On 09/05/2016 12:31 PM, David Sterba wrote:
On Fri, Sep 02, 2016 at 03:39:59PM -0400, Josef Bacik wrote:
In order to provide a better way to do subpage blocksizes we need to stop
allocating pages from a per fs btree inode and instead allocate our own pages.
This work depends on 3 generic patches that I've sent previously
remove mapping from balance_dirty_pages*()
writeback: allow for dirty metadata accounting
writeback: introduce super_operations->write_metadata
What's the status of the patches? I don't see them in current
linux-next. I can put them to our for-next for the time being so we can
test the btree_inode removal.
Still waiting on reviews. The first pass had no big objections, just need to
get people to look at the next pass.
This is a pretty big change but ultimately makes extent_buffer reclaim much
cleaner and will make the sub-pagesize blocksize work significantly cleaner.
I've been hammering on this for a few weeks now and seems to be pretty solid.
The preparatory patches are ok. The core patch is quite large, although
there are simple transformations to the new eb_info, the remaining
changes are still quite big for a review. But I don't see a way how to
split it, it's basically an all-or-nothing change. Overall it looks ok
so I'll put it to for-next.
Yeah I broke out what I could but unfortunately ripping out the btree_inode
means we have to make a bunch of other changes all at once, or we wouldn't be
able to bisect across it without things breaking horribly. Thanks,
Josef
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
