Re: [PATCH 0/7] cleanup __btrfs_map_block

[Qu Wenruo <quwenruo@cn.fujitsu.com>]

At 02/18/2017 09:28 AM, Liu Bo wrote:
> This is attempting to make __btrfs_map_block less scary :)
>

Great!!
That's also what I want to do in btrfs-progs.

And it would be better if we can take a step further to have a better 
API for both pure mirrored/striped profile and parity profile.

I tried to do it in btrfs-progs with the following patchset, while it 
needs more work to do in kernel.
https://patchwork.kernel.org/patch/9488519/

> The major changes are
>
> 1) split operations for discard out of __btrfs_map_block and we don't copy
> discard operations for the target device of dev replace since they're not
> as important as writes.
>
> 2) put dev replace stuff into helpers since they're basically
> self-independant.

Quite good starting point.

Thanks,
Qu
>
>
> Liu Bo (7):
>   Btrfs: create a helper for getting chunk map
>   Btrfs: separate DISCARD from __btrfs_map_block
>   Btrfs: introduce a function to get extra mirror from replace
>   Btrfs: handle operations for device replace separately
>   Btrfs: do not add extra mirror when dev_replace target dev is not
>     available
>   Btrfs: helper for ops that requires full stripe
>   Btrfs: convert BUG_ON to WARN_ON
>
>  fs/btrfs/extent_io.c |   3 +-
>  fs/btrfs/volumes.c   | 824 +++++++++++++++++++++++++++------------------------
>  fs/btrfs/volumes.h   |   2 +-
>  3 files changed, 445 insertions(+), 384 deletions(-)
>