On Sat, Apr 01 2017, Guoqing Jiang wrote:
> Since the value of first is always '1', we can
> set min_offset_diff directly.
Alternately, maybe the fact that first is always '1', is a sign of a bug
that should be fixed.
The calculation of "min_offset_diff" doesn't make much sense now. It is
just the last_offset_diff.
I think the correct fix is to set "first = 0;" at the end of the
rdev_for_each() loop.
Thanks for spotting this.
NeilBrown
>
> Signed-off-by: Guoqing Jiang <gqjiang@xxxxxxxx>
> ---
> drivers/md/raid10.c | 8 ++------
> 1 file changed, 2 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
> index 0f13d57..edaf8f4 100644
> --- a/drivers/md/raid10.c
> +++ b/drivers/md/raid10.c
> @@ -3702,7 +3702,6 @@ static int raid10_run(struct mddev *mddev)
> struct md_rdev *rdev;
> sector_t size;
> sector_t min_offset_diff = 0;
> - int first = 1;
> bool discard_supported = false;
>
> if (mddev->private == NULL) {
> @@ -3758,8 +3757,7 @@ static int raid10_run(struct mddev *mddev)
> diff = -diff;
> if (diff < 0)
> diff = 0;
> - if (first || diff < min_offset_diff)
> - min_offset_diff = diff;
> + min_offset_diff = diff;
>
> if (mddev->gendisk)
> disk_stack_limits(mddev->gendisk, rdev->bdev,
> @@ -4140,7 +4138,6 @@ static int raid10_start_reshape(struct mddev *mddev)
>
> unsigned long before_length, after_length;
> sector_t min_offset_diff = 0;
> - int first = 1;
> struct geom new;
> struct r10conf *conf = mddev->private;
> struct md_rdev *rdev;
> @@ -4169,8 +4166,7 @@ static int raid10_start_reshape(struct mddev *mddev)
> diff = -diff;
> if (diff < 0)
> diff = 0;
> - if (first || diff < min_offset_diff)
> - min_offset_diff = diff;
> + min_offset_diff = diff;
> }
> }
>
> --
> 2.6.6
Attachment:
signature.asc
Description: PGP signature
