- 11 Jun, 2015 10 commits
-
-
Joe Thornber authored
Previously REQ_DISCARD bios have been split into block sized chunks before submission to the thin target. There are a couple of issues with this: - If the block size is small, a large discard request can get broken up into a great many bios which is both slow and causes a lot of memory pressure. - The thin pool block size and the discard granularity for the underlying data device need to be compatible if we want to passdown the discard. This patch relaxes the block size granularity for thin devices. It makes use of the recent range locking added to the bio_prison to quiesce a whole range of thin blocks before unmapping them. Once a thin range has been unmapped the discard can then be passed down to the data device for those sub ranges where the data blocks are no longer used (ie. they weren't shared in the first place). This patch also doesn't make any apologies about open-coding portions of block core as a means to supporting async discard completions in the near-term -- if/when late bio splitting lands it'll all get cleaned up. Signed-off-by:
Joe Thornber <ejt@redhat.com> Signed-off-by:
Mike Snitzer <snitzer@redhat.com>
-
Joe Thornber authored
Removes a range of blocks from the btree. Signed-off-by:
-
Joe Thornber authored
Retrieve the next run of contiguously mapped blocks. Useful for working out where to break up IO. Signed-off-by:
-
Joe Thornber authored
Removes a range of leaf values from the tree. Signed-off-by:
-
Pekka Enberg authored
Use kvfree() instead of open-coding it. Signed-off-by:
Pekka Enberg <penberg@kernel.org> Signed-off-by:
-
Joe Thornber authored
The policy tick() method is normally called from interrupt context. Both the mq and smq policies do some bottom half work for the tick method in their map functions. However if no IO is going through the cache, then that bottom half work doesn't occur. With these policies this means recently hit entries do not age and do not get written back as early as we'd like. Fix this by introducing a new 'can_block' parameter to the tick() method. When this is set the bottom half work occurs immediately. 'can_block' is set when the tick method is called every second by the core target (not in interrupt context). Signed-off-by:
-
Mike Snitzer authored
Having the DM device name associated with the ERR or INFO message is very helpful. Signed-off-by: -
Joe Thornber authored
If a cache metadata operation fails (e.g. transaction commit) the cache's metadata device will abort the current transaction, set a new needs_check flag, and the cache will transition to "read-only" mode. If aborting the transaction or setting the needs_check flag fails the cache will transition to "fail-io" mode. Once needs_check is set the cache device will not be allowed to activate. Activation requires write access to metadata. Future work is needed to add proper support for running the cache in read-only mode. Once in fail-io mode the cache will report a status of "Fail". Also, add commit() wrapper that will disallow commits if in read_only or fail mode. Signed-off-by:
-
Joe Thornber authored
When the cache is idle, writeback work was only being issued every second. With this change outstanding writebacks are streamed constantly. This offers a writeback performance improvement. Signed-off-by:
-
Joe Thornber authored
The stochastic-multi-queue (smq) policy addresses some of the problems with the current multiqueue (mq) policy. Memory usage ------------ The mq policy uses a lot of memory; 88 bytes per cache block on a 64 bit machine. SMQ uses 28bit indexes to implement it's data structures rather than pointers. It avoids storing an explicit hit count for each block. It has a 'hotspot' queue rather than a pre cache which uses a quarter of the entries (each hotspot block covers a larger area than a single cache block). All these mean smq uses ~25bytes per cache block. Still a lot of memory, but a substantial improvement nontheless. Level balancing --------------- MQ places entries in different levels of the multiqueue structures based on their hit count (~ln(hit count)). This means the bottom levels generally have the most entries, and the top ones have very few. Having unbalanced levels like this reduces the efficacy of the multiqueue. SMQ does not maintain a hit count, instead it swaps hit entries with the least recently used entry from the level above. The over all ordering being a side effect of this stochastic process. With this scheme we can decide how many entries occupy each multiqueue level, resulting in better promotion/demotion decisions. Adaptability ------------ The MQ policy maintains a hit count for each cache block. For a different block to get promoted to the cache it's hit count has to exceed the lowest currently in the cache. This means it can take a long time for the cache to adapt between varying IO patterns. Periodically degrading the hit counts could help with this, but I haven't found a nice general solution. SMQ doesn't maintain hit counts, so a lot of this problem just goes away. In addition it tracks performance of the hotspot queue, which is used to decide which blocks to promote. If the hotspot queue is performing badly then it starts moving entries more quickly between levels. This lets it adapt to new IO patterns very quickly. Performance ----------- In my tests SMQ shows substantially better performance than MQ. Once this matures a bit more I'm sure it'll become the default policy. Signed-off-by:
-
- 29 May, 2015 25 commits
-
-
Joe Thornber authored
When considering whether to move a block to the cache we already give preferential treatment to discarded blocks, since they are cheap to promote (no read of the origin required since the data is junk). The same is true of blocks that are about to be completely overwritten, so we likewise boost their promotion chances. Signed-off-by:
-
Joe Thornber authored
Currently individual bios are deferred to the worker thread if they cannot be processed immediately (eg, a block is in the process of being moved to the fast device). This patch passes whole cells across to the worker. This saves reaquiring the cell, and also collects bios destined for the same block together, which allows them to be mapped with a single look up to the policy. This reduces the overhead of using dm-cache. Signed-off-by:
-
Joe Thornber authored
Rather than always releasing the prisoners in a cell, the client may want to promote one of them to be the new holder. There is a race here though between releasing an empty cell, and other threads adding new inmates. So this function makes the decision with its lock held. This function can have two outcomes: i) An inmate is promoted to be the holder of the cell (return value of 0). ii) The cell has no inmate for promotion and is released (return value of 1). Signed-off-by:
-
Joe Thornber authored
Signed-off-by:
-
Joe Thornber authored
We only allow non critical writeback if the origin is idle. It is up to the policy to decide what writeback work is critical. Signed-off-by:
-
Joe Thornber authored
Signed-off-by:
-
Joe Thornber authored
A little class that keeps track of the volume of io that is in flight, and the length of time that a device has been idle for. FIXME: rather than jiffes, may be best to use ktime_t (to support faster devices). Signed-off-by:
-
Joe Thornber authored
There is a race between a policy deciding to replace a cache entry, the core target writing back any dirty data from this block, and other IO threads doing IO to the same block. This sort of problem is avoided most of the time by the core target grabbing a bio prison cell before making the request to the policy. But for a demotion the core target doesn't know which block will be demoted, so can't do this in advance. Fix this demotion race by introducing a callback to the policy interface that allows the policy to grab the cell on behalf of the core target. Signed-off-by:
-
Milan Broz authored
A crypto driver can process requests synchronously or asynchronously and can use an internal driver queue to backlog requests. Add some comments to clarify internal logic and completion return codes. Signed-off-by:
Milan Broz <mbroz@redhat.com> Signed-off-by:
-
Lidong Zhong authored
Currently if there is a leg failure, the bio will be put into the hold list until userspace does a remove/replace on the leg. Doing so in a cluster config (clvmd) is problematic because there may be a temporary path failure that results in cluster raid1 remove/replace. Such recovery takes a long time due to a full resync. Update dm-raid1 to optionally ignore these failures so bios continue being issued without interrupton. To enable this feature userspace must pass "keep_log" when creating the dm-raid1 device. Signed-off-by:
Lidong Zhong <lzhong@suse.com> Tested-by:
Liuhua Wang <lwang@suse.com> Acked-by:
Heinz Mauelshagen <heinzm@redhat.com> Signed-off-by:
-
Geert Uytterhoeven authored
On 32-bit: drivers/md/dm-log-writes.c: In function ‘log_super’: drivers/md/dm-log-writes.c:323: warning: integer constant is too large for ‘long’ type Add a ULL suffix to WRITE_LOG_MAGIC to fix this. Also add a ULL suffix to WRITE_LOG_VERSION as it's stored in a __le64 field. Signed-off-by:
Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by:
-
Luis Henriques authored
Signed-off-by:
Luis Henriques <luis.henriques@canonical.com> Signed-off-by:
-
Heinz Mauelshagen authored
Add dm-raid access to the MD RAID0 personality to enable single zone striping. The following changes enable that access: - add type definition to raid_types array - make bitmap creation conditonal in super_validate(), because bitmaps are not allowed in raid0 - set rdev->sectors to the data image size in super_validate() to allow the raid0 personality to calculate the MD array size properly - use mdddev(un)lock() functions instead of direct mutex_(un)lock() (wrapped in here because it's a trivial change) - enhance raid_status() to always report full sync for raid0 so that userspace checks for 100% sync will succeed and allow for resize (and takeover/reshape once added in future paches) - enhance raid_resume() to not load bitmap in case of raid0 - add merge function to avoid data corruption (seen with readahead) that resulted from bio payloads that grew too large. This problem did not occur with the other raid levels because it either did not apply without striping (raid1) or was avoided via stripe caching. - raise version to 1.7.0 because of the raid0 API change Signed-off-by:
Jonathan Brassow <jbrassow@redhat.com> Signed-off-by:
-
Heinz Mauelshagen authored
- ensure maximum device limit in superblock - rename DMPF_* (print flags) to CTR_FLAG_* (constructor flags) and their respective struct raid_set member - use strcasecmp() in raid10_format_to_md_layout() as in the constructor Signed-off-by:
-
Heinz Mauelshagen authored
Remove comment above parse_raid_params() that claims "devices_handle_discard_safely" is a table line argument when it is actually is a module parameter. Also, backfill dm-raid target version 1.6.0 documentation. Signed-off-by:
-
Mike Snitzer authored
Leverage the block manager's read_only flag instead of duplicating it; access with new dm_bm_is_read_only() method. Signed-off-by: -
Mike Snitzer authored
The overwrite has only ever about optimizing away the need to zero a block if the entire block was being overwritten. As such it is only relevant when zeroing is enabled. Signed-off-by:
-
Mike Snitzer authored
Signed-off-by: -
Mike Snitzer authored
Introduce a single common method for cleaning up a DM device's mapped_device. No functional change, just eliminates duplication of delicate mapped_device cleanup code. Signed-off-by: -
Mike Snitzer authored
More often than not a request that is requeued _is_ mapped (meaning the clone request is allocated and clone->q is initialized). Rename dm_requeue_unmapped_original_request() to avoid potential confusion due to function name containing "unmapped". Also, remove dm_requeue_unmapped_request() since callers can easily call the dm_requeue_original_request() directly. Reviewed-by:
Christoph Hellwig <hch@lst.de> Signed-off-by:
-
Mike Snitzer authored
Do not allocate the io_pool mempool for blk-mq request-based DM (DM_TYPE_MQ_REQUEST_BASED) in dm_alloc_rq_mempools(). Also refine __bind_mempools() to have more precise awareness of which mempools each type of DM device uses -- avoids mempool churn when reloading DM tables (particularly for DM_TYPE_REQUEST_BASED). Signed-off-by: -
Mike Snitzer authored
-
Joe Thornber authored
dm_merge_bvec() was originally added in f6fccb ("dm: introduce merge_bvec_fn"). In that commit a value in sectors is converted to bytes using << 9, and then assigned to an int. This code made assumptions about the value of BIO_MAX_SECTORS. A later commit 148e51 ("dm: improve documentation and code clarity in dm_merge_bvec") was meant to have no functional change but it removed the use of BIO_MAX_SECTORS in favor of using queue_max_sectors(). At this point the cast from sector_t to int resulted in a zero value. The fallout being dm_merge_bvec() would only allow a single page to be added to a bio. This interim fix is minimal for the benefit of stable@ because the more comprehensive cleanup of passing a sector_t to all DM targets' merge function will impact quite a few DM targets. Signed-off-by: -
Junichi Nomura authored
dm-multipath accepts 0 path mapping. # echo '0 2097152 multipath 0 0 0 0' | dmsetup create newdev Such a mapping can be used to release underlying devices while still holding requests in its queue until working paths come back. However, once the multipath device is created over blk-mq devices, it rejects reloading of 0 path mapping: # echo '0 2097152 multipath 0 0 1 1 queue-length 0 1 1 /dev/sda 1' \ | dmsetup create mpath1 # echo '0 2097152 multipath 0 0 0 0' | dmsetup load mpath1 device-mapper: reload ioctl on mpath1 failed: Invalid argument Command failed With following kernel message: device-mapper: ioctl: can't change device type after initial table load. DM tries to inherit the current table type using dm_table_set_type() but it doesn't work as expected because of unnecessary check about whether the target type is hybrid or not. Hybrid type is for targets that work as either request-based or bio-based and not required for blk-mq or non blk-mq checking. Fixes: 65803c20 ("dm table: train hybrid target type detection to select blk-mq if appropriate") Signed-off-by:Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by:
-
Mike Snitzer authored
When stacking request-based dm device on non blk-mq device and device-mapper target could not map the request (error target is used, multipath target with all paths down, etc), the WARN_ON_ONCE() in free_rq_clone() will trigger when it shouldn't. The warning was added by commit aa6df8dd ("dm: fix free_rq_clone() NULL pointer when requeueing unmapped request"). But free_rq_clone() with clone->q == NULL is valid usage for the case where dm_kill_unmapped_request() initiates request cleanup. Fix this false warning by just removing the WARN_ON -- it only generated false positives and was never useful in catching the intended case (completing clone request not being mapped e.g. clone->q being NULL). Fixes: aa6df8dd ("dm: fix free_rq_clone() NULL pointer when requeueing unmapped request") Reported-by:
Bart Van Assche <bart.vanassche@sandisk.com> Reported-by:
-
- 27 May, 2015 3 commits
-
-
Mike Snitzer authored
Use BLK_MQ_RQ_QUEUE_BUSY to requeue a blk-mq request directly from the DM blk-mq device's .queue_rq. This cleans up the previous convoluted handling of request requeueing that would return BLK_MQ_RQ_QUEUE_OK (even though it wasn't) and then run blk_mq_requeue_request() followed by blk_mq_kick_requeue_list(). Also, document that DM blk-mq ontop of old request_fn devices cannot fail in clone_rq() since the clone request is preallocated as part of the pdu. Reported-by:
-
Mike Snitzer authored
Otherwise kmemleak reported: unreferenced object 0xffff88009b14e2b0 (size 16): comm "fio", pid 4274, jiffies 4294978034 (age 1253.210s) hex dump (first 16 bytes): 40 12 f3 99 01 88 ff ff 00 10 00 00 00 00 00 00 @............... backtrace: [<ffffffff81600029>] kmemleak_alloc+0x49/0xb0 [<ffffffff811679a8>] kmem_cache_alloc+0xf8/0x160 [<ffffffff8111c950>] mempool_alloc_slab+0x10/0x20 [<ffffffff8111cb37>] mempool_alloc+0x57/0x150 [<ffffffffa04d2b61>] __multipath_map.isra.17+0xe1/0x220 [dm_multipath] [<ffffffffa04d2cb5>] multipath_clone_and_map+0x15/0x20 [dm_multipath] [<ffffffffa02889b5>] map_request.isra.39+0xd5/0x220 [dm_mod] [<ffffffffa028b0e4>] dm_mq_queue_rq+0x134/0x240 [dm_mod] [<ffffffff812cccb5>] __blk_mq_run_hw_queue+0x1d5/0x380 [<ffffffff812ccaa5>] blk_mq_run_hw_queue+0xc5/0x100 [<ffffffff812ce350>] blk_sq_make_request+0x240/0x300 [<ffffffff812c0f30>] generic_make_request+0xc0/0x110 [<ffffffff812c0ff2>] submit_bio+0x72/0x150 [<ffffffff811c07cb>] do_blockdev_direct_IO+0x1f3b/0x2da0 [<ffffffff811c166e>] __blockdev_direct_IO+0x3e/0x40 [<ffffffff8120aa1a>] ext4_direct_IO+0x1aa/0x390 Fixes: e5863d9a ("dm: allocate requests in target when stacking on blk-mq devices") Reported-by: -
Junichi Nomura authored
When stacking request-based DM on blk_mq device, request cloning and remapping are done in a single call to target's clone_and_map_rq(). The clone is allocated and valid only if clone_and_map_rq() returns DM_MAPIO_REMAPPED. The "IS_ERR(clone)" check in map_request() does not cover all the !DM_MAPIO_REMAPPED cases that are possible (E.g. if underlying devices are not ready or unavailable, clone_and_map_rq() may return DM_MAPIO_REQUEUE without ever having established an ERR_PTR). Fix this by explicitly checking for a return that is not DM_MAPIO_REMAPPED in map_request(). Without this fix, DM core may call setup_clone() for a NULL clone and oops like this: BUG: unable to handle kernel NULL pointer dereference at 0000000000000068 IP: [<ffffffff81227525>] blk_rq_prep_clone+0x7d/0x137 ... CPU: 2 PID: 5793 Comm: kdmwork-253:3 Not tainted 4.0.0-nm #1 ... Call Trace: [<ffffffffa01d1c09>] map_tio_request+0xa9/0x258 [dm_mod] [<ffffffff81071de9>] kthread_worker_fn+0xfd/0x150 [<ffffffff81071cec>] ? kthread_parkme+0x24/0x24 [<ffffffff81071cec>] ? kthread_parkme+0x24/0x24 [<ffffffff81071fdd>] kthread+0xe6/0xee [<ffffffff81093a59>] ? put_lock_stats+0xe/0x20 [<ffffffff81071ef7>] ? __init_kthread_worker+0x5b/0x5b [<ffffffff814c2d98>] ret_from_fork+0x58/0x90 [<ffffffff81071ef7>] ? __init_kthread_worker+0x5b/0x5b Fixes: e5863d9a ("dm: allocate requests in target when stacking on blk-mq devices") Reported-by:
-
- 26 May, 2015 2 commits
-
-
Julia Lawall authored
Remove unneeded variable used to store return value. Generated by: scripts/coccinelle/misc/returnvar.cocci Signed-off-by:
Fengguang Wu <fengguang.wu@intel.com> Signed-off-by:
Julia Lawall <julia.lawall@lip6.fr> Signed-off-by:
Jens Axboe <axboe@fb.com>
-
Junichi Nomura authored
Without kicking queue, requeued request may stay forever in the queue if there are no other I/O activities to the device. The original error had been in v2.6.39 with commit 7eaceacc ("block: remove per-queue plugging"), which replaced conditional plugging by periodic runqueue. Commit 9d1deb83 in v4.1-rc1 removed the periodic runqueue and the problem started to manifest. Fixes: 9d1deb83 ("dm: don't schedule delayed run of the queue if nothing to do") Signed-off-by:
-
