raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
/*
|
|
|
|
* Copyright (C) 2015 Shaohua Li <shli@fb.com>
|
|
|
|
*
|
|
|
|
* This program is free software; you can redistribute it and/or modify it
|
|
|
|
* under the terms and conditions of the GNU General Public License,
|
|
|
|
* version 2, as published by the Free Software Foundation.
|
|
|
|
*
|
|
|
|
* This program is distributed in the hope it will be useful, but WITHOUT
|
|
|
|
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
|
|
|
|
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
|
|
|
|
* more details.
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
#include <linux/kernel.h>
|
|
|
|
#include <linux/wait.h>
|
|
|
|
#include <linux/blkdev.h>
|
|
|
|
#include <linux/slab.h>
|
|
|
|
#include <linux/raid/md_p.h>
|
2015-10-28 16:41:25 +01:00
|
|
|
#include <linux/crc32c.h>
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
#include <linux/random.h>
|
|
|
|
#include "md.h"
|
|
|
|
#include "raid5.h"
|
|
|
|
|
|
|
|
/*
|
|
|
|
* metadata/data stored in disk with 4k size unit (a block) regardless
|
|
|
|
* underneath hardware sector size. only works with PAGE_SIZE == 4096
|
|
|
|
*/
|
|
|
|
#define BLOCK_SECTORS (8)
|
|
|
|
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
/*
|
|
|
|
* reclaim runs every 1/4 disk size or 10G reclaimable space. This can prevent
|
|
|
|
* recovery scans a very long log
|
|
|
|
*/
|
|
|
|
#define RECLAIM_MAX_FREE_SPACE (10 * 1024 * 1024 * 2) /* sector */
|
|
|
|
#define RECLAIM_MAX_FREE_SPACE_SHIFT (2)
|
|
|
|
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
struct r5l_log {
|
|
|
|
struct md_rdev *rdev;
|
|
|
|
|
|
|
|
u32 uuid_checksum;
|
|
|
|
|
|
|
|
sector_t device_size; /* log device size, round to
|
|
|
|
* BLOCK_SECTORS */
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
sector_t max_free_space; /* reclaim run if free space is at
|
|
|
|
* this size */
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
|
|
|
|
sector_t last_checkpoint; /* log tail. where recovery scan
|
|
|
|
* starts from */
|
|
|
|
u64 last_cp_seq; /* log tail sequence */
|
|
|
|
|
|
|
|
sector_t log_start; /* log head. where new data appends */
|
|
|
|
u64 seq; /* log head sequence */
|
|
|
|
|
2015-10-05 09:31:06 +02:00
|
|
|
sector_t next_checkpoint;
|
|
|
|
u64 next_cp_seq;
|
|
|
|
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
struct mutex io_mutex;
|
|
|
|
struct r5l_io_unit *current_io; /* current io_unit accepting new data */
|
|
|
|
|
|
|
|
spinlock_t io_list_lock;
|
|
|
|
struct list_head running_ios; /* io_units which are still running,
|
|
|
|
* and have not yet been completely
|
|
|
|
* written to the log */
|
|
|
|
struct list_head io_end_ios; /* io_units which have been completely
|
|
|
|
* written to the log but not yet written
|
|
|
|
* to the RAID */
|
2015-09-02 22:49:46 +02:00
|
|
|
struct list_head flushing_ios; /* io_units which are waiting for log
|
|
|
|
* cache flush */
|
2015-10-05 09:31:07 +02:00
|
|
|
struct list_head finished_ios; /* io_units which settle down in log disk */
|
2015-09-02 22:49:46 +02:00
|
|
|
struct bio flush_bio;
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
|
|
|
|
struct kmem_cache *io_kc;
|
|
|
|
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
struct md_thread *reclaim_thread;
|
|
|
|
unsigned long reclaim_target; /* number of space that need to be
|
|
|
|
* reclaimed. if it's 0, reclaim spaces
|
|
|
|
* used by io_units which are in
|
|
|
|
* IO_UNIT_STRIPE_END state (eg, reclaim
|
|
|
|
* dones't wait for specific io_unit
|
|
|
|
* switching to IO_UNIT_STRIPE_END
|
|
|
|
* state) */
|
2015-09-02 22:49:47 +02:00
|
|
|
wait_queue_head_t iounit_wait;
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
struct list_head no_space_stripes; /* pending stripes, log has no space */
|
|
|
|
spinlock_t no_space_stripes_lock;
|
2015-10-05 09:31:09 +02:00
|
|
|
|
|
|
|
bool need_cache_flush;
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
};
|
|
|
|
|
|
|
|
/*
|
|
|
|
* an IO range starts from a meta data block and end at the next meta data
|
|
|
|
* block. The io unit's the meta data block tracks data/parity followed it. io
|
|
|
|
* unit is written to log disk with normal write, as we always flush log disk
|
|
|
|
* first and then start move data to raid disks, there is no requirement to
|
|
|
|
* write io unit with FLUSH/FUA
|
|
|
|
*/
|
|
|
|
struct r5l_io_unit {
|
|
|
|
struct r5l_log *log;
|
|
|
|
|
|
|
|
struct page *meta_page; /* store meta block */
|
|
|
|
int meta_offset; /* current offset in meta_page */
|
|
|
|
|
|
|
|
struct bio_list bios;
|
|
|
|
atomic_t pending_io; /* pending bios not written to log yet */
|
|
|
|
struct bio *current_bio;/* current_bio accepting new data */
|
|
|
|
|
|
|
|
atomic_t pending_stripe;/* how many stripes not flushed to raid */
|
|
|
|
u64 seq; /* seq number of the metablock */
|
|
|
|
sector_t log_start; /* where the io_unit starts */
|
|
|
|
sector_t log_end; /* where the io_unit ends */
|
|
|
|
struct list_head log_sibling; /* log->running_ios */
|
|
|
|
struct list_head stripe_list; /* stripes added to the io_unit */
|
|
|
|
|
|
|
|
int state;
|
|
|
|
};
|
|
|
|
|
|
|
|
/* r5l_io_unit state */
|
|
|
|
enum r5l_io_unit_state {
|
|
|
|
IO_UNIT_RUNNING = 0, /* accepting new IO */
|
|
|
|
IO_UNIT_IO_START = 1, /* io_unit bio start writing to log,
|
|
|
|
* don't accepting new bio */
|
|
|
|
IO_UNIT_IO_END = 2, /* io_unit bio finish writing to log */
|
2015-09-02 22:49:46 +02:00
|
|
|
IO_UNIT_STRIPE_END = 3, /* stripes data finished writing to raid */
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
};
|
|
|
|
|
|
|
|
static sector_t r5l_ring_add(struct r5l_log *log, sector_t start, sector_t inc)
|
|
|
|
{
|
|
|
|
start += inc;
|
|
|
|
if (start >= log->device_size)
|
|
|
|
start = start - log->device_size;
|
|
|
|
return start;
|
|
|
|
}
|
|
|
|
|
|
|
|
static sector_t r5l_ring_distance(struct r5l_log *log, sector_t start,
|
|
|
|
sector_t end)
|
|
|
|
{
|
|
|
|
if (end >= start)
|
|
|
|
return end - start;
|
|
|
|
else
|
|
|
|
return end + log->device_size - start;
|
|
|
|
}
|
|
|
|
|
|
|
|
static bool r5l_has_free_space(struct r5l_log *log, sector_t size)
|
|
|
|
{
|
|
|
|
sector_t used_size;
|
|
|
|
|
|
|
|
used_size = r5l_ring_distance(log, log->last_checkpoint,
|
|
|
|
log->log_start);
|
|
|
|
|
|
|
|
return log->device_size > used_size + size;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct r5l_io_unit *r5l_alloc_io_unit(struct r5l_log *log)
|
|
|
|
{
|
|
|
|
struct r5l_io_unit *io;
|
|
|
|
/* We can't handle memory allocate failure so far */
|
|
|
|
gfp_t gfp = GFP_NOIO | __GFP_NOFAIL;
|
|
|
|
|
|
|
|
io = kmem_cache_zalloc(log->io_kc, gfp);
|
|
|
|
io->log = log;
|
|
|
|
io->meta_page = alloc_page(gfp | __GFP_ZERO);
|
|
|
|
|
|
|
|
bio_list_init(&io->bios);
|
|
|
|
INIT_LIST_HEAD(&io->log_sibling);
|
|
|
|
INIT_LIST_HEAD(&io->stripe_list);
|
|
|
|
io->state = IO_UNIT_RUNNING;
|
|
|
|
return io;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void r5l_free_io_unit(struct r5l_log *log, struct r5l_io_unit *io)
|
|
|
|
{
|
|
|
|
__free_page(io->meta_page);
|
|
|
|
kmem_cache_free(log->io_kc, io);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void r5l_move_io_unit_list(struct list_head *from, struct list_head *to,
|
|
|
|
enum r5l_io_unit_state state)
|
|
|
|
{
|
|
|
|
struct r5l_io_unit *io;
|
|
|
|
|
|
|
|
while (!list_empty(from)) {
|
|
|
|
io = list_first_entry(from, struct r5l_io_unit, log_sibling);
|
|
|
|
/* don't change list order */
|
|
|
|
if (io->state >= state)
|
|
|
|
list_move_tail(&io->log_sibling, to);
|
|
|
|
else
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void __r5l_set_io_unit_state(struct r5l_io_unit *io,
|
|
|
|
enum r5l_io_unit_state state)
|
|
|
|
{
|
|
|
|
if (WARN_ON(io->state >= state))
|
|
|
|
return;
|
|
|
|
io->state = state;
|
|
|
|
}
|
|
|
|
|
2015-10-05 09:31:08 +02:00
|
|
|
static void r5l_io_run_stripes(struct r5l_io_unit *io)
|
|
|
|
{
|
|
|
|
struct stripe_head *sh, *next;
|
|
|
|
|
|
|
|
list_for_each_entry_safe(sh, next, &io->stripe_list, log_list) {
|
|
|
|
list_del_init(&sh->log_list);
|
|
|
|
set_bit(STRIPE_HANDLE, &sh->state);
|
|
|
|
raid5_release_stripe(sh);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
/* XXX: totally ignores I/O errors */
|
2015-10-05 09:31:09 +02:00
|
|
|
static void r5l_log_run_stripes(struct r5l_log *log)
|
|
|
|
{
|
|
|
|
struct r5l_io_unit *io, *next;
|
|
|
|
|
|
|
|
assert_spin_locked(&log->io_list_lock);
|
|
|
|
|
|
|
|
list_for_each_entry_safe(io, next, &log->running_ios, log_sibling) {
|
|
|
|
/* don't change list order */
|
|
|
|
if (io->state < IO_UNIT_IO_END)
|
|
|
|
break;
|
|
|
|
|
|
|
|
list_move_tail(&io->log_sibling, &log->finished_ios);
|
|
|
|
r5l_io_run_stripes(io);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
static void r5l_log_endio(struct bio *bio)
|
|
|
|
{
|
|
|
|
struct r5l_io_unit *io = bio->bi_private;
|
|
|
|
struct r5l_log *log = io->log;
|
2015-09-02 22:49:48 +02:00
|
|
|
unsigned long flags;
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
|
|
|
|
bio_put(bio);
|
|
|
|
|
|
|
|
if (!atomic_dec_and_test(&io->pending_io))
|
|
|
|
return;
|
|
|
|
|
2015-09-02 22:49:48 +02:00
|
|
|
spin_lock_irqsave(&log->io_list_lock, flags);
|
|
|
|
__r5l_set_io_unit_state(io, IO_UNIT_IO_END);
|
2015-10-05 09:31:09 +02:00
|
|
|
if (log->need_cache_flush)
|
|
|
|
r5l_move_io_unit_list(&log->running_ios, &log->io_end_ios,
|
|
|
|
IO_UNIT_IO_END);
|
|
|
|
else
|
|
|
|
r5l_log_run_stripes(log);
|
2015-09-02 22:49:48 +02:00
|
|
|
spin_unlock_irqrestore(&log->io_list_lock, flags);
|
|
|
|
|
2015-10-05 09:31:09 +02:00
|
|
|
if (log->need_cache_flush)
|
|
|
|
md_wakeup_thread(log->rdev->mddev->thread);
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static void r5l_submit_current_io(struct r5l_log *log)
|
|
|
|
{
|
|
|
|
struct r5l_io_unit *io = log->current_io;
|
|
|
|
struct r5l_meta_block *block;
|
|
|
|
struct bio *bio;
|
2015-09-02 22:49:48 +02:00
|
|
|
unsigned long flags;
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
u32 crc;
|
|
|
|
|
|
|
|
if (!io)
|
|
|
|
return;
|
|
|
|
|
|
|
|
block = page_address(io->meta_page);
|
|
|
|
block->meta_size = cpu_to_le32(io->meta_offset);
|
2015-10-28 16:41:25 +01:00
|
|
|
crc = crc32c_le(log->uuid_checksum, block, PAGE_SIZE);
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
block->checksum = cpu_to_le32(crc);
|
|
|
|
|
|
|
|
log->current_io = NULL;
|
2015-09-02 22:49:48 +02:00
|
|
|
spin_lock_irqsave(&log->io_list_lock, flags);
|
|
|
|
__r5l_set_io_unit_state(io, IO_UNIT_IO_START);
|
|
|
|
spin_unlock_irqrestore(&log->io_list_lock, flags);
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
|
|
|
|
while ((bio = bio_list_pop(&io->bios))) {
|
|
|
|
/* all IO must start from rdev->data_offset */
|
|
|
|
bio->bi_iter.bi_sector += log->rdev->data_offset;
|
|
|
|
submit_bio(WRITE, bio);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct r5l_io_unit *r5l_new_meta(struct r5l_log *log)
|
|
|
|
{
|
|
|
|
struct r5l_io_unit *io;
|
|
|
|
struct r5l_meta_block *block;
|
|
|
|
struct bio *bio;
|
|
|
|
|
|
|
|
io = r5l_alloc_io_unit(log);
|
|
|
|
|
|
|
|
block = page_address(io->meta_page);
|
|
|
|
block->magic = cpu_to_le32(R5LOG_MAGIC);
|
|
|
|
block->version = R5LOG_VERSION;
|
|
|
|
block->seq = cpu_to_le64(log->seq);
|
|
|
|
block->position = cpu_to_le64(log->log_start);
|
|
|
|
|
|
|
|
io->log_start = log->log_start;
|
|
|
|
io->meta_offset = sizeof(struct r5l_meta_block);
|
|
|
|
io->seq = log->seq;
|
|
|
|
|
|
|
|
bio = bio_kmalloc(GFP_NOIO | __GFP_NOFAIL, BIO_MAX_PAGES);
|
|
|
|
io->current_bio = bio;
|
|
|
|
bio->bi_rw = WRITE;
|
|
|
|
bio->bi_bdev = log->rdev->bdev;
|
|
|
|
bio->bi_iter.bi_sector = log->log_start;
|
|
|
|
bio_add_page(bio, io->meta_page, PAGE_SIZE, 0);
|
|
|
|
bio->bi_end_io = r5l_log_endio;
|
|
|
|
bio->bi_private = io;
|
|
|
|
|
|
|
|
bio_list_add(&io->bios, bio);
|
|
|
|
atomic_inc(&io->pending_io);
|
|
|
|
|
|
|
|
log->seq++;
|
|
|
|
log->log_start = r5l_ring_add(log, log->log_start, BLOCK_SECTORS);
|
|
|
|
io->log_end = log->log_start;
|
|
|
|
/* current bio hit disk end */
|
|
|
|
if (log->log_start == 0)
|
|
|
|
io->current_bio = NULL;
|
|
|
|
|
|
|
|
spin_lock_irq(&log->io_list_lock);
|
|
|
|
list_add_tail(&io->log_sibling, &log->running_ios);
|
|
|
|
spin_unlock_irq(&log->io_list_lock);
|
|
|
|
|
|
|
|
return io;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int r5l_get_meta(struct r5l_log *log, unsigned int payload_size)
|
|
|
|
{
|
|
|
|
struct r5l_io_unit *io;
|
|
|
|
|
|
|
|
io = log->current_io;
|
|
|
|
if (io && io->meta_offset + payload_size > PAGE_SIZE)
|
|
|
|
r5l_submit_current_io(log);
|
|
|
|
io = log->current_io;
|
|
|
|
if (io)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
log->current_io = r5l_new_meta(log);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void r5l_append_payload_meta(struct r5l_log *log, u16 type,
|
|
|
|
sector_t location,
|
|
|
|
u32 checksum1, u32 checksum2,
|
|
|
|
bool checksum2_valid)
|
|
|
|
{
|
|
|
|
struct r5l_io_unit *io = log->current_io;
|
|
|
|
struct r5l_payload_data_parity *payload;
|
|
|
|
|
|
|
|
payload = page_address(io->meta_page) + io->meta_offset;
|
|
|
|
payload->header.type = cpu_to_le16(type);
|
|
|
|
payload->header.flags = cpu_to_le16(0);
|
|
|
|
payload->size = cpu_to_le32((1 + !!checksum2_valid) <<
|
|
|
|
(PAGE_SHIFT - 9));
|
|
|
|
payload->location = cpu_to_le64(location);
|
|
|
|
payload->checksum[0] = cpu_to_le32(checksum1);
|
|
|
|
if (checksum2_valid)
|
|
|
|
payload->checksum[1] = cpu_to_le32(checksum2);
|
|
|
|
|
|
|
|
io->meta_offset += sizeof(struct r5l_payload_data_parity) +
|
|
|
|
sizeof(__le32) * (1 + !!checksum2_valid);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void r5l_append_payload_page(struct r5l_log *log, struct page *page)
|
|
|
|
{
|
|
|
|
struct r5l_io_unit *io = log->current_io;
|
|
|
|
|
|
|
|
alloc_bio:
|
|
|
|
if (!io->current_bio) {
|
|
|
|
struct bio *bio;
|
|
|
|
|
|
|
|
bio = bio_kmalloc(GFP_NOIO | __GFP_NOFAIL, BIO_MAX_PAGES);
|
|
|
|
bio->bi_rw = WRITE;
|
|
|
|
bio->bi_bdev = log->rdev->bdev;
|
|
|
|
bio->bi_iter.bi_sector = log->log_start;
|
|
|
|
bio->bi_end_io = r5l_log_endio;
|
|
|
|
bio->bi_private = io;
|
|
|
|
bio_list_add(&io->bios, bio);
|
|
|
|
atomic_inc(&io->pending_io);
|
|
|
|
io->current_bio = bio;
|
|
|
|
}
|
|
|
|
if (!bio_add_page(io->current_bio, page, PAGE_SIZE, 0)) {
|
|
|
|
io->current_bio = NULL;
|
|
|
|
goto alloc_bio;
|
|
|
|
}
|
|
|
|
log->log_start = r5l_ring_add(log, log->log_start,
|
|
|
|
BLOCK_SECTORS);
|
|
|
|
/* current bio hit disk end */
|
|
|
|
if (log->log_start == 0)
|
|
|
|
io->current_bio = NULL;
|
|
|
|
|
|
|
|
io->log_end = log->log_start;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void r5l_log_stripe(struct r5l_log *log, struct stripe_head *sh,
|
|
|
|
int data_pages, int parity_pages)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
int meta_size;
|
|
|
|
struct r5l_io_unit *io;
|
|
|
|
|
|
|
|
meta_size =
|
|
|
|
((sizeof(struct r5l_payload_data_parity) + sizeof(__le32))
|
|
|
|
* data_pages) +
|
|
|
|
sizeof(struct r5l_payload_data_parity) +
|
|
|
|
sizeof(__le32) * parity_pages;
|
|
|
|
|
|
|
|
r5l_get_meta(log, meta_size);
|
|
|
|
io = log->current_io;
|
|
|
|
|
|
|
|
for (i = 0; i < sh->disks; i++) {
|
|
|
|
if (!test_bit(R5_Wantwrite, &sh->dev[i].flags))
|
|
|
|
continue;
|
|
|
|
if (i == sh->pd_idx || i == sh->qd_idx)
|
|
|
|
continue;
|
|
|
|
r5l_append_payload_meta(log, R5LOG_PAYLOAD_DATA,
|
|
|
|
raid5_compute_blocknr(sh, i, 0),
|
|
|
|
sh->dev[i].log_checksum, 0, false);
|
|
|
|
r5l_append_payload_page(log, sh->dev[i].page);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (sh->qd_idx >= 0) {
|
|
|
|
r5l_append_payload_meta(log, R5LOG_PAYLOAD_PARITY,
|
|
|
|
sh->sector, sh->dev[sh->pd_idx].log_checksum,
|
|
|
|
sh->dev[sh->qd_idx].log_checksum, true);
|
|
|
|
r5l_append_payload_page(log, sh->dev[sh->pd_idx].page);
|
|
|
|
r5l_append_payload_page(log, sh->dev[sh->qd_idx].page);
|
|
|
|
} else {
|
|
|
|
r5l_append_payload_meta(log, R5LOG_PAYLOAD_PARITY,
|
|
|
|
sh->sector, sh->dev[sh->pd_idx].log_checksum,
|
|
|
|
0, false);
|
|
|
|
r5l_append_payload_page(log, sh->dev[sh->pd_idx].page);
|
|
|
|
}
|
|
|
|
|
|
|
|
list_add_tail(&sh->log_list, &io->stripe_list);
|
|
|
|
atomic_inc(&io->pending_stripe);
|
|
|
|
sh->log_io = io;
|
|
|
|
}
|
|
|
|
|
2015-09-02 22:49:48 +02:00
|
|
|
static void r5l_wake_reclaim(struct r5l_log *log, sector_t space);
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
/*
|
|
|
|
* running in raid5d, where reclaim could wait for raid5d too (when it flushes
|
|
|
|
* data from log to raid disks), so we shouldn't wait for reclaim here
|
|
|
|
*/
|
|
|
|
int r5l_write_stripe(struct r5l_log *log, struct stripe_head *sh)
|
|
|
|
{
|
|
|
|
int write_disks = 0;
|
|
|
|
int data_pages, parity_pages;
|
|
|
|
int meta_size;
|
|
|
|
int reserve;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
if (!log)
|
|
|
|
return -EAGAIN;
|
|
|
|
/* Don't support stripe batch */
|
|
|
|
if (sh->log_io || !test_bit(R5_Wantwrite, &sh->dev[sh->pd_idx].flags) ||
|
|
|
|
test_bit(STRIPE_SYNCING, &sh->state)) {
|
|
|
|
/* the stripe is written to log, we start writing it to raid */
|
|
|
|
clear_bit(STRIPE_LOG_TRAPPED, &sh->state);
|
|
|
|
return -EAGAIN;
|
|
|
|
}
|
|
|
|
|
|
|
|
for (i = 0; i < sh->disks; i++) {
|
|
|
|
void *addr;
|
|
|
|
|
|
|
|
if (!test_bit(R5_Wantwrite, &sh->dev[i].flags))
|
|
|
|
continue;
|
|
|
|
write_disks++;
|
|
|
|
/* checksum is already calculated in last run */
|
|
|
|
if (test_bit(STRIPE_LOG_TRAPPED, &sh->state))
|
|
|
|
continue;
|
|
|
|
addr = kmap_atomic(sh->dev[i].page);
|
2015-10-28 16:41:25 +01:00
|
|
|
sh->dev[i].log_checksum = crc32c_le(log->uuid_checksum,
|
|
|
|
addr, PAGE_SIZE);
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
kunmap_atomic(addr);
|
|
|
|
}
|
|
|
|
parity_pages = 1 + !!(sh->qd_idx >= 0);
|
|
|
|
data_pages = write_disks - parity_pages;
|
|
|
|
|
|
|
|
meta_size =
|
|
|
|
((sizeof(struct r5l_payload_data_parity) + sizeof(__le32))
|
|
|
|
* data_pages) +
|
|
|
|
sizeof(struct r5l_payload_data_parity) +
|
|
|
|
sizeof(__le32) * parity_pages;
|
|
|
|
/* Doesn't work with very big raid array */
|
|
|
|
if (meta_size + sizeof(struct r5l_meta_block) > PAGE_SIZE)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
set_bit(STRIPE_LOG_TRAPPED, &sh->state);
|
2015-09-04 23:14:16 +02:00
|
|
|
/*
|
|
|
|
* The stripe must enter state machine again to finish the write, so
|
|
|
|
* don't delay.
|
|
|
|
*/
|
|
|
|
clear_bit(STRIPE_DELAYED, &sh->state);
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
atomic_inc(&sh->count);
|
|
|
|
|
|
|
|
mutex_lock(&log->io_mutex);
|
|
|
|
/* meta + data */
|
|
|
|
reserve = (1 + write_disks) << (PAGE_SHIFT - 9);
|
|
|
|
if (r5l_has_free_space(log, reserve))
|
|
|
|
r5l_log_stripe(log, sh, data_pages, parity_pages);
|
|
|
|
else {
|
|
|
|
spin_lock(&log->no_space_stripes_lock);
|
|
|
|
list_add_tail(&sh->log_list, &log->no_space_stripes);
|
|
|
|
spin_unlock(&log->no_space_stripes_lock);
|
|
|
|
|
|
|
|
r5l_wake_reclaim(log, reserve);
|
|
|
|
}
|
|
|
|
mutex_unlock(&log->io_mutex);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
void r5l_write_stripe_run(struct r5l_log *log)
|
|
|
|
{
|
|
|
|
if (!log)
|
|
|
|
return;
|
|
|
|
mutex_lock(&log->io_mutex);
|
|
|
|
r5l_submit_current_io(log);
|
|
|
|
mutex_unlock(&log->io_mutex);
|
|
|
|
}
|
|
|
|
|
2015-09-02 22:49:49 +02:00
|
|
|
int r5l_handle_flush_request(struct r5l_log *log, struct bio *bio)
|
|
|
|
{
|
|
|
|
if (!log)
|
|
|
|
return -ENODEV;
|
|
|
|
/*
|
|
|
|
* we flush log disk cache first, then write stripe data to raid disks.
|
|
|
|
* So if bio is finished, the log disk cache is flushed already. The
|
|
|
|
* recovery guarantees we can recovery the bio from log disk, so we
|
|
|
|
* don't need to flush again
|
|
|
|
*/
|
|
|
|
if (bio->bi_iter.bi_size == 0) {
|
|
|
|
bio_endio(bio);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
bio->bi_rw &= ~REQ_FLUSH;
|
|
|
|
return -EAGAIN;
|
|
|
|
}
|
|
|
|
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
/* This will run after log space is reclaimed */
|
|
|
|
static void r5l_run_no_space_stripes(struct r5l_log *log)
|
|
|
|
{
|
|
|
|
struct stripe_head *sh;
|
|
|
|
|
|
|
|
spin_lock(&log->no_space_stripes_lock);
|
|
|
|
while (!list_empty(&log->no_space_stripes)) {
|
|
|
|
sh = list_first_entry(&log->no_space_stripes,
|
|
|
|
struct stripe_head, log_list);
|
|
|
|
list_del_init(&sh->log_list);
|
|
|
|
set_bit(STRIPE_HANDLE, &sh->state);
|
|
|
|
raid5_release_stripe(sh);
|
|
|
|
}
|
|
|
|
spin_unlock(&log->no_space_stripes_lock);
|
|
|
|
}
|
|
|
|
|
2015-10-05 09:31:06 +02:00
|
|
|
static sector_t r5l_reclaimable_space(struct r5l_log *log)
|
|
|
|
{
|
|
|
|
return r5l_ring_distance(log, log->last_checkpoint,
|
|
|
|
log->next_checkpoint);
|
|
|
|
}
|
|
|
|
|
2015-10-05 09:31:07 +02:00
|
|
|
static bool r5l_complete_finished_ios(struct r5l_log *log)
|
2015-10-05 09:31:06 +02:00
|
|
|
{
|
|
|
|
struct r5l_io_unit *io, *next;
|
|
|
|
bool found = false;
|
|
|
|
|
|
|
|
assert_spin_locked(&log->io_list_lock);
|
|
|
|
|
2015-10-05 09:31:07 +02:00
|
|
|
list_for_each_entry_safe(io, next, &log->finished_ios, log_sibling) {
|
2015-10-05 09:31:06 +02:00
|
|
|
/* don't change list order */
|
|
|
|
if (io->state < IO_UNIT_STRIPE_END)
|
|
|
|
break;
|
|
|
|
|
|
|
|
log->next_checkpoint = io->log_start;
|
|
|
|
log->next_cp_seq = io->seq;
|
|
|
|
|
|
|
|
list_del(&io->log_sibling);
|
|
|
|
r5l_free_io_unit(log, io);
|
|
|
|
|
|
|
|
found = true;
|
|
|
|
}
|
|
|
|
|
|
|
|
return found;
|
|
|
|
}
|
|
|
|
|
2015-09-02 22:49:48 +02:00
|
|
|
static void __r5l_stripe_write_finished(struct r5l_io_unit *io)
|
|
|
|
{
|
|
|
|
struct r5l_log *log = io->log;
|
|
|
|
unsigned long flags;
|
|
|
|
|
|
|
|
spin_lock_irqsave(&log->io_list_lock, flags);
|
|
|
|
__r5l_set_io_unit_state(io, IO_UNIT_STRIPE_END);
|
2015-10-05 09:31:06 +02:00
|
|
|
|
2015-10-05 09:31:07 +02:00
|
|
|
if (!r5l_complete_finished_ios(log)) {
|
2015-09-04 23:14:05 +02:00
|
|
|
spin_unlock_irqrestore(&log->io_list_lock, flags);
|
|
|
|
return;
|
|
|
|
}
|
2015-09-02 22:49:48 +02:00
|
|
|
|
2015-10-05 09:31:06 +02:00
|
|
|
if (r5l_reclaimable_space(log) > log->max_free_space)
|
2015-09-02 22:49:48 +02:00
|
|
|
r5l_wake_reclaim(log, 0);
|
|
|
|
|
|
|
|
spin_unlock_irqrestore(&log->io_list_lock, flags);
|
|
|
|
wake_up(&log->iounit_wait);
|
|
|
|
}
|
|
|
|
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
void r5l_stripe_write_finished(struct stripe_head *sh)
|
|
|
|
{
|
|
|
|
struct r5l_io_unit *io;
|
|
|
|
|
|
|
|
io = sh->log_io;
|
|
|
|
sh->log_io = NULL;
|
|
|
|
|
2015-09-02 22:49:48 +02:00
|
|
|
if (io && atomic_dec_and_test(&io->pending_stripe))
|
|
|
|
__r5l_stripe_write_finished(io);
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
}
|
|
|
|
|
2015-09-02 22:49:46 +02:00
|
|
|
static void r5l_log_flush_endio(struct bio *bio)
|
|
|
|
{
|
|
|
|
struct r5l_log *log = container_of(bio, struct r5l_log,
|
|
|
|
flush_bio);
|
|
|
|
unsigned long flags;
|
|
|
|
struct r5l_io_unit *io;
|
|
|
|
|
|
|
|
spin_lock_irqsave(&log->io_list_lock, flags);
|
2015-10-05 09:31:08 +02:00
|
|
|
list_for_each_entry(io, &log->flushing_ios, log_sibling)
|
|
|
|
r5l_io_run_stripes(io);
|
2015-10-05 09:31:07 +02:00
|
|
|
list_splice_tail_init(&log->flushing_ios, &log->finished_ios);
|
2015-09-02 22:49:46 +02:00
|
|
|
spin_unlock_irqrestore(&log->io_list_lock, flags);
|
|
|
|
}
|
|
|
|
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
/*
|
|
|
|
* Starting dispatch IO to raid.
|
|
|
|
* io_unit(meta) consists of a log. There is one situation we want to avoid. A
|
|
|
|
* broken meta in the middle of a log causes recovery can't find meta at the
|
|
|
|
* head of log. If operations require meta at the head persistent in log, we
|
|
|
|
* must make sure meta before it persistent in log too. A case is:
|
|
|
|
*
|
|
|
|
* stripe data/parity is in log, we start write stripe to raid disks. stripe
|
|
|
|
* data/parity must be persistent in log before we do the write to raid disks.
|
|
|
|
*
|
|
|
|
* The solution is we restrictly maintain io_unit list order. In this case, we
|
|
|
|
* only write stripes of an io_unit to raid disks till the io_unit is the first
|
|
|
|
* one whose data/parity is in log.
|
|
|
|
*/
|
|
|
|
void r5l_flush_stripe_to_raid(struct r5l_log *log)
|
|
|
|
{
|
2015-09-02 22:49:46 +02:00
|
|
|
bool do_flush;
|
2015-10-05 09:31:09 +02:00
|
|
|
|
|
|
|
if (!log || !log->need_cache_flush)
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
return;
|
|
|
|
|
|
|
|
spin_lock_irq(&log->io_list_lock);
|
2015-09-02 22:49:46 +02:00
|
|
|
/* flush bio is running */
|
|
|
|
if (!list_empty(&log->flushing_ios)) {
|
|
|
|
spin_unlock_irq(&log->io_list_lock);
|
|
|
|
return;
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
}
|
2015-09-02 22:49:46 +02:00
|
|
|
list_splice_tail_init(&log->io_end_ios, &log->flushing_ios);
|
|
|
|
do_flush = !list_empty(&log->flushing_ios);
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
spin_unlock_irq(&log->io_list_lock);
|
2015-09-02 22:49:46 +02:00
|
|
|
|
|
|
|
if (!do_flush)
|
|
|
|
return;
|
|
|
|
bio_reset(&log->flush_bio);
|
|
|
|
log->flush_bio.bi_bdev = log->rdev->bdev;
|
|
|
|
log->flush_bio.bi_end_io = r5l_log_flush_endio;
|
|
|
|
submit_bio(WRITE_FLUSH, &log->flush_bio);
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static void r5l_write_super(struct r5l_log *log, sector_t cp);
|
|
|
|
static void r5l_do_reclaim(struct r5l_log *log)
|
|
|
|
{
|
|
|
|
sector_t reclaim_target = xchg(&log->reclaim_target, 0);
|
2015-10-05 09:31:06 +02:00
|
|
|
sector_t reclaimable;
|
|
|
|
sector_t next_checkpoint;
|
|
|
|
u64 next_cp_seq;
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
|
|
|
|
spin_lock_irq(&log->io_list_lock);
|
|
|
|
/*
|
|
|
|
* move proper io_unit to reclaim list. We should not change the order.
|
|
|
|
* reclaimable/unreclaimable io_unit can be mixed in the list, we
|
|
|
|
* shouldn't reuse space of an unreclaimable io_unit
|
|
|
|
*/
|
|
|
|
while (1) {
|
2015-10-05 09:31:06 +02:00
|
|
|
reclaimable = r5l_reclaimable_space(log);
|
|
|
|
if (reclaimable >= reclaim_target ||
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
(list_empty(&log->running_ios) &&
|
|
|
|
list_empty(&log->io_end_ios) &&
|
2015-09-02 22:49:46 +02:00
|
|
|
list_empty(&log->flushing_ios) &&
|
2015-10-05 09:31:07 +02:00
|
|
|
list_empty(&log->finished_ios)))
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
break;
|
|
|
|
|
2015-10-05 09:31:06 +02:00
|
|
|
md_wakeup_thread(log->rdev->mddev->thread);
|
|
|
|
wait_event_lock_irq(log->iounit_wait,
|
|
|
|
r5l_reclaimable_space(log) > reclaimable,
|
|
|
|
log->io_list_lock);
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
}
|
2015-10-05 09:31:06 +02:00
|
|
|
|
|
|
|
next_checkpoint = log->next_checkpoint;
|
|
|
|
next_cp_seq = log->next_cp_seq;
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
spin_unlock_irq(&log->io_list_lock);
|
|
|
|
|
2015-10-05 09:31:06 +02:00
|
|
|
BUG_ON(reclaimable < 0);
|
|
|
|
if (reclaimable == 0)
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
return;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* write_super will flush cache of each raid disk. We must write super
|
|
|
|
* here, because the log area might be reused soon and we don't want to
|
|
|
|
* confuse recovery
|
|
|
|
*/
|
2015-10-05 09:31:06 +02:00
|
|
|
r5l_write_super(log, next_checkpoint);
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
|
|
|
|
mutex_lock(&log->io_mutex);
|
2015-10-05 09:31:06 +02:00
|
|
|
log->last_checkpoint = next_checkpoint;
|
|
|
|
log->last_cp_seq = next_cp_seq;
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
mutex_unlock(&log->io_mutex);
|
|
|
|
|
2015-10-05 09:31:06 +02:00
|
|
|
r5l_run_no_space_stripes(log);
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static void r5l_reclaim_thread(struct md_thread *thread)
|
|
|
|
{
|
|
|
|
struct mddev *mddev = thread->mddev;
|
|
|
|
struct r5conf *conf = mddev->private;
|
|
|
|
struct r5l_log *log = conf->log;
|
|
|
|
|
|
|
|
if (!log)
|
|
|
|
return;
|
|
|
|
r5l_do_reclaim(log);
|
|
|
|
}
|
|
|
|
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
static void r5l_wake_reclaim(struct r5l_log *log, sector_t space)
|
|
|
|
{
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
unsigned long target;
|
|
|
|
unsigned long new = (unsigned long)space; /* overflow in theory */
|
|
|
|
|
|
|
|
do {
|
|
|
|
target = log->reclaim_target;
|
|
|
|
if (new < target)
|
|
|
|
return;
|
|
|
|
} while (cmpxchg(&log->reclaim_target, target, new) != target);
|
|
|
|
md_wakeup_thread(log->reclaim_thread);
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
}
|
|
|
|
|
2015-10-04 18:20:12 +02:00
|
|
|
void r5l_quiesce(struct r5l_log *log, int state)
|
|
|
|
{
|
|
|
|
if (!log || state == 2)
|
|
|
|
return;
|
|
|
|
if (state == 0) {
|
|
|
|
log->reclaim_thread = md_register_thread(r5l_reclaim_thread,
|
|
|
|
log->rdev->mddev, "reclaim");
|
|
|
|
} else if (state == 1) {
|
|
|
|
/*
|
|
|
|
* at this point all stripes are finished, so io_unit is at
|
|
|
|
* least in STRIPE_END state
|
|
|
|
*/
|
|
|
|
r5l_wake_reclaim(log, -1L);
|
|
|
|
md_unregister_thread(&log->reclaim_thread);
|
|
|
|
r5l_do_reclaim(log);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-08-13 23:32:01 +02:00
|
|
|
struct r5l_recovery_ctx {
|
|
|
|
struct page *meta_page; /* current meta */
|
|
|
|
sector_t meta_total_blocks; /* total size of current meta and data */
|
|
|
|
sector_t pos; /* recovery position */
|
|
|
|
u64 seq; /* recovery position seq */
|
|
|
|
};
|
|
|
|
|
|
|
|
static int r5l_read_meta_block(struct r5l_log *log,
|
|
|
|
struct r5l_recovery_ctx *ctx)
|
|
|
|
{
|
|
|
|
struct page *page = ctx->meta_page;
|
|
|
|
struct r5l_meta_block *mb;
|
|
|
|
u32 crc, stored_crc;
|
|
|
|
|
|
|
|
if (!sync_page_io(log->rdev, ctx->pos, PAGE_SIZE, page, READ, false))
|
|
|
|
return -EIO;
|
|
|
|
|
|
|
|
mb = page_address(page);
|
|
|
|
stored_crc = le32_to_cpu(mb->checksum);
|
|
|
|
mb->checksum = 0;
|
|
|
|
|
|
|
|
if (le32_to_cpu(mb->magic) != R5LOG_MAGIC ||
|
|
|
|
le64_to_cpu(mb->seq) != ctx->seq ||
|
|
|
|
mb->version != R5LOG_VERSION ||
|
|
|
|
le64_to_cpu(mb->position) != ctx->pos)
|
|
|
|
return -EINVAL;
|
|
|
|
|
2015-10-28 16:41:25 +01:00
|
|
|
crc = crc32c_le(log->uuid_checksum, mb, PAGE_SIZE);
|
2015-08-13 23:32:01 +02:00
|
|
|
if (stored_crc != crc)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
if (le32_to_cpu(mb->meta_size) > PAGE_SIZE)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
ctx->meta_total_blocks = BLOCK_SECTORS;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int r5l_recovery_flush_one_stripe(struct r5l_log *log,
|
|
|
|
struct r5l_recovery_ctx *ctx,
|
|
|
|
sector_t stripe_sect,
|
|
|
|
int *offset, sector_t *log_offset)
|
|
|
|
{
|
|
|
|
struct r5conf *conf = log->rdev->mddev->private;
|
|
|
|
struct stripe_head *sh;
|
|
|
|
struct r5l_payload_data_parity *payload;
|
|
|
|
int disk_index;
|
|
|
|
|
|
|
|
sh = raid5_get_active_stripe(conf, stripe_sect, 0, 0, 0);
|
|
|
|
while (1) {
|
|
|
|
payload = page_address(ctx->meta_page) + *offset;
|
|
|
|
|
|
|
|
if (le16_to_cpu(payload->header.type) == R5LOG_PAYLOAD_DATA) {
|
|
|
|
raid5_compute_sector(conf,
|
|
|
|
le64_to_cpu(payload->location), 0,
|
|
|
|
&disk_index, sh);
|
|
|
|
|
|
|
|
sync_page_io(log->rdev, *log_offset, PAGE_SIZE,
|
|
|
|
sh->dev[disk_index].page, READ, false);
|
|
|
|
sh->dev[disk_index].log_checksum =
|
|
|
|
le32_to_cpu(payload->checksum[0]);
|
|
|
|
set_bit(R5_Wantwrite, &sh->dev[disk_index].flags);
|
|
|
|
ctx->meta_total_blocks += BLOCK_SECTORS;
|
|
|
|
} else {
|
|
|
|
disk_index = sh->pd_idx;
|
|
|
|
sync_page_io(log->rdev, *log_offset, PAGE_SIZE,
|
|
|
|
sh->dev[disk_index].page, READ, false);
|
|
|
|
sh->dev[disk_index].log_checksum =
|
|
|
|
le32_to_cpu(payload->checksum[0]);
|
|
|
|
set_bit(R5_Wantwrite, &sh->dev[disk_index].flags);
|
|
|
|
|
|
|
|
if (sh->qd_idx >= 0) {
|
|
|
|
disk_index = sh->qd_idx;
|
|
|
|
sync_page_io(log->rdev,
|
|
|
|
r5l_ring_add(log, *log_offset, BLOCK_SECTORS),
|
|
|
|
PAGE_SIZE, sh->dev[disk_index].page,
|
|
|
|
READ, false);
|
|
|
|
sh->dev[disk_index].log_checksum =
|
|
|
|
le32_to_cpu(payload->checksum[1]);
|
|
|
|
set_bit(R5_Wantwrite,
|
|
|
|
&sh->dev[disk_index].flags);
|
|
|
|
}
|
|
|
|
ctx->meta_total_blocks += BLOCK_SECTORS * conf->max_degraded;
|
|
|
|
}
|
|
|
|
|
|
|
|
*log_offset = r5l_ring_add(log, *log_offset,
|
|
|
|
le32_to_cpu(payload->size));
|
|
|
|
*offset += sizeof(struct r5l_payload_data_parity) +
|
|
|
|
sizeof(__le32) *
|
|
|
|
(le32_to_cpu(payload->size) >> (PAGE_SHIFT - 9));
|
|
|
|
if (le16_to_cpu(payload->header.type) == R5LOG_PAYLOAD_PARITY)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
for (disk_index = 0; disk_index < sh->disks; disk_index++) {
|
|
|
|
void *addr;
|
|
|
|
u32 checksum;
|
|
|
|
|
|
|
|
if (!test_bit(R5_Wantwrite, &sh->dev[disk_index].flags))
|
|
|
|
continue;
|
|
|
|
addr = kmap_atomic(sh->dev[disk_index].page);
|
2015-10-28 16:41:25 +01:00
|
|
|
checksum = crc32c_le(log->uuid_checksum, addr, PAGE_SIZE);
|
2015-08-13 23:32:01 +02:00
|
|
|
kunmap_atomic(addr);
|
|
|
|
if (checksum != sh->dev[disk_index].log_checksum)
|
|
|
|
goto error;
|
|
|
|
}
|
|
|
|
|
|
|
|
for (disk_index = 0; disk_index < sh->disks; disk_index++) {
|
|
|
|
struct md_rdev *rdev, *rrdev;
|
|
|
|
|
|
|
|
if (!test_and_clear_bit(R5_Wantwrite,
|
|
|
|
&sh->dev[disk_index].flags))
|
|
|
|
continue;
|
|
|
|
|
|
|
|
/* in case device is broken */
|
|
|
|
rdev = rcu_dereference(conf->disks[disk_index].rdev);
|
|
|
|
if (rdev)
|
|
|
|
sync_page_io(rdev, stripe_sect, PAGE_SIZE,
|
|
|
|
sh->dev[disk_index].page, WRITE, false);
|
|
|
|
rrdev = rcu_dereference(conf->disks[disk_index].replacement);
|
|
|
|
if (rrdev)
|
|
|
|
sync_page_io(rrdev, stripe_sect, PAGE_SIZE,
|
|
|
|
sh->dev[disk_index].page, WRITE, false);
|
|
|
|
}
|
|
|
|
raid5_release_stripe(sh);
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
error:
|
|
|
|
for (disk_index = 0; disk_index < sh->disks; disk_index++)
|
|
|
|
sh->dev[disk_index].flags = 0;
|
|
|
|
raid5_release_stripe(sh);
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int r5l_recovery_flush_one_meta(struct r5l_log *log,
|
|
|
|
struct r5l_recovery_ctx *ctx)
|
|
|
|
{
|
|
|
|
struct r5conf *conf = log->rdev->mddev->private;
|
|
|
|
struct r5l_payload_data_parity *payload;
|
|
|
|
struct r5l_meta_block *mb;
|
|
|
|
int offset;
|
|
|
|
sector_t log_offset;
|
|
|
|
sector_t stripe_sector;
|
|
|
|
|
|
|
|
mb = page_address(ctx->meta_page);
|
|
|
|
offset = sizeof(struct r5l_meta_block);
|
|
|
|
log_offset = r5l_ring_add(log, ctx->pos, BLOCK_SECTORS);
|
|
|
|
|
|
|
|
while (offset < le32_to_cpu(mb->meta_size)) {
|
|
|
|
int dd;
|
|
|
|
|
|
|
|
payload = (void *)mb + offset;
|
|
|
|
stripe_sector = raid5_compute_sector(conf,
|
|
|
|
le64_to_cpu(payload->location), 0, &dd, NULL);
|
|
|
|
if (r5l_recovery_flush_one_stripe(log, ctx, stripe_sector,
|
|
|
|
&offset, &log_offset))
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* copy data/parity from log to raid disks */
|
|
|
|
static void r5l_recovery_flush_log(struct r5l_log *log,
|
|
|
|
struct r5l_recovery_ctx *ctx)
|
|
|
|
{
|
|
|
|
while (1) {
|
|
|
|
if (r5l_read_meta_block(log, ctx))
|
|
|
|
return;
|
|
|
|
if (r5l_recovery_flush_one_meta(log, ctx))
|
|
|
|
return;
|
|
|
|
ctx->seq++;
|
|
|
|
ctx->pos = r5l_ring_add(log, ctx->pos, ctx->meta_total_blocks);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static int r5l_log_write_empty_meta_block(struct r5l_log *log, sector_t pos,
|
|
|
|
u64 seq)
|
|
|
|
{
|
|
|
|
struct page *page;
|
|
|
|
struct r5l_meta_block *mb;
|
|
|
|
u32 crc;
|
|
|
|
|
|
|
|
page = alloc_page(GFP_KERNEL | __GFP_ZERO);
|
|
|
|
if (!page)
|
|
|
|
return -ENOMEM;
|
|
|
|
mb = page_address(page);
|
|
|
|
mb->magic = cpu_to_le32(R5LOG_MAGIC);
|
|
|
|
mb->version = R5LOG_VERSION;
|
|
|
|
mb->meta_size = cpu_to_le32(sizeof(struct r5l_meta_block));
|
|
|
|
mb->seq = cpu_to_le64(seq);
|
|
|
|
mb->position = cpu_to_le64(pos);
|
2015-10-28 16:41:25 +01:00
|
|
|
crc = crc32c_le(log->uuid_checksum, mb, PAGE_SIZE);
|
2015-08-13 23:32:01 +02:00
|
|
|
mb->checksum = cpu_to_le32(crc);
|
|
|
|
|
|
|
|
if (!sync_page_io(log->rdev, pos, PAGE_SIZE, page, WRITE_FUA, false)) {
|
|
|
|
__free_page(page);
|
|
|
|
return -EIO;
|
|
|
|
}
|
|
|
|
__free_page(page);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
static int r5l_recovery_log(struct r5l_log *log)
|
|
|
|
{
|
2015-08-13 23:32:01 +02:00
|
|
|
struct r5l_recovery_ctx ctx;
|
|
|
|
|
|
|
|
ctx.pos = log->last_checkpoint;
|
|
|
|
ctx.seq = log->last_cp_seq;
|
|
|
|
ctx.meta_page = alloc_page(GFP_KERNEL);
|
|
|
|
if (!ctx.meta_page)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
r5l_recovery_flush_log(log, &ctx);
|
|
|
|
__free_page(ctx.meta_page);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* we did a recovery. Now ctx.pos points to an invalid meta block. New
|
|
|
|
* log will start here. but we can't let superblock point to last valid
|
|
|
|
* meta block. The log might looks like:
|
|
|
|
* | meta 1| meta 2| meta 3|
|
|
|
|
* meta 1 is valid, meta 2 is invalid. meta 3 could be valid. If
|
|
|
|
* superblock points to meta 1, we write a new valid meta 2n. if crash
|
|
|
|
* happens again, new recovery will start from meta 1. Since meta 2n is
|
|
|
|
* valid now, recovery will think meta 3 is valid, which is wrong.
|
|
|
|
* The solution is we create a new meta in meta2 with its seq == meta
|
|
|
|
* 1's seq + 10 and let superblock points to meta2. The same recovery will
|
|
|
|
* not think meta 3 is a valid meta, because its seq doesn't match
|
|
|
|
*/
|
|
|
|
if (ctx.seq > log->last_cp_seq + 1) {
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
ret = r5l_log_write_empty_meta_block(log, ctx.pos, ctx.seq + 10);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
log->seq = ctx.seq + 11;
|
|
|
|
log->log_start = r5l_ring_add(log, ctx.pos, BLOCK_SECTORS);
|
|
|
|
r5l_write_super(log, ctx.pos);
|
|
|
|
} else {
|
|
|
|
log->log_start = ctx.pos;
|
|
|
|
log->seq = ctx.seq;
|
|
|
|
}
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void r5l_write_super(struct r5l_log *log, sector_t cp)
|
|
|
|
{
|
|
|
|
struct mddev *mddev = log->rdev->mddev;
|
|
|
|
|
|
|
|
log->rdev->journal_tail = cp;
|
|
|
|
set_bit(MD_CHANGE_DEVS, &mddev->flags);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int r5l_load_log(struct r5l_log *log)
|
|
|
|
{
|
|
|
|
struct md_rdev *rdev = log->rdev;
|
|
|
|
struct page *page;
|
|
|
|
struct r5l_meta_block *mb;
|
|
|
|
sector_t cp = log->rdev->journal_tail;
|
|
|
|
u32 stored_crc, expected_crc;
|
|
|
|
bool create_super = false;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
/* Make sure it's valid */
|
|
|
|
if (cp >= rdev->sectors || round_down(cp, BLOCK_SECTORS) != cp)
|
|
|
|
cp = 0;
|
|
|
|
page = alloc_page(GFP_KERNEL);
|
|
|
|
if (!page)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
if (!sync_page_io(rdev, cp, PAGE_SIZE, page, READ, false)) {
|
|
|
|
ret = -EIO;
|
|
|
|
goto ioerr;
|
|
|
|
}
|
|
|
|
mb = page_address(page);
|
|
|
|
|
|
|
|
if (le32_to_cpu(mb->magic) != R5LOG_MAGIC ||
|
|
|
|
mb->version != R5LOG_VERSION) {
|
|
|
|
create_super = true;
|
|
|
|
goto create;
|
|
|
|
}
|
|
|
|
stored_crc = le32_to_cpu(mb->checksum);
|
|
|
|
mb->checksum = 0;
|
2015-10-28 16:41:25 +01:00
|
|
|
expected_crc = crc32c_le(log->uuid_checksum, mb, PAGE_SIZE);
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
if (stored_crc != expected_crc) {
|
|
|
|
create_super = true;
|
|
|
|
goto create;
|
|
|
|
}
|
|
|
|
if (le64_to_cpu(mb->position) != cp) {
|
|
|
|
create_super = true;
|
|
|
|
goto create;
|
|
|
|
}
|
|
|
|
create:
|
|
|
|
if (create_super) {
|
|
|
|
log->last_cp_seq = prandom_u32();
|
|
|
|
cp = 0;
|
|
|
|
/*
|
|
|
|
* Make sure super points to correct address. Log might have
|
|
|
|
* data very soon. If super hasn't correct log tail address,
|
|
|
|
* recovery can't find the log
|
|
|
|
*/
|
|
|
|
r5l_write_super(log, cp);
|
|
|
|
} else
|
|
|
|
log->last_cp_seq = le64_to_cpu(mb->seq);
|
|
|
|
|
|
|
|
log->device_size = round_down(rdev->sectors, BLOCK_SECTORS);
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
log->max_free_space = log->device_size >> RECLAIM_MAX_FREE_SPACE_SHIFT;
|
|
|
|
if (log->max_free_space > RECLAIM_MAX_FREE_SPACE)
|
|
|
|
log->max_free_space = RECLAIM_MAX_FREE_SPACE;
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
log->last_checkpoint = cp;
|
|
|
|
|
|
|
|
__free_page(page);
|
|
|
|
|
|
|
|
return r5l_recovery_log(log);
|
|
|
|
ioerr:
|
|
|
|
__free_page(page);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
int r5l_init_log(struct r5conf *conf, struct md_rdev *rdev)
|
|
|
|
{
|
|
|
|
struct r5l_log *log;
|
|
|
|
|
|
|
|
if (PAGE_SIZE != 4096)
|
|
|
|
return -EINVAL;
|
|
|
|
log = kzalloc(sizeof(*log), GFP_KERNEL);
|
|
|
|
if (!log)
|
|
|
|
return -ENOMEM;
|
|
|
|
log->rdev = rdev;
|
|
|
|
|
2015-10-05 09:31:09 +02:00
|
|
|
log->need_cache_flush = (rdev->bdev->bd_disk->queue->flush_flags != 0);
|
|
|
|
|
2015-10-28 16:41:25 +01:00
|
|
|
log->uuid_checksum = crc32c_le(~0, rdev->mddev->uuid,
|
|
|
|
sizeof(rdev->mddev->uuid));
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
|
|
|
|
mutex_init(&log->io_mutex);
|
|
|
|
|
|
|
|
spin_lock_init(&log->io_list_lock);
|
|
|
|
INIT_LIST_HEAD(&log->running_ios);
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
INIT_LIST_HEAD(&log->io_end_ios);
|
2015-09-02 22:49:46 +02:00
|
|
|
INIT_LIST_HEAD(&log->flushing_ios);
|
2015-10-05 09:31:07 +02:00
|
|
|
INIT_LIST_HEAD(&log->finished_ios);
|
2015-09-02 22:49:46 +02:00
|
|
|
bio_init(&log->flush_bio);
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
|
|
|
|
log->io_kc = KMEM_CACHE(r5l_io_unit, 0);
|
|
|
|
if (!log->io_kc)
|
|
|
|
goto io_kc;
|
|
|
|
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
log->reclaim_thread = md_register_thread(r5l_reclaim_thread,
|
|
|
|
log->rdev->mddev, "reclaim");
|
|
|
|
if (!log->reclaim_thread)
|
|
|
|
goto reclaim_thread;
|
2015-09-02 22:49:47 +02:00
|
|
|
init_waitqueue_head(&log->iounit_wait);
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
INIT_LIST_HEAD(&log->no_space_stripes);
|
|
|
|
spin_lock_init(&log->no_space_stripes_lock);
|
|
|
|
|
|
|
|
if (r5l_load_log(log))
|
|
|
|
goto error;
|
|
|
|
|
|
|
|
conf->log = log;
|
|
|
|
return 0;
|
|
|
|
error:
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
md_unregister_thread(&log->reclaim_thread);
|
|
|
|
reclaim_thread:
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
kmem_cache_destroy(log->io_kc);
|
|
|
|
io_kc:
|
|
|
|
kfree(log);
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
void r5l_exit_log(struct r5l_log *log)
|
|
|
|
{
|
raid5: log reclaim support
This is the reclaim support for raid5 log. A stripe write will have
following steps:
1. reconstruct the stripe, read data/calculate parity. ops_run_io
prepares to write data/parity to raid disks
2. hijack ops_run_io. stripe data/parity is appending to log disk
3. flush log disk cache
4. ops_run_io run again and do normal operation. stripe data/parity is
written in raid array disks. raid core can return io to upper layer.
5. flush cache of all raid array disks
6. update super block
7. log disk space used by the stripe can be reused
In practice, several stripes consist of an io_unit and we will batch
several io_unit in different steps, but the whole process doesn't
change.
It's possible io return just after data/parity hit log disk, but then
read IO will need read from log disk. For simplicity, IO return happens
at step 4, where read IO can directly read from raid disks.
Currently reclaim run if there is specific reclaimable space (1/4 disk
size or 10G) or we are out of space. Reclaim is just to free log disk
spaces, it doesn't impact data consistency. The size based force reclaim
is to make sure log isn't too big, so recovery doesn't scan log too
much.
Recovery make sure raid disks and log disk have the same data of a
stripe. If crash happens before 4, recovery might/might not recovery
stripe's data/parity depending on if data/parity and its checksum
matches. In either case, this doesn't change the syntax of an IO write.
After step 3, stripe is guaranteed recoverable, because stripe's
data/parity is persistent in log disk. In some cases, log disk content
and raid disks content of a stripe are the same, but recovery will still
copy log disk content to raid disks, this doesn't impact data
consistency. space reuse happens after superblock update and cache
flush.
There is one situation we want to avoid. A broken meta in the middle of
a log causes recovery can't find meta at the head of log. If operations
require meta at the head persistent in log, we must make sure meta
before it persistent in log too. The case is stripe data/parity is in
log and we start write stripe to raid disks (before step 4). stripe
data/parity must be persistent in log before we do the write to raid
disks. The solution is we restrictly maintain io_unit list order. In
this case, we only write stripes of an io_unit to raid disks till the
io_unit is the first one whose data/parity is in log.
The io_unit list order is important for other cases too. For example,
some io_unit are reclaimable and others not. They can be mixed in the
list, we shouldn't reuse space of an unreclaimable io_unit.
Includes fixes to problems which were...
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:32:00 +02:00
|
|
|
md_unregister_thread(&log->reclaim_thread);
|
raid5: add basic stripe log
This introduces a simple log for raid5. Data/parity writing to raid
array first writes to the log, then write to raid array disks. If
crash happens, we can recovery data from the log. This can speed up
raid resync and fix write hole issue.
The log structure is pretty simple. Data/meta data is stored in block
unit, which is 4k generally. It has only one type of meta data block.
The meta data block can track 3 types of data, stripe data, stripe
parity and flush block. MD superblock will point to the last valid
meta data block. Each meta data block has checksum/seq number, so
recovery can scan the log correctly. We store a checksum of stripe
data/parity to the metadata block, so meta data and stripe data/parity
can be written to log disk together. otherwise, meta data write must
wait till stripe data/parity is finished.
For stripe data, meta data block will record stripe data sector and
size. Currently the size is always 4k. This meta data record can be made
simpler if we just fix write hole (eg, we can record data of a stripe's
different disks together), but this format can be extended to support
caching in the future, which must record data address/size.
For stripe parity, meta data block will record stripe sector. It's
size should be 4k (for raid5) or 8k (for raid6). We always store p
parity first. This format should work for caching too.
flush block indicates a stripe is in raid array disks. Fixing write
hole doesn't need this type of meta data, it's for caching extension.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: NeilBrown <neilb@suse.com>
2015-08-13 23:31:59 +02:00
|
|
|
kmem_cache_destroy(log->io_kc);
|
|
|
|
kfree(log);
|
|
|
|
}
|