Improve write performance by preventing the delayed_list from dumping all its
stripes onto the handle_list in one shot. Delayed stripes are now further
delayed by being held on the 'hold_list'. The 'hold_list' is bypassed when:
* a STRIPE_IO_STARTED stripe is found at the head of 'handle_list'
* 'handle_list' is empty and i/o is being done to satisfy full stripe-width
write requests
* 'bypass_count' is less than 'bypass_threshold'. By default the threshold
is 1, i.e. every other stripe handled is a preread stripe provided the
top two conditions are false.
Benchmark data:
System: 2x Xeon 5150, 4x SATA, mem=1GB
Baseline: 2.6.24-rc7
Configuration: mdadm --create /dev/md0 /dev/sd[b-e] -n 4 -l 5 --assume-clean
Test1: dd if=/dev/zero of=/dev/md0 bs=1024k count=2048
* patched: +33% (stripe_cache_size = 256), +25% (stripe_cache_size = 512)
Test2: tiobench --size 2048 --numruns 5 --block 4096 --block 131072 (XFS)
* patched: +13%
* patched + preread_bypass_threshold = 0: +37%
Changes since v1:
* reduce bypass_threshold from (chunk_size / sectors_per_chunk) to (1) and
make it configurable. This defaults to fairness and modest performance
gains out of the box.
Changes since v2:
* [neilb@suse.de]: kill STRIPE_PRIO_HI and preread_needed as they are not
necessary, the important change was clearing STRIPE_DELAYED in
add_stripe_bio and this has been moved out to make_request for the hang
fix.
* [neilb@suse.de]: simplify get_priority_stripe
* [dan.j.williams@intel.com]: reset the bypass_count when ->hold_list is
sampled empty (+11%)
* [dan.j.williams@intel.com]: decrement the bypass_count at the detection
of stripes being naturally promoted off of hold_list +2%. Note, resetting
bypass_count instead of decrementing on these events yields +4% but that is
probably too aggressive.
Changes since v3:
* cosmetic fixups
Tested-by: James W. Laferriere <babydr@baby-dragons.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When a read bio is attached to the stripe and the corresponding block is
marked R5_UPTODATE, then a read (biofill) operation is scheduled to copy
the data from the stripe cache to the bio buffer. handle_stripe flags the
blocks to be operated on with the R5_Wantfill flag. If new read requests
arrive while raid5_run_ops is running they will not be handled until
handle_stripe is scheduled to run again.
Changelog:
* cleanup to_read and to_fill accounting
* do not fail reads that have reached the cache
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Acked-By: NeilBrown <neilb@suse.de>
handle_stripe will compute a block when a backing disk has failed, or when
it determines it can save a disk read by computing the block from all the
other up-to-date blocks.
Previously a block would be computed under the lock and subsequent logic in
handle_stripe could use the newly up-to-date block. With the raid5_run_ops
implementation the compute operation is carried out a later time outside
the lock. To preserve the old functionality we take advantage of the
dependency chain feature of async_tx to flag the block as R5_Wantcompute
and then let other parts of handle_stripe operate on the block as if it
were up-to-date. raid5_run_ops guarantees that the block will be ready
before it is used in another operation.
However, this only works in cases where the compute and the dependent
operation are scheduled at the same time. If a previous call to
handle_stripe sets the R5_Wantcompute flag there is no facility to pass the
async_tx dependency chain across successive calls to raid5_run_ops. The
req_compute variable protects against this case.
Changelog:
* remove the req_compute BUG_ON
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Acked-By: NeilBrown <neilb@suse.de>
When the raid acceleration work was proposed, Neil laid out the following
attack plan:
1/ move the xor and copy operations outside spin_lock(&sh->lock)
2/ find/implement an asynchronous offload api
The raid5_run_ops routine uses the asynchronous offload api (async_tx) and
the stripe_operations member of a stripe_head to carry out xor+copy
operations asynchronously, outside the lock.
To perform operations outside the lock a new set of state flags is needed
to track new requests, in-flight requests, and completed requests. In this
new model handle_stripe is tasked with scanning the stripe_head for work,
updating the stripe_operations structure, and finally dropping the lock and
calling raid5_run_ops for processing. The following flags outline the
requests that handle_stripe can make of raid5_run_ops:
STRIPE_OP_BIOFILL
- copy data into request buffers to satisfy a read request
STRIPE_OP_COMPUTE_BLK
- generate a missing block in the cache from the other blocks
STRIPE_OP_PREXOR
- subtract existing data as part of the read-modify-write process
STRIPE_OP_BIODRAIN
- copy data out of request buffers to satisfy a write request
STRIPE_OP_POSTXOR
- recalculate parity for new data that has entered the cache
STRIPE_OP_CHECK
- verify that the parity is correct
STRIPE_OP_IO
- submit i/o to the member disks (note this was already performed outside
the stripe lock, but it made sense to add it as an operation type
The flow is:
1/ handle_stripe sets STRIPE_OP_* in sh->ops.pending
2/ raid5_run_ops reads sh->ops.pending, sets sh->ops.ack, and submits the
operation to the async_tx api
3/ async_tx triggers the completion callback routine to set
sh->ops.complete and release the stripe
4/ handle_stripe runs again to finish the operation and optionally submit
new operations that were previously blocked
Note this patch just defines raid5_run_ops, subsequent commits (one per
major operation type) modify handle_stripe to take advantage of this
routine.
Changelog:
* removed ops_complete_biodrain in favor of ops_complete_postxor and
ops_complete_write.
* removed the raid5_run_ops workqueue
* call bi_end_io for reads in ops_complete_biofill, saves a call to
handle_stripe
* explicitly handle the 2-disk raid5 case (xor becomes memcpy), Neil Brown
* fix race between async engines and bi_end_io call for reads, Neil Brown
* remove unnecessary spin_lock from ops_complete_biofill
* remove test_and_set/test_and_clear BUG_ONs, Neil Brown
* remove explicit interrupt handling for channel switching, this feature
was absorbed (i.e. it is now implicit) by the async_tx api
* use return_io in ops_complete_biofill
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Acked-By: NeilBrown <neilb@suse.de>
handle_stripe5 and handle_stripe6 have very deep logic paths handling the
various states of a stripe_head. By introducing the 'stripe_head_state'
and 'r6_state' objects, large portions of the logic can be moved to
sub-routines.
'struct stripe_head_state' consumes all of the automatic variables that previously
stood alone in handle_stripe5,6. 'struct r6_state' contains the handle_stripe6
specific variables like p_failed and q_failed.
One of the nice side effects of the 'stripe_head_state' change is that it
allows for further reductions in code duplication between raid5 and raid6.
The following new routines are shared between raid5 and raid6:
handle_completed_write_requests
handle_requests_to_failed_array
handle_stripe_expansion
Changes:
* v2: fixed 'conf->raid_disk-1' for the raid6 'handle_stripe_expansion' path
* v3: removed the unused 'dirty' field from struct stripe_head_state
* v3: coalesced open coded bi_end_io routines into return_io()
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Acked-By: NeilBrown <neilb@suse.de>
If a bypass-the-cache read fails, we simply try again through the cache. If
it fails again it will trigger normal recovery precedures.
update 1:
From: NeilBrown <neilb@suse.de>
1/
chunk_aligned_read and retry_aligned_read assume that
data_disks == raid_disks - 1
which is not true for raid6.
So when an aligned read request bypasses the cache, we can get the wrong data.
2/ The cloned bio is being used-after-free in raid5_align_endio
(to test BIO_UPTODATE).
3/ We forgot to add rdev->data_offset when submitting
a bio for aligned-read
4/ clone_bio calls blk_recount_segments and then we change bi_bdev,
so we need to invalidate the segment counts.
5/ We don't de-reference the rdev when the read completes.
This means we need to record the rdev to so it is still
available in the end_io routine. Fortunately
bi_next in the original bio is unused at this point so
we can stuff it in there.
6/ We leak a cloned bio if the target rdev is not usable.
From: NeilBrown <neilb@suse.de>
update 2:
1/ When aligned requests fail (read error) they need to be retried
via the normal method (stripe cache). As we cannot be sure that
we can process a single read in one go (we may not be able to
allocate all the stripes needed) we store a bio-being-retried
and a list of bioes-that-still-need-to-be-retried.
When find a bio that needs to be retried, we should add it to
the list, not to single-bio...
2/ We were never incrementing 'scnt' when resubmitting failed
aligned requests.
[akpm@osdl.org: build fix]
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Replace all uses of kmem_cache_t with struct kmem_cache.
The patch was generated using the following script:
#!/bin/sh
#
# Replace one string by another in all the kernel sources.
#
set -e
for file in `find * -name "*.c" -o -name "*.h"|xargs grep -l $1`; do
quilt add $file
sed -e "1,\$s/$1/$2/g" $file >/tmp/$$
mv /tmp/$$ $file
quilt refresh
done
The script was run like this
sh replace kmem_cache_t "struct kmem_cache"
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
They are not needed. conf->failed_disks is the same as mddev->degraded and
conf->working_disks is conf->raid_disks - mddev->degraded.
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
There is a lot of commonality between raid5.c and raid6main.c. This patches
merges both into one module called raid456. This saves a lot of code, and
paves the way for online raid5->raid6 migrations.
There is still duplication, e.g. between handle_stripe5 and handle_stripe6.
This will probably be cleaned up later.
Cc: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Instead of checkpointing at each stripe, only checkpoint when a new write
would overwrite uncheckpointed data. Block any write to the uncheckpointed
area. Arbitrarily checkpoint at least every 3Meg.
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
We allow the superblock to record an 'old' and a 'new' geometry, and a
position where any conversion is up to. The geometry allows for changing
chunksize, layout and level as well as number of devices.
When using verion-0.90 superblock, we convert the version to 0.91 while the
conversion is happening so that an old kernel will refuse the assemble the
array. For version-1, we use a feature bit for the same effect.
When starting an array we check for an incomplete reshape and restart the
reshape process if needed. If the reshape stopped at an awkward time (like
when updating the first stripe) we refuse to assemble the array, and let
user-space worry about it.
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This patch provides the core of the resize/expand process.
sync_request notices if a 'reshape' is happening and acts accordingly.
It allocated new stripe_heads for the next chunk-wide-stripe in the target
geometry, marking them STRIPE_EXPANDING.
Then it finds which stripe heads in the old geometry can provide data needed
by these and marks them STRIPE_EXPAND_SOURCE. This causes stripe_handle to
read all blocks on those stripes.
Once all blocks on a STRIPE_EXPAND_SOURCE stripe_head are read, any that are
needed are copied into the corresponding STRIPE_EXPANDING stripe_head. Once a
STRIPE_EXPANDING stripe_head is full, it is marks STRIPE_EXPAND_READY and then
is written out and released.
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
We need to allow that different stripes are of different effective sizes, and
use the appropriate size. Also, when a stripe is being expanded, we must
block any IO attempts until the stripe is stable again.
Key elements in this change are:
- each stripe_head gets a 'disk' field which is part of the key,
thus there can sometimes be two stripe heads of the same area of
the array, but covering different numbers of devices. One of these
will be marked STRIPE_EXPANDING and so won't accept new requests.
- conf->expand_progress tracks how the expansion is progressing and
is used to determine whether the target part of the array has been
expanded yet or not.
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Before a RAID-5 can be expanded, we need to be able to expand the stripe-cache
data structure.
This requires allocating new stripes in a new kmem_cache. If this succeeds,
we copy cache pages over and release the old stripes and kmem_cache.
We then allocate new pages. If that fails, we leave the stripe cache at it's
new size. It isn't worth the effort to shrink it back again.
Unfortuanately this means we need two kmem_cache names as we, for a short
period of time, we have two kmem_caches. So they are raid5/%s and
raid5/%s-alt
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The remainder of this batch implements raid5 reshaping. Currently the only
shape change that is supported is added a device, but it is envisioned that
changing the chunksize and layout will also be supported, as well as changing
the level (e.g. 1->5, 5->6).
The reshape process naturally has to move all of the data in the array, and so
should be used with caution. It is believed to work, and some testing does
support this, but wider testing would be great for increasing my confidence.
You will need a version of mdadm newer than 2.3.1 to make use of raid5 growth.
This is because mdadm need to take a copy of a 'critical section' at the
start of the array incase there is a crash at an awkward moment. On restart,
mdadm will restore the critical section and allow reshape to continue.
I hope to release a 2.4-pre by early next week - it still needs a little more
polishing.
This patch:
Previously the array of disk information was included in the raid5 'conf'
structure which was allocated to an appropriate size. This makes it awkward
to change the size of that array. So we split it off into a separate
kmalloced array which will require a little extra indexing, but is much easier
to grow.
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
- replace open-coded hash chain with hlist macros
- Fix hash-table size at one page - it is already quite generous, so there
will never be a need to use multiple pages, so no need for __get_free_pages
No functional change.
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
There is this "FIXME" comment with a typo in it!! that been annoying me for
days, so I just had to remove it.
conf->disks[i].rdev should only be accessed if
- we know we hold a reference or
- the mddev->reconfig_sem is down or
- we have a rcu_readlock
handle_stripe was referencing rdev in three places without any of these. For
the first two, get an rcu_readlock. For the last, the same access
(md_sync_acct call) is made a little later after the rdev has been claimed
under and rcu_readlock, if R5_Syncio is set. So just use that access...
However R5_Syncio isn't really needed as the 'syncing' variable contains the
same information. So use that instead.
Issues, comment, and fix are identical in raid5 and raid6.
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
raid6 currently does not check the P/Q syndromes when doing a resync, it just
calculates the correct value and writes it. Doing the check can reduce writes
(often to 0) for a resync, and it is needed to properly implement the
echo check > sync_action
operation.
This patch implements the appropriate checks and tidies up some related code.
It also allows raid6 user-requested resync to bypass the intent bitmap.
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
There isn't really a need for raid5 attributes to be an a subdirectory,
so this patch moves them from
/sys/block/mdX/md/raid5/attribute
to
/sys/block/mdX/md/attribute
This suggests that all md personalities should co-operate about
namespace usage, but that shouldn't be a problem.
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
/sys/block/mdX/md/raid5/
contains raid5-related attributes.
Currently
stripe_cache_size
is number of entries in stripe cache, and is settable.
stripe_cache_active
is number of active entries, and in only readable.
Signed-off-by: Neil Brown <neilb@suse.de>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This patch changes the behaviour of raid5 when it gets a read error.
Instead of just failing the device, it tried to find out what should have
been there, and writes it over the bad block. For some media-errors, this
has a reasonable chance of fixing the error. If the write succeeds, and a
subsequent read succeeds as well, raid5 decided the address is OK and
conitnues.
Instead of failing a drive on read-error, we attempt to re-write the block,
and then re-read. If that all works, we allow the device to remain in the
array.
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Most awkward part of this is delaying write requests until bitmap updates have
been flushed.
To achieve this, we have a sequence number (seq_flush) which is incremented
each time the raid5 is unplugged.
If the raid thread notices that this has changed, it flushes bitmap changes,
and assigned the value of seq_flush to seq_write.
When a write request arrives, it is given the number from seq_write, and that
write request may not complete until seq_flush is larger than the saved seq
number.
We have a new queue for storing stripes which are waiting for a bitmap flush
and an extra flag for stripes to record if the write was 'degraded' and so
should not clear the a bit in the bitmap.
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.
Let it rip!
Dan Williams
Raz Ben-Jehuda(caro)
Christoph Lameter
NeilBrown
NeilBrown
Linus Torvalds