[2/4] btrfs: delete identified alien device in open_fs_devices
diff mbox series

Message ID 1570175403-4073-3-git-send-email-anand.jain@oracle.com
State New
Headers show
Series
  • btrfs: fix issues due to alien device
Related show

Commit Message

Anand Jain Oct. 4, 2019, 7:50 a.m. UTC
In open_fs_devices() we identify alien device but we don't reset its
the device::name. So progs device list does not show the device missing
as shown in the script below.

mkfs.btrfs -fq /dev/sdd && mount /dev/sdd /btrfs
mkfs.btrfs -fq -draid1 -mraid1 /dev/sdc /dev/sdb
sleep 3 # avoid racing with udev's useless scans if needed
btrfs dev add -f /dev/sdb /btrfs
mount -o degraded /dev/sdc /btrfs1

No missing device:
btrfs fi show -m /btrfs1
Label: none  uuid: 3eb7cd50-4594-458f-9d68-c243cc49954d
	Total devices 2 FS bytes used 128.00KiB
	devid    1 size 12.00GiB used 1.26GiB path /dev/sdc
	devid    2 size 12.00GiB used 1.26GiB path /dev/sdb

Signed-off-by: Anand Jain <anand.jain@oracle.com>
---
PS: Fundamentally its wrong approach that btrfs-progs deduces the device
missing state in the userland instead of obtaining it from the kernel.
I objected on the patch, but still those patches got merged, this bug is
one of its side effects. Ironically I wrote patches to read device_state
from the kernel using ioctl, procfs and sysfs but didn't get the due
attention till a merger.

 fs/btrfs/volumes.c | 13 ++++++++++---
 1 file changed, 10 insertions(+), 3 deletions(-)

Comments

Nikolay Borisov Oct. 4, 2019, 8:18 a.m. UTC | #1
On 4.10.19 г. 10:50 ч., Anand Jain wrote:
> In open_fs_devices() we identify alien device but we don't reset its
> the device::name. So progs device list does not show the device missing
> as shown in the script below.
> 
> mkfs.btrfs -fq /dev/sdd && mount /dev/sdd /btrfs
> mkfs.btrfs -fq -draid1 -mraid1 /dev/sdc /dev/sdb
> sleep 3 # avoid racing with udev's useless scans if needed
> btrfs dev add -f /dev/sdb /btrfs
> mount -o degraded /dev/sdc /btrfs1
> 
> No missing device:
> btrfs fi show -m /btrfs1
> Label: none  uuid: 3eb7cd50-4594-458f-9d68-c243cc49954d
> 	Total devices 2 FS bytes used 128.00KiB
> 	devid    1 size 12.00GiB used 1.26GiB path /dev/sdc
> 	devid    2 size 12.00GiB used 1.26GiB path /dev/sdb
> 
> Signed-off-by: Anand Jain <anand.jain@oracle.com>
> ---
> PS: Fundamentally its wrong approach that btrfs-progs deduces the device
> missing state in the userland instead of obtaining it from the kernel.
> I objected on the patch, but still those patches got merged, this bug is
> one of its side effects. Ironically I wrote patches to read device_state
> from the kernel using ioctl, procfs and sysfs but didn't get the due
> attention till a merger.
> 
>  fs/btrfs/volumes.c | 13 ++++++++++---
>  1 file changed, 10 insertions(+), 3 deletions(-)
> 
> diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
> index 06ec3577c6b4..05ade8c7342b 100644
> --- a/fs/btrfs/volumes.c
> +++ b/fs/btrfs/volumes.c
> @@ -803,10 +803,10 @@ static int btrfs_open_one_device(struct btrfs_fs_devices *fs_devices,
>  	disk_super = (struct btrfs_super_block *)bh->b_data;
>  	devid = btrfs_stack_device_id(&disk_super->dev_item);
>  	if (devid != device->devid)
> -		goto error_brelse;
> +		goto free_alien;
>  
>  	if (memcmp(device->uuid, disk_super->dev_item.uuid, BTRFS_UUID_SIZE))
> -		goto error_brelse;
> +		goto free_alien;
>  

Imo a better approach is to return a particular error code and do the
deletion in open_fs_devices. Otherwise it's not apparent why you use
list_for_each_entry_safe in one function to delete something in a
different one (whose name by the way doesn't suggest a deletion is going
on). Looking at the error I think enodev/enxio is appropriate.

>  	device->generation = btrfs_super_generation(disk_super);
>  
> @@ -845,6 +845,11 @@ static int btrfs_open_one_device(struct btrfs_fs_devices *fs_devices,
>  
>  	return 0;
>  
> +free_alien:
> +	fs_devices->num_devices--;
> +	list_del(&device->dev_list);
> +	btrfs_free_device(device);
> +
>  error_brelse:
>  	brelse(bh);
>  	blkdev_put(bdev, flags);
> @@ -1329,11 +1334,13 @@ static int open_fs_devices(struct btrfs_fs_devices *fs_devices,
>  				fmode_t flags, void *holder)
>  {
>  	struct btrfs_device *device;
> +	struct btrfs_device *tmp_device;
>  	struct btrfs_device *latest_dev = NULL;
>  
>  	flags |= FMODE_EXCL;
>  
> -	list_for_each_entry(device, &fs_devices->devices, dev_list) {
> +	list_for_each_entry_safe(device, tmp_device, &fs_devices->devices,
> +				 dev_list) {
>  		/* Just open everything we can; ignore failures here */
>  		if (btrfs_open_one_device(fs_devices, device, flags, holder))
>  			continue;
>
Anand Jain Oct. 6, 2019, 2:51 a.m. UTC | #2
>> diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
>> index 06ec3577c6b4..05ade8c7342b 100644
>> --- a/fs/btrfs/volumes.c
>> +++ b/fs/btrfs/volumes.c
>> @@ -803,10 +803,10 @@ static int btrfs_open_one_device(struct btrfs_fs_devices *fs_devices,
>>   	disk_super = (struct btrfs_super_block *)bh->b_data;
>>   	devid = btrfs_stack_device_id(&disk_super->dev_item);
>>   	if (devid != device->devid)
>> -		goto error_brelse;
>> +		goto free_alien;
>>   
>>   	if (memcmp(device->uuid, disk_super->dev_item.uuid, BTRFS_UUID_SIZE))
>> -		goto error_brelse;
>> +		goto free_alien;
>>   
> 
> Imo a better approach is to return a particular error code and do the
> deletion in open_fs_devices. Otherwise it's not apparent why you use
> list_for_each_entry_safe in one function to delete something in a
> different one (whose name by the way doesn't suggest a deletion is going
> on). Looking at the error I think enodev/enxio is appropriate.

  Agreed. And used -EUCLEAN I think that's most appropriate.
  Thanks, Anand

Patch
diff mbox series

diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index 06ec3577c6b4..05ade8c7342b 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -803,10 +803,10 @@  static int btrfs_open_one_device(struct btrfs_fs_devices *fs_devices,
 	disk_super = (struct btrfs_super_block *)bh->b_data;
 	devid = btrfs_stack_device_id(&disk_super->dev_item);
 	if (devid != device->devid)
-		goto error_brelse;
+		goto free_alien;
 
 	if (memcmp(device->uuid, disk_super->dev_item.uuid, BTRFS_UUID_SIZE))
-		goto error_brelse;
+		goto free_alien;
 
 	device->generation = btrfs_super_generation(disk_super);
 
@@ -845,6 +845,11 @@  static int btrfs_open_one_device(struct btrfs_fs_devices *fs_devices,
 
 	return 0;
 
+free_alien:
+	fs_devices->num_devices--;
+	list_del(&device->dev_list);
+	btrfs_free_device(device);
+
 error_brelse:
 	brelse(bh);
 	blkdev_put(bdev, flags);
@@ -1329,11 +1334,13 @@  static int open_fs_devices(struct btrfs_fs_devices *fs_devices,
 				fmode_t flags, void *holder)
 {
 	struct btrfs_device *device;
+	struct btrfs_device *tmp_device;
 	struct btrfs_device *latest_dev = NULL;
 
 	flags |= FMODE_EXCL;
 
-	list_for_each_entry(device, &fs_devices->devices, dev_list) {
+	list_for_each_entry_safe(device, tmp_device, &fs_devices->devices,
+				 dev_list) {
 		/* Just open everything we can; ignore failures here */
 		if (btrfs_open_one_device(fs_devices, device, flags, holder))
 			continue;