From: Dave Chinner <dchinner@xxxxxxxxxx> When doing a direct IO that spans the current EOF, and there are written blocks beyond EOF that extend beyond the current write, the only metadata update that needs to be done is a file size extension. However, we don't mark such iomaps as IOMAP_F_DIRTY to indicate that there is IO completion metadata updates required, and hence we may fail to correctly sync file size extensions made in IO completion when O_DSYNC writes are being used and the hardware supports FUA. Hence when setting IOMAP_F_DIRTY, we need to also take into account whether the iomap spans the current EOF. If it does, then we need to mark it dirty so that IO completion will call generic_write_sync() to flush the inode size update to stable storage correctly. Signed-off-by: Dave Chinner <dchinner@xxxxxxxxxx> --- V2: put the XFS code in the right function, tested on non-FUA capable hardware, too. fs/ext4/inode.c | 9 ++++++++- fs/xfs/xfs_iomap.c | 7 +++++++ include/linux/iomap.h | 2 ++ 3 files changed, 17 insertions(+), 1 deletion(-) diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index 516faa280ced..e9dc52537e5b 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -3523,9 +3523,16 @@ static int ext4_iomap_begin(struct inode *inode, loff_t offset, loff_t length, return ret; } + /* + * Writes that span EOF might trigger an IO size update on completion, + * so consider them to be dirty for the purposes of O_DSYNC even if + * there is no other metadata changes being made or are pending here. + */ iomap->flags = 0; - if (ext4_inode_datasync_dirty(inode)) + if (ext4_inode_datasync_dirty(inode) || + offset + length > i_size_read(inode)) iomap->flags |= IOMAP_F_DIRTY; + iomap->bdev = inode->i_sb->s_bdev; iomap->dax_dev = sbi->s_daxdev; iomap->offset = (u64)first_block << blkbits; diff --git a/fs/xfs/xfs_iomap.c b/fs/xfs/xfs_iomap.c index f780e223b118..32993c2acbd9 100644 --- a/fs/xfs/xfs_iomap.c +++ b/fs/xfs/xfs_iomap.c @@ -1049,6 +1049,13 @@ xfs_file_iomap_begin( trace_xfs_iomap_alloc(ip, offset, length, XFS_DATA_FORK, &imap); out_finish: + /* + * Writes that span EOF might trigger an IO size update on completion, + * so consider them to be dirty for the purposes of O_DSYNC even if + * there is no other metadata changes pending or have been made here. + */ + if ((flags & IOMAP_WRITE) && offset + length > i_size_read(inode)) + iomap->flags |= IOMAP_F_DIRTY; return xfs_bmbt_to_iomap(ip, iomap, &imap, shared); out_found: diff --git a/include/linux/iomap.h b/include/linux/iomap.h index 7aa5d6117936..24bd227d59f9 100644 --- a/include/linux/iomap.h +++ b/include/linux/iomap.h @@ -32,6 +32,8 @@ struct vm_fault; * * IOMAP_F_DIRTY indicates the inode has uncommitted metadata needed to access * written data and requires fdatasync to commit them to persistent storage. + * This needs to take into account metadata changes that *may* be made at IO + * completion, such as file size updates from direct IO. */ #define IOMAP_F_NEW 0x01 /* blocks have been newly allocated */ #define IOMAP_F_DIRTY 0x02 /* uncommitted metadata */
