On Fri, Nov 06 2015, rgoldwyn@xxxxxxx wrote:
> From: Guoqing Jiang <gqjiang@xxxxxxxx>
>
> For cluster raid, if one disk couldn't be reach in one node, then
> other nodes would receive the REMOVE message for the disk.
>
> In receiving node, we can't call md_kick_rdev_from_array to remove
> the disk from array synchronously since the disk might still be busy
> in this node. So let's set a ClusterRemove flag on the disk, then
> let the thread to do the removal job eventually.
Thanks.
I've applied this patch.
However
1/ it isn't against mainline.
2/ While the ClusterRemove flag is (currently) only used in a cluster
configuration, the functionality that it represents isn't necessarily
cluster specific. So I would prefer a more generic name (like
AutoRemove).
3/ similarly the test on mddev_is_cluster() in md_check_recovery()
doesn't really search much purpose.
Thanks,
NeilBrown
>
> Signed-off-by: Guoqing Jiang <gqjiang@xxxxxxxx>
> Signed-off-by: Goldwyn Rodrigues <rgoldwyn@xxxxxxxx>
> ---
> drivers/md/md-cluster.c | 7 +++++--
> drivers/md/md.c | 12 ++++++++++++
> drivers/md/md.h | 1 +
> 3 files changed, 18 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/md/md-cluster.c b/drivers/md/md-cluster.c
> index 3daa464..a681706 100644
> --- a/drivers/md/md-cluster.c
> +++ b/drivers/md/md-cluster.c
> @@ -443,8 +443,11 @@ static void process_remove_disk(struct mddev *mddev, struct cluster_msg *msg)
> struct md_rdev *rdev = md_find_rdev_nr_rcu(mddev,
> le32_to_cpu(msg->raid_slot));
>
> - if (rdev)
> - md_kick_rdev_from_array(rdev);
> + if (rdev) {
> + set_bit(ClusterRemove, &rdev->flags);
> + set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
> + md_wakeup_thread(mddev->thread);
> + }
> else
> pr_warn("%s: %d Could not find disk(%d) to REMOVE\n",
> __func__, __LINE__, le32_to_cpu(msg->raid_slot));
> diff --git a/drivers/md/md.c b/drivers/md/md.c
> index 44d0342..32ca592 100644
> --- a/drivers/md/md.c
> +++ b/drivers/md/md.c
> @@ -8222,6 +8222,18 @@ void md_check_recovery(struct mddev *mddev)
> goto unlock;
> }
>
> + if (mddev_is_clustered(mddev)) {
> + struct md_rdev *rdev;
> + /* kick the device if another node issued a
> + * remove disk.
> + */
> + rdev_for_each(rdev, mddev) {
> + if (test_and_clear_bit(ClusterRemove, &rdev->flags) &&
> + rdev->raid_disk < 0)
> + md_kick_rdev_from_array(rdev);
> + }
> + }
> +
> if (!mddev->external) {
> int did_change = 0;
> spin_lock(&mddev->lock);
> diff --git a/drivers/md/md.h b/drivers/md/md.h
> index 2ea0035..db54341 100644
> --- a/drivers/md/md.h
> +++ b/drivers/md/md.h
> @@ -172,6 +172,7 @@ enum flag_bits {
> * This device is seen locally but not
> * by the whole cluster
> */
> + ClusterRemove,
> };
>
> #define BB_LEN_MASK (0x00000000000001FFULL)
> --
> 1.8.5.6
Attachment:
signature.asc
Description: PGP signature
