Re: [PATCH 3/6] btrfs: split-brain case for scanned changing device with INCOMPAT_METADATA_UUID

[Nikolay Borisov <nborisov@suse.com>]

On 12.12.19 г. 13:01 ч., damenly.su@gmail.com wrote:
> From: Su Yue <Damenly_Su@gmx.com>
> 
> This patch adds the case for scanned changing device with
> INCOMPAT_METADATA_UUID.
> For this situation, the origin code only handles the case
> the devices already pulled into disk with INCOMPAT_METADATA_UUID set.
> There is an another case that the successful changed devices synced
> without INCOMPAT_METADATA_UUID.
> So add the check of Heather fsid of scanned device equals
> metadata_uuid of fs_devices which is with INCOMPAT_METADATA_UUID
> feature.
> 

This is hard for me to parse and correctly understand what you mean.

> Signed-off-by: Su Yue <Damenly_Su@gmx.com>
> ---
>  fs/btrfs/volumes.c | 29 ++++++++++++++++++++++++++---
>  1 file changed, 26 insertions(+), 3 deletions(-)
> 
> diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
> index b08b06a89a77..61b4a107bb58 100644
> --- a/fs/btrfs/volumes.c
> +++ b/fs/btrfs/volumes.c
> @@ -654,7 +654,6 @@ static struct btrfs_fs_devices *find_fsid_inprogress(
>  	return NULL;
>  }
>  
> -
>  static struct btrfs_fs_devices *find_fsid_changed(
>  					struct btrfs_super_block *disk_super)


find_fsid_changed handles the case where a device belongs to a
filesystem which had multiple successful fsid changed but it failed on
the last one.

>  {
> @@ -663,9 +662,14 @@ static struct btrfs_fs_devices *find_fsid_changed(
>  	/*
>  	 * Handles the case where scanned device is part of an fs that had
>  	 * multiple successful changes of FSID but curently device didn't
> -	 * observe it. Meaning our fsid will be different than theirs.
> +	 * observe it.
> +	 *
> +	 * Case 1: the devices already changed still owns the feature, their
> +	 * fsid must differ from the disk_super->fsid.

What do you mean by device to still owns the feature? Has the bit set or
something else?

>  	 */
>  	list_for_each_entry(fs_devices, &fs_uuids, fs_list) {
> +		if (fs_devices->fsid_change)
> +			continue;

Why do you do this?

>  		if (memcmp(fs_devices->metadata_uuid, fs_devices->fsid,
>  			   BTRFS_FSID_SIZE) != 0 &&
>  		    memcmp(fs_devices->metadata_uuid, disk_super->metadata_uuid,
> @@ -676,7 +680,26 @@ static struct btrfs_fs_devices *find_fsid_changed(
>  		}
>  	}
>  
> -	return NULL;
> +	/*
> +	 * Case 2: the synced devices doesn't have the metadata_uuid feature.
> +	 * NOTE: the fs_devices has same metadata_uuid and fsid in memory, but
> +	 * they differs in disk, because fs_id is copied to
> +	 * fs_devices->metadata_id while alloc_fs_devices if no metadata

It's not possible for the device to have metadata_uuid feature because
this function is called from device_list_add iff the device has
METADATA_UUID flag:

if (fsid_change_in_progress) {

if (!has_metadata_uuid) {
} else {
 find_fsid_changed <-- here we are sure our device has METADATA_UUID set.
}
}

> +	 * feature.
> +	 */
> +	list_for_each_entry(fs_devices, &fs_uuids, fs_list) {
> +		if (memcmp(fs_devices->metadata_uuid, fs_devices->fsid,
> +			   BTRFS_FSID_SIZE) == 0 &&
> +		    memcmp(fs_devices->fsid, disk_super->metadata_uuid,
> +			   BTRFS_FSID_SIZE) == 0 && !fs_devices->fsid_change)
> +			return fs_devices;
> +	}
> +
> +	/*
> +	 * Okay, can't found any fs_devices already synced, back to
> +	 * search devices unchanged or changing like the device.
> +	 */
> +	return find_fsid(disk_super->fsid, disk_super->metadata_uuid);
>  }
>  
>  static struct btrfs_fs_devices *find_fsid_changing_metada_uuid(
>