The case when array is already degraded has been omitted
by commit 7b99edab28 ("Assemble.c: respect force flag.").
Appropriative support has been added now.
Handlers for "run" and "force" have been divided into independent
routines. Especially force has to be as meaningless as possible.
It respects following rules:
- user agrees to start array as degraded (by --run) or is already
degraded
- raid456 module is in use
- some drives are missing (to limit potential abuses)
It doesn't allow to skip resync on dirty, but not degraded array.
This patch cleans up message generation for external array and makes it
consistent. Following code could be reused also for native.
In current implementation assemble_container_content is called once, in
both Incremental or Assembly mode. Thus makes that partial assembly is
not likely to happen. It is possible, but requires user input.
Partial assembly during reshape fails (sysfs_set_array
error - not yet investigated). For now I put FIXME to mark current
logic as known to be buggy because preexist_cnt contains both exp_cnt
and new_cnt which may produce an incorrect message.
Check for new disks and runstop is unnecessary, so has been removed.
This allows to print assemble status in every case, even if nothing new
happens.
Reported-by: Devon Beets <devon@sigmalabsinc.com>
Signed-off-by: Mariusz Tkaczyk <mariusz.tkaczyk@linux.intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
After raid1 to raid0 migration num_data_stripes value is
incorrect because was additionally divided by 2.
Create dedicated setters for num_data_stripes and num_domains
and propagate it across the code to unify alghoritms and
eliminate similar mistakes.
Signed-off-by: Mariusz Tkaczyk <mariusz.tkaczyk@linux.intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
Print error if chunk size is set as it is not supported.
Signed-off-by: Mateusz Grzonka <mateusz.grzonka@intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
Free memory allocated by structs dl and intel_super.
Allow __free_imsm_disk to decide if fd has to be closed and propagate it
across code instead of direct struct dl freeing.
Signed-off-by: Mateusz Grzonka <mateusz.grzonka@intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
In util.c, there is a building error:
'/md/metadata_version' directive writing 20 bytes into a
region of size between 0 and 255 [-Werror=format-overflow=]
In mapfile.c
It declares the fouth argument as 'int *' in map_update,
but in mdadm.h it's previously declared as an array 'int[4]'
Signed-off-by: Xiao Ni <xni@redhat.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
Spares in imsm belong to containers, not volumes, and must go into
a separate container when assembling the RAID.
Remove association spares with other arrays and make Examine print
separate containers for spares.
Auto assemble without config file already works like this. So make
creating a config file and assembling from it consistent with auto
assemble.
With this change, mdadm -Es will add this line to output if spares
are found:
ARRAY metadata=imsm UUID=00000000:00000000:00000000:00000000
Signed-off-by: Oleksandr Shchirskyi <oleksandr.shchirskyi@linux.intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
Arrays with external metadata do not have spare disks directly
assigned to volumes; spare disks belong to containers and are
moved to arrays when the array is degraded/reshaping.
Thus, the display of zero spare disks in volume details is
incorrect and can be confusing.
Signed-off-by: Oleksandr Shchirskyi <oleksandr.shchirskyi@linux.intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
"Sector Offset" field in Examine output was always printed in 512
byte sectors. Update it to support 4096 sector size.
Signed-off-by: Oleksandr Shchirskyi <oleksandr.shchirskyi@linux.intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
Fix regression caused by the patch 1f5d54a06
("Manage: Call validate_geometry when adding drive to external container")
- mdmonitor passes to Manage() routine dev name as min:mjr.
The open() used in validate_geometry_container()
in both ddf and imsm requires path, replace open calls by dev_open,
which allows to use dev path and min:mjr.
Signed-off-by: Blazej Kucman <blazej.kucman@intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
For super1.0 bitmap offset is -16. So it needs to use int type for bitmap offset.
Fixes: 1fe2e10073 (mdadm/bitmap: locate bitmap calcuate bitmap position wrongly)
Signed-off-by: Xiao Ni <xni@redhat.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
When adding drive to container call validate_geometry to verify whether
drive is supported and can be addded to container.
Remove unused parameters from validate_geometry_imsm_container().
There is no need to pass them.
Don't calculate freesize if it is not mandatory. Make it configurable.
Signed-off-by: Mariusz Tkaczyk <mariusz.tkaczyk@linux.intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
First namespace existence is not quaranted by NVMe specification.
Instead first the smallest one shall be chosen.
Signed-off-by: Mariusz Tkaczyk <mariusz.tkaczyk@linux.intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
Add method for reading sysfs attributes and propagate it across IMSM code.
Signed-off-by: Mariusz Tkaczyk <mariusz.tkaczyk@linux.intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
Each virtual device is linked with parent by "device". This patch adds
possibility to get previous device in sysfs tree.
Depending on device type, there is a different amount of virutal
layers. The best we can do is allow to directly specify how many
"device" links need to be resolved. This approach also allows to get
previous virtual device, which may contain some attributes.
Simplify fd2devname, this function doesn't require new functionality and
shall use generic fd2kname.
For nvme drives represented via nvme-subystem when path to block
device if requested, then return it without translation.
Signed-off-by: Mariusz Tkaczyk <mariusz.tkaczyk@linux.intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
When both --scan and device name is passed to --stop action,
then is executed only for given device. Scan is ignored.
Block the operation when both --scan and device name are passed.
Signed-off-by: Norbert Szulc <norbert.szulc@intel.com>
Signed-off-by: Mariusz Tkaczyk <mariusz.tkaczyk@linux.intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
There are some building errors if treating warning as errors.
Fix them in this patch.
Signed-off-by: Xiao Ni <xni@redhat.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
Expectation that size is always rounded is incorrect.
Just confirm that size is smaller to be certain that update is safe.
Signed-off-by: Mariusz Tkaczyk <mariusz.tkaczyk@linux.intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
This fixes growing containers which was broken with
commit 4ae96c8022 (mdadm: fix reshape from RAID5 to RAID6 with
backup file)
The issue being that containers use the function
wait_for_reshape_isms and expect a number value and not a
string value of "max". The change is to test for external
before setting the correct value.
Signed-off-by: Nigel Croxon <ncroxon@redhat.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
Make -ludev configurable, enabled by default.
To disable it, -DNO_LIBUDEV has to be set explicitly in CXFALGS.
This patch restores commit cab9c67d46 ("mdmonitor: set small delay
once") for configuration without libudev to bring minimal support in
such case.
Signed-off-by: Mariusz Tkaczyk <mariusz.tkaczyk@linux.intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
Add new UEFI TSata variable. Remove CSata variable.
This variable has been never exposed by UEFI.
Remove vulnerability to match different hbas with SATA variable.
Signed-off-by: Mariusz Tkaczyk <mariusz.tkaczyk@linux.intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
Add support for nvme devices which are represented
via nvme-subsystem.
Print warning when multi-path disk is added to RAID.
Signed-off-by: Oleksandr Shchirskyi <oleksandr.shchirskyi@intel.com>
Signed-off-by: Blazej Kucman <blazej.kucman@intel.com>
Signed-off-by: Mariusz Tkaczyk <mariusz.tkaczyk@linux.intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
Make it u64 to align it with curr_migr_init field from migration_area.
Name helpers as vol_curr_migr_unit for differentiation between those
fields. Add ommited fillers in struct migr_record.
Signed-off-by: Mariusz Tkaczyk <mariusz.tkaczyk@linux.intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
Current kernel sysfs interface for the bitmap is limited. It allows
the applying of the bitmap on non-active volumes only.
The reshape operation for a volume with a bitmap should be blocked.
Signed-off-by: Jakub Radtke <jakub.radtke@intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
For external metadata, bitmap should be added only when
explicitly set by the administrator.
They could be additional requirements to consider before
enabling the external metadata's functionality
(e.g., kernel support).
Signed-off-by: Jakub Radtke <jakub.radtke@intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
The patch updates the current bitmap functionality to handle adding
the bitmap on existing volumes.
Signed-off-by: Jakub Radtke <jakub.radtke@intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
Currently, the sysfs interface allows bitmap setup only when volume
is in an inactive state.
For external metadata to add bitmap to existing volume instead of
GROW, the UPDATE operation can be done.
The patch adds a "bitmap" argument to the allowed values for UPDATE.
Signed-off-by: Jakub Radtke <jakub.radtke@intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
When adding a spare to an existing array with bitmap, an additional
initialization (adding bitmap header and preparing the bitmap area)
is required.
Signed-off-by: Jakub Radtke <jakub.radtke@intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
This patch enables the bitmap functionality with the IMSM metadata format.
The functionality covers "internal" bitmap.
There are the following limitation to "internal" metadata:
- creating the bitmap using --grow is not supported
- parameter --bitmap-chunk is ignored (default value 64MiB is used)
Signed-off-by: Jakub Radtke <jakub.radtke@intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
The patch enables the implementation of a write-intent bitmap for external
metadata.
Configuration of the internal bitmaps for non-native metadata requires the
extension in superswitch to perform an additional sysfs setup before the
array is activated.
Signed-off-by: Jakub Radtke <jakub.radtke@intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
Current mdstat read functionality is not working correctly
for the volumes with the write-intent bitmap.
It affects rebuild and reshape use cases.
Signed-off-by: Jakub Radtke <jakub.radtke@intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
It can't remove the disk if there are not enough disks. For example, raid5 can't remove the
second disk. If the second disk is unplug from machine, it's better show missing and the raid
should be FAILED. It's better for administrator to monitor the raid.
Signed-off-by: Xiao Ni <xni@redhat.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
IMSM does more than comparing metadata and errors reported directly
from compare_super_imsm can be useful.
Add verbose flag to compare_super method and make all not critical
error printing configurable.
Signed-off-by: Mariusz Tkaczyk <mariusz.tkaczyk@linux.intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
If mdmonitor is awaken by event, wait for udev to finish
events processing, to eliminate the race between udev and mdadm
when spare has been added and need to be moved by mdmonitor
Signed-off-by: Oleksandr Shchirskyi <oleksandr.shchirskyi@intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
Partial Parity Log (PPL) was not documented in the man md.
Added brief info about PPL.
Signed-off-by: Oleksandr Shchirskyi <oleksandr.shchirskyi@intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
IMSM super keeps open descriptors in super->disks structure, they are
reliable and should be chosen if possible. The repeatedly called open
and close during reshape generates redundant udev change events on
each member drive.
Signed-off-by: Mariusz Tkaczyk <mariusz.tkaczyk@linux.intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
With kernel commit 480523feae58 ("md: only call set_in_sync() when it
is expected to succeed."), mddev->in_sync in clustered array is always
zero. It makes metadata resync_offset to always zero.
When assembling a clusterd array with "-U no-bitmap" option, kernel
md layer "mddev->resync_offset == 0" and "mddev->bitmap == NULL" will
trigger raid1 do sync on every bitmap chunk. the sync action is useless,
we should avoid it.
Related kernel flow:
```
md_do_sync
mddev->pers->sync_request
raid1_sync_request
md_bitmap_start_sync(mddev->bitmap, sector_nr, &sync_blocks, 1)
__bitmap_start_sync(bitmap, offset,&blocks1, degraded)
if (bitmap == NULL) {/* FIXME or bitmap set as 'failed' */
*blocks = 1024;
return 1; /* always resync if no bitmap */
}
```
Reprodusible steps:
```
node1 # mdadm -C /dev/md0 -b clustered -e 1.2 -n 2 -l mirror /dev/sd{a,b}
node1 # mdadm -Ss
(in another shell, executing & watching: watch -n 1 'cat /proc/mdstat')
node1 # mdadm -A -U no-bitmap /dev/md0 /dev/sd{a,b}
```
Signed-off-by: Zhao Heming <heming.zhao@suse.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
I've seen a case where the dev_roles list of a linear array
was corrupt. ->max_dev was > 128 and > raid_disks, and the
extra slots were '0', not 0xFFFE or 0xFFFF.
This caused problems when a 128th device was added.
So:
1/ make Grow_Add_device more robust so that if numbers
look wrong, it fails-safe.
2/ make examine_super1() report details if the dev_roles
array is corrupt.
Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
Reshaping a 3-disk RAID5 to 4-disk RAID6 will cause a hang of
the resync after the grow.
Adding a spare disk to avoid degrading the array when growing
is successful, but not successful when supplying a backup file
on the command line. If the reshape job is not already running,
set the sync_max value to max.
Signed-off-by: Nigel Croxon <ncroxon@redhat.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
Now it only adds bitmap offset based on cluster nodes. It's not right. It needs to
add per node bitmap space to find next node bitmap position.
Signed-off-by: Xiao Ni <xni@redhat.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
The dumped metadata files are shown as below
localhost:~ # ll -ih test/
total 16K
34565564 -rw-r--r-- 2 root root 1.0G Dec 14 21:15
scsi-0QEMU_QEMU_HARDDISK_drive-scsi0-0-0-3
34565563 -rw-r--r-- 2 root root 1.0G Dec 14 21:15
scsi-0QEMU_QEMU_HARDDISK_drive-scsi0-0-0-4
34565563 -rw-r--r-- 2 root root 1.0G Dec 14 21:15 sda
34565564 -rw-r--r-- 2 root root 1.0G Dec 14 21:15 sdb
It reports such error when trying to restore metadata for /dev/sda
localhost:~ # mdadm --restore=test /dev/sda
mdadm: test/scsi-0QEMU_QEMU_HARDDISK_drive-scsi0-0-0-4 is not the same
size as /dev/sda - cannot restore.
It's because the stb value has been changed to other metadata file in
the while statement.
Signed-off-by: Lidong Zhong <lidong.zhong@suse.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
If policy is set then mdmonitor is responsible for moving spares.
This logic is reduntant and potentialy dangerus, spare could be moved at
initrd stage depending on drives appearance order.
Remove it.
Signed-off-by: Mariusz Tkaczyk <mariusz.tkaczyk@linux.intel.com>
Signed-off-by: Jes Sorensen <jsorensen@fb.com>
Grow continue via service or fork is started during raid assembly.
If raid was assembled in initrd it will be newer restarted after
switch root.
Add udev support for starting mdadm-grow-continue service.
Signed-off-by: Mariusz Tkaczyk <mariusz.tkaczyk@intel.com>
Make install causes installation of binaries, udev and man.
This commit contains new target make install-bin, which
results in installation of binaries only.
Signed-off-by: Kinga Tanska <kinga.tanska@intel.com>
After kernel md module commit 480523feae581, in clustered env,
mddev->in_sync always zero, it will make array.state never set
up MD_SB_CLEAN. it causes "mdadm -D /dev/mdX" show state 'active'
all the time.
bitmap.c: add a new API IsBitmapDirty() to support inquiry bitmap
dirty or clean.
Signed-off-by: Zhao Heming <heming.zhao@suse.com>
oshchirs
Mateusz Grzonka
Mariusz Tkaczyk
Xiao Ni
Oleksandr Shchirskyi
Blazej Kucman
Norbert Szulc
Jes Sorensen
Nigel Croxon
Jakub Radtke
Zhao Heming
NeilBrown
Lidong Zhong
Tkaczyk Mariusz
Kinga Tanska