Adam Kwolek [Tue, 14 Jun 2011 13:59:18 +0000 (15:59 +0200)]
imsm: Metadata Attributes compatibility support
IMSM's meta data contains Attributes field that contains information about
supported features.
To assembly an array mdadm has to support all features specified by attributes.
The patch introduces new attributes support and validation of the attribuses
during an array assembly.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Tue, 14 Jun 2011 23:13:49 +0000 (09:13 +1000)]
imsm: FIX: Sometimes reshape cannot be finished
When array size is not aligned to copy area, number of migration unit
is increased in init_migr_record_imsm():7665 to reshape whole array.
During calculation of last migration unit, this should be in mind also,
otherwise checkpoint (max-1) is always written and reshape
is never finished in mdadm.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Tue, 14 Jun 2011 02:46:53 +0000 (12:46 +1000)]
imsm: FIX: Migration Raid0->Raid5 cannot be restarted correctly
When array raid0 is migrated to raid5, reshape cannot be continued
correctly due to wrong array parameters settings.
Raid disks number is set too big.
There is no need, during raid0->raid5 migration to increase
info->array.raid_disks, it is already set to final value using
designation map information.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Tue, 14 Jun 2011 02:42:22 +0000 (12:42 +1000)]
imsm: FIX: Raid5 data corruption data recovering from backup
Sporadicaly when Raid5's data are restored from backup area,
corruption occurs.
It doesn't happen if reshape process is beyond critical section.
Root cause of the problem is passing wrong starting point in
restore_stripes(). It was hard coded to 0 so far.
This causes that parity disks position in first stripe was always set
to the last raid disk. This position should depend on data position in array.
Proper start position was set and pointer for restoring data
(copy area address) is adjusted to passed start parameter.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Tue, 14 Jun 2011 02:42:16 +0000 (12:42 +1000)]
imsm: FIX: Disable automatic metadata rollback for broken reshape
mdmon cannot rollback metadata changes automatically.
It can break reshape process in the way that in case of reshape break
user will not be able to deal with broken reshape due to lack of information
about reshape geometry.
mdadm (process that invokes reshape) doesn't make any rollback to allow
for user action. mdmon should not do this either unless it knows for sure
it is save. such knowledge is not available for automatic rollback.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Thu, 9 Jun 2011 03:00:55 +0000 (13:00 +1000)]
FIX: Move buffer to next location
When no output file is given save_stripes() should collect amount of stripes
in passed buffer. Currently all stripes are saved in the same area in passed
buffer. This causes that last stripe is returned on buffer begin only.
Increase buffer (buf) pointer when save_stripes() is about switch to next
stripe operation. This allows for proper buffer filling as input parameter
length directs.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Thu, 9 Jun 2011 03:00:55 +0000 (13:00 +1000)]
imsm: FIX: Move reshape_progress forward
When array under reshape is assembled, reshape position used in sysfs_set_array()
should be set to position after recovered from backup area.
This avoids data corruption due to reshape the same array area again.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Thu, 9 Jun 2011 03:00:55 +0000 (13:00 +1000)]
imsm: FIX: Detect failed devices during recover_backup_imsm()
Detect in recover_backup_imsm() if not opened disks number is smaller
than allowed degradation for given raid level. This allows for reshape restart
on degraded array.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Thu, 9 Jun 2011 03:00:55 +0000 (13:00 +1000)]
imsm: FIX: Use metadata information for restore_stripes() and save_stripes()
For raid0 reshape imsm uses degraded raid4 for this operation.
Using real raid level (raid0) for stripe calculation causes no need
for parity calculation and can speed up reshape process.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Thu, 9 Jun 2011 03:00:55 +0000 (13:00 +1000)]
imsm: FIX: Do not use pba_of_lba0 for copy position calculation
imsm_manage_reshape() should not shift start copy position.
This offset is passed to manage reshape function /and it is used/
as input parameter in offsets table already.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Thu, 9 Jun 2011 03:00:55 +0000 (13:00 +1000)]
imsm: FIX: Check layout for level migration
When user doesn't specify raid 5 layout for raid0->rai5 migration,
layout structure member is uninitialized. Earlier it cannot be determined
if it is correct or not.
In metadata handle proper verification is placed.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Thu, 9 Jun 2011 03:00:55 +0000 (13:00 +1000)]
imsm: FIX: Max position could not be rounded to MB
When rounded array size information from metadata is used for number
of migration units calculation it can occurs that result of units
can be smaller (-1) than required due to used (rounded) array size).
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Thu, 9 Jun 2011 03:00:55 +0000 (13:00 +1000)]
imsm: FIX: Detect migration end during migration record saving
Checkpoint should be saved when migration is in progress only.
End of reshape (based on passes status) should be detected and it should
not cause abort of reshape/check-pointing/ operation.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
NeilBrown [Thu, 9 Jun 2011 02:42:02 +0000 (12:42 +1000)]
Fix some fall-out from recent memset-zero for getinfo_super
container_content_imsm was setting info->next before calling
getinfo_super_imsm_container which now zeros everything.
So move that assignment to afterwards.
So both imsm and ddf were assuming info->disk.raid_disk means
something but it doesn't. So fix those.
Adam Kwolek [Wed, 8 Jun 2011 07:13:30 +0000 (17:13 +1000)]
Do not use backup file for external metadata
When external metatdata handler supports manage_reshape()
and recover_backup() functions in super switch backup file is not required
and can be omitted. For backup purposes metadata specific mechanisms
are used.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Wed, 8 Jun 2011 07:13:26 +0000 (17:13 +1000)]
FIX: Enable metadata updates for raid0
When raid0 is takeovered to degraded raid4, metadata updates has to be
applied via mdmon (raid4 has to be monitored).
It is not possible due to no update_tail pointer initialization
in supertype structure.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Wed, 8 Jun 2011 07:12:39 +0000 (17:12 +1000)]
imsm: Add metadata update type for general migration check-pointing
There are 2 places for keeping checkpoint information:
- metadata (per volume information used during volume
initialization and rebuilding).
- migration record (per container information used during
migration/reshape)
During reshape both checkpoints has to contains the same information.
To do this mdadm will send metadta update with checkpoint information.
Note: Checkpoint information consistence is not critical. During general
migration restart, information from migration record is used only.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Wed, 8 Jun 2011 07:11:49 +0000 (17:11 +1000)]
imsm: Disable checkpoint updating by mdmon for general migration
imsm contains 2 check-pointing mechanism. One (per array) is used for
initialization and rebuild and second (per container) is used for general
migration (reshape). First is controlled by mdmon, second by mdadm.
To avoid conflicts disable mdmon checkpoints updating for general
migration.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Wed, 8 Jun 2011 07:11:23 +0000 (17:11 +1000)]
imsm: Implement recover_backup_imsm() for imsm metadata
Add ability to restore data backed up in General Migration Copy Area
in case of unexpected reshape interruption.
Function restores data during an array assembly and then reshape
is continues from next checkpoint.
Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Wed, 8 Jun 2011 07:11:11 +0000 (17:11 +1000)]
Add reshape restart support for external metadata
Patch introduces support for reshape process restart for external metadata
using metadata specific data handling methods.
It introduces recover_backup() function that restores array to stable state
It is equivalent to Grow_restart() functionality for native metadata.
Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Wed, 8 Jun 2011 07:09:50 +0000 (17:09 +1000)]
imsm: update blocks_per_migr_unit() to support migration record
blocks_per_migr_unit() has to use information from migration record
for general migration case. This causes to pass intel_super pointer
to this function and some other interfaces changes.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Wed, 8 Jun 2011 07:09:09 +0000 (17:09 +1000)]
imsm: Check if array degradation has been changed
Before reshaping every "migration unit", check if array is still usable.
In failed disks number is greater than allowed degradation level, reshape
has to be aborted.
Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Before reshape is started, mdadm should check again if there is only one
array (in container) under reshape. Then function "divides" array in to
"migration units" that can fits migration copy area and enters main loop.
It checks if current "migration unit" requires to be backed up.
If necessary mdadm saves it to copy area and updates migration record.
Then MD-driver is directed to perform reshape step (by "migration unit" size)
and checkpoint is moved forward. In this way reshape is executed until
array ends.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Wed, 8 Jun 2011 07:07:10 +0000 (17:07 +1000)]
imsm: Add wait_for_reshape_imsm() implementation
After each checkpoint mdadm should set new reshaped area and wait
until md finishes reshape. Function wait_for_reshape_imsm() sets
new reshape range and waits for job completion.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Wed, 8 Jun 2011 06:46:37 +0000 (16:46 +1000)]
imsm: check migration compatibility
Under Windows IMSM can reshape arrays in 2 directions
(ascending and decsending).
Under Linux one (ascending) direction is supported at this moment.
Block loading metadata when decsending reshape is detected
Windows also uses optimalization area during reshaping array.
Linux does not support it.
The patch blocks this operation also.
Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Wed, 8 Jun 2011 06:46:35 +0000 (16:46 +1000)]
imsm: Add support for copy area and backup operations
This patch adds methods of manipulating migration record:
init_migr_record_imsm() - initiate migration record at the beginning of
the reshape process
write_imsm_migr_rec() - saves migration record to array.
Migration record is stored on 2 first disks in array only.
save_backup_imsm() - saves critical data stripes to Migration Copy Area
and updates the current migration unit status.
Uses restore_stripes() to format a destination stripe, and to write it
to the Migration Copy Area.
save_checkpoint_imsm() - Updates the current unit status in the
migration record.
Migration record is written to 2 first array disks only (similar to reading
operation).
Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Wed, 8 Jun 2011 06:28:40 +0000 (16:28 +1000)]
Define dummy functions to mdmon.c
Definitions are necessary to compile mdmon.
Metadata specific source code is compiled to mdmon.
Functions used for reshape check pointing:
- restore_stripes()
- save_stripes
- abort_reshape
are not used in mdmon, but they are compiled in it.
To enable mdmon compilation, dummy functions are used.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Wed, 8 Jun 2011 06:24:48 +0000 (16:24 +1000)]
Support restore_stripes() from the given buffer
For external metadata backup location and saving methods depends
on metadata specific implementation details. Currently restore_stripes()
function is able to restore data only from the given backup file handles
and it is used only for assembling partially reshaped arrays.
As this function will be very helpful for external metadata backup
mechanism, add the support for restoring data from the given source buffer.
Add possibility for save_stripes() to work without designation targets.
Save_stripes() can now prepare data for restore_stripes() only.
Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Wed, 8 Jun 2011 06:19:06 +0000 (16:19 +1000)]
imsm: Add migration record to intel_super
IMSM for securing reshape process uses special disk area outside metadata
for reshaped area backup purposes. If just reshaped array area requires
backup, bunch of array stripes prepared for reshape is stored in to
Migration Copy Area. In case of reshape interruption, Option ROM during
restart or mdadm during reshape restart (when no reboot occurs) will
restore Migration Copy Area to designation array. Reshape can be
continued from stable array stable state.
This patch adds support for IMSM migration record structure.
IMSM migration record is stored on the first two disks of IMSM volume
during the migration.
Add function for reading migration record, so mdadm can read (if present)
migration record. Migration record has to be cleared every time
MIGR_GEN_MIGR is started.
Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
NeilBrown [Mon, 23 May 2011 07:21:37 +0000 (17:21 +1000)]
Restore ability to create imsm array from specific devices.
A recent change to improve error messages make it not possible to
create an array from devices that are 'busy'. However if they are
made busy by a container, then the create should be allowed.
NeilBrown [Tue, 10 May 2011 06:30:40 +0000 (16:30 +1000)]
Give suitable error for mdadm /dev/md0 --stop
Options like --stop must come before the device that is being
stopped. If (in --misc mode) a device does not have an option,
nothing will be done to it, which can be confusing.
So report an error in this case.
NeilBrown [Tue, 10 May 2011 03:09:37 +0000 (13:09 +1000)]
Grow: allow auto-readonly arrays to be reshaped.
In an array is auto-readonly then a reshape will not start.
But auto-readonly is only wanted until something is explicitly
done to acknowledge that the array is really wanted.
So it is perfectly correct to switch an auto-readonly array to
'clean' if a reshape has been requested.
NeilBrown [Tue, 10 May 2011 02:53:51 +0000 (12:53 +1000)]
Grow: handle abort/restart of grow while being monitored.
If a device fails while the grow is being monitored but the array is
still functional, the Grow will appear to abort and then almost
instantly restart from where it was up to.
So if it appears to abort, wait up to 10 seconds for a restart (it
should be much much less than this.
NeilBrown [Tue, 10 May 2011 00:44:00 +0000 (10:44 +1000)]
Grow: When setting component size make sure components are ready.
If you change the size of a member of an array (e.g. it might be a
dm device that can be resized, or on a smart storage device), md
doesn't notice and so the space cannot be used without explicitly
telling md that the device is bigger.
This change causes "mdadm --grow --size=...." to make sure each
component device is making at least that much space available if it
can.
Normally usage of "--size=max" will cause all devices to make max
space available, the md will use as much as it can of that.
When level is changed from raid0 to raid5 memory is required for replace device
smaller device/array object.
This memory is allocated in manager context in prepare_update()
Prepare_update() is called in manager context so memory allocation are
allowed here. This allows us to look for spare devices for meta update.
Introducing raid0->raid5 level migration metadata update structure
is prepared for future use.
Adding spare device is required to hold additional raid5 parity.
Mdadm just checks for spares, but it is not included in update.
If there are no spares available, abort. Otherwise we will create
degraded array what should be not allowed.
Mdmon will decide what spare device is used for parity.
Adam Kwolek [Mon, 2 May 2011 06:12:03 +0000 (16:12 +1000)]
imsm: FIX: Do not write check-point '0'
When 2 arrays are configured in container and arrays are reassembled during
rebuild or initialization, checkpoint for one array can be reset. It depends
on arrays assembly order.
Scenario:
1. Create 2 arrays (e.g. raid5)
2. Add spare to container
3. Degrade arrays /rebuild starts on array #1 and continues to n%/
4. Reassembly arrays
5. Rebuild starts on array #2 /because of assembly order/ from 0%
6. On first checkpoint stored for array #2 (non 0 value), checkpoint
for array #1 is cleared /it is delayed rebuild in md, so progress is 0/
7. Rebuild on #1 starts from n% /it was configured before checkpoint
was cleared/.
Any next reassembly during rebuild of #2 array (after p.6) causes
checkpoint information lost for array #1.
Solution is not store checkpoint for progress == 0.
Checkpoint is set to 0 when rebuild/initialization starts.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Tue, 19 Apr 2011 07:25:43 +0000 (17:25 +1000)]
FIX: Check correctly raid disks during reshape restart
During reshape restart info->array.raid_disks contains new raid_disks number
It cannot be compared against old disks number. Such check will always fail.
Check raid disks array field against final disks number for restart.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Mon, 18 Apr 2011 00:31:43 +0000 (10:31 +1000)]
FIX: Count correctly added devices
When array is in reshape state raid_disks field contains final disks number.
To know how many disks were added, disk.raid_disk index has to be compared
against old disk number computed using delta_disks.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Mon, 18 Apr 2011 00:31:15 +0000 (10:31 +1000)]
FIX: Set proper raid disks during migration
During migration raid_disks field contains new disks number now.
It should be set old disks number first and then new disks number
to allow md to calculate e.g. delta_disks parameter.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Mon, 18 Apr 2011 00:31:06 +0000 (10:31 +1000)]
FIX: Fiddle raid_disks number when restarting reshape
When restarting a reshape, the value of 'raid_disks' is the *new*
value. The old value is found by subtracting delta_disks.
So before calling analyse_change we must set raid_disks to be the
old value, and then reset it afterwards.
All other fields are cleanly separated with the main field being
the 'old' value and a new_* field available.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Thu, 14 Apr 2011 07:50:17 +0000 (17:50 +1000)]
FIX: Use successfully loaded metadata only
Values greater than 0, means error. We exit from loop on error
with empty super-block pointer when sd pointer is valid.
This cannot be detected by check condition as error.
For sure we shouldn't go forward with error condition.
It leads to throwing exception with core file when metadata handler
wants to access non existing super-block.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Adam Kwolek [Mon, 11 Apr 2011 05:00:13 +0000 (15:00 +1000)]
FIX: Raid0 expansion cannot be restarted
When raid0 expansion is restarted, mdadm refuses to correctly assemble
array because critical section cannot be restored from backup file.
mdadm exits with information:
mdadm: Failed to restore critical section for reshape - sorry.
For raid0 new level is 0, current array level is 4.
Function Grow_restart() doesn't allow for level change.
Grow_restart really shouldn't be checking for level changes.
As they are always instantaneous they should never appear
in the metadata so it doesn't mean anything to check for them.
Signed-off-by: Adam Kwolek <adam.kwolek@intel.com> Signed-off-by: NeilBrown <neilb@suse.de>
Mike Frysinger [Mon, 11 Apr 2011 04:54:42 +0000 (14:54 +1000)]
mdadm/mdmon: use CFLAGS when linking
People often put flags that control ABI options into CFLAGS (like -mcpu)
and don't duplicate them in LDFLAGS because most build systems nowadays
(like autotools) use both when linking. So make that work with mdadm's
custom build system too.
Signed-off-by: Mike Frysinger <vapier@gentoo.org> Signed-off-by: NeilBrown <neilb@suse.de>