[PATCH 0/2] block: add info messages when opening a block device O_WRITE and O_EXCL concurrently

[PATCH 0/2] block: add info messages when opening a block device O_WRITE and O_EXCL concurrently

[zhangyi (F)]

Open an exclusive-opened block device for write or open a write-opened
block deviece exclusively cannot make sure it is exclusive enough. So
it may corrupt the block device when some one writing data through the
counterpart raw block device, such as corrupt a mounted file system.
And it is hard to find the bad process who corrupt the device after
something bad happens.

Although we can watch the block device in the user space, we still want
the kernel can give us some messages if that happens for some special
cases. So this patch set want to add some info messages to hint the
potential data corruption.

zhangyi (F) (2):
  block: add info when opening an exclusive opened block device for
    write
  block: add info when opening a write opend block device exclusively

 fs/block_dev.c     | 38 +++++++++++++++++++++++++++++++++++---
 include/linux/fs.h |  1 +
 2 files changed, 36 insertions(+), 3 deletions(-)

-- 
2.7.4

[PATCH 1/2] block: add info when opening an exclusive opened block device for write

[zhangyi (F)]

Opening an exclusive opened block device for write make the exclusive
open isn't exclusive enough, it may lead to data corruption when some
one writing data through the counterpart raw block device, such as
corrupt a mounted file system. This patch add an info message when
opening an exclusive opened block device for write to hint the
potential data corruption.

Note that there are some legal cases such as file system or device
mapper online resize, so this message is just a hint and isn't always
mean that a risky written happens.

Signed-off-by: zhangyi (F) <yi.zhang@huawei.com>
---
 fs/block_dev.c | 22 +++++++++++++++++++---
 1 file changed, 19 insertions(+), 3 deletions(-)

diff --git a/fs/block_dev.c b/fs/block_dev.c
index bb28e2e..d92aa45 100644
--- a/fs/block_dev.c
+++ b/fs/block_dev.c
@@ -1667,13 +1667,13 @@ int blkdev_get(struct block_device *bdev, fmode_t mode, void *holder)
 
 	res = __blkdev_get(bdev, mode, 0);
 
+	mutex_lock(&bdev->bd_mutex);
+	spin_lock(&bdev_lock);
+
 	if (whole) {
 		struct gendisk *disk = whole->bd_disk;
 
 		/* finish claiming */
-		mutex_lock(&bdev->bd_mutex);
-		spin_lock(&bdev_lock);
-
 		if (!res) {
 			BUG_ON(!bd_may_claim(bdev, whole, holder));
 			/*
@@ -1710,6 +1710,22 @@ int blkdev_get(struct block_device *bdev, fmode_t mode, void *holder)
 
 		mutex_unlock(&bdev->bd_mutex);
 		bdput(whole);
+	} else {
+		if (!res && (mode & FMODE_WRITE) && bdev->bd_holders) {
+			char name[BDEVNAME_SIZE];
+
+			/*
+			 * Open an exclusive opened device for write may
+			 * probability corrupt the device, such as a
+			 * mounted file system, give a hint here.
+			 */
+			pr_info_ratelimited("VFS: Open an exclusive opened "
+				    "block device for write %s [%d %s].\n",
+				    bdevname(bdev, name), current->pid,
+				    current->comm);
+		}
+		spin_unlock(&bdev_lock);
+		mutex_unlock(&bdev->bd_mutex);
 	}
 
 	return res;
-- 
2.7.4

[PATCH 2/2] block: add info when opening a write opend block device exclusively

[zhangyi (F)]

Just like open an exclusive opened block device for write, open a block
device exclusively which has been opened for write by some other
processes may also lead to potential data corruption. This patch record
the write openers and give a hint if that happens.

Signed-off-by: zhangyi (F) <yi.zhang@huawei.com>
---
 fs/block_dev.c     | 20 ++++++++++++++++++--
 include/linux/fs.h |  1 +
 2 files changed, 19 insertions(+), 2 deletions(-)

diff --git a/fs/block_dev.c b/fs/block_dev.c
index d92aa45..c278195 100644
--- a/fs/block_dev.c
+++ b/fs/block_dev.c
@@ -1606,6 +1606,8 @@ static int __blkdev_get(struct block_device *bdev, fmode_t mode, int for_part)
 	bdev->bd_openers++;
 	if (for_part)
 		bdev->bd_part_count++;
+	if (mode & FMODE_WRITE)
+		bdev->bd_write_openers++;
 	mutex_unlock(&bdev->bd_mutex);
 	disk_unblock_events(disk);
 	/* only one opener holds refs to the module and disk */
@@ -1654,6 +1656,7 @@ int blkdev_get(struct block_device *bdev, fmode_t mode, void *holder)
 {
 	struct block_device *whole = NULL;
 	int res;
+	char name[BDEVNAME_SIZE];
 
 	WARN_ON_ONCE((mode & FMODE_EXCL) && !holder);
 
@@ -1673,6 +1676,19 @@ int blkdev_get(struct block_device *bdev, fmode_t mode, void *holder)
 	if (whole) {
 		struct gendisk *disk = whole->bd_disk;
 
+		/*
+		 * Open an write opened block device exclusively, the
+		 * writing process may probability corrupt the device,
+		 * such as a mounted file system, give a hint here.
+		 */
+		if (!res && (bdev->bd_write_openers >
+		    ((mode & FMODE_WRITE) ? 1 : 0)) && !bdev->bd_holders) {
+			pr_info_ratelimited("VFS: Open an write opened "
+				"block device exclusively %s [%d %s].\n",
+				bdevname(bdev, name), current->pid,
+				current->comm);
+		}
+
 		/* finish claiming */
 		if (!res) {
 			BUG_ON(!bd_may_claim(bdev, whole, holder));
@@ -1712,8 +1728,6 @@ int blkdev_get(struct block_device *bdev, fmode_t mode, void *holder)
 		bdput(whole);
 	} else {
 		if (!res && (mode & FMODE_WRITE) && bdev->bd_holders) {
-			char name[BDEVNAME_SIZE];
-
 			/*
 			 * Open an exclusive opened device for write may
 			 * probability corrupt the device, such as a
@@ -1848,6 +1862,8 @@ static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part)
 	struct block_device *victim = NULL;
 
 	mutex_lock_nested(&bdev->bd_mutex, for_part);
+	if (mode & FMODE_WRITE)
+		bdev->bd_write_openers--;
 	if (for_part)
 		bdev->bd_part_count--;
 
diff --git a/include/linux/fs.h b/include/linux/fs.h
index dd28e76..0dc066d 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -463,6 +463,7 @@ struct request_queue;
 struct block_device {
 	dev_t			bd_dev;  /* not a kdev_t - it's a search key */
 	int			bd_openers;
+	int			bd_write_openers;
 	struct inode *		bd_inode;	/* will die */
 	struct super_block *	bd_super;
 	struct mutex		bd_mutex;	/* open/close mutex */
-- 
2.7.4