Skip to contentSkip to conversation

bdev: Refresh bdev size for disks without partitioning

PATCH In review Jan Kara 6 participants 17 messages Oct 21, 2019 – Nov 5, 2019

Subsystems (2)
lore
Share & tools
lore

Jan Kara

author reviewer
#0 lore raw

cover bdev: Refresh bdev size for disks without partitioning

Hello,

I've been debugging for quite a long time strange issues with encrypted DVDs not being readable on Linux (bko#194965). In the end I've tracked down the problem to the fact that block device size is not updated when the media is inserted in case the DVD device is already open. This is IMO a bug in block device code as the size gets properly update in case the device has partition scanning enabled. The following series fixes the problem by refreshing disk size on each open even for devices with partition scanning disabled.

Honza

author reviewer
#1 lore raw

PATCH 1/2 bdev: Factor out bdev revalidation into a common helper

Factor out code handling revalidation of bdev on disk change into a common helper.

---
 fs/block_dev.c | 26 ++++++++++++++------------
 1 file changed, 14 insertions(+), 12 deletions(-)
fs/block_dev.c +14 -12
index 9c073dbdc1b0..88c6d35ec71d 100644
@@ -1512,6 +1512,14 @@ EXPORT_SYMBOL(bd_set_size);
 
 static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part);
 
+static void bdev_disk_changed(struct block_device *bdev, bool invalidate)
+{
+	if (invalidate)
+		invalidate_partitions(bdev->bd_disk, bdev);
+	else
+		rescan_partitions(bdev->bd_disk, bdev);
+}
+
 /*
  * bd_mutex locking:
  *
@@ -1594,12 +1602,9 @@ static int __blkdev_get(struct block_device *bdev, fmode_t mode, int for_part)
 			 * The latter is necessary to prevent ghost
 			 * partitions on a removed medium.
 			 */
-			if (bdev->bd_invalidated) {
-				if (!ret)
-					rescan_partitions(disk, bdev);
-				else if (ret == -ENOMEDIUM)
-					invalidate_partitions(disk, bdev);
-			}
+			if (bdev->bd_invalidated &&
+			    (!ret || ret == -ENOMEDIUM))
+				bdev_disk_changed(bdev, ret == -ENOMEDIUM);
 
 			if (ret)
 				goto out_clear;
@@ -1632,12 +1637,9 @@ static int __blkdev_get(struct block_device *bdev, fmode_t mode, int for_part)
 			if (bdev->bd_disk->fops->open)
 				ret = bdev->bd_disk->fops->open(bdev, mode);
 			/* the same as first opener case, read comment there */
-			if (bdev->bd_invalidated) {
-				if (!ret)
-					rescan_partitions(bdev->bd_disk, bdev);
-				else if (ret == -ENOMEDIUM)
-					invalidate_partitions(bdev->bd_disk, bdev);
-			}
+			if (bdev->bd_invalidated &&
+			    (!ret || ret == -ENOMEDIUM))
+				bdev_disk_changed(bdev, ret == -ENOMEDIUM);
 			if (ret)
 				goto out_unlock_bdev;
 		}
Signature · 1 line
2.16.4
#2 lore raw

Re: [PATCH 1/2] bdev: Factor out bdev revalidation into a common helper

Quoted 30 lines · Jan Kara

Factor out code handling revalidation of bdev on disk change into a common helper.

---
 fs/block_dev.c | 26 ++++++++++++++------------
 1 file changed, 14 insertions(+), 12 deletions(-)
fs/block_dev.c +8 -0
index 9c073dbdc1b0..88c6d35ec71d 100644
@@ -1512,6 +1512,14 @@ EXPORT_SYMBOL(bd_set_size);
 
 static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part);
 
+static void bdev_disk_changed(struct block_device *bdev, bool invalidate)
+{
+	if (invalidate)
+		invalidate_partitions(bdev->bd_disk, bdev);
+	else
+		rescan_partitions(bdev->bd_disk, bdev);
+}
+
 /*
  * bd_mutex locking:
  *
@@ -1594,12 +1602,9 @@ static int __blkdev_get(struct block_device *bdev, fmode_t mode, int for_part)
 			 * The latter is necessary to prevent ghost
 			 * partitions on a removed medium.

fs/block_dev.c static int __blkdev_get(struct block_device *bdev, fmode_t mode, int for_part) · line 1607

Quoted 10 lines
fs/block_dev.c +3 -6
 			 */
-			if (bdev->bd_invalidated) {
-				if (!ret)
-					rescan_partitions(disk, bdev);
-				else if (ret == -ENOMEDIUM)
-					invalidate_partitions(disk, bdev);
-			}
+			if (bdev->bd_invalidated &&
+			    (!ret || ret == -ENOMEDIUM))
+				bdev_disk_changed(bdev, ret == -ENOMEDIUM);

This is a little confusing since from its name, bdev_disk_changed() seem to imply that a new disk is present but this is called only if bdev is invalidated... What about calling this simply bdev_revalidate_disk(), since rescan_partitions() will call the disk revalidate method.

Also, it seems to me that this function could be used without the complex "if ()" condition by slightly modifying it:

static void bdev_revalidate_disk(struct block_device *bdev,
bool invalidate)

{
	if (bdev->bd_invalidated && invalidate)
		invalidate_partitions(bdev->bd_disk, bdev);
	else
		rescan_partitions(bdev->bd_disk, bdev);
}

Otherwise, this all looks fine to me.

Quoted 20 lines
fs/block_dev.c +3 -6
 			if (ret)
 				goto out_clear;
@@ -1632,12 +1637,9 @@ static int __blkdev_get(struct block_device *bdev, fmode_t mode, int for_part)
 			if (bdev->bd_disk->fops->open)
 				ret = bdev->bd_disk->fops->open(bdev, mode);
 			/* the same as first opener case, read comment there */
-			if (bdev->bd_invalidated) {
-				if (!ret)
-					rescan_partitions(bdev->bd_disk, bdev);
-				else if (ret == -ENOMEDIUM)
-					invalidate_partitions(bdev->bd_disk, bdev);
-			}
+			if (bdev->bd_invalidated &&
+			    (!ret || ret == -ENOMEDIUM))
+				bdev_disk_changed(bdev, ret == -ENOMEDIUM);
 			if (ret)
 				goto out_unlock_bdev;
 		}
Signature · 2 lines
Damien Le Moal
Western Digital Research
author reviewer
#3 lore raw

Re: [PATCH 1/2] bdev: Factor out bdev revalidation into a common helper

Quoted 38 lines · Damien Le Moal
Quoted 37 lines · Jan Kara

Factor out code handling revalidation of bdev on disk change into a common helper.

---
 fs/block_dev.c | 26 ++++++++++++++------------
 1 file changed, 14 insertions(+), 12 deletions(-)
fs/block_dev.c +8 -6
index 9c073dbdc1b0..88c6d35ec71d 100644
@@ -1512,6 +1512,14 @@ EXPORT_SYMBOL(bd_set_size);
 
 static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part);
 
+static void bdev_disk_changed(struct block_device *bdev, bool invalidate)
+{
+	if (invalidate)
+		invalidate_partitions(bdev->bd_disk, bdev);
+	else
+		rescan_partitions(bdev->bd_disk, bdev);
+}
+
 /*
  * bd_mutex locking:
  *
@@ -1594,12 +1602,9 @@ static int __blkdev_get(struct block_device *bdev, fmode_t mode, int for_part)
 			 * The latter is necessary to prevent ghost
 			 * partitions on a removed medium.
 			 */
-			if (bdev->bd_invalidated) {
-				if (!ret)
-					rescan_partitions(disk, bdev);
-				else if (ret == -ENOMEDIUM)
-					invalidate_partitions(disk, bdev);
-			}
Quoted 8 lines

fs/block_dev.c static int __blkdev_get(struct block_device *bdev, fmode_t mode, int for_part) · line 1607

Quoted 3 lines
fs/block_dev.c +3 -0
+			if (bdev->bd_invalidated &&
+			    (!ret || ret == -ENOMEDIUM))
+				bdev_disk_changed(bdev, ret == -ENOMEDIUM);

This is a little confusing since from its name, bdev_disk_changed() seem to imply that a new disk is present but this is called only if bdev is invalidated... What about calling this simply bdev_revalidate_disk(), since rescan_partitions() will call the disk revalidate method.

Honestly, the whole disk revalidation code is confusing to me :) I had to draw a graph of which function calls which to understand what's going on in that code and I think it could really use some refactoring. But that's besides current point :)

Your "only if bdev is invalidated" is true but not actually a full story. ->bd_invalidated effectively gets set only through check_disk_change(). All other places that end up calling flush_disk() clear bd_invalidated shortly afterwards. So the function I've created is a direct counterpart to check_disk_change() that just needs to happen after the device is successfully open. I wanted to express that in the name - hence bdev_disk_changed(). So yes, bdev_disk_changed() should be called exactly when the new disk is present. It is bd_invalidated that is actually misnamed.

Now I'm not really tied to bdev_disk_changed(). bdev_revalidate_disk() seems a bit confusing to me though because the disk has actually been already revalidated in check_disk_change() and the function won't revalidate the disk for devices with partition scan disabled.

Quoted 5 lines

Also, it seems to me that this function could be used without the complex "if ()" condition by slightly modifying it:

static void bdev_revalidate_disk(struct block_device *bdev,
bool invalidate)

Quoted 8 lines
{
	if (bdev->bd_invalidated && invalidate)
		invalidate_partitions(bdev->bd_disk, bdev);
	else
		rescan_partitions(bdev->bd_disk, bdev);
}

Otherwise, this all looks fine to me.

Well, but you don't want to call rescan_partitions() if bd_invalidated is not set. But yes, we could move bd_invalidated check into bdev_disk_changed(), but then it would seem less clear why this function is getting called. This ties somewhat to the discussion above. Hum, I guess if we call the function just bdev_revalidate(), the name won't be confusing and it would then make sense to move the bd_invalidated condition in there.

Honza

Signature · 2 lines
Jan Kara <[email protected]>
SUSE Labs, CR
#4 lore raw

Re: [PATCH 1/2] bdev: Factor out bdev revalidation into a common helper

Quoted 48 lines · Jan Kara
Quoted 38 lines · Damien Le Moal
Quoted 37 lines · Jan Kara

Factor out code handling revalidation of bdev on disk change into a common helper.

---
 fs/block_dev.c | 26 ++++++++++++++------------
 1 file changed, 14 insertions(+), 12 deletions(-)
fs/block_dev.c +8 -6
index 9c073dbdc1b0..88c6d35ec71d 100644
@@ -1512,6 +1512,14 @@ EXPORT_SYMBOL(bd_set_size);
 
 static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part);
 
+static void bdev_disk_changed(struct block_device *bdev, bool invalidate)
+{
+	if (invalidate)
+		invalidate_partitions(bdev->bd_disk, bdev);
+	else
+		rescan_partitions(bdev->bd_disk, bdev);
+}
+
 /*
  * bd_mutex locking:
  *
@@ -1594,12 +1602,9 @@ static int __blkdev_get(struct block_device *bdev, fmode_t mode, int for_part)
 			 * The latter is necessary to prevent ghost
 			 * partitions on a removed medium.
 			 */
-			if (bdev->bd_invalidated) {
-				if (!ret)
-					rescan_partitions(disk, bdev);
-				else if (ret == -ENOMEDIUM)
-					invalidate_partitions(disk, bdev);
-			}
Quoted 8 lines

fs/block_dev.c static int __blkdev_get(struct block_device *bdev, fmode_t mode, int for_part) · line 1607

Quoted 3 lines
fs/block_dev.c +3 -0
+			if (bdev->bd_invalidated &&
+			    (!ret || ret == -ENOMEDIUM))
+				bdev_disk_changed(bdev, ret == -ENOMEDIUM);

This is a little confusing since from its name, bdev_disk_changed() seem to imply that a new disk is present but this is called only if bdev is invalidated... What about calling this simply bdev_revalidate_disk(), since rescan_partitions() will call the disk revalidate method.

Quoted 4 lines

Honestly, the whole disk revalidation code is confusing to me :) I had to draw a graph of which function calls which to understand what's going on in that code and I think it could really use some refactoring. But that's besides current point :)

OK. I guess I got confused too...

Quoted 10 lines

Your "only if bdev is invalidated" is true but not actually a full story. ->bd_invalidated effectively gets set only through check_disk_change(). All other places that end up calling flush_disk() clear bd_invalidated shortly afterwards. So the function I've created is a direct counterpart to check_disk_change() that just needs to happen after the device is successfully open. I wanted to express that in the name - hence bdev_disk_changed(). So yes, bdev_disk_changed() should be called exactly when the new disk is present. It is bd_invalidated that is actually misnamed.

Got it.

Quoted 17 lines

Now I'm not really tied to bdev_disk_changed(). bdev_revalidate_disk() seems a bit confusing to me though because the disk has actually been already revalidated in check_disk_change() and the function won't revalidate the disk for devices with partition scan disabled.>

Quoted 12 lines

Also, it seems to me that this function could be used without the complex "if ()" condition by slightly modifying it:

static void bdev_revalidate_disk(struct block_device *bdev,
bool invalidate)

{
	if (bdev->bd_invalidated && invalidate)
		invalidate_partitions(bdev->bd_disk, bdev);
	else
		rescan_partitions(bdev->bd_disk, bdev);
}
Quoted 8 lines
Quoted 1 line

Otherwise, this all looks fine to me.

Well, but you don't want to call rescan_partitions() if bd_invalidated is not set. But yes, we could move bd_invalidated check into bdev_disk_changed(), but then it would seem less clear why this function is getting called. This ties somewhat to the discussion above. Hum, I guess if we call the function just bdev_revalidate(), the name won't be confusing and it would then make sense to move the bd_invalidated condition in there.

Indeed, we don't want another partition scan. Missed that one. For the function name, since the goal is to revalidate only the bdev capacity, what about bdev_revalidate_capacity() then ? But looking at the code and seeing the partitions functions calls does not clarify things though. Oh well, keep the name you proposed and we can cleanup everything with a refactor :)

Cheers.

Signature · 2 lines
Damien Le Moal
Western Digital Research
author reviewer
#5 lore raw

PATCH 2/2 bdev: Refresh bdev size for disks without partitioning

In reply to Jan Kara

Currently, block device size in not updated on second and further open for block devices where partition scan is disabled. This is particularly annoying for example for DVD drives as that means block device size does not get updated once the media is inserted into a drive if the device is already open when inserting the media. This is actually always the case for example when pktcdvd is in use.

Fix the problem by revalidating block device size on every open even for devices with partition scan disabled.

---
 fs/block_dev.c | 19 ++++++++++---------
 1 file changed, 10 insertions(+), 9 deletions(-)
fs/block_dev.c +10 -9
index 88c6d35ec71d..d612468ee66b 100644
@@ -1403,11 +1403,7 @@ static void flush_disk(struct block_device *bdev, bool kill_dirty)
 		       "resized disk %s\n",
 		       bdev->bd_disk ? bdev->bd_disk->disk_name : "");
 	}
-
-	if (!bdev->bd_disk)
-		return;
-	if (disk_part_scan_enabled(bdev->bd_disk))
-		bdev->bd_invalidated = 1;
+	bdev->bd_invalidated = 1;
 }
 
 /**
@@ -1514,10 +1510,15 @@ static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part);
 
 static void bdev_disk_changed(struct block_device *bdev, bool invalidate)
 {
-	if (invalidate)
-		invalidate_partitions(bdev->bd_disk, bdev);
-	else
-		rescan_partitions(bdev->bd_disk, bdev);
+	if (disk_part_scan_enabled(bdev->bd_disk)) {
+		if (invalidate)
+			invalidate_partitions(bdev->bd_disk, bdev);
+		else
+			rescan_partitions(bdev->bd_disk, bdev);
+	} else {
+		check_disk_size_change(bdev->bd_disk, bdev, !invalidate);
+		bdev->bd_invalidated = 0;
+	}
 }
 
 /*
Signature · 1 line
2.16.4
#6 lore raw
Quoted 35 lines · Jan Kara

Currently, block device size in not updated on second and further open for block devices where partition scan is disabled. This is particularly annoying for example for DVD drives as that means block device size does not get updated once the media is inserted into a drive if the device is already open when inserting the media. This is actually always the case for example when pktcdvd is in use.

Fix the problem by revalidating block device size on every open even for devices with partition scan disabled.

---
  fs/block_dev.c | 19 ++++++++++---------
  1 file changed, 10 insertions(+), 9 deletions(-)
fs/block_dev.c +1 -5
index 88c6d35ec71d..d612468ee66b 100644
@@ -1403,11 +1403,7 @@ static void flush_disk(struct block_device *bdev, bool kill_dirty)
  		       "resized disk %s\n",
  		       bdev->bd_disk ? bdev->bd_disk->disk_name : "");
  	}
-
-	if (!bdev->bd_disk)
-		return;
-	if (disk_part_scan_enabled(bdev->bd_disk))
-		bdev->bd_invalidated = 1;
+	bdev->bd_invalidated = 1;
  }
 
  /**
@@ -1514,10 +1510,15 @@ static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part);
 
  static void bdev_disk_changed(struct block_device *bdev, bool invalidate)

fs/block_dev.c static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part); · line 1517

Quoted 10 lines
fs/block_dev.c +5 -4
  {
-	if (invalidate)
-		invalidate_partitions(bdev->bd_disk, bdev);
-	else
-		rescan_partitions(bdev->bd_disk, bdev);
+	if (disk_part_scan_enabled(bdev->bd_disk)) {
+		if (invalidate)
+			invalidate_partitions(bdev->bd_disk, bdev);
+		else
+			rescan_partitions(bdev->bd_disk, bdev);

Maybe use the new common helper to replace above.

Thanks,
Guoqing

Quoted 7 lines
fs/block_dev.c +4 -0
+	} else {
+		check_disk_size_change(bdev->bd_disk, bdev, !invalidate);
+		bdev->bd_invalidated = 0;
+	}
  }
 
  /*
author reviewer
#7 lore raw
Quoted 40 lines · Guoqing Jiang
Quoted 37 lines · Jan Kara

Currently, block device size in not updated on second and further open for block devices where partition scan is disabled. This is particularly annoying for example for DVD drives as that means block device size does not get updated once the media is inserted into a drive if the device is already open when inserting the media. This is actually always the case for example when pktcdvd is in use.

Fix the problem by revalidating block device size on every open even for devices with partition scan disabled.

---
  fs/block_dev.c | 19 ++++++++++---------
  1 file changed, 10 insertions(+), 9 deletions(-)
fs/block_dev.c +1 -8
index 88c6d35ec71d..d612468ee66b 100644
@@ -1403,11 +1403,7 @@ static void flush_disk(struct block_device *bdev, bool kill_dirty)
  		       "resized disk %s\n",
  		       bdev->bd_disk ? bdev->bd_disk->disk_name : "");
  	}
-
-	if (!bdev->bd_disk)
-		return;
-	if (disk_part_scan_enabled(bdev->bd_disk))
-		bdev->bd_invalidated = 1;
+	bdev->bd_invalidated = 1;
  }
  /**
@@ -1514,10 +1510,15 @@ static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part);
  static void bdev_disk_changed(struct block_device *bdev, bool invalidate)
  {
-	if (invalidate)
-		invalidate_partitions(bdev->bd_disk, bdev);
-	else
Quoted 8 lines

fs/block_dev.c static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part); · line 1517

Quoted 6 lines
fs/block_dev.c +5 -1
-		rescan_partitions(bdev->bd_disk, bdev);
+	if (disk_part_scan_enabled(bdev->bd_disk)) {
+		if (invalidate)
+			invalidate_partitions(bdev->bd_disk, bdev);
+		else
+			rescan_partitions(bdev->bd_disk, bdev);

Maybe use the new common helper to replace above.

What do you mean exactly? Because there's only this place that has the code pattern here...

Honza

Quoted 7 lines
Quoted 6 lines
fs/block_dev.c +4 -0
+	} else {
+		check_disk_size_change(bdev->bd_disk, bdev, !invalidate);
+		bdev->bd_invalidated = 0;
+	}
  }
  /*
Signature · 2 lines
Jan Kara <[email protected]>
SUSE Labs, CR
#8 lore raw
Quoted 47 lines · Jan Kara
Quoted 38 lines · Guoqing Jiang
Quoted 36 lines · Jan Kara

Currently, block device size in not updated on second and further open for block devices where partition scan is disabled. This is particularly annoying for example for DVD drives as that means block device size does not get updated once the media is inserted into a drive if the device is already open when inserting the media. This is actually always the case for example when pktcdvd is in use.

Fix the problem by revalidating block device size on every open even for devices with partition scan disabled.

---
   fs/block_dev.c | 19 ++++++++++---------
   1 file changed, 10 insertions(+), 9 deletions(-)
fs/block_dev.c +1 -7
index 88c6d35ec71d..d612468ee66b 100644
@@ -1403,11 +1403,7 @@ static void flush_disk(struct block_device *bdev, bool kill_dirty)
   		       "resized disk %s\n",
   		       bdev->bd_disk ? bdev->bd_disk->disk_name : "");
   	}
-
-	if (!bdev->bd_disk)
-		return;
-	if (disk_part_scan_enabled(bdev->bd_disk))
-		bdev->bd_invalidated = 1;
+	bdev->bd_invalidated = 1;
   }
   /**
@@ -1514,10 +1510,15 @@ static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part);
   static void bdev_disk_changed(struct block_device *bdev, bool invalidate)
   {
-	if (invalidate)
-		invalidate_partitions(bdev->bd_disk, bdev);
Quoted 8 lines

fs/block_dev.c static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part); · line 1517

Quoted 7 lines
fs/block_dev.c +5 -2
-	else
-		rescan_partitions(bdev->bd_disk, bdev);
+	if (disk_part_scan_enabled(bdev->bd_disk)) {
+		if (invalidate)
+			invalidate_partitions(bdev->bd_disk, bdev);
+		else
+			rescan_partitions(bdev->bd_disk, bdev);

Maybe use the new common helper to replace above.

Quoted 2 lines

What do you mean exactly? Because there's only this place that has the code pattern here...

The above looks same as the new function.

+static void bdev_disk_changed(struct block_device *bdev, bool invalidate) +{
+ if (invalidate)
+ invalidate_partitions(bdev->bd_disk, bdev);
+ else
+ rescan_partitions(bdev->bd_disk, bdev);
+}

So maybe just call it in the 'if' branch, am I misunderstand something?

static void bdev_disk_changed(struct block_device *bdev, bool invalidate)

{
	if (disk_part_scan_enabled(bdev->bd_disk))	
		bdev_disk_changed(bdev, invalidate)
	else {
		check_disk_size_change(bdev->bd_disk, bdev, !invalidate);
		bdev->bd_invalidated = 0;
	}
}

Thanks,
Guoqing

#9 lore raw
Quoted 12 lines · Guoqing Jiang
Quoted 12 lines
Quoted 12 lines
Quoted 12 lines

   static void bdev_disk_changed(struct block_device *bdev, bool invalidate)
   {
-    if (invalidate)
-        invalidate_partitions(bdev->bd_disk, bdev);
-    else
-        rescan_partitions(bdev->bd_disk, bdev);
+    if (disk_part_scan_enabled(bdev->bd_disk)) {
+        if (invalidate)
+            invalidate_partitions(bdev->bd_disk, bdev); +        else
+            rescan_partitions(bdev->bd_disk, bdev);

Quoted 6 lines
Quoted 4 lines
Quoted 1 line

Maybe use the new common helper to replace above.

What do you mean exactly? Because there's only this place that has the code
pattern here...

The above looks same as the new function.

The above is changed in the new bdev_disk_changed() function. Patch 1/2 factors out the pattern and this patch changes the behaviour

Signature · 9 lines
Johannes Thumshirn                            SUSE Labs Filesystems
[email protected]                                +49 911 74053 689
SUSE Software Solutions Germany GmbH
Maxfeldstr. 5
90409 Nürnberg
Germany
(HRB 36809, AG Nürnberg)
Geschäftsführer: Felix Imendörffer
Key fingerprint = EC38 9CAB C2C4 F25D 8600 D0D0 0393 969D 2D76 0850
#10 lore raw
Quoted 13 lines · Johannes Thumshirn
Quoted 12 lines · Guoqing Jiang
Quoted 12 lines
Quoted 12 lines
Quoted 12 lines

   static void bdev_disk_changed(struct block_device *bdev, bool
invalidate)
   {
-    if (invalidate)
-        invalidate_partitions(bdev->bd_disk, bdev);
-    else
-        rescan_partitions(bdev->bd_disk, bdev);
+    if (disk_part_scan_enabled(bdev->bd_disk)) {
+        if (invalidate)
+            invalidate_partitions(bdev->bd_disk, bdev); +        else
+            rescan_partitions(bdev->bd_disk, bdev);

Quoted 8 lines
Quoted 5 lines
Quoted 4 lines
Quoted 1 line

Maybe use the new common helper to replace above.

What do you mean exactly? Because there's only this place that has the code
pattern here...

The above looks same as the new function.

The above is changed in the new bdev_disk_changed() function. Patch 1/2 factors out the pattern and this patch changes the behaviour

Oops, yes, sorry for the noise.

Thanks,
Guoqing

author reviewer
#11 lore raw
In reply to Guoqing Jiang
Quoted 63 lines · Guoqing Jiang
Quoted 47 lines · Jan Kara
Quoted 38 lines · Guoqing Jiang
Quoted 36 lines · Jan Kara

Currently, block device size in not updated on second and further open for block devices where partition scan is disabled. This is particularly annoying for example for DVD drives as that means block device size does not get updated once the media is inserted into a drive if the device is already open when inserting the media. This is actually always the case for example when pktcdvd is in use.

Fix the problem by revalidating block device size on every open even for devices with partition scan disabled.

---
   fs/block_dev.c | 19 ++++++++++---------
   1 file changed, 10 insertions(+), 9 deletions(-)
fs/block_dev.c +1 -7
index 88c6d35ec71d..d612468ee66b 100644
@@ -1403,11 +1403,7 @@ static void flush_disk(struct block_device *bdev, bool kill_dirty)
   		       "resized disk %s\n",
   		       bdev->bd_disk ? bdev->bd_disk->disk_name : "");
   	}
-
-	if (!bdev->bd_disk)
-		return;
-	if (disk_part_scan_enabled(bdev->bd_disk))
-		bdev->bd_invalidated = 1;
+	bdev->bd_invalidated = 1;
   }
   /**
@@ -1514,10 +1510,15 @@ static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part);
   static void bdev_disk_changed(struct block_device *bdev, bool invalidate)
   {
-	if (invalidate)
-		invalidate_partitions(bdev->bd_disk, bdev);
Quoted 8 lines

fs/block_dev.c static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part); · line 1517

Quoted 7 lines
fs/block_dev.c +5 -2
-	else
-		rescan_partitions(bdev->bd_disk, bdev);
+	if (disk_part_scan_enabled(bdev->bd_disk)) {
+		if (invalidate)
+			invalidate_partitions(bdev->bd_disk, bdev);
+		else
+			rescan_partitions(bdev->bd_disk, bdev);

Maybe use the new common helper to replace above.

Quoted 2 lines

What do you mean exactly? Because there's only this place that has the code pattern here...

The above looks same as the new function.

fs/block_dev.c +7 -0
+static void bdev_disk_changed(struct block_device *bdev, bool invalidate)
+{
+	if (invalidate)
+		invalidate_partitions(bdev->bd_disk, bdev);
+	else
+		rescan_partitions(bdev->bd_disk, bdev);
+}
Quoted 1 line

So maybe just call it in the 'if' branch, am I misunderstand something?

I think you misparsed the diff ;) The code that you mention as duplicated just gets reindented by the patch.

Quoted 9 lines

static void bdev_disk_changed(struct block_device *bdev, bool invalidate)

{
	if (disk_part_scan_enabled(bdev->bd_disk))	
		bdev_disk_changed(bdev, invalidate)
	else {
		check_disk_size_change(bdev->bd_disk, bdev, !invalidate);
		bdev->bd_invalidated = 0;
	}
}

This has infinite recursion in it - you call bdev_disk_changed() from bdev_disk_changed().

Honza

Signature · 2 lines
Jan Kara <[email protected]>
SUSE Labs, CR
#12 lore raw
Quoted 74 lines · Jan Kara
Quoted 60 lines · Guoqing Jiang
Quoted 46 lines · Jan Kara
Quoted 37 lines · Guoqing Jiang
Quoted 36 lines · Jan Kara

Currently, block device size in not updated on second and further open for block devices where partition scan is disabled. This is particularly annoying for example for DVD drives as that means block device size does not get updated once the media is inserted into a drive if the device is already open when inserting the media. This is actually always the case for example when pktcdvd is in use.

Fix the problem by revalidating block device size on every open even for devices with partition scan disabled.

---
    fs/block_dev.c | 19 ++++++++++---------
    1 file changed, 10 insertions(+), 9 deletions(-)
fs/block_dev.c +1 -7
index 88c6d35ec71d..d612468ee66b 100644
@@ -1403,11 +1403,7 @@ static void flush_disk(struct block_device *bdev, bool kill_dirty)
    		       "resized disk %s\n",
    		       bdev->bd_disk ? bdev->bd_disk->disk_name : "");
    	}
-
-	if (!bdev->bd_disk)
-		return;
-	if (disk_part_scan_enabled(bdev->bd_disk))
-		bdev->bd_invalidated = 1;
+	bdev->bd_invalidated = 1;
    }
    /**
@@ -1514,10 +1510,15 @@ static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part);
    static void bdev_disk_changed(struct block_device *bdev, bool invalidate)
    {
-	if (invalidate)
-		invalidate_partitions(bdev->bd_disk, bdev);
Quoted 8 lines

fs/block_dev.c static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part); · line 1517

Quoted 7 lines
fs/block_dev.c +5 -2
-	else
-		rescan_partitions(bdev->bd_disk, bdev);
+	if (disk_part_scan_enabled(bdev->bd_disk)) {
+		if (invalidate)
+			invalidate_partitions(bdev->bd_disk, bdev);
+		else
+			rescan_partitions(bdev->bd_disk, bdev);

Maybe use the new common helper to replace above.

Quoted 2 lines

What do you mean exactly? Because there's only this place that has the code pattern here...

The above looks same as the new function.

fs/block_dev.c +7 -0
+static void bdev_disk_changed(struct block_device *bdev, bool invalidate)
+{
+	if (invalidate)
+		invalidate_partitions(bdev->bd_disk, bdev);
+	else
+		rescan_partitions(bdev->bd_disk, bdev);
+}
Quoted 1 line

So maybe just call it in the 'if' branch, am I misunderstand something?

I think you misparsed the diff ;) The code that you mention as duplicated just gets reindented by the patch.

Quoted 9 lines

static void bdev_disk_changed(struct block_device *bdev, bool invalidate)

{
	if (disk_part_scan_enabled(bdev->bd_disk))	
		bdev_disk_changed(bdev, invalidate)
	else {
		check_disk_size_change(bdev->bd_disk, bdev, !invalidate);
		bdev->bd_invalidated = 0;
	}
}
Quoted 2 lines

This has infinite recursion in it - you call bdev_disk_changed() from bdev_disk_changed().

Right :-[, I need some coffee to refresh, sorry for the noise.

BRs,
Guoqing

maintainer
#13 lore raw
In reply to Jan Kara
Quoted 9 lines · Jan Kara

Hello,

I've been debugging for quite a long time strange issues with encrypted DVDs not being readable on Linux (bko#194965). In the end I've tracked down the problem to the fact that block device size is not updated when the media is inserted in case the DVD device is already open. This is IMO a bug in block device code as the size gets properly update in case the device has partition scanning enabled. The following series fixes the problem by refreshing disk size on each open even for devices with partition scanning disabled.

It's really confusing to have different behavior for partition vs whole device. This series looks good to me, the size change code is really hard to follow.

I don't see any serious objections here, I'm going to queue this up for 5.4.

Signature · 1 line
Jens Axboe
author reviewer
#14 lore raw
Quoted 11 lines · Jens Axboe
Quoted 9 lines · Jan Kara

Hello,

I've been debugging for quite a long time strange issues with encrypted DVDs not being readable on Linux (bko#194965). In the end I've tracked down the problem to the fact that block device size is not updated when the media is inserted in case the DVD device is already open. This is IMO a bug in block device code as the size gets properly update in case the device has partition scanning enabled. The following series fixes the problem by refreshing disk size on each open even for devices with partition scanning disabled.

Quoted 5 lines

It's really confusing to have different behavior for partition vs whole device. This series looks good to me, the size change code is really hard to follow.

I don't see any serious objections here, I'm going to queue this up for 5.4.

Thanks Jens! I'll look into refactoring the size change / revalidation code so that it's easier to understand what's going on...

Honza

Signature · 2 lines
Jan Kara <[email protected]>
SUSE Labs, CR
author reviewer
#16 lore raw
Quoted 8 lines · Christoph Hellwig
Quoted 2 lines · Jan Kara

Thanks Jens! I'll look into refactoring the size change / revalidation code so that it's easier to understand what's going on...

I actualy have a series for this. I've started rebasing it on top of you work and will need to do some testing. My current WIP is here:

http://git.infradead.org/users/hch/block.git/shortlog/refs/heads/partition-cleanup

Cool. Thanks for that! Skimming over the series it looks good to me. The only remaining thing I wanted to look into is bdev->bd_invalidated handling. Because the only thing it actually indicates in practice is that DISK_EVENT_MEDIA_CHANGE was set in check_disk_change(). All the other call chains end up clearing bdev->bd_invalidated before it has any effect. And that just looks very cryptic to me... So my plan was to at least move the setting of bdev->bd_invalidated to check_disk_change() and rename it to something saner if I cannot come up with anything better for propagating the information from check_disk_change() to __blkdev_get().

Honza

Signature · 2 lines
Jan Kara <[email protected]>
SUSE Labs, CR