X-Git-Url: http://ftp.safe.ca/?a=blobdiff_plain;f=fs%2Fblock_dev.c;h=26e5f50266205a7e7d2069bb92dd5600e77fe61c;hb=578454ff7eab61d13a26b568f99a89a2c9edc881;hp=e59440c7e1cf3e0c352f4900a91bebf540833c32;hpb=1a3cbbc5a5e8a66934aa0947896a4aca6fd77298;p=safe%2Fjmp%2Flinux-2.6

diff --git a/fs/block_dev.c b/fs/block_dev.c
index e59440c..26e5f50 100644
--- a/fs/block_dev.c
+++ b/fs/block_dev.c
@@ -245,37 +245,14 @@ struct super_block *freeze_bdev(struct block_device *bdev)
 	sb = get_active_super(bdev);
 	if (!sb)
 		goto out;
-	if (sb->s_flags & MS_RDONLY) {
-		sb->s_frozen = SB_FREEZE_TRANS;
-		up_write(&sb->s_umount);
+	error = freeze_super(sb);
+	if (error) {
+		deactivate_super(sb);
+		bdev->bd_fsfreeze_count--;
 		mutex_unlock(&bdev->bd_fsfreeze_mutex);
-		return sb;
-	}
-
-	sb->s_frozen = SB_FREEZE_WRITE;
-	smp_wmb();
-
-	sync_filesystem(sb);
-
-	sb->s_frozen = SB_FREEZE_TRANS;
-	smp_wmb();
-
-	sync_blockdev(sb->s_bdev);
-
-	if (sb->s_op->freeze_fs) {
-		error = sb->s_op->freeze_fs(sb);
-		if (error) {
-			printk(KERN_ERR
-				"VFS:Filesystem freeze failed\n");
-			sb->s_frozen = SB_UNFROZEN;
-			deactivate_locked_super(sb);
-			bdev->bd_fsfreeze_count--;
-			mutex_unlock(&bdev->bd_fsfreeze_mutex);
-			return ERR_PTR(error);
-		}
+		return ERR_PTR(error);
 	}
-	up_write(&sb->s_umount);
-
+	deactivate_super(sb);
  out:
 	sync_blockdev(bdev);
 	mutex_unlock(&bdev->bd_fsfreeze_mutex);
@@ -296,40 +273,22 @@ int thaw_bdev(struct block_device *bdev, struct super_block *sb)
 
 	mutex_lock(&bdev->bd_fsfreeze_mutex);
 	if (!bdev->bd_fsfreeze_count)
-		goto out_unlock;
+		goto out;
 
 	error = 0;
 	if (--bdev->bd_fsfreeze_count > 0)
-		goto out_unlock;
+		goto out;
 
 	if (!sb)
-		goto out_unlock;
-
-	BUG_ON(sb->s_bdev != bdev);
-	down_write(&sb->s_umount);
-	if (sb->s_flags & MS_RDONLY)
-		goto out_unfrozen;
-
-	if (sb->s_op->unfreeze_fs) {
-		error = sb->s_op->unfreeze_fs(sb);
-		if (error) {
-			printk(KERN_ERR
-				"VFS:Filesystem thaw failed\n");
-			sb->s_frozen = SB_FREEZE_TRANS;
-			bdev->bd_fsfreeze_count++;
-			mutex_unlock(&bdev->bd_fsfreeze_mutex);
-			return error;
-		}
-	}
-
-out_unfrozen:
-	sb->s_frozen = SB_UNFROZEN;
-	smp_wmb();
-	wake_up(&sb->s_wait_unfrozen);
+		goto out;
 
-	if (sb)
-		deactivate_locked_super(sb);
-out_unlock:
+	error = thaw_super(sb);
+	if (error) {
+		bdev->bd_fsfreeze_count++;
+		mutex_unlock(&bdev->bd_fsfreeze_mutex);
+		return error;
+	}
+out:
 	mutex_unlock(&bdev->bd_fsfreeze_mutex);
 	return 0;
 }
@@ -406,16 +365,23 @@ static loff_t block_llseek(struct file *file, loff_t offset, int origin)
  
 int blkdev_fsync(struct file *filp, struct dentry *dentry, int datasync)
 {
-	struct block_device *bdev = I_BDEV(filp->f_mapping->host);
+	struct inode *bd_inode = filp->f_mapping->host;
+	struct block_device *bdev = I_BDEV(bd_inode);
 	int error;
 
-	error = sync_blockdev(bdev);
-	if (error)
-		return error;
-	
-	error = blkdev_issue_flush(bdev, NULL);
+	/*
+	 * There is no need to serialise calls to blkdev_issue_flush with
+	 * i_mutex and doing so causes performance issues with concurrent
+	 * O_SYNC writers to a block device.
+	 */
+	mutex_unlock(&bd_inode->i_mutex);
+
+	error = blkdev_issue_flush(bdev, GFP_KERNEL, NULL, BLKDEV_IFL_WAIT);
 	if (error == -EOPNOTSUPP)
 		error = 0;
+
+	mutex_lock(&bd_inode->i_mutex);
+
 	return error;
 }
 EXPORT_SYMBOL(blkdev_fsync);
@@ -694,11 +660,144 @@ static bool bd_may_claim(struct block_device *bdev, struct block_device *whole,
 }
 
 /**
+ * bd_prepare_to_claim - prepare to claim a block device
+ * @bdev: block device of interest
+ * @whole: the whole device containing @bdev, may equal @bdev
+ * @holder: holder trying to claim @bdev
+ *
+ * Prepare to claim @bdev.  This function fails if @bdev is already
+ * claimed by another holder and waits if another claiming is in
+ * progress.  This function doesn't actually claim.  On successful
+ * return, the caller has ownership of bd_claiming and bd_holder[s].
+ *
+ * CONTEXT:
+ * spin_lock(&bdev_lock).  Might release bdev_lock, sleep and regrab
+ * it multiple times.
+ *
+ * RETURNS:
+ * 0 if @bdev can be claimed, -EBUSY otherwise.
+ */
+static int bd_prepare_to_claim(struct block_device *bdev,
+			       struct block_device *whole, void *holder)
+{
+retry:
+	/* if someone else claimed, fail */
+	if (!bd_may_claim(bdev, whole, holder))
+		return -EBUSY;
+
+	/* if someone else is claiming, wait for it to finish */
+	if (whole->bd_claiming && whole->bd_claiming != holder) {
+		wait_queue_head_t *wq = bit_waitqueue(&whole->bd_claiming, 0);
+		DEFINE_WAIT(wait);
+
+		prepare_to_wait(wq, &wait, TASK_UNINTERRUPTIBLE);
+		spin_unlock(&bdev_lock);
+		schedule();
+		finish_wait(wq, &wait);
+		spin_lock(&bdev_lock);
+		goto retry;
+	}
+
+	/* yay, all mine */
+	return 0;
+}
+
+/**
+ * bd_start_claiming - start claiming a block device
+ * @bdev: block device of interest
+ * @holder: holder trying to claim @bdev
+ *
+ * @bdev is about to be opened exclusively.  Check @bdev can be opened
+ * exclusively and mark that an exclusive open is in progress.  Each
+ * successful call to this function must be matched with a call to
+ * either bd_claim() or bd_abort_claiming().  If this function
+ * succeeds, the matching bd_claim() is guaranteed to succeed.
+ *
+ * CONTEXT:
+ * Might sleep.
+ *
+ * RETURNS:
+ * Pointer to the block device containing @bdev on success, ERR_PTR()
+ * value on failure.
+ */
+static struct block_device *bd_start_claiming(struct block_device *bdev,
+					      void *holder)
+{
+	struct gendisk *disk;
+	struct block_device *whole;
+	int partno, err;
+
+	might_sleep();
+
+	/*
+	 * @bdev might not have been initialized properly yet, look up
+	 * and grab the outer block device the hard way.
+	 */
+	disk = get_gendisk(bdev->bd_dev, &partno);
+	if (!disk)
+		return ERR_PTR(-ENXIO);
+
+	whole = bdget_disk(disk, 0);
+	put_disk(disk);
+	if (!whole)
+		return ERR_PTR(-ENOMEM);
+
+	/* prepare to claim, if successful, mark claiming in progress */
+	spin_lock(&bdev_lock);
+
+	err = bd_prepare_to_claim(bdev, whole, holder);
+	if (err == 0) {
+		whole->bd_claiming = holder;
+		spin_unlock(&bdev_lock);
+		return whole;
+	} else {
+		spin_unlock(&bdev_lock);
+		bdput(whole);
+		return ERR_PTR(err);
+	}
+}
+
+/* releases bdev_lock */
+static void __bd_abort_claiming(struct block_device *whole, void *holder)
+{
+	BUG_ON(whole->bd_claiming != holder);
+	whole->bd_claiming = NULL;
+	wake_up_bit(&whole->bd_claiming, 0);
+
+	spin_unlock(&bdev_lock);
+	bdput(whole);
+}
+
+/**
+ * bd_abort_claiming - abort claiming a block device
+ * @whole: whole block device returned by bd_start_claiming()
+ * @holder: holder trying to claim @bdev
+ *
+ * Abort a claiming block started by bd_start_claiming().  Note that
+ * @whole is not the block device to be claimed but the whole device
+ * returned by bd_start_claiming().
+ *
+ * CONTEXT:
+ * Grabs and releases bdev_lock.
+ */
+static void bd_abort_claiming(struct block_device *whole, void *holder)
+{
+	spin_lock(&bdev_lock);
+	__bd_abort_claiming(whole, holder);		/* releases bdev_lock */
+}
+
+/**
  * bd_claim - claim a block device
  * @bdev: block device to claim
  * @holder: holder trying to claim @bdev
  *
- * Try to claim @bdev.
+ * Try to claim @bdev which must have been opened successfully.  This
+ * function may be called with or without preceding
+ * blk_start_claiming().  In the former case, this function is always
+ * successful and terminates the claiming block.
+ *
+ * CONTEXT:
+ * Might sleep.
  *
  * RETURNS:
  * 0 if successful, -EBUSY if @bdev is already claimed.
@@ -706,11 +805,14 @@ static bool bd_may_claim(struct block_device *bdev, struct block_device *whole,
 int bd_claim(struct block_device *bdev, void *holder)
 {
 	struct block_device *whole = bdev->bd_contains;
-	int res = -EBUSY;
+	int res;
+
+	might_sleep();
 
 	spin_lock(&bdev_lock);
 
-	if (bd_may_claim(bdev, whole, holder)) {
+	res = bd_prepare_to_claim(bdev, whole, holder);
+	if (res == 0) {
 		/* note that for a whole device bd_holders
 		 * will be incremented twice, and bd_holder will
 		 * be set to bd_claim before being set to holder
@@ -719,10 +821,13 @@ int bd_claim(struct block_device *bdev, void *holder)
 		whole->bd_holder = bd_claim;
 		bdev->bd_holders++;
 		bdev->bd_holder = holder;
-		res = 0;
 	}
 
-	spin_unlock(&bdev_lock);
+	if (whole->bd_claiming)
+		__bd_abort_claiming(whole, holder);	/* releases bdev_lock */
+	else
+		spin_unlock(&bdev_lock);
+
 	return res;
 }
 EXPORT_SYMBOL(bd_claim);
@@ -1338,6 +1443,7 @@ EXPORT_SYMBOL(blkdev_get);
 
 static int blkdev_open(struct inode * inode, struct file * filp)
 {
+	struct block_device *whole = NULL;
 	struct block_device *bdev;
 	int res;
 
@@ -1360,22 +1466,25 @@ static int blkdev_open(struct inode * inode, struct file * filp)
 	if (bdev == NULL)
 		return -ENOMEM;
 
+	if (filp->f_mode & FMODE_EXCL) {
+		whole = bd_start_claiming(bdev, filp);
+		if (IS_ERR(whole)) {
+			bdput(bdev);
+			return PTR_ERR(whole);
+		}
+	}
+
 	filp->f_mapping = bdev->bd_inode->i_mapping;
 
 	res = blkdev_get(bdev, filp->f_mode);
-	if (res)
-		return res;
 
-	if (filp->f_mode & FMODE_EXCL) {
-		res = bd_claim(bdev, filp);
-		if (res)
-			goto out_blkdev_put;
+	if (whole) {
+		if (res == 0)
+			BUG_ON(bd_claim(bdev, filp) != 0);
+		else
+			bd_abort_claiming(whole, filp);
 	}
 
-	return 0;
-
- out_blkdev_put:
-	blkdev_put(bdev, filp->f_mode);
 	return res;
 }
 
@@ -1586,27 +1695,34 @@ EXPORT_SYMBOL(lookup_bdev);
  */
 struct block_device *open_bdev_exclusive(const char *path, fmode_t mode, void *holder)
 {
-	struct block_device *bdev;
-	int error = 0;
+	struct block_device *bdev, *whole;
+	int error;
 
 	bdev = lookup_bdev(path);
 	if (IS_ERR(bdev))
 		return bdev;
 
+	whole = bd_start_claiming(bdev, holder);
+	if (IS_ERR(whole)) {
+		bdput(bdev);
+		return whole;
+	}
+
 	error = blkdev_get(bdev, mode);
 	if (error)
-		return ERR_PTR(error);
+		goto out_abort_claiming;
+
 	error = -EACCES;
 	if ((mode & FMODE_WRITE) && bdev_read_only(bdev))
-		goto blkdev_put;
-	error = bd_claim(bdev, holder);
-	if (error)
-		goto blkdev_put;
+		goto out_blkdev_put;
 
+	BUG_ON(bd_claim(bdev, holder) != 0);
 	return bdev;
-	
-blkdev_put:
+
+out_blkdev_put:
 	blkdev_put(bdev, mode);
+out_abort_claiming:
+	bd_abort_claiming(whole, holder);
 	return ERR_PTR(error);
 }