Re: [PATCH] Btrfs: avoid losing data raid profile when deleting a device

[Liu Bo <bo.li.liu@oracle.com>]

On Tue, Oct 10, 2017 at 09:57:46AM +0300, Nikolay Borisov wrote:
> 
> 
> On  9.10.2017 21:01, Liu Bo wrote:
> > We've avoided data losing raid profile when doing balance, but it
> > turns out that deleting a device could also result in the same
> > problem.
> > 
> > This fixes the problem by creating an empty data chunk before
> > relocating the data chunk.
> > 
> > Metadata/System chunk are supposed to have non-zero bytes all the time
> > so their raid profile is persistent.
> 
> This patch introduces new warning:
> 
> fs/btrfs/volumes.c:3523:29: note: ‘trans’ was declared here
>   struct btrfs_trans_handle *trans;
>

Not sure how I missed this, thanks for pointing it out.

> 
> > 
> > Reported-by: James Alandt <James.Alandt@wdc.com>
> > Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
> > ---
> >  fs/btrfs/volumes.c | 87 ++++++++++++++++++++++++++++++++++++++++++------------
> >  1 file changed, 68 insertions(+), 19 deletions(-)
> > 
> > diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
> > index 4a72c45..3f48bcd 100644
> > --- a/fs/btrfs/volumes.c
> > +++ b/fs/btrfs/volumes.c
> > @@ -144,6 +144,8 @@ static int __btrfs_map_block(struct btrfs_fs_info *fs_info,
> >  			     u64 logical, u64 *length,
> >  			     struct btrfs_bio **bbio_ret,
> >  			     int mirror_num, int need_raid_map);
> > +static int btrfs_may_alloc_data_chunk(struct btrfs_fs_info *fs_info,
> > +				      u64 chunk_offset);
> 
> Also there is no need to have this forward declaration, the function can
> just as well be put right before __btrfs_balance. Let's try and keep
> changes minimal.
>

OK.

> >  
> >  DEFINE_MUTEX(uuid_mutex);
> >  static LIST_HEAD(fs_uuids);
> > @@ -3476,7 +3478,6 @@ static int __btrfs_balance(struct btrfs_fs_info *fs_info)
> >  	u32 count_meta = 0;
> >  	u32 count_sys = 0;
> >  	int chunk_reserved = 0;
> > -	u64 bytes_used = 0;
> >  
> >  	/* step one make some room on all the devices */
> >  	devices = &fs_info->fs_devices->devices;
> > @@ -3635,28 +3636,22 @@ static int __btrfs_balance(struct btrfs_fs_info *fs_info)
> >  			goto loop;
> >  		}
> >  
> > -		ASSERT(fs_info->data_sinfo);
> > -		spin_lock(&fs_info->data_sinfo->lock);
> > -		bytes_used = fs_info->data_sinfo->bytes_used;
> > -		spin_unlock(&fs_info->data_sinfo->lock);
> > -
> > -		if ((chunk_type & BTRFS_BLOCK_GROUP_DATA) &&
> > -		    !chunk_reserved && !bytes_used) {
> > -			trans = btrfs_start_transaction(chunk_root, 0);
> > -			if (IS_ERR(trans)) {
> > -				mutex_unlock(&fs_info->delete_unused_bgs_mutex);
> > -				ret = PTR_ERR(trans);
> > -				goto error;
> > -			}
> > -
> > -			ret = btrfs_force_chunk_alloc(trans, fs_info,
> > -						      BTRFS_BLOCK_GROUP_DATA);
> > -			btrfs_end_transaction(trans);
> > +		if (!chunk_reserved) {
> > +			/*
> > +			 * We may be relocating the only data chunk we have,
> > +			 * which could potentially end up with losing data's
> > +			 * raid profile, so lets allocate an empty one in
> > +			 * advance.
> > +			 */
> > +			ret = btrfs_may_alloc_data_chunk(fs_info,
> > +							 found_key.offset);
> >  			if (ret < 0) {
> >  				mutex_unlock(&fs_info->delete_unused_bgs_mutex);
> > +				ret = PTR_ERR(trans);

I'll remove this ret = PTR_ERR(trans);

-liubo

> >  				goto error;
> > +			} else if (ret == 1) {
> > +				chunk_reserved = 1;
> >  			}
> > -			chunk_reserved = 1;
> >  		}
> >  
> >  		ret = btrfs_relocate_chunk(fs_info, found_key.offset);
> > @@ -4327,6 +4322,48 @@ int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info)
> >  }
> >  
> >  /*
> > + * return 1 : allocate a data chunk successfully,
> > + * return <0: errors during allocating a data chunk,
> > + * return 0 : no need to allocate a data chunk.
> > + */
> > +static int btrfs_may_alloc_data_chunk(struct btrfs_fs_info *fs_info,
> > +				      u64 chunk_offset)
> > +{
> > +	struct btrfs_block_group_cache *cache;
> > +	u64 bytes_used;
> > +	u64 chunk_type;
> > +
> > +	cache = btrfs_lookup_block_group(fs_info, chunk_offset);
> > +	ASSERT(cache);
> > +	chunk_type = cache->flags;
> > +	btrfs_put_block_group(cache);
> > +
> > +	if (chunk_type & BTRFS_BLOCK_GROUP_DATA) {
> > +		spin_lock(&fs_info->data_sinfo->lock);
> > +		bytes_used = fs_info->data_sinfo->bytes_used;
> > +		spin_unlock(&fs_info->data_sinfo->lock);
> > +
> > +		if (!bytes_used) {
> > +			struct btrfs_trans_handle *trans;
> > +			int ret;
> > +
> > +			trans =	btrfs_join_transaction(fs_info->tree_root);
> > +			if (IS_ERR(trans))
> > +				return PTR_ERR(trans);
> > +
> > +			ret = btrfs_force_chunk_alloc(trans, fs_info,
> > +						      BTRFS_BLOCK_GROUP_DATA);
> > +			btrfs_end_transaction(trans);
> > +			if (ret < 0)
> > +				return ret;
> > +
> > +			return 1;
> > +		}
> > +	}
> > +	return 0;
> > +}
> > +
> > +/*
> >   * shrinking a device means finding all of the device extents past
> >   * the new size, and then following the back refs to the chunks.
> >   * The chunk relocation code actually frees the device extent
> > @@ -4419,6 +4456,18 @@ int btrfs_shrink_device(struct btrfs_device *device, u64 new_size)
> >  		chunk_offset = btrfs_dev_extent_chunk_offset(l, dev_extent);
> >  		btrfs_release_path(path);
> >  
> > +		/*
> > +		 * We may be relocating the only data chunk we have,
> > +		 * which could potentially end up with losing data's
> > +		 * raid profile, so lets allocate an empty one in
> > +		 * advance.
> > +		 */
> > +		ret = btrfs_may_alloc_data_chunk(fs_info, chunk_offset);
> > +		if (ret < 0) {
> > +			mutex_unlock(&fs_info->delete_unused_bgs_mutex);
> > +			goto done;
> > +		}
> > +
> >  		ret = btrfs_relocate_chunk(fs_info, chunk_offset);
> >  		mutex_unlock(&fs_info->delete_unused_bgs_mutex);
> >  		if (ret && ret != -ENOSPC)
> > 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html