In rare circumstances, the short period that *hot_remove_disk()
waits isn't long enough to IO to complete. This particularly happens
when a device is failing and many retries are still happening.
We don't want to increase the normal wait time for "mdadm --remove"
as that might be use just to test if a device is active or not, and a
delay would be problematic.
So allow "--force" to mean that mdadm should try extra hard for a
--remove to complete, waiting up to 5 seconds.
Note that this patch fixes a comment which claim the previous
wait time was half a second, where it was really 50msec.
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Jes Sorensen <Jes.Sorensen@gmail.com>
The new hot_remove_disk() will retry HOT_REMOVE_DISK
several times in the face of EBUSY.
However we sometimes remove a device by writing "remove" to the
"state" attributed. This should be retried as well.
So introduce sys_hot_remove_disk() to repeat this action a few times.
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Jes Sorensen <Jes.Sorensen@gmail.com>
HOT_REMOVE_DISK can fail with EBUSY if there are outstanding
IO request that have not completed yet. It can sometimes
be helpful to wait a little while for these to complete.
We already do this in impose_level() when reshaping a device,
but not in Manage.c in response to an explicit --remove request.
So create hot_remove_disk() to central this code, and call it
where-ever it makes sense to wait for a HOT_REMOVE_DISK to succeed.
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Jes Sorensen <Jes.Sorensen@gmail.com>
We currently use '1' to indicate that a flag (writemostly or failfast)
needs to be set, and '2' to indicate that it needs to be cleared.
Using magic number like this is not a best-practice.
So replaced them with values from a enum.
No functional change.
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Jes Sorensen <Jes.Sorensen@redhat.com>
Allow per-device "failfast" flag to be set when creating an
array or adding devices to an array.
When re-adding a device which had the failfast flag, it can be removed
using --nofailfast.
failfast status is printed in --detail and --examine output.
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Jes Sorensen <Jes.Sorensen@redhat.com>
A 'faulty' drive is being removed from a container after it has been
released by an array, however there is a race there. The drive is
released asynchronously by a monitor but sometimes it doesn't happen
before container checks it. It results in a container refusing to remove
a drive as it still seems to be a part of some array.
It seems 'ping_monitor' could be a solution here to assure monitor has
had a chance to process the events, however it doesn't resolve the
problem - sometimes an array has to request a release of the drive few
times (as the array is busy) and single 'ping_monitor' call is not
sufficient. As there is no way to query monitor progress, it forces us
to retry a check several times before an error is returned.
Signed-off-by: Tomasz Majchrzak <tomasz.majchrzak@intel.com>
Signed-off-by: Jes Sorensen <Jes.Sorensen@redhat.com>
dev_st is only ever assigned if array->not_persistent == 0, so move
the second use of it into the same scope where the assignment is
made.
Signed-off-by: Jes Sorensen <Jes.Sorensen@redhat.com>
sysfs_read() allocates and populates a struct mdinfo, however the code
forgot to free it again, before dropping the reference to the pointer.
Reviewed-by: NeilBrown <neilb@suse.com>
Signed-off-by: Jes Sorensen <Jes.Sorensen@redhat.com>
Commit d180d2aa2a ("Manage: fix test for 'is array failed'.")
introduced a regression which would not allow to re-add new
drivers to a failed array.
Fixes: d180d2aa2a ("Manage: fix test for 'is array failed'.")
Signed-off-by: Hannes Reinecke <hare@suse.de>
Cc: Coly Li <colyli@suse.de>
Cc: Neil Brown <neilb@suse.com>
Signed-off-by: Jes Sorensen <Jes.Sorensen@redhat.com>
sysfs_read() may return NULL, so we should check the validity of the
pointer before dereferencing it.
Signed-off-by: Jes Sorensen <Jes.Sorensen@redhat.com>
2.6.28+ kernels handle this themselves and issuing the event here can
cause a race.
Reported-by: Sebastian Parschauer <sebastian.riemer@profitbricks.com>
Suggested-by: NeilBrown <neilb@suse.com>
Signed-off-by: Jes Sorensen <Jes.Sorensen@redhat.com>
This patch tries recreates missing/faulty journal in mdadm.
Example:
./mdadm --fail /dev/md1 /dev/sdb2
mdadm: set /dev/sdb2 faulty in /dev/md1
./mdadm --stop /dev/md1
mdadm: stopped /dev/md1
./mdadm -A --scan --force
mdadm: Journal is missing or stale, starting array read only.
mdadm: /dev/md/1 has been started with 15 drives.
./mdadm --add-journal /dev/md1 /dev/sdb2
mdadm: added /dev/sdb2
Signed-off-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
If sysfs node existed, we should try to write "re-add" to it.
Signed-off-by: Guoqing Jiang <gqjiang@suse.com>
Signed-off-by: NeilBrown <neilb@suse.com>
If it is a cluster raid, the disc.state need to be
changed accordingly when do re-add.
Signed-off-by: Guoqing Jiang <gqjiang@suse.com>
Signed-off-by: NeilBrown <neilb@suse.com>
When parse_cluster_confirm_arg return 0, it means the
arg are parsed successfully, so change !rv to rv.
Signed-off-by: Guoqing Jiang <gqjiang@suse.com>
Signed-off-by: NeilBrown <neilb@suse.com>
If the array is reshaping to more devices, then stopping
during that initial critical section is a bad idea.
So check for it and wait a bit.
Should probably handle final critical section of a reduction
too.
same-size reshape should be handled correctly already.
Signed-off-by: NeilBrown <neilb@suse.de>
A race can allow 'completed' to read as 2^63-1, which takes
a long time to count up to.
So guard against that possibility.
Signed-off-by: NeilBrown <neilb@suse.com>
A 'devnm' never starts with '/', so this test is pointless.
The code should use the passed-in devname unless it is clearly
not usable. So fix it to do that.
Signed-off-by: NeilBrown <neilb@suse.de>
These both have the same value, and have done since the
'devnm' concept was introduced.
So discard the pointless duplicate.
Signed-off-by: NeilBrown <neilb@suse.de>
A clustered disk is added by the traditional --add sequence.
However, other nodes need to acknowledge that they can "see"
the device. This is done by --cluster-confirm:
--cluster-confirm SLOTNUM:/dev/whatever (if disk is found)
or
--cluster-confirm SLOTNUM:missing (if disk is not found)
The node initiating the --add, has the disk state tagged with
MD_DISK_CLUSTER_ADD and the one confirming tag the disk with
MD_DISK_CANDIDATE.
Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>
Signed-off-by: Guoqing Jiang <gqjiang@suse.com>
Signed-off-by: NeilBrown <neilb@suse.de>
avail_size1 requires ->sb, so we must only call it if ->sb
was loaded.
If ->sb wasn't loaded, then we are only proceding on the basis that
the kernel might be able to work something out - we don't need to
do any tests on size.
Reported-by: Christoffer Hammarström <christoffer.hammarstrom@linuxgods.com>
Signed-off-by: NeilBrown <neilb@suse.de>
URL: https://bugs.debian.org/784874
We 'active_disks' does not count spares, so if array is rebuilding,
this will not necessarily find all devices, so may report an array
as failed when it isn't.
Counting up to nr_disks is better.
Signed-off-by: NeilBrown <neilb@suse.de>
It is best to keep strings all together so that they
are easier to search for in the source code.
If a string is so long that it looks ugly one line,
them maybe it should be broken into multiple lines
for display too.
Only strings which contain a newline can be broken
into multiple lines:
"It is OK to\n"
"break this string\n"
Signed-off-by: NeilBrown <neilb@suse.de>
"--remove detached" and others stopped working a while
back when I refactored some code.
For 'remove' and 'fail', the device may not exist so
if it is "MM:mm", (e.g. added by "detached"), just parse
out the numbers.
Reported-by: Killian De Volder <killian.de.volder@megasoft.be>
Signed-off-by: NeilBrown <neilb@suse.de>
The only use 'struct stat stb' to get the 'rdev', and sometimes
we don't even use 'stat'.
So make 'rdev' a stand-alone variable, and only declare stb'
when we actually need it.
Signed-off-by: NeilBrown <neilb@suse.de>
1/ enough_fd doesn't use avail_disks any more, so discard it.
2/ Manage_Add increments 'found' at the wrong place, so it can
waste time before calling enough().
Signed-off-by: NeilBrown <neilb@suse.de>
--add-spare is like --add, but a --re-add is never attempted.
So it is equivalent to two separate commands:
--zero-metadata
--add
Signed-off-by: NeilBrown <neilb@suse.de>
We really need to make sure assemble_container_content()
gets called to finished the assembly of these.
Reported-by: Francis Moreau <francis.moro@gmail.com>
Signed-off-by: NeilBrown <neilb@suse.de>
Current "mdadm --run /dev/mdX" will not handle external metadata
properly. mdmon won't be started etc.
So use the code from "mdadm -IRs" instead - that already does all
the right things.
Reported-by: Francis Moreau <francis.moro@gmail.com>
Signed-off-by: NeilBrown <neilb@suse.de>
This call to validate_geometry is really rather gratuitous.
It is purely about the fact that super0 cannot use more than 4TB.
So just make it an explicit test - less confusing that way.
With this, validate_geometry is only called from Create, which
makes it easier to reason about.
Also validate_geometry is now never passed NULL for the 'chunk'
parameter, so we can remove those annoying tests for NULL.
Signed-off-by: NeilBrown <neilb@suse.de>
If we stop too soon after reshape starts (probably only during
testing), we can get confused by the status of the reshape.
If that might be happening - sleep a bit longer.
Also allow for reshape going unusually slowly (again, probably only
during testing).
Signed-off-by: NeilBrown <neilb@suse.de>
It is possible for 'sync_completed' to be further ahead than
we deduced from 'reshape_position'. However we cannot read it while
the array is frozen, so it is hard to know.
Once that array is unfrozen, check and if sync_completed is ahead of
'sync_max', push 'sync_max' well ahead if 'sync_completed' so it
will all synchronise up properly.
Signed-off-by: NeilBrown <neilb@suse.de>
To be able to revert-reshape of raid4/5/6 which is changing
the number of devices, the reshape must has been stopped on a multiple
of the old and new stripe sizes.
The kernel only enforces the new stripe size multiple.
So we enforce the old-stripe-size multiple by careful use of
"sync_max" and monitoring "reshape_position".
Signed-off-by: NeilBrown <neilb@suse.de>
Manage_runstop has an open-coded version of use_udev() which is no
longer correct. So make it use use_udev() explicitly.
Signed-off-by: NeilBrown <neilb@suse.de>
A RAID10 array can have 'sets' of devices which are reported by
--detail.
They can now be collectively failed or removed.
Signed-off-by: NeilBrown <neilb@suse.de>
When stopping an mdmon array, at reshape might be being aborted
which inhibets O_EXCL. So if that is possible, call flush_mdmon
to make sure mdmon isn't still busy.
Reported-by: Pawel Baldysiak <pawel.baldysiak@intel.com>
Signed-off-by: NeilBrown <neilb@suse.de>
There are a number of fields which should not
be left uninitialised. e.g. attempt_re_add can get
confused if ->writemostly is not set correctly.
Signed-off-by: NeilBrown <neilb@suse.de>
NeilBrown
Tomasz Majchrzak
Jes Sorensen
Hannes Reinecke
Song Liu
Guoqing Jiang
NeilBrown
Lukasz Dorau