Skip to contentSkip to conversation

fix a race between del_gendisk and BLKRRPART v2

PATCH In review Christoph Hellwig 4 participants 9 messages May 14, 2021 – May 20, 2021

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

Hi all,

this is based of a patch from Gulam and suggestions from Ming and fixes a race between del_gendisk and BLKRRPART while also removing a global lock.

Changes since v1:
- fix the GENHD_FL_UP check in __blkdev_get
- don't change where remove_inode_hash is called for now
- improve the commit message

Diffstat:

 block/genhd.c         |   11 +----------
 fs/block_dev.c        |   18 ++++++++----------
 include/linux/genhd.h |    2 --
 3 files changed, 9 insertions(+), 22 deletions(-)
#1 lore raw

PATCH 1/2 block: prevent block device lookups at the beginning of del_gendisk

As an artifact of how gendisk lookup used to work in earlier kernels, GENHD_FL_UP is only cleared very late in del_gendisk, and a global lock is used to prevent opens from succeeding while del_gendisk is tearing down the gendisk. Switch to clearing the flag early and under bd_mutex so that callers can use bd_mutex to stabilize the flag, which removes the need for the global mutex.

---
 block/genhd.c         | 11 +----------
 fs/block_dev.c        | 15 +++++----------
 include/linux/genhd.h |  2 --
 3 files changed, 6 insertions(+), 22 deletions(-)
block/genhd.c +1 -10
index 39ca97b0edc6..9f8cb7beaad1 100644
@@ -29,8 +29,6 @@
 
 static struct kobject *block_depr;
 
-DECLARE_RWSEM(bdev_lookup_sem);
-
 /* for extended dynamic devt allocation, currently only one major is used */
 #define NR_EXT_DEVT		(1 << MINORBITS)
 static DEFINE_IDA(ext_devt_ida);
@@ -609,13 +607,8 @@ void del_gendisk(struct gendisk *disk)
 	blk_integrity_del(disk);
 	disk_del_events(disk);
 
-	/*
-	 * Block lookups of the disk until all bdevs are unhashed and the
-	 * disk is marked as dead (GENHD_FL_UP cleared).
-	 */
-	down_write(&bdev_lookup_sem);
-
 	mutex_lock(&disk->part0->bd_mutex);
+	disk->flags &= ~GENHD_FL_UP;
 	blk_drop_partitions(disk);
 	mutex_unlock(&disk->part0->bd_mutex);
 
@@ -629,8 +622,6 @@ void del_gendisk(struct gendisk *disk)
 	remove_inode_hash(disk->part0->bd_inode);
 
 	set_capacity(disk, 0);
-	disk->flags &= ~GENHD_FL_UP;
-	up_write(&bdev_lookup_sem);
 
 	if (!(disk->flags & GENHD_FL_HIDDEN)) {
 		sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
fs/block_dev.c +5 -10
index eb265d72fce8..580bae995b87 100644
@@ -1298,6 +1298,9 @@ static int __blkdev_get(struct block_device *bdev, fmode_t mode)
 	struct gendisk *disk = bdev->bd_disk;
 	int ret = 0;
 
+	if (!(disk->flags & GENHD_FL_UP))
+		return -ENXIO;
+
 	if (!bdev->bd_openers) {
 		if (!bdev_is_partition(bdev)) {
 			ret = 0;
@@ -1332,8 +1335,7 @@ static int __blkdev_get(struct block_device *bdev, fmode_t mode)
 			whole->bd_part_count++;
 			mutex_unlock(&whole->bd_mutex);
 
-			if (!(disk->flags & GENHD_FL_UP) ||
-			    !bdev_nr_sectors(bdev)) {
+			if (!bdev_nr_sectors(bdev)) {
 				__blkdev_put(whole, mode, 1);
 				bdput(whole);
 				return -ENXIO;
@@ -1364,16 +1366,12 @@ struct block_device *blkdev_get_no_open(dev_t dev)
 	struct block_device *bdev;
 	struct gendisk *disk;
 
-	down_read(&bdev_lookup_sem);
 	bdev = bdget(dev);
 	if (!bdev) {
-		up_read(&bdev_lookup_sem);
 		blk_request_module(dev);
-		down_read(&bdev_lookup_sem);
-
 		bdev = bdget(dev);
 		if (!bdev)
-			goto unlock;
+			return NULL;
 	}
 
 	disk = bdev->bd_disk;
@@ -1383,14 +1381,11 @@ struct block_device *blkdev_get_no_open(dev_t dev)
 		goto put_disk;
 	if (!try_module_get(bdev->bd_disk->fops->owner))
 		goto put_disk;
-	up_read(&bdev_lookup_sem);
 	return bdev;
 put_disk:
 	put_disk(disk);
 bdput:
 	bdput(bdev);
-unlock:
-	up_read(&bdev_lookup_sem);
 	return NULL;
 }
 
include/linux/genhd.h +0 -2
index 7e9660ea967d..6fc26f7bdf71 100644
@@ -306,8 +306,6 @@ static inline void bd_unlink_disk_holder(struct block_device *bdev,
 }
 #endif /* CONFIG_SYSFS */
 
-extern struct rw_semaphore bdev_lookup_sem;
-
 dev_t blk_lookup_devt(const char *name, int partno);
 void blk_request_module(dev_t devt);
 #ifdef CONFIG_BLOCK
Signature · 1 line
2.30.2
Reviewed-by on v1 1/2
#2 lore raw

Re: [PATCH 1/2] block: prevent block device lookups at the beginning of del_gendisk

Quoted 8 lines · Christoph Hellwig

As an artifact of how gendisk lookup used to work in earlier kernels, GENHD_FL_UP is only cleared very late in del_gendisk, and a global lock is used to prevent opens from succeeding while del_gendisk is tearing down the gendisk. Switch to clearing the flag early and under bd_mutex so that callers can use bd_mutex to stabilize the flag, which removes the need for the global mutex.

Signature · 1 line
Ming
Reviewed-by Ming Lei Tested-by
#3 lore raw

RE: [PATCH 1/2] block: prevent block device lookups at the beginning of del_gendisk

-----Original Message-----
From: Ming Lei <[email protected]>
Sent: Saturday, May 15, 2021 12:05 PM
To: Christoph Hellwig <[email protected]>
Cc: [email protected]; Gulam Mohamed <[email protected]>; [email protected]
Subject: Re: [PATCH 1/2] block: prevent block device lookups at the beginning of del_gendisk

Quoted 8 lines · Christoph Hellwig

As an artifact of how gendisk lookup used to work in earlier kernels, GENHD_FL_UP is only cleared very late in del_gendisk, and a global lock is used to prevent opens from succeeding while del_gendisk is tearing down the gendisk. Switch to clearing the flag early and under bd_mutex so that callers can use bd_mutex to stabilize the flag, which removes the need for the global mutex.

I have tested these patches on latest kernel in our environment and didn't find any issue. The fix is working fine.

--
Ming

#4 lore raw

PATCH 2/2 block: fix a race between del_gendisk and BLKRRPART

In reply to Christoph Hellwig

From: Gulam Mohamed <[email protected]>

When BLKRRPART is called concurrently with del_gendisk, the partitions rescan can create a stale partition that will never be be cleaned up.

Fix this by checking the the disk is up before rescanning partitions while under bd_mutex.

[hch: split from a larger patch]
---
 fs/block_dev.c | 3 +++
 1 file changed, 3 insertions(+)
fs/block_dev.c +3 -0
index 580bae995b87..4494411fa4d3 100644
@@ -1244,6 +1244,9 @@ int bdev_disk_changed(struct block_device *bdev, bool invalidate)
 
 	lockdep_assert_held(&bdev->bd_mutex);
 
+	if (!(disk->flags & GENHD_FL_UP))
+		return -ENXIO;
+
 rescan:
 	if (bdev->bd_part_count)
 		return -EBUSY;
Signature · 1 line
2.30.2
Reviewed-by on v1 2/2
#5 lore raw

Re: [PATCH 2/2] block: fix a race between del_gendisk and BLKRRPART

Quoted 20 lines · Christoph Hellwig

From: Gulam Mohamed <[email protected]>

When BLKRRPART is called concurrently with del_gendisk, the partitions rescan can create a stale partition that will never be be cleaned up.

Fix this by checking the the disk is up before rescanning partitions while under bd_mutex.

[hch: split from a larger patch]
---
 fs/block_dev.c | 3 +++
 1 file changed, 3 insertions(+)
fs/block_dev.c +0 -0
index 580bae995b87..4494411fa4d3 100644

fs/block_dev.c int bdev_disk_changed(struct block_device *bdev, bool invalidate) · line 1252

Quoted 9 lines
fs/block_dev.c +3 -0
@@ -1244,6 +1244,9 @@ int bdev_disk_changed(struct block_device *bdev, bool invalidate)
 
 	lockdep_assert_held(&bdev->bd_mutex);
 
+	if (!(disk->flags & GENHD_FL_UP))
+		return -ENXIO;
+
 rescan:
 	if (bdev->bd_part_count)
 		return -EBUSY;
Signature · 1 line
Ming
Reviewed-by Ming Lei Tested-by
#6 lore raw

RE: [PATCH 2/2] block: fix a race between del_gendisk and BLKRRPART

-----Original Message-----
From: Ming Lei <[email protected]>
Sent: Saturday, May 15, 2021 12:05 PM
To: Christoph Hellwig <[email protected]>
Cc: [email protected]; Gulam Mohamed <[email protected]>; [email protected]
Subject: Re: [PATCH 2/2] block: fix a race between del_gendisk and BLKRRPART

Quoted 29 lines · Christoph Hellwig

From: Gulam Mohamed <[email protected]>

When BLKRRPART is called concurrently with del_gendisk, the partitions rescan can create a stale partition that will never be be cleaned up.

Fix this by checking the the disk is up before rescanning partitions while under bd_mutex.

[hch: split from a larger patch]
---
 fs/block_dev.c | 3 +++
 1 file changed, 3 insertions(+)
fs/block_dev.c index renamed from fs/block_dev.c +0 -0

580bae995b87..4494411fa4d3 100644

fs/block_dev.c +3 -0
@@ -1244,6 +1244,9 @@ int bdev_disk_changed(struct block_device *bdev, 
 bool invalidate)
 
 	lockdep_assert_held(&bdev->bd_mutex);
 
+	if (!(disk->flags & GENHD_FL_UP))
+		return -ENXIO;
+
 rescan:
 	if (bdev->bd_part_count)
Quoted 1 line

return -EBUSY;

--
Ming

#7 lore raw
In reply to Christoph Hellwig

Jens,

can you take a look at these?

Quoted 10 lines · Christoph Hellwig

Hi all,

this is based of a patch from Gulam and suggestions from Ming and fixes a race between del_gendisk and BLKRRPART while also removing a global lock.

Changes since v1:
- fix the GENHD_FL_UP check in __blkdev_get
- don't change where remove_inode_hash is called for now
- improve the commit message

Quoted 5 lines

Diffstat:

 block/genhd.c         |   11 +----------
 fs/block_dev.c        |   18 ++++++++----------
 include/linux/genhd.h |    2 --
 3 files changed, 9 insertions(+), 22 deletions(-)

---end quoted text---

maintainer
#8 lore raw
Quoted 3 lines · Christoph Hellwig

Jens,

can you take a look at these?

Looks good to me, applied.

Signature · 1 line
Jens Axboe