On Tue, Oct 05, 2021 at 03:22:58PM +0900, Naohiro Aota wrote:

As discussed in the Zoned Storage page [1],  the kernel page cache does not
guarantee that cached dirty pages will be flushed to a block device in
sequential sector order. Thus, we must use O_DIRECT for writing to a zoned
device to ensure the write ordering.

[1] https://zonedstorage.io/linux/overview/#zbd-support-restrictions

As a writng buffer is embedded in some other struct (e.g., "char data[]" in
struct extent_buffer), it is difficult to allocate the struct so that the
writng buffer is aligned.

This series introduces btrfs_{pread,pwrite} to wrap around pread/pwrite,
which allocates an aligned bounce buffer, copy the buffer contents, and
proceeds the IO. And, it now opens a zoned device with O_DIRECT.

Since the allocation and copying are costly, it is better to do them only
when necessary. But, it is cumbersome to call fcntl(F_GETFL) to determine
the file is opened with O_DIRECT or not every time doing an IO.

As zoned device forces to use zoned btrfs, I decided to use the zoned flag
to determine if it is direct-IO or not. This can cause a false-positive (to
use the bounce buffer when a file is *not* opened with O_DIRECT) in case of
emulated zoned mode on a non-zoned device or a regular file. Considering
the emulated zoned mode is mostly for debugging or testing, I believe this
is acceptable.

* Changes
v2
  - Rebased on the latest "devel" branch
  - Add patch to fix segfault in several cases
  - drop ZONED flag from BTRFS_CONVERT_ALLOWED_FEATURES

Patches 1 to 3 are preparation to fix some issues in the current code.

Patches 4 and 5 wraps pread/pwrite with newly introduced function
btrfs_pread/btrfs_pwrite.

Patch 6 deals with the zoned flag while reading the initial trees.

Patch 7 finally opens a zoned device with O_DIRECT.

Naohiro Aota (7):
  btrfs-progs: mkfs: do not set zone size on non-zoned mode
  btrfs-progs: set eb->fs_info properly
  btrfs-progs: drop ZONED flag from BTRFS_CONVERT_ALLOWED_FEATURES
  btrfs-progs: introduce btrfs_pwrite wrapper for pwrite
  btrfs-progs: introduce btrfs_pread wrapper for pread
  btrfs-progs: temporally set zoned flag for initial tree reading
  btrfs-progs: use direct-io for zoned device

Is this still supposed to work?

  $ ./mkfs.btrfs -f -O zoned -d single -m single img
  ...
  ERROR: 16384 is not aligned to 1048576
  ERROR: error during mkfs: Input/output error

On commit below this patchset it works and creates a filesystem with
zoned mode and zone size 256M.