| From: |
| John Garry <john.g.garry-AT-oracle.com> |
| To: |
| axboe-AT-kernel.dk, brauner-AT-kernel.org, djwong-AT-kernel.org, viro-AT-zeniv.linux.org.uk, jack-AT-suse.cz, dchinner-AT-redhat.com, hch-AT-lst.de, cem-AT-kernel.org |
| Subject: |
| [PATCH v6 0/7] block atomic writes for xfs |
| Date: |
| Mon, 30 Sep 2024 12:54:31 +0000 |
| Message-ID: |
| <20240930125438.2501050-1-john.g.garry@oracle.com> |
| Cc: |
| linux-block-AT-vger.kernel.org, linux-kernel-AT-vger.kernel.org, linux-xfs-AT-vger.kernel.org, linux-fsdevel-AT-vger.kernel.org, hare-AT-suse.de, martin.petersen-AT-oracle.com, catherine.hoang-AT-oracle.com, mcgrof-AT-kernel.org, ritesh.list-AT-gmail.com, ojaswin-AT-linux.ibm.com, John Garry <john.g.garry-AT-oracle.com> |
| Archive-link: |
| Article |
This series expands atomic write support to filesystems, specifically
XFS.
Flag FS_XFLAG_ATOMICWRITES is added as an enabling flag for atomic writes.
XFS can be formatted for atomic writes as follows:
mkfs.xfs -i atomicwrites=1 -b size=16384 /dev/sda
Support can be enabled through xfs_io command:
$xfs_io -c "chattr +W" filename
$xfs_io -c "lsattr -v" filename
[atomic-writes] filename
$xfs_io -c statx filename
...
stat.stx_atomic_write_unit_min = 16384
stat.stx_atomic_write_unit_max = 16384
stat.stx_atomic_write_segments_max = 1
...
Dependence on forcealign is gone for the moment. This means that we can
only write a single FS block atomically. Since we can now have FS block
size > PAGE_SIZE for XFS, we can write atomically write > 4K blocks on
x86.
Using the large FS block size has downsides, so we still want the
forcealign feature.
Baseline is v6.12-rc1
Basic xfsprogs support at:
https://github.com/johnpgarry/xfsprogs-dev/tree/atomic-wr...
Patches for this series can be found at:
https://github.com/johnpgarry/linux/tree/atomic-writes-v6...
Changes since v5:
- Drop forcealign dependency
- Can support atomically writing a single FS block
- XFS_DIFLAG2_ATOMICWRITES is inherited from parent directories
- Add RB tags from Darrick (thanks!)
Changes since v4:
- Drop iomap extent-based zeroing and use single bio to cover multiple
extents
- Move forcealign to another series
- Various change in ioctl, sb, inode validation
- Add patch to tweak generic_atomic_write_valid() API
John Garry (7):
block/fs: Pass an iocb to generic_atomic_write_valid()
fs: Export generic_atomic_write_valid()
fs: iomap: Atomic write support
xfs: Support FS_XFLAG_ATOMICWRITES
xfs: Support atomic write for statx
xfs: Validate atomic writes
xfs: Support setting FMODE_CAN_ATOMIC_WRITE
block/fops.c | 8 +++----
fs/iomap/direct-io.c | 26 ++++++++++++++++++++---
fs/iomap/trace.h | 3 ++-
fs/read_write.c | 5 +++--
fs/xfs/libxfs/xfs_format.h | 11 ++++++++--
fs/xfs/libxfs/xfs_inode_buf.c | 38 ++++++++++++++++++++++++++++++++++
fs/xfs/libxfs/xfs_inode_util.c | 6 ++++++
fs/xfs/libxfs/xfs_sb.c | 2 ++
fs/xfs/xfs_buf.c | 15 +++++++++++++-
fs/xfs/xfs_buf.h | 5 ++++-
fs/xfs/xfs_buf_mem.c | 2 +-
fs/xfs/xfs_file.c | 19 +++++++++++++++++
fs/xfs/xfs_inode.h | 22 ++++++++++++++++++++
fs/xfs/xfs_ioctl.c | 37 +++++++++++++++++++++++++++++++++
fs/xfs/xfs_iops.c | 24 +++++++++++++++++++++
fs/xfs/xfs_mount.h | 2 ++
fs/xfs/xfs_reflink.c | 4 ++++
fs/xfs/xfs_super.c | 4 ++++
include/linux/fs.h | 2 +-
include/linux/iomap.h | 1 +
include/uapi/linux/fs.h | 1 +
21 files changed, 221 insertions(+), 16 deletions(-)
--
2.31.1
