| Message ID | 20211129024930.1574325-1-naohiro.aota@wdc.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | btrfs: fix re-dirty process of tree-log nodes | expand |
On 29/11/2021 03:51, Naohiro Aota wrote: > For zoned btrfs, we re-dirty a freeing tree node to ensure btrfs write > the region and not to leave a write hole on a zoned device. Current > code failed to re-dirty a node when the tree-log tree's depth >= > 2. This leads to a transaction abort with -EAGAIN. > I'd rephrase the above to: For zoned btrfs, we re-dirty a freed tree node to ensure btrfs can write the region and does not leave a hole on write on a zoned device. The current code fails to failed to re-dirty a node when the tree-log tree's depth was greater or equal to 2. This lead to a transaction abort with -EAGAIN. > Fix the issue by properly re-dirty a node on walking up the tree. > > Link: https://github.com/kdave/btrfs-progs/issues/415 > Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com> > --- Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Shouldn't we mark this for stable (v5.15)? After all the reporter did hit the problem on 5.15.
On Mon, Nov 29, 2021 at 11:49:30AM +0900, Naohiro Aota wrote: > For zoned btrfs, we re-dirty a freeing tree node to ensure btrfs write > the region and not to leave a write hole on a zoned device. Current > code failed to re-dirty a node when the tree-log tree's depth >= > 2. This leads to a transaction abort with -EAGAIN. > > Fix the issue by properly re-dirty a node on walking up the tree. > > Link: https://github.com/kdave/btrfs-progs/issues/415 Can you please add more information from the issue that's relevant to the problem? Eg. the stacktraces, reproducer etc.
On Mon, Nov 29, 2021 at 07:57:39AM +0000, Johannes Thumshirn wrote: > On 29/11/2021 03:51, Naohiro Aota wrote: > > For zoned btrfs, we re-dirty a freeing tree node to ensure btrfs write > > the region and not to leave a write hole on a zoned device. Current > > code failed to re-dirty a node when the tree-log tree's depth >= > > 2. This leads to a transaction abort with -EAGAIN. > > > > I'd rephrase the above to: > > For zoned btrfs, we re-dirty a freed tree node to ensure btrfs > can write the region and does not leave a hole on write on a zoned > device. The current code fails to failed to re-dirty a node when the > tree-log tree's depth was greater or equal to 2. This lead to a > transaction abort with -EAGAIN. > > > > Fix the issue by properly re-dirty a node on walking up the tree. > > > > Link: https://github.com/kdave/btrfs-progs/issues/415 > > Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com> > > --- > > Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> > > Shouldn't we mark this for stable (v5.15)? After all the reporter did > hit the problem on 5.15. > Oops, I've added the fixes tag and the stable tag to the commit, but somehow I forgot to update the patch file. I'll resend the patch with your fix with for the commit log. Thanks.
On Mon, Nov 29, 2021 at 01:19:39PM +0100, David Sterba wrote: > On Mon, Nov 29, 2021 at 11:49:30AM +0900, Naohiro Aota wrote: > > For zoned btrfs, we re-dirty a freeing tree node to ensure btrfs write > > the region and not to leave a write hole on a zoned device. Current > > code failed to re-dirty a node when the tree-log tree's depth >= > > 2. This leads to a transaction abort with -EAGAIN. > > > > Fix the issue by properly re-dirty a node on walking up the tree. > > > > Link: https://github.com/kdave/btrfs-progs/issues/415 > > Can you please add more information from the issue that's relevant to > the problem? Eg. the stacktraces, reproducer etc. Sure. I'll revise the patch with the fixes tags. Thanks,
diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c index de79e15a7c6a..a42e132f35f5 100644 --- a/fs/btrfs/tree-log.c +++ b/fs/btrfs/tree-log.c @@ -2890,6 +2890,8 @@ static noinline int walk_up_log_tree(struct btrfs_trans_handle *trans, path->nodes[*level]->len); if (ret) return ret; + btrfs_redirty_list_add(trans->transaction, + next); } else { if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &next->bflags)) clear_extent_buffer_dirty(next); @@ -2970,6 +2972,8 @@ static int walk_log_tree(struct btrfs_trans_handle *trans, next->start, next->len); if (ret) goto out; + btrfs_redirty_list_add(trans->transaction, + next); } else { if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &next->bflags)) clear_extent_buffer_dirty(next); @@ -3420,8 +3424,6 @@ static void free_log_tree(struct btrfs_trans_handle *trans, EXTENT_DIRTY | EXTENT_NEW | EXTENT_NEED_WAIT); extent_io_tree_release(&log->log_csum_range); - if (trans && log->node) - btrfs_redirty_list_add(trans->transaction, log->node); btrfs_put_root(log); }
For zoned btrfs, we re-dirty a freeing tree node to ensure btrfs write the region and not to leave a write hole on a zoned device. Current code failed to re-dirty a node when the tree-log tree's depth >= 2. This leads to a transaction abort with -EAGAIN. Fix the issue by properly re-dirty a node on walking up the tree. Link: https://github.com/kdave/btrfs-progs/issues/415 Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com> --- fs/btrfs/tree-log.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)