On 7/26/19 4:47 AM, yuyufen wrote:
Hi, Guoqing
On 2019/7/24 17:09, Guoqing Jiang wrote:
When a disk is added to array, the following path is called in mdadm.
Manage_subdevs -> sysfs_freeze_array
-> Manage_add
-> sysfs_set_str(&info, NULL, "sync_action","idle")
Then from kernel side, Manage_add invokes the path (add_new_disk ->
validate_super = super_1_validate) to set In_sync flag.
Since In_sync means "device is in_sync with rest of array", and the new
added disk need to resync thread to help the synchronization of data.
And md_reap_sync_thread would call spare_active to set In_sync for the
new added disk finally. So don't set In_sync if array is in frozen.
Signed-off-by: Guoqing Jiang <guoqing.jiang@xxxxxxxxxxxxxxx>
---
drivers/md/md.c | 11 +++++++++--
1 file changed, 9 insertions(+), 2 deletions(-)
diff --git a/drivers/md/md.c b/drivers/md/md.c
index 0ced0933d246..d0223316064d 100644
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -1826,8 +1826,15 @@ static int super_1_validate(struct mddev *mddev, struct md_rdev *rdev)
if (!(le32_to_cpu(sb->feature_map) &
MD_FEATURE_RECOVERY_BITMAP))
rdev->saved_raid_disk = -1;
- } else
- set_bit(In_sync, &rdev->flags);
+ } else {
+ /*
+ * If the array is FROZEN, then the device can't
+ * be in_sync with rest of array.
+ */
+ if (!test_bit(MD_RECOVERY_FROZEN,
+ &mddev->recovery))
+ set_bit(In_sync, &rdev->flags);
+ }
rdev->raid_disk = role;
break;
}
super_1_validate() set rdev with 'In_sync', while add_new_disk() will clear the flag bit again after
that.
clear_bit(In_sync, &rdev->flags); /* just to be sure */
So, I think there is no bad influence without test FROZEN. Am I ignoring anything?
Yes, you are right about it.
Though there is a small time window that the disk could be showed as 'U', I guess it is better to
reduce the window so we can get the correct status as much as possible.
Thanks,
Guoqing
