100644 include/mtd/ubiblk-user.h

diff --git a/Documentation/ioctl/ioctl-number.txt b/Documentation/ioctl/ioctl-number.txt
index 845a191..b24df7f 100644
--- a/Documentation/ioctl/ioctl-number.txt
+++ b/Documentation/ioctl/ioctl-number.txt
@@ -150,6 +150,7 @@ Code  Seq#(hex)	Include File		Comments
 'M'	00-0F	drivers/video/fsl-diu-fb.h	conflict!
 'N'	00-1F	drivers/usb/scanner.h
 'O'     00-06   mtd/ubi-user.h		UBI
+'O'     10-11   mtd/ubiblk-user.h       ubiblk
 'P'	all	linux/soundcard.h	conflict!
 'P'	60-6F	sound/sscape_ioctl.h	conflict!
 'P'	00-0F	drivers/usb/class/usblp.c	conflict!
diff --git a/drivers/mtd/ubi/Kconfig b/drivers/mtd/ubi/Kconfig
index 4dcc752..977934a 100644
--- a/drivers/mtd/ubi/Kconfig
+++ b/drivers/mtd/ubi/Kconfig
@@ -60,4 +60,20 @@ config MTD_UBI_DEBUG
 	help
 	  This option enables UBI debugging.
 
+config MTD_UBI_UBIBLK
+	tristate "Read-only block transition layer on top of UBI"
+	help
+	   Read-only block interface on top of UBI.
+
+	   This option adds ubiblk, which creates a read-ony block device from
+	   UBI volumes.  It makes it possible to use R/O block filesystems on
+	   top of UBI volumes (and hence, on top of MTDs while avoiding bad
+	   blocks).
+
+	   ubiblk devices are created by invoking appropriate ioctl to the
+	   ubiblk_ctrl device node.
+
+	   The devices are named ubiblkX_Y where X is the UBI number and Y is
+	   the Volume ID.
+
 endif # MTD_UBI
diff --git a/drivers/mtd/ubi/Makefile b/drivers/mtd/ubi/Makefile
index c9302a5..354b2df 100644
--- a/drivers/mtd/ubi/Makefile
+++ b/drivers/mtd/ubi/Makefile
@@ -5,3 +5,4 @@ ubi-y += misc.o
 
 ubi-$(CONFIG_MTD_UBI_DEBUG) += debug.o
 obj-$(CONFIG_MTD_UBI_GLUEBI) += gluebi.o
+obj-$(CONFIG_MTD_UBI_UBIBLK) += ubiblk.o
diff --git a/drivers/mtd/ubi/ubiblk.c b/drivers/mtd/ubi/ubiblk.c
new file mode 100644
index 0000000..6758f00
--- /dev/null
+++ b/drivers/mtd/ubi/ubiblk.c
@@ -0,0 +1,697 @@
+/*
+ * Copyright (c) Free Electrons, 2011
+ * Copyright (c) International Business Machines Corp., 2006
+ * Copyright © 2003-2010 David Woodhouse <dwmw2@...radead.org>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See
+ * the GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ *
+ * Author: David Wagner
+ * Some code taken from gluebi.c
+ *	(Artem Bityutskiy (Битюцкий Артём), Joern Engel)
+ * Some code taken from cdev.c (Artem Bityutskiy (Битюцкий Артём))
+ * Some code taken from mtd_blkdevs.c (David Woodhouse)
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/vmalloc.h>
+#include <linux/mtd/ubi.h>
+#include <linux/blkdev.h>
+#include <linux/miscdevice.h>
+#include <linux/kthread.h>
+#include <linux/mutex.h>
+#include <linux/list.h>
+#include <mtd/ubiblk-user.h>
+#include "ubi.h"
+
+#define BLK_SIZE 512
+
+#define UBIBLK_MAX_DEVS (UBI_MAX_DEVICES * UBI_MAX_VOLUMES)
+
+/**
+ * struct ubiblk_dev - represents a ubiblk device, proxying a UBI volume
+ * @desc: open UBI volume descriptor
+ * @vi: UBI volume information
+ * @ubi_num: UBI device number
+ * @vol_id: UBI volume identifier
+ * @gd: the disk (block device) created by ubiblk
+ * @rq: the request queue to @gd
+ * @req_task: the thread processing @rq requests
+ * @vol_lock: protects write access to the elements of this structure
+ * @queue_lock: avoids concurrent accesses to the request queue
+ * @list: linked list structure
+ */
+struct ubiblk_dev {
+	struct ubi_volume_desc *desc;
+	struct ubi_volume_info *vi;
+	int ubi_num;
+	int vol_id;
+
+	struct gendisk *gd;
+	struct request_queue *rq;
+	struct task_struct *req_task;
+
+	struct mutex vol_lock;
+
+	spinlock_t queue_lock;
+
+	struct list_head list;
+};
+
+/* Linked list of all ubiblk_dev instances */
+static LIST_HEAD(ubiblk_devs);
+
+/* Avoid concurrent access to the above list */
+static struct mutex devlist_lock;
+
+static int ubiblk_major;
+static const struct block_device_operations ubiblk_ops;
+
+/* The device receiving the ioctls */
+static struct miscdevice ctrl_dev;
+
+static struct ubiblk_dev *ubiblk_find_dev(struct ubi_volume_info *vi)
+{
+	struct ubiblk_dev *dev;
+	struct list_head *list_ptr;
+
+	/* TODO: use list_for_each_entry ? */
+	list_for_each(list_ptr, &ubiblk_devs) {
+		dev = list_entry(list_ptr, struct ubiblk_dev, list);
+		if (dev && dev->ubi_num == vi->ubi_num &&
+		    dev->vol_id == vi->vol_id)
+			break;
+		dev = NULL;
+	}
+	return dev;
+}
+
+/**
+ * do_ubiblk_request - Read a LEB and fill the request buffer with the
+ * requested sector
+ *
+ * @req: the request data structure
+ * @dev: the ubiblk device on which the request is issued
+ */
+static int do_ubiblk_request(struct request *req, struct ubiblk_dev *dev)
+{
+	unsigned long start, len, read_bytes;
+	int offset;
+	int leb;
+	int ret;
+
+	start = blk_rq_pos(req) << 9;
+	len = blk_rq_cur_bytes(req);
+	read_bytes = 0;
+
+	/* We are always reading. No need to handle writing for now */
+
+	leb = start / dev->vi->usable_leb_size;
+	offset = start % dev->vi->usable_leb_size;
+
+	do {
+		if (offset + len > dev->vi->usable_leb_size)
+			len = dev->vi->usable_leb_size - offset;
+
+		if (unlikely(blk_rq_pos(req) + blk_rq_cur_sectors(req) >
+		    get_capacity(req->rq_disk))) {
+			dev_err(disk_to_dev(dev->gd),
+				"attempting to read too far\n");
+			return -EIO;
+		}
+
+		/* Read (len) bytes of LEB (leb) from (offset) and put the
+		 * result in the buffer given by the request.
+		 * If the request is overlapping on several lebs, (read_bytes)
+		 * will be > 0 and the data will be put in the buffer at
+		 * offset (read_bytes)
+		 */
+		ret = ubi_read(dev->desc, leb, req->buffer + read_bytes,
+			       offset, len);
+
+		if (ret) {
+			dev_err(disk_to_dev(dev->gd), "ubi_read error\n");
+			return ret;
+		}
+
+		read_bytes += len;
+
+		len = blk_rq_cur_bytes(req) - read_bytes;
+		leb++;
+		offset = 0;
+	} while (read_bytes < blk_rq_cur_bytes(req));
+
+	return 0;
+}
+
+/**
+ * ubi_ubiblk_request - wakes the processing thread
+ *
+ * @rq: the request queue which device is to be awaken
+ */
+static void ubi_ubiblk_request(struct request_queue *rq)
+{
+	struct ubiblk_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->req_task);
+}
+
+/**
+ * ubiblk_open - open a UBI volume (get the volume descriptor)
+ *
+ * @bdev: the corresponding block device
+ * @mode: opening mode (don't care as long as ubiblk is read-only)
+ */
+static int ubiblk_open(struct block_device *bdev, fmode_t mode)
+{
+	struct ubiblk_dev *dev = bdev->bd_disk->private_data;
+	int err = 0;
+
+	dev_dbg(disk_to_dev(dev->gd), "opened mode=%d\n", mode);
+
+	dev->desc = ubi_open_volume(dev->ubi_num, dev->vol_id,
+					UBI_READONLY);
+	if (IS_ERR(dev->desc)) {
+		dev_err(disk_to_dev(dev->gd), "failed to open");
+
+		err = PTR_ERR(dev->desc);
+		dev->desc = NULL;
+		goto out_err;
+	}
+
+	dev->vi = kzalloc(sizeof(struct ubi_volume_info), GFP_KERNEL);
+	if (!dev->vi) {
+		err = -ENOMEM;
+		goto out_close;
+	}
+	ubi_get_volume_info(dev->desc, dev->vi);
+
+	return 0;
+
+out_close:
+	ubi_close_volume(dev->desc);
+	dev->desc = NULL;
+out_err:
+	return err;
+}
+
+/**
+ * ubiblk_release - close a UBI volume (close the volume descriptor)
+ *
+ * @gd: the disk that was previously opened
+ * @mode: don't care
+ */
+static int ubiblk_release(struct gendisk *gd, fmode_t mode)
+{
+	struct ubiblk_dev *dev = gd->private_data;
+	dev_dbg(disk_to_dev(dev->gd), "released, mode=%d\n", mode);
+
+	kfree(dev->vi);
+	dev->vi = NULL;
+	if (dev->desc) {
+		ubi_close_volume(dev->desc);
+		dev->desc = NULL;
+	}
+
+	return 0;
+}
+
+/**
+ * ubi_ubiblk_thread - loop on the block request queue and wait for new
+ * requests ; run them with do_ubiblk_request(). Mostly copied from
+ * mtd_blkdevs.c
+ *
+ * @arg: the ubiblk device which request queue to process
+ */
+static int ubi_ubiblk_thread(void *arg)
+{
+	struct ubiblk_dev *dev = arg;
+	struct request_queue *rq = dev->rq;
+	struct request *req = NULL;
+
+	spin_lock_irq(rq->queue_lock);
+
+	while (!kthread_should_stop()) {
+		int res;
+
+		if (!req && !(req = blk_fetch_request(rq))) {
+			set_current_state(TASK_INTERRUPTIBLE);
+
+			if (kthread_should_stop())
+				set_current_state(TASK_RUNNING);
+
+			spin_unlock_irq(rq->queue_lock);
+			schedule();
+			spin_lock_irq(rq->queue_lock);
+			continue;
+		}
+
+		spin_unlock_irq(rq->queue_lock);
+
+		mutex_lock(&dev->vol_lock);
+		res = do_ubiblk_request(req, dev);
+		mutex_unlock(&dev->vol_lock);
+
+		spin_lock_irq(rq->queue_lock);
+
+		if (!__blk_end_request_cur(req, res))
+			req = NULL;
+	}
+
+	if (req)
+		__blk_end_request_all(req, -EIO);
+
+	spin_unlock_irq(rq->queue_lock);
+
+	return 0;
+}
+
+/**
+ * ubiblk_create - create a ubiblk device proxying a UBI volume
+ *
+ * @vi: the UBI volume information data structure
+ *
+ * An UBI volume has been created ; create a corresponding ubiblk device:
+ * Initialize the locks, the structure, the block layer infos and start a
+ * req_task.
+ */
+static int ubiblk_create(struct ubi_volume_info *vi)
+{
+	struct ubiblk_dev *dev;
+	struct gendisk *gd;
+	int ret = 0;
+
+	mutex_lock(&devlist_lock);
+	/* Check that the volume isn't already proxyfied */
+	if (ubiblk_find_dev(vi)) {
+		ret = -EEXIST;
+		goto out_devlist;
+	}
+
+	dev = kzalloc(sizeof(struct ubiblk_dev), GFP_KERNEL);
+	if (!dev) {
+		ret = -ENOMEM;
+		goto out_devlist;
+	}
+
+	list_add(&dev->list, &ubiblk_devs);
+
+	mutex_init(&dev->vol_lock);
+	mutex_lock(&dev->vol_lock);
+
+	dev->ubi_num = vi->ubi_num;
+	dev->vol_id = vi->vol_id;
+
+	dev->desc = ubi_open_volume(dev->ubi_num, dev->vol_id,
+					UBI_READONLY);
+	if (IS_ERR(dev->desc)) {
+		pr_err("ubi_open_volume failed\n");
+		ret = PTR_ERR(dev->desc);
+		goto out_vol;
+	}
+
+	dev->vi = kzalloc(sizeof(struct ubi_volume_info), GFP_KERNEL);
+	if (!dev->vi) {
+		ret = -ENOMEM;
+		goto out_info;
+	}
+	ubi_get_volume_info(dev->desc, dev->vi);
+
+	/* Initialize the gendisk of this ubiblk device */
+	gd = alloc_disk(1);
+	if (!gd) {
+		pr_err("alloc_disk failed\n");
+		ret = -ENODEV;
+		goto out_disk;
+	}
+
+	gd->fops = &ubiblk_ops;
+	gd->major = ubiblk_major;
+	gd->first_minor = dev->ubi_num * UBI_MAX_VOLUMES + dev->vol_id;
+	gd->private_data = dev;
+	sprintf(gd->disk_name, "ubiblk%d_%d", dev->ubi_num, dev->vol_id);
+	set_capacity(gd,
+		     (dev->vi->size *
+		      dev->vi->usable_leb_size) >> 9);
+	set_disk_ro(gd, 1);
+	dev->gd = gd;
+
+	spin_lock_init(&dev->queue_lock);
+	dev->rq = blk_init_queue(ubi_ubiblk_request, &dev->queue_lock);
+	if (!dev->rq) {
+		pr_err("blk_init_queue failed\n");
+		ret = -ENODEV;
+		goto out_queue;
+	}
+	dev->rq->queuedata = dev;
+	blk_queue_logical_block_size(dev->rq, BLK_SIZE);
+	dev->gd->queue = dev->rq;
+
+	/* Stolen from mtd_blkdevs.c */
+	/* Create processing req_task
+	 *
+	 * The processing of the request has to be done in process context (it
+	 * might sleep) but blk_run_queue can't block ; so we need to separate
+	 * the event of a request being added to the queue (which triggers the
+	 * callback ubi_ubiblk_request - that is set with blk_init_queue())
+	 * and the processing of that request.
+	 *
+	 * Thus, the sole purpose of ubi_ubiblk_reuqest is to wake the kthread
+	 * up so that it will process the request queue
+	 */
+	dev->req_task = kthread_run(ubi_ubiblk_thread, dev, "%s%d_%d",
+				  "kubiblk", dev->ubi_num, dev->vol_id);
+	if (IS_ERR(dev->req_task)) {
+		ret = PTR_ERR(dev->req_task);
+		goto out_thread;
+	}
+
+	add_disk(dev->gd);
+
+	dev_info(disk_to_dev(dev->gd),
+		 "created from ubi%d:%d(%s)\n", dev->ubi_num, dev->vol_id,
+		 dev->vi->name);
+
+	kfree(dev->vi);
+	dev->vi = NULL;
+	ubi_close_volume(dev->desc);
+	dev->desc = NULL;
+	mutex_unlock(&dev->vol_lock);
+
+	mutex_unlock(&devlist_lock);
+
+	return 0;
+
+out_thread:
+	blk_cleanup_queue(dev->rq);
+out_queue:
+	put_disk(dev->gd);
+out_disk:
+	kfree(dev->vi);
+	dev->vi = NULL;
+out_info:
+	ubi_close_volume(dev->desc);
+	dev->desc = NULL;
+out_vol:
+	mutex_unlock(&dev->vol_lock);
+out_devlist:
+	mutex_unlock(&devlist_lock);
+
+	return ret;
+}
+
+/**
+ * ubiblk_remove - destroy a ubiblk device
+ * @vi: the UBI volume information data structure
+ *
+ * A UBI volume has been removed or we are requested to unproxify a volume ;
+ * destroy the corresponding ubiblk device
+ */
+static int ubiblk_remove(struct ubi_volume_info *vi)
+{
+	struct ubiblk_dev *dev = NULL;
+
+	mutex_lock(&devlist_lock);
+
+	dev = ubiblk_find_dev(vi);
+
+	if (!dev) {
+		mutex_unlock(&devlist_lock);
+		pr_warn("trying to remove %s, but it isn't handled\n",
+			vi->name);
+		return -ENODEV;
+	}
+
+	if (dev->desc) {
+		mutex_unlock(&devlist_lock);
+		return -EBUSY;
+	}
+
+	del_gendisk(dev->gd);
+	blk_cleanup_queue(dev->rq);
+	kthread_stop(dev->req_task);
+	put_disk(dev->gd);
+
+	kfree(dev->vi);
+
+	list_del(&dev->list);
+	kfree(dev);
+
+	mutex_unlock(&devlist_lock);
+	pr_info("unproxyfied %s\n", vi->name);
+	return 0;
+}
+
+/**
+ * ubiblk_resize - resize a ubiblk device
+ * @vi: the UBI volume information data structure
+ *
+ * A UBI volume has been resized, change the ubiblk device geometry accordingly
+ */
+static int ubiblk_resize(struct ubi_volume_info *vi)
+{
+	struct ubiblk_dev *dev;
+
+	/* We don't touch the list, but we better lock it: it could be that the
+	 * device gets removed between the time the device has been found and
+	 * the time we access dev->gd
+	 */
+	mutex_lock(&devlist_lock);
+	dev = ubiblk_find_dev(vi);
+	if (!dev) {
+		mutex_unlock(&devlist_lock);
+		pr_warn("trying to resize %s, which isn't handled\n",
+			vi->name);
+		return -ENODEV;
+	}
+
+	mutex_lock(&dev->vol_lock);
+	set_capacity(dev->gd,
+		     (vi->size * vi->usable_leb_size) >> 9);
+	dev_dbg(disk_to_dev(dev->gd), "resized to %d LEBs\n", vi->size);
+	mutex_unlock(&dev->vol_lock);
+
+	mutex_unlock(&devlist_lock);
+	return 0;
+}
+
+/**
+ * ubiblk_notify - dispatches the UBI notifications
+ * copied from gluebi.c
+ * @nb: unused
+ * @notification_type: the notification type sent by UBI
+ * @ns_ptr: contains the notifications' additional informations
+ */
+static int ubiblk_notify(struct notifier_block *nb,
+			 unsigned long notification_type, void *ns_ptr)
+{
+	struct ubi_notification *nt = ns_ptr;
+
+	switch (notification_type) {
+	case UBI_VOLUME_REMOVED:
+		ubiblk_remove(&nt->vi);
+		break;
+	case UBI_VOLUME_RESIZED:
+		ubiblk_resize(&nt->vi);
+		break;
+	default:
+		break;
+	}
+	return NOTIFY_OK;
+}
+
+static const struct block_device_operations ubiblk_ops = {
+	.owner = THIS_MODULE,
+	.open = ubiblk_open,
+	.release = ubiblk_release,
+};
+
+static struct notifier_block ubiblk_notifier = {
+	.notifier_call = ubiblk_notify,
+};
+
+
+/**
+ * ubiblk_ctrl_ioctl - ioctl handling for proxying/unproxying a UBI volume
+ * @file: the file on which the ioctl was invoked (usunsed)
+ * @cmd: the ioctl type
+ * @arg: additional command informations
+ */
+static long ubiblk_ctrl_ioctl(struct file *file, unsigned int cmd,
+			      unsigned long arg)
+{
+	int err = 0;
+	void __user *argp = (void __user *)arg;
+
+	struct ubi_volume_desc *desc;
+	struct ubi_volume_info vi;
+	struct ubiblk_ctrl_req req;
+
+	if (!capable(CAP_SYS_RESOURCE))
+		return -EPERM;
+
+	err = copy_from_user(&req, argp, sizeof(struct ubiblk_ctrl_req));
+	if (err)
+		return -EFAULT;
+
+	if (req.ubi_num < 0 || req.vol_id < 0)
+		return -EINVAL;
+
+	desc = ubi_open_volume(req.ubi_num, req.vol_id, UBI_READONLY);
+	if (IS_ERR(desc)) {
+		dev_err(ctrl_dev.this_device, "opening ubi%d:%d failed\n",
+			req.ubi_num, req.vol_id);
+		return PTR_ERR(desc);
+	}
+
+	ubi_get_volume_info(desc, &vi);
+
+	switch (cmd) {
+	case UBIBLK_IOCADD:
+		dev_info(ctrl_dev.this_device, "proxying ubi%d:%d\n",
+			 req.ubi_num, req.vol_id);
+		err = ubiblk_create(&vi);
+		break;
+	case UBIBLK_IOCDEL:
+		dev_info(ctrl_dev.this_device, "unproxying ubi%d:%d\n",
+			 req.ubi_num, req.vol_id);
+		err = ubiblk_remove(&vi);
+		break;
+
+	default:
+		err = -ENOTTY;
+		break;
+	}
+
+	ubi_close_volume(desc);
+
+	return err;
+}
+
+#ifdef CONFIG_COMPAT
+static long ubiblk_ctrl_compat_ioctl(struct file *file, unsigned int cmd,
+				     unsigned long arg)
+{
+	unsigned long translated_arg = (unsigned long)compat_ptr(arg);
+
+	return ubiblk_ctrl_ioctl(file, cmd, translated_arg);
+}
+#endif
+
+/* ubiblk control device (receives ioctls) */
+static const struct file_operations ubiblk_ctrl_ops = {
+	.owner = THIS_MODULE,
+	.unlocked_ioctl = ubiblk_ctrl_ioctl,
+#ifdef CONFIG_COMPAT
+	.compat_ioctl = ubiblk_ctrl_compat_ioctl,
+#endif
+	.llseek = no_llseek,
+};
+static struct miscdevice ctrl_dev = {
+	.minor = MISC_DYNAMIC_MINOR,
+	.name = "ubiblk_ctrl",
+	.fops = &ubiblk_ctrl_ops,
+};
+
+/**
+ * ubi_ubiblk_init - initialize the module
+ *
+ * Get a major number and register to UBI notifications ; register the ioctl
+ * handler device
+ */
+static int __init ubi_ubiblk_init(void)
+{
+	int ret = 0;
+
+	ret = register_blkdev(0, "ubiblk");
+	if (ret <= 0) {
+		pr_err("can't register the major\n");
+		return -ENODEV;
+	}
+	ubiblk_major = ret;
+
+	mutex_init(&devlist_lock);
+
+	ret = misc_register(&ctrl_dev);
+	if (ret) {
+		pr_err("can't register control device\n");
+		goto out_unreg_blk;
+	}
+
+	ret = ubi_register_volume_notifier(&ubiblk_notifier, 1);
+	if (ret < 0)
+		goto out_unreg_ctrl;
+
+	pr_info("major=%d\n", ubiblk_major);
+
+	return ret;
+
+out_unreg_ctrl:
+	misc_deregister(&ctrl_dev);
+out_unreg_blk:
+	unregister_blkdev(ubiblk_major, "ubiblk");
+
+	return ret;
+}
+
+/*
+ * ubi_ubiblk_exit - end of life
+ *
+ * unregister the block device major, unregister from UBI notifications,
+ * unregister the ioctl handler device stop the threads and free the memory.
+ */
+static void __exit ubi_ubiblk_exit(void)
+{
+	struct list_head *list_ptr, *next;
+	struct ubiblk_dev *dev;
+
+	ubi_unregister_volume_notifier(&ubiblk_notifier);
+	misc_deregister(&ctrl_dev);
+
+	list_for_each_safe(list_ptr, next, &ubiblk_devs) {
+		dev = list_entry(list_ptr, struct ubiblk_dev, list);
+
+		/* TODO: it shouldn't happen, right ? */
+		if (dev->desc)
+			ubi_close_volume(dev->desc);
+
+		del_gendisk(dev->gd);
+		blk_cleanup_queue(dev->rq);
+		kthread_stop(dev->req_task);
+		put_disk(dev->gd);
+
+		kfree(dev->vi);
+		list_del(&dev->list); /* really needed ? */
+		kfree(dev);
+	}
+
+	unregister_blkdev(ubiblk_major, "ubiblk");
+	pr_info("end of life\n");
+}
+
+module_init(ubi_ubiblk_init);
+module_exit(ubi_ubiblk_exit);
+MODULE_DESCRIPTION("Read-only block transition layer on top of UBI");
+MODULE_AUTHOR("David Wagner");
+MODULE_LICENSE("GPL");
diff --git a/include/mtd/Kbuild b/include/mtd/Kbuild
index 192f8fb..d0d59d8 100644
--- a/include/mtd/Kbuild
+++ b/include/mtd/Kbuild
@@ -3,3 +3,4 @@ header-y += mtd-abi.h
 header-y += mtd-user.h
 header-y += nftl-user.h
 header-y += ubi-user.h
+header-y += ubiblk-user.h
diff --git a/include/mtd/ubiblk-user.h b/include/mtd/ubiblk-user.h
new file mode 100644
index 0000000..61df415
--- /dev/null
+++ b/include/mtd/ubiblk-user.h
@@ -0,0 +1,47 @@
+/*
+ * Copyright © Free Electrons, 2011
+ * Copyright © International Business Machines Corp., 2006
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See
+ * the GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ *
+ * Author: David Wagner
+ * Some code taken from ubi-user.h
+ */
+
+#ifndef __UBIBLK_USER_H__
+#define __UBIBLK_USER_H__
+
+#include <linux/types.h>
+
+/**
+ * ubiblk_ctrl_req - additional ioctl data structure
+ * @ubi_num: UBI device number
+ * @vol_id: UBI volume identifier
+ */
+struct ubiblk_ctrl_req {
+	__s32 ubi_num;
+	__s32 vol_id;
+} __packed;
+
+/* ioctl commands of the UBI control character device */
+#define UBIBLK_CTRL_IOC_MAGIC 'O'
+
+/* Create a ubiblk device from a UBI volume */
+#define UBIBLK_IOCADD _IOW(UBIBLK_CTRL_IOC_MAGIC, 0x10, struct ubiblk_ctrl_req)
+/* Delete a ubiblk device */
+#define UBIBLK_IOCDEL _IOW(UBIBLK_CTRL_IOC_MAGIC, 0x11, struct ubiblk_ctrl_req)
+/* If you add ioctls here, please note that UBI uses 'O'/0x00-0x06 */
+
+#endif
-- 
1.7.0.4

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists