On Tue, 7 Dec 2010 16:07:35 +0000 "Labun, Marcin" [off-list ref]
wrote:

quoted

From 4bd19fb7b8a4258bf6cf34288be635bdb9af3dbe Mon Sep 17 00:00:00 2001

From: Marcin Labun <redacted>
Date: Wed, 30 Nov 2010 03:55:18 +0100
Subject: [PATCH 1/2] IMSM: Fix problem in mdmon monitor of using removed disk from in imsm container.

Manager thread shall pass the information to monitor thread (mdmon)
that some devices are removed from container. Otherwise, monitor (mdmon)
might use such devices (spares) to rebuild the array that has gone degraded.

This problem happens for imsm containers, since a list of the container disks
is maintained in intel_super structure. When array goes degraded, the list is
searched to find a spare disks to start rebuild.
Without this fix the rebuild could be stared on the spare device that was
a member of the container, but has been removed from it.

New super type function handler has been introduced to prepare metadata
format specific information about removed devices.
int (*remove_from_super)(struct supertype *st, mdu_disk_info_t *dinfo,
                         int fd);
The message prepared in remove_from_super is later processed
by proceess_update handler in monitor thread.

I don't like this.  There is unnecessary complexity.

adding a disk and removing a disk are very different sorts of operations.
When adding a disk, you need to pass extra information about how the disk
might be used - whether it is already part of the array, or if it is a fresh
spare or whatever.
When removing a device there is none of that.  Just remove the device.

So when mdadm removes a device from a container it should
  - get a lock so mdmon won't assign the device as spare
  - check that the device is still a spare
  - remove the device from the container
  - unlock
  - ping mdmon

mdmon should notice that the device has gone and should update the metadata
accordingly.

So you may still need a 'remove_from_super' method, but it will not send a
metadata update request to mdmon.
Rather it will be run by mdmon when it notices the device is gone.

It is probably appropriate to pass an mdu_disk_info_t or maybe just a device
number.  I don't think there is any need to pass an 'fd'.

Does that approach seem OK to you?

Thanks,
NeilBrown