On Tue, Jan 10, 2017 at 4:11 PM, Toshi Kani <toshi.k...@hpe.com> wrote:
> DAX IO path does not support iostat, but its metadata IO path does.
> Therefore, iostat shows metadata IO statistics only, which has been
> confusing to users.
>
> Add iostat support to the DAX read/write path.
>
> Note, iostat still does not support the DAX mmap path as it allows
> user applications to access directly.
>
> Signed-off-by: Toshi Kani <toshi.k...@hpe.com>
> Cc: Andrew Morton <a...@linux-foundation.org>
> Cc: Dan Williams <dan.j.willi...@intel.com>
> Cc: Alexander Viro <v...@zeniv.linux.org.uk>
> Cc: Dave Chinner <da...@fromorbit.com>
> Cc: Ross Zwisler <ross.zwis...@linux.intel.com>
> ---
> v4: Rebased to 4.10, applied the v3 change to new dax_iomap_rw().
> ---
> fs/dax.c | 13 +++++++++++++
> 1 file changed, 13 insertions(+)
>
> diff --git a/fs/dax.c b/fs/dax.c
> index 5c74f60..4d5f4c0 100644
> --- a/fs/dax.c
> +++ b/fs/dax.c
> @@ -1058,12 +1058,22 @@ dax_iomap_rw(struct kiocb *iocb, struct iov_iter
> *iter,
> {
> struct address_space *mapping = iocb->ki_filp->f_mapping;
> struct inode *inode = mapping->host;
> + struct gendisk *disk = inode->i_sb->s_bdev->bd_disk;
> loff_t pos = iocb->ki_pos, ret = 0, done = 0;
> unsigned flags = 0;
> + unsigned long start = 0;
>
> if (iov_iter_rw(iter) == WRITE)
> flags |= IOMAP_WRITE;
>
> + if (blk_queue_io_stat(disk->queue)) {
> + int sec = iov_iter_count(iter) >> 9;
> +
> + start = jiffies;
> + generic_start_io_acct(iov_iter_rw(iter),
> + (!sec) ? 1 : sec, &disk->part0);
> + }
> +
> while (iov_iter_count(iter)) {
> ret = iomap_apply(inode, pos, iov_iter_count(iter), flags,
> ops,
> iter, dax_iomap_actor);
> @@ -1073,6 +1083,9 @@ dax_iomap_rw(struct kiocb *iocb, struct iov_iter *iter,
> done += ret;
> }
>
> + if (start)
> + generic_end_io_acct(iov_iter_rw(iter), &disk->part0, start);
> +
I think we can afford to add a separate flag that indicates whether we
called generic_start_io_acct(). Just in case 'start' is '0' after
'jiffies' rolls over.
