4.4-stable review patch. If anyone has any objections, please let me know.
------------------
From: BingJing Chang bingjingc@synology.com
[ Upstream commit bda3153998f3eb2cafa4a6311971143628eacdbc ]
During assemble, the spare marked for replacement is not checked. conf->fullsync cannot be updated to be 1. As a result, recovery will treat it as a clean array. All recovering sectors are skipped. Original device is replaced with the not-recovered spare.
mdadm -C /dev/md0 -l10 -n4 -pn2 /dev/loop[0123] mdadm /dev/md0 -a /dev/loop4 mdadm /dev/md0 --replace /dev/loop0 mdadm -S /dev/md0 # stop array during recovery
mdadm -A /dev/md0 /dev/loop[01234]
After reassemble, you can see recovery go on, but it completes immediately. In fact, recovery is not actually processed.
To solve this problem, we just add the missing logics for replacment spares. (In raid1.c or raid5.c, they have already been checked.)
Reported-by: Alex Chen alexchen@synology.com Reviewed-by: Alex Wu alexwu@synology.com Reviewed-by: Chung-Chiang Cheng cccheng@synology.com Signed-off-by: BingJing Chang bingjingc@synology.com Signed-off-by: Shaohua Li shli@fb.com Signed-off-by: Sasha Levin alexander.levin@microsoft.com Signed-off-by: Greg Kroah-Hartman gregkh@linuxfoundation.org --- drivers/md/raid10.c | 7 +++++++ 1 file changed, 7 insertions(+)
--- a/drivers/md/raid10.c +++ b/drivers/md/raid10.c @@ -3691,6 +3691,13 @@ static int run(struct mddev *mddev) disk->rdev->saved_raid_disk < 0) conf->fullsync = 1; } + + if (disk->replacement && + !test_bit(In_sync, &disk->replacement->flags) && + disk->replacement->saved_raid_disk < 0) { + conf->fullsync = 1; + } + disk->recovery_disabled = mddev->recovery_disabled - 1; }