diff options
| author | Linus Torvalds <torvalds@linux-foundation.org> | 2018-01-31 22:05:47 +0300 |
|---|---|---|
| committer | Linus Torvalds <torvalds@linux-foundation.org> | 2018-01-31 22:05:47 +0300 |
| commit | 0be600a5add76e8e8b9e1119f2a7426ff849aca8 (patch) | |
| tree | d5fcc2b119f03143f9bed1b9aa5cb85458c8bd03 | |
| parent | 040639b7fcf73ee39c15d38257f652a2048e96f2 (diff) | |
| parent | 9614e2ba9161c7f5419f4212fa6057d2a65f6ae6 (diff) | |
| download | linux-0be600a5add76e8e8b9e1119f2a7426ff849aca8.tar.xz | |
Merge tag 'for-4.16/dm-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm
Pull device mapper updates from Mike Snitzer:
- DM core fixes to ensure that bio submission follows a depth-first
tree walk; this is critical to allow forward progress without the
need to use the bioset's BIOSET_NEED_RESCUER.
- Remove DM core's BIOSET_NEED_RESCUER based dm_offload infrastructure.
- DM core cleanups and improvements to make bio-based DM more efficient
(e.g. reduced memory footprint as well leveraging per-bio-data more).
- Introduce new bio-based mode (DM_TYPE_NVME_BIO_BASED) that leverages
the more direct IO submission path in the block layer; this mode is
used by DM multipath and also optimizes targets like DM thin-pool
that stack directly on NVMe data device.
- DM multipath improvements to factor out legacy SCSI-only (e.g.
scsi_dh) code paths to allow for more optimized support for NVMe
multipath.
- A fix for DM multipath path selectors (service-time and queue-length)
to select paths in a more balanced way; largely academic but doesn't
hurt.
- Numerous DM raid target fixes and improvements.
- Add a new DM "unstriped" target that enables Intel to workaround
firmware limitations in some NVMe drives that are striped internally
(this target also works when stacked above the DM "striped" target).
- Various Documentation fixes and improvements.
- Misc cleanups and fixes across various DM infrastructure and targets
(e.g. bufio, flakey, log-writes, snapshot).
* tag 'for-4.16/dm-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm: (69 commits)
dm cache: Documentation: update default migration_throttling value
dm mpath selector: more evenly distribute ties
dm unstripe: fix target length versus number of stripes size check
dm thin: fix trailing semicolon in __remap_and_issue_shared_cell
dm table: fix NVMe bio-based dm_table_determine_type() validation
dm: various cleanups to md->queue initialization code
dm mpath: delay the retry of a request if the target responded as busy
dm mpath: return DM_MAPIO_DELAY_REQUEUE if QUEUE_IO or PG_INIT_REQUIRED
dm mpath: return DM_MAPIO_REQUEUE on blk-mq rq allocation failure
dm log writes: fix max length used for kstrndup
dm: backfill missing calls to mutex_destroy()
dm snapshot: use mutex instead of rw_semaphore
dm flakey: check for null arg_name in parse_features()
dm thin: extend thinpool status format string with omitted fields
dm thin: fixes in thin-provisioning.txt
dm thin: document representation of <highest mapped sector> when there is none
dm thin: fix documentation relative to low water mark threshold
dm cache: be consistent in specifying sectors and SI units in cache.txt
dm cache: delete obsoleted paragraph in cache.txt
dm cache: fix grammar in cache-policies.txt
...
31 files changed, 1409 insertions, 671 deletions
diff --git a/Documentation/device-mapper/cache-policies.txt b/Documentation/device-mapper/cache-policies.txt index d3ca8af21a31..86786d87d9a8 100644 --- a/Documentation/device-mapper/cache-policies.txt +++ b/Documentation/device-mapper/cache-policies.txt @@ -60,7 +60,7 @@ Memory usage: The mq policy used 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 +smq uses 28bit indexes to implement its 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 @@ -84,7 +84,7 @@ resulting in better promotion/demotion decisions. Adaptability: The mq policy maintained a hit count for each cache block. For a -different block to get promoted to the cache it's hit count has to +different block to get promoted to the cache its hit count has to exceed the lowest currently in the cache. This meant it could take a long time for the cache to adapt between varying IO patterns. diff --git a/Documentation/device-mapper/cache.txt b/Documentation/device-mapper/cache.txt index cdfd0feb294e..ff0841711fd5 100644 --- a/Documentation/device-mapper/cache.txt +++ b/Documentation/device-mapper/cache.txt @@ -59,7 +59,7 @@ Fixed block size The origin is divided up into blocks of a fixed size. This block size is configurable when you first create the cache. Typically we've been using block sizes of 256KB - 1024KB. The block size must be between 64 -(32KB) and 2097152 (1GB) and a multiple of 64 (32KB). +sectors (32KB) and 2097152 sectors (1GB) and a multiple of 64 sectors (32KB). Having a fixed block size simplifies the target a lot. But it is something of a compromise. For instance, a small part of a block may be @@ -119,7 +119,7 @@ doing here to avoid migrating during those peak io moments. For the time being, a message "migration_threshold <#sectors>" can be used to set the maximum number of sectors being migrated, -the default being 204800 sectors (or 100MB). +the default being 2048 sectors (1MB). Updating on-disk metadata ------------------------- @@ -143,11 +143,6 @@ the policy how big this chunk is, but it should be kept small. Like the dirty flags this data is lost if there's a crash so a safe fallback value should always be possible. -For instance, the 'mq' policy, which is currently the default policy, -uses this facility to store the hit count of the cache blocks. If -there's a crash this information will be lost, which means the cache -may be less efficient until those hit counts are regenerated. - Policy hints affect performance, not correctness. Policy messaging diff --git a/Documentation/device-mapper/dm-raid.txt b/Documentation/device-mapper/dm-raid.txt index 32df07e29f68..390c145f01d7 100644 --- a/Documentation/device-mapper/dm-raid.txt +++ b/Documentation/device-mapper/dm-raid.txt @@ -343,5 +343,8 @@ Version History 1.11.0 Fix table line argument order (wrong raid10_copies/raid10_format sequence) 1.11.1 Add raid4/5/6 journal write-back support via journal_mode option -1.12.1 fix for MD deadlock between mddev_suspend() and md_write_start() available +1.12.1 Fix for MD deadlock between mddev_suspend() and md_write_start() available 1.13.0 Fix dev_health status at end of "recover" (was 'a', now 'A') +1.13.1 Fix deadlock caused by early md_stop_writes(). Also fix size an + state races. +1.13.2 Fix raid redundancy validation and avoid keeping raid set frozen diff --git a/Documentation/device-mapper/snapshot.txt b/Documentation/device-mapper/snapshot.txt index ad6949bff2e3..b8bbb516f989 100644 --- a/Documentation/device-mapper/snapshot.txt +++ b/Documentation/device-mapper/snapshot.txt @@ -49,6 +49,10 @@ The difference between persistent and transient is with transient snapshots less metadata must be saved on disk - they can be kept in memory by the kernel. +When loading or unloading the snapshot target, the corresponding +snapshot-origin or snapshot-merge target must be suspended. A failure to +suspend the origin target could result in data corruption. + * snapshot-merge <origin> <COW device> <persistent> <chunksize> diff --git a/Documentation/device-mapper/thin-provisioning.txt b/Documentation/device-mapper/thin-provisioning.txt index 1699a55b7b70..4bcd4b7f79f9 100644 --- a/Documentation/device-mapper/thin-provisioning.txt +++ b/Documentation/device-mapper/thin-provisioning.txt @@ -112,9 +112,11 @@ $low_water_mark is expressed in blocks of size $data_block_size. If free space on the data device drops below this level then a dm event will be triggered which a userspace daemon should catch allowing it to extend the pool device. Only one such event will be sent. -Resuming a device with a new table itself triggers an event so the -userspace daemon can use this to detect a situation where a new table -already exceeds the threshold. + +No special event is triggered if a just resumed device's free space is below +the low water mark. However, resuming a device always triggers an +event; a userspace daemon should verify that free space exceeds the low +water mark when handling this event. A low water mark for the metadata device is maintained in the kernel and will trigger a dm event if free space on the metadata device drops below @@ -274,7 +276,8 @@ ii) Status <transaction id> <used metadata blocks>/<total metadata blocks> <used data blocks>/<total data blocks> <held metadata root> - [no_]discard_passdown ro|rw + ro|rw|out_of_data_space [no_]discard_passdown [error|queue]_if_no_space + needs_check|- transaction id: A 64-bit number used by userspace to help synchronise with metadata @@ -394,3 +397,6 @@ ii) Status If the pool has encountered device errors and failed, the status will just contain the string 'Fail'. The userspace recovery tools should then be used. + + In the case where <nr mapped sectors> is 0, there is no highest + mapped sector and the value of <highest mapped sector> is unspecified. diff --git a/Documentation/device-mapper/unstriped.txt b/Documentation/device-mapper/unstriped.txt new file mode 100644 index 000000000000..0b2a306c54ee --- /dev/null +++ b/Documentation/device-mapper/unstriped.txt @@ -0,0 +1,124 @@ +Introduction +============ + +The device-mapper "unstriped" target provides a transparent mechanism to +unstripe a device-mapper "striped" target to access the underlying disks +without having to touch the true backing block-device. It can also be +used to unstripe a hardware RAID-0 to access backing disks. + +Parameters: +<number of stripes> <chunk size> <stripe #> <dev_path> <offset> + +<number of stripes> + The number of stripes in the RAID 0. + +<chunk size> + The amount of 512B sectors in the chunk striping. + +<dev_path> + The block device you wish to unstripe. + +<stripe #> + The stripe number within the device that corresponds to physical + drive you wish to unstripe. This must be 0 indexed. + + +Why use this module? +==================== + +An example of undoing an existing dm-stripe +------------------------------------------- + +This small bash script will setup 4 loop devices and use the existing +striped target to combine the 4 devices into one. It then will use +the unstriped target ontop of the striped device to access the +individual backing loop devices. We write data to the newly exposed +unstriped devices and verify the data written matches the correct +underlying device on the striped array. + +#!/bin/bash + +MEMBER_SIZE=$((128 * 1024 * 1024)) +NUM=4 +SEQ_END=$((${NUM}-1)) +CHUNK=256 +BS=4096 + +RAID_SIZE=$((${MEMBER_SIZE}*${NUM}/512)) +DM_PARMS="0 ${RAID_SIZE} striped ${NUM} ${CHUNK}" +COUNT=$((${MEMBER_SIZE} / ${BS})) + +for i in $(seq 0 ${SEQ_END}); do + dd if=/dev/zero of=member-${i} bs=${MEMBER_SIZE} count=1 oflag=direct + losetup /dev/loop${i} member-${i} + DM_PARMS+=" /dev/loop${i} 0" +done + +echo $DM_PARMS | dmsetup create raid0 +for i in $(seq 0 ${SEQ_END}); do + echo "0 1 unstriped ${NUM} ${CHUNK} ${i} /dev/mapper/raid0 0" | dmsetup create set-${i} +done; + +for i in $(seq 0 ${SEQ_END}); do + dd if=/dev/urandom of=/dev/mapper/set-${i} bs=${BS} count=${COUNT} oflag=direct + diff /dev/mapper/set-${i} member-${i} +done; + +for i in $(seq 0 ${SEQ_END}); do + dmsetup remove set-${i} +done + +dmsetup remove raid0 + +for i in $(seq 0 ${SEQ_END}); do + losetup -d /dev/loop${i} + rm -f member-${i} +done + +Another example +--------------- + +Intel NVMe drives contain two cores on the physical device. +Each core of the drive has segregated access to its LBA range. +The current LBA model has a RAID 0 128k chunk on each core, resulting +in a 256k stripe across the two cores: + + Core 0: Core 1: + __________ __________ + | LBA 512| | LBA 768| + | LBA 0 | | LBA 256| + ---------- ---------- + +The purpose of this unstriping is to provide better QoS in noisy +neighbor environments. When two partitions are created on the +aggregate drive without this unstriping, reads on one partition +can affect writes on another partition. This is because the partitions +are striped across the two cores. When we unstripe this hardware RAID 0 +and make partitions on each new exposed device the two partitions are now +physically separated. + +With the dm-unstriped target we're able to segregate an fio script that +has read and write jobs that are independent of each other. Compared to +when we run the test on a combined drive with partitions, we were able +to get a 92% reduction in read latency using this device mapper target. + + +Example dmsetup usage +===================== + +unstriped ontop of Intel NVMe device that has 2 cores +----------------------------------------------------- +dmsetup create nvmset0 --table '0 512 unstriped 2 256 0 /dev/nvme0n1 0' +dmsetup create nvmset1 --table '0 512 unstriped 2 256 1 /dev/nvme0n1 0' + +There will now be two devices that expose Intel NVMe core 0 and 1 +respectively: +/dev/mapper/nvmset0 +/dev/mapper/nvmset1 + +unstriped ontop of striped with 4 drives using 128K chunk size +-------------------------------------------------------------- +dmsetup create raid_disk0 --table '0 512 unstriped 4 256 0 /dev/mapper/striped 0' +dmsetup create raid_disk1 --table '0 512 unstriped 4 256 1 /dev/mapper/striped 0' +dmsetup create raid_disk2 --table '0 512 unstriped 4 256 2 /dev/mapper/striped 0' +dmsetup create raid_disk3 --table '0 512 unstriped 4 256 3 /dev/mapper/striped 0' diff --git a/drivers/md/Kconfig b/drivers/md/Kconfig index 83b9362be09c..2c8ac3688815 100644 --- a/drivers/md/Kconfig +++ b/drivers/md/Kconfig @@ -269,6 +269,13 @@ config DM_BIO_PRISON source "drivers/md/persistent-data/Kconfig" +config DM_UNSTRIPED + tristate "Unstriped target" + depends on BLK_DEV_DM + ---help--- + Unstripes I/O so it is issued solely on a single drive in a HW + RAID0 or dm-striped target. + config DM_CRYPT tristate "Crypt target support" depends on BLK_DEV_DM diff --git a/drivers/md/Makefile b/drivers/md/Makefile index f701bb211783..63255f3ebd97 100644 --- a/drivers/md/Makefile +++ b/drivers/md/Makefile @@ -43,6 +43,7 @@ obj-$(CONFIG_BCACHE) += bcache/ obj-$(CONFIG_BLK_DEV_MD) += md-mod.o obj-$(CONFIG_BLK_DEV_DM) += dm-mod.o obj-$(CONFIG_BLK_DEV_DM_BUILTIN) += dm-builtin.o +obj-$(CONFIG_DM_UNSTRIPED) += dm-unstripe.o obj-$(CONFIG_DM_BUFIO) += dm-bufio.o obj-$(CONFIG_DM_BIO_PRISON) += dm-bio-prison.o obj-$(CONFIG_DM_CRYPT) += dm-crypt.o diff --git a/drivers/md/dm-bufio.c b/drivers/md/dm-bufio.c index c546b567f3b5..414c9af54ded 100644 --- a/drivers/md/dm-bufio.c +++ b/drivers/md/dm-bufio.c @@ -662,7 +662,7 @@ static void submit_io(struct dm_buffer *b, int rw, bio_end_io_t *end_io) sector = (b->block << b->c->sectors_per_block_bits) + b->c->start; - if (rw != WRITE) { + if (rw != REQ_OP_WRITE) { n_sectors = 1 << b->c->sectors_per_block_bits; offset = 0; } else { @@ -740,7 +740,7 @@ static void __write_dirty_buffer(struct dm_buffer *b, b->write_end = b->dirty_end; if (!write_list) - submit_io(b, WRITE, write_endio); + submit_io(b, REQ_OP_WRITE, write_endio); else list_add_tail(&b->write_list, write_list); } @@ -753,7 +753,7 @@ static void __flush_write_list(struct list_head *write_list) struct dm_buffer *b = list_entry(write_list->next, struct dm_buffer, write_list); list_del(&b->write_list); - submit_io(b, WRITE, write_endio); + submit_io(b, REQ_OP_WRITE, write_endio); cond_resched(); } blk_finish_plug(&plug); @@ -1123,7 +1123,7 @@ static void *new_read(struct dm_bufio_client *c, sector_t block, return NULL; if (need_submit) - submit_io(b, READ, read_endio); + submit_io(b, REQ_OP_READ, read_endio); wait_on_bit_io(&b->state, B_READING, TASK_UNINTERRUPTIBLE); @@ -1193,7 +1193,7 @@ void dm_bufio_prefetch(struct dm_bufio_client *c, dm_bufio_unlock(c); if (need_submit) - submit_io(b, READ, read_endio); + submit_io(b, REQ_OP_READ, read_endio); dm_bufio_release(b); cond_resched(); @@ -1454,7 +1454,7 @@ retry: old_block = b->block; __unlink_buffer(b); __link_buffer(b, new_block, b->list_mode); - submit_io(b, WRITE, write_endio); + submit_io(b, REQ_OP_WRITE, write_endio); wait_on_bit_io(&b->state, B_WRITING, TASK_UNINTERRUPTIBLE); __unlink_buffer(b); @@ -1716,7 +1716,7 @@ struct dm_bufio_client *dm_bufio_client_create(struct block_device *bdev, unsign if (!DM_BUFIO_CACHE_NAME(c)) { r = -ENOMEM; mutex_unlock(&dm_bufio_clients_lock); - goto bad_cache; + goto bad; } } @@ -1727,7 +1727,7 @@ struct dm_bufio_client *dm_bufio_client_create(struct block_device *bdev, unsign if (!DM_BUFIO_CACHE(c)) { r = -ENOMEM; mutex_unlock(&dm_bufio_clients_lock); - goto bad_cache; + goto bad; } } } @@ -1738,27 +1738,28 @@ struct dm_bufio_client *dm_bufio_client_create(struct block_device *bdev, unsign if (!b) { r = -ENOMEM; - goto bad_buffer; + goto bad; } __free_buffer_wake(b); } + c->shrinker.count_objects = dm_bufio_shrink_count; + c->shrinker.scan_objects = dm_bufio_shrink_scan; + c->shrinker.seeks = 1; + c->shrinker.batch = 0; + r = register_shrinker(&c->shrinker); + if (r) + goto bad; + mutex_lock(&dm_bufio_clients_lock); dm_bufio_client_count++; list_add(&c->client_list, &dm_bufio_all_clients); __cache_size_refresh(); mutex_unlock(&dm_bufio_clients_lock); - c->shrinker.count_objects = dm_bufio_shrink_count; - c->shrinker.scan_objects = dm_bufio_shrink_scan; - c->shrinker.seeks = 1; - c->shrinker.batch = 0; - register_shrinker(&c->shrinker); - return c; -bad_buffer: -bad_cache: +bad: while (!list_empty(&c->reserved_buffers)) { struct dm_buffer *b = list_entry(c->reserved_buffers.next, struct dm_buffer, lru_list); @@ -1767,6 +1768,7 @@ bad_cache: } dm_io_client_destroy(c->dm_io); bad_dm_io: + mutex_destroy(&c->lock); kfree(c); bad_client: return ERR_PTR(r); @@ -1811,6 +1813,7 @@ void dm_bufio_client_destroy(struct dm_bufio_client *c) BUG_ON(c->n_buffers[i]); dm_io_client_destroy(c->dm_io); + mutex_destroy(&c->lock); kfree(c); } EXPORT_SYMBOL_GPL(dm_bufio_client_destroy); diff --git a/drivers/md/dm-core.h b/drivers/md/dm-core.h index 6a14f945783c..3222e21cbbf8 100644 --- a/drivers/md/dm-core.h +++ b/drivers/md/dm-core.h @@ -91,8 +91,7 @@ struct mapped_device { /* * io objects are allocated from here. */ - mempool_t *io_pool; - + struct bio_set *io_bs; struct bio_set *bs; /* @@ -130,8 +129,6 @@ struct mapped_device { struct srcu_struct io_barrier; }; -void dm_init_md_queue(struct mapped_device *md); -void dm_init_normal_md_queue(struct mapped_device *md); int md_in_flight(struct mapped_device *md); void disable_write_same(struct mapped_device *md); void disable_write_zeroes(struct mapped_device *md); diff --git a/drivers/md/dm-crypt.c b/drivers/md/dm-crypt.c index 2ad429100d25..8168f737590e 100644 --- a/drivers/md/dm-crypt.c +++ b/drivers/md/dm-crypt.c @@ -2193,6 +2193,8 @@ static void crypt_dtr(struct dm_target *ti) kzfree(cc->cipher_auth); kzfree(cc->authenc_key); + mutex_destroy(&cc->bio_alloc_lock); + /* Must zero key material before freeing */ kzfree(cc); } @@ -2702,8 +2704,7 @@ static int crypt_ctr(struct dm_target *ti, unsigned int argc, char **argv) goto bad; } - cc->bs = bioset_create(MIN_IOS, 0, (BIOSET_NEED_BVECS | - BIOSET_NEED_RESCUER)); + cc->bs = bioset_create(MIN_IOS, 0, BIOSET_NEED_BVECS); if (!cc->bs) { ti->error = "Cannot allocate crypt bioset"; goto bad; diff --git a/drivers/md/dm-delay.c b/drivers/md/dm-delay.c index 288386bfbfb5..1783d80c9cad 100644 --- a/drivers/md/dm-delay.c +++ b/drivers/md/dm-delay.c @@ -229,6 +229,8 @@ static void delay_dtr(struct dm_target *ti) if (dc->dev_write) dm_put_device(ti, dc->dev_write); + mutex_destroy(&dc->timer_lock); + kfree(dc); } diff --git a/drivers/md/dm-flakey.c b/drivers/md/dm-flakey.c index b82cb1ab1eaa..1b907b15f5c3 100644 --- a/drivers/md/dm-flakey.c +++ b/drivers/md/dm-flakey.c @@ -70,6 +70,11 @@ static int parse_features(struct dm_arg_set *as, struct flakey_c *fc, arg_name = dm_shift_arg(as); argc--; + if (!arg_name) { + ti->error = "Insufficient feature arguments"; + return -EINVAL; + } + /* * drop_writes */ diff --git a/drivers/md/dm-io.c b/drivers/md/dm-io.c index b4357ed4d541..a8d914d5abbe 100644 --- a/drivers/md/dm-io.c +++ b/drivers/md/dm-io.c @@ -58,8 +58,7 @@ struct dm_io_client *dm_io_client_create(void) if (!client->pool) goto bad; - client->bios = bioset_create(min_ios, 0, (BIOSET_NEED_BVECS | - BIOSET_NEED_RESCUER)); + client->bios = bioset_create(min_ios, 0, BIOSET_NEED_BVECS); if (!client->bios) goto bad; diff --git a/drivers/md/dm-kcopyd.c b/drivers/md/dm-kcopyd.c index eb45cc3df31d..e6e7c686646d 100644 --- a/drivers/md/dm-kcopyd.c +++ b/drivers/md/dm-kcopyd.c @@ -477,8 +477,10 @@ static int run_complete_job(struct kcopyd_job *job) * If this is the master job, the sub jobs have already * completed so we can free everything. */ - if (job->master_job == job) + if (job->master_job == job) { + mutex_destroy(&job->lock); mempool_free(job, kc->job_pool); + } fn(read_err, write_err, context); if (atomic_dec_and_test(&kc->nr_jobs)) @@ -750,6 +752,7 @@ int dm_kcopyd_copy(struct dm_kcopyd_client *kc, struct dm_io_region *from, * followed by SPLIT_COUNT sub jobs. */ job = mempool_alloc(kc->job_pool, GFP_NOIO); + mutex_init(&job->lock); /* * set up for the read. @@ -811,7 +814,6 @@ int dm_kcopyd_copy(struct dm_kcopyd_client *kc, struct dm_io_region *from, if (job->source.count <= SUB_JOB_SIZE) dispatch_job(job); else { - mutex_init(&job->lock); job->progress = 0; split_job(job); } diff --git a/drivers/md/dm-log-writes.c b/drivers/md/dm-log-writes.c index 189badbeddaf..3362d866793b 100644 --- a/drivers/md/dm-log-writes.c +++ b/drivers/md/dm-log-writes.c @@ -594,7 +594,7 @@ static int log_mark(struct log_writes_c *lc, char *data) return -ENOMEM; } - block->data = kstrndup(data, maxsize, GFP_KERNEL); + block->data = kstrndup(data, maxsize - 1, GFP_KERNEL); if (!block->data) { DMERR("Error copying mark data"); kfree(block); diff --git a/drivers/md/dm-mpath.c b/drivers/md/dm-mpath.c index ef57c6d1c887..7d3e572072f5 100644 --- a/drivers/md/dm-mpath.c +++ b/drivers/md/dm-mpath.c @@ -64,36 +64,30 @@ struct priority_group { /* Multipath context */ struct multipath { - struct list_head list; - struct dm_target *ti; - - const char *hw_handler_name; - char *hw_handler_params; + unsigned long flags; /* Multipath state flags */ spinlock_t lock; - - unsigned nr_priority_groups; - struct list_head priority_groups; - - wait_queue_head_t pg_init_wait; /* Wait for pg_init completion */ + enum dm_queue_mode queue_mode; struct pgpath *current_pgpath; struct priority_group *current_pg; struct priority_group *next_pg; /* Switch to this PG if set */ - unsigned long flags; /* Multipath state flags */ + atomic_t nr_valid_paths; /* Total number of usable paths */ + unsigned nr_priority_groups; + struct list_head priority_groups; + const char *hw_handler_name; + char *hw_handler_params; + wait_queue_head_t pg_init_wait; /* Wait for pg_init completion */ unsigned pg_init_retries; /* Number of times to retry pg_init */ unsigned pg_init_delay_msecs; /* Number of msecs before pg_init retry */ - - atomic_t nr_valid_paths; /* Total number of usable paths */ atomic_t pg_init_in_progress; /* Only one pg_init allowed at once */ atomic_t pg_init_count; /* Number of times pg_init called */ - enum dm_queue_mode queue_mode; - struct mutex work_mutex; struct work_struct trigger_event; + struct dm_target *ti; struct work_struct process_queued_bios; struct bio_list queued_bios; @@ -135,10 +129,10 @@ static struct pgpath *alloc_pgpath(void) { struct pgpath *pgpath = kzalloc(sizeof(*pgpath), GFP_KERNEL); - if (pgpath) { - pgpath->is_active = true; - INIT_DELAYED_WORK(&pgpath->activate_path, activate_path_work); - } + if (!pgpath) + return NULL; + + pgpath->is_active = true; return pgpath; } @@ -193,13 +187,8 @@ static struct multipath *alloc_multipath(struct dm_target *ti) if (m) { INIT_LIST_HEAD(&m->priority_groups); spin_lock_init(&m->lock); - set_bit(MPATHF_QUEUE_IO, &m->flags); atomic_set(&m->nr_valid_paths, 0); - atomic_set(&m->pg_init_in_progress, 0); - atomic_set(&m->pg_init_count, 0); - m->pg_init_delay_msecs = DM_PG_INIT_DELAY_DEFAULT; INIT_WORK(&m->trigger_event, trigger_event); - init_waitqueue_head(&m->pg_init_wait); mutex_init(&m->work_mutex); m->queue_mode = DM_TYPE_NONE; @@ -221,13 +210,26 @@ static int alloc_multipath_stage2(struct dm_target *ti, struct multipath *m) m->queue_mode = DM_TYPE_MQ_REQUEST_BASED; else m->queue_mode = DM_TYPE_REQUEST_BASED; - } else if (m->queue_mode == DM_TYPE_BIO_BASED) { + + } else if (m->queue_mode == DM_TYPE_BIO_BASED || + m->queue_mode == DM_TYPE_NVME_BIO_BASED) { INIT_WORK(&m->process_queued_bios, process_queued_bios); - /* - * bio-based doesn't support any direct scsi_dh management; - * it just discovers if a scsi_dh is attached. - */ - set_bit(MPATHF_RETAIN_ATTACHED_HW_HANDLER, &m->flags); + + if (m->queue_mode == DM_TYPE_BIO_BASED) { + /* + * bio-based doesn't support any direct scsi_dh management; + * it just discovers if a scsi_dh is attached. + */ + set_bit(MPATHF_RETAIN_ATTACHED_HW_HANDLER, &m->flags); + } + } + + if (m->queue_mode != DM_TYPE_NVME_BIO_BASED) { + set_bit(MPATHF_QUEUE_IO, &m->flags); + atomic_set(&m->pg_init_in_progress, 0); + atomic_set(&m->pg_init_count, 0); + m->pg_init_delay_msecs = DM_PG_INIT_DELAY_DEFAULT; + init_waitqueue_head(&m->pg_init_wait); } dm_table_set_type(ti->table, m->queue_mode); @@ -246,6 +248,7 @@ static void free_multipath(struct multipath *m) kfree(m->hw_handler_name); kfree(m->hw_handler_params); + mutex_destroy(&m->work_mutex); kfree(m); } @@ -264,29 +267,23 @@ static struct dm_mpath_io *get_mpio_from_bio(struct bio *bio) return dm_per_bio_data(bio, multipath_per_bio_data_size()); } -static struct dm_bio_details *get_bio_details_from_bio(struct bio *bio) +static struct dm_bio_details *get_bio_details_from_mpio(struct dm_mpath_io *mpio) { /* dm_bio_details is immediately after the dm_mpath_io in bio's per-bio-data */ - struct dm_mpath_io *mpio = get_mpio_from_bio(bio); void *bio_details = mpio + 1; - return bio_details; } -static void multipath_init_per_bio_data(struct bio *bio, struct dm_mpath_io **mpio_p, - struct dm_bio_details **bio_details_p) +static void multipath_init_per_bio_data(struct bio *bio, struct dm_mpath_io **mpio_p) { struct dm_mpath_io *mpio = get_mpio_from_bio(bio); - struct dm_bio_details *bio_details = get_bio_details_from_bio(bio); + struct dm_bio_details *bio_details = get_bio_details_from_mpio(mpio); - memset(mpio, 0, sizeof(*mpio)); - memset(bio_details, 0, sizeof(*bio_details)); - dm_bio_record(bio_details, bio); + mpio->nr_bytes = bio->bi_iter.bi_size; + mpio->pgpath = NULL; + *mpio_p = mpio; - if (mpio_p) - *mpio_p = mpio; - if (bio_details_p) - *bio_details_p = bio_details; + dm_bio_record(bio_details, bio); } /*----------------------------------------------- @@ -340,6 +337,9 @@ static void __switch_pg(struct multipath *m, struct priority_group *pg) { m->current_pg = pg; + if (m->queue_mode == DM_TYPE_NVME_BIO_BASED) + return; + /* Must we initialise the PG first, and queue I/O till it's ready? */ if (m->hw_handler_name) { set_bit(MPATHF_PG_INIT_REQUIRED, &m->flags); @@ -385,7 +385,8 @@ static struct pgpath *choose_pgpath(struct multipath *m, size_t nr_bytes) unsigned bypassed = 1; if (!atomic_read(&m->nr_valid_paths)) { - clear_bit(MPATHF_QUEUE_IO, &m->flags); + if (m->queue_mode != DM_TYPE_NVME_BIO_BASED) + clear_bit(MPATHF_QUEUE_IO, &m->flags); goto failed; } @@ -516,12 +517,10 @@ static int multipath_clone_and_map(struct dm_target *ti, struct request *rq, return DM_MAPIO_KILL; } else if (test_bit(MPATHF_QUEUE_IO, &m->flags) || test_bit(MPATHF_PG_INIT_REQUIRED, &m->flags)) { - if (pg_init_all_paths(m)) - return DM_MAPIO_DELAY_REQUEUE; - return DM_MAPIO_REQUEUE; + pg_init_all_paths(m); + return DM_MAPIO_DELAY_REQUEUE; } - memset(mpio, 0, sizeof(*mpio)); mpio->pgpath = pgpath; mpio->nr_bytes = nr_bytes; @@ -530,12 +529,23 @@ static int multipath_clone_and_map(struct dm_target *ti, struct request *rq, clone = blk_get_request(q, rq->cmd_flags | REQ_NOMERGE, GFP_ATOMIC); if (IS_ERR(clone)) { /* EBUSY, ENODEV or EWOULDBLOCK: requeue */ - bool queue_dying = blk_queue_dying(q); - if (queue_dying) { + if (blk_queue_dying(q)) { atomic_inc(&m->pg_init_in_progress); activate_or_offline_path(pgpath); + return DM_MAPIO_DELAY_REQUEUE; } - return DM_MAPIO_DELAY_REQUEUE; + + /* + * blk-mq's SCHED_RESTART can cover this requeue, so we + * needn't deal with it by DELAY_REQUEUE. More importantly, + * we have to return DM_MAPIO_REQUEUE so that blk-mq can + * get the queue busy feedback (via BLK_STS_RESOURCE), + * otherwise I/O merging can suffer. + */ + if (q->mq_ops) + return DM_MAPIO_REQUEUE; + else + return DM_MAPIO_DELAY_REQUEUE; } clone->bio = clone->biotail = NULL; clone->rq_disk = bdev->bd_disk; @@ -557,9 +567,9 @@ static void multipath_release_clone(struct request *clone) /* * Map cloned bios (bio-based multipath) */ -static int __multipath_map_bio(struct multipath *m, struct bio *bio, struct dm_mpath_io *mpio) + +static struct pgpath *__map_bio(struct multipath *m, struct bio *bio) { - size_t nr_bytes = bio->bi_iter.bi_size; struct pgpath *pgpath; unsigned long flags; bool queue_io; @@ -568,7 +578,7 @@ static int __multipath_map_bio(struct multipath *m, struct bio *bio, struct dm_m pgpath = READ_ONCE(m->current_pgpath); queue_io = test_bit(MPATHF_QUEUE_IO, &m->flags); if (!pgpath || !queue_io) - pgpath = choose_pgpath(m, nr_bytes); + pgpath = choose_pgpath(m, bio->bi_iter.bi_size); if ((pgpath && queue_io) || (!pgpath && test_bit(MPATHF_QUEUE_IF_NO_PATH, &m->flags))) { @@ -576,14 +586,62 @@ static int __multipath_map_bio(struct multipath *m, struct bio *bio, struct dm_m spin_lock_irqsave(&m->lock, flags); bio_list_add(&m->queued_bios, bio); spin_unlock_irqrestore(&m->lock, flags); + /* PG_INIT_REQUIRED cannot be set without QUEUE_IO */ if (queue_io || test_bit(MPATHF_PG_INIT_REQUIRED, &m->flags)) pg_init_all_paths(m); else if (!queue_io) queue_work(kmultipathd, &m->process_queued_bios); - return DM_MAPIO_SUBMITTED; + + return ERR_PTR(-EAGAIN); } + return pgpath; +} + +static struct pgpath *__map_bio_nvme(struct multipath *m, struct bio *bio) +{ + struct pgpath *pgpath; + unsigned long flags; + + /* Do we need to select a new pgpath? */ + /* + * FIXME: currently only switching path if no path (due to failure, etc) + * - which negates the point of using a path selector + */ + pgpath = READ_ONCE(m->current_pgpath); + if (!pgpath) + pgpath = choose_pgpath(m, bio->bi_iter.bi_size); + + if (!pgpath) { + if (test_bit(MPATHF_QUEUE_IF_NO_PATH, &m->flags)) { + /* Queue for the daemon to resubmit */ + spin_lock_irqsave(&m->lock, flags); + bio_list_add(&m->queued_bios, bio); + spin_unlock_irqrestore(&m->lock, flags); + queue_work(kmultipathd, &m->process_queued_bios); + + return ERR_PTR(-EAGAIN); + } + return NULL; + } + + return pgpath; +} + +static int __multipath_map_bio(struct multipath *m, struct bio *bio, + struct dm_mpath_io *mpio) +{ + struct pgpath *pgpath; + + if (m->queue_mode == DM_TYPE_NVME_BIO_BASED) + pgpath = __map_bio_nvme(m, bio); + else + pgpath = __map_bio(m, bio); + + if (IS_ERR(pgpath)) + return DM_MAPIO_SUBMITTED; + if (!pgpath) { if (must_push_back_bio(m)) return DM_MAPIO_REQUEUE; @@ -592,7 +650,6 @@ static int __multipath_map_bio(struct multipath *m, struct bio *bio, struct dm_m } mpio->pgpath = pgpath; - mpio->nr_bytes = nr_bytes; bio->bi_status = 0; bio_set_dev(bio, pgpath->path.dev->bdev); @@ -601,7 +658,7 @@ static int __multipath_map_bio(struct multipath *m, struct bio *bio, struct dm_m if (pgpath->pg->ps.type->start_io) pgpath->pg->ps.type->start_io(&pgpath->pg->ps, &pgpath->path, - nr_bytes); + mpio->nr_bytes); return DM_MAPIO_REMAPPED; } @@ -610,8 +667,7 @@ static int multipath_map_bio(struct dm_target *ti, struct bio *bio) struct multipath *m = ti->private; struct dm_mpath_io *mpio = NULL; - multipath_init_per_bio_data(bio, &mpio, NULL); - + multipath_init_per_bio_data(bio, &mpio); return __multipath_map_bio(m, bio, mpio); } @@ -619,7 +675,8 @@ static void process_queued_io_list(struct multipath *m) { if (m->queue_mode == DM_TYPE_MQ_REQUEST_BASED) dm_mq_kick_requeue_list(dm_table_get_md(m->ti->table)); - else if (m->queue_mode == DM_TYPE_BIO_BASED) + else if (m->queue_mode == DM_TYPE_BIO_BASED || + m->queue_mode == DM_TYPE_NVME_BIO_BASED) queue_work(kmultipathd, &m->process_queued_bios); } @@ -649,7 +706,9 @@ static void process_queued_bios(struct work_struct *work) blk_start_plug(&plug); while ((bio = bio_list_pop(&bios))) { - r = __multipath_map_bio(m, bio, get_mpio_from_bio(bio)); + struct dm_mpath_io *mpio = get_mpio_from_bio(bio); + dm_bio_restore(get_bio_details_from_mpio(mpio), bio); + r = __multipath_map_bio(m, bio, mpio); switch (r) { case DM_MAPIO_KILL: bio->bi_status = BLK_STS_IOERR; @@ -752,34 +811,11 @@ static int parse_path_selector(struct dm_arg_set *as, struct priority_group *pg, return 0; } -static struct pgpath *parse_path(struct dm_arg_set *as, struct path_selector *ps, - struct dm_target *ti) +static int setup_scsi_dh(struct block_device *bdev, struct multipath *m, char **error) { - int r; - struct pgpath *p; - struct multipath *m = ti->private; - struct request_queue *q = NULL; + struct request_queue *q = bdev_get_queue(bdev); const char *attached_handler_name; - - /* we need at least a path arg */ - if (as->argc < 1) { - ti->error = "no device given"; - return ERR_PTR(-EINVAL); - } - - p = alloc_pgpath(); - if (!p) - return ERR_PTR(-ENOMEM); - - r = dm_get_device(ti, dm_shift_arg(as), dm_table_get_mode(ti->table), - &p->path.dev); - if (r) { - ti->error = "error getting device"; - goto bad; - } - - if (test_bit(MPATHF_RETAIN_ATTACHED_HW_HANDLER, &m->flags) || m->hw_handler_name) - q = bdev_get_queue(p->path.dev->bdev); + int r; if (test_bit(MPATHF_RETAIN_ATTACHED_HW_HANDLER, &m->flags)) { retain: @@ -811,26 +847,59 @@ retain: char b[BDEVNAME_SIZE]; printk(KERN_INFO "dm-mpath: retaining handler on device %s\n", - bdevname(p->path.dev->bdev, b)); + bdevname(bdev, b)); goto retain; } if (r < 0) { - ti->error = "error attaching hardware handler"; - dm_put_device(ti, p->path.dev); - goto bad; + *error = "error attaching hardware handler"; + return r; } if (m->hw_handler_params) { r = scsi_dh_set_params(q, m->hw_handler_params); if (r < 0) { - ti->error = "unable to set hardware " - "handler parameters"; - dm_put_device(ti, p->path.dev); - goto bad; + *error = "unable to set hardware handler parameters"; + return r; } } } + return 0; +} + +static struct pgpath *parse_path(struct dm_arg_set *as, struct path_selector *ps, + struct dm_target *ti) +{ + int r; + struct pgpath *p; + struct multipath *m = ti->private; + + /* we need at least a path arg */ + if (as->argc < 1) { + ti->error = "no device given"; + return ERR_PTR(-EINVAL); + } + + p = alloc_pgpath(); + if (!p) + return ERR_PTR(-ENOMEM); + + r = dm_get_device(ti, dm_shift_arg(as), dm_table_get_mode(ti->table), + &p->path.dev); + if (r) { + ti->error = "error getting device"; + goto bad; + } + + if (m->queue_mode != DM_TYPE_NVME_BIO_BASED) { + INIT_DELAYED_WORK(&p->activate_path, activate_path_work); + r = setup_scsi_dh(p->path.dev->bdev, m, &ti->error); + if (r) { + dm_put_device(ti, p->path.dev); + goto bad; + } + } + r = ps->type->add_path(ps, &p->path, as->argc, as->argv, &ti->error); if (r) { dm_put_device(ti, p->path.dev); @@ -838,7 +907,6 @@ retain: } return p; - bad: free_pgpath(p); return ERR_PTR(r); @@ -933,7 +1001,8 @@ static int parse_hw_handler(struct dm_arg_set *as, struct multipath *m) if (!hw_argc) return 0; - if (m->queue_mode == DM_TYPE_BIO_BASED) { + if (m->queue_mode == DM_TYPE_BIO_BASED || + m->queue_mode == DM_TYPE_NVME_BIO_BASED) { dm_consume_args(as, hw_argc); DMERR("bio-based multipath doesn't allow hardware handler args"); return 0; @@ -1022,6 +1091,8 @@ static int parse_features(struct dm_arg_set *as, struct multipath *m) if (!strcasecmp(queue_mode_name, "bio")) m->queue_mode = DM_TYPE_BIO_BASED; + else if (!strcasecmp(queue_mode_name, "nvme")) + m->queue_mode = DM_TYPE_NVME_BIO_BASED; else if (!strcasecmp(queue_mode_name, "rq")) m->queue_mode = DM_TYPE_REQUEST_BASED; else if (!strcasecmp(queue_mode_name, "mq")) @@ -1122,7 +1193,7 @@ static int multipath_ctr(struct dm_target *ti, unsigned argc, char **argv) ti->num_discard_bios = 1; ti->num_write_same_bios = 1; ti->num_write_zeroes_bios = 1; - if (m->queue_mode == DM_TYPE_BIO_BASED) + if (m->queue_mode == DM_TYPE_BIO_BASED || m->queue_mode == DM_TYPE_NVME_BIO_BASED) ti->per_io_data_size = multipath_per_bio_data_size(); else ti->per_io_data_size = sizeof(struct dm_mpath_io); @@ -1151,16 +1222,19 @@ static void multipath_wait_for_pg_init_completion(struct multipath *m) static void flush_multipath_work(struct multipath *m) { - set_bit(MPATHF_PG_INIT_DISABLED, &m->flags); - smp_mb__after_atomic(); + if (m->hw_handler_name) { + set_bit(MPATHF_PG_INIT_DISABLED, &m->flags); + smp_mb__after_atomic(); + + flush_workqueue(kmpath_handlerd); + multipath_wait_for_pg_init_completion(m); + + clear_bit(MPATHF_PG_INIT_DISABLED, &m->flags); + smp_mb__after_atomic(); + } - flush_workqueue(kmpath_handlerd); - multipath_wait_for_pg_init_completion(m); flush_workqueue(kmultipathd); flush_work(&m->trigger_event); - - clear_bit(MPATHF_PG_INIT_DISABLED, &m->flags); - smp_mb__after_atomic(); } static void multipath_dtr(struct dm_target *ti) @@ -1496,7 +1570,10 @@ static int multipath_end_io(struct dm_target *ti, struct request *clone, if (error && blk_path_error(error)) { struct multipath *m = ti->private; - r = DM_ENDIO_REQUEUE; + if (error == BLK_STS_RESOURCE) + r = DM_ENDIO_DELAY_REQUEUE; + else + r = DM_ENDIO_REQUEUE; if (pgpath) fail_path(pgpath); @@ -1521,7 +1598,7 @@ static int multipath_end_io(struct dm_target *ti, struct request *clone, } static int multipath_end_io_bio(struct dm_target *ti, struct bio *clone, - blk_status_t *error) + blk_status_t *error) { struct multipath *m = ti->private; struct dm_mpath_io *mpio = get_mpio_from_bio(clone); @@ -1546,9 +1623,6 @@ static int multipath_end_io_bio(struct dm_target *ti, struct bio *clone, goto done; } - /* Queue for the daemon to resubmit */ - dm_bio_restore(get_bio_details_from_bio(clone), clone); - spin_lock_irqsave(&m->lock, flags); bio_list_add(&m->queued_bios, clone); spin_unlock_irqrestore(&m->lock, flags); @@ -1656,6 +1730,9 @@ static void multipath_status(struct dm_target *ti, status_type_t type, case DM_TYPE_BIO_BASED: DMEMIT("queue_mode bio "); break; + case DM_TYPE_NVME_BIO_BASED: + DMEMIT("queue_mode nvme "); + break; case DM_TYPE_MQ_REQUEST_BASED: DMEMIT("queue_mode mq "); break; diff --git a/drivers/md/dm-queue-length.c b/drivers/md/dm-queue-length.c index 23f178641794..969c4f1a3633 100644 --- a/drivers/md/dm-queue-length.c +++ b/drivers/md/dm-queue-length.c @@ -195,9 +195,6 @@ static struct dm_path *ql_select_path(struct path_selector *ps, size_t nr_bytes) if (list_empty(&s->valid_paths)) goto out; - /* Change preferred (first in list) path to evenly balance. */ - list_move_tail(s->valid_paths.next, &s->valid_paths); - list_for_each_entry(pi, &s->valid_paths, list) { if (!best || (atomic_read(&pi->qlen) < atomic_read(&best->qlen))) @@ -210,6 +207,9 @@ static struct dm_path *ql_select_path(struct path_selector *ps, size_t nr_bytes) if (!best) goto out; + /* Move most recently used to least preferred to evenly balance. */ + list_move_tail(&best->list, &s->valid_paths); + ret = best->path; out: spin_unlock_irqrestore(&s->lock, flags); diff --git a/drivers/md/dm-raid.c b/drivers/md/dm-raid.c index e5ef0757fe23..7ef469e902c6 100644 --- a/drivers/md/dm-raid.c +++ b/drivers/md/dm-raid.c @@ -29,6 +29,9 @@ */ #define MIN_RAID456_JOURNAL_SPACE (4*2048) +/* Global list of all raid sets */ +static LIST_HEAD(raid_sets); + static bool devices_handle_discard_safely = false; /* @@ -105,8 +108,6 @@ struct raid_dev { #define CTR_FLAG_JOURNAL_DEV (1 << __CTR_FLAG_JOURNAL_DEV) #define CTR_FLAG_JOURNAL_MODE (1 << __CTR_FLAG_JOURNAL_MODE) -#define RESUME_STAY_FROZEN_FLAGS (CTR_FLAG_DELTA_DISKS | CTR_FLAG_DATA_OFFSET) - /* * Definitions of various constructor flags to * be used in checks of valid / invalid flags @@ -209,6 +210,8 @@ struct raid_dev { #define RT_FLAG_UPDATE_SBS 3 #define RT_FLAG_RESHAPE_RS 4 #define RT_FLAG_RS_SUSPENDED 5 +#define RT_FLAG_RS_IN_SYNC 6 +#define RT_FLAG_RS_RESYNCING 7 /* Array elements of 64 bit needed for rebuild/failed disk bits */ #define DISKS_ARRAY_ELEMS ((MAX_RAID_DEVICES + (sizeof(uint64_t) * 8 - 1)) / sizeof(uint64_t) / 8) @@ -224,8 +227,8 @@ struct rs_layout { struct raid_set { struct dm_target *ti; + struct list_head list; - uint32_t bitmap_loaded; uint32_t stripe_cache_entries; unsigned long ctr_flags; unsigned long runtime_flags; @@ -270,6 +273,19 @@ static void rs_config_restore(struct raid_set *rs, struct rs_layout *l) mddev->new_chunk_sectors = l->new_chunk_sectors; } +/* Find any raid_set in active slot for @rs on global list */ +static struct raid_set *rs_find_active(struct raid_set *rs) +{ + struct raid_set *r; + struct mapped_device *md = dm_table_get_md(rs->ti->table); + + list_for_each_entry(r, &raid_sets, list) + if (r != rs && dm_table_get_md(r->ti->table) == md) + return r; + + return NULL; +} + /* raid10 algorithms (i.e. formats) */ #define ALGORITHM_RAID10_DEFAULT 0 #define ALGORITHM_RAID10_NEAR 1 @@ -572,7 +588,7 @@ static const char *raid10_md_layout_to_format(int layout) } /* Return md raid10 algorithm for @name */ -static int raid10_name_to_format(const char *name) +static const int raid10_name_to_format(const char *name) { if (!strcasecmp(name, "near")) return ALGORITHM_RAID10_NEAR; @@ -675,15 +691,11 @@ static struct raid_type *get_raid_type_by_ll(const int level, const int layout) return NULL; } -/* - * Conditionally change bdev capacity of @rs - * in case of a disk add/remove reshape - */ -static void rs_set_capacity(struct raid_set *rs) +/* Adjust rdev sectors */ +static void rs_set_rdev_sectors(struct raid_set *rs) { struct mddev *mddev = &rs->md; struct md_rdev *rdev; - struct gendisk *gendisk = dm_disk(dm_table_get_md(rs->ti->table)); /* * raid10 sets rdev->sector to the device size, which @@ -692,8 +704,16 @@ static void rs_set_capacity(struct raid_set *rs) rdev_for_each(rdev, mddev) if (!test_bit(Journal, &rdev->flags)) rdev->sectors = mddev->dev_sectors; +} - set_capacity(gendisk, mddev->array_sectors); +/* + * Change bdev capacity of @rs in case of a disk add/remove reshape + */ +static void rs_set_capacity(struct raid_set *rs) +{ + struct gendisk *gendisk = dm_disk(dm_table_get_md(rs->ti->table)); + + set_capacity(gendisk, rs->md.array_sectors); revalidate_disk(gendisk); } @@ -744,6 +764,7 @@ static struct raid_set *raid_set_alloc(struct dm_target *ti, struct raid_type *r mddev_init(&rs->md); + INIT_LIST_HEAD(&rs->list); rs->raid_disks = raid_devs; rs->delta_disks = 0; @@ -761,6 +782,9 @@ static struct raid_set *raid_set_alloc(struct dm_target *ti, struct raid_type *r for (i = 0; i < raid_devs; i++) md_rdev_init(&rs->dev[i].rdev); + /* Add @rs to global list. */ + list_add(&rs->list, &raid_sets); + /* * Remaining items to be initialized by further RAID params: * rs->md.persistent @@ -773,6 +797,7 @@ static struct raid_set *raid_set_alloc(struct dm_target *ti, struct raid_type *r return rs; } +/* Free all @rs allocations and remove it from global list. */ static void raid_set_free(struct raid_set *rs) { int i; @@ -790,6 +815,8 @@ static void raid_set_free(struct raid_set *rs) dm_put_device(rs->ti, rs->dev[i].data_dev); } + list_del(&rs->list); + kfree(rs); } @@ -1002,7 +1029,7 @@ static int validate_raid_redundancy(struct raid_set *rs) !rs->dev[i].rdev.sb_page) rebuild_cnt++; - switch (rs->raid_type->level) { + switch (rs->md.level) { case 0: break; case 1: @@ -1017,6 +1044,11 @@ static int validate_raid_redundancy(struct raid_set *rs) break; case 10: copies = raid10_md_layout_to_copies(rs->md.new_layout); + if (copies < 2) { + DMERR("Bogus raid10 data copies < 2!"); + return -EINVAL; + } + if (rebuild_cnt < copies) break; @@ -1576,6 +1608,24 @@ static sector_t __rdev_sectors(struct raid_set *rs) return 0; } +/* Check that calculated dev_sectors fits all component devices. */ +static int _check_data_dev_sectors(struct raid_set *rs) +{ + sector_t ds = ~0; + struct md_rdev *rdev; + + rdev_for_each(rdev, &rs->md) + if (!test_bit(Journal, &rdev->flags) && rdev->bdev) { + ds = min(ds, to_sector(i_size_read(rdev->bdev->bd_inode))); + if (ds < rs->md.dev_sectors) { + rs->ti->error = "Component device(s) too small"; + return -EINVAL; + } + } + + return 0; +} + /* Calculate the sectors per device and per array used for @rs */ static int rs_set_dev_and_array_sectors(struct raid_set *rs, bool use_mddev) { @@ -1625,7 +1675,7 @@ static int rs_set_dev_and_array_sectors(struct raid_set *rs, bool use_mddev) mddev->array_sectors = array_sectors; mddev->dev_sectors = dev_sectors; - return 0; + return _check_data_dev_sectors(rs); bad: rs->ti->error = "Target length not divisible by number of data devices"; return -EINVAL; @@ -1674,8 +1724,11 @@ static void do_table_event(struct work_struct *ws) struct raid_set *rs = container_of(ws, struct raid_set, md.event_work); smp_rmb(); /* Make sure we access most actual mddev properties */ - if (!rs_is_reshaping(rs)) + if (!rs_is_reshaping(rs)) { + if (rs_is_raid10(rs)) + rs_set_rdev_sectors(rs); rs_set_capacity(rs); + } dm_table_event(rs->ti->table); } @@ -1860,7 +1913,7 @@ static bool rs_reshape_requested(struct raid_set *rs) if (rs_takeover_requested(rs)) return false; - if (!mddev->level) + if (rs_is_raid0(rs)) return false; change = mddev->new_layout != mddev->layout || @@ -1868,7 +1921,7 @@ static bool rs_reshape_requested(struct raid_set *rs) rs->delta_disks; /* Historical case to support raid1 reshape without delta disks */ - if (mddev->level == 1) { + if (rs_is_raid1(rs)) { if (rs->delta_disks) return !!rs->delta_disks; @@ -1876,7 +1929,7 @@ static bool rs_reshape_requested(struct raid_set *rs) mddev->raid_disks != rs->raid_disks; } - if (mddev->level == 10) + if (rs_is_raid10(rs)) return change && !__is_raid10_far(mddev->new_layout) && rs->delta_disks >= 0; @@ -2340,7 +2393,7 @@ static int super_init_validation(struct raid_set *rs, struct md_rdev *rdev) DMERR("new device%s provided without 'rebuild'", new_devs > 1 ? "s" : ""); return -EINVAL; - } else if (rs_is_recovering(rs)) { + } else if (!test_bit(__CTR_FLAG_REBUILD, &rs->ctr_flags) && rs_is_recovering(rs)) { DMERR("'rebuild' specified while raid set is not in-sync (recovery_cp=%llu)", (unsigned long long) mddev->recovery_cp); return -EINVAL; @@ -2640,12 +2693,19 @@ static int rs_adjust_data_offsets(struct raid_set *rs) * Make sure we got a minimum amount of free sectors per device */ if (rs->data_offset && - to_sector(i_size_read(rdev->bdev->bd_inode)) - rdev->sectors < MIN_FREE_RESHAPE_SPACE) { + to_sector(i_size_read(rdev->bdev->bd_inode)) - rs->md.dev_sectors < MIN_FREE_RESHAPE_SPACE) { rs->ti->error = data_offset ? "No space for forward reshape" : "No space for backward reshape"; return -ENOSPC; } out: + /* + * Raise recovery_cp in case data_offset != 0 to + * avoid false recovery positives in the constructor. + */ + if (rs->md.recovery_cp < rs->md.dev_sectors) + rs->md.recovery_cp += rs->dev[0].rdev.data_offset; + /* Adjust data offsets on all rdevs but on any raid4/5/6 journal device */ rdev_for_each(rdev, &rs->md) { if (!test_bit(Journal, &rdev->flags)) { @@ -2682,14 +2742,14 @@ static int rs_setup_takeover(struct raid_set *rs) sector_t new_data_offset = rs->dev[0].rdev.data_offset ? 0 : rs->data_offset; if (rt_is_raid10(rs->raid_type)) { - if (mddev->level == 0) { + if (rs_is_raid0(rs)) { /* Userpace reordered disks -> adjust raid_disk indexes */ __reorder_raid_disk_indexes(rs); /* raid0 -> raid10_far layout */ mddev->layout = raid10_format_to_md_layout(rs, ALGORITHM_RAID10_FAR, rs->raid10_copies); - } else if (mddev->level == 1) + } else if (rs_is_raid1(rs)) /* raid1 -> raid10_near layout */ mddev->layout = raid10_format_to_md_layout(rs, ALGORITHM_RAID10_NEAR, rs->raid_disks); @@ -2777,6 +2837,23 @@ static int rs_prepare_reshape(struct raid_set *rs) return 0; } +/* Get reshape sectors from data_offsets or raid set */ +static sector_t _get_reshape_sectors(struct raid_set *rs) +{ + struct md_rdev *rdev; + sector_t reshape_sectors = 0; + + rdev_for_each(rdev, &rs->md) + if (!test_bit(Journal, &rdev->flags)) { + reshape_sectors = (rdev->data_offset > rdev->new_data_offset) ? + rdev->data_offset - rdev->new_data_offset : + rdev->new_data_offset - rdev->data_offset; + break; + } + + return max(reshape_sectors, (sector_t) rs->data_offset); +} + /* * * - change raid layout @@ -2788,6 +2865,7 @@ static int rs_setup_reshape(struct raid_set *rs) { int r = 0; unsigned int cur_raid_devs, d; + sector_t reshape_sectors = _get_reshape_sectors(rs); struct mddev *mddev = &rs->md; struct md_rdev *rdev; @@ -2804,13 +2882,13 @@ static int rs_setup_reshape(struct raid_set *rs) /* * Adjust array size: * - * - in case of adding disks, array size has + * - in case of adding disk(s), array size has * to grow after the disk adding reshape, * which'll hapen in the event handler; * reshape will happen forward, so space has to * be available at the beginning of each disk * - * - in case of removing disks, array size + * - in case of removing disk(s), array size * has to shrink before starting the reshape, * which'll happen here; * reshape will happen backward, so space has to @@ -2841,7 +2919,7 @@ static int rs_setup_reshape(struct raid_set *rs) rdev->recovery_offset = rs_is_raid1(rs) ? 0 : MaxSector; } - mddev->reshape_backwards = 0; /* adding disks -> forward reshape */ + mddev->reshape_backwards = 0; /* adding disk(s) -> forward reshape */ /* Remove disk(s) */ } else if (rs->delta_disks < 0) { @@ -2874,6 +2952,15 @@ static int rs_setup_reshape(struct raid_set *rs) mddev->reshape_backwards = rs->dev[0].rdev.data_offset ? 0 : 1; } + /* + * Adjust device size for forward reshape + * because md_finish_reshape() reduces it. + */ + if (!mddev->reshape_backwards) + rdev_for_each(rdev, &rs->md) + if (!test_bit(Journal, &rdev->flags)) + rdev->sectors += reshape_sectors; + return r; } @@ -2890,7 +2977,7 @@ static void configure_discard_support(struct raid_set *rs) /* * XXX: RAID level 4,5,6 require zeroing for safety. */ - raid456 = (rs->md.level == 4 || rs->md.level == 5 || rs->md.level == 6); + raid456 = rs_is_raid456(rs); for (i = 0; i < rs->raid_disks; i++) { struct request_queue *q; @@ -2915,7 +3002,7 @@ static void configure_discard_support(struct raid_set *rs) * RAID1 and RAID10 personalities require bio splitting, * RAID0/4/5/6 don't and process large discard bios properly. */ - ti->split_discard_bios = !!(rs->md.level == 1 || rs->md.level == 10); + ti->split_discard_bios = !!(rs_is_raid1(rs) || rs_is_raid10(rs)); ti->num_discard_bios = 1; } @@ -2935,10 +3022,10 @@ static void configure_discard_support(struct raid_set *rs) static int raid_ctr(struct dm_target *ti, unsigned int argc, char **argv) { int r; - bool resize; + bool resize = false; struct raid_type *rt; unsigned int num_raid_params, num_raid_devs; - sector_t calculated_dev_sectors, rdev_sectors; + sector_t calculated_dev_sectors, rdev_sectors, reshape_sectors; struct raid_set *rs = NULL; const char *arg; struct rs_layout rs_layout; @@ -3021,7 +3108,10 @@ static int raid_ctr(struct dm_target *ti, unsigned int argc, char **argv) goto bad; } - resize = calculated_dev_sectors != rdev_sectors; + + reshape_sectors = _get_reshape_sectors(rs); + if (calculated_dev_sectors != rdev_sectors) + resize = calculated_dev_sectors != (reshape_sectors ? rdev_sectors - reshape_sectors : rdev_sectors); INIT_WORK(&rs->md.event_work, do_table_event); ti->private = rs; @@ -3105,19 +3195,22 @@ static int raid_ctr(struct dm_target *ti, unsigned int argc, char **argv) goto bad; } - /* - * We can only prepare for a reshape here, because the - * raid set needs to run to provide the repective reshape - * check functions via its MD personality instance. - * - * So do the reshape check after md_run() succeeded. - */ - r = rs_prepare_reshape(rs); - if (r) - return r; + /* Out-of-place space has to be available to allow for a reshape unless raid1! */ + if (reshape_sectors || rs_is_raid1(rs)) { + /* + * We can only prepare for a reshape here, because the + * raid set needs to run to provide the repective reshape + * check functions via its MD personality instance. + * + * So do the reshape check after md_run() succeeded. + */ + r = rs_prepare_reshape(rs); + if (r) + return r; - /* Reshaping ain't recovery, so disable recovery */ - rs_setup_recovery(rs, MaxSector); + /* Reshaping ain't recovery, so disable recovery */ + rs_setup_recovery(rs, MaxSector); + } rs_set_cur(rs); } else { /* May not set recovery when a device rebuild is requested */ @@ -3144,7 +3237,6 @@ static int raid_ctr(struct dm_target *ti, unsigned int argc, char **argv) mddev_lock_nointr(&rs->md); r = md_run(&rs->md); rs->md.in_sync = 0; /* Assume already marked dirty */ - if (r) { ti->error = "Failed to run raid array"; mddev_unlock(&rs->md); @@ -3248,25 +3340,27 @@ static int raid_map(struct dm_target *ti, struct bio *bio) } /* Return string describing the current sync action of @mddev */ -static const char *decipher_sync_action(struct mddev *mddev) +static const char *decipher_sync_action(struct mddev *mddev, unsigned long recovery) { - if (test_bit(MD_RECOVERY_FROZEN, &mddev->recovery)) + if (test_bit(MD_RECOVERY_FROZEN, &recovery)) return "frozen"; - if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) || - (!mddev->ro && test_bit(MD_RECOVERY_NEEDED, &mddev->recovery))) { - if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)) + /* The MD sync thread can be done with io but still be running */ + if (!test_bit(MD_RECOVERY_DONE, &recovery) && + (test_bit(MD_RECOVERY_RUNNING, &recovery) || + (!mddev->ro && test_bit(MD_RECOVERY_NEEDED, &recovery)))) { + if (test_bit(MD_RECOVERY_RESHAPE, &recovery)) return "reshape"; - if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) { - if (!test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) + if (test_bit(MD_RECOVERY_SYNC, &recovery)) { + if (!test_bit(MD_RECOVERY_REQUESTED, &recovery)) return "resync"; - else if (test_bit(MD_RECOVERY_CHECK, &mddev->recovery)) + else if (test_bit(MD_RECOVERY_CHECK, &recovery)) return "check"; return "repair"; } - if (test_bit(MD_RECOVERY_RECOVER, &mddev->recovery)) + if (test_bit(MD_RECOVERY_RECOVER, &recovery)) return "recover"; } @@ -3283,7 +3377,7 @@ static const char *decipher_sync_action(struct mddev *mddev) * 'A' = Alive and in-sync raid set component _or_ alive raid4/5/6 'write_through' journal device * '-' = Non-existing device (i.e. uspace passed '- -' into the ctr) */ -static const char *__raid_dev_status(struct raid_set *rs, struct md_rdev *rdev, bool array_in_sync) +static const char *__raid_dev_status(struct raid_set *rs, struct md_rdev *rdev) { if (!rdev->bdev) return "-"; @@ -3291,85 +3385,108 @@ static const char *__raid_dev_status(struct raid_set *rs, struct md_rdev *rdev, return "D"; else if (test_bit(Journal, &rdev->flags)) return (rs->journal_dev.mode == R5C_JOURNAL_MODE_WRITE_THROUGH) ? "A" : "a"; - else if (!array_in_sync || !test_bit(In_sync, &rdev->flags)) + else if (test_bit(RT_FLAG_RS_RESYNCING, &rs->runtime_flags) || + (!test_bit(RT_FLAG_RS_IN_SYNC, &rs->runtime_flags) && + !test_bit(In_sync, &rdev->flags))) return "a"; else return "A"; } -/* Helper to return resync/reshape progress for @rs and @array_in_sync */ -static sector_t rs_get_progress(struct raid_set *rs, - sector_t resync_max_sectors, bool *array_in_sync) +/* Helper to return resync/reshape progress for @rs and runtime flags for raid set in sync / resynching */ +static sector_t rs_get_progress(struct raid_set *rs, unsigned long recovery, + sector_t resync_max_sectors) { - sector_t r, curr_resync_completed; + sector_t r; struct mddev *mddev = &rs->md; - curr_resync_completed = mddev->curr_resync_completed ?: mddev->recovery_cp; - *array_in_sync = false; + clear_bit(RT_FLAG_RS_IN_SYNC, &rs->runtime_flags); + clear_bit(RT_FLAG_RS_RESYNCING, &rs->runtime_flags); if (rs_is_raid0(rs)) { r = resync_max_sectors; - *array_in_sync = true; + set_bit(RT_FLAG_RS_IN_SYNC, &rs->runtime_flags); } else { - r = mddev->reshape_position; - - /* Reshape is relative to the array size */ - if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) || - r != MaxSector) { - if (r == MaxSector) { - *array_in_sync = true; - r = resync_max_sectors; - } else { - /* Got to reverse on backward reshape */ - if (mddev->reshape_backwards) - r = mddev->array_sectors - r; - - /* Devide by # of data stripes */ - sector_div(r, mddev_data_stripes(rs)); - } - - /* Sync is relative to the component device size */ - } else if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) - r = curr_resync_completed; + if (test_bit(MD_RECOVERY_NEEDED, &recovery) || + test_bit(MD_RECOVERY_RESHAPE, &recovery) || + test_bit(MD_RECOVERY_RUNNING, &recovery)) + r = mddev->curr_resync_completed; else r = mddev->recovery_cp; - if ((r == MaxSector) || - (test_bit(MD_RECOVERY_DONE, &mddev->recovery) && - (mddev->curr_resync_completed == resync_max_sectors))) { + if (r >= resync_max_sectors && + (!test_bit(MD_RECOVERY_REQUESTED, &recovery) || + (!test_bit(MD_RECOVERY_FROZEN, &recovery) && + !test_bit(MD_RECOVERY_NEEDED, &recovery) && + !test_bit(MD_RECOVERY_RUNNING, &recovery)))) { /* * Sync complete. */ - *array_in_sync = true; - r = resync_max_sectors; - } else if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) { + /* In case we have finished recovering, the array is in sync. */ + if (test_bit(MD_RECOVERY_RECOVER, &recovery)) + set_bit(RT_FLAG_RS_IN_SYNC, &rs->runtime_flags); + + } else if (test_bit(MD_RECOVERY_RECOVER, &recovery)) { + /* + * In case we are recovering, the array is not in sync + * and health chars should show the recovering legs. + */ + ; + + } else if (test_bit(MD_RECOVERY_SYNC, &recovery) && + !test_bit(MD_RECOVERY_REQUESTED, &recovery)) { + /* + * If "resync" is occurring, the raid set + * is or may be out of sync hence the health + * characters shall be 'a'. + */ + set_bit(RT_FLAG_RS_RESYNCING, &rs->runtime_flags); + + } else if (test_bit(MD_RECOVERY_RESHAPE, &recovery) && + !test_bit(MD_RECOVERY_REQUESTED, &recovery)) { + /* + * If "reshape" is occurring, the raid set + * is or may be out of sync hence the health + * characters shall be 'a'. + */ + set_bit(RT_FLAG_RS_RESYNCING, &rs->runtime_flags); + + } else if (test_bit(MD_RECOVERY_REQUESTED, &recovery)) { /* * If "check" or "repair" is occurring, the raid set has * undergone an initial sync and the health characters * should not be 'a' anymore. */ - *array_in_sync = true; + set_bit(RT_FLAG_RS_IN_SYNC, &rs->runtime_flags); + } else { struct md_rdev *rdev; /* + * We are idle and recovery is needed, prevent 'A' chars race + * caused by components still set to in-sync by constrcuctor. + */ + if (test_bit(MD_RECOVERY_NEEDED, &recovery)) + set_bit(RT_FLAG_RS_RESYNCING, &rs->runtime_flags); + + /* * The raid set may be doing an initial sync, or it may * be rebuilding individual components. If all the * devices are In_sync, then it is the raid set that is * being initialized. */ + set_bit(RT_FLAG_RS_IN_SYNC, &rs->runtime_flags); rdev_for_each(rdev, mddev) if (!test_bit(Journal, &rdev->flags) && - !test_bit(In_sync, &rdev->flags)) - *array_in_sync = true; -#if 0 - r = 0; /* HM FIXME: TESTME: https://bugzilla.redhat.com/show_bug.cgi?id=1210637 ? */ -#endif + !test_bit(In_sync, &rdev->flags)) { + clear_bit(RT_FLAG_RS_IN_SYNC, &rs->runtime_flags); + break; + } } } - return r; + return min(r, resync_max_sectors); } /* Helper to return @dev name or "-" if !@dev */ @@ -3385,7 +3502,7 @@ static void raid_status(struct dm_target *ti, status_type_t type, struct mddev *mddev = &rs->md; struct r5conf *conf = mddev->private; int i, max_nr_stripes = conf ? conf->max_nr_stripes : 0; - bool array_in_sync; + unsigned long recovery; unsigned int raid_param_cnt = 1; /* at least 1 for chunksize */ unsigned int sz = 0; unsigned int rebuild_disks; @@ -3405,17 +3522,18 @@ static void raid_status(struct dm_target *ti, status_type_t type, /* Access most recent mddev properties for status output */ smp_rmb(); + recovery = rs->md.recovery; /* Get sensible max sectors even if raid set not yet started */ resync_max_sectors = test_bit(RT_FLAG_RS_PRERESUMED, &rs->runtime_flags) ? mddev->resync_max_sectors : mddev->dev_sectors; - progress = rs_get_progress(rs, resync_max_sectors, &array_in_sync); + progress = rs_get_progress(rs, recovery, resync_max_sectors); resync_mismatches = (mddev->last_sync_action && !strcasecmp(mddev->last_sync_action, "check")) ? atomic64_read(&mddev->resync_mismatches) : 0; - sync_action = decipher_sync_action(&rs->md); + sync_action = decipher_sync_action(&rs->md, recovery); /* HM FIXME: do we want another state char for raid0? It shows 'D'/'A'/'-' now */ for (i = 0; i < rs->raid_disks; i++) - DMEMIT(__raid_dev_status(rs, &rs->dev[i].rdev, array_in_sync)); + DMEMIT(__raid_dev_status(rs, &rs->dev[i].rdev)); /* * In-sync/Reshape ratio: @@ -3466,7 +3584,7 @@ static void raid_status(struct dm_target *ti, status_type_t type, * v1.10.0+: */ DMEMIT(" %s", test_bit(__CTR_FLAG_JOURNAL_DEV, &rs->ctr_flags) ? - __raid_dev_status(rs, &rs->journal_dev.rdev, 0) : "-"); + __raid_dev_status(rs, &rs->journal_dev.rdev) : "-"); break; case STATUSTYPE_TABLE: @@ -3622,24 +3740,19 @@ static void raid_io_hints(struct dm_target *ti, struct queue_limits *limits) blk_limits_io_opt(limits, chunk_size * mddev_data_stripes(rs)); } -static void raid_presuspend(struct dm_target *ti) -{ - struct raid_set *rs = ti->private; - - md_stop_writes(&rs->md); -} - static void raid_postsuspend(struct dm_target *ti) { struct raid_set *rs = ti->private; if (!test_and_set_bit(RT_FLAG_RS_SUSPENDED, &rs->runtime_flags)) { + /* Writes have to be stopped before suspending to avoid deadlocks. */ + if (!test_bit(MD_RECOVERY_FROZEN, &rs->md.recovery)) + md_stop_writes(&rs->md); + mddev_lock_nointr(&rs->md); mddev_suspend(&rs->md); mddev_unlock(&rs->md); } - - rs->md.ro = 1; } static void attempt_restore_of_faulty_devices(struct raid_set *rs) @@ -3816,10 +3929,33 @@ static int raid_preresume(struct dm_target *ti) struct raid_set *rs = ti->private; struct mddev *mddev = &rs->md; - /* This is a resume after a suspend of the set -> it's already started */ + /* This is a resume after a suspend of the set -> it's already started. */ if (test_and_set_bit(RT_FLAG_RS_PRERESUMED, &rs->runtime_flags)) return 0; + if (!test_bit(__CTR_FLAG_REBUILD, &rs->ctr_flags)) { + struct raid_set *rs_active = rs_find_active(rs); + + if (rs_active) { + /* + * In case no rebuilds have been requested + * and an active table slot exists, copy + * current resynchonization completed and + * reshape position pointers across from + * suspended raid set in the active slot. + * + * This resumes the new mapping at current + * offsets to continue recover/reshape without + * necessarily redoing a raid set partially or + * causing data corruption in case of a reshape. + */ + if (rs_active->md.curr_resync_completed != MaxSector) + mddev->curr_resync_completed = rs_active->md.curr_resync_completed; + if (rs_active->md.reshape_position != MaxSector) + mddev->reshape_position = rs_active->md.reshape_position; + } + } + /* * The superblocks need to be updated on disk if the * array is new or new devices got added (thus zeroed @@ -3851,11 +3987,10 @@ static int raid_preresume(struct dm_target *ti) mddev->resync_min = mddev->recovery_cp; } - rs_set_capacity(rs); - /* Check for any reshape request unless new raid set */ - if (test_and_clear_bit(RT_FLAG_RESHAPE_RS, &rs->runtime_flags)) { + if (test_bit(RT_FLAG_RESHAPE_RS, &rs->runtime_flags)) { /* Initiate a reshape. */ + rs_set_rdev_sectors(rs); mddev_lock_nointr(mddev); r = rs_start_reshape(rs); mddev_unlock(mddev); @@ -3881,21 +4016,15 @@ static void raid_resume(struct dm_target *ti) attempt_restore_of_faulty_devices(rs); } - mddev->ro = 0; - mddev->in_sync = 0; - - /* - * Keep the RAID set frozen if reshape/rebuild flags are set. - * The RAID set is unfrozen once the next table load/resume, - * which clears the reshape/rebuild flags, occurs. - * This ensures that the constructor for the inactive table - * retrieves an up-to-date reshape_position. - */ - if (!(rs->ctr_flags & RESUME_STAY_FROZEN_FLAGS)) - clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); - if (test_and_clear_bit(RT_FLAG_RS_SUSPENDED, &rs->runtime_flags)) { + /* Only reduce raid set size before running a disk removing reshape. */ + if (mddev->delta_disks < 0) + rs_set_capacity(rs); + mddev_lock_nointr(mddev); + clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); + mddev->ro = 0; + mddev->in_sync = 0; mddev_resume(mddev); mddev_unlock(mddev); } @@ -3903,7 +4032,7 @@ static void raid_resume(struct dm_target *ti) static struct target_type raid_target = { .name = "raid", - .version = {1, 13, 0}, + .version = {1, 13, 2}, .module = THIS_MODULE, .ctr = raid_ctr, .dtr = raid_dtr, @@ -3912,7 +4041,6 @@ static struct target_type raid_target = { .message = raid_message, .iterate_devices = raid_iterate_devices, .io_hints = raid_io_hints, - .presuspend = raid_presuspend, .postsuspend = raid_postsuspend, .preresume = raid_preresume, .resume = raid_resume, diff --git a/drivers/md/dm-rq.c b/drivers/md/dm-rq.c index b7d175e94a02..aeaaaef43eff 100644 --- a/drivers/md/dm-rq.c +++ b/drivers/md/dm-rq.c @@ -315,6 +315,10 @@ static void dm_done(struct request *clone, blk_status_t error, bool mapped) /* The target wants to requeue the I/O */ dm_requeue_original_request(tio, false); break; + case DM_ENDIO_DELAY_REQUEUE: + /* The target wants to requeue the I/O after a delay */ + dm_requeue_original_request(tio, true); + break; default: DMWARN("unimplemented target endio return value: %d", r); BUG(); @@ -713,7 +717,6 @@ int dm_old_init_request_queue(struct mapped_device *md, struct dm_table *t) /* disable dm_old_request_fn's merge heuristic by default */ md->seq_rq_merge_deadline_usecs = 0; - dm_init_normal_md_queue(md); blk_queue_softirq_done(md->queue, dm_softirq_done); /* Initialize the request-based DM worker thread */ @@ -821,7 +824,6 @@ int dm_mq_init_request_queue(struct mapped_device *md, struct dm_table *t) err = PTR_ERR(q); goto out_tag_set; } - dm_init_md_queue(md); return 0; diff --git a/drivers/md/dm-service-time.c b/drivers/md/dm-service-time.c index 7b8642045c55..f006a9005593 100644 --- a/drivers/md/dm-service-time.c +++ b/drivers/md/dm-service-time.c @@ -282,9 +282,6 @@ static struct dm_path *st_select_path(struct path_selector *ps, size_t nr_bytes) if (list_empty(&s->valid_paths)) goto out; - /* Change preferred (first in list) path to evenly balance. */ - list_move_tail(s->valid_paths.next, &s->valid_paths); - list_for_each_entry(pi, &s->valid_paths, list) if (!best || (st_compare_load(pi, best, nr_bytes) < 0)) best = pi; @@ -292,6 +289,9 @@ static struct dm_path *st_select_path(struct path_selector *ps, size_t nr_bytes) if (!best) goto out; + /* Move most recently used to least preferred to evenly balance. */ + list_move_tail(&best->list, &s->valid_paths); + ret = best->path; out: spin_unlock_irqrestore(&s->lock, flags); diff --git a/drivers/md/dm-snap.c b/drivers/md/dm-snap.c index a0613bd8ed00..216035be5661 100644 --- a/drivers/md/dm-snap.c +++ b/drivers/md/dm-snap.c @@ -47,7 +47,7 @@ struct dm_exception_table { }; struct dm_snapshot { - struct rw_semaphore lock; + struct mutex lock; struct dm_dev *origin; struct dm_dev *cow; @@ -439,9 +439,9 @@ static int __find_snapshots_sharing_cow(struct dm_snapshot *snap, if (!bdev_equal(s->cow->bdev, snap->cow->bdev)) continue; - down_read(&s->lock); + mutex_lock(&s->lock); active = s->active; - up_read(&s->lock); + mutex_unlock(&s->lock); if (active) { if (snap_src) @@ -909,7 +909,7 @@ static int remove_single_exception_chunk(struct dm_snapshot *s) int r; chunk_t old_chunk = s->first_merging_chunk + s->num_merging_chunks - 1; - down_write(&s->lock); + mutex_lock(&s->lock); /* * Process chunks (and associated exceptions) in reverse order @@ -924,7 +924,7 @@ static int remove_single_exception_chunk(struct dm_snapshot *s) b = __release_queued_bios_after_merge(s); out: - up_write(&s->lock); + mutex_unlock(&s->lock); if (b) flush_bios(b); @@ -983,9 +983,9 @@ static void snapshot_merge_next_chunks(struct dm_snapshot *s) if (linear_chunks < 0) { DMERR("Read error in exception store: " "shutting down merge"); - down_write(&s->lock); + mutex_lock(&s->lock); s->merge_failed = 1; - up_write(&s->lock); + mutex_unlock(&s->lock); } goto shut; } @@ -1026,10 +1026,10 @@ static void snapshot_merge_next_chunks(struct dm_snapshot *s) previous_count = read_pending_exceptions_done_count(); } - down_write(&s->lock); + mutex_lock(&s->lock); s->first_merging_chunk = old_chunk; s->num_merging_chunks = linear_chunks; - up_write(&s->lock); + mutex_unlock(&s->lock); /* Wait until writes to all 'linear_chunks' drain */ for (i = 0; i < linear_chunks; i++) @@ -1071,10 +1071,10 @@ static void merge_callback(int read_err, unsigned long write_err, void *context) return; shut: - down_write(&s->lock); + mutex_lock(&s->lock); s->merge_failed = 1; b = __release_queued_bios_after_merge(s); - up_write(&s->lock); + mutex_unlock(&s->lock); error_bios(b); merge_shutdown(s); @@ -1173,7 +1173,7 @@ static int snapshot_ctr(struct dm_target *ti, unsigned int argc, char **argv) s->exception_start_sequence = 0; s->exception_complete_sequence = 0; INIT_LIST_HEAD(&s->out_of_order_list); - init_rwsem(&s->lock); + mutex_init(&s->lock); INIT_LIST_HEAD(&s->list); spin_lock_init(&s->pe_lock); s->state_bits = 0; @@ -1338,9 +1338,9 @@ static void snapshot_dtr(struct dm_target *ti) /* Check whether exception handover must be cancelled */ (void) __find_snapshots_sharing_cow(s, &snap_src, &snap_dest, NULL); if (snap_src && snap_dest && (s == snap_src)) { - down_write(&snap_dest->lock); + mutex_lock(&snap_dest->lock); snap_dest->valid = 0; - up_write(&snap_dest->lock); + mutex_unlock(&snap_dest->lock); DMERR("Cancelling snapshot handover."); } up_read(&_origins_lock); @@ -1371,6 +1371,8 @@ static void snapshot_dtr(struct dm_target *ti) dm_exception_store_destroy(s->store); + mutex_destroy(&s->lock); + dm_put_device(ti, s->cow); dm_put_device(ti, s->origin); @@ -1458,7 +1460,7 @@ static void pending_complete(void *context, int success) if (!success) { /* Read/write error - snapshot is unusable */ - down_write(&s->lock); + mutex_lock(&s->lock); __invalidate_snapshot(s, -EIO); error = 1; goto out; @@ -1466,14 +1468,14 @@ static void pending_complete(void *context, int success) e = alloc_completed_exception(GFP_NOIO); if (!e) { - down_write(&s->lock); + mutex_lock(&s->lock); __invalidate_snapshot(s, -ENOMEM); error = 1; goto out; } *e = pe->e; - down_write(&s->lock); + mutex_lock(&s->lock); if (!s->valid) { free_completed_exception(e); error = 1; @@ -1498,7 +1500,7 @@ out: full_bio->bi_end_io = pe->full_bio_end_io; increment_pending_exceptions_done_count(); - up_write(&s->lock); + mutex_unlock(&s->lock); /* Submit any pending write bios */ if (error) { @@ -1694,7 +1696,7 @@ static int snapshot_map(struct dm_target *ti, struct bio *bio) /* FIXME: should only take write lock if we need * to copy an exception */ - down_write(&s->lock); + mutex_lock(&s->lock); if (!s->valid || (unlikely(s->snapshot_overflowed) && bio_data_dir(bio) == WRITE)) { @@ -1717,9 +1719,9 @@ static int snapshot_map(struct dm_target *ti, struct bio *bio) if (bio_data_dir(bio) == WRITE) { pe = __lookup_pending_exception(s, chunk); if (!pe) { - up_write(&s->lock); + mutex_unlock(&s->lock); pe = alloc_pending_exception(s); - down_write(&s->lock); + mutex_lock(&s->lock); if (!s->valid || s->snapshot_overflowed) { free_pending_exception(pe); @@ -1754,7 +1756,7 @@ static int snapshot_map(struct dm_target *ti, struct bio *bio) bio->bi_iter.bi_size == (s->store->chunk_size << SECTOR_SHIFT)) { pe->started = 1; - up_write(&s->lock); + mutex_unlock(&s->lock); start_full_bio(pe, bio); goto out; } @@ -1764,7 +1766,7 @@ static int snapshot_map(struct dm_target *ti, struct bio *bio) if (!pe->started) { /* this is protected by snap->lock */ pe->started = 1; - up_write(&s->lock); + mutex_unlock(&s->lock); start_copy(pe); goto out; } @@ -1774,7 +1776,7 @@ static int snapshot_map(struct dm_target *ti, struct bio *bio) } out_unlock: - up_write(&s->lock); + mutex_unlock(&s->lock); out: return r; } @@ -1810,7 +1812,7 @@ static int snapshot_merge_map(struct dm_target *ti, struct bio *bio) chunk = sector_to_chunk(s->store, bio->bi_iter.bi_sector); - down_write(&s->lock); + mutex_lock(&s->lock); /* Full merging snapshots are redirected to the origin */ if (!s->valid) @@ -1841,12 +1843,12 @@ redirect_to_origin: bio_set_dev(bio, s->origin->bdev); if (bio_data_dir(bio) == WRITE) { - up_write(&s->lock); + mutex_unlock(&s->lock); return do_origin(s->origin, bio); } out_unlock: - up_write(&s->lock); + mutex_unlock(&s->lock); return r; } @@ -1878,7 +1880,7 @@ static int snapshot_preresume(struct dm_target *ti) down_read(&_origins_lock); (void) __find_snapshots_sharing_cow(s, &snap_src, &snap_dest, NULL); if (snap_src && snap_dest) { - down_read(&snap_src->lock); + mutex_lock(&snap_src->lock); if (s == snap_src) { DMERR("Unable to resume snapshot source until " "handover completes."); @@ -1888,7 +1890,7 @@ static int snapshot_preresume(struct dm_target *ti) "source is suspended."); r = -EINVAL; } - up_read(&snap_src->lock); + mutex_unlock(&snap_src->lock); } up_read(&_origins_lock); @@ -1934,11 +1936,11 @@ static void snapshot_resume(struct dm_target *ti) (void) __find_snapshots_sharing_cow(s, &snap_src, &snap_dest, NULL); if (snap_src && snap_dest) { - down_write(&snap_src->lock); - down_write_nested(&snap_dest->lock, SINGLE_DEPTH_NESTING); + mutex_lock(&snap_src->lock); + mutex_lock_nested(&snap_dest->lock, SINGLE_DEPTH_NESTING); __handover_exceptions(snap_src, snap_dest); - up_write(&snap_dest->lock); - up_write(&snap_src->lock); + mutex_unlock(&snap_dest->lock); + mutex_unlock(&snap_src->lock); } up_read(&_origins_lock); @@ -1953,9 +1955,9 @@ static void snapshot_resume(struct dm_target *ti) /* Now we have correct chunk size, reregister */ reregister_snapshot(s); - down_write(&s->lock); + mutex_lock(&s->lock); s->active = 1; - up_write(&s->lock); + mutex_unlock(&s->lock); } static uint32_t get_origin_minimum_chunksize(struct block_device *bdev) @@ -1995,7 +1997,7 @@ static void snapshot_status(struct dm_target *ti, status_type_t type, switch (type) { case STATUSTYPE_INFO: - down_write(&snap->lock); + mutex_lock(&snap->lock); if (!snap->valid) DMEMIT("Invalid"); @@ -2020,7 +2022,7 @@ static void snapshot_status(struct dm_target *ti, status_type_t type, DMEMIT("Unknown"); } - up_write(&snap->lock); + mutex_unlock(&snap->lock); break; @@ -2086,7 +2088,7 @@ static int __origin_write(struct list_head *snapshots, sector_t sector, if (dm_target_is_snapshot_merge(snap->ti)) continue; - down_write(&snap->lock); + mutex_lock(&snap->lock); /* Only deal with valid and active snapshots */ if (!snap->valid || !snap->active) @@ -2113,9 +2115,9 @@ static int __origin_write(struct list_head *snapshots, sector_t sector, pe = __lookup_pending_exception(snap, chunk); if (!pe) { - up_write(&snap->lock); + mutex_unlock(&snap->lock); pe = alloc_pending_exception(snap); - down_write(&snap->lock); + mutex_lock(&snap->lock); if (!snap->valid) { free_pending_exception(pe); @@ -2158,7 +2160,7 @@ static int __origin_write(struct list_head *snapshots, sector_t sector, } next_snapshot: - up_write(&snap->lock); + mutex_unlock(&snap->lock); if (pe_to_start_now) { start_copy(pe_to_start_now); diff --git a/drivers/md/dm-stats.c b/drivers/md/dm-stats.c index 29bc51084c82..56059fb56e2d 100644 --- a/drivers/md/dm-stats.c +++ b/drivers/md/dm-stats.c @@ -228,6 +228,7 @@ void dm_stats_cleanup(struct dm_stats *stats) dm_stat_free(&s->rcu_head); } free_percpu(stats->last); + mutex_destroy(&stats->mutex); } static int dm_stats_create(struct dm_stats *stats, sector_t start, sector_t end, diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c index aaffd0c0ee9a..5fe7ec356c33 100644 --- a/drivers/md/dm-table.c +++ b/drivers/md/dm-table.c @@ -866,7 +866,8 @@ EXPORT_SYMBOL(dm_consume_args); static bool __table_type_bio_based(enum dm_queue_mode table_type) { return (table_type == DM_TYPE_BIO_BASED || - table_type == DM_TYPE_DAX_BIO_BASED); + table_type == DM_TYPE_DAX_BIO_BASED || + table_type == DM_TYPE_NVME_BIO_BASED); } static bool __table_type_request_based(enum dm_queue_mode table_type) @@ -909,13 +910,33 @@ static bool dm_table_supports_dax(struct dm_table *t) return true; } +static bool dm_table_does_not_support_partial_completion(struct dm_table *t); + +struct verify_rq_based_data { + unsigned sq_count; + unsigned mq_count; +}; + +static int device_is_rq_based(struct dm_target *ti, struct dm_dev *dev, + sector_t start, sector_t len, void *data) +{ + struct request_queue *q = bdev_get_queue(dev->bdev); + struct verify_rq_based_data *v = data; + + if (q->mq_ops) + v->mq_count++; + else + v->sq_count++; + + return queue_is_rq_based(q); +} + static int dm_table_determine_type(struct dm_table *t) { unsigned i; unsigned bio_based = 0, request_based = 0, hybrid = 0; - unsigned sq_count = 0, mq_count = 0; + struct verify_rq_based_data v = {.sq_count = 0, .mq_count = 0}; struct dm_target *tgt; - struct dm_dev_internal *dd; struct list_head *devices = dm_table_get_devices(t); enum dm_queue_mode live_md_type = dm_get_md_type(t->md); @@ -923,6 +944,14 @@ static int dm_table_determine_type(struct dm_table *t) /* target already set the table's type */ if (t->type == DM_TYPE_BIO_BASED) return 0; + else if (t->type == DM_TYPE_NVME_BIO_BASED) { + if (!dm_table_does_not_support_partial_completion(t)) { + DMERR("nvme bio-based is only possible with devices" + " that don't support partial completion"); + return -EINVAL; + } + /* Fallthru, also verify all devices are blk-mq */ + } BUG_ON(t->type == DM_TYPE_DAX_BIO_BASED); goto verify_rq_based; } @@ -937,8 +966,8 @@ static int dm_table_determine_type(struct dm_table *t) bio_based = 1; if (bio_based && request_based) { - DMWARN("Inconsistent table: different target types" - " can't be mixed up"); + DMERR("Inconsistent table: different target types" + " can't be mixed up"); return -EINVAL; } } @@ -959,8 +988,18 @@ static int dm_table_determine_type(struct dm_table *t) /* We must use this table as bio-based */ t->type = DM_TYPE_BIO_BASED; if (dm_table_supports_dax(t) || - (list_empty(devices) && live_md_type == DM_TYPE_DAX_BIO_BASED)) + (list_empty(devices) && live_md_type == DM_TYPE_DAX_BIO_BASED)) { t->type = DM_TYPE_DAX_BIO_BASED; + } else { + /* Check if upgrading to NVMe bio-based is valid or required */ + tgt = dm_table_get_immutable_target(t); + if (tgt && !tgt->max_io_len && dm_table_does_not_support_partial_completion(t)) { + t->type = DM_TYPE_NVME_BIO_BASED; + goto verify_rq_based; /* must be stacked directly on NVMe (blk-mq) */ + } else if (list_empty(devices) && live_md_type == DM_TYPE_NVME_BIO_BASED) { + t->type = DM_TYPE_NVME_BIO_BASED; + } + } return 0; } @@ -980,7 +1019,8 @@ verify_rq_based: * (e.g. request completion process for partial completion.) */ if (t->num_targets > 1) { - DMWARN("Request-based dm doesn't support multiple targets yet"); + DMERR("%s DM doesn't support multiple targets", + t->type == DM_TYPE_NVME_BIO_BASED ? "nvme bio-based" : "request-based"); return -EINVAL; } @@ -997,28 +1037,29 @@ verify_rq_based: return 0; } - /* Non-request-stackable devices can't be used for request-based dm */ - list_for_each_entry(dd, devices, list) { - struct request_queue *q = bdev_get_queue(dd->dm_dev->bdev); - - if (!queue_is_rq_based(q)) { - DMERR("table load rejected: including" - " non-request-stackable devices"); - return -EINVAL; - } + tgt = dm_table_get_immutable_target(t); + if (!tgt) { + DMERR("table load rejected: immutable target is required"); + return -EINVAL; + } else if (tgt->max_io_len) { + DMERR("table load rejected: immutable target that splits IO is not supported"); + return -EINVAL; + } - if (q->mq_ops) - mq_count++; - else - sq_count++; + /* Non-request-stackable devices can't be used for request-based dm */ + if (!tgt->type->iterate_devices || + !tgt->type->iterate_devices(tgt, device_is_rq_based, &v)) { + DMERR("table load rejected: including non-request-stackable devices"); + return -EINVAL; } - if (sq_count && mq_count) { + if (v.sq_count && v.mq_count) { DMERR("table load rejected: not all devices are blk-mq request-stackable"); return -EINVAL; } - t->all_blk_mq = mq_count > 0; + t->all_blk_mq = v.mq_count > 0; - if (t->type == DM_TYPE_MQ_REQUEST_BASED && !t->all_blk_mq) { + if (!t->all_blk_mq && + (t->type == DM_TYPE_MQ_REQUEST_BASED || t->type == DM_TYPE_NVME_BIO_BASED)) { DMERR("table load rejected: all devices are not blk-mq request-stackable"); return -EINVAL; } @@ -1079,7 +1120,8 @@ static int dm_table_alloc_md_mempools(struct dm_table *t, struct mapped_device * { enum dm_queue_mode type = dm_table_get_type(t); unsigned per_io_data_size = 0; - struct dm_target *tgt; + unsigned min_pool_size = 0; + struct dm_target *ti; unsigned i; if (unlikely(type == DM_TYPE_NONE)) { @@ -1089,11 +1131,13 @@ static int dm_table_alloc_md_mempools(struct dm_table *t, struct mapped_device * if (__table_type_bio_based(type)) for (i = 0; i < t->num_targets; i++) { - tgt = t->targets + i; - per_io_data_size = max(per_io_data_size, tgt->per_io_data_size); + ti = t->targets + i; + per_io_data_size = max(per_io_data_size, ti->per_io_data_size); + min_pool_size = max(min_pool_size, ti->num_flush_bios); } - t->mempools = dm_alloc_md_mempools(md, type, t->integrity_supported, per_io_data_size); + t->mempools = dm_alloc_md_mempools(md, type, t->integrity_supported, + per_io_data_size, min_pool_size); if (!t->mempools) return -ENOMEM; @@ -1705,6 +1749,20 @@ static bool dm_table_all_devices_attribute(struct dm_table *t, return true; } +static int device_no_partial_completion(struct dm_target *ti, struct dm_dev *dev, + sector_t start, sector_t len, void *data) +{ + char b[BDEVNAME_SIZE]; + + /* For now, NVMe devices are the only devices of this class */ + return (strncmp(bdevname(dev->bdev, b), "nvme", 3) == 0); +} + +static bool dm_table_does_not_support_partial_completion(struct dm_table *t) +{ + return dm_table_all_devices_attribute(t, device_no_partial_completion); +} + static int device_not_write_same_capable(struct dm_target *ti, struct dm_dev *dev, sector_t start, sector_t len, void *data) { @@ -1820,6 +1878,8 @@ void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q, } blk_queue_write_cache(q, wc, fua); + if (dm_table_supports_dax(t)) + queue_flag_set_unlocked(QUEUE_FLAG_DAX, q); if (dm_table_supports_dax_write_cache(t)) dax_write_cache(t->md->dax_dev, true); diff --git a/drivers/md/dm-thin.c b/drivers/md/dm-thin.c index f91d771fff4b..629c555890c1 100644 --- a/drivers/md/dm-thin.c +++ b/drivers/md/dm-thin.c @@ -492,6 +492,11 @@ static void pool_table_init(void) INIT_LIST_HEAD(&dm_thin_pool_table.pools); } +static void pool_table_exit(void) +{ + mutex_destroy(&dm_thin_pool_table.mutex); +} + static void __pool_table_insert(struct pool *pool) { BUG_ON(!mutex_is_locked(&dm_thin_pool_table.mutex)); @@ -1717,7 +1722,7 @@ static void __remap_and_issue_shared_cell(void *context, bio_op(bio) == REQ_OP_DISCARD) bio_list_add(&info->defer_bios, bio); else { - struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));; + struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook)); h->shared_read_entry = dm_deferred_entry_inc(info->tc->pool->shared_read_ds); inc_all_io_entry(info->tc->pool, bio); @@ -4387,6 +4392,8 @@ static void dm_thin_exit(void) dm_unregister_target(&pool_target); kmem_cache_destroy(_new_mapping_cache); + + pool_table_exit(); } module_init(dm_thin_init); diff --git a/drivers/md/dm-unstripe.c b/drivers/md/dm-unstripe.c new file mode 100644 index 000000000000..65f838fa2e99 --- /dev/null +++ b/drivers/md/dm-unstripe.c @@ -0,0 +1,219 @@ +/* + * Copyright (C) 2017 Intel Corporation. + * + * This file is released under the GPL. + */ + +#include "dm.h" + +#include <linux/module.h> +#include <linux/init.h> +#include <linux/blkdev.h> +#include <linux/bio.h> +#include <linux/slab.h> +#include <linux/bitops.h> +#include <linux/device-mapper.h> + +struct unstripe_c { + struct dm_dev *dev; + sector_t physical_start; + + uint32_t stripes; + + uint32_t unstripe; + sector_t unstripe_width; + sector_t unstripe_offset; + + uint32_t chunk_size; + u8 chunk_shift; +}; + +#define DM_MSG_PREFIX "unstriped" + +static void cleanup_unstripe(struct unstripe_c *uc, struct dm_target *ti) +{ + if (uc->dev) + dm_put_device(ti, uc->dev); + kfree(uc); +} + +/* + * Contruct an unstriped mapping. + * <number of stripes> <chunk size> <stripe #> <dev_path> <offset> + */ +static int unstripe_ctr(struct dm_target *ti, unsigned int argc, char **argv) +{ + struct unstripe_c *uc; + sector_t tmp_len; + unsigned long long start; + char dummy; + + if (argc != 5) { + ti->error = "Invalid number of arguments"; + return -EINVAL; + } + + uc = kzalloc(sizeof(*uc), GFP_KERNEL); + if (!uc) { + ti->error = "Memory allocation for unstriped context failed"; + return -ENOMEM; + } + + if (kstrtouint(argv[0], 10, &uc->stripes) || !uc->stripes) { + ti->error = "Invalid stripe count"; + goto err; + } + + if (kstrtouint(argv[1], 10, &uc->chunk_size) || !uc->chunk_size) { + ti->error = "Invalid chunk_size"; + goto err; + } + + // FIXME: must support non power of 2 chunk_size, dm-stripe.c does + if (!is_power_of_2(uc->chunk_size)) { + ti->error = "Non power of 2 chunk_size is not supported yet"; + goto err; + } + + if (kstrtouint(argv[2], 10, &uc->unstripe)) { + ti->error = "Invalid stripe number"; + goto err; + } + + if (uc->unstripe > uc->stripes && uc->stripes > 1) { + ti->error = "Please provide stripe between [0, # of stripes]"; + goto err; + } + + if (dm_get_device(ti, argv[3], dm_table_get_mode(ti->table), &uc->dev)) { + ti->error = "Couldn't get striped device"; + goto err; + } + + if (sscanf(argv[4], "%llu%c", &start, &dummy) != 1) { + ti->error = "Invalid striped device offset"; + goto err; + } + uc->physical_start = start; + + uc->unstripe_offset = uc->unstripe * uc->chunk_size; + uc->unstripe_width = (uc->stripes - 1) * uc->chunk_size; + uc->chunk_shift = fls(uc->chunk_size) - 1; + + tmp_len = ti->len; + if (sector_div(tmp_len, uc->chunk_size)) { + ti->error = "Target length not divisible by chunk size"; + goto err; + } + + if (dm_set_target_max_io_len(ti, uc->chunk_size)) { + ti->error = "Failed to set max io len"; + goto err; + } + + ti->private = uc; + return 0; +err: + cleanup_unstripe(uc, ti); + return -EINVAL; +} + +static void unstripe_dtr(struct dm_target *ti) +{ + struct unstripe_c *uc = ti->private; + + cleanup_unstripe(uc, ti); +} + +static sector_t map_to_core(struct dm_target *ti, struct bio *bio) +{ + struct unstripe_c *uc = ti->private; + sector_t sector = bio->bi_iter.bi_sector; + + /* Shift us up to the right "row" on the stripe */ + sector += uc->unstripe_width * (sector >> uc->chunk_shift); + + /* Account for what stripe we're operating on */ + sector += uc->unstripe_offset; + + return sector; +} + +static int unstripe_map(struct dm_target *ti, struct bio *bio) +{ + struct unstripe_c *uc = ti->private; + + bio_set_dev(bio, uc->dev->bdev); + bio->bi_iter.bi_sector = map_to_core(ti, bio) + uc->physical_start; + + return DM_MAPIO_REMAPPED; +} + +static void unstripe_status(struct dm_target *ti, status_type_t type, + unsigned int status_flags, char *result, unsigned int maxlen) +{ + struct unstripe_c *uc = ti->private; + unsigned int sz = 0; + + switch (type) { + case STATUSTYPE_INFO: + break; + + case STATUSTYPE_TABLE: + DMEMIT("%d %llu %d %s %llu", + uc->stripes, (unsigned long long)uc->chunk_size, uc->unstripe, + uc->dev->name, (unsigned long long)uc->physical_start); + break; + } +} + +static int unstripe_iterate_devices(struct dm_target *ti, + iterate_devices_callout_fn fn, void *data) +{ + struct unstripe_c *uc = ti->private; + + return fn(ti, uc->dev, uc->physical_start, ti->len, data); +} + +static void unstripe_io_hints(struct dm_target *ti, + struct queue_limits *limits) +{ + struct unstripe_c *uc = ti->private; + + limits->chunk_sectors = uc->chunk_size; +} + +static struct target_type unstripe_target = { + .name = "unstriped", + .version = {1, 0, 0}, + .module = THIS_MODULE, + .ctr = unstripe_ctr, + .dtr = unstripe_dtr, + .map = unstripe_map, + .status = unstripe_status, + .iterate_devices = unstripe_iterate_devices, + .io_hints = unstripe_io_hints, +}; + +static int __init dm_unstripe_init(void) +{ + int r; + + r = dm_register_target(&unstripe_target); + if (r < 0) + DMERR("target registration failed"); + + return r; +} + +static void __exit dm_unstripe_exit(void) +{ + dm_unregister_target(&unstripe_target); +} + +module_init(dm_unstripe_init); +module_exit(dm_unstripe_exit); + +MODULE_DESCRIPTION(DM_NAME " unstriped target"); +MODULE_AUTHOR("Scott Bauer <scott.bauer@intel.com>"); +MODULE_LICENSE("GPL"); diff --git a/drivers/md/dm-zoned-metadata.c b/drivers/md/dm-zoned-metadata.c index 70485de37b66..969954915566 100644 --- a/drivers/md/dm-zoned-metadata.c +++ b/drivers/md/dm-zoned-metadata.c @@ -2333,6 +2333,9 @@ static void dmz_cleanup_metadata(struct dmz_metadata *zmd) /* Free the zone descriptors */ dmz_drop_zones(zmd); + + mutex_destroy(&zmd->mblk_flush_lock); + mutex_destroy(&zmd->map_lock); } /* diff --git a/drivers/md/dm-zoned-target.c b/drivers/md/dm-zoned-target.c index 6d7bda6f8190..caff02caf083 100644 --- a/drivers/md/dm-zoned-target.c +++ b/drivers/md/dm-zoned-target.c @@ -827,6 +827,7 @@ err_fwq: err_cwq: destroy_workqueue(dmz->chunk_wq); err_bio: + mutex_destroy(&dmz->chunk_lock); bioset_free(dmz->bio_set); err_meta: dmz_dtr_metadata(dmz->metadata); @@ -861,6 +862,8 @@ static void dmz_dtr(struct dm_target *ti) dmz_put_zoned_device(ti); + mutex_destroy(&dmz->chunk_lock); + kfree(dmz); } diff --git a/drivers/md/dm.c b/drivers/md/dm.c index 8c26bfc35335..d6de00f367ef 100644 --- a/drivers/md/dm.c +++ b/drivers/md/dm.c @@ -60,18 +60,73 @@ void dm_issue_global_event(void) } /* - * One of these is allocated per bio. + * One of these is allocated (on-stack) per original bio. */ +struct clone_info { + struct dm_table *map; + struct bio *bio; + struct dm_io *io; + sector_t sector; + unsigned sector_count; +}; + +/* + * One of these is allocated per clone bio. + */ +#define DM_TIO_MAGIC 7282014 +struct dm_target_io { + unsigned magic; + struct dm_io *io; + struct dm_target *ti; + unsigned target_bio_nr; + unsigned *len_ptr; + bool inside_dm_io; + struct bio clone; +}; + +/* + * One of these is allocated per original bio. + * It contains the first clone used for that original. + */ +#define DM_IO_MAGIC 5191977 struct dm_io { + unsigned magic; struct mapped_device *md; blk_status_t status; atomic_t io_count; - struct bio *bio; + struct bio *orig_bio; unsigned long start_time; spinlock_t endio_lock; struct dm_stats_aux stats_aux; + /* last member of dm_target_io is 'struct bio' */ + struct dm_target_io tio; }; +void *dm_per_bio_data(struct bio *bio, size_t data_size) +{ + struct dm_target_io *tio = container_of(bio, struct dm_target_io, clone); + if (!tio->inside_dm_io) + return (char *)bio - offsetof(struct dm_target_io, clone) - data_size; + return (char *)bio - offsetof(struct dm_target_io, clone) - offsetof(struct dm_io, tio) - data_size; +} +EXPORT_SYMBOL_GPL(dm_per_bio_data); + +struct bio *dm_bio_from_per_bio_data(void *data, size_t data_size) +{ + struct dm_io *io = (struct dm_io *)((char *)data + data_size); + if (io->magic == DM_IO_MAGIC) + return (struct bio *)((char *)io + offsetof(struct dm_io, tio) + offsetof(struct dm_target_io, clone)); + BUG_ON(io->magic != DM_TIO_MAGIC); + return (struct bio *)((char *)io + offsetof(struct dm_target_io, clone)); +} +EXPORT_SYMBOL_GPL(dm_bio_from_per_bio_data); + +unsigned dm_bio_get_target_bio_nr(const struct bio *bio) +{ + return container_of(bio, struct dm_target_io, clone)->target_bio_nr; +} +EXPORT_SYMBOL_GPL(dm_bio_get_target_bio_nr); + #define MINOR_ALLOCED ((void *)-1) /* @@ -93,8 +148,8 @@ static int dm_numa_node = DM_NUMA_NODE; * For mempools pre-allocation at the table loading time. */ struct dm_md_mempools { - mempool_t *io_pool; struct bio_set *bs; + struct bio_set *io_bs; }; struct table_device { @@ -103,7 +158,6 @@ struct table_device { struct dm_dev dm_dev; }; -static struct kmem_cache *_io_cache; static struct kmem_cache *_rq_tio_cache; static struct kmem_cache *_rq_cache; @@ -170,14 +224,9 @@ static int __init local_init(void) { int r = -ENOMEM; - /* allocate a slab for the dm_ios */ - _io_cache = KMEM_CACHE(dm_io, 0); - if (!_io_cache) - return r; - _rq_tio_cache = KMEM_CACHE(dm_rq_target_io, 0); if (!_rq_tio_cache) - goto out_free_io_cache; + return r; _rq_cache = kmem_cache_create("dm_old_clone_request", sizeof(struct request), __alignof__(struct request), 0, NULL); @@ -212,8 +261,6 @@ out_free_rq_cache: kmem_cache_destroy(_rq_cache); out_free_rq_tio_cache: kmem_cache_destroy(_rq_tio_cache); -out_free_io_cache: - kmem_cache_destroy(_io_cache); return r; } @@ -225,7 +272,6 @@ static void local_exit(void) kmem_cache_destroy(_rq_cache); kmem_cache_destroy(_rq_tio_cache); - kmem_cache_destroy(_io_cache); unregister_blkdev(_major, _name); dm_uevent_exit(); @@ -486,18 +532,69 @@ out: return r; } -static struct dm_io *alloc_io(struct mapped_device *md) +static void start_io_acct(struct dm_io *io); + +static struct dm_io *alloc_io(struct mapped_device *md, struct bio *bio) { - return mempool_alloc(md->io_pool, GFP_NOIO); + struct dm_io *io; + struct dm_target_io *tio; + struct bio *clone; + + clone = bio_alloc_bioset(GFP_NOIO, 0, md->io_bs); + if (!clone) + return NULL; + + tio = container_of(clone, struct dm_target_io, clone); + tio->inside_dm_io = true; + tio->io = NULL; + + io = container_of(tio, struct dm_io, tio); + io->magic = DM_IO_MAGIC; + io->status = 0; + atomic_set(&io->io_count, 1); + io->orig_bio = bio; + io->md = md; + spin_lock_init(&io->endio_lock); + + start_io_acct(io); + + return io; } static void free_io(struct mapped_device *md, struct dm_io *io) { - mempool_free(io, md->io_pool); + bio_put(&io->tio.clone); +} + +static struct dm_target_io *alloc_tio(struct clone_info *ci, struct dm_target *ti, + unsigned target_bio_nr, gfp_t gfp_mask) +{ + struct dm_target_io *tio; + + if (!ci->io->tio.io) { + /* the dm_target_io embedded in ci->io is available */ + tio = &ci->io->tio; + } else { + struct bio *clone = bio_alloc_bioset(gfp_mask, 0, ci->io->md->bs); + if (!clone) + return NULL; + + tio = container_of(clone, struct dm_target_io, clone); + tio->inside_dm_io = false; + } + + tio->magic = DM_TIO_MAGIC; + tio->io = ci->io; + tio->ti = ti; + tio->target_bio_nr = target_bio_nr; + + return tio; } static void free_tio(struct dm_target_io *tio) { + if (tio->inside_dm_io) + return; bio_put(&tio->clone); } @@ -510,17 +607,15 @@ int md_in_flight(struct mapped_device *md) static void start_io_acct(struct dm_io *io) { struct mapped_device *md = io->md; - struct bio *bio = io->bio; - int cpu; + struct bio *bio = io->orig_bio; int rw = bio_data_dir(bio); io->start_time = jiffies; - cpu = part_stat_lock(); - part_round_stats(md->queue, cpu, &dm_disk(md)->part0); - part_stat_unlock(); + generic_start_io_acct(md->queue, rw, bio_sectors(bio), &dm_disk(md)->part0); + atomic_set(&dm_disk(md)->part0.in_flight[rw], - atomic_inc_return(&md->pending[rw])); + atomic_inc_return(&md->pending[rw])); if (unlikely(dm_stats_used(&md->stats))) dm_stats_account_io(&md->stats, bio_data_dir(bio), @@ -531,7 +626,7 @@ static void start_io_acct(struct dm_io *io) static void end_io_acct(struct dm_io *io) { struct mapped_device *md = io->md; - struct bio *bio = io->bio; + struct bio *bio = io->orig_bio; unsigned long duration = jiffies - io->start_time; int pending; int rw = bio_data_dir(bio); @@ -752,15 +847,6 @@ int dm_set_geometry(struct mapped_device *md, struct hd_geometry *geo) return 0; } -/*----------------------------------------------------------------- - * CRUD START: - * A more elegant soln is in the works that uses the queue - * merge fn, unfortunately there are a couple of changes to - * the block layer that I want to make for this. So in the - * interests of getting something for people to use I give - * you this clearly demarcated crap. - *---------------------------------------------------------------*/ - static int __noflush_suspending(struct mapped_device *md) { return test_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); @@ -780,8 +866,7 @@ static void dec_pending(struct dm_io *io, blk_status_t error) /* Push-back supersedes any I/O errors */ if (unlikely(error)) { spin_lock_irqsave(&io->endio_lock, flags); - if (!(io->status == BLK_STS_DM_REQUEUE && - __noflush_suspending(md))) + if (!(io->status == BLK_STS_DM_REQUEUE && __noflush_suspending(md))) io->status = error; spin_unlock_irqrestore(&io->endio_lock, flags); } @@ -793,7 +878,8 @@ static void dec_pending(struct dm_io *io, blk_status_t error) */ spin_lock_irqsave(&md->deferred_lock, flags); if (__noflush_suspending(md)) - bio_list_add_head(&md->deferred, io->bio); + /* NOTE early return due to BLK_STS_DM_REQUEUE below */ + bio_list_add_head(&md->deferred, io->orig_bio); else /* noflush suspend was interrupted. */ io->status = BLK_STS_IOERR; @@ -801,7 +887,7 @@ static void dec_pending(struct dm_io *io, blk_status_t error) } io_error = io->status; - bio = io->bio; + bio = io->orig_bio; end_io_acct(io); free_io(md, io); @@ -847,7 +933,7 @@ static void clone_endio(struct bio *bio) struct mapped_device *md = tio->io->md; dm_endio_fn endio = tio->ti->type->end_io; - if (unlikely(error == BLK_STS_TARGET)) { + if (unlikely(error == BLK_STS_TARGET) && md->type != DM_TYPE_NVME_BIO_BASED) { if (bio_op(bio) == REQ_OP_WRITE_SAME && !bio->bi_disk->queue->limits.max_write_same_sectors) disable_write_same(md); @@ -1005,7 +1091,7 @@ static size_t dm_dax_copy_from_iter(struct dax_device *dax_dev, pgoff_t pgoff, /* * A target may call dm_accept_partial_bio only from the map routine. It is - * allowed for all bio types except REQ_PREFLUSH. + * allowed for all bio types except REQ_PREFLUSH and REQ_OP_ZONE_RESET. * * dm_accept_partial_bio informs the dm that the target only wants to process * additional n_sectors sectors of the bio and the rest of the data should be @@ -1055,7 +1141,7 @@ void dm_remap_zone_report(struct dm_target *ti, struct bio *bio, sector_t start) { #ifdef CONFIG_BLK_DEV_ZONED struct dm_target_io *tio = container_of(bio, struct dm_target_io, clone); - struct bio *report_bio = tio->io->bio; + struct bio *report_bio = tio->io->orig_bio; struct blk_zone_report_hdr *hdr = NULL; struct blk_zone *zone; unsigned int nr_rep = 0; @@ -1122,67 +1208,15 @@ void dm_remap_zone_report(struct dm_target *ti, struct bio *bio, sector_t start) } EXPORT_SYMBOL_GPL(dm_remap_zone_report); -/* - * Flush current->bio_list when the target map method blocks. - * This fixes deadlocks in snapshot and possibly in other targets. - */ -struct dm_offload { - struct blk_plug plug; - struct blk_plug_cb cb; -}; - -static void flush_current_bio_list(struct blk_plug_cb *cb, bool from_schedule) -{ - struct dm_offload *o = container_of(cb, struct dm_offload, cb); - struct bio_list list; - struct bio *bio; - int i; - - INIT_LIST_HEAD(&o->cb.list); - - if (unlikely(!current->bio_list)) - return; - - for (i = 0; i < 2; i++) { - list = current->bio_list[i]; - bio_list_init(¤t->bio_list[i]); - - while ((bio = bio_list_pop(&list))) { - struct bio_set *bs = bio->bi_pool; - if (unlikely(!bs) || bs == fs_bio_set || - !bs->rescue_workqueue) { - bio_list_add(¤t->bio_list[i], bio); - continue; - } - - spin_lock(&bs->rescue_lock); - bio_list_add(&bs->rescue_list, bio); - queue_work(bs->rescue_workqueue, &bs->rescue_work); - spin_unlock(&bs->rescue_lock); - } - } -} - -static void dm_offload_start(struct dm_offload *o) -{ - blk_start_plug(&o->plug); - o->cb.callback = flush_current_bio_list; - list_add(&o->cb.list, ¤t->plug->cb_list); -} - -static void dm_offload_end(struct dm_offload *o) -{ - list_del(&o->cb.list); - blk_finish_plug(&o->plug); -} - -static void __map_bio(struct dm_target_io *tio) +static blk_qc_t __map_bio(struct dm_target_io *tio) { int r; sector_t sector; - struct dm_offload o; struct bio *clone = &tio->clone; + struct dm_io *io = tio->io; + struct mapped_device *md = io->md; struct dm_target *ti = tio->ti; + blk_qc_t ret = BLK_QC_T_NONE; clone->bi_end_io = clone_endio; @@ -1191,44 +1225,37 @@ static void __map_bio(struct dm_target_io *tio) * anything, the target has assumed ownership of * this io. */ - atomic_inc(&tio->io->io_count); + atomic_inc(&io->io_count); sector = clone->bi_iter.bi_sector; - dm_offload_start(&o); r = ti->type->map(ti, clone); - dm_offload_end(&o); - switch (r) { case DM_MAPIO_SUBMITTED: break; case DM_MAPIO_REMAPPED: /* the bio has been remapped so dispatch it */ trace_block_bio_remap(clone->bi_disk->queue, clone, - bio_dev(tio->io->bio), sector); - generic_make_request(clone); + bio_dev(io->orig_bio), sector); + if (md->type == DM_TYPE_NVME_BIO_BASED) + ret = direct_make_request(clone); + else + ret = generic_make_request(clone); break; case DM_MAPIO_KILL: - dec_pending(tio->io, BLK_STS_IOERR); free_tio(tio); + dec_pending(io, BLK_STS_IOERR); break; case DM_MAPIO_REQUEUE: - dec_pending(tio->io, BLK_STS_DM_REQUEUE); free_tio(tio); + dec_pending(io, BLK_STS_DM_REQUEUE); break; default: DMWARN("unimplemented target map return value: %d", r); BUG(); } -} -struct clone_info { - struct mapped_device *md; - struct dm_table *map; - struct bio *bio; - struct dm_io *io; - sector_t sector; - unsigned sector_count; -}; + return ret; +} static void bio_setup_sector(struct bio *bio, sector_t sector, unsigned len) { @@ -1272,28 +1299,49 @@ static int clone_bio(struct dm_target_io *tio, struct bio *bio, return 0; } -static struct dm_target_io *alloc_tio(struct clone_info *ci, - struct dm_target *ti, - unsigned target_bio_nr) +static void alloc_multiple_bios(struct bio_list *blist, struct clone_info *ci, + struct dm_target *ti, unsigned num_bios) { struct dm_target_io *tio; - struct bio *clone; + int try; - clone = bio_alloc_bioset(GFP_NOIO, 0, ci->md->bs); - tio = container_of(clone, struct dm_target_io, clone); + if (!num_bios) + return; - tio->io = ci->io; - tio->ti = ti; - tio->target_bio_nr = target_bio_nr; + if (num_bios == 1) { + tio = alloc_tio(ci, ti, 0, GFP_NOIO); + bio_list_add(blist, &tio->clone); + return; + } - return tio; + for (try = 0; try < 2; try++) { + int bio_nr; + struct bio *bio; + + if (try) + mutex_lock(&ci->io->md->table_devices_lock); + for (bio_nr = 0; bio_nr < num_bios; bio_nr++) { + tio = alloc_tio(ci, ti, bio_nr, try ? GFP_NOIO : GFP_NOWAIT); + if (!tio) + break; + + bio_list_add(blist, &tio->clone); + } + if (try) + mutex_unlock(&ci->io->md->table_devices_lock); + if (bio_nr == num_bios) + return; + + while ((bio = bio_list_pop(blist))) { + tio = container_of(bio, struct dm_target_io, clone); + free_tio(tio); + } + } } -static void __clone_and_map_simple_bio(struct clone_info *ci, - struct dm_target *ti, - unsigned target_bio_nr, unsigned *len) +static blk_qc_t __clone_and_map_simple_bio(struct clone_info *ci, + struct dm_target_io *tio, unsigned *len) { - struct dm_target_io *tio = alloc_tio(ci, ti, target_bio_nr); struct bio *clone = &tio->clone; tio->len_ptr = len; @@ -1302,16 +1350,22 @@ static void __clone_and_map_simple_bio(struct clone_info *ci, if (len) bio_setup_sector(clone, ci->sector, *len); - __map_bio(tio); + return __map_bio(tio); } static void __send_duplicate_bios(struct clone_info *ci, struct dm_target *ti, unsigned num_bios, unsigned *len) { - unsigned target_bio_nr; + struct bio_list blist = BIO_EMPTY_LIST; + struct bio *bio; + struct dm_target_io *tio; + + alloc_multiple_bios(&blist, ci, ti, num_bios); - for (target_bio_nr = 0; target_bio_nr < num_bios; target_bio_nr++) - __clone_and_map_simple_bio(ci, ti, target_bio_nr, len); + while ((bio = bio_list_pop(&blist))) { + tio = container_of(bio, struct dm_target_io, clone); + (void) __clone_and_map_simple_bio(ci, tio, len); + } } static int __send_empty_flush(struct clone_info *ci) @@ -1327,32 +1381,22 @@ static int __send_empty_flush(struct clone_info *ci) } static int __clone_and_map_data_bio(struct clone_info *ci, struct dm_target *ti, - sector_t sector, unsigned *len) + sector_t sector, unsigned *len) { struct bio *bio = ci->bio; struct dm_target_io *tio; - unsigned target_bio_nr; - unsigned num_target_bios = 1; - int r = 0; + int r; - /* - * Does the target want to receive duplicate copies of the bio? - */ - if (bio_data_dir(bio) == WRITE && ti->num_write_bios) - num_target_bios = ti->num_write_bios(ti, bio); - - for (target_bio_nr = 0; target_bio_nr < num_target_bios; target_bio_nr++) { - tio = alloc_tio(ci, ti, target_bio_nr); - tio->len_ptr = len; - r = clone_bio(tio, bio, sector, *len); - if (r < 0) { - free_tio(tio); - break; - } - __map_bio(tio); + tio = alloc_tio(ci, ti, 0, GFP_NOIO); + tio->len_ptr = len; + r = clone_bio(tio, bio, sector, *len); + if (r < 0) { + free_tio(tio); + return r; } + (void) __map_bio(tio); - return r; + return 0; } typedef unsigned (*get_num_bios_fn)(struct dm_target *ti); @@ -1379,56 +1423,50 @@ static bool is_split_required_for_discard(struct dm_target *ti) return ti->split_discard_bios; } -static int __send_changing_extent_only(struct clone_info *ci, +static int __send_changing_extent_only(struct clone_info *ci, struct dm_target *ti, get_num_bios_fn get_num_bios, is_split_required_fn is_split_required) { - struct dm_target *ti; unsigned len; unsigned num_bios; - do { - ti = dm_table_find_target(ci->map, ci->sector); - if (!dm_target_is_valid(ti)) - return -EIO; - - /* - * Even though the device advertised support for this type of - * request, that does not mean every target supports it, and - * reconfiguration might also have changed that since the - * check was performed. - */ - num_bios = get_num_bios ? get_num_bios(ti) : 0; - if (!num_bios) - return -EOPNOTSUPP; + /* + * Even though the device advertised support for this type of + * request, that does not mean every target supports it, and + * reconfiguration might also have changed that since the + * check was performed. + */ + num_bios = get_num_bios ? get_num_bios(ti) : 0; + if (!num_bios) + return -EOPNOTSUPP; - if (is_split_required && !is_split_required(ti)) - len = min((sector_t)ci->sector_count, max_io_len_target_boundary(ci->sector, ti)); - else - len = min((sector_t)ci->sector_count, max_io_len(ci->sector, ti)); + if (is_split_required && !is_split_required(ti)) + len = min((sector_t)ci->sector_count, max_io_len_target_boundary(ci->sector, ti)); + else + len = min((sector_t)ci->sector_count, max_io_len(ci->sector, ti)); - __send_duplicate_bios(ci, ti, num_bios, &len); + __send_duplicate_bios(ci, ti, num_bios, &len); - ci->sector += len; - } while (ci->sector_count -= len); + ci->sector += len; + ci->sector_count -= len; return 0; } -static int __send_discard(struct clone_info *ci) +static int __send_discard(struct clone_info *ci, struct dm_target *ti) { - return __send_changing_extent_only(ci, get_num_discard_bios, + return __send_changing_extent_only(ci, ti, get_num_discard_bios, is_split_required_for_discard); } -static int __send_write_same(struct clone_info *ci) +static int __send_write_same(struct clone_info *ci, struct dm_target *ti) { - return __send_changing_extent_only(ci, get_num_write_same_bios, NULL); + return __send_changing_extent_only(ci, ti, get_num_write_same_bios, NULL); } -static int __send_write_zeroes(struct clone_info *ci) +static int __send_write_zeroes(struct clone_info *ci, struct dm_target *ti) { - return __send_changing_extent_only(ci, get_num_write_zeroes_bios, NULL); + return __send_changing_extent_only(ci, ti, get_num_write_zeroes_bios, NULL); } /* @@ -1441,17 +1479,17 @@ static int __split_and_process_non_flush(struct clone_info *ci) unsigned len; int r; - if (unlikely(bio_op(bio) == REQ_OP_DISCARD)) - return __send_discard(ci); - else if (unlikely(bio_op(bio) == REQ_OP_WRITE_SAME)) - return __send_write_same(ci); - else if (unlikely(bio_op(bio) == REQ_OP_WRITE_ZEROES)) - return __send_write_zeroes(ci); - ti = dm_table_find_target(ci->map, ci->sector); if (!dm_target_is_valid(ti)) return -EIO; + if (unlikely(bio_op(bio) == REQ_OP_DISCARD)) + return __send_discard(ci, ti); + else if (unlikely(bio_op(bio) == REQ_OP_WRITE_SAME)) + return __send_write_same(ci, ti); + else if (unlikely(bio_op(bio) == REQ_OP_WRITE_ZEROES)) + return __send_write_zeroes(ci, ti); + if (bio_op(bio) == REQ_OP_ZONE_REPORT) len = ci->sector_count; else @@ -1468,34 +1506,33 @@ static int __split_and_process_non_flush(struct clone_info *ci) return 0; } +static void init_clone_info(struct clone_info *ci, struct mapped_device *md, + struct dm_table *map, struct bio *bio) +{ + ci->map = map; + ci->io = alloc_io(md, bio); + ci->sector = bio->bi_iter.bi_sector; +} + /* * Entry point to split a bio into clones and submit them to the targets. */ -static void __split_and_process_bio(struct mapped_device *md, - struct dm_table *map, struct bio *bio) +static blk_qc_t __split_and_process_bio(struct mapped_device *md, + struct dm_table *map, struct bio *bio) { struct clone_info ci; + blk_qc_t ret = BLK_QC_T_NONE; int error = 0; if (unlikely(!map)) { bio_io_error(bio); - return; + return ret; } - ci.map = map; - ci.md = md; - ci.io = alloc_io(md); - ci.io->status = 0; - atomic_set(&ci.io->io_count, 1); - ci.io->bio = bio; - ci.io->md = md; - spin_lock_init(&ci.io->endio_lock); - ci.sector = bio->bi_iter.bi_sector; - - start_io_acct(ci.io); + init_clone_info(&ci, md, map, bio); if (bio->bi_opf & REQ_PREFLUSH) { - ci.bio = &ci.md->flush_bio; + ci.bio = &ci.io->md->flush_bio; ci.sector_count = 0; error = __send_empty_flush(&ci); /* dec_pending submits any data associated with flush */ @@ -1506,32 +1543,95 @@ static void __split_and_process_bio(struct mapped_device *md, } else { ci.bio = bio; ci.sector_count = bio_sectors(bio); - while (ci.sector_count && !error) + while (ci.sector_count && !error) { error = __split_and_process_non_flush(&ci); + if (current->bio_list && ci.sector_count && !error) { + /* + * Remainder must be passed to generic_make_request() + * so that it gets handled *after* bios already submitted + * have been completely processed. + * We take a clone of the original to store in + * ci.io->orig_bio to be used by end_io_acct() and + * for dec_pending to use for completion handling. + * As this path is not used for REQ_OP_ZONE_REPORT, + * the usage of io->orig_bio in dm_remap_zone_report() + * won't be affected by this reassignment. + */ + struct bio *b = bio_clone_bioset(bio, GFP_NOIO, + md->queue->bio_split); + ci.io->orig_bio = b; + bio_advance(bio, (bio_sectors(bio) - ci.sector_count) << 9); + bio_chain(b, bio); + ret = generic_make_request(bio); + break; + } + } } /* drop the extra reference count */ dec_pending(ci.io, errno_to_blk_status(error)); + return ret; } -/*----------------------------------------------------------------- - * CRUD END - *---------------------------------------------------------------*/ /* - * The request function that just remaps the bio built up by - * dm_merge_bvec. + * Optimized variant of __split_and_process_bio that leverages the + * fact that targets that use it do _not_ have a need to split bios. */ -static blk_qc_t dm_make_request(struct request_queue *q, struct bio *bio) +static blk_qc_t __process_bio(struct mapped_device *md, + struct dm_table *map, struct bio *bio) +{ + struct clone_info ci; + blk_qc_t ret = BLK_QC_T_NONE; + int error = 0; + + if (unlikely(!map)) { + bio_io_error(bio); + return ret; + } + + init_clone_info(&ci, md, map, bio); + + if (bio->bi_opf & REQ_PREFLUSH) { + ci.bio = &ci.io->md->flush_bio; + ci.sector_count = 0; + error = __send_empty_flush(&ci); + /* dec_pending submits any data associated with flush */ + } else { + struct dm_target *ti = md->immutable_target; + struct dm_target_io *tio; + + /* + * Defend against IO still getting in during teardown + * - as was seen for a time with nvme-fcloop + */ + if (unlikely(WARN_ON_ONCE(!ti || !dm_target_is_valid(ti)))) { + error = -EIO; + goto out; + } + + tio = alloc_tio(&ci, ti, 0, GFP_NOIO); + ci.bio = bio; + ci.sector_count = bio_sectors(bio); + ret = __clone_and_map_simple_bio(&ci, tio, NULL); + } +out: + /* drop the extra reference count */ + dec_pending(ci.io, errno_to_blk_status(error)); + return ret; +} + +typedef blk_qc_t (process_bio_fn)(struct mapped_device *, struct dm_table *, struct bio *); + +static blk_qc_t __dm_make_request(struct request_queue *q, struct bio *bio, + process_bio_fn process_bio) { - int rw = bio_data_dir(bio); struct mapped_device *md = q->queuedata; + blk_qc_t ret = BLK_QC_T_NONE; int srcu_idx; struct dm_table *map; map = dm_get_live_table(md, &srcu_idx); - generic_start_io_acct(q, rw, bio_sectors(bio), &dm_disk(md)->part0); - /* if we're suspended, we have to queue this io for later */ if (unlikely(test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags))) { dm_put_live_table(md, srcu_idx); @@ -1540,12 +1640,27 @@ static blk_qc_t dm_make_request(struct request_queue *q, struct bio *bio) queue_io(md, bio); else bio_io_error(bio); - return BLK_QC_T_NONE; + return ret; } - __split_and_process_bio(md, map, bio); + ret = process_bio(md, map, bio); + dm_put_live_table(md, srcu_idx); - return BLK_QC_T_NONE; + return ret; +} + +/* + * The request function that remaps the bio to one target and + * splits off any remainder. + */ +static blk_qc_t dm_make_request(struct request_queue *q, struct bio *bio) +{ + return __dm_make_request(q, bio, __split_and_process_bio); +} + +static blk_qc_t dm_make_request_nvme(struct request_queue *q, struct bio *bio) +{ + return __dm_make_request(q, bio, __process_bio); } static int dm_any_congested(void *congested_data, int bdi_bits) @@ -1626,20 +1741,9 @@ static const struct dax_operations dm_dax_ops; static void dm_wq_work(struct work_struct *work); -void dm_init_md_queue(struct mapped_device *md) -{ - /* - * Initialize data that will only be used by a non-blk-mq DM queue - * - must do so here (in alloc_dev callchain) before queue is used - */ - md->queue->queuedata = md; - md->queue->backing_dev_info->congested_data = md; -} - -void dm_init_normal_md_queue(struct mapped_device *md) +static void dm_init_normal_md_queue(struct mapped_device *md) { md->use_blk_mq = false; - dm_init_md_queue(md); /* * Initialize aspects of queue that aren't relevant for blk-mq @@ -1653,9 +1757,10 @@ static void cleanup_mapped_device(struct mapped_device *md) destroy_workqueue(md->wq); if (md->kworker_task) kthread_stop(md->kworker_task); - mempool_destroy(md->io_pool); if (md->bs) bioset_free(md->bs); + if (md->io_bs) + bioset_free(md->io_bs); if (md->dax_dev) { kill_dax(md->dax_dev); @@ -1681,6 +1786,10 @@ static void cleanup_mapped_device(struct mapped_device *md) md->bdev = NULL; } + mutex_destroy(&md->suspend_lock); + mutex_destroy(&md->type_lock); + mutex_destroy(&md->table_devices_lock); + dm_mq_cleanup_mapped_device(md); } @@ -1734,10 +1843,10 @@ static struct mapped_device *alloc_dev(int minor) md->queue = blk_alloc_queue_node(GFP_KERNEL, numa_node_id); if (!md->queue) goto bad; + md->queue->queuedata = md; + md->queue->backing_dev_info->congested_data = md; - dm_init_md_queue(md); - - md->disk = alloc_disk_node(1, numa_node_id); + md->disk = alloc_disk_node(1, md->numa_node_id); if (!md->disk) goto bad; @@ -1820,17 +1929,22 @@ static void __bind_mempools(struct mapped_device *md, struct dm_table *t) { struct dm_md_mempools *p = dm_table_get_md_mempools(t); - if (md->bs) { - /* The md already has necessary mempools. */ - if (dm_table_bio_based(t)) { - /* - * Reload bioset because front_pad may have changed - * because a different table was loaded. - */ + if (dm_table_bio_based(t)) { + /* + * The md may already have mempools that need changing. + * If so, reload bioset because front_pad may have changed + * because a different table was loaded. + */ + if (md->bs) { bioset_free(md->bs); - md->bs = p->bs; - p->bs = NULL; + md->bs = NULL; + } + if (md->io_bs) { + bioset_free(md->io_bs); + md->io_bs = NULL; } + + } else if (md->bs) { /* * There's no need to reload with request-based dm * because the size of front_pad doesn't change. @@ -1842,13 +1956,12 @@ static void __bind_mempools(struct mapped_device *md, struct dm_table *t) goto out; } - BUG_ON(!p || md->io_pool || md->bs); + BUG_ON(!p || md->bs || md->io_bs); - md->io_pool = p->io_pool; - p->io_pool = NULL; md->bs = p->bs; p->bs = NULL; - + md->io_bs = p->io_bs; + p->io_bs = NULL; out: /* mempool bind completed, no longer need any mempools in the table */ dm_table_free_md_mempools(t); @@ -1894,6 +2007,7 @@ static struct dm_table *__bind(struct mapped_device *md, struct dm_table *t, { struct dm_table *old_map; struct request_queue *q = md->queue; + bool request_based = dm_table_request_based(t); sector_t size; lockdep_assert_held(&md->suspend_lock); @@ -1917,12 +2031,15 @@ static struct dm_table *__bind(struct mapped_device *md, struct dm_table *t, * This must be done before setting the queue restrictions, * because request-based dm may be run just after the setting. */ - if (dm_table_request_based(t)) { + if (request_based) dm_stop_queue(q); + + if (request_based || md->type == DM_TYPE_NVME_BIO_BASED) { /* - * Leverage the fact that request-based DM targets are - * immutable singletons and establish md->immutable_target - * - used to optimize both dm_request_fn and dm_mq_queue_rq + * Leverage the fact that request-based DM targets and + * NVMe bio based targets are immutable singletons + * - used to optimize both dm_request_fn and dm_mq_queue_rq; + * and __process_bio. */ md->immutable_target = dm_table_get_immutable_target(t); } @@ -1962,13 +2079,18 @@ static struct dm_table *__unbind(struct mapped_device *md) */ int dm_create(int minor, struct mapped_device **result) { + int r; struct mapped_device *md; md = alloc_dev(minor); if (!md) return -ENXIO; - dm_sysfs_init(md); + r = dm_sysfs_init(md); + if (r) { + free_dev(md); + return r; + } *result = md; return 0; @@ -2026,6 +2148,7 @@ int dm_setup_md_queue(struct mapped_device *md, struct dm_table *t) switch (type) { case DM_TYPE_REQUEST_BASED: + dm_init_normal_md_queue(md); r = dm_old_init_request_queue(md, t); if (r) { DMERR("Cannot initialize queue for request-based mapped device"); @@ -2043,15 +2166,10 @@ int dm_setup_md_queue(struct mapped_device *md, struct dm_table *t) case DM_TYPE_DAX_BIO_BASED: dm_init_normal_md_queue(md); blk_queue_make_request(md->queue, dm_make_request); - /* - * DM handles splitting bios as needed. Free the bio_split bioset - * since it won't be used (saves 1 process per bio-based DM device). - */ - bioset_free(md->queue->bio_split); - md->queue->bio_split = NULL; - - if (type == DM_TYPE_DAX_BIO_BASED) - queue_flag_set_unlocked(QUEUE_FLAG_DAX, md->queue); + break; + case DM_TYPE_NVME_BIO_BASED: + dm_init_normal_md_queue(md); + blk_queue_make_request(md->queue, dm_make_request_nvme); break; case DM_TYPE_NONE: WARN_ON_ONCE(true); @@ -2130,7 +2248,6 @@ EXPORT_SYMBOL_GPL(dm_device_name); static void __dm_destroy(struct mapped_device *md, bool wait) { - struct request_queue *q = dm_get_md_queue(md); struct dm_table *map; int srcu_idx; @@ -2141,7 +2258,7 @@ static void __dm_destroy(struct mapped_device *md, bool wait) set_bit(DMF_FREEING, &md->flags); spin_unlock(&_minor_lock); - blk_set_queue_dying(q); + blk_set_queue_dying(md->queue); if (dm_request_based(md) && md->kworker_task) kthread_flush_worker(&md->kworker); @@ -2752,11 +2869,12 @@ int dm_noflush_suspending(struct dm_target *ti) EXPORT_SYMBOL_GPL(dm_noflush_suspending); struct dm_md_mempools *dm_alloc_md_mempools(struct mapped_device *md, enum dm_queue_mode type, - unsigned integrity, unsigned per_io_data_size) + unsigned integrity, unsigned per_io_data_size, + unsigned min_pool_size) { struct dm_md_mempools *pools = kzalloc_node(sizeof(*pools), GFP_KERNEL, md->numa_node_id); unsigned int pool_size = 0; - unsigned int front_pad; + unsigned int front_pad, io_front_pad; if (!pools) return NULL; @@ -2764,16 +2882,19 @@ struct dm_md_mempools *dm_alloc_md_mempools(struct mapped_device *md, enum dm_qu switch (type) { case DM_TYPE_BIO_BASED: case DM_TYPE_DAX_BIO_BASED: - pool_size = dm_get_reserved_bio_based_ios(); + case DM_TYPE_NVME_BIO_BASED: + pool_size = max(dm_get_reserved_bio_based_ios(), min_pool_size); front_pad = roundup(per_io_data_size, __alignof__(struct dm_target_io)) + offsetof(struct dm_target_io, clone); - - pools->io_pool = mempool_create_slab_pool(pool_size, _io_cache); - if (!pools->io_pool) + io_front_pad = roundup(front_pad, __alignof__(struct dm_io)) + offsetof(struct dm_io, tio); + pools->io_bs = bioset_create(pool_size, io_front_pad, 0); + if (!pools->io_bs) + goto out; + if (integrity && bioset_integrity_create(pools->io_bs, pool_size)) goto out; break; case DM_TYPE_REQUEST_BASED: case DM_TYPE_MQ_REQUEST_BASED: - pool_size = dm_get_reserved_rq_based_ios(); + pool_size = max(dm_get_reserved_rq_based_ios(), min_pool_size); front_pad = offsetof(struct dm_rq_clone_bio_info, clone); /* per_io_data_size is used for blk-mq pdu at queue allocation */ break; @@ -2781,7 +2902,7 @@ struct dm_md_mempools *dm_alloc_md_mempools(struct mapped_device *md, enum dm_qu BUG(); } - pools->bs = bioset_create(pool_size, front_pad, BIOSET_NEED_RESCUER); + pools->bs = bioset_create(pool_size, front_pad, 0); if (!pools->bs) goto out; @@ -2801,10 +2922,10 @@ void dm_free_md_mempools(struct dm_md_mempools *pools) if (!pools) return; - mempool_destroy(pools->io_pool); - if (pools->bs) bioset_free(pools->bs); + if (pools->io_bs) + bioset_free(pools->io_bs); kfree(pools); } diff --git a/drivers/md/dm.h b/drivers/md/dm.h index 36399bb875dd..114a81b27c37 100644 --- a/drivers/md/dm.h +++ b/drivers/md/dm.h @@ -49,7 +49,6 @@ struct dm_md_mempools; /*----------------------------------------------------------------- * Internal table functions. *---------------------------------------------------------------*/ -void dm_table_destroy(struct dm_table *t); void dm_table_event_callback(struct dm_table *t, void (*fn)(void *), void *context); struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index); @@ -206,7 +205,8 @@ void dm_kcopyd_exit(void); * Mempool operations */ struct dm_md_mempools *dm_alloc_md_mempools(struct mapped_device *md, enum dm_queue_mode type, - unsigned integrity, unsigned per_bio_data_size); + unsigned integrity, unsigned per_bio_data_size, + unsigned min_pool_size); void dm_free_md_mempools(struct dm_md_mempools *pools); /* diff --git a/include/linux/device-mapper.h b/include/linux/device-mapper.h index a5538433c927..da83f64952e7 100644 --- a/include/linux/device-mapper.h +++ b/include/linux/device-mapper.h @@ -28,6 +28,7 @@ enum dm_queue_mode { DM_TYPE_REQUEST_BASED = 2, DM_TYPE_MQ_REQUEST_BASED = 3, DM_TYPE_DAX_BIO_BASED = 4, + DM_TYPE_NVME_BIO_BASED = 5, }; typedef enum { STATUSTYPE_INFO, STATUSTYPE_TABLE } status_type_t; @@ -221,14 +222,6 @@ struct target_type { #define dm_target_is_wildcard(type) ((type)->features & DM_TARGET_WILDCARD) /* - * Some targets need to be sent the same WRITE bio severals times so - * that they can send copies of it to different devices. This function - * examines any supplied bio and returns the number of copies of it the - * target requires. - */ -typedef unsigned (*dm_num_write_bios_fn) (struct dm_target *ti, struct bio *bio); - -/* * A target implements own bio data integrity. */ #define DM_TARGET_INTEGRITY 0x00000010 @@ -291,13 +284,6 @@ struct dm_target { */ unsigned per_io_data_size; - /* - * If defined, this function is called to find out how many - * duplicate bios should be sent to the target when writing - * data. - */ - dm_num_write_bios_fn num_write_bios; - /* target specific data */ void *private; @@ -329,35 +315,9 @@ struct dm_target_callbacks { int (*congested_fn) (struct dm_target_callbacks *, int); }; -/* - * For bio-based dm. - * One of these is allocated for each bio. - * This structure shouldn't be touched directly by target drivers. - * It is here so that we can inline dm_per_bio_data and - * dm_bio_from_per_bio_data - */ -struct dm_target_io { - struct dm_io *io; - struct dm_target *ti; - unsigned target_bio_nr; - unsigned *len_ptr; - struct bio clone; -}; - -static inline void *dm_per_bio_data(struct bio *bio, size_t data_size) -{ - return (char *)bio - offsetof(struct dm_target_io, clone) - data_size; -} - -static inline struct bio *dm_bio_from_per_bio_data(void *data, size_t data_size) -{ - return (struct bio *)((char *)data + data_size + offsetof(struct dm_target_io, clone)); -} - -static inline unsigned dm_bio_get_target_bio_nr(const struct bio *bio) -{ - return container_of(bio, struct dm_target_io, clone)->target_bio_nr; -} +void *dm_per_bio_data(struct bio *bio, size_t data_size); +struct bio *dm_bio_from_per_bio_data(void *data, size_t data_size); +unsigned dm_bio_get_target_bio_nr(const struct bio *bio); int dm_register_target(struct target_type *t); void dm_unregister_target(struct target_type *t); @@ -500,6 +460,11 @@ void dm_table_set_type(struct dm_table *t, enum dm_queue_mode type); int dm_table_complete(struct dm_table *t); /* + * Destroy the table when finished. + */ +void dm_table_destroy(struct dm_table *t); + +/* * Target may require that it is never sent I/O larger than len. */ int __must_check dm_set_target_max_io_len(struct dm_target *ti, sector_t len); @@ -585,6 +550,7 @@ do { \ #define DM_ENDIO_DONE 0 #define DM_ENDIO_INCOMPLETE 1 #define DM_ENDIO_REQUEUE 2 +#define DM_ENDIO_DELAY_REQUEUE 3 /* * Definitions of return values from target map function. @@ -592,7 +558,7 @@ do { \ #define DM_MAPIO_SUBMITTED 0 #define DM_MAPIO_REMAPPED 1 #define DM_MAPIO_REQUEUE DM_ENDIO_REQUEUE -#define DM_MAPIO_DELAY_REQUEUE 3 +#define DM_MAPIO_DELAY_REQUEUE DM_ENDIO_DELAY_REQUEUE #define DM_MAPIO_KILL 4 #define dm_sector_div64(x, y)( \ |