[BACKGROUND]
Currently btrfs never uses bio_split() to split its bio against RAID
stripe boundaries.
Instead inside btrfs we check our stripe boundary everytime we allocate
a new bio, and ensure the new bio never cross stripe boundaries.
[PROBLEMS]
Although this works fine, it's against the common practice used in
stacked drivers, and is making the effort to convert to iomap harder.
There is also an hidden burden, every time we allocate a new bio, we uses
BIO_MAX_BVECS, but since we know the boundaries, for RAID0/RAID10 we can
only fit at most 16 pages (fixed 64K stripe size, and 4K page size),
wasting the 256 slots we allocated.
[CHALLENGES]
To change the situation, this patchset attempts to improve the situation
by moving the bio split into btrfs_map_bio() time, so upper layer should
no longer bother the bio split against RAID stripes or even chunk
boundaries.
But there are several challenges:
- Conflicts in various endio functions
There is a special endio function, end_workqueue_bio(), that if a bio
has this endio function, and get split, we will double free the
btrfs_end_io_wq_cache.
And some hidden RAID56 endio also has this problem.
For RAID56 problems, it's remaining to be solved in the v1 version.
But for end_workqueue_bio() it can be moved after bios been split.
- Checksum verification
Currently we rely on btrfs_bio::csum to contain the checksum for the
whole bio.
If one bio get split, csum will no longer points to the correct
location for the split bio.
This can be solved by introducing btrfs_bio::offset_to_original, and
use that new member to calculate where we should read csum from.
For the parent bio, it still has btrfs_bio::csum for the whole bio,
thus it can still free it correctly.
- Independent endio for each split bio
Unlike stack drivers, for RAID10 btrfs needs to try its best effort to
read every sectors, to handle the following case:
Dev 1 (missing) | D1 (X) |
Dev 2 (OK) | D1 (V) |
Dev 3 (OK) | D2 (V) |
Dev 4 (OK) | D2 (X) |
In the above RAID10 case, dev1 is missing, and although dev4 is fine,
its D2 sector is corrupted (by bit rot or whatever).
If we use bio_chain(), read bio for both D1 and D2 will be split, and
since D1 is missing, the whole D1 and D2 read will be marked as error,
thus we will try to read from dev2 and dev4.
But D2 in dev4 has csum mismatch, we can only rebuild D1 and D2
correctly from dev2:D1 and dev3:D2.
This patchset resolve this by saving bi_iter into btrfs_bio::iter, and
uses that at endio to iterate only the split part of an bio.
Other than this, existing read/write page endio functions can handle
them properly without problem.
[RFC]
The patchset is only lightly tested, as there is still some endio
conflicts in the ancient RAID56 code.
But despite that, regular buffered read/write and repair should work
without problem.
This patchset is sent because above mentioned challenges, all the
solutions need extra review/feedback, not only from btrfs community but
also block layer community, to determine if this is really the best
solution.
Qu Wenruo (11):
btrfs: update an stale comment on btrfs_submit_bio_hook()
btrfs: refactor btrfs_map_bio()
btrfs: move btrfs_bio_wq_end_io() calls into submit_stripe_bio()
btrfs: introduce btrfs_bio_split() helper
btrfs: save bio::bi_iter into btrfs_bio::iter before submitting
btrfs: make end_bio_extent_readpage() to handle split bio properly
btrfs: make end_bio_extent_*_writepage() to handle split biot properly
btrfs: allow btrfs_map_bio() to split bio according to chunk stripe
boundaries
btrfs: remove bio split operations in btrfs_submit_direct()
btrfs: remove btrfs_bio_ctrl::len_to_stripe_boundary
btrfs: temporarily disable RAID56
fs/btrfs/compression.c | 7 +-
fs/btrfs/ctree.h | 5 +-
fs/btrfs/disk-io.c | 9 +-
fs/btrfs/extent_io.c | 183 +++++++++++++++++++++++++++--------------
fs/btrfs/extent_io.h | 3 +-
fs/btrfs/inode.c | 153 +++++++++++-----------------------
fs/btrfs/raid56.c | 2 +
fs/btrfs/volumes.c | 138 ++++++++++++++++++++++++-------
fs/btrfs/volumes.h | 72 +++++++++++++++-
9 files changed, 358 insertions(+), 214 deletions(-)
--
2.34.0
--
dm-devel mailing list
dm-devel@xxxxxxxxxx
https://listman.redhat.com/mailman/listinfo/dm-devel
