Re: mdraid10 regression in 2.6.27.4 (possibly earlier)

I'd just like to note that I have the same problem with 2.6.27.
/proc/mdstat is stuck saying
      
md4 : active raid10 sde3[6](S) sdf3[4] sdd3[2] sdc3[1] sdb3[0] sda3[5]
      131837184 blocks 256K chunks 2 near-copies [6/5] [UUU_UU]
      bitmap: 31/126 pages [124KB], 512KB chunk

while the kernel complains every second or so:
Nov  2 17:28:32: RAID10 conf printout:
Nov  2 17:28:32:  --- wd:5 rd:6
Nov  2 17:28:32:  disk 0, wo:0, o:1, dev:sdb3
Nov  2 17:28:32:  disk 1, wo:0, o:1, dev:sdc3
Nov  2 17:28:32:  disk 2, wo:0, o:1, dev:sdd3
Nov  2 17:28:32:  disk 4, wo:0, o:1, dev:sdf3
Nov  2 17:28:32:  disk 5, wo:0, o:1, dev:sda3
Nov  2 17:28:33: RAID10 conf printout:
Nov  2 17:28:33:  --- wd:5 rd:6
Nov  2 17:28:33:  disk 0, wo:0, o:1, dev:sdb3
Nov  2 17:28:33:  disk 1, wo:0, o:1, dev:sdc3
Nov  2 17:28:33:  disk 2, wo:0, o:1, dev:sdd3
Nov  2 17:28:33:  disk 4, wo:0, o:1, dev:sdf3
Nov  2 17:28:33:  disk 5, wo:0, o:1, dev:sda3
Nov  2 17:28:38: RAID10 conf printout:
Nov  2 17:28:38:  --- wd:5 rd:6
Nov  2 17:28:38:  disk 0, wo:0, o:1, dev:sdb3
Nov  2 17:28:38:  disk 1, wo:0, o:1, dev:sdc3
Nov  2 17:28:38:  disk 2, wo:0, o:1, dev:sdd3
Nov  2 17:28:38:  disk 4, wo:0, o:1, dev:sdf3
Nov  2 17:28:38:  disk 5, wo:0, o:1, dev:sda3
Nov  2 17:28:38: RAID10 conf printout:
Nov  2 17:28:38:  --- wd:5 rd:6
Nov  2 17:28:38:  disk 0, wo:0, o:1, dev:sdb3
Nov  2 17:28:38:  disk 1, wo:0, o:1, dev:sdc3
Nov  2 17:28:38:  disk 2, wo:0, o:1, dev:sdd3
Nov  2 17:28:38:  disk 4, wo:0, o:1, dev:sdf3
Nov  2 17:28:38:  disk 5, wo:0, o:1, dev:sda3
Nov  2 17:28:43: RAID10 conf printout:
Nov  2 17:28:43:  --- wd:5 rd:6
Nov  2 17:28:43:  disk 0, wo:0, o:1, dev:sdb3
Nov  2 17:28:43:  disk 1, wo:0, o:1, dev:sdc3
Nov  2 17:28:43:  disk 2, wo:0, o:1, dev:sdd3
Nov  2 17:28:43:  disk 4, wo:0, o:1, dev:sdf3
Nov  2 17:28:43:  disk 5, wo:0, o:1, dev:sda3
Nov  2 17:28:44: RAID10 conf printout:
Nov  2 17:28:44:  --- wd:5 rd:6
Nov  2 17:28:44:  disk 0, wo:0, o:1, dev:sdb3
Nov  2 17:28:44:  disk 1, wo:0, o:1, dev:sdc3
Nov  2 17:28:44:  disk 2, wo:0, o:1, dev:sdd3
Nov  2 17:28:44:  disk 4, wo:0, o:1, dev:sdf3
Nov  2 17:28:44:  disk 5, wo:0, o:1, dev:sda3

This is Not Good (tm).

Quad-core Phenom, 64-bit kernel and userland.

And an extra datapoint.

Booting into 2.6.26.5 triggers an instant resync of the spare disks, so 
it means we have a regression between 2.6.26.5 and 2.6.27.4

Likewise, 2.6.26.6 fixed the problem for me.  And I can confirm that it's
broken in plain 2.6.27.

There are 8 commits to drivers/md/raid10.c in that interval.
commit 0310fa216decc3ecfab41f327638fa48a81f3735
    Allow raid10 resync to happening in larger chunks.
commit 1e24b15b267293567a8d752721c7ae63f281325a
    Merge branch 'for-linus' of git://neil.brown.name/md
commit 388667bed591b2359713bb17d5de0cf56e961447
    md: raid10: wake up frozen array
commit 8a392625b665c676a77c62f8608d10ff430bcb83
    Merge branch 'for-linus' of git://neil.brown.name/md
commit f233ea5c9e0d8b95e4283bf6a3436b88f6fd3586
    md: Make mddev->array_size sector-based.
commit cc371e66e340f35eed8dc4651c7c18e754c7fb26
    Add bvec_merge_data to handle stacked devices and ->merge_bvec()
commit 199050ea1ff2270174ee525b73bc4c3323098897
    rationalise return value for ->hot_add_disk method.
commit 6c2fce2ef6b4821c21b5c42c7207cb9cf8c87eda
    Support adding a spare to a live md array with external metadata.

It's not obvious which of these is the problem; they mostly look pretty simple
and safe, at least as far as raid10 ic concerned.