On 2021/7/8 下午7:54, David Sterba wrote:

On Mon, Jul 05, 2021 at 10:00:57AM +0800, Qu Wenruo wrote:

quoted

This allows later decompress functions to get rid of kmap()/kunmap()
pairs.

And since all other filesystems are getting rid of HIGHMEM, it should
not be a problem for btrfs.

Although we removed the HIGHMEM allocation, we still keep the
kmap()/kunmap() pairs.
They will be removed when involved functions are refactored later.

I've sent the highmem cleanup and will drop this patch, there should be
minimal conflicts in the followup changes.

Thanks, that's great.

I'll update the patchset later after I have shaken out all bugs in
subpage compressed read path.
(Optimistic ETA, around v5.14-rc3)

As previously without a way to write compressed extent under subpage
case, we have no coverage for subpage compressed read.

I hope not to submit new fixes just for subpage compressed read path
after the initial enablement.

Thanks,
Qu